Decoding Arenavirus Pathogenesis: Essential Roles for Alpha-Dystroglycan-Virus Interactions and the Immune Response

Abstract

Pathogenesis following a virus infection results from interactions between the virus and its host. The outcome is determined by tipping the balance between virulence of the virus or susceptibility/resistance of the host to favor one or the other. This review focuses on two important members of the Old World arenavirus family: Lassa fever virus (LFV), a robust human pathogen that causes a severe acute hemorrhagic disease; and lymphocytic choriomeningitis virus (LCMV), also a human pathogen but better known in the context of its rodent model. Research with this model has uncovered and illuminated many of our current concepts in immunobiology and viral pathogenesis. Presented here are recent advances that form the framework for a better understanding of how viruses induce and maintain persistent infection as well as for the pathogenesis associated with acute LFV infection. A major component for understanding the pathogenesis of these arenaviruses revolves around study of the interaction of virus with its receptor, alpha-dystroglycan (α-DG).

Introduction

Viruses are studied for the diseases they cause, and also used as probes for investigating basic biologic mechanism(s). The arenavirus family contains a number of viruses like Lassa fever virus (LFV), Junin, Machupo, Guanarito, and Sabia -- important human pathogens that cause hemorrhagic fever and a high incidence of death. Also in this group is lymphocytic choriomeningitis virus (LCMV), which can cause human disease but is best known for its use in research that has successfully uncovered much of our current knowledge of viral pathogenesis, viral persistence and a variety of areas of immunobiology including: immunologic tolerance, major histocompatibility complex (MHC)-based recognition, CD8 and CD4 T cell activity, T cell exhaustion, T and B cell-mediated immune memory, immune-complex disease and virus-induced alteration of cellular differentiation (“luxury”) functions.

The arenaviruses were initially subdivided into two major groups based on serologic typing and phylogenetic evidence. This grouping into Old World arenaviruses (LFV and LCMV) and New World arenaviruses (Junin, Machupo, Guanarito, and Sabia) (Fig. 1A) represents a distinction that has held up to study by genome sequencing and monoclonal antibodies. The natural reservoirs of arenaviruses are selected rodent species, in which the viruses are most often maintained as an asymptomatic, persistent infection that is transmitted primarily by vertical/congenital routes. Spread to humans is by contact with excreta from infected rodents or with infected blood from patients. In the case of LCMV, transmission has also occurred via laboratory accidents involving infected cell culture material (i.e., LCMV infected cultures fail to exhibit evidence of cytopathology), contact with asymptomatic pets (e.g., hamsters) that are persistently infected with the virus, and by transplant-associated infection (in immunocompromised patients). Indeed, arenaviruses cause little or no toxicity to the cells they infect. The cell and tissue injury -- as well as resultant disease -- associated with infection are instead caused largely by activity of the immune system of the host, whose antiviral response produces factors that act against and damage virus-infected cells. An additional factor in pathology is the displacement of cellular molecules that are normally attached to cellular receptors by viral proteins; this can result in conformational changes that cause the cell membrane to become fragile as well as interference with normal signaling events.

Fig. 1.

Panel A, top: high-resolution electron microscopy images of the LCMV arenavirus. The left-hand image shows the virion in its entirety, with the GP1 spikes protruding from its surface. The central image shows several virions, and illustrates the arrangement of individual spikes around a central core; each core has five spikes 27 mm in length, that are generated from and equally spaced around its center. The right image shows ribosomes that are often incorporated into the virion; these play no role in virus replication, but reflect the polymorphism and sloppiness of arenavirus budding (Oldstone, 2002). Panel A, bottom: schematic diagrams show the four genes of the arenaviruses and their order. Arrows indicate the ambisense strategy of transcription. The L (long) RNA encodes the Z (similar to matrix protein of paramyxoviruses), and L (polymerase) proteins. The S (short) RNA encodes a common GP that is cleaved post-translationally to produce GP1 (spike GP), GP2 (transmembrane GP), and NP (nucleoprotein). Together NP and L form the transcriptional complex. Panel B: schematic illustration of the structure of DG and the molecules that bind to its α and β chains. Panel C: organ selection of variants (●: CTL⁻P⁺; o: CTL⁺P⁻) and their differential expression in (left panel) or absence from (right panel) the white pulp of the spleen when analyzed by using in situ viral nucleic acid hybridization (see Ahmed and Oldstone, 1988; Sevilla et al., 2000; Smelt et al., 2001). Panel D: difference in affinity of CTL⁻P⁺ Cl 13 virus versus CTL⁺P⁻ ARM 53b virus for purified α-DG. CTL⁻P⁺ Cl 13 binds at 2.5 log higher affinity than CTL⁺P⁻ ARM 53b. Panel E: selective amounts of α-DG on DCs (CD11c⁺) but not T (CD4, CD8) or B (B200) cells isolated from the spleen (see Kunz et al., 2001; Sevilla et al., 2000; Smelt et al., 2001). Panel F, top: schematic illustration of linear structure of the DG molecule, with its α and β subunits and the location of the binding site for LCMV, LFV and the ECM molecule laminin indicated. Panel F, bottom: schematic depiction of the actions of the glycosyltransferase LARGE as it binds to and modifies α-DG, and of the consequences of this modification for interactions between the cell, ECM components and virus.

