Hematogenous transmission of virus to the CNS involving either BMEC infection, damage to the tight junctions, or both, results in changes to BBB integrity that are likely an essential mechanism of subsequent ictogenesis and epileptogenesis (Löscher and Friedman, 2020). One hallmark of a damaged BBB is the extravasation of albumin from the blood to the brain parenchyma (Friedman et al., 2009). In the brain parenchyma, albumin can be taken up or bound to neurons, astrocytes, and microglial cells. In astrocytes, albumin can be taken up via transforming growth factor-beta (TGF-β) receptors. This is followed by downregulation of inward rectifying potassium channels (Kir 4.1), water channels (aquaporin 4; AQP4), and glutamate transporters in these astrocytes (Löscher and Friedman, 2020). As a result, the buffering of extracellular potassium and glutamate is reduced, which facilitates N-methyl-D-aspartate (NMDA) receptor-mediated neuronal hyperexcitability and eventually induces epileptiform activity (Löscher and Friedman, 2020). TGF-β receptor signaling is further associated with transcriptional changes involved in inflammation, alterations in extracellular matrix (specifically the perineuronal nets around inhibitory interneurons), excitatory synaptogenesis, and pathological plasticity, all considered important mechanisms that contribute to lowering the seizure threshold during epileptogenesis (Löscher and Friedman, 2020). As a proof-of-concept that albumin extravasation plays a crucial role in the generation of seizures, the angiotensin II type 1 (AT1) receptor antagonist, losartan, which blocks brain TGF-β receptor signaling, was shown to prevent epilepsy in different models of epileptogenesis (Swissa et al., 2019).

For many decades, the limbic system in the temporal lobes, including the hippocampal formation and parahippocampal areas such as the piriform, perirhinal, and entorhinal cortices, have been known to play a crucial role in the development of seizures and epilepsy (Walter, 1969; Meldrum, 1975; Ribak et al., 1992; Engel, 1996; Löscher and Ebert, 1996; Chatzikonstantinou, 2014; Scharfman, 2019). The hippocampus is considered by many to be the generator of TLE, the most common type of epilepsy in adults and a frequent consequence of viral infections (Vezzani et al., 2016). TLE is typically associated with hippocampal sclerosis, a neuropathological condition with severe neuronal cell loss and gliosis in the hippocampus, specifically in the CA1 (Cornu Ammonis area 1) region and subiculum of the hippocampus proper and in the hilus of the dentate gyrus (Blümcke et al., 2002). Hippocampal sclerosis was first described in 1880 by Wilhelm Sommer as an etiological component of epilepsy (Sommer, 1880). In addition to neuron loss, aberrant sprouting of dentate granule cell mossy fibers in mesial TLE is thought to underlie the creation of aberrant circuitry that promotes the generation or spread of spontaneous seizure activity (Sutula and Dudek, 2007; Scharfman, 2019). Surgical removal of the sclerotic hippocampus in drug-resistant patients often improves or even cures TLE (Löscher et al., 2020). Thus, these structural changes in the hippocampal formation provide a mechanism by which viral infections could induce seizures and epilepsy.

As discussed above, some viruses may be more epileptogenic due to their anatomic distribution, as in the case of HSV, with a propensity to affect the temporal lobes, including the hippocampus (Theodore, 2014). HSV causes widespread inflammation, edema, and parenchymal necrosis (Theodore, 2014). Experimental corneal inoculation of HSV-1 in BALB/c mice led to increased CA3 pyramidal cell excitability and aberrant mossy fiber sprouting in the hippocampus as well as clinical seizures (Wu et al., 2003). Remarkably, after initial infection, HSV can establish persistent latent infections in the CNS, acting as a continuous source of HSE recurrence (Zhang et al., 2020).

Concerning the neurotropic virus HHV-6, several studies and a recent meta-analysis suggest a pathogenic role of HHV-6B infection in the development of mesial TLE, especially when associated with hippocampal sclerosis and a history of febrile seizures (Wipfler et al., 2018; Bartolini et al., 2019; Wang and Li, 2021). HHV-6, which is ubiquitous and infects most people when they are children, establishes latent infections in the CNS, especially in the hippocampus and amygdala, and is associated with neurologic diseases, including TLE (Wang and Li, 2021). In a meta-analysis of studies that detected HHV-6 genomic DNA or protein in brain samples from the hippocampus of people with mesial TLE, HHV-6 DNA was detected in 19.6% of all TLE patients compared to 10.3% of all controls (P < 0.05) (Wipfler et al., 2018).

Transcriptional analysis of the amygdala in patients with hippocampal sclerosis revealed higher expression of CCL2 and glial fibrillary acidic protein (GFAP) in HHV-6 positive samples and a positive correlation between viral load and protein expression (Kawamura et al., 2015). As described above, CCL2 is a chemokine that participates in the migration and CNS infiltration of monocytes, in which HHV-6 can establish latent infection (Bartolini et al., 2019). Overexpression of GFAP and CCL2 is associated with neuronal loss and gliosis and has been previously described in resected epileptogenic tissue from the hippocampus (Xu et al., 2011). However, the casual relationship and possible pathological role of HHV-6 in TLE are yet to be elucidated. Infections with ZIKV have also been reported to cause alterations in temporal lobe structures such as the hippocampus, leading to memory and behavioral deficits and seizures (Stanelle-Bertram et al., 2018; Büttner et al., 2019; Raper et al., 2020). This will be discussed in more detail below.

Upon viral invasion of the CNS, activation of the innate and adaptive immune response is critical to control viral replication and spread (Libbey and Fujinami, 2014; Figure 3). However, an exuberant innate response to the infection may cause considerable acute bystander pathology, while failing to adequately control viral replication which may lead to persistent smoldering inflammation that results in chronic neuropathology (DePaula-Silva et al., 2021; Figure 3). In general, as illustrated in Figure 5, inflammation plays a prominent role in the mechanisms underlying increased neuronal excitability in both early and late seizures associated with virus infection (Vezzani et al., 2016). Furthermore, oxidative stress is thought to contribute to these processes (Figure 5). As shown in Figures 1, 3, 5, initiation of neuroinflammation may either be the result of neuroinvasion, host danger signal response mediated effects or both. As described above, encephalitis is defined as inflammation of parenchymal CNS tissue that occurs in response to viral replication (Vezzani et al., 2016). Once a virus enters the brain parenchyma, inflammation may result from two mechanisms that are not mutually exclusive. First, viruses may directly infect neurons leading to unconstrained neuronal lysis and death and the release of proinflammatory cytokines and cellular products that act as endogenous danger signals (such as ATP or mitochondria-derived DNA N-formyl peptides) (Vezzani et al., 2016; Di Virgilio et al., 2020; Das et al., 2021). Second, viral PAMPs may activate PRRs on microglia and astrocytes, leading to cytokine and chemokine production that recruits innate immune effectors that drive immunopathology. These inflammatory responses drive acute injury but are also associated with the formation of a residual pathological state marked by continued BBB dysfunction and injury, neuronal death, and persistent neuronal hyperexcitability, all of which may contribute to ictogenesis and epileptogenesis.

Viruses may also trigger post-infectious encephalitis or encephalomyelitis, even in the absence of neuroinvasion during the initial infection. Such delayed responses are elicited following the development of T cell- and/or antibody-mediated recognition of self epitopes (Vezzani et al., 2016; Popkirov et al., 2017; Joubert and Dalmau, 2019). Molecular mimicry, epitope spreading, and unmasking of autoreactive lymphocytes (see Figure 1) are the primary mechanisms by which infectious agents induce autoimmunity (Powell and Black, 2001; Cusick et al., 2012; Pape et al., 2019; Gupta and Weaver, 2021).

During the acute response to CNS infection, brain resident cells recruit peripheral immune cells to sites of viral infection (Manglani and McGavern, 2018). Among the acute responders, CNS infiltration of monocytes and neutrophils is a hallmark of CNS inflammation, including viral infection (Terry et al., 2012). These cells engage in several potent effector functions including the production and secretion of numerous pro-inflammatory mediators and reactive oxygen species that drive tissue damage (Terry et al., 2012). Monocytes that migrate into the infected brain also differentiate into macrophages, dendritic cells, and, arguably, microglial populations (see below). In addition to invasion of blood-borne immune cells such as monocytes and neutrophils, brain resident innate immune cells, including microglia and astrocytes, also produce proinflammatory cytokines and reactive oxygen species that contribute to inflammation and CNS injury (Figure 5).

It has been proposed that the IL-1 cytokine system may play a pivotal role in the development of fSE and mesial TLE (Dube et al., 2005; Dube et al., 2010). IL-1β is the primary cytokine responsible for mediating febrile responses in humans and it is a powerful proconvulsant implicated in ictogenesis and epileptogenesis (Dube et al., 2010; Vezzani et al., 2016; Vezzani et al., 2019). At least in part, this effect of IL-1 β is related to its suppressive action on inhibitory GABA currents and enhancement of NMDA-mediated neuronal Ca²⁺ influx, resulting in increased glutamatergic excitation (Huang et al., 2011; Mishra et al., 2012; Vezzani and Viviani, 2015). The effects of IL-1β are mediated via IL-1 receptor type 1 (IL-1R1), which is enriched in cortical and hippocampal neurons where it co-localizes and physically associates with the NR2B (GluN2B) subunit of the NMDA receptor (Vezzani and Viviani, 2015). IL-1R1 is activated by IL-1β that is released from neurons, glia, brain endothelial cells, and infiltrating monocytes following inflammasome activation (Labzin et al., 2016; Vezzani et al., 2019). Elevation of IL-1β induces robust release of other proinflammatory cytokines, including IL-6 and CXCL8 (Heida and Pittman, 2005; Vezzani et al., 2016). A recent study that examined the association between plasma cytokines and fSE in children, as well as their potential as biomarkers of acute hippocampal injury, found that levels of CXCL8 and epidermal growth factor (EGF) were significantly elevated after fSE in comparison to controls (Gallentine et al., 2017). However, individual cytokine levels were not predictive of MRI changes in the hippocampus.

