Pathogenesis of flavivirus encephalitis.

Introduction

Within the flavivirus family, viruses that cause natural infections of the central nervous system (CNS) principally include members of the Japanese encephalitis virus (JEV) serogroup and the tick-borne encephalitis virus (TBEV) serocomplex. Neuroinvasion follows infection of host organisms in the periphery by bites of chronically infected mosquito and tick vectors. Syndromes that result from CNS infection in humans range from mild aseptic meningitis to acute encephalitis of variable morbidity and mortality and are often complicated by neurologic sequelae among survivors. The pathogenesis of these diseases involves complex interactions of viruses, which differ in neurovirulence potential, and a number of host factors, which govern susceptibility to infection and the capacity to mount effective antiviral immune responses both in the periphery and within the CNS. Animal models have been instrumental for providing insight into how virus-specific and host factors influence the course of disease. Rodent models have been used in classic experiments on pathogenesis and continue to be relied upon for studies of viral neurovirulence determinants and immune system requirements for a successful antiviral response, particularly because of readily available knockout strains. Nonhuman primates have also been useful for some studies of peripheral immune responses to encephalitic viruses and for quantitating the disease burden caused by replication of virus in the CNS (monkey neurovirulence testing) and have also been applied as models for immunization and challenge. This review summarizes progress in the field of flavivirus neuropathogenesis since previous reviews on this topic (Monath 1986, Monath 1996). Mosquito-borne and tick-borne viruses are considered together, although it is important to note that there are differences in the pathogenesis of these two groups of viruses.

HostFactors

It is generally acknowledged that the ratio of apparent to inapparent infections with flaviviruses is quite low (on the order of 1:100 to 1:300), implying that a number of host factors are involved in protection against CNS disease. The most well known of these are age, genetic factors, and preexisting flavivirus immunity. The effect of age has been recognized in both clinical and experimental studies of flavivirus encephalitis (Eldadah 1967a, Grossberg 1966, Luby 1967, O'Leary 1942, Powell 1978, Weiner 1970), in some cases accompanied by effects of gender (Andersen and Hanson, 1974). In terms of human infections, clinical disease with JEV is primarily a pediatric entity, suggesting that certain features of the developing nervous system predispose to pathogenesis of encephalitis. Experimental evidence in support of this is based on the known propensity of arboviruses to infect and spread more rapidly through nervous tissue of young rodents; however, resistance to fatal infection occurs abruptly soon after the weanling stage. This resistance occurs in conjunction with neural ontogeny and is associated with restricted replication of virus in neurons as a function of their degree of differentiation, with low levels of virus being observed in mature neurons in some models (Hase 1993, Ogata 1991). However, this phenomenon is very dependent on the intrinsic level of viral neurovirulence, being most apparent with strains of lower virulence (Eldadah 1967a, Ogata 1991, Oliver 1997), and is also affected by the dose and route of inoculation (Eldadah 1967a, Fitzgeorge 1980). Changes in the expression of cellular receptors and⧸or intracellular factors required for efficient replication have been invoked as explanations for the age-related resistance. In support of these hypotheses, it is known that the replication block can be overcome by the neuroadaptation of flaviviruses, which results in genetic changes in the envelope region, as well as in nonstructural regions of the genome (Chambers 2001, McMinn 1997, Schlesinger 1996).

Studies with the mouse model of Sindbis virus encephalitis suggest that the differential expression of neural genes in young versus older mice is the determinant of age-related resistance. Candidate genes involved in this process include apoptotic regulators, interferon-responsive genes, and other classes that are regulated developmentally (Labrada ; see Section IX). These factors are presumed to operate by activating innate antiviral effector systems in neurons and promoting survival of these cells after viral infection.

Despite these experimental data, age-dependent resistance to encephalitic viruses in areas endemic for human disease is influenced by the effects of immunization and recurrent inapparent infections with homologous and heterologous viruses, such that the cumulative immune responses in adults may confer protection or ameliorate the severity of disease (Kurane 2002, Solomon 2002). However, susceptibility of young children (below the age of 9 months) to yellow fever virus (YFV) 17D vaccine-associated encephalitis is well documented (Freestone, 1994), and resistance to this adverse event in older individuals is clearly not based on immunity from recurrent subclinical infections. This supports the phenomenon of age-related susceptibility to encephalitis. A predisposition of elderly individuals to develop encephalitis from West Nile Virus (WNV) and St. Louis encephalitis (SLE) virus has also been observed. In this case, lack of previous cross-reactive immunity may be an important factor, but comorbid illnesses may contribute to the risk of complications, and the decline in immune function with advanced age is also likely to play a role (Grubeck-Loebenstein and Wick, 2002). In support of this, immunocompromised individuals encountering flavivirus infections appear to be less able to mount effective immune responses, as in reports of HIV-infected subjects with complicated CNS disease (Neogi 1998, Okhuysen 1993, Szilak 2000, Wasay 2000) and in transplant recipients sustaining WNV infection (Iwamoto ). YFV 17D vaccine appears to be tolerated in HIV infection, provided immunosupression is not severe, although this isuue deserves further investigation (Kengsakul 2002, Receveur 2000).

Coexistent infection with other infectious agents has been suggested as a modifying factor for flavivirus encephalitis. The association between cysticercosis and JEV infection was investigated in a controlled study that did not provide evidence for predisposition to viral encephalitis (Azad ). Association of TBEV encephalitis with borrelia infection has been described (Korenberg ), but probably reflects coexistence of these pathogens in the tick vector. Autopsy studies have suggested that herpes simplex infection may predispose to JEV encephalitis by altering the integrity of the blood–brain barrier (Hayashi and Arita, 1977). Experimentally, infection of mice with Toxocara canis or Trichinella spiralis predisposes to JEV encephalitis as a result of T-cell immunosuppression (Cypess 1973, Gupta 1987, Lubiniecki 1974, Pavri 1975), suggesting the possibility of increased disease severity in endemic areas where parasitic pathogens and encephalitic viruses coexist.

The susceptibility of mice to flavivirus encephalitis is controlled genetically and is associated with host factors that map to chromosome 5 at the oligoadenylate synthetase (OAS) gene cluster (Mashimo 2002, Perelygin 2002, Sangster 1994). Although OAS is involved in the activation of RNase L, the mechanism of resistance associated with the gene is not known and could conceivably relate to other potential functions of OAS proteins in cellular responses to viral injury (Samuel, 2002). The importance of the OAS gene in human susceptibility to these viruses requires investigation, as differences exist between murine and human gene clusters. Evidence shows that 2′, 5′-OAS is activated in response to peripheral flavivirus infection in humans (Bonnevie-Nielsen 1989, Bonnevie-Nielsen 1995), but studies to correlate this with the effectiveness of the antiviral responses are not yet available.

The effect of different class I HLA-A and B alleles on the immune response to flavivirus infection has been investigated for dengue viruses, where associations have been found with increased and decreased susceptibility to dengue hemorrhagic fever (Loke 2001, Stephens 2002), suggesting a relationship between the extent of T-cell activation and severe disease. It is unclear whether similar results will be found for encephalitic viruses, as these associations occurred in the context of secondary infections and presumably affect the strength of the memory rather than the primary T-cell response.

Interactions between flavivirus untranslated regions (UTRs) and intracellular proteins indicate another potential level of host-mediated control over virus infection (Brinton, 2000). Multiple proteins have been reported to bind to the 3′ UTR of the positive strand of various flaviviruses, including EF-1α and Mov34, which bind to the positive strand 3′ UTR of WNV and JEV, respectively (Blackwell 1995, Blackwell 1997, Ta 2000), and as many as eight cellular proteins in the case of dengue (De Nova-Ocampo ). Several proteins have also been reported to bind to the 3′ UTR of the negative strand of dengue virus (Yocupicio-Monroy ) and four proteins in the case of West Nile virus (Li ). These proteins have been suggested to participate in flavivirus replication by influencing viral transcription and⧸or translation in conjunction with host intracellular membranes. There is some evidence that interactions of the viral RNA with such proteins are in fact important for virus replication (Li ). It is not known whether compartmentalization of their interactions in different cell types is a determinant of tissue tropism in infected hosts.

ArthropodFactorsAffectingPathogenesis

It has become evident from studies on arthropod vectors that certain components of their saliva influence pathogenesis in vertebrate hosts. For mosquitoes, flaviviruses are deposited principally in the extravascular tissue during probing, as virus that is injected intravascularly is reingested rapidly during the blood meal (Burke 2001, Turell 1992, Turell 1995). Instead of a rapid dissemination of virus via the bloodstream, flaviviruses may then undergo some replication locally in subcutaneous tissues accompanied by spread to regional lymph nodes through lymphatic channels. Dendritic cells in the skin are likely to serve as a vehicle for the transport of virus to lymphoid tissues.

Components of the insect saliva modulate the earliest steps in flavivirus infection by altering the local host immune response. Feeding by Aedes aegypti or Culex pipiens mosquitoes or administration of sialokinin-I, a mosquito salivary protein, downregulates interferon-γ (IFN-γ) production and upregulates the TH2 cytokines interleukin (IL)-4 and IL-10 (Zeidner ). Salivary gland extracts from Dermacentor and Ixodes ticks decrease natural killer activity (Kubes 1994, Kubes 2002), suppress the antiviral actions of interferons in cell culture (Hajnicka ), and enhance the transmission of TBEV in rodents (Labuda ). Overall, these findings suggest that insect factors facilitate flavivirus transmission by interfering with aspects of both innate and adaptive responses. Whether these effects are required to establish systemic infection and neuroinvasion is not known.

ExtraneuralInfection

Flavivirus neuropathogenesis involves both neuroinvasiveness (capacity to enter the CNS) and neurovirulence (replication within the CNS) (Monath, 1986), both of which can be manipulated experimentally. In rodent models, neurovirulence is an inherent property of most of these viruses, and the quantity of virus needed to cause infection in the CNS is usually quite small. In classic studies of arbovirus pathogenesis (Albrecht 1968, Huang 1963), distinctions were made among neuropathogenic phenotypes on the basis of replication efficiency and pathogenic potential in the peripheral tissues versus the CNS, with various phenotypes being distinguished (high peripheral susceptibility, low neurotropism; low neuroinvasiveness, high neurotropism; and high neuroinvasiveness, high neurotropism). These phenotypes were related to different clinical outcomes, which ranged from inapparent infection to acute encephalitis of varying severity, and are influenced in the rodent model by host age and species (Kozuch ). These concepts have held up over time and have been supplemented by additional data concerning viral determinants of virulence and host innate and adaptive immune mechansisms. A main principle that applies is the relationship between peripheral virus burden and the propensity to cause neuroinvasion. Viruses with a low capacity to replicate in the periphery generally can be classified as low in neuroinvasive potential, regardless of their intrinsic level of neurovirulence. A relationship between systemic virus burden and viremia is also apparent (Albrecht, 1968), with the potential of the virus to generate viremia being a correlate of neuroinvasion as it applies to most naturally acquired encephalitic infections. Aerosol-acquired infections are probably an exception, but some of these may also cause systemic infection with viremia after gaining access to the lower respiratory tract. Mucosal infection of the alimentary tract has also been implicated both experimentally and in naturally acquired cases of TBEV encephalitis (Gresikova 1975, Odeola 1977). Furthermore, data from a number of studies indicate that such factors as the time of onset, magnitude, and duration of the viremia, as well as the integrity of the host innate immune system, also influence the risk of entry into the CNS, prior to the onset of the virus-specific immune response (see Section VIII). Type I interferons and macrophages in particular have been identified as important factors in limiting infection and clearing systemic virus. Thus this process is largely a balance between the replicative efficiency of the virus and the effectiveness of early host defenses in clearing viremia.