Interestingly, although the arenaviruses naturally and preferentially exist in a persistent state within their natural rodent hosts, the first arenavirus was isolated from a human (CG) who died of infection during an outbreak of St. Louis encephalitis. Specifically, LCMV was isolated by Charles Armstrong from a monkey that had been used as a vehicle for passaging the infectious agent from this patient. The newly isolated virus passed through a Berkefeld N filter and, when injected intracerebrally (i.c.) into sentinel adult mice, caused rapid onset (within 6 to 9 days) of a central nervous system (CNS) disease that was characterized clinically by seizures and death (Amrstrong and Lillie, 1934). The histologic portrait was of lymphocytic infiltration in the leptomeninges and choroid plexus which led to the virus’ name lymphocytic choriomeningitis virus (LCMV). On the basis of the clinical incubation course, routes that caused infection, results of non-overlapping antibody neutralization assays, and histopathologic picture of CNS infection, LCMV was swiftly separated from the virus that was responsible for St. Louis encephalitis. Two years later, Eric Traub identified a persistent viral agent dwelling in white mice without causing them apparent harm (Traub, 1935). That virus, isolated by Traub through inoculation of other adult mice which did not previously carry the virus, mimicked Charles Armstrong’s isolate. Both the prototype neurotropic LCMV Armstrong virus (ARM 53b) and the prototype viral isolate from persistently infected mice (Cl 13) were cloned and sequenced four decades later (Salvato et al., 1988, 1989).

Arenavirus binding to a cellular receptor and its entry into the cell are initiated by the virus envelope glycoprotein (GP). This protein is first synthesized as a single polypeptide, the GP precursor (GPC), and then undergoes proteolytic processing by the cellular proprotein SKI-1/S1P (convertase subtilisin kexin isozyme-1/site-1 protease) resulting in two proteins: GP1 and GP2. GP1 is located on the GP spikes that surround the virus, and interacts with its receptor on the host cell’s plasma membrane (Fig. 1A). GP2 is the transmembrane protein that anchors GP1 to the virus surface. Virus binding to cellular receptors is the key determinant of the physiologic outcome of infection. Alpha-dystroglycan (α-DG) has been identified as the cellular receptor for the Old World arenaviruses (LFV, LCMV, Mopeia, Mobala, and presumably Lujo virus), as well as for the New World Clade C arenaviruses Latino and Oliveros (Fig. 1B) (Cao et al., 1998). The receptor for the New World Clade A and B arenavirus is transferin receptor-1 (Radoshitzky et al., 2007). In this review, we will discuss the mechanisms of infection and pathogenesis of the two best studies of the Old World arenaviruses, LFV and LCMV.

i. In vivo selection of LCMV variants and molecular and biologic mapping: a single amino acid change in viral GP1 and another in the viral polymerase determine whether the virus causes an acute self-limiting infection or a persistent infection

When LCMV infects its natural murine host, virus variants are generated via organ-specific selection (Fig. 1C) (Ahmed and Oldstone, 1988). Variants selected in the CNS are markedly dissimilar to those selected in lymphoid tissues and cells (spleen, lymph nodes, dendritic cells [DCs], CD4 and CD8 T cells) at both the biological and chemical levels. Inoculation with ARM 53b -- the virus cloned from the original Armstrong isolate after multiple mouse-brain passages -- into adult immunocompetent mice subcutaneously (s.c.), intraperitoneally (i.p.), or intravascularly (i.v.) leads to rapid emergence of virus-specific MHC-restricted T cell responses in which CD8 T cells attack and remove the virus-infected cells that serve as “factories” that produce infectious virus. This results in rapid purging of the virus and clearance of the infection. The result is quite different when ARM 53b is administered intracerebrally (i.c.). Again, virus-specific CD8 T cells are generated, as i.c. inoculation breaks through the blood-brain-barrier and the virus is thus again introduced peripherally. At the same time, however, brain cells -- in particular those of the leptomeningeal and choroid plexus cells -- become infected. Ultimately, the virus-specific CD8 T cells migrate to the CNS and interact directly with virus-infected cells (in the leptomeninges, choroid plexus, and endothelium), causing a lethal damage. Specifically, they release cytokines and chemokines that attract polymorphonuclear and mononuclear myeloid cells. The latter cell types can destroy cells and in addition they release cytokines/chemokines, which leads to an increase in intracerebral pressure that kills the host (Kim et al., 2009). In the case of s.c., i.p., and i.v. inoculation of ARM 53b, CNS injury and host death are avoided because the cytotoxic T lymphocytes (CTLs) are generated before the virus can infect CNS cells.