The nuclear protein high mobility group box 1 (HMGB1), which is released by neurons and macrophages/monocytes in response to exogenous and endogenous inflammatory stimuli and during unconstrained cell death, is thought to play a critical role as a danger signal in virus infection-induced inflammatory responses in the CNS (Wang et al., 2006; Vezzani et al., 2016; Walker et al., 2022). Furthermore, HMGB1 has been implicated in the generation of seizures and epilepsy (Ravizza et al., 2018). As with IL-1β, TNF-α, and IL-6, HMGB1 has pro-ictogenic properties in animal models and affects neuronal function by inducing rapid post-translational changes in glutamate receptor subunit composition and/or phosphorylation (Vezzani et al., 2016). HMGB1 physiologically interacts with nucleosomes, transcription factors, and histones within the nucleus of nearly every cell type but is rapidly translocated to the cytoplasm and released following brain injury and during seizures (Jiang et al., 2020; Murao et al., 2021). Several viruses that cause encephalitis and seizures, including WNV, SARS, TBEV, and IAV, can induce the release of HMGB1 (Wang et al., 2006; Ding et al., 2021). HMGB1 binds to and activates the receptor for advanced glycation end products (RAGE), toll-like receptor 4 (TLR4), and TLR2 (Jiang et al., 2020), inducing signal transduction cascades that drive inflammation. Indeed, activation of IL-1R1 and HMGB1 receptors expressed by microglia and astrocytes orchestrates inflammatory events that result in the release of cytokines and chemokines, induction of the prostaglandin-synthesizing enzyme cyclooxygenase 2 (COX-2), and activation of the complement system, and may thereby subsequently lead to recruitment of leukocytes to the brain (Vezzani et al., 2016).

Whereas the processes illustrated in Figure 5 and discussed above would be relevant for all viruses that cause encephalitis and/or sterile inflammation, there are also neuropathophysiological processes and outcomes specific to individual viruses. For instance, the transactivator of transcription (Tat) protein is a major viral protein in HIV that can directly drive neurotoxicity (Atluri et al., 2015). Tat is vital for HIV replication and influences transcription initiation and elongation at the HIV promoter. In addition, however, Tat injures neurons via several different mechanisms, including induction of inflammatory cytokines, impairment of mitochondrial function, and activation of ionotropic glutamate receptors (Atluri et al., 2015). Indeed, Haughey et al. (2001) reported that HIV-1 Tat potentiates the excitotoxicity of glutamate by phosphorylating NMDA receptors, a process that is critically involved in neuronal hyperexcitability, seizures, and epileptogenesis (Ghasemi and Schachter, 2011; Hanada, 2020). The effect of prolonged exposure to endogenously produced Tat in the brain was investigated using a transgenic mouse model constitutively expressing the HIV-1 Tat gene (Zucchini et al., 2013). Stimulus-evoked glutamate exocytosis in the hippocampus and cortex of these mice was significantly increased and was associated with increased seizure susceptibility. In addition to the effects associated with the Tat protein, the HIV type 1 envelope glycoprotein gp120 activates macrophages, which release neurotoxins that affect the glutamate system, leading to activation of voltage-dependent calcium channels and modulation of NMDA signals (Potter et al., 2013).

Concerning the SARS-CoV-2 virus, the specific mechanisms by which this virus affects the CNS remain unclear (Pröbstel and Schirmer, 2021). As described above, infection with SARS-CoV-2 may result in psychiatric and neurological symptoms, including seizures; more than 35% of COVID-19 patients develop such symptoms, particularly during severe manifestation of the disease (Tavkar et al., 2021). It is well accepted that the entry of SARS-CoV-2 into a host cell is mediated by ACE2, which functions as an entry receptor (Hoffmann et al., 2020). Membrane-bound ACE2 is a zinc-containing metalloenzyme located on the surface of cells. SARS-CoV-2 downregulates ACE2, with a consequent loss of its catalytic activity (Pacheco-Herrero et al., 2021). Inflammation and thrombosis have been associated with enhanced and unimpeded angiotensin II effects through the ACE2-AT1 receptor axis (Pacheco-Herrero et al., 2021).

In the CNS, ACE2 is expressed in the majority of brain regions (e.g., the amygdala, cortex, frontal cortex, substantia nigra, and hippocampus) but mostly at low levels (Chen et al., 2020). Analysis of human and mouse brains showed that ACE2 is expressed predominantly in neurons but also in non-neuronal cells, including astrocytes, oligodendrocytes, endothelial cells, and pericytes (Tavkar et al., 2021). The expression of ACE2 makes CNS cells susceptible to SARS-CoV-2 infection, provided that the virus enters the brain. As summarized in Table 2 and Figure 2, current evidence points to two plausible mechanisms of brain invasion by SARS-CoV-2: (i) entry into the CNS via axonal transport along infected olfactory nerves and then dissemination via trans-synaptic transmission to other brain areas (Montalvan et al., 2020; Yachou et al., 2020; Meinhardt et al., 2021); note that as many as 65% of COVID-19 affected individuals reported hyposmia, anosmia, and ageusia, suggesting the possibility of transsynaptic spread not only via the olfactory route but also along lingual and glossopharyngeal nerves (Figure 2; Sulzer et al., 2020); (ii) entry into the CNS via a hematogenous pathway, either through the infiltration of infected blood cells (usually leukocytes) or through infection of endothelial cells at the BBB. The hematogenous pathway may also involve infection of epithelial cells of the choroid plexus, the building blocks of the blood-CSF barrier (Montalvan et al., 2020; Murta et al., 2020; Vargas et al., 2020; Yachou et al., 2020). Another intriguing mechanism via which SARS-CoV-2 may spread is through the vagus nerve from infected lungs (Jarrahi et al., 2020). Using human brain organoids derived from induced pluripotent stem cells as a valuable tool for investigating SARS-CoV-2 neurotropism, it was found that choroid plexus organoids showed a high rate of infection and supported productive viral replication, consistent with the finding that the choroid plexus exhibits high ACE2 expression (Jacob et al., 2020; Pellegrini et al., 2020). Besides epithelial cells of choroid plexus, neurons, astrocytes, and neural progenitor cells in brain organoids are also susceptible to SARS-CoV-2 infection, although the infection rates for these cell types remain under debate (Tavkar et al., 2021). Overall, replication of SARS-CoV-2 in the CNS remains a controversial issue.

Concerning the mechanisms of neurological symptoms such as seizures, many groups argue that the devastating neurological damage caused by SARS-CoV-2 is not a consequence of direct infection of neural cells but rather a result of the severe peripheral hyper-inflammation associated with COVID-19 (Pacheco-Herrero et al., 2021; Tavkar et al., 2021). Among the various consequences of such inflammation, impairment of BBB may be involved in CNS symptoms, as discussed above and illustrated in Figures 3, 5. Furthermore, it has been suggested that endothelial dysfunction in several organs, including the CNS, may be triggered by the interaction between SARS-CoV-2 and ACE2 receptors that are expressed by endothelial cells (Pacheco-Herrero et al., 2021). In patients with COVID-19, magnetic resonance imaging (MRI) detected lesions that are compatible with a cerebral small-vessel disease and with disruption of the BBB (Nassir et al., 2021).

More recently, Wenzel et al. (2021) reported structural changes in cerebral small vessels of patients with COVID-19 and elucidated potential mechanisms underlying the vascular pathology. Both in patients and two animal models of SARS-CoV-2 infection, an increase in string vessels was observed in the brain. These structures represent the endothelial cell-free remnants of lost capillaries. Furthermore, the authors also found evidence that BMECs are infected and that the death of BMECs in COVID-19 is secondary to SARS-CoV-2 infection. The SARS-CoV-2 genome encodes two viral proteases that are responsible for processing viral polyproteins into the individual components of the replication and transcription complexes. Wenzel et al. (2021) found that one of them, SARS-CoV-2 M^pro, cleaves the host protein nuclear factor (NF)-κB essential modulator (NEMO), which is known to modulate cell survival and prevent apoptosis and necroptosis.

