Manipulation of cell surface macromolecules by flaviviruses.

Cell surface macromolecules play a crucial role in the biology and pathobiology of flaviviruses, both as receptors for virus entry and as signaling molecules for cell-cell interactions in the processes of vascular permeability and inflammation. This review examines the cell tropism and pathogenesis of flaviviruses from the standpoint of cell surface molecules, which have been implicated as receptors in both virus-cell as well as cell-cell interactions. The emerging picture is one that encompasses extensive regulation and interplay among the invading virus, viral immune complexes, Fc receptors, major histocompatibility complex antigens, and adhesion molecules.

Introduction

Flaviviruses comprise a rich and diverse family of agents that infect a variety of hosts and cause a wide spectrum of disease. Three disease types are recognized for flaviviruses, namely encephalitis, hemorrhagic fever, and fever–arthralgia–rash. Disease distinctions are not absolute, and overlapping pathologies among various flavivirus members are often observed. The ability of flaviviruses to cause such divergent clinical syndromes, associated with virus replication in a number of different organs, has profound implications for the types of cell surface molecules the virus recognizes as receptors. Mutational analyses of the flaviviral E protein have demonstrated a striking ability of flaviviruses to adapt to different cells and receptors. Given the considerable homologies among them, flaviviruses show a remarkable capacity to cause vastly different diseases with a minimum of alterations in the E protein.

The cell surface molecules, which act as receptors for flaviviruses, are only starting to be identified. In addition to providing the molecules involved in virus attachment and penetration, the host cell erects a battery of surface structures that mediate communication with other cells and trigger host defense and pathological processes. Many of these are modulated by flavivirus infection and contribute to the overall picture of pathogenesis.

The Flavivirus Receptor Binding Protein

The flavivirus E protein is a multifunctional protein involved in cell receptor binding (Anderson 1992, Chen 1996, He 1995) and virus entry via fusion with a host cell membrane (Rice, 1996). Some of the functional activities of the E protein, notably membrane fusion, are regulated by interaction with a second viral protein, prM. It is believed that the association of prM with E stabilizes certain pH-sensitive epitopes on the E protein, thereby preventing the conformational changes that normally occur at acidic pH and activate the fusogenic activity of the E protein (Allison 1995, Guirakhoo 1992, Heinz 1994). In addition to its normal role in flavivirus assembly, the prM protein has also been included in novel recombinant formulations in which it is generally coexpressed with the E protein; the resultant E⧸prM complexes have been shown to be immunogenic and protective as vaccines against challenge with several flaviviruses, including Japanese encephalitis virus (Mason ), yellow fever virus (Pincus ), dengue virus (Fonseca ), and tick-borne encephalitis (TBE) virus (Heinz ).

In TBE virus, the majority of extracellular virus is largely free of prM protein due to a late intracellular processing event that generates a carboxy-terminal fragment designated M and which together with the E and C proteins are believed to constitute the protein components of the mature virus particle (Heinz ). Cleavage of prM to M enhances low pH–dependent virus–cell fusion (Guirakhoo ) and infectivity (Guirakhoo 1992, Heinz 1994, Randolph 1990, Shapiro 1972, Wengler 1989). Dengue virions containing prM are still infectious (Randolph ) and bind to permissive cells in a manner that can be blocked using E-specific antibodies (He 1995, Wang 1999). Virus particles containing mainly E and prM also show antibody-enhanced binding to Fc receptor-bearing K562 cells as well as to platelets (Wang ). Thus, in addition to being requisite precursors to mature virus particles, virus particles containing prM possess many properties associated with mature virus particles.

Flaviviruses appear to gain entry to the cell by the endocytic pathway (Rice, 1996). At low pH, the E protein undergoes a conformational change (Allison ) involving dissociation of the E dimer (Stiasny ), thereby exposing a hidden fusion peptide, followed by reorganization of E into a trimer (Allison ), in which the fusion peptide is brought close to the membrane-anchoring carboxy terminus (Ferlenghi ). Remarkably similar structural features and conformational rearrangements have been noted between the flavivirus E protein and the alphavirus E1 (Heinz 2001, Lescar 2001, Pletnev 2001, Strauss 2001), suggesting a common evolutionary origin for these two virion surface proteins.

Considerable homology exists among flaviviral E proteins, raising the possibility that different flaviviruses may have similar receptor-binding motifs. For example, many mosquito-borne flaviviruses contain an RGD sequence (e.g., residues 388–390 of the Murray Valley encephalitis virus E protein), which has been implicated in virulence (Lobigs ) and receptor binding by analogy with integrin-binding motifs (Rey ). Mutagenesis studies of the yellow fever virus (Van der Most ) and Murray Valley encephalitis virus (Hurrelbrink and McMinn, 2001) RGD motifs, however, have cast doubt on the role of integrins in flavivirus attachment or entry.

Studies with TBE virus have identified important determinants for pathogenicity within the suspected receptor-binding site on the upper-lateral surface of domain III (Mandl ). Acquisition of heparan sulfate-binding mutations by passaging TBE in cell culture has also implicated amino acids in this region in receptor binding (Mandl ). The selection of virus mutants on the basis of weak binding to brain membranes has been used with several neurotropic flaviviruses (Holbrook 2001, Ni 1998, Ni 2000) and has identified a variety of mutations within domain III as well as other regions of E. For dengue virus, blocking of virus cell binding correlates more closely to virus neutralization for mAb 3H5 than for mAb 1B7 (Wang ). This may suggest that mAb 3H5 neutralizes dengue virus predominantly by blocking virus–cell attachment, whereas mAb 1B7 neutralizes dengue virus largely by a postattachment mechanism. The mAb 3H5-binding site on the dengue viral E protein has been partly characterized (Hiramatsu 1996, Megret 1992, Trirawatanapong 1992) and probably encompasses, at a minimum, residues 383–385 (Hiramatsu ) within domain III. More recent data involving a larger number of monoclonal antibodies indicate that mAbs that interact with domain III are in fact the most effective blockers of virus–cell attachment (Crill and Roehrig, 2001). A putative heparan sulfate-binding site on the dengue-2 E protein is also located within this region (Chen ), and comparative sequencing of dengue type 2 genomes has implicated amino acid 390 of the E protein as a major determinant of pathogenicity (Leitmeyer ). The pH-dependent conformational “hinge” region (between domains I and II) of the E protein has also been implicated in virulence, receptor interaction, and⧸or membrane fusion (Hurrelbrink 2001, Lee 1997, Monath 2002). Further mutagenesis studies will undoubtedly help define the sites of the E protein involved in flavivirus–cell macromolecule recognition.