Organ-specific virus variants were generated by injecting cloned ARM 53b into newborn mice less than 24 hours old. This resulted in a life-long persistent infection. Virus variants isolated from the CNS of adult mice, in which the virus persistently infects only neurons, were similar to the wild-type ARM strain biologically. That is, these variants (CTL⁺P⁻ viruses) induced a potent and specific anti-LCMV CTL response in adult mice, and the response cleared this infection within two weeks (Fig. 1C). In contrast, isolates cloned from lymphoid tissues and cells, failed to generate a CTL response that was sufficient to clear the virus when injected i.v. into adult immunocompetent mice, and a persistent infection followed (CTL⁻P⁺ viruses) (Fig. 1C). Viral clones isolated from the CNS of mice originally infected with CTL⁻P⁺ virus generated revertants with characteristics of the original ARM 53b virus (CTL⁺P⁻ virus). Thus, tissue and cell-specific selection is important in the evolution of these viruses, and suggests a mechanism to account for the emergence of viral variants in nature.

An analysis of the sequences of over 50 of the lymphoid and CNS variants revealed that CTL⁻P⁺ lymphoid variants commonly encode an aliphatic small amino acid (predominantly leucine, but occasionally isoleucine or valine), at position 260 of the GP1 protein. In contrast, CTL⁺P⁻ variants have a bulky aromatic phenylalanine at this position of GP1. In addition, sequencing of the ARM 53b CTL⁺P⁻ and Cl 13 CTL⁻P⁺ variants revealed a difference in the polymerase gene (L) at residue 1079 with ARM 53b encoding a lysine and Cl 13 a glutamine (Salvato et al., 1991). Although variants other than Cl 13 and ARM 53b were found to encode alternative amino acids at various positions of the GP1 or polymerase protein (Salvato et al., 1989, 1991; Sevilla et al., 2000), these variants consistently featured the residue at GP1 position 260 that are consistent with their CTL response and permissiveness status (i.e., positions GP1 260/L 1079: leucine/glutamine in CTL⁻P⁺; phenylalanine/lysine in CTL⁺P⁻). Multiple studies, including reverse genetics-based screens (Emonet, Sullivan, de la Torre, and Oldstone, unpublished data), have shown that GP1 residue 260 is required for binding to the α-DG receptor on DCs in order to initiate virus-induced immunosuppression. Our knowledge of how residue 1079 causes persistent infection associated with CTL⁻P⁺ remains more limited; given that the field still lacks a polymerase assay, it has not been possible to map the domains and functions of the LCMV polymerase protein.

Additional studies have indicated that although the CTL⁻P⁺ variants and CTL⁺P⁻ variants replicate to yield similar titers in the spleen, (10⁻⁶ to 10⁻⁷ PFU at day 3 to 5 following infection with 1 × 10⁵ PFU i.p.), replication occurred at different sites within this organ (Sevilla et al., 2000; Smelt et al., 2001). In situ hybridization of viral nucleic acid, as well as fluorescence immunochemical microscopy for viral antigens carried out at early time points after infection (days 1–3), revealed that the CTL⁻P⁺ variants localized to the white pulp of the spleen whereas the CTL⁺P⁻ variants localized primarily to the red pulp and only minimally to the white pulp (Fig. 1C). Virus overlay assays were then carried out. To this end, cultures containing cells or tissues permissive to virus infection were prepared and subjected to dounce homogenization, and the proteins were separated by polyacrylamide gel electrophoresis and transferred to nylon membranes. These blots were incubated with either labeled of unlabeled LCMV (Cl 13), and then with a monoclonal antibody to LCMV GP. A broad band centered at roughly 116 kD was revealed (Fig. 1D). A range of enzymatic tests that followed identified the material as a glycoprotein that did not contain lipid. With this information in hand and a marker to identify the putative glycoprotein receptor, analysis involving a series of chromatographic steps using ion-exchange and lectin columns resulted in purification of a single band that peptide profiling and mass spectrometry revealed to be alpha-dystroglycan (α-DG) (Fig. 1B). Subsequent experiments formally established α-DG as the receptor for LCMV, as well as for other Old World and the New World Clade C arenaviruses. The designation of α-DG as the receptor for these viruses was based on three stringent criteria: 1) binding of the arenaviruses to purified α-DG; 2) impaired binding between the arenavirus and α-DG in the presence of soluble α-DG; and, most importantly, 3) the conversion of cells lacking the α-DG gene from a permissive to a non-permissiveness state with respect to infection, and restoration of the permissive state by reintroduction of the α-DG gene (by either transfection or adenovirus-mediated infection) (Cao et al., 1998; Kunz et al., 2001; Smelt et al., 2001; Spiropoulou et al., 2002).

