|Home | About | Journals | Submit | Contact Us | Français|
Dengue virus (DENV) causes disease ranging from dengue fever (DF), a self-limited febrile illness, to the potentially lethal dengue hemorrhagic fever and dengue shock syndrome (DHF/DSS). Epidemiological studies have suggested that DHF/DSS usually occurs in patients who, prior to infection, have acquired DENV-reactive antibodies, either from a previous infection with a heterologous DENV serotype or, in the case of infants, passively from an immune mother. Therefore, it has been hypothesized that subneutralizing levels of DENV-specific antibodies exacerbate disease, a phenomenon termed antibody-dependent enhancement (ADE). To date, the mechanism of ADE and its contribution to pathology remain elusive. Here, we demonstrate that administration of DENV-specific antibodies to DENV-infected mice can sufficiently increase severity of disease so that a mostly non-lethal illness becomes a fatal disease resembling human DHF/DSS. Antibodies promote massive infection of liver sinusoidal endothelial cells (LSECs), which results in increased systemic levels of virus. Our findings demonstrate that a subprotective humoral response may, under some circumstances, have pathological consequences.
The four serotypes of dengue virus (DENV1-4), a flavivirus transmitted to humans by Aedes mosquitoes, induce a spectrum of disease ranging from dengue fever (DF), an acute, self-limited febrile illness, to the life-threatening dengue hemorrhagic fever and dengue shock syndrome (DHF/DSS), characterized by plasma leakage, low platelet counts, liver damage, elevated cytokine levels (“cytokine storm”) and, in the most severe cases, death due to shock (Halstead, 2007). Two and a half billion people in tropical and subtropical regions are at risk of infection, and it is estimated that 50–100 million cases occur annually, of which 500,000 are severe and 15,000 are fatal (Rico-Hesse, 2007). Epidemiological studies suggest that DHF/DSS occurs predominantly during either secondary infection with a heterologous serotype (Halstead et al., 1967) or primary infection in 6- to 9-month-old infants of DENV-immune mothers (Halstead, 1982). Therefore, it appears that nearly all severe dengue cases occur in patients who have acquired DENV-reactive antibody prior to infection, either actively from a previous infection, or passively from an immune mother. Accordingly, it has been hypothesized that subneutralizing levels of DENV-specific antibodies exacerbate disease by increasing infection of cells bearing Fc-γ receptors (FcγRs), a phenomenon termed antibody-dependent enhancement of infection (ADE) (Halstead, 2003). However, little is known about the mechanism of ADE in vivo and its contribution to pathology, as increased disease severity due to antibodies has never been demonstrated in vivo.
Here, we demonstrate that administering anti-DENV antibodies to DENV-infected mice can sufficiently exacerbate disease so that a mostly non-lethal illness turns into a fatal disease resembling human DHF/DSS. Massively enhanced infection of liver sinusoidal endothelial cells (LSECs) occurred in mice treated with DENV-specific antibodies. Following enhanced infection of LSECs, mice exhibited increased levels of virus in other tissues, cytokine storm, low platelet counts, increased vascular permeability, intestinal hemorrhage and ultimately death.
Epidemiological observations and in vitro studies have provided the majority of the evidence for the occurrence of ADE. In vitro, subneutralizing concentrations of DENV-specific antibodies enhanced viral replication in mononuclear phagocytes (Halstead et al., 1977). In vivo, passive transfer of immune serum (Halstead, 1979) or antibodies (Goncalvez et al., 2007) into DENV-infected rhesus monkeys resulted in increased viremia, although signs of disease were not apparent. However, to date, it has not been determined whether antibodies alone can exacerbate DENV-induced disease.