Cell Targets for Flaviviruses

Dendritic Cells

Transmission of flaviviruses to humans generally occurs via the bite of an infected mosquito or tick. In the case of dengue, inoculated virus is thought to first replicate in skin Langerhans (dendritic) cells (Palucka 2000, Taweechaisupapong 1996a, Taweechaisupapong 1996b, Wu 2000). Dendritic cells have also been shown to be involved in the transport of intradermally inoculated West Nile virus to local draining lymph nodes, with a subsequent accumulation of leukocytes (Johnston ). It is likely that dendritic cells will prove to be efficient carriers of a wide number of flaviviruses from their cutaneous site of infection to lymphoid and possibly other tissues.

Given the importance of dendritic cells in initiating immune responses (Banchereau ), they probably play a pivotal role in stimulating host defense against invading flaviviruses. Dengue virus infection of immature myeloid dendritic cells has been shown to induce their maturation accompanied by the expression of major histocompatibility complex (MHC) class I and II antigens; the costimulatory molecules CD40, CD80, and CD86; and the dendritic cell marker CD83 (Libraty ). Such changes were seen in both dengue-infected and bystander cells, indicating that upregulation of cell surface molecules could be a consequence of virus infection as well as virus-induced cytokine expression. Similarly, Langerhans cells infected with West Nile virus, as well as an alphavirus, Semliki Forest virus, express increased cell surface MHC class II and appear to undergo maturation to a cell type similar to lymphoid dendritic cells (Johnston ). The efficient presentation of both MHC class I– and II–associated viral peptides on the surface of dendritic cells permits the generation of potent cytotoxic and helper T cell responses (see also Section V,A).

Monocytes and Macrophages

Monocytes and macrophages have long been recognized as major targets of flavivirus replication in the human host (Halstead 1989, Halstead 1977b, Scott 1980). They are also important host cells for the antibody-enhanced replication of certain flaviviruses (see Section IV,C). Because of their presence in the circulation, blood monocytes may be particularly important to the pathogenesis of hemorrhagic viruses, such as dengue. Because most of the pathological changes associated with dengue virus are hemostatic in nature, it is suspected that blood cells, particularly virus-infected blood monocytes, orchestrate many of these effects.

Dengue virus–infected human monocytes have been shown to be potent sources of vasoactive cytokines such as tumor necrosis factor (TNF)-α (Anderson ) and interleukin (IL)-1β (Chang and Shaio, 1994). Monocytes are also known producers of several other vasoactive mediators, including IL-6, platelet-activating factor (PAF), prostaglandins, thromboxanes, leukotrienes, and nitric oxide (Bulger 2000, Funk 2001, Lefer 1989, Maruo 1992, Montrucchio 2000, Szabo 1999), any of which could have powerful effects on endothelial cell physiology. A crucial aspect in understanding dengue pathogenesis will be the identification of additional vasoactive mediators, which trigger the key dysfunctional events in vascular integrity.

Various tissue macrophages are undoubtedly important in the pathogenesis of flaviviral diseases but have, to date, not received much attention. Skin mononuclear cells, pulmonary, splenic, and thymic macrophages and liver Kupffer cells have been recognized carriers of viral antigen (Halstead, 1989). In the liver, virus or viral antigen has been found in Kupffer cells and hepatocytes in infections with yellow fever (Monath ) and dengue (Bhamarapravati 1967, Hall 1991, Halstead 1989, Rosen 1989). Destruction of Kupffer cells, possibly by apoptosis, has been reported in the liver of some patients with fatal dengue (Huerre ). Primary cultures of Kupffer cells apparently undergo an abortive infection with dengue virus in which viral antigen but no progeny virus is produced (Marianneau ).

Endothelial Cells

Many flaviviruses invade either visceral or central nervous system tissues following initial replication in dendritic cells, monocytes, or macrophages. Often this necessitates a transfer of virus across blood vessel endothelial layers.

For neurotropic flaviviruses, endothelial cells of the cerebral microvasculature constitute a barrier that must be overcome in order to gain access to the central nervous system. How this occurs remains uncertain. Transendothelial passage of virus may direct infection of cerebral microvascular endothelial cells, may transport across the endothelial layer, or both (Dropulic and Masters, 1990). Japanese encephalitis virus has been observed electron microscopically to traverse mouse cerebral endothelial cells by transcytosis (Liou and Hsu, 1998). Alternatively, virus may spread from blood vessels to the olfactory neuroepithelium and from there to olfactory neurons (McMinn 1996, Monath 1983).

Even normally nonneurotropic flaviviruses may occasionally invade the central nervous system under certain conditions. Modulation of the blood–brain barrier by anesthetics (Ben-Nathan ) or lipopolysaccharide (Lustig ) has been reported to facilitate neuroinvasion by a normally noninvasive strain of West Nile virus. Flaviviruses may also trigger the production of soluble factors that perturb the integrity of the blood–brain barrier, leading to increased leakage of proteins and cells into the central nervous system (Chaturvedi ). These studies indicate that even nonneurotropic flaviviruses may infect tissues of the central nervous system or otherwise affect the integrity of the blood–brain barrier under special circumstances.

Transendothelial migration of individual leukocytes (e.g., lymphocytes, monocytes, neutrophils, eosinophils) is regulated in a highly specific manner by the differential expression of selected adhesion molecules on endothelial cells (reviewed in Crockett 1998, Lowell 1999. Flaviviruses, including dengue (Anderson ) and West Nile (Shen ) viruses, activate endothelial cell adhesion molecule expression by either direct (virus-mediated) or indirect (cytokine-mediated) mechanisms (see Section V,C). In the presence of leukocyte-attracting chemokines, such virus-triggered activation of the vascular endothelium may contribute toward the migration of leukocytes into extravascular tissues. In addition to being a mechanism for virus dissemination, this process may also be a factor in phenomena such as leukopenia and particularly neutropenia (loss of circulating leukocytes, neutrophils) often observed in flavivirus, particularly dengue, infection (reviewed in Halstead, 1989). Due to the lack of suitable animal models for severe dengue disease, i.e., dengue hemorrhagic fever (DHF) or dengue shock syndrome (DSS), there are difficulties in assessing the roles of such events, particularly the identification of adhesion molecules mediating the transendothelial migration of neutrophils using blocking antibodies against specific integrins, as has been performed for other disease states (Doerschuk 1990, Gao 1994, Issekutz 1993, Laberge 1995, Springer 1995).

The hallmark feature of increased vascular permeability in hemorrhagic flavivirus (e.g., dengue) infection suggests that vascular endothelial cells may mediate the fluid leakage and hemorrhaging that occur in DHF⧸DSS. Endothelial cells line the inner surface of blood vessels and play essential roles in maintaining an antithrombogenic surface and regulating vascular permeability. Increased vascular permeability can arise from a variety of mediators associated with acute inflammation and shock (Bulger 2000, Funk 2001, Lefer 1989, Michel 1988, Montrucchio 2000, Schnittler 1990). It is thought that vascular permeability is largely controlled by changes in endothelial cell–cell contact, which result in gap formation, thus allowing for fluid exchange between blood and interstitial tissue fluid (Michel, 1988). An electron microscopic study of endothelium from DHF biopsy samples revealed the occasional presence of gaps (Sahaphong ), thus providing evidence that endothelial cell features may indeed be perturbed during DHF⧸DSS.