CellularReceptors forFlaviviruses

The cellular receptors that mediate attachment and entry of flaviviruses have been only partially characterized, and to date there has been no definitive identification of the molecules required for these processes in either peripheral or CNS tissues. Although several groups have used biochemical approaches to identify candidate protein receptors (Kopecky 1999, Martinez-Barragan 2001, Ramos-Castaneda 1997) on mammalian cells, their physiologic relevance remains unclear, as there is heterogeneity in the proteins that bind flaviviruses in different cell types. Most recently, CD209 or DC-SIGN (the dendritic cell-specific intercellular adhesion molecule 3–grabbing nonintegrin) has been proposed as a cell surface ligand for dengue virus (Navarro-Sanchez 2003, Tassaneetrithep 2003). Additional studies are required to evaluate the in vivo significance of DC-SIGN as an attachment or entry ligand and whether this is a common determinant of tropism for other flaviviruses. One possibility to reconcile these observations is that multiple independent cellular receptor molecules are utilized either during the spread of flaviviruses within the host or during entry into the CNS. A number of early studies have described binding of flaviviruses to mouse brain substances (reviewed in Albrecht, 1968), suggesting that this tissue is enriched for a unique receptor activity, which may enhance tropism for this organ, particularly in developing brain (Kimura-Kuroda ). In this regard, it has been shown that mouse and monkey brain membrane-receptor preparations preferentially bind neurovirulent strains of flaviviruses, but not attenuated variants (Ni 1998, Ni 2000). Receptor variability in vivo may be a general mechanism for promoting wide tissue tropisms of arthropod-borne viruses, which require cycling in both arthropod and vertebrate hosts. Some data suggest that different cell surface proteins may be utilized for the entry of insect versus vertebrate cells (Martinez-Barragan 2001, Munoz 1998). Tissue tropism in mosquitos has also been observed to correlate with expression of a specific receptor molecule (Yazi-Mendoza ).

Heparan sulfate has been proposed as a flavivirus receptor based on studies showing the dependence of dengue virus infectivity on binding of the E protein to heparan sulfate on target cells (Chen 1997, Hilgard 2000). Subsequent reports have demonstrated that infectivity of TBEV, yellow fever virus (YFV), JEV, and Murray Valley encephalitis (MVE) viruses is affected by cell surface interactions with glycosaminoglycans that are proposed to mediate initial low-affinity binding to the cell surface (Germi 2002, Mandl 2001, Lee 2000, Su 2001), but the role of heparin as an authentic receptor for virulent flaviviruses remains uncertain. For instance, the serial passage of JEV and MVE viruses in cell culture results in selection for viruses that exhibit increased binding to heparin but decreased virulence in vivo (Lee and Lobigs, 2002). Similar observations have been made with alphaviruses (Bernard 2000, Klimstra 1998), where heparin binding was associated with cell culture adaptation of primary virus isolates and attenuation of viral virulence. Thus enhanced binding to glycosaminoglycans is a marker for attenuation of JEV and MVE viruses in the mouse model and correlates with rapid clearance of the glycosaminoglycan-binding variants from the circulation compared to more pathogenic strains (Lee and Lobigs, 2002). The mechanism responsible for this process has not been defined. The relationship of this observation to classic studies on neuroinvasion is also unclear, as virulent strains were originally characterized by their ability to undergo rapid uptake from the circulation, presumably as a result of highly efficient binding and entry to target cells (Albrecht, 1968). Wild-type and glycosaminoglycan-binding variants may differ, however, with respect to their entry into cells that are permissive for replication or are involved in virus clearance (Lee and Lobigs, 2002).

Antibody-dependent enhancement (ADE) of infection has been reported for some encephalitic flaviviruses. ADE is believed to occur through FcγR I (CD64) and FcγR II (CD32), although a second type of ADE that requires complement has also been described (Cardosa 1986, Cardosa 1983). In tissue culture, ADE occurs with several flaviviruses in cells of myeloid lineage (Diamond 2000b, Halstead 1977, Halstead 1980, Brandt 1982, Schlesinger 1983, Cardosa 1983, Cardosa 1986, Brandriss 1984, Halstead 1984). Most of these data are relevant to the pathogenesis of dengue infection, and the significance for encephalitic viruses is less certain. However, there is some experimental evidence that ADE may be involved (Hawkes, 1964), and this notion is consistent with the concepts described for dengue viruses based on the wide antibody cross-reactivity among flaviviruses with respect to the E protein. Neutralizing homologous or cross-reactive antiflavivirus antibodies can enhance neurovirulence and mortality associated with YFV and JEV infection (Gould 1987, Gould 1989, Lobigs 2003b). However, in some cases, the pathogenesis was associated with complement-mediated cytolysis and not with enhancement of infection in vivo (Gould ). The strongest data in support of ADE in dengue infections are epidemiological in nature. In this regard, there is not abundant evidence to support a phenomenon of enhancement of JEV or other encephalitic viruses by preexisting cross-reactive antibodies in natural infections. There is evidence of cross-protection in experimental models, particularly among JEV serogroup members, and for amelioration of JEV encephalitis by prior immunity to related viruses, including dengue (Kurane 2002, Solomon 2002). Experimentally, the protective effect is presumably antibody mediated and is most apparent following infection with live virus or transfer of serum from animals infected with virus (Broom 2000, Tesh 2002). Antibody reponses elicited by inactivated virus do not exhibit much cross-protection in animals and may lead to enhanced infections (Broom ). Consistent with these observations, recipients of inactivated JEV vaccine or live-attenuated dengue vaccine did not generate neutralizing activity in sera against WNV (Kanesa-Thasan ). Severe forms of TBEV encephalitis have been observed after passive immunization with hyperimmune globulin (Waldvogel ), but it is not clear if this represents an enhancement phenomenon as opposed to either failure of antibody to penetrate the CNS or suppression of peripheral or CNS immune responses by high-titer immunoglobulin. The entire issue is somewhat limited by the fact that the encephalitic viruses may not necessarily target cells with abundance of Fc receptors, such as monocyte–macrophages, which are generally considered more important for the pathogenesis of dengue viruses.

CellularTropism ofEncephaliticFlaviviruses

In cell culture, flaviviruses readily infect a variety of cell types, including epithelial, endothelial, and fibroblasts (Avirutnan 1998, Bielefeldt-Ohmann 1998, Diamond 2000b, Kurane 1992), but the relationship of these findings to in vivo replication is uncertain. After peripheral inoculation, flaviviruses probably do not replicate extensively in the skin, but are spread from local lymph nodes by immature dendritic or Langerhans cells, which are permissive for infection (Byrne 2001, McMinn 1996, Johnston 2000, Libraty 2001, Wu 2000, Johnston 1996). Within 1 day of infection, epidermal Langerhans cells that express viral antigens migrate from the skin to the draining lymph node (Byrne 2001, Wu 2000) while expressing maturation markers such as B7-1, B7-2, class II MHC molecules, CD11b, and CD83 (Ho ). These cells produce tumor necrosis factor (TNF)-α and IFN-α (Ho 2001, Libraty 2001) and become more resistant to flavivirus infection (Wu ). Thus infected dendritic cells probably serve to promote antigen presentation in the lymph node and also participate in the spread of infection to lymphoid compartments. The consequences of DC infection, whether apoptosis (as for alphaviruses) or persistent infection, as in the case of Kunjin virus replicons (Varnavski and Khromykh, 1999), and their effects on subsequent shaping of the immune response remain important areas for further investigation. For instance, data from other viral models indicate that there is a quantitative requirement for activated dendritic cells in order to induce T-cell responses (Ludewig ). Survival versus death of these cells as a result of virus infection may have an important impact on this requirement. After replication in lymphoid tissue, encephalitic viruses are believed to exit via efferent lymphatics (Malkova and Frankova, 1959) and gain access to the circulation, whereby systemic infection is established.

Although a tropism of encephalitic viruses for lymphoid tissues has been observed, the identities of the cell types in other compartments that support replication to the levels needed to generate a viremia sufficient to cause neuroinvasion have not been determined definitively. Replication in various peripheral tissues occurs, but vascular endothelial cells have not necessarily been implicated as important sites of replication (Albrecht, 1968 and references therein). However, it should be noted that dengue, JEV, and probably other flaviviruses can enter and, in some cases, establish infection in endothelial cells (Dropulic 1990, Liou 1998) and modulate their activation state and cytokine production (Anderson 1997, Avirutnan 1998, Huang 2000, Bosch 2002). In vitro studies with endothelial cells are complicated by the fact that variable responses can be observed depending on the cell types and assay systems; however, given the potential role of these cells in immune activation, further studies on the effects of flavivirus infection on cytokine and chemokine production by these cells are needed.

In the CNS, neurons are the primary targets for encephalitic flaviviruses (Eldadah 1967b, Weiner 1970, Iwasaki 1986; Kimura-Kuroda 1992, Wang 1997, Xiao 2001). Although viral replication and antigen production have also been observed in cultured oligodendrocytes (Jordan ) and astrocytes (Chen 2000, Liu 1988, Suri 1995), the significance of these reports is difficult to judge, as there is scant evidence for infection of glial cells in vivo. Neurotropism and neurovirulence are governed to a large extent by determinants in the viral E protein, as indicated by an abundance of genetic data indicating that mutations in this protein modulate these phenotypes (McMinn 1997, Ni 1998, Ni 2000), presumably through their effects on receptor targeting and postreceptor events involved in virus entry. In the absence of an experimental system to manipulate the virus receptor on neurons, one cannot conclude whether high neurotropism of these viruses is dependent solely on the binding of the E protein. Comparison of sequence data from virulent and attenuated strains of encephalitic viruses also suggests that the nonstructural region and 3′ UTRs contain determinants that influence pathogenesis. It is important in these types of studies to differentiate effects of genetic mutations on overall replication fitness of the virus versus specific effects in terms of interactions of viral proteins and RNA structures with host factors that affect pathogenesis uniquely. Thus the molecular basis for neurotropism is not understood adequately, and further studies using genetic clones of well-characterized viruses in animal models should help address this issue. Furthermore, the use of primate models to investigate the issue is needed greatly, as the observations in rodent models may not have direct correlates to human infections.