ii. Molecular characterization of the Old World arenavirus receptor, alpha-dystroglycan (α-DG), and how it dictates arenavirus pathogenesis

Dystroglycan (DG) is a cell surface molecule that links proteins of the extra-cellular matrix (ECM) -- laminin, agrin, perlecan, and neurexin -- with the internal actin-cytoskeletal machinery (Barresi and Campbell, 2006; Ervasti and Campbell, 1991; Kunz, 2009) (Fig. 1B). DG, which is encoded by a single gene, is post-translationally processed to form an α subunit (α-DG) of 653 amino acids and a β subunit (β-DG) of 242 amino acids. α-DG is non-covalently bound to β-DG, and β-DG binds intracellularly to the cytoskeleton adapter protein dystrophin, which links the DG complex to the actin-cytoskeleton network. α-DG contains the binding site for Old World arenaviruses, and the β subunit is not essential for the binding or entry of virus, because entry occurs in the presence of either a deletion mutant in which the β-DG cytoplasmic domain is absent or that of a C-terminal fusion between α-DG and the transmembrane domain of the PDGF receptor (Kunz et al., 2003). The above-described studies indicate that virus infection occurs regardless of whether the actin-based cytoskeleton is intact. In several cell types, DG-associated proteins -- including dystrophin, utrophin, sarcospans and α, β, γ and δ, and the sacroglycans -- appear to be crucial for the stability and cellular trafficking of the DG complex (Fig. 1B), since mutations in the encoding genes lead to loss of the DG receptor and membrane stability, and to the development of neuromuscular diseases (Cohn, 2005; Kanagawa and Toda, 2006).

α-DG consists of N- and C-terminal globular protein heads that are separated by a mucin domain which can itself be divided into functional N- and C-terminal regions (Fig. 1F). The binding site for LCMV, LFV and the ECM molecule laminin has been mapped to an 18 amino acid domain spanning residues 316–334, although it remains unclear whether identical amino acids within this site are responsible for the interactions with the two viruses and with laminin. Interestingly, binding to Ca⁺⁺, Mg⁺⁺, and heparin is necessary for binding to laminin but not for binding to virus (Kunz et al., 2001). However, virus binding to α-DG requires that it displaces the ECM molecule. Competition assays have revealed that CTL⁻P⁺ viruses out-compete the ECM molecule laminin for binding by a factor of 4- to 6-fold; CTL⁺P⁻ viruses, in contrast, consistently fail to displace laminin (Kunz et al., 2001). Further, when purified α-DG is placed on a membrane, CTL⁻P⁺ viruses bind α-DG with an affinity 2 to 2.5 logs higher than that of CTL⁺P⁻ viruses (Fig. 1D) (Kunz et al., 2001; Sevilla et al., 2000).

These observations describe two issues that are important for understanding arenavirus-associated pathogenesis. First, those cells in the body that express high levels of α-DG are most likely to bear the burden and direct consequences of LCMV/LFV tropism and infection. Second, since binding of α-DG to ECM molecules like laminin stabilizes the cell’s membrane, interference with this interaction by high affinity-binding arenaviruses may destabilize the membrane. Furthermore, this displacement may result in a loss of appropriate signals and thus significant disruption of cell function. As discussed below, both viral tropism and ECM displacement play critical roles in arenavirus pathogenesis.

The glycosyltransferase enzyme LARGE is an essential player in arenavirus infection and pathogenesis (Kanagawa et al., 2004; Kunz et al., 2005a, 2005b). In order to bind either ECM molecules or arenaviruses, α-DG must be post-translationally modified by LARGE. This post-translational modification occurs within the N-terminal region of the mucin domain (Fig. 1F). Additionally, the N-terminal domain of α-DG interacts with LARGE. This two-fold interaction with LARGE is believed to cause a change in the conformation of α-DG and to lead to the modification of a phosphorylated O-mannosyl glycan to the mucin-like domain, thereby creating the binding site for arenaviruses and ECM proteins (Yoshida-Moriguchi et al., 2010). Interestingly, a recent genome survey in West Africa, where LFV is endemic, revealed that polymorphisms in two genes that have been functionally associated with the α-DG/DG receptor complex are positively selected (Sabeti et al., 2007). The strongest signal was in a 400 kb region that lies entirely within the LARGE gene. The second strongest signal was in dystrophin, the adapter protein that links the C-terminus of β-DG to the actin cytoskeleton network. The hypothesis currently under investigation is that those individuals with a polymorphism in LARGE and/or dystrophin are likely to be less susceptible to severe LFV infection, because the virus is less effective at binding to α-DG, and thus also less effective at entering the cells. Consequently, virus replication is less efficient, and the immune system has a competitive advantage in controlling virus infection.