However, other findings suggest that SARS-CoV-2-related neurological complications may be a direct result of the neurovirulent properties of the virus (Shehata et al., 2021). Overall, it has been postulated that there are several different mechanisms involved in COVID-19–associated CNS dysfunction, including activation of inflammatory and thrombotic pathways and, in a few patients, a direct viral effect on the brain endothelium and the brain parenchyma (Bodro et al., 2021). However, further studies are needed to clarify the relative contribution of each of these mechanisms. A recent landmark study used three independent approaches to probe the capacity of SARS-CoV-2 to infect the brain (Song et al., 2021). In the first, transgenic mice overexpressing human ACE2 were found to support SARS-CoV-2 neuroinvasion. After intranasal administration, the virus was widely present in neural cells throughout the forebrain. In the second, using human brain organoids, clear evidence of infection with accompanying metabolic changes in infected and neighboring neurons was found. In this study, neuronal infection could be prevented by blocking ACE2. Finally, in autopsies from patients who died of COVID-19, SARS-CoV-2 was detected in cortical neurons. Remarkably, none of the regions of positive viral staining showed lymphocyte or leukocyte infiltration, indicating that SARS-CoV-2 did not invoke an immune response typical of other neurotropic viruses. These findings provide compelling evidence that the brain is a site for the high replicative potential for SARS-CoV-2 and that neurons can become a target of SARS-CoV-2 infection, with devastating consequences for localized ischemia in the brain and cell death, highlighting SARS-CoV-2 neurotropism.

The lipid-binding protein apolipoprotein E (ApoE) is the most abundant apolipoprotein in the brain (Flowers and Rebeck, 2020). It is produced predominantly by astrocytes and to some extent microglia. In addition, neurons upregulate ApoE expression in response to excitotoxic injury (Liao et al., 2017). As a major component of very low-density lipoproteins in the brain, ApoE facilitates the transfer of cholesterol and phospholipid between cells. ApoE has been linked with immune responses and neuroinflammation, metabolism, synaptic plasticity, transcriptional regulation, and vascular function by modulating cerebral blood flow, neuronal-vascular coupling, and BBB integrity (Liao et al., 2017). There are three major isoforms (ApoE2, ApoE3, and ApoE4) in humans (Liao et al., 2017). The most common isoform (77–78%) in the general population is E3, whereas E2 is evident in 7–8%, and E4 in 14–16% of individuals (Weisgraber and Mahley, 1996). ApoE4, a strong genetic risk factor for Alzheimer’s disease, is known to lead to BBB dysfunction (Montagne et al., 2020) and has been associated with increased risk for severe COVID-19 (Kuo et al., 2020). Recently, Wang et al. (2021a) tested the neurotropism of SARS-CoV-2 in human induced pluripotent stem cell (hiPSC) models and observed low-grade infection of neurons and astrocytes. They then generated isogenic ApoE3/3 and ApoE4/4 hiPSCs and found an increased rate of SARS-CoV-2 infection in ApoE4/4 neurons and astrocytes. ApoE4 astrocytes exhibited enlarged size and elevated nuclear fragmentation upon SARS-CoV-2 infection. These findings suggest that ApoE4 may play a causal role in COVID-19 severity.

Interestingly, ApoE4 has also been associated with seizures. For instance, spontaneous seizures were observed in aged ApoE4 targeted replacement (TR) mice but not in age-matched ApoE2 TR or ApoE3 TR mice (Hunter et al., 2012). In mice with overexpression of ApoE4 (but not ApoE2 or ApoE3), intranasal administration of kainate induced more severe seizures, increased microglial activation, and triggered more hippocampal damage than in wild-type mice (Zhang et al., 2012). In a case-control genetic association study in patients with mesial TLE and hippocampal sclerosis, ApoE4 carriers had an earlier onset of epilepsy than non-carriers (Leal et al., 2017). Thus, in summary, ApoE4 may play a role in seizures observed in viral infections, including COVID-19.

As described above, it is a matter of debate whether SARS-COV-2 can enter the brain, but several studies indicate that the SARS-CoV-2 S1 protein can be released from viral membranes, can cross the BBB, and is present in brain cells including neurons (Meinhardt et al., 2021; Rhea et al., 2021; Petrovszki et al., 2022). Thus, Datta et al. (2021) tested the hypothesis that SARS-CoV-2 S1 protein can directly induce neuronal injury. The latter authors found that the S1 protein accumulates in endolysosomes of human cortical neurons and induces aberrant endolysosome morphology and neuritic varicosities, which could contribute to the high incidence of neurological disorders associated with COVID-19.

Emerging data suggest that ∼10–40% of patients fail to fully recover after acute COVID-19 infection (Doyle, 2021; Nalbandian et al., 2021). Patients who report symptoms persisting for weeks or months after the acute illness have been termed “long haulers” or described as having “long-COVID.” Long-COVID comprises a variety of symptoms, of which the neurological component prevails, often characterized as post-infectious fatigue syndrome (Sandler et al., 2021). Furthermore, new-onset seizures in people with COVID-19 can potentially extend beyond the acute phase of the infection (Doyle, 2021). The most widely accepted theory on the genesis of these symptoms builds upon the development of microvascular dysfunction similar to that seen in numerous vascular diseases such as diabetes. This can occur through the peripheral activation of ACE2 receptors or through the exacerbating effects of pro-inflammatory cytokines that remain in circulation even after the infection diminishes (Nalbandian et al., 2021). However, at least in part, some of the mechanisms of CNS symptoms discussed above for the acute infection may also play a role in post-COVID symptoms.

As discussed above, accumulating evidence suggests a pathogenic role of HHV-6B infection in the development of mesial TLE, and a relationship between viral load and markers that directly (CCL2) or indirectly (GFAP) reflect inflammatory or otherwise injurious processes. How might these observations associating mesial TLE with increased HHV-6 viral detection and increased markers of neuroinflammation and astrocyte activation be mechanistically associated with epilepsy? Inflammation and HHV-6 infection have each been demonstrated to induce dysregulation of glutamate homeostasis in astrocytes, which is hypothesized to play a central role in the pathogenesis of epilepsy (Leibovitch and Jacobson, 2015). In vitro, HHV-6 infection of primary astrocytes has been shown to downregulate levels of glutamate transporter expression, which supports the concomitant observation of decreased glutamate uptake in infected versus uninfected astrocytes (Fotheringham et al., 2008). Inflammatory cytokines, such as IL-1β, can also inhibit the astrocyte reuptake of glutamate (Vezzani and Baram, 2007). Because HHV-6–infected astrocytes have been demonstrated in mesial TLE, and because the virus can induce a metabolic dysregulation that is considered to contribute to epileptogenesis, this mechanism is biologically plausible (Leibovitch and Jacobson, 2015). Interestingly, ApoE4 has been suggested to increase viral load and seizure frequency in mesial TLE patients that are positive for HHV-6B DNA and protein in temporal lobe brain samples resected during epilepsy surgery (Huang et al., 2015).

Astrocytes exert many essential complex functions in the healthy CNS that are necessary to maintain synaptic and neural circuit homeostasis (Sofroniew and Vinters, 2010). Astrocytes respond to all forms of CNS insults, including viral encephalitis, through a process referred to as reactive astrogliosis, which has become a pathological hallmark of CNS structural lesions. Astrogliosis is a common step in the sequence of events that converts a normal brain into an epileptic brain after an acquired insult (Klein et al., 2018). Astrogliosis is involved in inflammatory processes as well as dysregulation of astroglial potassium and gap junction channels, which together alter glioneuronal communication and, by impairing uptake and redistribution of extracellular K⁺ accumulated during neuronal activity, can contribute to or cause seizures (Klein et al., 2018). Astrocytic compartmentalization of synapses also plays an essential role in neurotransmitter homeostasis by concentrating high levels of transporters for glutamate, GABA, and glycine that serve to clear these neurotransmitters from the synaptic space (Sofroniew and Vinters, 2010). During neuroinflammation, high levels of cytokines such as IL-6 lead to decreased glutamate uptake from the synaptic space by downregulating the excitatory amino acid transporter 2 (EAAT2; formerly glutamate transporter 1) on astrocytes, leading to glutamate accumulation and consequent neuronal hyperexcitability (Verhoog et al., 2020).

Within this context, it is important to note that astrocytes are thought to play a crucial role in flavivirus infections of the CNS by mediating the mechanisms that underlie neurological sequelae such as seizures and epilepsy (Potokar et al., 2019; Zheng et al., 2020; Ashraf et al., 2021). Indeed, given the anatomic position of astroglia and their homeostatic role in the CNS, one can predict that virus invasion may lead to important functional consequences for the entire CNS upon the interaction of astrocytes with viruses. Furthermore, in comparison to neurons, infected astrocytes produce orders of magnitude more virus, as demonstrated for ZIKV, TBEV, and WNV (Tavkar et al., 2021). This is highly relevant for the spread of infection through the CNS, especially because astrocytes are also more resilient to the lytic effects of flavivirus infection. Interestingly, different flavivirus strains appear to exert different effects on specific astrocyte responses (Potokar et al., 2019; Ashraf et al., 2021).

For example, TEBV, an important human pathogen that may result in dangerous neuroinfections (meningitis, meningoencephalitis, myelitis) and is endemic in Europe and Asia, replicates in astrocytes but does not typically affect astrocyte viability (Palus et al., 2014; Potokar et al., 2019). TBEV infection induces several morphologic and functional changes in infected rat and human astrocytes, including astrocyte activation as indicated by increased production of GFAP (Tavkar et al., 2021). Upon activation by TEBV infection, astrocytes release inflammatory cytokines and chemokines that may enhance neuronal excitability (Figure 5). TBEV infection of astrocytes may also alter the permeability of the BBB, as shown in mice (Ruzek et al., 2011). One of the key molecules that degrade the integrity of the BBB is matrix metalloproteinase 9 (MMP-9), which is overproduced in TBEV-infected astrocytes in vitro and increased in the serum and CSF of TBEV-infected patients (Potokar et al., 2019).