ImmuneResponses toFlaviviruses andTheirRole inPathogenesis

Susceptibility to flavivirus encephalitis implies a failure at some stage of the immune response that theoretically may be defined in either qualitative or quantitative terms. There is substantial clinical and experimental evidence for a correlation between protection against encephalitic disease and the presence of virus-specific antibodies, but the molecular and cellular basis for the development of this response has not been defined thoroughly. In studies that have shown protection by antibodies, the roles of other immune system components in the process have not often been assessed. Furthermore, there is increasing evidence that flaviviruses have evolved mechanisms to manipulate the effector functions of both innate and adaptive immune responses. The magnitude and importance of these responses probably vary from one experimental model to another and account for differences observed in studies that have examined the immune system in the context of a either a primary or a memory response.

Innate Immunity

Interferons

In vitro and in vivo studies have demonstrated that interferon-dependent responses are relevant to protection against flavivirus infections. Dendritic cells in the skin may be the first cells to produce type I interferons (IFN-α or -β) in response to flavivirus infection and initiate this antiviral response (Libraty ). These interferons inhibit flavivirus infection by preventing translation and replication of infectious viral RNA at least partially through an RNase L-, Mx1-, and Protein Kinase R (PKR)-independent mechanism (Anderson 2002, Diamond 2000a, Diamond 2001). These studies have been supported by experiments in immunodeficient and therapeutic mouse models of disease. Pretreatment of mice with IFN-α or its inducers prevents or ameliorates flavivirus infections (Brooks 1999, Harrington 1977, Leyssen 2003b), and mice that are deficient in type I IFNs or their receptor have increased susceptibility to flaviviruses (Johnson 1999, Lobigs 2003a). The role of type II IFN (IFN-γ) in protection versus immunopathogenesis of flavivirus encephalitis is less clear, as this cytokine has a multitude of effects on the host response to these viruses. In part, this includes induction of proinflammatory and antiviral molecules, including nitric oxide (Lin ), and enhancement of the phagocytic activity of monocytes⧸macrophages through increased Fc receptor expression (Rothman and Ennis, 1999). Data from various models support the importance of IFN-γ production in the context of a TH1 virus-specific immune response for the control of infection with encephalitic viruses (Johnson 1999, Liu 2001, Lobigs 2003a). Distinctions should be made, however, concerning the effects of IFN-γ in the periphery and in the context of the CNS immune response (see Section X).

Flaviviruses appear to be capable of attenuating some of the IFN-dependent antiviral effector mechanisms. Treatment of cells or animals with IFN-α as few as 4 h after infection with dengue or SLE viruses resulted in almost a complete loss of antiviral activity (Brooks 1999, Diamond 2000a). Similarly, IFN-α treatment of patients with documented JEV encephalitis had no significant effect on outcome (Solomon ), despite anecdotal reports of its benefit (Harinasuta ). The mechanisms by which the antiviral effect is avoided remain uncharacterized, but data suggest that it may act at a very early step in infection. Future studies should also help determine if the role of type I IFNs in these infections lies only in innate intracellular effector defenses or whether there are effects on the quality and magnitude of the subsequent cell-mediated immune response through their immunoregulatory effects.

Macrophages

Activation of macrophages and modulation of their effector functions are integral parts of flavivirus pathogenesis (Rothman 1999, Spain-Santana 2001). In addition to their role in nonspecific defense, macrophages are targets of infection by some flaviviruses and have the potential to contribute to pathogenesis through antibody-dependent enhancement of infection mediated by Fc and complement receptors (Cardosa 1986, Hawkes 1964, Peiris 1981, Gollins 1984). The preponderance of data supports the protective role of macrophages in control of infection by means of cytokine production and antigen presentation to B and T cells (Kulkarni 1991a, Marianneau 1999). Classic studies have shown that abrogation of phagocytic activity of macrophages results in higher viremia, neuroinvasion, and more severe encephalitis (Ben-Nathan 1996a, Khozinsky 1985, Monath 1971, Monath 1971, Zisman 1971). Some of the protective effect provided appears to be mediated by the stimulation of inducible nitric oxide synthetase (NOS-2) to produce nitric oxide (NO) and other reactive oxygen intermediates such as peroxynitrites (Saxena ). Pretreatment of macrophages with agents that induce NO synthesis have been shown to inhibit JEV infection in vitro (Lin ). Moreover, treatment of mice with a NOS-2 inhibitor increased mortality after JEV infection (Lin ). However, other studies suggest that the inflammatory actions of NO and other reactive oxygen intermediates may, in some cases, contribute to flavivirus pathogenesis. The in vivo administration of a competitive inhibitor of NOS-2 improved survival in mice infected with TBEV (Kreil 1995, Kreil 1996). Activation of macrophages in response to flavivirus infection promotes not only production of NO, but also release of TNF-α, IL-1β, IL-8, and other mediators of acute inflammation that may contribute to tissue damage where macrophages accumulate (Atrasheuskaya 2003, Raghupathy 1998, Bosch 2002, Rothman 1999). These various studies indicate that the behavior of macrophages is fundamental to the pathogenesis of flavivirus disease, but the contribution to virus clearance versus deleterious effects driven by IFN-γ and other proinflammatory stimuli depends on regulation of their activity, and the properties of this innate defense may therefore vary from one context to another.

Natural Killer Cells

Natural killer (NK) lymphocytes lyse infected cells by releasing cytotoxic granules that contain perforin and granzymes or by binding to apoptosis-inducing receptors on target cells (Orange ). NK cell activation is finely regulated through a balance of activating (Ly49D, Ly49H, and NKG2D) and inhibitory receptors [killer cell immunoglobulin-like receptors (KIR), immunoglobulin-like inhibitory receptors (ILT), and CD94-NKG2A] (Smith, ). A decrease in expression of class I MHC molecules on a cell may prompt NK cell activation by attenuating the inhibitory signals. Thus NK cell target recognition occurs after ligation of activating receptors and repression of inhibitory receptors on the cell surface. NK responses have been analyzed in various experimental models of flavivirus infection, but as noted (Hill ), the characterization of the responding cells has been limited. Studies of NK cell activity during WNV infection have revealed blunted cytolytic activity against virus-infected cells, associated with upregulation of MHC antigen and ICAM-1 expression on the target cells by interferon-independent mechanisms (Müllbacher ). However, NK cell-dependent lysis of dengue virus-infected target cells by both natural killer and antibody-dependent cell-mediated cytotoxicity has been observed (Kurane ). Infection of mice with Langat, WNV, and TBEV transiently activated and then suppressed NK cell activity (Vargin and Semenov, 1986). Despite these conflicting observations, the bulk of evidence currently suggests that flaviviruses have evolved a mechanism to evade NK cell responses through an augmentation of cell surface class I MHC expression (King 1988, Liu 1988, Liu 1989, King 1989, Liu 1989), driven by a TAP-dependent process (Momburg 2001, Mullbacher 1995, Lobigs 1999) and NF-κB-dependent transcriptional activation of MHC class I genes (Kesson and King, 2001). Thus, flaviviruses may overcome susceptibility to NK cell-mediated lysis at the expense of increased class I MHC expression and later recognition by virus-specific cytotoxic lymphocytes (CTLs). Consistent with this hypothesis, splenocytes from WNV-immunized mice had poor NK cell lytic activity (Momburg ), and mice that are genetically deficient in NK cells demonstrate no increased morbidity or mortality compared to wild-type controls in response to WNV infection (M. Engle, W. Yokoyama, and M. Diamond, unpublished results). Some residual NK cell function may still be important during flavivirus pathogenesis, as suggested by studies with perforin and Fas knockout mice, which are partially protected from encephalitic disease, through events that operate at the level of neuroinvasion. This may involve the cytotoxic activities of NK cells and⧸or CTLs (Lincon Luna ). In addition, the lack of a substantial effect of IFN-α in the therapy of flavivirus encephalitis is also consistent with inhibition of the NK response, as type I interferons are normally potent activators of these cells.

Natural Antibody

Natural antibodies are primarily of the IgM class, although activity of IgG has also been described. They are secreted constitutively by CD5+ B-1 cells without specific stimulation, have widely variable binding avidities, and represent an initial nonspecific defense against pathogens (Baumgarth 2000, Casali 1989, Ochsenbein 1999a) through direct neutralization of some bacteria and viruses in the circulation (Gobet 1988, Ochsenbein 1999a), enhancement of phagocytosis (Navin ), and complement activation (Baumgarth ). Although the role of natural antibody in flavivirus infection remains unexplored, mice that genetically lack secreted IgM, (sIgM −⧸−), but in which cell surface IgM and IgG responses are intact have increased mortality in certain viral infections, involving a defect in the antiviral IgG responses (Baumgarth 2000, Boes 1998). Such mice are also very susceptible to infection with WNV (Fig. 1 ; M. Engle and M. Diamond, unpublished data). This observation, along with other data, suggests an important role for natural antibody and complement during the early antiviral defense against flaviviruses, although a virus-specific IgM response is likely to be more important (see Section VII,A,5).

Fig 1

Soluble IgM-deficient mice are highly sensitive to peripheral infection with WNV. Mice were infected by subcutaneous inoculation of the footpad with 100 plaque-forming units of WNV NY 99 and monitored for mortality from CNS disease.

Complement

The complement system is an important innate defense for limiting infection by fungal, bacterial, and viral pathogens. Complement inhibits viruses by several mechanisms (Volanakis, 2002), including lysis of enveloped viral particles and virus-infected cells by the C5–C9 membrane attack complex; recruitment and activation of monocytes and granulocytes by C3a and C5a; clearance of virus from circulation after opsonization by proteolytic fragments of C3, C3b, and C3bi, followed by uptake into cells that express complement receptor; and C3-facilitated uptake of antigen and presentation by macrophages and dendritic cells (Ochsenbein and Zinkernagel, 2000) during priming of T and B lymphocytes (Da Costa 1999, Ochsenbein 1999b, Kopf 2002). Preliminary studies indicate that complement plays an essential role in limiting WNV infection. Mice that are genetically deficient in C3 uniformly succumb to infection even at low inoculating doses (Fig. 2 ; E. Mehlhop, M. Engle, and M. Diamond, unpublished data). Additional studies must be performed to determine which individual mechanisms are most critical for this complement-mediated control. A deficiency of C3 could exacerbate WNV infection because of depressed C5–C9 lytic or C3 opsonic activity that results in a failure to clear virus from the circulation. Alternatively, C3 may play important roles in linking the innate and adaptive immune responses (Barrington 2001, Carroll 1998, Ochsenbein 2000) against WNV. C3 is required for normal IgG production and T-cell priming against influenza and herpes viruses (Da Costa 1999, Fischer 1996; Kopf ), and a deficiency in C3 decreases opsonization and viral antigen presentation, leading to deficits in the adaptive B-and T-cell responses (Ochsenbein and Zinkernagel, 2000). Although the lytic and proinflammatory activity of complement may contribute to the defense against WNV, preliminary data indicate that a deficiency in either C3 compromises the antiviral B-cell immune response, as mice that lack C3 have markedly depressed titers of WNV-specific IgG (E. Mehlhop, M. Engle, and M. Diamond, unpublished data).