Although dengue virus infects endothelial cells in vitro (Andrews 1978, Avirutnan 1998, Killen 1993), there is no evidence that endothelial cell infection occurs clinically, as neither virus particles nor viral antigen has been detected in the endothelium of tissue specimens (Halstead 1988, Halstead 1989, Sahaphong 1980), in contrast to that seen in cases of ebola (Zaki ) or hantaan hemorrhagic fever (Gavrilovskaya 1999, Wang 1997). It is likely that dengue virus mediates endothelial cell activation via an indirect route, involving blood monocytes, which are a major cell target for dengue virus infection (Halstead 1977b, Scott 1980). A major candidate event in such a route is the activation of endothelial cell adhesion molecules by a factor(s) (particularly TNF-α) produced by dengue virus–infected blood monocytes (Anderson ).

TNF is a key cytokine in a variety of normal and pathological immune responses, including immunoregulation, regulation of cell proliferation, cytotoxicity, and in the mediation of endotoxic shock (Fiers 1991, Tartaglia 1992, Tracey 1993, Vassalli 1992). Monocyte-derived TNF-α appears to play a pivotal role in dengue-associated endothelial cell activation (Anderson ) and may be an important effector in the manifestation of DHF⧸DSS. Support for the clinical significance of this observation comes from observations of elevated TNF levels in the sera of patients with severe dengue disease (Green 1999b, Hober 1993, Vitarana 1991, Yadav 1991). Taken together, current evidence indicates that dengue virus represents a rather unique group of viruses that target monocytes, thereby triggering the production of factors such as TNF-α, which in turn affect other cell targets, including endothelial cells. While the overall picture of endothelial cell dysfunction in DHF⧸DSS is obviously more complex than can be explained by any single factor, the role of TNF in dengue pathogenesis would seem to merit particular attention.

Current knowledge of endothelial cell responses observed in endotoxic shock may be instructive for the understanding of vascular leakage in DHF⧸DSS. Plasma leakage induced by endotoxin (lipopolysaccharide, LPS) from gram-negative bacteria encompasses a complex cascade of processes, including activation and functional alteration of endothelial cells. Major mediators of endothelial cell perturbation in endotoxic shock are LPS itself, as well as cytokines such as TNF-α and IL-1β (Bevilacqua, 1993). These factors can modulate endothelial cell function to varying degrees by activating cytokine and vasoactive factor release (Rink 1996, Shanley 1995), upregulating adhesion molecule expression (Bevilacqua 1993, Luscinskas 1991, Moser 1989, Smith 1989), and mediating transendothelial migration of specific leukocytes (Issekutz 1995, Luscinskas 1991, Morzycki 1990, Moser 1989, Smith 1989). Additional factors, particularly lipid mediators such as PAF, leukotrienes, thromboxanes, and prostaglandins, may contribute to further endothelial cell dysfunction, including vascular leakage (Bulger 2000, Funk 2001, Lefer 1989, Montrucchio 2000). While the involvement of these vasoactive mediators is recognized in endotoxic shock, more needs to be learned of their role in the vascular dysfunction that occurs in severe dengue disease.

Lymphocytes

Although lymphocytes are potently involved in the host response and immunopathology of flavivirus (especially dengue) diseases, their role as virus-permissive host cells is unclear. Dengue virus has been identified in circulating B cells from acutely ill dengue patients by immunocytochemistry and by recovery of infectious virus after passage in mosquitoes (King ). In vitro studies showed that cells and cultured cell lines of both B and T cell derivation could be infected with dengue virus (Bielefeldt-Ohmann 2001, Kurane 1990, Marchette 1978, Mentor 1997, Sung 1975, Takasaki 2001, Theofilopoulos 1976). Continued passage of dengue virus in lymphoblastoid (Raji) cells can give rise to dengue virus variants capable of replication in human lymphocytes (Brandt ). Interestingly, lymphocytes do not appear to undergo antibody-enhanced dengue virus infection (Brandt 1979, Kurane 1990), even though B cells do have Fc receptors (Dijstelbloem ; see Section IV,C).

Neural Cells

The initial stages of pathogenesis for neurotropic flaviviruses appear to be common for flaviviruses in general in that the virus progresses from the subcutaneous site of inoculation to lymph nodes, followed by viremia and replication in extraneural tissues. Invasion into the central nervous system is marked by high virus titers in the brain and detectable virus or viral antigen in neurons (Albrecht, 1968). Cell destruction in tick-borne encephalitis may be less extensive than that seen in herpes simplex type 1 encephalitis (Studahl ), although this is variable and may involve considerable inflammation (Chu 1999, Matthews 2000, Suzuki 2000). Susceptible cell types include both neurons and glial cells (Chu 1999, Ramos 1998, Steele 2000).

Basophils⧸Mast Cells

As notorious producers of vasoactive mediators, mast cells have been a source of controversial speculation for years in dengue pathogenesis. Cells resembling degranulated mast cells have been reported in skin perivascular infiltrates from DHF⧸DSS cases (Bhamarapravati ). Dengue patients showed elevated levels of urinary histamine (a major granule product of mast cells), which correlated with disease severity (Tuchinda ), suggesting that mast cells may have a contributory role in the pathogenesis of dengue. Although antihistamine treatment does not resolve shock in severely dengue-diseased patients (Halstead, 1989), histamine is only one of several potent vasoactive factors produced by mast cells (Benyon 1991, Bradding 1993, Galli 1984, Grabbe 1994, Marshall 1994, Moller 1991, Moller 1993, Moller 1998, Nilsson 1995, Schwartz 1984), some of which could cause vascular dysfunction in dengue infection. DHF⧸DSS patients have been reported to have elevated serum levels of IgE (Pavri ), which has been speculated to relate to IgE-triggered histamine release in the manifestation of shock (Pavri and Prasad, 1980).

Mast cells reside mainly in the tissues, often closely associated with blood vessels (Alving 1991, Anton 1998, Pesci 1996, Pulimood 1998, Selye 1966, Selye 1968). They are present in large numbers in the skin (Marshall ), where transmission of insect-borne flaviviruses occurs. Basophils, however, comprise about 1% of total circulating cells and would be accessible to virus in the blood. Dengue virus infects basophil⧸mast cell–like KU812 cells in an antibody-enhanced manner, coupled with the release of vasoactive cytokines, IL-1β and IL-6 (King 2000, King 2002). This cell line, which can be differentiated easily toward either a basophil or mast cell phenotype (Saito ), may provide further insights into potential roles for basophils and mast cells in dengue disease.

Dengue patients show increased serum levels of anaphylatoxins C3a and C5a (Malasit, 1987), which can attract (Nilsson ) and activate (Kownatzki, 1982) mast cells. Among the expected mast cell secretion products would be vasoactive factors, including histamine, which has been detected in elevated amounts in the urine of dengue patients (Tuchinda ).