Upon WNV infection, astrocytes also release MMPs and pro-inflammatory cytokines, leading to disruption of the BBB and recruitment of leukocytes (Ashraf et al., 2021). Analysis of autopsied neural tissues from humans with WNV encephalomyelitis revealed WNV infection of both neurons and glia (van Marle et al., 2007). In human astrocytes and neurons, WNV replicates efficiently but distinctively with a higher and faster replication rate in astrocytes (Cheeran et al., 2005). Astrocytes have an active role in the spread of WNV in the CNS and in the maintenance of WNV neuroinvasive potential. Among the WNV-induced functional changes in astrocytes is the expression of endoplasmic reticulum stress-related genes linked to WNV neurovirulence (van Marle et al., 2007). WNV-infected astrocytes also upregulate the expression of several chemokines, but only after infection with the replication-competent virus and not with an inactivated virus (Potokar et al., 2019). In an experimental murine model of WNV-induced seizures, intranasal inoculation with WNV caused limbic seizures in B6 mice, but not in IFN-γ-deficient (IFN-γ^–/–) mice (Getts et al., 2007). Both strains showed similar levels of virus in the brain, as well as similar concentrations of TNF-α and IL-6, both of which alter neuronal excitability. However, TNF-α deficient mice infected intranasally with WNV still developed severe limbic seizures, similar to B6 wild-type mice (Getts et al., 2007). While the absence of seizures in the infected IFN-γ^–/– mice was shown to be associated with the influence of this cytokine on excitatory circuit development, rather than a direct effect on synaptic function, per se, the observation highlights the complicated relationship between inflammation and CNS function. Finally, in patients with WNV encephalitis, increased infiltration of monocytes into the brain was found (Ashhurst et al., 2013), which, as discussed elsewhere in this review, appears to be a common outcome of CNS infection.

In addition to the profound impact on fetuses and neonates (fetal growth restriction, abnormalities of the CNS, including microcephaly) caused by intrauterine infections with ZIKV during pregnancy, this virus can also cause neurologic symptoms in adults (Guillain-Barré syndrome, myelitis, encephalitis, and neuralgia) (Potokar et al., 2019). Following infection of immunocompetent pregnant mice with ZIKV, we found the virus particularly in glial cells, such as astrocytes, oligodendrocytes, and microglia, most profoundly in the brainstem and cerebellum of the maternal brain (Stanelle-Bertram et al., 2018). Interestingly, the male offspring from ZIKV infected mothers were more likely to suffer from impairment of learning and memory compared to females, likely as a result of more severe neuropathological alterations in the hippocampus compared to their female littermates (Stanelle-Bertram et al., 2018). Furthermore, in a study in which perinatal infection was simulated by using neonatal mice, seizures were observed following subcutaneous inoculation of 1-day-old immunocompetent B6 mice with ZIKV PRVABC59 (Manangeeswaran et al., 2016). The seizures were associated with ZIKV infection in the brain, neurodegeneration in the hippocampus and cerebellum, and infiltration of brain tissue with CD4⁺ and CD8⁺ T cells. In a study with ZIKV infection in 3-days-old Swiss mice, the animals developed frequent seizures during the acute phase, which were reduced by inhibiting TNF-α (Nem de Oliveira Souzaem et al., 2018). During adulthood, ZIKV replication persisted in neonatally infected mice, and the animals showed increased susceptibility to chemically induced seizures and neurodegeneration, predominantly in the hippocampus, thalamus, striatum, and cortex. Both cell death and impaired proliferation of neural precursors were shown to underlie ZIKV-induced neuropathology (Nem de Oliveira Souzaem et al., 2018). In a subsequent study from the same group, the effects of ZIKV infection on neuronal networks (determined from electrophysiological activity) and how different mechanisms can trigger epilepsy in ZIKV Swiss mice were examined (Pinheiro et al., 2020).

Astrocytes, together with microglia, are proposed to be major ZIKV targets in fetal brain development (Potokar et al., 2019). Primary fetal human astrocytes particularly stand out for their susceptibility to ZIKV infection in comparison with neurons and neural progenitor cells. As is the case for TBEV, astrocytes are also proposed to serve as a reservoir for ZIKV, and they apparently induce neuroinflammation through pro-inflammatory cytokines mediating synaptic and cognitive changes (Potokar et al., 2019).

As with other flaviviruses, astrocytes are also an important player in altered BBB permeability in response to JEV. Upon infection with JEV, astrocytes release vascular endothelial growth factor (VEGF), IL-6, and MMPs (Potokar et al., 2019). In addition to affecting the BBB, astrocytes are also involved in neuroinflammatory responses in the JEV-infected CNS that may underlie ictogenesis.

The family of small, positive-sense, single-stranded, non-enveloped RNA viruses known as the Picornaviridae includes numerous human pathogens with known and potential neurovirulence (Rotbart, 1995; Buenz and Howe, 2006), including members of the Enterovirus genus such as poliovirus, the echoviruses, the Coxsackie viruses, and the rhinoviruses. The global ubiquity of these viruses, the high transmissibility, and the widespread exposure experienced by children make picornaviruses an important component of emerging or re-emerging infections associated with neurological disease (Fischer et al., 2022). For example, enterovirus 71 (EV71), the causative pathogen in hand, foot, and mouth disease, was originally isolated in California in 1969 from a 9-month old girl with encephalitis (Schmidt et al., 1974). Further outbreaks of this and related serotypes occurred across the US, South America, Europe, and Asia, with hundreds of thousands of infections in Asia-Pacific countries since the 1990s and thousands of deaths due to encephalitis or encephalomyelitis (Puenpa et al., 2019). Notably, while seizures are reported in some of these patients (Bissel et al., 2015), a predominant outcome for children with neurologic manifestations is death, suggesting that neurovirulent picornaviruses induce severe neuropathology. As we and others have discussed, several picornavirus proteins, including the structural proteins VP1, VP2, and VP3 and the non-structural proteins 2A and 3C directly engage pro-apoptotic mechanisms in infected cells (Buenz and Howe, 2006) and co-opt antiviral mechanisms (Wang et al., 2018). However, seizures are clearly a component of picornaviral infections in less severe cases, including a broad propensity to febrile seizures, acute seizures, and late spontaneous seizures (Table 1).

Picornavirus neurotropism is obviously well established for human poliovirus (Whitton et al., 2005). The human poliovirus receptor CD155 is enriched in anterior horn motor neurons (Gromeier et al., 2000) and mediates cellular entry, as proven by neuronal infection and development of paralytic poliomyelitis in mice transgenically expressing CD155 (Ren et al., 1990). Other picornaviruses exploit different cellular receptors. For example, both EV71 and coxsackievirus A16 (CVA16) utilize scavenger receptor class B, member 2 (SCARB2; aka CD36L2) to enter cells. This protein, widely and highly expressed in the brain, gut, and immune system, localizes to neurons, and transgenic expression of human SCARB2 in mice renders the host susceptible to CNS infection with EV71 (Fujii et al., 2013). While the pathophysiological relevance is not clear, it is notable that mutations in SCARB2 are associated with epilepsy (Rubboli et al., 2011).

Given the broad expression of picornavirus receptors, the development of focal neurological sequelae must depend upon cell-intrinsic responses to infection or cell-specific sensitivity to innate and/or adaptive immune responses elicited by CNS infection. Concerning the former, one potential mechanism of neuronal specificity arises from the rapid and robust shutdown of host cell translation that is a hallmark of picornavirus infection (Etchison et al., 1982) and is mediated by viral protease cleavage of cap-dependent translation factor eIF-4G (Whitton et al., 2005). While cap-dependent translation is important to all cells, neurons may exhibit a unique sensitivity to translation inhibition. For example, evidence from ischemia-reperfusion models indicates that vulnerable neuronal populations in the hippocampus selectively undergo apoptosis in response to downregulated protein synthesis (Ayuso et al., 2013). Likewise, specific neuronal populations may be uniquely sensitive to the activation of stress pathways activated by translation inhibition, such as NFκB activation due to loss of IκBα translation and suppression of AKT signaling as part of the integrated stress response (Kapur et al., 2017). In parallel, suppression of glutamate transporter expression and local neuroinflammatory responses that result in the release of factors such as TNFα may combine to drive both hyperexcitability and accelerated neuronal cell death (Guerrini et al., 1995; Kaltschmidt et al., 1995; McCoy and Tansey, 2008). Finally, concerning neuron-specific sensitivity to infection-induced neuroinflammatory responses, robust evidence obtained using the mouse picornavirus TMEV, outlined below, indicates that innate immune cell-mediated acute antiviral responses lead to both neuronal cell death and dysregulation of electrophysiological homeostasis.