Fig 2

C3-deficient mice are highly sensitive to peripheral infection with WNV. Mice were infected as described in Fig. 1 and monitored for mortality from CNS disease.

Adaptive Immunity

The roles of both humoral and cellular immunity during the pathogenesis of flavivirus disease have been studied in various models involving active and passive immunization in normal and immunodeficient animals, and understanding of the immunologic basis for protection against acute encephalitis is beginning to advance. B cells, CD4+ cells, and CD8+ T cells have all been implicated in contributing to protection, although as stated earlier, their relative importance seems to vary depending on the context of the experimental model under investigation.

B Cells and Antibody

Antibody responses to the E and NS1 proteins involve many epitopes, and both neutralizing and nonneutralizing antibodies (including those against NS1) can prevent fatal encephalitis, as demonstrated in many passive transfer and active immunization experiments in experimental animal models (Brandriss 1986, Diamond 2003, Gould 1987, Henchal 1988, Kimura-Kuroda 1988, Putnak 1990, Schlesinger 1985, Schlesinger 1987, Zhang 1989). A protective effect can occur even after establishment of infection in the CNS. Mechanisms of antibody-mediated protection during flavivirus infection include direct neutralization of receptor binding (Crill and Roehrig, 2001), blocking of virus uncoating (Gollins and Porterfield, 1984), and Fc receptor-dependent virus clearance via the reticuloendothelial system. Most neutralizing antibodies recognize the structural E protein, and antibody epitopes appear to be broadly distributed over its surface (Heinz 1986, Roehrig 1989); however, these do not all represent potent sites for neutralization. A subset of neutralization epitopes is found on the prM protein (Colombage 1998, Falconar 1999, Pincus 1992). The presence of nonneutralizing, yet protective antibodies against NS1 is also well documented (Cane 1988, Schlesinger 1986, Henchal 1988, Falgout 1990, Putnak 1990, Despres 1991). Antibodies to NS1 are proposed to mediate the lysis of virus-infected cells, which express this protein on their surface, by complement-mediated lysis and⧸or antibody-dependent cellular cytotoxicity (Hill 1993, Kurane 1984, Schlesinger 1990). These humoral responses are believed to be important components of the protective immune response to flavivirus infection, but antibodies to NS1 are not often measured in experimental models of infection or immunization and challenge.

Although the neutralizing antibody responses are considered correlates of protection (Markoff, 2000), this process is probably also a function of additional innate and adaptive immune effector systems whose roles in the control of infection are less easily demonstrated. In this regard, challenge in the context of passively administered virus-specific antibodies is not necessarily associated with sterilizing immunity, indicating that antiviral defenses other than antibodies are involved in protection (Kreil 1998a, Kreil 1998b). Definitive experiments to determine the extent of immune activation in this setting are likely to expand our understanding of the correlates of protection and the immunological basis for a successful antiviral response.

Mice that lack B cells are very vulnerable to flavivirus infections and encephalitis (Diamond 2003, Liu 2001), purportedly as a consequence of lacking antibodies. However, these models must be explored further to determine whether T-cell responses are otherwise fully activated and effective, as it is possible that B cells could influence CD4+ and CD8+ T-cell responses through antigen presentation or other immunoregulatory events, as has been observed in other neurotropic viral infections (Bergmann ).

Factors that drive B-cell activation and maturation during flavivirus infections are not well understood. Multiple T helper epitopes have been identified in the E protein of JEV and MVE viruses, some of which are dominant, broadly reactive among JE serogroup viruses, and prime for the neutralizing antibody (Kutubuddin 1991, Mathews 1991, Mathews 1992, Roehrig 1992). Immunization with plasmid DNA encoding the E region is more effective than viral E antigen itself, presumably due to the induction of T-cell responses to the E protein (Chen ). Immunization of HIV-infected individuals with inactivated flavivirus vaccines has suggested that CD4+ T cells may not be critical for the induction of protective antibodies (Panasiuk ). However, others have observed that antibody responses to inactivated or live viral vaccines are weak in these subjects (Rojanasuphot 1998, Sibailly 1997). Immunization of CD4 knockout or class II knockout mice with YFV markedly diminished or abrogated the neutralizing antibody response, respectively (Chambers and Liang, unpublished data), suggesting that there is dependence on functional CD4+ T cells to generate long-lasting B-cell memory against flaviviruses. Differences in immunization schedules and numbers of residual CD4+ lineage T cells in these various reports may explain the discrepancies. Studies on the role of dendritic cells in early B-cell activation and the nature of the toll-like receptor signals induced in response to viral antigens on these cells may provide new insights into the determinants that establish and drive the memory response to flavivirus antigens.

An IgM response to flaviviruses is a feature of most clinical and experimental infections (Martin ) and has been reported to be a correlate of protection during clinical JEV encephalitis in some studies (Burke 1985a, Burke 1985b, Ravi 1993). Flavivirus infections typically elicit IgM responses that can often persist for prolonged periods (Edelman 1976, Monath 1971, Roehrig 2003). However, experimentally, the initial IgM response to encephalitic viruses may possess variable neutralization activity and protective capacity (Diamond ; Hofmann et al., 1978; Ishii ); in some studies, the complement-fixing activity was limited compared to that provided by the ensuing IgG response (Lee 1961, Ishii 1968). This contrasts with what has been observed with YF 17D vaccination (Monath, 1971) and may reflect differences in early B-cell activation among these infections or differences between humans and mice with respect to the process. However, the role of a vigorous IgM response may be in providing temporary neutralizing activity, while more importantly activating complement-dependent pathways involved in programming virus-specific B- and T-cell responses. Studies with the WNV model (see later) suggest a critical early function of both IgM and complement in the control of extraneural infection. The neutralizing activity of sera rises in conjunction with the appearance of IgG, and this response includes a variety of biological activities, including hemagglutination inhibition, complement fixation, and virus neutralization. The memory or “antigenic sin” response elicited by cross-reactive antigens involves all of these responses, but is weakest for the neutralization response (Innis, 1997), suggesting that there is some hierarchical pattern of B-cell epitopes or some selectivity in recognizing the most cross-reactive antigens. This phenomenon is probably evolutionarily adaptive for recurrent infections with heterologous flaviviruses, but is capitalized upon by dengue viruses during the pathogenesis of dengue hemorrhagic fever and shock syndrome. In any case, the anamnestic antibody response has been demonstrated to be critical to a defense against encephalitic viruses in the context of immunization and challenge models, which are surrogates for the efficacy of vaccines (Konishi 1999, Pan 2001). It nevertheless remains unclear at the present time the extent to which antibody responses alone contribute to the control of acute infections.

T Cells

T-cell responses to flavivirus proteins have been best studied for members of the DEN and JEV serogroups. Both CD4+ and CD8+ T lymphocyte responses involve broad flavivirus cross-reactivity, although this varies significantly among different MHC haplotypes (Hill 1992, Kulkarni 1992, Kurane 1991, Uren 1987; reviewed in Hill ). Multiple epitopes for T-cell responses have been identified on both structural and nonstructural proteins; however, genetic factors restrict the number of targets and fine specificities differ considerably. Determinants for class I responses are more frequent within the viral nonstructural region, particularly the NS3 protein, which contains dominant epitopes in both humans and mice. In contrast, viral structural protein antigens elicit class II responses more consistently. The role of virus-specific CD4+ T cells in flavivirus encephalitis is not well understood, although experimental models indicate a requirement for such cells in protection against acute disease (see Section X). Some of the cytotoxic T-cell response to dengue and JE viruses is contained in the CD4+ compartment and is probably mediated by Fas⧸FasL interactions (Aihara 2000, Gagnon 1999), which has implications for possible immunopathogenic responses both in the periphery and in the CNS.

Flavivirus-specific CD8+ T cells have multiple effector functions, including cytotoxic activity and production of IFN-γ (Douglas 1994, Kesson 1987, Kurane 1989, Kurane 1991, Kurane 1995, Kulkarni 1991b, Murali-Krishna 1996, Takada 2000), suggesting that polarization of the immune response toward a TH1 phenotype is involved in control of these viruses. Although the importance of type I interferons in protection against these viruses would suggest its involvement in promoting this T-cell response, as has been described for other viruses (Cousens ), the role of IL-12 in driving this process has received only limited study (Chen 2001, Phillpotts 2003), and its importance may vary from one context to another (Dalod ). Cellular immunity clearly contributes to the control of virus infection in experimental animal models, but this varies depending on the context examined and with respect to the virulence of the challenge virus. T-cell-deficient mice fail to generate protective immunity after a sublethal challenge with YFV strains (Bradish ). Moreover, animals that are treated with drugs that impair T-cell function develop a rapidly progressive flavivirus encephalitis (Camenga 1974, Cole 1968, Nathanson 1970). The adoptive transfer of immune spleen cells can protect against encephalitis, but the lymphocyte subpopulations that mediate this protection have not been very well characterized in classic studies (Bradish 1980, Camenga 1974, Jacoby 1980).

More recent studies have, to some degree, clarified the role of CD8+ T cells in these infections. Quantitation of the CD8+ T-cell response to YFV in experimental mice (van der Most ) reveals activation by immunodominant epitopes and is supported by the observation that CD8+ knockout mice exhibit a defect in the clearance of infectious YFV from the CNS (Liu and Chambers, 2001; T. J. Chambers, unpublished data) and also have increased mortality after WNV infection (B. Shrestha and M. Diamond, manuscript in preparation). Human recipients of YFV 17D vaccine exhibit an increase in CD8+ T cells as well, and epitopes have been mapped to multiple proteins (Chu 1999, Co 2002). While data from most of these models would indicate that CTL responses are primarily protective in vivo, their potential for immunopathogenic effects requires further investigation. Some studies with MVE virus (Licon Luna ) and dengue virus (Rothman and Ennis, 1999) suggest that the cytotoxicity of CTL may contribute to the disease pathogenesis. Studies with knockout mice continue to provide novel information on the role of T cells in flavivirus pathogenesis and immunity. However, an important limitation of these experiments is that the effect of gene knockout on immunologic development and function is not known. Better attempts to assess the entire range of properties of the immune response in these types of experimental models are needed, as gene knockouts could have multiple effects beyond simply loss of the targeted function.