Analysis of cells of the immune system indicates that dendritic cells DCs primarily express the cellular receptor α-DG (Fig. 1E). Biochemical study of spleens revealed that over 99% of α-DG was associated with DCs, and less than 1% was associated with CD4 or CD8 T cells or with B cells. By selection, virus strains and variants that bind α-DG with high affinity are associated with virus replication in the white pulp of the spleen, with preferential replication in a majority of DCs. The consequence of mature DC infection is two-fold (Fig. 2). First, MHC class I and II proteins, as well as the B7.1 and B7.2 costimulatory molecules are markedly down-regulated (Fig. 2A) (Sevilla et al., 2000). Second, there is a decrease in trafficking of MHC molecules to the DC surface (Fig. 2D). The outcome is a failure of DCs to present viral antigens, and thus their failure to fully arm and expand immune-specific T and B cells (Sevilla et al., 2000). As a result there is a global suppression of anti-LCMV T cell as well as CTL responses to other viruses including VSV and HSV, and suppression of antibody responses to soluble or particulate antigens. In addition, the virus induces infected DCs to produce negative regulators of the immune system (Fig. 2B,C) (Brooks et al., 2006, 2008a, 2008b; Ejrnaes et al., 2006). Several of these factors will be discussed below. The bottom line is that the receptor (α-DG)-virus interaction on DCs in vivo is an early and essential step in the commencement of virus-induced immunosuppression and viral persistence.

Fig. 2.

Panel A: schematic illustration of the ability of CTL⁻P⁺ virus Cl 13 to push aside ECM molecules from α-DG so that the virus can bind to its receptor site, preferentially infect DCs and, directed by its polymerase gene, replicate in DCs. Consequences of virus replication are: 1) the production of negative regulators of the immune response, such as IL-10; 2) the down-regulation of MHC class I and II receptors (shown in Panel D for individual DCs) and their costimulatory molecules B7.1 and B7.2; 3) a reduction in the development, migration, and expansion of new DCs due to the infection of precursor cells; and 4) a global suppression of the immune response. Also indicated is the inability of ARM 53b to displace the ECM molecule laminin. In the Cl 13 virus, the capacity to bind and enter DCs maps to the leucine residue at position 260 of the GP1 protein, whereas the enhanced replication of virus, once in the DC, is associated with glutamine at position 1079 of the polymerase gene. Panel B shows that the negative immune regulator IL-10 is produced primarily by Cl 13-infected DCs and not during ARM 53b infection. Panel C illustrates that IL-10 abrogates the function of CD4 and CD8 T cells, a process that is reversed when antibodies to the IL-10 receptor (rec) are applied (see Brooks et al., 2006, 2008a, 2008b). Panel E: table summarizing restoration of anti-LCMV T cell function in the context of therapy with antibody to IL-10rec (R) (based on data in Panel C); this works by purging the virus and terminating the infection in vivo.

Viruses have additional mechanisms for dismantling the host’s DC-T cell network, for example, interference with DC maturation and migration (Sevilla et al., 2000, 2004). This phenomenon was first observed when LCMV Cl 13-infected hematopoietic progenitor cells were found to be impaired in their development into DCs, both in vivo and in vitro. This is a consequence of their production of IFN-α and IFN-β (Sevilla et al., 2004) as DC progenitor cells in mice deficient for the receptor for IFNα/β still undergo differentiation into DCs even in the context of virus infection. Other experiments in which administration of the growth factor FLT-3 ligand resulted in dramatic enhancement of the number of mature DCs. When either uninfected or LCMV ARM-infected mice were exposed to FLT-3 ligand for 10 days, the number of common (c) DCs in the spleen was increased over 25-fold. In contrast, when FLT-3 ligand was administered during Cl 13 infection, production of cDCs was not significantly enhanced (< 2-fold increase) (Sevilla et al., 2004).