Platelets

Evidence for platelet involvement in dengue pathogenesis comes from at least two (probably related) sources. First, thrombocytopenia (loss of circulating platelets) is one of the most consistent clinical features of severe dengue infection (Halstead, 1989). Second, viral immune complexes have been detected on platelets from dengue patients (Boonpucknavig 1979, Phanichyakarn 1977a). Functional studies on platelets in dengue-diseased individuals have been sparse, but include a markedly reduced half-life (Mitrakul ), deficient ADP release (Mitrakul ), increased adhesiveness (Doury ), increased tagging by complement fragments (Malasit, 1987), and increased release of β-thromboglobulin and platelet factor 4 (Srichaikul ). There is also evidence for platelet activation in dengue patients (Doury 1976, Krishnamurti 2001, Srichaikul 1989). Although these results relate to a variety of platelet functions, they do indicate a general alteration in platelet physiology, which is consistent with platelet involvement and triggering of thrombocytopenia in dengue disease.

Dengue virus has been recovered from washed patient platelets (Scott ), and virus has been reported to bind to platelets in the absence of antibody as assayed using immunofluorescence and immunoperoxidase techniques (Funahara ). However, the levels of antibody-independent bound virus are very low compared to the levels of virus bound in the presence of dengue-specific antibodies (Wang ). As noted earlier, dengue immune complexes have been demonstrated on platelets from dengue patients (Boonpucknavig 1979, Phanichyakarn 1977a). Weiss and Halstead (1965) originally proposed the possibility that dengue virus interactions with platelets might be involved in the thrombocytopenia observed in severe dengue disease. The finding that dengue virus binding to platelets is dependent on a virus-specific antibody is consistent with epidemiological and experimental data linking preexisting host antibodies to an increased risk of DHF⧸DSS (reviewed in Halstead, 1990).

Several other viruses have been shown to bind directly to platelets (Bik 1982, Danon 1959, Forghani 1983, Larke 1970, Lee 1993, Zucker-Franklin 1990). Platelet association may stabilize or protect blood-borne viruses (Larke and Wheelock, 1970) and may function as a mechanism of hematogenous dissemination (Forghani and Schmidt, 1983). Virus binding to platelets has been suggested to be a contributing mechanism to thromobocytopenia arising from infections with vaccinia (Bik ), chikungunya (Larke and Wheelock, 1970), and rubella (Bayer ). Thrombocytopenia in these virus infections is generally much milder than that observed in severe dengue disease.

Levels of dengue virus in the blood can exceed 107 infectious units⧸ml (Gubler 1988, Monath 1994). Such high viremic titers are likely necessary to ensure infection and transmission of the obligate mosquito intermediary host (Monath, 1994). Assuming a reasonable particle:infectivity ratio of 100:1, virus particle titers in blood may rival normal platelet counts (3 × 108⧸ml). Such parity between numbers of virus particles and platelets suggests that antibody-enhanced binding of virus to platelets may have a profound effect on platelets. Circulating virus-immune complexes are detected in DHF⧸DSS, and levels of immune complexes have been correlated with severity of disease (Ruangjirachuporn ) and some of these are platelet associated (Boonpucknavig 1979, Phanichyakarn 1977a). These observations suggest that sufficient binding of virus immune complexes to platelets may occur to tag the majority of circulating platelets. Such an event could lead to immune clearance by the reticuloendothelial system, thereby precipitating the thrombocytopenia frequently associated with severe dengue disease.

It is likely that molecules other than Fc receptors on the platelet surface may mediate antibody-enhanced binding of dengue virus (Wang ). Drug-induced thrombocytopenias provide interesting examples in this regard. It is known that given the appropriate accessory ligand (i.e., drug), IgG can bind to platelets through either the Fc receptor or other surface proteins. A variety of clinical thrombocytopenias are known that involve an immune component in pathogenesis. Many of these reflect activities of host antibodies, which react with proteins on the surface of platelets. These antibodies may be autoimmune in nature (i.e., antibodies that bind to platelet surface molecules) or dependent on a third party ligand (drug or protein), which then induces binding of the antibody–ligand complex to either the platelet Fcγ receptor or to another surface protein. For example, a number of individuals are susceptible to drug-dependent thrombocytopenia when administered drugs such as heparin or quinine⧸quinidine (Aster 1989, Hackett 1982). While heparin-dependent antibodies bind to the platelet Fcγ receptor (Adelman 1989, Chong 1989a, Chong 1989b, Kelton 1988), quinine⧸quinidine-dependent antibodies bind to platelet protein heterodimers GPIIb⧸IIIa and GPIa⧸IX (Berndt 1985, Chong 1983, Christie 1987, Devine 1995). This latter category of immune-mediated thrombocytopenia may be relevant to the understanding of dengue-associated thrombocytopenia, as patient antibodies mediate dengue virus binding to platelets via a platelet surface protein other than the Fcγ receptor (Wang ).

Communication between platelets and endothelial cells is a frequent intermediate step in certain events such as platelet adhesion, aggregation, and regulation of vascular permeability. How this occurs in dengue infection and what the effects are on endothelial cell function are unknown. Binding of viruses to platelets can have potentially profound immunological effects [e.g., the stimulation of TGF-β release by platelets bound by Epstein–Barr virus (Ahmad and Menezes, 1997)]. In light of reports of altered platelet function in dengue patients, discussed earlier, there is a tantalizing need to determine the immunological consequences of antibody-enhanced dengue virus binding to platelets in terms of platelet as well as endothelial cell physiological responses.

Many products of complement activation can also be deposited on platelets (Devine, 1992). In view of evidence for complement activation in severe dengue disease (Halstead 1989, Malasit 1987), binding of complement products might play a role in the immune destruction of platelets leading to thrombocytopenia. Platelets display surface receptors, e.g., C1q receptor (Peerschke 1987, Peerschke 1998), membrane cofactor protein (Seya ), and decay-accelerating factor (Devine ), for specific components of complement activation. In addition, the platelet surface can act as a substrate for the deposition of C3dg and C5b-9 (Devine, 1992). Fragments of C3 have been detected on the platelets of DHF⧸DSS patients (Malasit, 1987).

In addition to immune complex deposition on platelets, thrombocytopenia associated with DHF⧸DSS might also arise by the immune destruction of platelets through antiplatelet autoantibodies. Antiplatelet autoantibodies have been reported in the sera of dengue patients (Lin ), although they have also been detected in patients recovering from a variety of viral infections (Imbach, 1994). Antiplatelet antibodies are strongly linked to the pathogenesis of immune-mediated thrombocytopenias, such as idiopathic thrombocytopenic purpura (Winkelstein and Kiss, 1997).