Based on a large series of subsequent studies of the groups of Robert S. Fujinami, H. Steve White, and Karen S. Wilcox at the University of Utah, which were reviewed by Libbey and Fujinami (2011) and DePaula-Silva et al. (2017, 2021), it was suggested that infiltrating monocytes (CD45^hi CD11b⁺) present in the brain of B6 mice at day 3 post-infection are an important source of IL-6, which critically contributes to the development of acute seizures in the TMEV-induced seizure model. Furthermore, the production of high levels of TNF-α by microglia during the acute phase of the infection was found to play a role (Cusick et al., 2013). When mice deficient in TNF receptors, TNF-α or IL-6 were infected with TMEV, the incidence of acute seizures was significantly decreased, whereas IL-1R1 deficient mice did not differ from wild-type controls (Kirkman et al., 2010; Patel et al., 2017). From these data and the known effects of IL-6 and TNF-α on neuronal activity, it was suggested that IL-6 and TNF-α secreted in the brain by infiltrated monocytes and resident microglia during TMEV infection in B6 mice may contribute to enhanced glutamatergic excitation and decreased GABAergic inhibition and lead to a more seizure prone state (DePaula-Silva et al., 2021). In support of this hypothesis, TMEV infection of B6 mice depleted of monocytes resulted in a significant decrease in the number of mice experiencing seizures, substantiating a role for infiltrating monocytes in the development of acute seizures in the TMEV-induced seizure model (Cusick et al., 2013). However, at least in part, the experimental methods used to reduce monocyte invasion and distinguish monocytes/macrophages from microglia were not specific, so the exact role and interplay of these and other immune cells in the TMEV model remained elusive.

The interesting data reported by the University of Utah groups prompted W. Löscher’s group to establish the TMEV model in B6 mice in Hannover, Germany. Unexpectedly, it took several years to reproduce the seizure phenotype in our laboratory (Bröer et al., 2016). Indeed, the BeAn strain of TMEV was used in thousands of SJL/J and B6 mice by Wolfgang Baumgärtner’s group at the Department of Pathology at the University of Veterinary Medicine in Hannover over a period of ∼15 years in numerous studies on mechanisms involved in virus-induced demyelination but seizures were never observed in B6 mice. Thus, we hypothesized that either the substrain of B6 mice or the BeAn substrain used in these experiments may have been responsible for the lack of seizures. This hypothesis was addressed by comparing two B6 and two BeAn substrains, including the mouse and virus substrains used in the original studies of Fujinami and White (Bröer et al., 2016). In addition, we compared the potency of the BeAn and DA TMEV strains to induce seizures and epilepsy in mice. The idea behind this approach was to study what is and what is not necessary for the development of acute and late seizures after brain infection in mice. Receiver operating characteristic (ROC) curve analysis was used to determine which virus-induced brain alterations are associated with seizure development. In B6 mice infected with different TMEV virus (sub)strains, the severity of hippocampal neurodegeneration, amount of MAC3-positive microglia/macrophages, and expression of ISG15 were almost perfect at discriminating seizing from non-seizing B6 mice, whereas T-lymphocyte brain infiltration was not found to be a crucial factor (Bröer et al., 2016).

The potential role of blood-borne monocyte brain invasion for the seizure phenotype induced by TMEV infection of B6 mice suggested by the University of Utah groups (Libbey and Fujinami, 2011; DePaula-Silva et al., 2017, 2018, 2021) prompted us to perform a series of studies using either genetic or pharmacological strategies. The outcome of these studies is summarized in Table 3. First, to better differentiate brain-resident myeloid cells, including microglia, from invading monocytes in the TMEV encephalitis model of TLE, we compared virus-induced effects in B6 WT vs. B6-based Cx3cr1-cre^ER±tdTomato^St/Wt mice, in which long-lived CX3CR1⁺ cells such as microglia can be distinguished from infiltrating monocytes by the expression of the red fluorescent protein tdTomato (Käufer et al., 2018a). When using flow cytometry to differentiate blood-borne monocytes (CD45^highCD11b⁺) from resident microglia (CD45^lowCD11⁺) in the brain, the Cx3cr1-cre^ER±tdTomato^St/Wt reporter mice provided qualitative proof that activated myeloid cells present in the CNS after TMEV infection consist of microglia and infiltrating monocytes (Table 3), although concerning CD45 and CD11b expression, some microglia become indistinguishable from monocytes during CNS infection (Käufer et al., 2018a).

Next, we used two pharmacological approaches to determine the impact of invading monocytes vs. resident microglia for early seizures and hippocampal damage induced by TMEV in B6 mice. When using systemic administration of liposome-encapsulated clodronate liposomes as a selective and widely used approach for monocyte depletion, almost complete depletion of monocytic cells was achieved in the spleen and blood of Theiler’s virus-infected B6 mice, which was associated with a 70% decrease in the number of brain-infiltrating monocytes as assessed by flow cytometry (Waltl et al., 2018a). As shown in Table 3, significantly fewer clodronate liposome-treated mice exhibited seizures than liposome controls. The severity of seizures was not affected by monocyte depletion, but the seizure burden (the number of seizures per mouse observed over 7 days after infection) was markedly reduced (Waltl et al., 2018a). However, the development of hippocampal damage was not prevented or reduced by monocyte depletion (Table 3).

Surprisingly, clodronate liposome treatment did not reduce the increased Iba1 and Mac3 labeling in the hippocampus of infected mice, indicating that activated microglia may contribute to hippocampal damage (Waltl et al., 2018a). Thus, our next pharmacological approach used prolonged administration of PLX5622, a specific inhibitor of colony-stimulating factor 1 receptor that depletes microglia (Waltl et al., 2018b). As shown in Table 3, microglia depletion accelerated the occurrence of seizures, exacerbated hippocampal damage, and led to neurodegeneration in the spinal cord, which is normally not observed in B6 mice. These data suggested that microglia are required early after infection to limit virus distribution and persistence, most likely by modulating T cell activation (Waltl et al., 2018b). An antiviral role of microglia has also been demonstrated for ZIKV, HSV, JEV, WNV, and several other virus infections (Terry et al., 2012; Chen et al., 2019). Interestingly, TNF-α expression in the brain of TMEV-infected mice was not affected by microglia depletion, suggesting that CNS and/or infiltrating cells other than microglia are also secreting this cytokine (Waltl et al., 2018b). More recently, our data have been partially confirmed by the University of Utah groups (DePaula-Silva et al., 2018; Sanchez et al., 2019).

In an additional series of experiments, we used genetic approaches (Ccr2-KO and Cx3cr1-KO mice) to study the role of invading monocytes vs. activated microglia for early seizures and hippocampal damage (Käufer et al., 2018a). CCR2 and CX3CR1 are two chemokine receptors that regulate the responses of myeloid cells, such as monocytes and microglia, during inflammation (Prinz and Priller, 2010). Based on their differential expression of the chemokine receptors CCR2 and CX3CR1 in mice, so-called “inflammatory” (or “classic”) monocytes (CCR2⁺CX3CR1^low), which are highly mobile and rapidly recruited to inflamed tissues, can be distinguished from patrolling (non-classic) monocytes (CCR2^–CX3CR1^high), which are larger in size and patrol along vascular endothelium such as the BBB (Prinz et al., 2011; Prinz and Priller, 2017). Brain-resident microglia produce the myelo-attractant cytokine CCL2 (also known as MCP1), a CCR2 ligand that promotes the transmigration of CCR2⁺ monocytes (and T cells) across the BBB via CCL2/CCR2 crosstalk (Prinz and Priller, 2010; Howe et al., 2017). Mice devoid of the Ccr2 gene exhibit markedly reduced recruitment of monocytes and reduced pathology in several brain disease models, including autoimmune encephalitis, MS, stroke, and status epilepticus (Prinz and Priller, 2010; Chu et al., 2014; Varvel et al., 2016). Interestingly, in SJL mice, in which infection with TMEV induces severe spinal cord demyelination (Figure 6A), the use of Ccr2-KO mice reduced monocyte infiltration, demyelination, and long-term disease severity (Bennett et al., 2007).

As shown in Table 3, in B6 mice, the lack of CCR2 or CX3CR1 receptors was associated with a significant reduction of monocyte invasion and almost complete prevention of hippocampal damage but did not prevent seizure development after viral CNS infection (Käufer et al., 2018a). These data are compatible with the hypothesis that CNS inflammatory mechanism(s) other than the infiltrating myeloid cells trigger the development of seizures during viral encephalitis. It is also important to note that the consequences of pharmacological vs. genetic manipulation of monocyte invasion and microglia activation strikingly differed (Table 3). Furthermore, the interplay between microglia and invading monocytes in this model is more complex than previously proposed by other groups (Libbey and Fujinami, 2011; DePaula-Silva et al., 2017, 2018, 2021).

All studies described thus far examined the role of various manipulations on the occurrence of early seizures and hippocampal damage in the TMEV model in B6 mice. As described above, a fraction of the mice also develops spontaneous recurrent seizures, i.e., epilepsy after a latent period of several weeks (Figure 6B). In the experiments of the Löscher group, the incidence of epilepsy was determined by continuous (24/7) video-EEG monitoring, resulting in an epilepsy incidence of 33%, while the incidence of early seizures was 77% (Anjum et al., 2018). When determining the development of epilepsy in mice following treatment with clodronate liposomes or in Ccr2-KO and Cx3cr1-KO mice, no significant difference from controls was observed (Käufer et al., 2018b; Waltl et al., 2018c). This would suggest that–as outlined above–the mechanisms underlying early and late seizures are different. In this respect, it is interesting to note that although there are significant increases in amplitude and frequency of spontaneous and miniature excitatory currents (mediated by glutamate) in hippocampal CA3 neurons recorded in brain slices prepared during the acute infection period and during chronic epilepsy 2 months after infection, the patterns of changes observed are markedly different during these two periods, suggesting that there are underlying changes in the network over time (Smeal et al., 2012). In addition to the changes in excitatory currents of CA3 neurons both during the acute infection and 2 months later shown by Smeal et al. (2012), additional experiments disclosed a decrease in CA3 inhibitory network activity (mediated by GABA) during the acute infection, but not at the 2-month time point, again suggesting different mechanisms of seizure generation during the acute infection and during chronic epilepsy (Smeal et al., 2015).