Human studies on immune responses to JEV or JEV structural protein antigens encoded in recombinant vaccinia virus revealed proliferation and induction of cytolytic activity in the CD8+ T-cell compartment (Konishi 1995, Konishi 1998a), similar to what has been observed in mice (Konishi 1997, Konishi 1998b). The significance of these responses is uncertain, as another study found no correlation between T-cell proliferation and either the antibody response or the clinical outcome (Desai 1995a, Desai 1994). Inactivated JEV vaccine induced CD4+, class II-restricted T cells with cytolytic activity (Aihara ), suggesting that it is not capable of inducing high levels of CD8+ CTLs; in mice, this vaccine elicts a TH2 immune response (Ramakrishna ). As noted earlier, in the context of clinical infections with flaviviruses, the integrity of cellular immune function appears to be important (Iwamoto 2003, Okhuysen 1993, Neogi 1998, Szilak 2000). However, better understanding of the effector properties of T cells and their role in protection in humans and in experimental animals is needed.

Neuroinvasion

Neuroinvasiveness is a critical step in the pathogenesis of flavivirus encephalitis and is affected by both viral and host factors. In terms of viral factors, characterization of various virulent and attenuated strains of JEV, TBEV, YFV, and WNV has revealed that viral determinants of neuroinvasiveness map principally the E protein (reviewed in McMinn, 1997). The mechanisms associated with these genetic determinants have not been completely determined, but are believed to relate to increased viral infectivity toward important target cells through enhanced binding and penetration. Entry into the CNS has been proposed to involve a number of potential processes, none of which has been definitively demonstrated in vivo. The proposed pathways include (1) transport across the cerebrovascular endothelium, or infection of these and other cells constituting the blood–brain barrier (Dropulic 1990, Liou 1998); (2) access to the CNS after loss of blood–brain barrier integrity (Kobiler 1989, Lustig 1992); and (3) entry through the olfactory epithelium (McMinn 1996, Monath 1983). There are regions of the CNS where lack of a blood–brain barrier may constitute sites of vulnerability to infection in the presence of viremia (i.e., the choroid plexus and the circumventricular organs). Some of these may be supported by nonspecific defenses that are sufficient to withstand a low level of viremia, as the choroid is a site of IFN-α expression and OAS induction (Asada-Kubota 1997, Khan 1989). Entry of virus into the CNS by passage across the small cerebrovascular vessels is consistent with the accumulation of perivascular infiltrates of inflammatory cells, which is a hallmark of flavivirus encephalitis (Johnson ). However, access through the olfactory bulb is believed to occur either after infection by the aerosol route or intranasally (Hambleton 1983, Nir 1965, Myint 1999, Raengsakulrach 1999) or in the context of hematogenous dissemination of virus (McMinn 1996, Monath 1986, Monath 1983). The olfactory bulb is especially vulnerable to infection because of the exposure of its nerve terminals within the olfactory mucosa, and this route is exploited by other neurotropic viruses (Fazakerley, 2002).

Disruption of the blood–brain barrier facilitates the entry of noninvasive flaviviruses into the CNS (Lustig ), suggesting that neuroinvasion may be influenced by host factors that alter the permeability of this barrier as a result of systemic infection (Chaturvedi 1991, Mathur 1992, Kaiser 2000), including IFN-γ, TNF-α, and possibly effector functions of CTLs and NK cells (Lincon Luna ). Neuroinvasion is also influenced by physical stress and other agents, including inhalational anesthetics (Ben-Nathan 1989, Ben-Nathan 1992, Ben-Nathan 1996b, Ben-Nathan 2000). Immunosuppression by corticosterone and other endogenous immunomodulators is an important factor, as involution of lymphoid tissue has been observed and probably facilitates the generation of viremia by attenuated viral strains. Perturbation of the blood–brain barrier may also be involved in this process (Ben-Nathan ). Early viremia and sustained viremia are correlated with neuroinvasion in animal models, consistent with the belief that replication to high titers in peripheral tissues is an important property of invasive strains, at least in immunologically normal hosts (Albrecht 1968, Huang 1963, Monath 1986). However, the timing of peak viremia is probably critical, as levels sufficient to cause neuroinvasion are dissipated concurrent with the appearance of adaptive immune responses in the periphery (Diamond 2003, Halevy 1994). This is also evident from studies of immunodeficient mice lacking T- and B-cell responses, which exhibit neuroinvasion in the presence of viremia (Chambers 2001, Halevy 1994, Lin 1998, Johnson 1999, Charlier 2002, Diamond 2003). At present it appears that the mechanisms responsible for entry into the CNS may vary, depending on a specific virus and the level of host immunocompetence. Continued research with the use of knockout strains harboring single and multiple defects in these redundant immune system defenses will improve the understanding of how this process occurs.

The mechanisms of flavivirus spread within the CNS are not established. Studies with alphaviruses in the mouse model suggest that this occurs in a circuit-specific manner among olfactory neurons that are undergoing developmental synaptogenesis (Oliver and Fazakerley, 1998). Spread of MVE virus in immature mice exhibited a pattern consistent with these observations, but occurred throughout the CNS (McMinn ). An unusual accumulation of WNV particles in myelin lamellae was observed in spinal cord cultures (Shahar ), suggesting the possibility of virus transport by mechanisms not only involving the axoplasm.

Neuropathology

Flaviviruses, particularly members of the JEV and TBEV serogroups, cause viral encephalitis in many vertebrate species, a dead-end transmission pathway believed to reflect an evolutionarily conserved capacity of these viruses to grow in the CNS of arthropods and vertebrates (Monath, 1986). In contrast to the noncytopathic nature of infection in arthropods, a spectrum of acute and chronic CNS pathologic changes occur in vertebrates and have been documented extensively (Chu 1999, Manuelidis 1956, Dominguez 1963, Nathanson 1966, Vince 1969, Zlotnik 1971, Zlotnik 1976, Reyes 1981, Pogodina 1983, Levenbook 1987; Manuelidis, 1956; Nathanson et al., 1966; Pogodina et al., 1983; Reyes et al., 1981; Vince and Grevic, 1969; Zlotnik et al., 1971, 1976). Virus can be demonstrated within neurons throughout the brain and spinal cord, but infection of other cell types has been less well characterized. In humans, neuroinvasive flaviviruses cause an acute, often fatal encephalomyelitis (Monath, 1986) associated with characteristic inflammatory changes and often targeted to specific regions (Johnson 1985, Zimmerman 1946, Miyake 1964, Suzuki 1966). The pathological changes have been characterized in many experimental animal models, as well as in fatal human cases. These viruses can evoke inflammatory infiltrates extending from the meningeal layers into the brain substance, with features typical of other viral encephalitides, including leptomeningitis, perivascular lymphocytic accumulation, parenchymal infiltrates, and microglial nodules associated with neuronophagia in regions of viral infection (Fig 3, Fig 4, Fig 5 ). The neuropathology can, in some cases, include the destruction of vascular structures with focal hemorrhage, suggesting a vasculitis. Loss of regional blood flow, as well as disruption of the blood–brain barrier, has been described in clinical cases of TBEV (Gunther 1998, Kaiser 2000, Gunther 1996, Kaiser 2002). The nature and degree of the inflammatory disease depend on many factors, including the virulence of the virus, the route of infection, and the age and immunocompetence of the host. Varying levels of CNS inflammation without frank evidence of neuronal damage have been described in some models, including human autopsy cases, suggesting that viral infection can induce lethal pathophysiology prior to or in the absence of recruitment of the peripheral immune response (reviewed in Monath, 1986). Although cytopathic effects have been observed primarily in virus-infected neurons, other noninfected cells can also exhibit pathologic changes, presumably through bystander injury (see later). A number of studies have documented the distribution of the neuropathology and the clinical manifestations that characterize flavivirus infection of the CNS. For instance, infection in cortical regions typically gives rise to depressed consciouness and seizures, but the involvement of subcortical regions, including the midbrain and brainstem, as well as thalamic and basal ganglial involvement, can give rise to a variety of movement disorders (Asher 1975, Misra 1997a, Pradhan 1999, Ogata 2000, Kalita 2000, Murgod 2001). WNV encephalitis in the United States and JEV encephalitis have both been characterized by a poliomyelitis-like syndrome, suggesting infection of lower motor neurons in association with flaccid paralysis (Glass 2002, Solomon 1998, Misra 1997b, Leis 2002). Lower motor neuron disease is also typical of TBEV infection. Factors governing the differences in neuronal susceptibility to flaviviruses are not known, but may be similar to those that operate in the case of other neurotropic arboviruses, such as Sindbis, where the neuronal response to viral injury may be variable (Griffin and Hardwick, 1999). Neuronal death associated with flavivirus infection has classically been ascribed to degenerative necrosis. Pathologic changes that accompany this process include vacuolization and proliferation of intracellular membranes, which produces a characteristic ultrastructural appearance (Fig. 6 ; Murphy ). Whether there is a greater propensity to cause necrosis versus apoptosis requires further study.

Fig 3

Yellow fever virus meningoencephalitis in the rhesus monkey showing leptomeningeal accumulation of acute inflammatory cells. Courtesy of the United States Army Medical Research Institute of Infectious Diseases (USAMRIID). (See Color Insert.)

Fig 4

Yellow fever encephalitis in the rhesus monkey showing focus of perivascular infiltrate with mononuclear cells in the cerebral cortex. Courtesy of USAMRIID. (See Color Insert.)

Fig 5

Yellow fever encephalitis in the rhesus monkey showing microglial nodule with neuronophagia of a cortical neuron stained for viral antigen. Courtesy of USAMRIID. (See Color Insert.)

Fig 6

Electron micrograph of a mouse CNS neuron infected with SLE virus showing characteristic cytoplasmic pathology but integrity of the nuclear (N) envelope. Arrow indicates virions within inner and outer nuclear membranes. From Murphy , with permission.