Cell Targets: An Overview

While this brief discussion of cell targets for flaviviruses is by no means complete, it highlights some of the major interactions as they relate to pathogenesis. Because pathogenesis is probably best understood for dengue, Fig. 1 illustrates the interactions of hemorrhagic flavivirus (e.g., dengue) with cell targets both within and outside the vascular system.

F

Model showing surface interactions of hemorrhagic flavivirus (dengue) with extra- and intravascular cell targets. Intravascularly, the presence of subneutralizing levels of antivirus antibodies stimulates virus attachment to platelets and infection of monocytes. This results in immune complex deposition on platelets and secretion of vasoactive factors from virus-infected monocytes. Among such vasoactive factors are cytokines, particularly TNF-α, which activates increased surface expression of adhesion molecules on endothelial cells. Extravascularly, virus infection of tissue macrophages, mast cells, and dendritic cells may result in the release of additional factors, which contribute to endothelial cell perturbation.

Cell Surface Macromolecules Involved in Flavivirus Attachment

Glycosaminoglycans

Glycosaminoglycans and proteoglycans (i.e., proteins bearing glycosaminoglycans) are important cell surface molecules involved in a variety of ligand recognition and cell signaling processes (Gallo, 2000). Because glycosaminoglycans are widely distributed on cells, they are attractive candidates as virus receptors. Some degree of specificity (i.e., virus tropism) may arise from the compositional heterogeneity of glycosaminoglycans, as well as quantitative differences in the degree of expression on various cell types.

Flaviviruses seem to share, with a large number of virus families, the ability to bind glycosaminoglycans (Birkmann 2001, Dechecchi 2000, Dechecchi 2001, Duisit 1999, Feldman 1999, Feldman 2000, Giroglou 2001, Goodfellow 2001, Heil 2001, Hsiao 1999, Hulst 2000, Hulst 2001, Lin 2000, Liu 2002, Patel 1993, Rue 2002, Shukla 1999, Shukla 2001). Glycosaminoglycans such as heparin and its structural analogues have been investigated for their ability to bind dengue virus and thereby to gain insights as to the structural requirements for dengue receptors. Potential glycosaminoglycan-binding motifs have been identified on the dengue viral E protein at two sites, the best characterized of which appears to be composed of amino acids 188, 284–295, and 305–310 and which may also play a role in virus–cell attachment (Chen ). Heparin (minimum of 10 carbohydrates) and an uncharacterized highly sulfated heparin sulfate isolated from bovine liver were found to show the best binding to dengue E protein (Chen ). Attachment of dengue virus to human hepatoma cells has also been reported to be inhibited by heparin (Hilgard and Stockert, 2000). A further study involving a panel of natural and synthetic polyanionic, sulfated compounds suggested that binding of the dengue E protein required a highly sulfated (and highly charged) oligosaccharide with a minimum size of 39Å and a high degree of structural flexibility (Marks ).

The role of glycosaminoglycans in natural (i.e., nontissue culture-adapted) strains of flaviviruses needs to be studied further. It has long been recognized that dengue virus passaged in various host cell types can give rise to virus variants with altered cell specificity (Brandt 1979, Halstead 1984a, Halstead 1984b, Halstead 1984c). Passage-dependent mutations of the dengue virus E protein at a number of different amino acid residues have been documented (Lee ). Following passage of TBE virus in cultured BHK-21 cells, virus mutants were selected that contained more positively charged amino acids in the putative receptor-binding region of the E protein, resulting in dependence on cell surface heparan sulfate (Mandl ). Such mutants were diminished in their neurovirulence in mice as well as in their replication in primary chicken cells and plaque formation in porcine kidney cells (Mandl ). A large number of other viruses have also been shown to undergo loss of virulence upon adaptation to cell culture associated with heparan sulfate utilization (Bernard 2000, Byrnes 2000, Klimstra 1998, Klimstra 1999, Lee 2000, Neff 1998, Sa-Carvalho 1997).

CD14

CD14 and the Toll-like receptor (TLR) pattern recognition receptors are involved in the innate response to lipopolysaccharide and other microbial products (Diamond 2000, Imler 2000). A role for CD14 and TLR4 has been found for respiratory syncytial virus (RSV) (Kurt-Jones ), suggesting that these receptors may have a broader involvement in host response than previously thought. A possible role for CD14 in dengue infection has been postulated on the basis of inhibition of dengue virus infection of human monocytes with bacterial lipopolysaccharide (Chen ). However, this has been disputed (Bielefeldt-Ohmann ) and requires further investigation.

Fc Receptors

As indicated earlier, flaviviruses are capable of initiating infection of appropriate host cells through as yet largely unidentified primary receptors. In addition, a number of flaviviruses are capable of using subneutralizing levels of virus-specific antibodies to attach to and gain entry to cells bearing Fc and⧸or complement receptors (Cardosa 1983, Halstead 1982, Halstead 1977a, Schlesinger 1981a) by a process known as antibody-dependent enhancement (ADE) of infection (Table I ). ADE has been documented for dengue (Halstead ), West Nile (Peiris and Porterfield, 1979), yellow fever (Schlesinger and Brandriss, 1981b), tick-borne encephalitis (Phillpotts ) and Japanese encephalitis (Cecilia and Ghosh, 1988) viruses. Early work with dengue virus and monocytes differentiated between trypsin-sensitive and trypsin-resistant cell surface molecules as the putative receptors for antibody-independent and antibody-dependent infection, respectively (Daughaday ).

TABLE I

FcRs for Antibody-Enhanced Infection of Dengue Virus

		Dengue virus replication
Cell	FcγRa	Ab independent	Ab enhanced	FcγR for ADE
Monocyte	I,II,III	Yesb	Yesb	I, IIf
Dendritic cells	II	Yesc	Noc	None
Mast cell⧸basophil	I,II	Nod	Yesd	Unknown
Kupffer cell	I,II,III	Noe	Unknown	Unknown

Compiled from van de Winkel 1991, Dijstelbloem 2001, Okayama 2000, Anselmino 1989, Tuijnman 1993.

From Halstead and O'Rourke (1977).

From Wu and Libraty .

From King .

Abortive infection, but expressing viral antigen (Marianneau ).

From Littaua and Kontny .

To date, dengue virus appears to be the only flavivirus in which strong evidence exists for antibody-dependent enhancement as a major contributing factor to severe disease (Halstead 1980, Thein 1997). Severe dengue disease, encompasing conditions known as dengue hemorrhagic fever⧸dengue shock syndrome, involves several well-defined hemostatic abnormalities, including the leakage of plasma into interstitial spaces, as well as thrombocytopenia and bleeding (Halstead 1990, Kurane 1994). The potential to cause severe hemorrhagic disease is a general property of dengue viruses and is not limited to any one viral serotype (Gubler 1998, Rigau-Perez 1998). Although different strains of dengue may influence the severity of hemorrhagic symptoms (Leitmeyer 1999, Rico-Hesse 1997), it is also generally accepted that pathogenesis depends on immunopathological processes (Rothman and Ennis, 1999). Thus the roles of prior immunity, antibody-enhanced virus infection, and immune-mediated pathologic effects on the vascular system are key points in understanding the pathogenesis of dengue hemorrhagic disease.