In toto, the accumulated data indicates that Cl 13 (CTL⁻P⁺) virus uses a 4-pronged strategy to ablate DC function and thereby establish viral persistence. First, in the cases of both LCMV ARM and Cl 13 infections, a small but active set of LCMV-specific CD8⁺ CTLs is generated early (day 4, 5). These CTLs destroy, at this time, the DCs that present viral antigen on their surfaces; removal of CD8 T cells with a specific monoclonal antibody against CD8 T cells is sufficient to prevent the development of persistent infection (Borrow et al., 1995). Concurrently, Cl 13 can outcompete the ECM protein laminin for binding to α-DG receptors on DCs, and then infects many of the remaining DCs (Kunz et al., 2001; Sevilla et al., 2000). Once virus is internalized into and replicating within DCs, selective distortion of the cell’s transcriptional and translational machinery ensues, resulting in the down-regulation of molecules that are required to present the antigen (e.g., MHC class I and II complexes, costimulatory molecules) and to thereby arm and expand T cell populations that could eliminate the viral infection (Kunz et al., 2001; Sevilla et al., 2000; Smelt et al., 2001). Third, the infected DCs are induced by the virus to produce immune response suppressors such as IL-10 which might also act to down-regulate class I MHC (Brooks et al., 2006, 2008a, 2008b; Ejrnaes et al., 2006). Such molecules play a commanding role in turning off the function of (exhausting) most of the antiviral T cells that remain in the periphery (Fig. 2C). Fourth, the virus prevents infected progenitor cells from maturing into DCs and migrating to the spleen thereby preventing the host from producing new DCs to reinforce the original pool (Sevilla et al., 2004). Precisely at which stage of differentiation DC progenitor cell(s) are infected is currently unknown, but this question is being actively investigated in our laboratory (Ng, Nussenzweig, and Oldstone, unpublished observations).

Recently others have suggested that CTL⁻P⁺ Cl 13 virus preferentially, and most effectively, replicates in DC stromal cells:fibroblastic reticular cells (FRC) (Mueller et al., 2007). These cells are suspected to express α-DG on their surfaces. FRCs have also been reported to interact with the PD-1, a negative regulator of the immune response (Mueller et al., 2007). The precise molecular events or molecules engaged in the suspected immunosuppression are unknown, and given that these FRCs comprise 0.4% or less of the total cell number in the spleen, they are difficult to investigate. Nevertheless, in the interest of clarifying the roles and functions of FRCs, we have recently isolated and immortalized pure populations of these cells (Ng, Nyak, and Oldstone, unpublished data).

Over the last few years, attention has focused on the molecules that are induced during persistent virus infection and negatively regulate (suppress) the host’s immune response. The two that have been best characterized, IL-10 (Brooks et al., 2006, 2008a, 2008b; Ejrnaes et al., 2006) and PD-1 (Barber et al., 2006; Blackburn et al., 2010; Ha et al., 2008; Mueller et al., 2007), independently abrogate antiviral CD8 and CD4 T cell function. Both molecules were found originally during LCMV Cl 13 infections, and have now also been identified in humans who suffer persistent infection with HIV, HBV and HCV (Brockman et al., 2009; Kaufmann and Walker, 2008; Martin-Blondel et al., 2009; Rutebemberwa et al., 2008). Recent experiments have revealed the importance of the activities of IL-10 and PD-1 molecules with respect to the course of virus infection. In fact, in the LCMV model, the application of antibodies against the IL-10 receptor (Brooks et al., 2006, 2008a, 2008b; Ejrnaes et al., 2006) and PD-L1 (Barber et al., 2006; Blackburn et al., 2010; Ha et al., 2008) to unresponsive “exhausted” T cells led to an in vivo rescue of T cell function that was sufficient to purge the persisting virus and control the viral infection (Fig. 2E). Recovery of exhausted T cells in vitro has also been achieved with antibodies against the IL-10 receptor and PD-1 in persistent infections in humans (Brockman et al., 2009; Kaufmann and Walker, 2008; Martin-Blondel et al., 2009; Rutebemberwa et al., 2008), and in vivo with an antibody against PD-1 in the case of SIV infection (Velu et al., 2009). In addition, depletion and reconstitution studies performed in vivo have established that TGFβ (Tinoco et al., 2009), IL-21 (Elsaesser et al., 2009), and LAG3 (Blackburn et al., 2009) are important for suppressing the immune response during persistent infection. Other suppressor molecules, namely, CD160, 2B4, CTLA-4, PIR-B, GP49, KLRG1, NKG2A, NKG2D, and BTLA were identified as inhibitory receptors, primarily on exhausted CD8 T cells (Blackburn et al., 2009). All these molecules, as well as IL-10, PD-1, IL-21, TGFβ, and LAG3, were found to be elevated -- by gene chip analysis of gene expression in Cl 13-infected DCs, CD4 T cells or CD8 T cells, during the course of infection (at an initiation or during maintained persistent infection) -- when compared to molecules harvested at similar time points from similar but uninfected or ARM 53b infected populations of immune cells.