To investigate the effect of DENV-specific antibodies in vivo, type I and II interferon (IFN) receptor-deficient 129/Sv mice (AG129) were administered 200 μl of DENV1-, 2-, 3- or 4-immune serum (obtained from AG129 mice infected with 1.5×106 FACS infectious units (FIU) of DENV) or naïve serum one day before and one day after i.v. infection with 5×108 genomic equivalents (GE) (approximately 104 PFU) of the DENV2 strain S221, a triple-plaque-purified clone isolated from a mouse-passaged DENV2 strain (Shresta et al., 2006). Mice administered DENV1-, 3- or 4-immune serum died from severe disease 4 to 5 days after infection, without manifesting neurologic abnormalities, whereas mice treated with naïve serum were sacrificed when they developed signs of paralysis 7 to 12 days after infection (fig. 1A). Half of the mice administered DENV2-immune serum were protected from disease and survived much longer than control mice (33% survived over 30 days), whereas the other half exhibited the early, severe disease phenotype. All immune sera contained DENV2-reactive antibodies (fig. 1B), although the DENV2-immune serum contained about 10 times more than the others. The ability of the sera to reduce infection of C6/36 mosquito cells by the DENV2 strain S221 was measured (fig. 1C), and the DENV2-immune serum neutralized virus 15 to 20 times better than the DENV1-, 3- or 4-immune sera. Naïve serum at low dilutions was able to prevent infection of C6/36 cells; however, this effect is likely independent of antibodies, as similar observations were made with serum obtained from μMt mice, which lack antibodies (data not shown). Notably, DENV2-immune serum obtained from AG129 mice infected with a higher dose of virus (4.0×107 FIU, instead of 1.5×106 FIU as in fig. 1), completely protected against disease in recipient mice infected with 5×108 GE of S221 i.v. (data not shown). These results demonstrate that while heterologous immune serum exacerbates disease in DENV-infected mice, homologous serum can protect from disease.
To confirm that antibodies were responsible for the decreased survival time observed in mice treated with immune serum, the capacity of purified monoclonal antibodies (mAbs) to enhance disease was tested. AG129 mice were administered 15 μg of mouse monoclonal antibody 2H2 (IgG2a anti-membrane protein (prM/M), DENV1-4 reactive), 4G2 (IgG2a anti-envelope protein (E), pan-flavivirus-reactive) or 3H5 (IgG1 anti-E, DENV2-specific) and infected with S221 (anti-DENV-treated mice). Isotype-treated mice were administered an isotype-matched antibody of irrelevant specificity. Anti-DENV-treated mice exhibited severe disease and died 4 to 5 days after infection, whereas isotype-treated mice survived longer and were sacrificed when they developed neurological symptoms around 8–13 days after infection (fig. 2A), demonstrating that ADE is not restricted to one IgG subclass or DENV epitope.
To determine the relationship between disease outcome and the amount of antibody administered, doses ranging from 0.5 to 500 μg of the neutralizing mAb 4G2 were administered to S221-infected mice. While mice treated with 2 and 15 μg exhibited severe disease and decreased survival time, mice administered 500 μg were protected from severe disease and survived longer than control mice (fig. 2B). These results show that, depending on the concentration, a neutralizing antibody can either increase or decrease disease severity.
Decreased survival following administration of DENV-reactive mAbs was also observed in AG129 mice infected intradermally (i.d.) with 5×108 GE S221 (fig. 2C); in AG129 mice infected i.v. with 5×107 or 5×106 GE S221 (fig. 2D); and in 129/Sv mice lacking only the type I IFN receptor (A129), although a higher amount of antibody and a 200-fold greater dose of virus were necessary to induce severe disease in these mice compared to AG129 (fig. 2E). Therefore, ADE-induced severe dengue disease occurs under various experimental conditions.
Further experiments were performed in AG129 mice infected with 5×108 GE S221 i.v. on day 0 and treated on day -1 and 1 with 15 μg 2H2 (anti-DENV-treated mice) or an isotype control unless stated otherwise. At this dose, it was possible to assess how antibodies turn a mostly non-lethal disease in isotype-treated mice into a lethal disease resembling DHF/DSS in anti-DENV-treated mice.
Elevated levels of various cytokines (IL-6, IL-10, IL-8, IFN-γ) (Chaturvedi et al., 2000), presence of TNF (Green et al., 1999), increased vascular permeability (Halstead, 2007), thrombocytopenia (Halstead, 2007), increased hematocrit (Kittigul et al., 2007), and gastrointestinal bleeding (Chiu et al., 2005) are hallmarks of DHF/DSS in humans. Anti-DENV-treated mice had significantly higher levels of TNF, IL-6 and IL-10 compared to isotype-treated mice 90 hours after infection (fig. 3A). In addition, IL-1β, IL-9, IL-12(p40), IL-17, KC, G-CSF and RANTES were elevated in anti-DENV-treated mice compared to isotype-treated mice 72 and/or 90 hours after infection (supplementary fig. S1). Neither IL-6 neutralization nor IL-10 receptor blockade had any effect on the survival time of anti-DENV-treated mice, but neutralization of TNF prevented early death (fig. 3B). Shortly before succumbing, anti-DENV-treated mice exhibited low platelet counts (fig. 3C), elevated hematocrit (fig. 3D), increased vascular permeability in the liver, as measured by extravasation of Evans Blue (fig. 3E), and gastrointestinal hemorrhage (fig. 3F), paralleling DHF/DSS symptoms.