While the pathogenesis of severe dengue disease is not completely understood, it is clear from laboratory and epidemiological studies that a considerable risk factor is prior immunity. Severe dengue disease, DHF⧸DSS, rarely occurs in seronegative individuals suffering their first dengue infection, but instead occurs in individuals who have preexisting dengue viral antibodies, either from a previous infection or from passive antibody transfer, e.g., following maternal transmission of antibodies to the fetus (Kliks 1988, Kliks 1989). Estimates suggest that 99% of children suffering from DHF⧸DSS have preexisting immunity from a prior dengue virus infection (Halstead, 1988). Consequently, from this and other studies, it has been calculated that prior exposure to dengue increases the risk for hemorrhagic disease in a second dengue infection by at least 15-fold (Halstead 1980, Thein 1997). Preexisting serum antibodies can potentiate virus infection by the mechanism of antibody-dependent enhancement, giving rise to amplified virus replication and to an increased potential for the development of hemorrhagic symptoms (Halstead, 1989). Viremic titers are higher in secondary dengue infections in both humans (Gubler ) and experimental monkeys (Halstead ). Antibody-enhanced dengue virus infection of human blood monocytes is necessary for the production of endothelial cell activators (Anderson ), thereby providing a link between antibody-dependent enhancement and alteration of endothelial cell properties, which might contribute to vascular permeability in dengue infection.

For certain other viruses, e.g., influenza (Tamura ) and HIV (Takeda 1990, Takeda 1992), distinct “neutralizing” and “antibody-enhancing” epitopes have been identified on the respective viral attachment proteins. Surprisingly, no systematic approach has yet been undertaken to identify regions on the E protein that are essential for ADE, even though this issue was raised as a challenge to research on dengue many years ago (Halstead, 1988).

Human Fcγ receptors are currently categorized into three classes: FcγRI (CD64), FcγRII (CD32), and FcγRIII (CD16). While FcγRI shows high affinity for monomeric IgG, FcγRII and FcγRIII bind monomeric IgG poorly and are more likely involved in binding immune complexes (Dijstelbloem ). FcγRII is the most widely distributed, being expressed on most circulating leukocytes (van de Winkel and Anderson, 1991). Monocytes express all three FcγRs to varying degrees (van de Winkel and Anderson, 1991), although FcγRI and FcγRII predominate, whereas FcγRIII appears to be limited to a subpopulation (∼10%) of monocytes (Anderson 1990, Passlick 1989). FcγRIII constitutes the major FcγR on macrophages (Fanger ), although FcγRI and FcγRII are also present (Tuijnman 1993, van de Winkel 1991). It is also important to recognize that FcR expression on cells, including macrophages, can vary depending on the microenvironment (Tomita ).

Although strong evidence exists for FcγR involvement in ADE of dengue virus, the participating FcγRs in vivo have not yet been identified rigorously. In cultured cell lines (monocytic U937 or erythroleukemic K562 cells), FcγRI (Kontny ) and FcγRII (Littaua ) have been shown to mediate ADE of dengue virus infection. That FcγRI has the ability to mediate ADE of dengue has been demonstrated using COS cells transfected with FcγRI (Schlesinger and Chapman, 1999).

Dengue and DHF patients show elevated serum levels of interferon (IFN)-γ (Kurane ). Because IFN-γ can upregulate both MHC class I and II molecules as well as FcγR (particularly FcγRI) expression in monocytes (Erbe 1990, Perussia 1983), the chances for ADE may be increased, thereby creating a vicious cycle involving positive cytokine feedback and virus amplification (Kurane and Ennis, 1992). IFN-γ has been shown to enhance ADE of dengue virus infection of human monocytic U937 cells (Kontny ), although any enhancing effect on dengue infection of peripheral blood monocytes may be negated by the antiviral properties of IFN-γ (Sittisombut ).

Mast cells and basophils express mainly FcγRII (Anselmino 1989, Okayama 2001a, Wedi 1996) and some (IFN-γ–inducible) FcγRI (Okayama 2000, Okayama 2001b) as well as the high-affinity FcϵRI for IgE (Guo 1992, Sperr 1994). As noted previously, the basophil⧸mast cell KU812 cell line exhibits antibody-enhanced dengue virus infection and produces vasoactive cytokines (King ).

Although FcγR-mediated ADE of flaviviruses has been examined extensively as a mechanism for virus amplification, the biological consequences for the participating host cell are not well understood. Because FcγR-mediated cell signaling is complex, the functional effects of virus–antibody interactions with cell surface FcγRs need to be investigated. Monocytes infected with dengue virus in the presence of antibody release cytokines such as TNF-α (Anderson ). Induction of TNF-α requires infectious virus (Anderson ), suggesting that virus replication (or perhaps expression of one or more crucial viral genes) is responsible for the stimulation of TNF-α release. Therefore, in this case, the FcγR is likely facilitating antibody-enhanced virus replication rather than providing a signal triggered by virus binding to the FcγR. Similarly, antibody-enhanced dengue virus infection of KU812 basophil⧸mast cells produces IL-1β, IL-6 (King ), and selected chemokines (King et al., 2002). Suppressive effects of antibody-enhanced flavivirus or alphavirus infection on monocyte cytokine secretion have also been reported (Lidbury 2000, Yang 2001).

Both activating (FcγRI, FcγRIIa, and FcγRIIIa) and inhibitory (FcγRIIb) forms of FcγRs exist, which mediate signal transduction via a cytoplasmic immunoreceptor tyrosine-based activation motif (ITAM) or inhibitory (ITIM) motif, respectively (Dijstelbloem ). The ITAM and associated molecules are necessary for the endocytosis of FcR-bound immune complexes (Amigorena and Bonnerot, 1999) and therefore play a likely role in the initiating events of antibody-enhanced flavivirus infection. Although not necessary for FcγRII, an accessory subunit (homo- or heterodimeric γ or ζ chains) is required for signaling through FcγRI and FcγRIIIa (Ravetch, 1994). A further FcγR (FcγRIIIb) lacks transmembrane and cytoplasmic domains and is instead anchored to the cell surface membrane via a glycosylphosphatidylinositol (GPI) linkage (Selvaraj 1988, Simmons 1988). It apparently does not participate in signal transduction and has been speculated to sequester and accumulate immune complexes at specific sites on the cell surface (Huizinga 1988, Selvaraj 1988).