In addition to the biased gene chip approach, in the case of the LCMV Cl 13 model a non-biased forward genetic approach has been used to identify novel negative regulators of the antiviral immune response. This method uses ethylnitrosourea (ENU) to mutagenize mice that are then challenged with 1 × 10⁶ PFU of LCMV Cl 13 i.v. At 15 or 30 days post-infection, non-mutanized normal mice will have generated 4 to 6 logs of virus in their sera. In the genetic screen, mice that have been ENU-mutagenized and infected with virus are examined for limited or undetectable virus at those predetermined time points, indicating that genes encoding negative regulators of the antiviral immune response have been knocked out. Thereafter, breeding studies are performed to ensure that loss of the suppressor factor is genetically transmitted, i.e., that the mutation is in the germline. Once clearance of the virus is established as genetically caused, deep sequencing or positional cloning is carried out. To date, we have inoculated over 2,000 ENU mice and obtained 12 hits that are currently under investigation (Popkin, Beutler, and Oldstone, unpublished results). Based on these data, we calculate that roughly 100 such negative regulator molecules exist in the mouse and likely human genomes.

iii. Probing for mechanism(s) responsible for LFV pathogenesis and for its clinical manifestations

LFV is endemic to West Africa, particularly Guinea, Liberia, Sierra Leone, and Nigeria (Ogbu et al., 2007). An estimated 300,000 residents in these localities incur infection each year, and approximately 20,000 to 30,000 deaths result from this (Fichet-Calvet and Rogers, 2009; Ogbu et al., 2007). For about 80–90% of those infected with LFV, the disease is mild, and recovery is associated with an anti-LFV immune response, although approximately 20% of these survivors suffer from permanent hearing loss (Cummins et al., 1990). This hearing deficiency is believed to result from a sensory neuropathy, and is the most common cause of hearing defects in West Africa. In addition, approximately 20% of those infected with LFV suffer a severe disorder affecting multiple organ systems and, during epidemics, this can cause a 50% or higher level of fatality. Of patients hospitalized for LFV infection, 15 to 20% die from the illness. As a clinical marker, those having serum titers of 8.5 log 10 PFU/ml of LFV or higher usually succumb to the infection. The incubation period from contact with the virus to the onset of illness can be as short as five or as long as 21 days. In addition to fever, malaise, progressively worsening sore throat, abdominal pain, and diarrhea, patients’ suffer from leakage of blood from small blood vessels (e.g., from capillaries due to needle punctures made during hospital care) and from internal bleeding. As bleeding worsens patients become delirious or confused, and some convulse before dying. Thus, as shown in Figure 3, three cardinal clinical manifestations are in need of physiologic or molecular explanation (Fig. 3A). In those individuals who are severely ill (many of whom succumb to this disease) the first sign is a high virus titer (>8.5 logs PFU/ml of serum) in the absence of an effective immune response. Second, vascular leakage occurs and leads to shock. In recovering patients, blood levels of LFV during acute infection are lower than 8.5 logs PFU/ml, and third, with recovery comes a protective immune response, but often, also the neurosensory hearing loss (Cummins et al., 1990).

Fig. 3.

Experimental evidence supporting an important role for α-DG in LFV pathogenesis. Panel A: correlation of serum titer (>8.5 logs PFU/ml) of LFV with vascular leakage, shock and high mortality over the course of infection. Based on similarities to the LCMV Cl 13 model (Fig. 1 and 2), LFV replication and the absence of an effective anti-LFV immune response are believed the result of the ability of LFV to displace the ECM molecule laminin, bind to α-DG and infect large numbers of DCs. α-DG is concentrated on antigen-presenting cells, and serves as a receptor for LFV binding and entry (see Cao et al., 1998; Spiropoulou et al., 2002). The consequences of LFV replication in antigen-presenting cells likely lead to a dysfunction of their activity and the generation of suppressive factors that abort the host’s immune response. Panels B and C: displacement of the ECM molecule laminin leads to enhanced membrane fragility and instability (Panel B) (see Han et al., 2009, for details) and cause a loss of signals that are required for the formation and maintenance of the myelin sheath (Panel C) (see Rambukkana et al., 2003, for details). The effects noted in Panels B and C have been documented with inactivated non-replicating virus.