Several flaviviruses have been shown to induce apoptosis, both in vitro (Isaeva 1998, Liao 1998, Parquet 2001, Prikhod'ko 2002) and in vivo, in the rodent CNS (Andrews 1999, Despres 1998, Isaeva 1998, Duarte dos Santos 2000, Xiao 2001; Fig. 7 ). In this regard, reports describing primarily degenerative pathology at the light or electronmicroscopic level, typically involving cytoplasmic changes such as chromatolysis, swelling, and dissolution of Nissl substance, also mentioned nuclear pathology, including pyknosis, disruption of the nuclear envelope, and alterations in chromatin (Dominguez 1963, Manuelidis 1956, Mathews 2000). These findings are of interest given reports that neuronotropic viruses such as Sindbis virus may not induce typical apoptotic morphology in neurons (Griffin 1999, Sammin 1999, Havert 2000, Kerr 2002) and that cell death may occur through both apoptotic and necrotic mechanisms (Havert 2000, Sammin 1999, Nargi-Aizenman 2001), depending on the integrity of the apoptotic pathways of a given population of neurons, their profile of apoptotic modulators, and the presence of excitotoxic stimuli. The most detailed in vivo characterization of neuronal apoptosis by flaviviruses has been reported for dengue. Neuroadapted dengue virus induces apoptosis in infected neurons as well as in noninfected cells (Després ), suggesting that indirect mechanisms of cellular injury occur in areas of heavy virus burden. It is important to note that these findings were demonstrated in very young mice, whose neurons are highly susceptible to apoptotic stimuli, and may not reflect the response of more mature cells in older mice. However, studies with WNV in the adult hamster model provide evidence that highly neurovirulent strains are potent inducers of apoptosis (Xiao ; Fig. 7). A relationship between viral virulence and the extent of apoptosis has not been clearly established for flaviviruses; however, the general capacity of neuroadapted strains to produce high virus burdens in association with cytopathology suggests that the situation is likely to resemble that of alphaviruses and other neuronotropic viruses (Lewis 1996, Oberhaus 1997, Theerasurakarn 1998). Although the molecular details of this process are not fully known, the expression of some flavivirus proteins appears to directly induce apoptotic cell death of neurons, including the WNV capsid protein (Yang ), the Langat virus NS3 protein, which causes apoptosis through the activation of caspases 3, 8, and 9 (Prikhod'ko ), and the E proteins of neuroadapted dengue-2 virus (Duarte dos Santos ) and JEV, which appear to stimulate an ER unfolded protein response (Su ) and a component of oxidative stress (Raung ). However, there are many other potential mechanisms for provoking apoptosis by neuronotropic RNA viruses, including signaling through interferon-α-dependent pathways, phospholipase A2 activation, activation of NF-κB and p53-regulated genes (Fazakerley, 2001), and activation of apoptosis during viral entry (Jan ). Some of these processes may also be involved in the pathogenesis of encephalitic flaviviruses.

Fig 7

WNV infection in the Syrian golden hamster. (A) Viral antigen-positive neurons in the cerebral cortex. (B) TUNEL-positive apoptotic neurons in the cortex. Courtesy of Dr. Shu-Yan Xiao. From Xiao et al., 2001. (See Color Insert.)

The role of apoptotic regulators bcl-2, bax, bcl-X, and related gene products in modulating neuronal death has been only partially characterized for flaviviruses. Forced expression of bcl-2-related genes promotes the survival of neuronal cell lines infected with JEV and dengue virus and facilitates viral persistence, primarily by restricting virus-induced cytopathic effects and not viral replication (Liao 1997, Liao 1998). These outcomes vary in neuronal versus nonneuronal cell lines, and differential effects of bcl-2 and bcl-X appear to operate. Such findings suggest that the effects of apoptosis modulators on flaviviruses are similar to those observed with alphaviruses, where it is known that the expression of these proteins can vary in their effects from one cell type to another. Conclusions about the role of these proteins in different in vivo models therefore require careful evaluation (Griffin 1999, Levine 2002). Furthermore, because neuronotropic viruses can induce both necrosis and apoptosis, neuronal death may require assessment by several criteria. Neuronal injury as a result of bystander effects may also be a factor during flavivirus neuropathogenesis given that microglial activation and elaboration of inflammatory mediators, including IL-1β and TNF-α, occur in the CNS during these infections (Andrews 1999, Ravi 1997, Liu 2001) and may accompany the production of nitric oxide and peroxynitrite, which can cause neurotoxicity. Other potential mechanisms include excitatory cell death due to the activation of NMDA receptors, which has been implicated in the pathogenesis of Sindbis virus and HIV (Nargi-Aizenman and Griffin, 2001). Thus, although it is likely that the neurovirulence phenotype of flaviviruses is linked to the extent of neuronal cell death caused during the encephalitis, there appear to be multiple independent mechanisms by which neuronotropic viruses cause cell death. This process can be affected by the region of the brain affected, the degree of neuronal maturity, the factors that regulate cell death signaling receptors and their pathways, and levels of apoptosis modulators and other innate responses of virus-infected neurons (Liang 1998, Griffin 1999, Levine 2002). In some cases, changes in the expression of neurotrophins may also be involved in the CNS response to viral injury (Zocher ); however, the relevance of this phenomenon to other types of viral encephalitis has yet to be widely investigated. The process is also subject to additional influence by the properties of the immune response recruited into the CNS, including CD4+ and CD8+ T cells, which may be involved in cytotoxicity toward virus-infected and, in the case of CD4+ cells, perhaps even noninfected cells (Despres 1998, Gagnon 1999), through Fas-dependent mechanisms under certain circumstances (Medana ). The effects of nonspecific inflammation, such as release of toxic substances from neutrophils (Andrews ), may also contribute to cellular injury.

TheCentralNervousSystemImmuneResponse

Flavivirus infections induce a CNS inflammatory response of variable intensity. Data from most experimental models suggest that this inflammation is a requirement for protection from lethal infection with neurovirulent strains. The characteristics of the inflammation in experimental models have been shown to be affected by numerous factors, which include the endogenous CNS response, as well as the adequacy of the peripheral immune response and its timely recruitment into the CNS. In some cases of encephalitis, relatively scant inflammatory disease has been noted. This has also been observed experimentally as, for example, with TBEV under conditions where viral neuroinvasion and CNS involvement progress rapidly (Vince and Grevic, 1969). Immunosuppression also dramatically reduces the intensity of the CNS inflammation (Hirsch 1967, Leyssen 2003a). However, acute inflammation may become severe in response to a heavy antigen load and has been implicated in immunopathologic reactions in the CNS (reviewed in Monath, 1986).

Innate Responses

Viral infections of the CNS commonly result in the induction of innate responses, which include activation and proliferation of microglia and activation of astrocytes and cerebrovascular endothelium, with ensuing production of chemokines and proinflammatory cytokines (Benveniste, 1997). A consequence of this activation is the conditioning of cells in the CNS parenchyma and the blood–brain barrier to accommodate and modulate the influx of activated lymphocytes from the periphery by the upregulation of adhesion molecules and class I and II antigens. Activation of innate responses within the CNS during flavivirus encephalitis has been suggested by several human and experimental animal model studies in which the expression of chemokine and cytokine genes or their proteins has been analyzed. The production of IL-8 and macrophage inhibitory factor (MIF) has been detected in cerebrospinal fluid (CSF) or brain tissue during early stages of encephalitis with JEV and MVE viruses (Andrews 1999, Singh 2000, Suzuki 2000), with levels of IL-8 correlating directly with the number of neutrophils in the CSF. TNF-α and IL-1β are also elicited in response to flavivirus infection of the CNS (Liu 2001, Ravi 1997, Suzuki 2000), presumably representing intrinsic responses of microglia and astrocytes to the acute injury (Benveniste, 1997). Thus the early stage of the encephalitis involves the endogenous expression of mediators that results in the recruitment of nonspecific acute inflammatory cells with the potential for the production of neurotoxic substances, such as reactive oxygen intermediates, and probably facilitates further stimuli that intensify the inflammation. For instance, IL-1β and TNF-α can also mediate the release of IL-8 from astrocytes (Aloisi ). The induction of the nonspecific acute inflammation may be a deleterious process because levels of IL-8 are predictive of fatal disease (Suzuki ) and treatment of infected mice with inhibitors of NOS-2 in the acute stage of encephalitis lessened mortality in association with reduced inflammation (Andrews ). Because viral infections of the CNS induce the expression of chemokines, which have been implicated in inflammatory cell recruitment, (Liu 2000, Liu 2001, Liu 2001), the intense level of inflammation observed in flavivirus encephalitis may be driven by the induction of one or more chemokines and their receptors. The expression of monocyte and T-cell chemokines has not been reported in these infections; however, studies with other models suggest that MCP-1, IP-10, RANTES, and other chemokines are involved in the trafficking of leukocytes into areas of virus infection in the CNS. In conjunction with the effects of TNF-α and IFN-γ, both of which can lead to loss of integrity of the blood–brain barrier, this collection of stimuli may be sufficient to drive the commitment phase of the inflammatory response, during which the unrestricted entry of T lymphocytes then proceeds (Hickey, 1999).

The intrinsic defensive response of the CNS to viral injury also includes the induction of other classes of genes likely to influence the antiviral activity of this compartment through direct effects and by shaping the virus-specific immune response to viral injury (Johnston 2001, Labrada 2002). These include IFN-α and IFN-regulated genes such as ISG12 (Labrada ). IFN-α itself does not seem to be strongly upregulated in acute encephalitis. However, IFN-α has been reported in the CSF and brain tissue of human cases of encephalitis with JEV serogroup viruses and, in such cases, appeared to represent a marker of severe infection with a fatal outcome (Burke 1987, Luby 1969, Leport 1984). The role of type I interferons in the CNS response to injury is complex, and differences in the activities of IRFs and IGSFs toward target response elements in neurons and glia can affect the range of genes involved in the response, including MHC antigens, the OAS1b protein, and antiapoptotic factors (Baron-Delage 2000, Hirsch 1986, Njenga 1997, Massa 1999, Lucas 2003). Studies of viral infection in IFN-α and IFN-α-regulated gene knockout mice should help clarify the role of this defense system in directly controlling the viral infection of neurons versus effects on promoting the activity of the immune response recruited from the periphery. The importance of IL-12 in the innate CNS response to flavivirus encephalitis has not been investigated extensively; however, the outcome of CNS disease appears to vary among different viruses when this cytokine is used as an experimental therapy (Phillpotts ). Because of the importance of IL-12 in antiviral defense against other CNS viral infections, through its ability to stimulate NOS (Reiss ), more studies are needed to determine if it is implicated in protection against flaviviruses.