The roles of activating and inhibitory FcRs in viral ADE have not yet been ascertained. Activating FcRs are expressed on monocytes, macrophages, granulocytes, natural killer (NK) cells, and platelets but not on most lymphocytes (Dijstelbloem ). Inhibitory FcRs, however, are found on B cells, dendritic cells, and macrophages (Dijstelbloem ). Interestingly, ADE of dengue virus is best documented for monocytes⧸macrophages and related cell lines (Halstead, 1989). In contrast, lymphocytic cells (Brandt 1979, Kurane 1990) and dendritic cells (Wu ) do not appear to support antibody-enhanced dengue virus infection. Whether this is due to differential expression of activating versus inhibitory FcRs remains to be investigated.

FcRs for IgE (primarily the high-affinity FcϵRI) are expressed on cells such as monocytes, macrophages, mast cells, basophils, and dendritic cells and are structurally related to FcγRs (Ravetch, 1994). Their role in binding IgE and⧸or immune-complexed flaviviruses, such as dengue, remains unexplored. Similarly unexplored is the potential role of the neonatal Fc IgG receptor (FcRn), structurally related to MHC class I and involved in IgG transport across cells (Ghetie and Ward, 2000). In addition to being expressed on certain epithelial and endothelial cells, FcRn is also expressed functionally on monocytes, macrophages, and dendritic cells (Zhu ).

Complement Receptors

In addition to the FcγR, the antibody-complexed flavivirus has been shown to be taken up by a macrophage cell line using the complement receptor-3 (Cardosa ). In the case studied—West Nile virus infection of mouse P388D1 macrophages—ADE was mediated by the presence of antiviral IgM and was inhibited with a CR3-blocking antibody. This mode of ADE was, however, found to be quantitatively less productive than the more commonly studied route of ADE, i.e., involving FcγR-mediated uptake route of IgG–virus complexes (Cardosa ).

Virus Binding Proteins Identified on Cells

The recent demonstration of DC-SIGN as a functional dengue virus receptor on human dendritic cells represents an important advance in the definitive identification of flavivirus receptors (Navarro_Sanchez 2003, Tassaneetrithep 2003). Several studies have identified cell surface proteins that bind flaviviruses, generally assayed by virus overlay blots of SDS–PAGE-resolved cell proteins (Table II ). Further work is required to confirm the involvement of these and other proteins as receptors in flavivirus infection.

TABLE II

Flavivirus Binding Proteins on Cells

Cell	Virus	Binding protein(s)	Reference
Human erythroleukemic K562 cells	Dengue-2	100 kDa	Rothwell et al. (1996)
Human and mouse neuroblastoma cells	Dengue-2	65 kDa	Ramos-Castaneda et al. (1997)
Human monocytic, B and T cell lines	Dengue-2	32, 45, 72 kDa	Bielefeldt-Ohmann 1998, Bielefeldt-Ohmann 2001
Monkey kidney Vero cells	Dengue-4	44, 74 kDa	Martinez-Barragan and del Angel (2001)
Mosquito C6⧸36 cells	Dengue-4	40, 45 kDa	Salas-Benito and del Angel (1997)
Mosquito C6⧸36 cells	Dengue-2	65, 80 kDa	Munoz et al. (1998)
Human hepatoma HuH-7 cells	Dengue-1	33- and 37-kDa proteoglycans	Hilgard and Stockert (2000)
Pig kidney PS cells	TBE	35 kDa	Kopecky et al. (1999)
Human dendritic cells	Dengue	DC-SIGN	Hilgard 2000, Navarro_Sanchez 2003, Navarro_Sanchez 2003
Vero cells; mouse neuroblastoma cells	West Nile	105 kDa	Kopecky 1999, Chu 2003

Cell Surface Macromolecules Modulated by Flavivirus Infection

A number of flaviviruses are able to stimulate the expression of cell surface molecules. Notable among these are adhesion molecules and major histocompatibility antigens. Multiple mechanisms appear to be involved, including virus- and cytokine-dependent pathways.

MHC Class I

Flavivirus infection of a number of cell types causes an increase in cell surface MHC class I expression (King 1988, King 1989, Libraty 2001, Liu 1989, Lobigs 1996, Shen 1995a, Shen 1997). Evidence for both virus-dependent (Lobigs ) and cytokine-dependent (Libraty 2001, Shen 1997) mechanisms has been reported. One process appears to be driven by the amount of flaviviral peptides generated by proteolysis and imported into the transporter associated with antigen processing (TAP), which results in increased cell surface expression of peptide-loaded MHC class I (Momburg ). The upregulation of MHC class I molecules by flaviviruses is perhaps reminiscent of that observed in infections by coronaviruses (Suzumura ) but stands in contrast to the virus-manipulated downregulation of MHC class I by viruses such as herpesviruses (Jennings 1985, Ploegh 1998), adenoviruses (Sparer and Gooding, 1998), poxviruses (Boshkov ), and HIV (Scheppler ). Although enhanced MHC class I expression would be expected to lead to greater cytotoxic T (Tc) cell–mediated cytolysis, it would render cells less susceptible to recognition by NK cells. Evidence has been presented that flavivirus-infected cells in fact show reduced susceptibility to NK cells at the cost of enhanced Tc cell–mediated lysis (Lobigs ). It has been suggested that such a response may permit flaviviruses to evade an early NK cell response and thereby allow for substantial amplification of virus during the viremic phase of infection (Momburg ). Nevertheless, evidence shows that NK cells are activated during dengue infection (Green ), and NK cell–mediated cytotoxicity has been reported to correlate with the severity of disease (Homchampa ).

Dendritic cells also undergo upregulation of MHC class I molecules following infection with dengue virus (Libraty ). Compared to other antigen-presenting cells, dendritic cells have superior T cell–stimulating activities (McKinney 1989, Timares 1998). Because antigen presentation via dendritic cell MHC class I can provoke exceptionally strong proliferation in CD8-bearing T cells (Bhardwaj 1994, Elbe 1994, McKinney 1989), much of the overall cytotoxic T cell response arising in flavivirus infection may be dictated at the level of the dendritic cell.

MHC Class II

West Nile virus infection induces MHC class II expression in mouse macrophages (Shen ), mouse astrocytes (Liu ), rat Schwann cells (Argall ), and human myoblasts (Bao ). Upregulation of dendritic cell MHC class II occurs in response to dengue (Libraty ) and West Nile (Johnston ) virus infection. Given the potent ability of dendritic cells to activate T cells (Banchereau ), the communication between dendritic cell MHC class II–peptide complexes and recognition molecules on CD4-expressing T cells should provide insights into some of the molecular processes underlying T cell activation.

Adhesion Molecules

Adhesion molecules are expressed on a variety of cells and mediate a spectrum of processes (Ley 2001, Roebuck 1999, Springer 1995). From the standpoint of flaviviruses, the most significant processes likely concern adhesion molecules on vascular endothelial cells, as these cells regulate permeability as well as transendothelial migration of leukocytes (Springer, 1995). Of particular importance are intercellular adhesion molecule 1 (ICAM-1; CD54), vascular cell adhesion molecule-1 (VCAM-1; CD106), and E-selectin (CD 62E), which are upregulated on the surface of the endothelium by inflammatory cytokines, cellular stress, and virus infection (Roebuck and Finnegan, 1999).