In the liver of anti-DENV-treated mice, levels of viral RNA were 8-fold, 42-fold and 10-fold higher than isotype-treated mice 48, 72 and 90 hours after infection, respectively (fig. 4A). In the small intestine, levels of viral RNA were 3.3-fold higher at 72 hours, and 10-fold higher at 90 hours (fig. 4A). In the spleen and serum, respectively, RNA levels were 4.2-fold and 2.8-fold higher at 72 hours (fig. 4A). In all other organs examined, no significant differences between anti-DENV-treated and isotype-treated mice were observed until 90 hours after infection, at which time viral RNA levels were higher in the kidney, stomach, large intestine and brain of anti-DENV-treated mice (fig. 4B). At 90 hours post-infection, infectious virus was detected in all the organs by plaque assay on BHK cells (supplementary fig. S2), and the differences observed in viral RNA levels were confirmed. In summary, viral RNA levels were first increased in the liver of anti-DENV-treated mice by 48 hours, followed by a general increase in other organs, most notably the small intestine.
In vitro, ADE of DENV infection requires the presence of FcγRs on target cells (Halstead et al., 1977), suggesting the involvement of the Fc portion of the antibody. To test this in vivo, mice were infected with 5×108 GE S221 and, one day later, treated with equimolar amounts of either an isotype control antibody, the anti-DENV antibody 2H2, the F(ab′)2 of 2H2, intact 2H2 and an FcγR-blocking antibody or the F(ab′)2 of 2H2 chemically coupled to the isotype control. Viral RNA levels in the liver were elevated 72 hours after infection only in the mice treated with either intact anti-DENV antibody or the F(ab′)2 of 2H2 chemically coupled to the isotype control (fig. 4C). In addition, blocking FcγRs II and III prevented the increase in viral load in mice administered intact 2H2 (fig. 4C). These results demonstrate that ADE requires an interaction between the Fc portion of the virus-bound antibodies and FcγRs. The DENV-reactivity and the presence (or absence) of the Fc portion of the isotype control, the anti-DENV antibody 2H2, the F(ab′)2 of 2H2 and the F(ab′)2 of 2H2 chemically coupled to the isotype control were confirmed by ELISA on DENV2-coated plates using a secondary antibody specific for either the Fc portion (supplementary fig. S3, left panel) or the F(ab′)2 portion (supplementary fig. S3, right panel).
The liver and small intestine were the tissues in which levels of viral RNA were most elevated in anti-DENV-treated mice relative to isotype-treated mice. To investigate infection of the cells in those tissues, total mononuclear cells were purified, stained for various surface molecules and intracellular DENV membrane protein (DENV prM, using clone 2H2), and analyzed by flow cytometry. In the liver, DENV prM+ cells were CD45− and CD31+, which is phenotypically consistent with liver sinusoidal endothelial cells (LSECs) (Katz et al., 2004) (fig. 5). A large percentage of LSECs expressed DENV prM at 48 and 72 hours in anti-DENV-treated mice, whereas the percentage of DENV prM+ LSECs was low in isotype-treated mice (fig. 5 upper and lower left part). Expression of DENV prM was negligible in CD45+ and CD45−CD31− cell populations in both groups of mice. In the small intestine (fig. 6 upper and lower left part), the DENV prM+ cells were MHC-II+, CD103−, Ly-6C−, CD11clo-int, CD11b+ and F4/80+, consistent with the phenotype of lamina propria macrophages (LPMs) (Uematsu et al., 2008) (MHC-II+ cell populations in fig. 6 and MHC-II− cell populations in supplementary fig. S4). At 48 hours, a small percentage of LPMs expressed DENV prM in both isotype-treated and anti-DENV-treated mice. This percentage was similar in both groups, indicating that infection of LPMs is not directly affected by anti-DENV antibodies. By 72 hours, the percentage of prM+ LPMs had increased in both groups, but was three times higher in anti-DENV-treated mice.