In addition to epilepsy as a long-term outcome of TMEV infection in B6 mice, these animals also exert behavioral and cognitive alterations, such as increased anxiety, decreased pentylenetetrazole seizure threshold, and impaired learning and memory (Umpierre et al., 2014; Barker-Haliski et al., 2015). Treatment of mice with minocycline, but not valproic acid, during the acute phase of the TMEV infection improved long-term behavioral outcomes in the TMEV model (Barker-Haliski et al., 2016), but epilepsy was not monitored in this study. Minocycline was used in this study to directly suppress microglial activation and overexpression of inflammatory cytokines.

In summary, our data on the TMEV model suggest that hippocampal damage is not critically involved in ictogenesis and epileptogenesis, because genetic manipulations that completely prevented the damage did not modify the incidence of early or late seizures (Käufer et al., 2018a,b). The mismatch between findings of genetic versus pharmacological manipulations in TMEV-infected B6 mice illustrated in Table 3 deserves further study.

In parallel with the elegant work of the Utah and Hannover groups, the Howe group also came to recognize the importance of infiltrating inflammatory monocytes. Our initial studies probed the neuropathological and behavioral sequelae of TMEV encephalitis in B6 mice, revealing that pyramidal neurons in the CA1 region of the hippocampus were selectively lost by 4 days after intracerebral inoculation with Daniel’s strain of TMEV and that mice tested in the Morris water maze starting at 11 days after infection exhibited a profound disruption in the ability to form spatial memories (Buenz et al., 2006). We showed that memory impairment was associated with damage to the CA1 region in two ways. First, the increasing hippocampal injury was associated with a graded loss in the ability to learn the maze; second, mice with any amount of hippocampal damage converted from a spatial memory strategy to a cue-based escape strategy. At the time, we focused on the role of neurotropic viruses in the direct killing of hippocampal neurons (Buenz and Howe, 2006) and we postulated that low-level neurovirulence amongst the human picornaviruses results in widespread erosion of cognitive reserves in humans, potentially explaining the development of memory and cognitive impairments with age in the absence of clear etiology.

Unexpectedly, however, our follow-up studies indicated that apoptosis of hippocampal neurons during acute TMEV encephalitis occurred independently of direct cellular infection (Buenz et al., 2009). Indeed, while many infected mice exhibited nearly complete loss of all CA1 neurons in the dorsal hippocampus, only a small fraction of these neurons expressed TMEV antigen before death. This is consistent with our contention that only 20–2000 cells in the brain are infected with TMEV in the immediate aftermath of inoculation (Howe et al., 2017). Moreover, we showed that CA1 neurons exhibited evidence of oxidative injury and apoptotic processes as early as 2 days after inoculation, with peak neuronal death occurring within 4 days, a timeline that is inconsistent with any effect of antiviral adaptive immune-mediated mechanisms (Dethlefs et al., 1997; Mendez-Fernandez et al., 2003). Given the virus-independent nature of the CA1 pyramidal neuron death, we next sought to protect these neurons by interfering with the apoptotic cascade. We observed that calpain was specifically activated in CA1 neurons as early as 2 days after TMEV inoculation, prompting us to treat infected animals with ritonavir, a drug designed as an HIV protease inhibitor that also suppresses calpain (Howe et al., 2016). We found that ritonavir therapy almost completely prevented the loss of CA1 pyramidal neurons, without impacting viral fitness or eventual viral clearance. Moreover, we found that calpain inhibition preserved cognitive performance in the Morris water maze, protected novel object recognition learning, and completely prevented the development of acute, high Racine score seizures. Critically, this therapeutic effect was achieved even when therapy was started at 36 h after inoculation, a timepoint at which mice already exhibit low Racine score events and encephalitis is well established (Howe et al., 2016).

In parallel with the ritonavir work, we published two studies showing that inflammatory monocytes are the primary driver of hippocampal injury and cognitive impairment in the TMEV model. In the first, we showed that inflammatory monocytes infiltrate the TMEV inoculated brain within hours (Howe et al., 2012a)–indeed, in our most recent work we have observed these cells in the hippocampus within 3 h of inoculation. We defined these cells as CD45^hiCD11b⁺ cells that are positive for Ly6C and Ly6B but negative for Ly6G. We also established that the LysM:GFP mouse generated by David Sacks (Faust et al., 2000) (not based on the LysM-Cre line) permitted the clear delineation of infiltrating inflammatory monocytes (GFP^mid), infiltrating neutrophils (GFP^hi), and microglia (GFP^neg). Furthermore, we showed that immunodepletion of monocytes but not neutrophils preserved cognitive performance in the Morris water maze and protected the hippocampus from injury.

In the second study, we showed that despite equivalent viral load and acute encephalitis, SJL mice do not exhibit any injury to the hippocampus and this effect was genetically dominant, as the F1 offspring of SJL × B6 mice also showed hippocampal preservation (Howe et al., 2012b). Relevant to the discussion above regarding the direct viral killing of neurons, we also observed large numbers of intact CA1 pyramidal neurons loaded with TMEV antigen at 3 days after inoculation; these neurons eventually clear the virus non-lytically and remain intact. Strikingly, we found that SJL mice exhibited a markedly truncated inflammatory monocyte response at 24 h after inoculation, while neutrophil infiltration levels were the same or greater than B6 mice at the same timepoint. The B6 response profile was recapitulated in B10.S mice (Patick et al., 1990) (a C57BL/10 congenic line that expresses H-2^s and is therefore histocompatible with SJL mice) and we used bone marrow reconstitution to create chimeric animals with a B10.S nervous system and SJL immune system or an SJL nervous system with a B10.S immune system. We found that reconstitution of SJL mice with a B10.S immune system resulted in robust inflammatory monocyte infiltration and consequent hippocampal injury that was indistinguishable from B10.S mice reconstituted with B10.S bone marrow. Finally, we showed that adoptive transfer of Ly6C⁺Ly6G^– B6 peritoneal exudate monocytes (induced by mineral oil) into B6 × SJL F1 hosts at 18 h after TMEV inoculation led to profound hippocampal injury and abrogation of scent-based novel object recognition learning.

Given the acute timing of the inflammatory monocyte response and the rapid initiation of hippocampal injury and behavioral seizures in B6 mice, we sought to identify the molecular and cellular sources driving leukocyte recruitment to the CNS. We found that by 3 h after intracranial inoculation of TMEV the hippocampus exhibited a profound upregulation of inflammatory chemokine transcripts that was quickly followed by upregulation of inflammatory cytokine RNA (Howe et al., 2017). Moreover, we observed that serum CCL2 levels peak at 3 h after infection and this was temporally associated with high levels of CCL2 in the brain and hippocampus. Genetic deletion of CCR2 essentially abrogated inflammatory monocyte infiltration, while systemic immunodepletion of CCL2 but not CCL7 also truncated the monocytic response during acute TMEV encephalitis. Unexpectedly, we found that CCL2 was predominantly expressed by hippocampal neurons at 6 h after TMEV inoculation and we showed that neuron-specific deletion of CCL2 (Syn-Cre × CCL2-RFP^fl/fl) resulted in complete suppression of serum and hippocampal CCL2 levels at this timepoint and greatly attenuated inflammatory monocyte infiltration at 24 h after inoculation.