In the case of dengue, activation of endothelial cells occurs in vitro via TNF-α released from antibody-enhanced dengue virus infection of monocytes (Anderson ). Such activation involves upregulation of adhesion molecules E-selectin, ICAM-1, and VCAM-1. Evidence that similar activation processes occur in vivo comes from clinical studies showing elevated serum levels of TNF-α (Green 1999b, Hober 1993, Vitarana 1991, Yadav 1991) and soluble VCAM-1 (Murgue ) in dengue- and DHF⧸DSS-infected patients. Surprisingly, serum levels of soluble ICAM-1 were actually found to be lower than those of control subjects, although this may reflect plasma protein loss through leakage (Bethell ). Moreover, the function of soluble forms of ICAM-1 remains unclear, and their expression appeazrs to be regulated differently from that of membrane-bound ICAM-1 (Komatsu 1997, van Den Engel 2000).

Two phases of ICAM-1 upregulation have been noted in West Nile and Kunjin virus infection of human embryonic fibroblasts, namely an early (∼2 h postinfection) virus-dependent process and a later (∼24 h postinfection) event that is mediated by type 1 interferon (Shen ).

For neurotropic flaviviruses, such as West Nile virus in the mouse, the development of encephalitis has been correlated with viremia (Weiner ), suggesting virus penetration of the blood–brain barrier. The endothelium of the brain microvasculature normally represents a block between circulating virus and the central nervous system. Expression of endothelial cell adhesion molecules, thereby facilitating leukocyte adherence and diapedesis through the endothelium, may be an important mode of dissemination of virus-infected monocytes or other leukocytes into the brain. West Nile virus infection of human endothelial cells causes the upregulation of E-selectin, ICAM-1, and VCAM-1 (Shen ), which could mediate the transendothelial migration of leukocytes. Upregulation of these adhesion molecules was observed to occur early (2–4 h) in infection and appeared to be triggered by the virus rather than by cytokines (Shen ).

Further studies are required to clarify the role of endothelial cell adhesion molecule expression in the neuroinvasion of certain flaviviruses. Assuming such a role is confirmed, it will be incumbent to identify the mechanisms by which either free or cell-borne flaviviruses are stimulated to cross the vascular endothelial layer. For virus-infected leukocytes, such stimulation likely arises, at least in part, from chemokines produced by cells of the central nervous system. Astrocytes infected with JE virus have been reported to release chemokines (RANTES and MCP-1), which may play a role in the transendothelial migration of leukocytes (including those possibly carrying virus) across the blood–brain barrier (Chen ). Thus, once neural infection is initiated, the process could be amplified by the production of leukocyte-attracting chemokines at the site of infection.

Other Cell Surface Macromolecular Modifications Triggered by Flavivirus Infection

Complement Deposition

Complement activation is well documented in dengue disease (Nishioka 1974, Phanichyakarn 1977b, Russell 1969), with peak activation and the production of C3a and C5a occurring at the time of vascular leakage and⧸or shock (Malasit, 1987). Complement activation is likely to be largely mediated by immune complexes consisting of IgG and virus (Bokisch 1973a, Bokisch 1973b, Shaio 1992, Sobel 1975), although the low levels of circulating immune complexes detected in patients have stimulated thought as to other possible mechanisms (Malasit, 1987). Receptors for C3a and C5a are found on a wide variety of cells, including many human peripheral blood leukocytes (Chenoweth 1978, Fureder 1995, Kretzschmar 1993, Nilsson 1996, van Epps 1984). C5a receptors have been reported on endothelial cells, although at lower levels than myeloid cells (Zwirner ).

Although endothelial cells do not appear to be major targets for dengue virus in vivo (Halstead 1988, Halstead 1989, Sahaphong 1980), endothelial cells infected with dengue virus in vitro can become a substrate for deposition of C3dg and C5b-9, provided the dengue antibody is present (Avirutnan ). The presence of complement activation products on the endothelial cell surface could be a contributing factor to vascular permeability (Saadi ). Furthermore, anaphylotoxins and⧸or deposition of sublytic C5b-9 on the endothelial cell surface has the potential to activate the expression of adhesion molecules (Foreman ), cytokines (Saadi ), chemokines (Selvan ), cyclooxygenase-2 (Bustos ), tissue factor (Saadi and Platt 1995), heparan sulfate proteoglycan proteinases (Ihrcke and Platt, 1996), and even functional or morphological changes such as permeability loss and gap formation (Saadi et al., 1995).

Thus, in addition to being activated by leukocyte-derived cytokines (Anderson ), endothelial cells may also be coaxed toward a more permeability-enhancing state by virus infection and virus-mediated complement deposition. At present, the lack of evidence for in vivo infection of endothelial cells by dengue virus would suggest that the cytokine-mediated pathway is dominant. Figure 2 shows a model illustrating the potential role of endothelial cell perturbation by monocyte-derived cytokines and complement activation products in initiating vascular permeability and leukocyte extravasation in severe hemorrhagic flavivirus disease.

F

Model depicting possible events in endothelial cell surface perturbation during hemorrhagic flavivirus (dengue) infection. Endothelial cell activation, leading to upregulation of adhesion molecules (E-selectin, VCAM-1, ICAM-1), can be triggered by monocyte-derived cytokines (Anderson ) or by deposition of C5b-9 and other products of complement activation (Avirutnan ). C5b-9 is represented as a membrane attack complex pore structure, although the deposition of C5b-9 on dengue-infected cells appears associated with sublytic, rather than lytic, responses (Avirutnan ). Increased adhesion molecule expression, along with uncharacterized vasoactive factors, can lead to endothelial leakage and can mediate rolling, adhesion, and transendothelial migration of leukocytes into extravascular tissues. Similar processes may also contribute to the invasion of cell-borne neurotropic flaviviruses through the endothelial blood–brain barrier.

Conclusions

Much remains to be learned about the primary receptors for flaviviruses, though much knowledge has been gained about the initial interactions of flaviviruses with cell surface structures. The ability of flaviviruses to affect cell entry through heparan sulfate–type proteoglycans, as well as their dexterity to adjust mutationally to different receptors, depending on host cell type, illustrates the plasticity of the viral E protein to adapt to changing conditions and to ensure successful virus replication. Beyond this, certain flaviviruses, notably dengue virus, are masters at exploiting host antibody and Fc receptor–bearing cells to dramatically amplify viral replication. Flavivirus replication is coupled to altered cellular expression of cytokines, chemokines, and cell surface molecules, which shape the host response and immunopathogenesis associated with flavivirus infections. Ongoing and future characterization of the cell surface structures that mediate these events will be helpful in understanding the mechanisms of flavivirus-induced disease and in developing therapeutic and⧸or preventive strategies.