Virus-Specific Responses

Compared to information available on other neurotropic viruses, there has been relatively little work done to characterize the properties of virus-specific T cells recruited into the CNS in response to flavivirus infection (Scheider-Schaulies et al., 1997). This response is subject to a number of influences, including previous immunologic experience with related viruses, the level of immunocompetence, immunogenetic host factors, and the virulence of the infecting virus. Available information comes from a limited number of human and animal model experiments, which partially characterized lymphocytes or soluble markers of activated T cells and cytokines in the CNS (Burke 1985b, Johnson 1985, Johnson 1986, Carson 2003, Iwasaki 1993, Kuno 1993), and other studies that have evaluated the requirements for these cells in the control of infection of this compartment in mouse models (Liu and Chambers, 2001; Liu et al., 1989a; Murali-Krishna 1996, Liu 2001, van der Most 2000, van der Most 2003). T cells bearing both CD4+ and CD8+ surface markers have been visualized in perivascular infiltrates, in CSF, and in brain parenchyma during human flavivirus encephalitis (Johnson 1985, Johnson 1986, Sampson 2000). These cells include a larger proportion of CD4+ to CD8+ T cells and moderate numbers of B cells and macrophages. However, the composition of cells differed regionally in the CNS, with macrophages and T cells more abundant in brain parenchyma and B cells more common within perivascular infiltrates. Factors governing the distribution of these cells and their effector functions are not well understood. Data from other experimental models suggest that CD4+ T cells are important in directing the recruitment of lymphocytes (Hickey, 1999) and in maintaining the effector function of CD8+ T cells within the brain parenchyma (Stohlman ). CD8+ T cells have clearly been isolated from the brains of flavivirus-infected mice and, in some cases, demonstrated to have cytolytic activity (Liu et al., 1989a); however, the contributions of such cells to both virus clearance and cellular injury within this compartment have not been defined. This remains a fundamental question, as the role of CD4+ and CD8+ T cells in different viral infections of the CNS and different experimental paradigms can vary considerably (Fazakerley 2002, Schneider-Schaulies 1997). For instance, virus clearance from mouse brains acutely infected with JEV has been reported by virus-specific CD8+ T cells adoptively transferred into the CNS; however, CD4+ T cells were also required (Murali-Krishna ). The relationship of this adoptive response to the natural immune response recruited from the periphery remains unclear, as the route of transfer may not reflect the normal pathway of lymphocyte trafficking. A requirement for CNS-associated CD4+ and CD8+ T cells was also observed in an immunization⧸challenge model of dengue virus (van der Most ). However, in the context of a memory response to YFV in the mouse model, CD4+ T cells and B cells were required for the control of viral infection, whereas CD8+ T cells were not required, although CD8-deficient mice exhibited a defect in virus clearance (Liu and Chambers, 2001; T. J. Chambers, unpublished data). Part of the effector activity of CD8+ T cells may be mediated by the production of IFN-γ, which has a range of effects on the immunological properties of the CNS, including the upregulation of class I and II antigens on glial cells and injured neurons, activation of microglial, priming of astrocytes for cytokine production, and increasing permeability of the blood–brain barrier, as well as antiviral activity in the brain parenchyma (Popko 1997, Kundig 1993). IFN-γ knockout mice are defective in the clearance of YFV from the CNS and exhibit decreased inflammatory cell recruitment to this compartment (Liu and Chambers, 2001), indicating an important role of this cytokine in flavivirus encephalitis. IFN-γ may have a primary role in these processes because elimination of virus-infected neurons by CTLs is very tightly constrained by the absence of constitutive expression of MHC class I in these cells and the lack of susceptibility to perforin-mediated lysis (Medana ; Neumann ). Although killing can occur through Fas–FasL interactions (Medana ), induction of FasL expression by neurons can also confer protection against CTL attack under certain conditions (Medana ) and even induce apoptosis in infiltrating lymphocytes (Flugel ). Virus-specific CD8+ T cells have been shown to persist in the brains of mice recovering from dengue encephalitis for prolonged periods and differentiate into effector–memory cells that lose CTL activity while still producing IFN-γ (van der Most ). Thus it appears that cellular immune-mediated mechanisms of neuronal cell death are subject to tight regulation in the presence of acute viral injury.

The role of antibody responses in the control of flavivirus encephalitis has been investigated in many classic studies, and it has been demonstrated repeatedly that immune serum can arrest viral infection in the CNS (Roehrig ). Mechanisms responsible for this antibody-mediated control of infection remain unclear; however, studies with YFV encephalitis indicate that the Fc region and the IgG subclass are critical for protection in the mouse model (Schlesinger 1995, Schlesinger 1993). These data suggest that direct effector functions associated with IgG, including interactions with cells bearing Fc receptors, such as microglia and recruited macrophages, are involved in this process. It is also conceivable that antibody–dependent cellular cytotoxicity directed against cell surface NS1 or even complement-mediated lysis also contributes to the protective effect. The antibody-mediated control of neuronal infection is a well-established mechanism for alphaviruses and coronaviruses, in which case the role of antibody in preventing reactivation of viral infection has been demonstrated (Levine 1991, Levine 1992, Griffin 1997, Lin 1999). In the case of Sindbis virus, there is inhibition of viral release and eventual sequestering of viral RNA within neurons without clearance. This process occurs preferentially at cortical sites of infection and obligates the local retention of virus-specific B cells (Tyor ) in contrast to infection in the spinal cord where the cytokine-mediated (IFN-γ) elimination of viral RNA apparently predominates (Binder and Griffin, 2001). Antibody-mediated mechanisms may also operate in flavivirus encephalitis, as reactivation or recrudescence of CNS disease has been documented in some experimental animal systems and in clinical cases (Section XII). Factors responsible for the prolonged survival of B cells in the CNS are not known, particularly whether these cells represent a resident population of memory cells or require replenishment from peripheral sites (Tschen ).

The beneficial role of antibody responses within the CNS is also supported by clinical studies of JEV and TBEV encephalitis (Burke 1985a, Burke 1985c, Gunther 1997, Han 1988, Ravi 1993, Burke 1985a, Burke 1985c, Gunther 1997, Kaiser 2000, Potula 2003). High levels of IgM correlate with an improved outcome in some cases, presumably reflecting the recruitment of virus-specific B cells into the CNS, as was reported in JEV cases (Burke 1985a, Gunther 1997), and the presence of IgG1 in particular is associated with the control of infection (Thakare ), although IgM can be elevated in severe cases with a fatal outcome (Desai et al., 1994a). The functional activities provided by IgM in the CNS that result in clinical benefit remain unclear, as experimentally, IgM may have limited neutralization activity against encephalitic viruses at least early in infection (Diamond et al., 2003; Hoffman 1978, Diamond 2003). The contribution of complement-fixing activity could be important, as discussed earlier, in view of the fact that complement proteins may be upregulated in the CNS as a result of viral infection (Johnston ).

The inflammatory response to flavivirus infection of the CNS may, in some cases, cause deleterious effects on neuronal function and survival. Immune complexes and autoantibodies to neurofilaments and myelin basic protein in the CSF and serum have been reported in severe JEV and TBEV infections (Thakare 1991, Fokina 1991, Desai 1995b, Thakare 1988) and may reflect an immunopathogenic process rather than nonspecific reactions to viral injury, particularly because of their association with poor outcome. Mechanisms leading to resolution of the acute inflammation in CNS viral infections remain undefined (Bradl and Flugel, 2002). At least three potential factors could be involved, including the expression of IL-4, IL-10, and TGF-β, which have been observed in the CNS of Sindbis virus-infected mice (Wesselingh ) and are known to have immunomodulatory activities that reduce CNS inflammation; induction of apoptosis of infiltrating T cells; or possibly entry of NKT cells, which have been associated with the suppression of inflammatory responses. Further studies are needed to determine whether these or other factors are required to downregulate potentially harmful immune responses in cases where viral disease is eventually controlled.

Neuropathogenesis: WestNileVirus as aModel

Similar to JEV, infections with WNV can be characterized as protean, involving an extraordinary host range, and an abundance of pathologic and virologic data has been obtained from experiments with WNV in birds (Steele 2000, Kramer 2001), horses (Bunning ), and humans (Asnis ). WNV has emerged as dimorphic in its clinical disease, with the apparent shift from typical WN fever to disease of greater severity, including more frequent cases of acute encephalitis in conjunction with the emergence of New World lineage I strains (Cernescu 1997, Asnis 2000). Studies with virulent strains such as New York 1999 WNV have demonstrated the neuropathogenic potential inherent in this virus. It is therefore important to note that lineage differences (type I versus type II) and perhaps genetic variation within lineages may influence the development of CNS disease.

After peripheral inoculation in the mouse model, WNV is believed to infect Langerhans dendritic cells (Johnston ), which migrate to draining lymph nodes, and within 12 to 24 h of infection, viral replication is observed in secondary lymphoid tissue (Diamond 2003, McMinn 1996). Infectious virus is detected in serum within 24 to 48 h of infection (Kramer 2001, Diamond 2003). The course of the early infection is slightly different in wild-type (WT) versus B-cell-deficient mice, as the peak of viremia occurs later in the latter case (day 2 versus days 4 to 6; Fig. 8 ). Shortly afterward, in WT mice, infectious virus is detected in visceral organs such as the spleen, kidney, and heart but not the liver (Diamond 2003, Weiner 1970, Kramer 2001), which may reflect restricted tropism or a high level of reticuloendothelial clearance in this organ. The levels of infectious virus in visceral tissues and serum peaked by day 4 after infection and thereafter diminished (Diamond 2003, Xiao 2001, Kramer 2001), concurrent with a rise in the titer of neutralizing antibodies (Fig. 9 ). In B-cell-deficient mice, replication in peripheral lymphoid tissue and visceral organs follows kinetics similar to those of WT mice; however, virus is not cleared from these sites. These data indicate the profound susceptibility of mice to WNV in the absence of antibody-producing B cells.

Fig 8

Wild-type or B-cell-deficient (uMT) mice were inoculated with WNV in the footpad, and the virus content in serum, peripheral tissues, and brain was measured serially using plaque assay or quantitative polymerase chain reaction. From Diamond , with permission.

Fig 9

Neutralizing IgM and IgG antibody responses in acute WNV encephalitis in the mouse model. From Diamond , with permission.

Virus can first be detected in the CNS by 4 days after peripheral subcutaneous infection in both WT and B-cell-deficient mice. Infectious virus is detected simultaneously in multiple sites in the brain, as well as in the inferior and superior spinal cord, suggesting a hematogenous route of dissemination and⧸or rapid spread within the CNS (Diamond ). However, the route of peripheral inoculation influences the rate of spread to the brain and spinal cord. Infection via an intraperitoneal or intravenous route results in the spread of infectious virus to the brain within 2 days of infection (Kramer and Bernard, 2001), with these animals succumbing to infection several days earlier than those infected by a subcutaneous route. These differences probably reflect the fact that the interaction of virus with peripheral tissues results in the engagement of nonspecific and virus-specific defenses that have an impact on the course of disease. Models of infection that bypass the physiologic route of inoculation must be considered with this limitation in mind, as the balance of viral replication and dissemination versus immune activation can be skewed greatly by intraperitoneal or intravenous injection. Regardless of the route of administration, the course of disease after the onset of CNS infection is rapid and leads to fatal encephalitis beginning by days 9 to 10 postinfection, associated with high burdens of brain-associated virus.