Finally, we have recently determined that the size of the inflammatory monocyte response during acute TMEV encephalitis effectively controls the extent of hippocampal injury, the loss of spatial learning, and the induction of high-grade Racine score behavioral seizures in B6 mice (Howe et al., 2022). In this work, we used different amounts of initial TMEV inoculum to drive different levels of encephalitis. We found that introducing 12,500 plaque-forming units of TMEV into the brain elicited encephalitis at 24 h, which was 90% less intense than our standard inoculum of 200,000 plaque-forming units in terms of absolute numbers of infiltrating inflammatory monocytes. While at first glance this seems obvious, it is critical to note that at 24 h the total load of infectious virus in the brain was equivalent between the two inocula. This means that the initial viral exposure, not the amount of replicated virus, set the pace for the downstream encephalitic response. Indeed, animals inoculated with the lower amount of virus exhibited essentially no increase in CCL2 at 24 h and exhibited no increase in TNFα or IL6 in the hippocampus at 24 or 72 h after inoculation. These mice had limited hippocampal injury and showed complete preservation of spatial learning in the Barnes maze. Assessment of behavioral seizures through the first 10 days after inoculation revealed that the low virus group exhibited no high-level Racine seizures at any timepoint. EEG analysis confirmed reduced ictal activity. Notably, however, the low virus group still developed low-grade Racine seizures and did exhibit EEG abnormalities. Looking at microglial activation in these animals, we found that there was equivalent upregulation of Iba-1⁺ microglia in the hippocampus between viral inocula and these microglia showed equivalent upregulation of activation markers such as CD44. Within this context, we also found that mice inoculated with the lower amount of TMEV were more resistant to kainic acid-induced status epilepticus at 24 h relative to mice receiving the standard inoculum, but were still more sensitive than uninfected mice. These findings suggest that microglial activation acts as a binary switch during acute CNS viral infection while the infiltrating monocyte response (and encephalitis, stricto sensu) is graded. Moreover, while microglial activation during CNS viral infection primes the brain for ictogenesis, the full induction of acute clinical-grade seizures requires infiltration of inflammatory monocytes. This may have profound implications for considerations of ictogenesis during viral encephalitis. For example, even a small amount of viral invasion into the CNS may trigger microglial activation that confers a decrease in seizure threshold without rising to the level of clinical manifestations. In the context of a patient with other factors predisposing to ictogenesis, this microglial effect may be sufficient to push the system over into a seizure state. Likewise, the critical role of inflammatory monocyte infiltration in driving seizures during viral encephalitis provides an opportunity to consider therapeutic approaches that prevent these cells from invading the CNS. In numerous experiments over many years, we have never observed a detrimental effect on viral clearance associated with suppression of inflammatory monocyte responses, while we have repeatedly observed neuroprotective effects of reducing inflammatory monocyte infiltration. This leads us to strongly favor the development of new therapies to inhibit these cells or the application of unconventional therapies such as monocyte adsorption apheresis in patients with viral encephalitis.

Following the acute innate response to intracerebral inoculation with TMEV, a robust adaptive response is mounted, leading to infiltration of virus-specific CD8⁺ cytotoxic T lymphocytes, which play a significant role in viral clearance from the host (Libbey and Fujinami, 2011). In B6 mice, the earliest wave of anti-viral T cells arrives in the brain around 4 days after inoculation, peaking around day 7 (Deb and Howe, 2008). This response is marked by nearly complete restriction to recognition of a peptide derived from the VP2 capsid protein presented on the D^b MHC class I molecule (Johnson et al., 1999; Howe et al., 2007). While antiviral CD8⁺ T cells recognize and kill infected cells, within the CNS it is vital to host survival and function to clear virus non-lytically via mechanisms such as IFNγ (Rodriguez et al., 2003). The involvement of CD8⁺ cytotoxic T lymphocytes and viral clearance in the development of acute seizures in the TMEV-induced seizure model was assessed through the use of OT-I transgenic mice (B6 background), in which the majority of the CD8⁺ T cells carry an ovalbumin-specific T-cell receptor (Kirkman et al., 2010). The number of TMEV-infected OT-I mice experiencing acute seizures was comparable to wild-type B6 mice, suggesting that the seizures were not influenced by TMEV-specific CD8⁺ T cells. In TMEV-infected B6 mice, the acute symptomatic seizures resolve by ∼day 10 post-infection (Figure 6B), which was also observed in OT-I mice, indicating that the cessation of seizures was also not due to the clearance of virus by the CD8⁺ T-cell response (Kirkman et al., 2010). Similarly, in our studies, we found no significant correlation (by ROC analysis) between T lymphocyte brain infiltration and acute symptomatic seizures (Bröer et al., 2016). More recently, RAG1^–/– mice, which are deficient in mature T and B cells, were compared with B6 mice infected with TMEV (DePaula-Silva et al., 2018). As expected, CD4⁺ and CD8⁺ T cells were absent from the brains of RAG1^–/– mice, but the number of RAG1^–/– mice experiencing seizures was similar to control mice, further substantiating that lymphocytes are not playing a role in the development of acute seizures following TMEV infection (DePaula-Silva et al., 2018).

When we depleted microglia by prolonged treatment with PLX5662 from 21 days before to 6 or 7 days after TMEV infection of B6 mice, an unfavorable hippocampal and spinal cord ratio between Tregs and effector T cells was observed, thus reducing antiviral immunity in these regions (Waltl et al., 2018b). This possibility was substantiated by a marked increase in brain mRNA expression of the immunosuppressive cytokine IL-10 in the brain of infected PLX5622-treated mice, which is released by Tregs and suppresses the activation of cytotoxic T cells. These data thus added to the concept of microglia–T cell crosstalk (Schetters et al., 2017). Recently, it has been proposed that a dysregulated microglia-T-cell interplay during viral infection may result in altered phagocytosis of neuronal synapses by microglia that causes neurocognitive impairment (Chhatbar and Prinz, 2021).

The role of glutamate receptors and transporters in the TMEV model in B6 mice was studied by the University of Utah groups. Based on previous data from these groups indicating that the inflammatory cytokines IL-6 and TNF-α play a role in seizure development in the TMEV model and that infiltrating monocytes are major producers of these cytokines, the potential role of the metabotropic glutamate receptor 5 (mGluR5) was examined (Hanak et al., 2019). mGluR5 is a G-protein coupled receptor that has been shown to reduce IL-6 and TNF-α production in microglia and macrophages and to provide neuroprotection in other disease models (Loane et al., 2014; Zhang et al., 2015). Hanak et al. (2019) found that pharmacological stimulation of mGluR5 with the selective positive allosteric modulator VU0360172 not only reduced acute seizure outcomes in TMEV-infected B6 mice but also reduced the percent of microglia and macrophages producing TNF-α 3 days post-infection. Immunofluorescence confocal imaging showed a significant decrease in mGluR5 immunoreactivity in the CA1 and CA3 regions of the hippocampus with no significant changes seen in the dentate or cerebral cortex (control brain region) in TMEV-infected B6 mice with seizures compared to controls (Hanak et al., 2019).

Concerning ionotropic glutamate receptors, i.e., NMDA, kainate, and AMPA receptors, Libbey et al. (2016) determined the effects of three antagonists, MK-801, GYKI-52466, and NBQX, on acute seizure development in the TMEV-induced seizure model in B6 mice. Surprisingly, they found that only the AMPA receptor antagonist NBQX affected acute seizure development, resulting in a significantly higher number of mice experiencing seizures, an increase in the number of seizures per mouse, a greater cumulative seizure score per mouse, and a significantly higher mortality rate among the mice. This proconvulsant effect of NBQX observed in the TMEV-induced seizure model was unexpected, because NBQX has previously been shown to be a potent anticonvulsant in a variety of animal seizure models (Catarzi et al., 2007).

In another study, the role of glutamate transporters was examined (Loewen et al., 2019). Glutamate transporters such as GLT-1 expressed by glial cells contribute significantly to the control of extracellular glutamate levels, and the expression profile and function of these glutamate transporters have been implicated in epilepsy (Peterson and Binder, 2020). TMEV-infected seizing B6 mice show evidence of reactive astrogliosis, which has been associated with decreases in glutamate transporter expression and function in sclerotic tissue (Proper et al., 2002; Dossi et al., 2018). However, pharmacological and genetic methods used to modulate the glial glutamate transporters, while effective in other models, were not sufficient to reduce the number or severity of behavioral seizures in TMEV-infected B6 mice (Loewen et al., 2019).

Overall, the TMEV encephalitis model of acute and late seizures in B6 mice is a powerful tool for testing the inflammatory mechanisms that drive ictogenesis. In addition to providing a robust platform for manipulating the host immune response during viral encephalitis to alter seizure biology, the TMEV model also serves as a new, biologically and clinically relevant platform for testing established and novel therapeutics in the context of seizures that evolve without the introduction of ictogenic pharmacological agents or electrical stimulation (Metcalf et al., 2021). While a tremendous amount of progress has been made in understanding the acute phase of TMEV encephalitis, much work remains to discover the cellular and molecular mechanisms that link inflammation to ictogenesis and, critically, to identify the mechanisms that lead from the acute phase of the disease to the development of late, spontaneous seizures (DePaula-Silva et al., 2021).

In principle, the use of ASMs does not differ between early and late seizures, but control of acute symptomatic seizures during viral infection requires simultaneous treatment of the underlying etiology (Koppel, 2009; Gunawardane and Fields, 2018; Löscher and Klein, 2021a). Preferred medications for the treatment of acute symptomatic seizures or status epilepticus are those available for intravenous use, such as benzodiazepines, fosphenytoin or phenytoin, valproate, levetiracetam, and phenobarbital (Gunawardane and Fields, 2018). Prophylactic treatment with ASMs should be started as soon as possible after the onset of infection and continued as long as needed (Singhi, 2011). If not adequately treated, early seizures may progress to status epilepticus. However, prophylactic ASMs are not administered in all cases after onset of infection. Instead, ASMs are often withheld until the first-post-infection seizure.

Although the use of ASMs after the onset of infection appears clinically plausible, there is insufficient evidence to support or refute the routine use of ASMs for the prevention of early seizures in viral encephalitis. Concerning other causes of early seizures, only a few placebo-controlled clinical studies on the treatment of early seizures exist. In a study in which treatment with intravenous phenytoin or placebo was started within 24 h of traumatic brain injury, 3.6% of the patients assigned to phenytoin had seizures between the onset of treatment and day 7, as compared with 14.2% of patients assigned to placebo (P < 0.0001; Temkin et al., 1990). Although phenytoin is the standard of care to prevent acute symptomatic seizures after brain injury, a recent meta-analysis of clinical studies indicated that levetiracetam has similar efficacy to phenytoin in preventing such seizures (Zhao et al., 2018).