Much of our understanding of the pathogenesis of LFV is theoretical and draws on lessons we have learned from LCMV Cl 13-like virus infection based on the fact that both LFV and the LCMV CTL⁻P⁺ viruses utilize α-DG as a receptor for attachment to and entry into cells (Cao et al., 1998; Spiropoulou et al., 2002). DCs are the major components of the immune system that initiate the innate and adoptive antiviral immune responses, and they are the essential point of contact for the virus. Further, among the various cell populations that constitute the immune response, DCs express the greatest amount of α-DG on their surfaces (Fig. 1E) (Sevilla et al., 2000). Specifically, more than 99% of all α-DG found in the immune system is present on DCs which are the major antigen-presenting cells in the body. These cells are responsible for the processing and expression of viral peptides with MHC molecules in order to present viral antigens and, in association with costimulatory molecules for arming and expanding the virus-specific immune response. As has been carefully worked out and established in the case of LCMV Cl 13 (see ii. above), we know that those viral strains that bind α-DG with the highest affinity and displace ECM molecules like laminin preferentially infect DCs and alter their ability to initiate an effective immune response (Oldstone, 2002; Sevilla et al., 2000; Smelt et al., 2001). The consequence is suppression of both the innate and adaptive antiviral immune responses, which leads to heightened virus replication. A similar scenario is postulated, although not yet proven to occur, in LFV-infected individuals who fail to develop an effective antiviral immune response. This failure, coupled with an elevation in viral titer as a consequence of unchecked replication, results in the host being overwhelmed by the infection. Those LFV-infected individuals who mount an immune response most often survive the infection. The success or failure to mount an effective immune response that clears the LFV infection is a likely attributable to positive genetic selection (Sabeti et al., 2007). Recent biochemical analysis of α-DG revealed that the cellular glycosyltransferase LARGE is essential for adding a sugar unit to α-DG, thereby changing its conformation to promote binding to LFV (Kunz et al., 2005a). This information reinforces the above-mentioned observations from the genomic survey in West Africa (Sabeti et al., 2007) that identified positively selected polymorphisms in LARGE and dystrophin (the protein that links the β-DG to the actin-cytoskeleton network), a status associated with migration and stability of the α-DG complex. Thus, if LFV binding to α-DG is decreased and causes diminished entry of virus into DCs as a result of defects in LARGE or dystrophin, the result may be a selective advantage to the host. Then, as LFV replicates in DCs, the antiviral immune response that is mounted has a better chance of controlling the virus infection.

Although tropism and enhanced or limited binding and entry of LFV into antigen-presenting cells is postulated to exert control over the vigor of an antiviral adoptive immune response and susceptibility in one person or resistance in another, it is tempting to postulate that the associated vascular leakage and neurosensory loss may result as consequences of a distinct but different mechanism. That sequence may be one in which viruses displaces the ECM and causes membrane instability (Han et al., 2009) (Fig. 3B), perhaps by disrupting GAP junctions on endothelial cells, and/or by disturbing signaling emanating from the ECM. Vascular leakage is associated with LFV replication in the liver, an organ that produces many blood-clotting proteins whose production is diminished during infection with LFV. Additionally, we have observed that binding of inactivated virus (non-replicative) to membranes displaces ECM laminin and leads to membrane instability and leakage, as shown in Fig. 3B. This is similar to a mechanism in which binding of inactive LFV to the cell surface of Schwann cells disrupts myelin formation, leading to demyelination (Fig. 3C) (Rambukkana et al., 2003). In this paradigm, Schwann cells are cocultured with rat axonal cells. Under normal circumstances, ECM laminin signals lead via the Schwann cells to the co-localization and deposition of myelin and laminin around axons. However, when inactive LFV is bound to Schwann cells, myelin formation is aborted, as judged by both electron microscopy and immunohistochemical staining for myelin basic protein (Fig. 3B). Similarly, laminin does not form along myelin sheaths, but rather forms aggregates. Like LCMV, LFV is primarily a non-lytic virus and does not destroy cells. Cell or tissue damage is instead caused by immunopathologic mechanisms, by the distortion of ECM signals, or by virus replication in differentiated cells (such as DCs) and selectively interference with their homeostatic function.

Conclusions

Arenaviruses per se cause little direct cell or tissue injury. Injury accompanying arenavirus infection depends primarily on two factors. The first is production of the host antiviral immune response which is meant to recognize, attack and remove infected cells that essentially serve as virus-producing factories. The cell and tissue injury that can result from this process cause the clinical manifestations of infection during viral clearance. The pathologic outcome of tissue destruction occurs when the immune response is too vigorous or takes place in a vulnerable area such as the CNS. DCs are the major antigen-presenting cells and arm and expand the T and B cell populations needed to attack virus-infected cells and to clear the infection. Viruses use a multipronged strategy to disrupt DC function, thereby quieting the antiviral immune response and promoting their own long-term survival. This review documents how α-DG, the receptor for LCMV and LFV, contributes to the pathogenesis of these viral infections. Among the cells of the immune system, DCs express the highest levels of α-DG and are the sentinel cells that LCMV, and presumably also LFV, infect. The resultant infection of DCs compromises DC function. When persistent infection arises, the consequences for the host are alterations to the (“luxury”) functions of specialized cells, including DCs, which lead to disordered homeostasis and disease (reviewed in Oldstone, 2002). The second determinant of injury is the displacement of laminin or other ECM molecules that bind to the same site on α-DG that LCMV and LFV seek. In pushing aside ECM molecules, the virus destabilizes membranes and causes interference with ECM signals that are required to maintain homeostasis. The end result is disease manifestations that typify acute infection. The next level of understanding of the molecular pathogenesis of these viruses will require the identification of the normal functions of such ECM signals.