Neutralizing activity associated with the IgM class is detectable by day 4 postinfection in the mouse model (Fig. 9), and this antibody response can confer partial protection against virus challenge in naïve mice (Diamond ). Neutralizing IgG is detectable by day 8 and reaches levels of activity that are 10-fold higher than that of IgM by day 12 postinfection. The importance of secreted IgM in protection against disease is indicated from these data, as well as data presented earlier (Fig. 1). The relative importance of the mechanisms involved, including neutralizing activity per se or complement-assisted priming of the B-cell response to facilitate IgG production, is currently under investigation.

VirusPersistence

Persistent infection in vitro and in vivo has been described in a number of experimental and clinical settings involving flaviviruses. The phenomenon of persistent infection in cell cultures of vertebrate and arthropod origin is well documented (Jarman 1968, Katz 1968, Schmaljohn 1977, Schmaljohn 1979, Igarashi 1979, Loginova 1980, Brinton 1982, Shah 1987, Randolph 1988a, Randolph 1988b, Zhang 1993, Chen 1996, Lancaster 1998, Poidinger 1991, Vlaycheva 2002), and the findings are analogous to many other models of viral infections in vitro where this process can be observed. Flavivirus persistence in vitro typically arises following a cytocidal infection, with survival of a residual population of cells that harbor low levels of replicating virus for long periods of time. The majority of cells in such cultures usually express viral antigen, but only a minority actually are productive of infectious virus. The cultures are generally resistant to superinfection with homologous but not heterologous virus, although in some cases, superinfection can drive virus production in quiescent antigen-positive cells (Schmaljohn and Blair, 1979). The persistence is not dependent on the expression of IFN-α, rather there is evidence for the involvement of host cell antiapoptotic pathways (see later). The infection is not necessarily deleterious for host cells, but in some cases, reduced growth efficiency occurs. Viruses detected in persistently infected cultures frequently undergo phenotypic alterations, including a reduction in plaque size, temperature sensitivity (Randolph and Hardy, 1988b; Shah and Gadkari, 1987), host-range restriction (Randolph and Hardy, 1988b), and loss of neurovirulence for mice (Igarashi, 1979). The genetic basis for the phenotypic change of viral variants has, in some cases, been demonstrated (Vlaycheva and Chambers, 2002). However, these infections are also associated with the emergence of defective viral particles or defective RNAs (Brinton 1982, Debnath 1991, Lancaster 1998) that form the basis for superinfection resistance, and perhaps in part for the attenuation of mouse neurovirulence in flavivirus–resistant strains of mice. Some persistently infected cultures exhibit alterations in the composition of viral proteins or produce truncated forms of NS1 (Chen ), which presumably are believed to reflect deletions in the viral genome associated with the generation of defective interfering viruses. The genome of defective MVE virus in Vero cells was shown to lack a large segment encompassing the prM-E region and a portion of the NS1 region, which leads to the production of truncated NS1 (Lancaster 1998, Blitvich 1999). This finding is entirely consistent with the fact that replicons such as those engineered for Kunjin virus are designed to delete the corresponding region of the genome and are able to establish persistent replication in cell culture without any typical viral cytopathic effects (Varnavski and Khromykh, 1999). At least part of the explanation for the loss of cytocidal activity therefore may result from the deletion of structural proteins, which may provoke an ER overload response and⧸or trigger apoptosis (Duarte dos Santos et al., 2000; Prikhod'ko 2001, Su 2002). The production of infectious virus from persistently infected cultures can be reduced by treatment with a neutralizing antibody (Randolph and Hardy, 1988a), but it is not clear whether this shifts the cells to preferentially harbor defective viral genomes or whether these are actually eliminated.

The expression of apoptotic modulators is an important factor influencing the establishment and maintenance of persistent infection. In a model of JEV infection, the persistence in some cell lines was facilitated by the expression of bcl-2 (Liao ), indicating that the mechanism is analogous to that of alphaviruses, where lytic infection can be converted to persistence in the presence of antiapoptotic proteins (Levine ). However, the mechanisms that confer resistance to JEV-induced apoptosis must differ in some way from those of Sindbis and other viruses, as bcl-2 could fully protect N18 neuroblastoma cells against Sindbis, but not JEV. It is likely that the difference involves divergent pathways for activating the apoptotic process (Jan 2000, Su 2002). It is also possible that some cell lines represent variants that are defective in their apoptotic pathways, allowing virus to adapt in the absence of apoptotic events. At present, there is no evidence that flaviviruses encode proteins with antiapoptotic properties that might influence the process of the persistence.

The relevance of data on persistently infected cells in vitro to the issue of persistent infection in animal models and in apparent human cases is not straightforward. Although evidence for such clinical entities has been reported, the situation is complicated by the fact that there may be overlap between infections that have protracted convalescence and those that have frank neurological sequelae because flavivirus encephalitis is associated with neurologic complications that confer long-term disability among those who survive acute infection (Richter 1961, Vaneeva 1969, Kumar 1993, Hubalek 1999, Baruah 2002, Greve 2002, Haglund 2003, Finley 1980, Huy 1994). In some cases, there is radiographic and pathologic evidence of permanent neurologic injury, including reduced cerebral blood flow, areas of abnormal signal density, and cellular dropout and gliosis (Gunther 1998, Ishii 1977, Shoji 1990, Kumar 1997). Second, the possibility that a recrudescent latent infection really represents reinfection in a host with suboptimal immunity is a difficult question to evaluate. Also, the criteria for defining persistence are somewhat arbitrary. The presence of prolonged expression of IgM has been invoked as a basis for concluding persistence, but only infrequently has the presence of infectious virus been documented in situations where the duration of CNS disease exceeds that which is expected in acute uncomplicated flavivirus encephalitis (2 to 3 weeks).

Mechanisms of flavivirus persistence in vivo may theoretically also involve the formation of DI particles, which reduce infectious viral load, but like their in vitro counterparts, may not be sufficiently cytopathic to cause neuronal cell death. Such defective viruses may be capable of stimulating immune responses and thus be detected serologically. Experimentally, there is certainly evidence that DI particles can inhibit the production of infectious virus in the CNS (Smith 1981, Barrett 1984, Atkinson 1986). The failure to detect infectious virus in both clinical and experimental situations where prolonged evidence of viral activity in the CNS was suspected is consistent with this hypothesis (Pogodina 1983, Ravi 1993). In a study of viral RNA in the brains of mice with the Flv resistance allele, a reduction in genome-length RNA rather than appearance of DI RNAs was observed (Urosevic ), even though flavivirus–resistance in mice is associated with production of D1 virus (Smith, 1981). Thus, data implicating the role of DI particles in in vitro persistence may not correspond directly to an in vivo process.

Encephalitic flaviviruses that have been implicated in the occurrence of persistent infection of the CNS include members of the JEV serogroup and TBEV serogroup (Slavin 1943, Zlotnik 1971, Ogawa 1973, Iliyenko 1974, Pogodina 1983), with the latter having a particular propensity to cause chronic infections. In humans, there have been reports of various types of persistent infection. These include (1) reactivation of latent disease, as in children with JEV encephalitis who experienced recurrent infection months after the primary infection and, in some cases, gave positive virus isolations (Sharma ); (2) chronic progressive disease with cognitive and motor disturbances that resembled subacute sclerosing panencephalitis, years after a primary infection with Russian spring–summer encephalitis (RSSE) virus (Ogawa ); (3) prolonged infection in primary cases of acute encephalitis, with elevation of CSF IgM for as long as 10 months and, in some cases, with virus isolated from spinal fluid (Ravi ); and (4) evidence of prolonged circulation of virus-infected cells without clear-cut CNS disease (Southam ). Protracted cases of TBEV encephalitis, with either cognitive dysfunction or spinal nerve paralysis, occur frequently (Haglund and Gunther, 2003), although the relationship of these outcomes to persistent viral infection is not known. These different types of clinical infections have been mimicked by various experimental models, and it has been documented that virus can be recovered after a protracted phase of infection. For instance, strains of WNV with differing neurovirulence properties were capable of causing prolonged encephalitis of variable severity, ranging up to 5 months in duration in monkeys (Pogodina ). Eventually the viruses were cleared or were rendered replication defective and detectable only by the presence of viral antigen. Some viruses recovered from brains of these animals had undergone attenuation of mouse neurovirulence, indicating a selection for genetic variants and⧸or defective interfering particles. Mechanisms preventing the efficient clearance of virus were not determined, but did not involve defects in the production of neutralizing antibodies. Immunosuppression, either by cytotoxic drugs or even that associated with pregnancy, has been demonstrated to cause prolonged flavivirus disease (Zlotnik 1971, Mathur 1986), and reactivation of infection can be elicited during or after the primary infection by immune depletion (Mathur ). In some cases, as in JEV infection of mice, viral latency was established in T lymphocytes, which could be subsequently activated by immunosuppression to produce infectious virus (Mathur ). These latter findings are consistent with reports of recurrent JEV infection observed in children, but the stimulus for reactivation in such cases is not known (Sharma ).

In the CNS, experimental persistent infections have been related to immune response factors, as well as the ability of neurons to survive viral infection by mechanisms involving apoptotic modulators. Immunological tolerance is also a potential factor that has been associated with other viruses such as LCMV that establish persistent CNS infection, but this does not seem to play any obvious role in flavivirus persistence. Failure to clear brain-associated virus may result from a limited effectiveness of the innate CNS defenses and the peripheral immune responses that are activated in response to flavivirus infections. In addition, elimination of viral RNA may require CD8+ T-cell functions [as described for alphaviruses (Binder and Griffin, 2001)], which are severely restricted due to a lack of class I expression on neurons. The generally resistant state of differentiated neurons toward apoptotic stimuli is also likely to be an important factor (Griffin and Hardwick, 1999). Together, these processes may facilitate the persistence of viral RNA and perhaps foster the evolution of DI particles and noncytopathic-attenuated viral variants. The host immune response, together with properties of the neuronal cellular environment, influences the likelihood of virus perisistence and may both contribute to differences in phenotypes of persisting viruses. For instance, WNV variants that emerged from persistent infection in monkeys tended to acquire attenuation phenotypes (Pogodina ), whereas it has been observed in persistent infection of mice with Sindbis virus that neurovirulent mutants arise (Levine and Griffin, 1992, 1993). Selection pressures imposed by neutralizing antibodies that are produced intraparenchymally by virus-specific B cells may facilitate the emergence of viral variants in some cases, but tissue-specific adaptations are also involved. At present, it is reasonable to believe that the phenomenon of flavivirus persistence in the CNS is a function of many variables, which include the heterogeneity of the neuronal response to injury, encompassing innate interferon-regulated antiviral defenses (Johnston ) and the competence for and propensity towards apoptosis. Interaction of virus–infected cells with virus-specific T cells possessing cytokine-mediated effector functions that can eliminate viral RNA and B cells that can provide antibodies capable of downregulating viral replication influences this process in a region-specific manner. The balance among these factors results in a spectrum of outcomes that may range from either clearance to merely suppression of viral infection, with variable consequences for long-term neurologic function.