Concerning the treatment of febrile seizures in children, most febrile seizures are self-limited (simple febrile seizures); however, when seizures last longer than 5 min (complex febrile seizures or fSE), a benzodiazepine should be administered to break the seizure (Löscher and Klein, 2021a). A 2018 Cochrane review concluded that intravenous lorazepam and diazepam have similar rates of seizure cessation and respiratory depression (McTague et al., 2018). When intravenous access is unavailable, buccal midazolam or rectal diazepam is acceptable.

In contrast to febrile seizures, FIRES is very difficult to treat. Treatment modalities for these patients include, among others, various ASMs, ketogenic diet, intravenous corticosteroids, intravenous immunoglobulin, and burst-suppression coma (Hon et al., 2018). More recently, based on our initial report of efficacy in a child with FIRES (Kenney-Jung et al., 2016; Clarkson et al., 2019), the IL-1R antagonist anakinra has been increasingly used in the treatment of these patients, with mixed results (Farias-Moeller et al., 2018; Löscher and Klein, 2021a; Yamanaka et al., 2021). Furthermore, the IL-6 receptor antagonist tocilizumab has been used as an alternative (Stredny et al., 2020; Yamanaka et al., 2021).

In the TMEV model in B6 mice, the efficacy of various ASMs and anti-inflammatory compounds to suppress acute symptomatic seizures has been tested. In studies by Barker-Haliski et al. (2015, 2016), valproate, but not carbamazepine or minocycline, reduced the acute seizure burden. In a recent study by Metcalf et al. (2021), several prototype ASMs were effective, including lacosamide, phenytoin, ezogabine, phenobarbital, tiagabine, gabapentin, levetiracetam, topiramate, and valproate. Of these, phenobarbital and valproate had the greatest effect (>95% seizure burden reduction). The prototype anti-inflammatory drugs celecoxib, dexamethasone, and prednisone also moderately reduced seizure burden. Furthermore, cannabidiol reduced seizures in the TMEV model (Patel et al., 2019). The TMEV model in B6 mice is currently utilized by the NIH/NINDS-funded Epilepsy Therapy Screening Program (ETSP) as a tool for evaluating the antiseizure effect of novel compounds (Wilcox et al., 2020).

However, it is important to note that transient treatment of acute symptomatic seizures with ASMs has no proven effect on the subsequent development of epilepsy (Löscher and Klein, 2021a). Similarly, it is not entirely clear whether treatment of FIRES with anakinra improves the long-term prognosis of affected patients (Yamanaka et al., 2021), though several studies report adequate seizure control with ASMs as long as anakinra therapy is maintained (Kenney-Jung et al., 2016; Yamanaka et al., 2021).

Prevention of epilepsy after neurological insults such as viral encephalitis is an unmet clinical need (Pitkänen and Lukasiuk, 2011; Löscher et al., 2013; Löscher and Klein, 2021a). In principle, there are at least three strategies to prevent epilepsy after brain infections or other epileptogenic brain insults (Vezzani et al., 2016), (1) prevention of the initial insult; (2) initial insult modification (diminishing the long-term consequences of the insult by reducing the severity or duration of the initial brain insult); and (3) “true” antiepileptogenesis or disease modification after the insult by interfering with the mechanisms underlying epileptogenesis. Ad 1, wearing masks in public, maintaining a safe distance from others, and getting vaccinated are examples of prevention of the initial insult, i.e., the infection. Ad 2, intervention with appropriate treatment of the CNS infection that modifies the initial insult and thereby reduces the risk of long-term consequences. A novel emerging strategy is targeting phosphatidylserine receptors such as the TAM tyrosine kinase receptor family (TYRO3, AXL, and MERTK), which confers potent protection against neuroinvasion and CNS lesion development during neuroinvasive virus infection (Wang et al., 2021b). Our findings showing preservation of the hippocampus and inhibition of ictogenesis in mice treated with a calpain inhibitor during the acute phase of TMEV encephalitis is another potential example (Howe et al., 2016). Ad 3, antiepileptogenesis or disease modification after the infection includes treatments that directly target the complex mechanisms underlying epileptogenesis and/or actively repair damaged neural circuits.

As discussed above, the available experimental evidence supports the idea that inflammation in the brain caused by viral infections may contribute to acute seizures and epilepsy development using partially overlapping mechanisms. This highlights the possibility of identifying common targets for therapeutic interventions which may not only suppress the symptoms of the disease but also interfere with key pathogenic mechanisms. Thus, one can envisage the use of specific anti-inflammatory drugs blocking the key pathogenic inflammatory mechanisms (Vezzani, 2014, 2015). The advantage of this approach is that some of these drugs are already available in the clinic for the treatment of auto-inflammatory or autoimmune diseases, such as the IL-1R antagonist anakinra, the IL-1α antagonist canakinumab, the IL-6 receptor antagonist tocilizumab, and the TNF-α antibody infliximab or the TNFα receptor fusion protein etanercept. Preclinical research is important to determine the therapeutic potential of such novel treatments. The challenge, however, is to design an intervention that blocks the detrimental arm of brain inflammation without interfering with the homeostatic mechanisms; in this context, the implementation of resolving anti-inflammatory mechanisms rather than the prevention of the inflammatory cascade may be a better strategy to avoid complications surrounding viral control and clearance (Vezzani et al., 2016). Preventing inflammation may also be difficult due to the rapid onset and amplification of the pathogenic cascade after the first inciting event.

Animal models are important in the search for novel treatments that provide disease-modifying efficacy after viral infections. In a study by Barker-Haliski et al. (2016), in which TMEV-infected B6 mice were treated during the acute phase with minocycline to suppress microglial activation and overexpression of inflammatory cytokines, minocycline improved long-term behavioral outcomes and normalized seizure threshold. However, because late spontaneous seizures are relatively rare in the TMEV model (Figure 6B), necessitating continuous (24/7) video-EEG monitoring of large groups of infected mice, the potential effect of minocycline on the development of epilepsy was not examined by Barker-Haliski et al. (2016). As discussed above, we used continuous video-EEG monitoring in the TMEV model in B6 mice to determine the potential antiepileptogenic effect of pharmacologic and genetic manipulations interfering with monocyte invasion and found no significant effects on the development of epilepsy (Käufer et al., 2018b; Waltl et al., 2018c). The disparity between this finding and evidence that inflammatory monocytes contribute to early ictogenesis further underscores the need for more extensive research in this arena.

Febrile seizures in children often, but not always, occur in the context of an ongoing systemic virus or bacterial infection (Vezzani et al., 2016). This clinical setting has been reproduced in immature rodents by systemic administration of lipopolysaccharide (LPS) to mimic Gram-negative infections or poly(I:C) to mimic viral infections. This acute challenge imposed in a specific developmental window (postnatal day 7–14) increases the susceptibility to subsequent seizures induced by kainate, pilocarpine, or pentylenetetrazole (Galic et al., 2009; Riazi et al., 2010; Galic et al., 2012). The increased susceptibility to provoked seizures was maintained in the animals until adulthood. Systemic inflammation, not the fever per se, transiently induced IL-1α and TNF-α in the hippocampus and neocortex, and prevention of this brain response with minocycline, or using an IL-1R antagonist (anakinra) or TNF-α inactivating antibodies, precluded both the acute and long-term reduction in seizure threshold, as well as the comorbidities (anxiety-like behavior and learning and memory deficits) observed in adulthood (Riazi et al., 2010; Galic et al., 2012). The transient inflammatory challenge permanently altered the expression of glutamate receptor subtypes and the Na-K-2Cl co-transporter NKCC1 in the rat forebrain, which may have implications for the observed long-term pathophysiological outcomes (Reid et al., 2013; Riazi et al., 2015).

Interestingly, transient treatment with anakinra after brain injury has been reported to provide disease-modifying or antiepileptogenic effects in non-infectious post-status epilepticus models of acquired epilepsy (Noé et al., 2013; Dyomina et al., 2020). Similarly, the immunomodulator fingolimod, which is established as an MS therapy, was reported to exert antiepileptogenic, neuroprotective effects, and anti-inflammatory effects in post-status epilepticus models of TLE (Gao et al., 2012; Pitsch et al., 2019). The main pharmacologic effect of fingolimod is immunomodulation of lymphocyte homing, thereby reducing the numbers of T and B cells in circulation and, as a consequence, reducing lymphocyte migration into the CNS (Chun et al., 2019). In addition, fingolimod acts on CNS resident cells and inhibits the activation of astrocytes and microglia (Bascunana et al., 2020). Thus, this drug may be an interesting candidate for epilepsy prevention studies in viral encephalitis models, including the TMEV model.

However, as shown in Figure 4, epileptogenesis is a complex multifactorial process, so it seems unlikely that affecting neuroinflammation alone will be sufficient to halt this process. We have proposed previously that multi-targeted cocktails of drugs may provide a more effective strategy for epilepsy prevention after brain insults (Löscher, 2021; Löscher and Klein, 2021b). Proof of concept of this strategy has been achieved for the intrahippocampal kainate mouse model of TLE (Schidlitzki et al., 2020; Welzel et al., 2021). However, it remains to be evaluated whether this strategy is also viable for epilepsy developing during viral infections.