To assess viral replication in situ, frozen tissue sections were stained for DENV non-structural protein 3 (NS3), which is absent from the virion, but produced by the host cell during viral genome translation. In the liver, NS3 co-localized with cells expressing CD31, confirming productive infection of LSECs (fig. 5 lower right part). Co-localization of NS3 with CD68 was observed only where CD31 was expressed as well, likely because CD68+ macrophages are located in the liver sinusoids, surrounded by CD31+ LSECs. In the small intestine, all NS3 localized to cells expressing F4/80 and CD68 (fig. 6 lower right part), confirming productive infection of LPMs. To determine the effect of anti-DENV antibodies on infection of LSECs by non-mouse-passaged DENV strains, immunofluorescent staining was performed on liver sections from anti-DENV-treated and isotype-treated mice 72 hours after infection with four clinical isolates: DENV1 Julia (Nicaragua), DENV2 PL046 (Taiwan), DENV3 UNC3001 (Sri Lanka) or DENV4 H241 (Philippines). Anti-DENV antibodies greatly increased infection of LSECs by all clinical isolates (fig. 7).
Using the first in vivo model for ADE-induced severe dengue disease, we have shown that anti-DENV antibodies enhance infection of LSECs, ultimately resulting in increased infection of other tissues, low platelet counts, elevated hematocrit, cytokine storm, intestinal hemorrhage and early death, which was prevented by neutralizing TNF. As this model is based on passive transfer of anti-DENV antibodies, it is analogous to DHF/DSS in infants, which is believed to result from ADE mediated by DENV-specific maternal antibodies. With respect to secondary infections, this work separates the antibody component from other aspects of a secondary immune response, enabling the role of antibodies in DENV-induced disease to be studied in isolation.
Although many of the symptoms of human DHF/DSS were observed in the anti-DENV-treated mice in our study, the fact that we used IFN receptor-deficient mice must be taken into account when interpreting our results because of the potential role of IFN in both pathogenesis and protection. This study would ideally have been performed using immunocompetent mice, but wildtype mice did not permit detectable viral replication and did not exhibit signs of disease. A potential explanation for this is that although DENV is able to inhibit IFN signaling in human cells (Jones et al., 2005; Munoz-Jordan et al., 2003), viruses that are able to block IFN signaling in human cells have been shown to fail to do so in mouse cells (Young et al., 2001). Accordingly, it seems that the virus used in the present study is unable to sufficiently disrupt IFN signaling to allow replication in wildtype mice. Thus, although studying DENV infection in complete absence of IFN signaling has limitations, IFN receptor-deficient mice permit viral replication and develop disease much like that observed in humans, making these mice useful for studying DENV-induced disease. The use of a mouse-passaged virus for these studies could potentially be another limitation. However, increased infection of LSECs via ADE was not restricted to the mouse-passaged dengue strain, as Ab-mediated increased infection of LSECs was confirmed with all the DENV strains tested, including clinical isolates from each serotype. Interestingly, overt disease was only observed in mice infected with the mouse-passaged strain S221, likely due to its increased ability to replicate in mice relative to the clinical isolates.
The present study shows in vivo that anti-DENV antibodies specifically enhance infection of LSECs, as only the liver exhibited increased viral loads during the first 48 hours after infection. Anti-DENV antibodies seem to directly and exclusively enhance infection of LSECs, subsequently resulting in the increased infection observed later in other organs. The finding that LPMs are infected to the same level at 48 hours regardless of the presence of antibody provides support for this scenario. Only after increased replication in LSECs has occurred is a higher level of infection in LPMs detected in anti-DENV-treated mice. Similarly, in other organs, higher viral loads in anti-DENV-treated mice were detected only after increased replication in LSECs. A role for LSECs in permitting ADE seems reasonable, considering that LSECs specialize in filtering small particles, including particles the size of DENV virions, out of the bloodstream (Elvevold et al., 2008), promote active antigen uptake via FcγRs (Elvevold et al., 2008), and constitutively express L-SIGN (Bashirova et al., 2001) and mannose receptors (Elvevold et al., 2008), which have been demonstrated to facilitate DENV infection (Miller et al., 2008; Tassaneetrithep et al., 2003).
ADE in vitro results from increased attachment of virus to the surface of cells through interaction of antibody-virus complexes with cellular FcγRs (Halstead et al., 1977), which subsequently results in enhanced infection. Cells that lack attachment factors, and consequently do not permit efficient attachment of virus in the absence of antibodies, are most susceptible to ADE (Boonnak et al., 2008). This likely occurs because, in the absence of attachment factors, binding of antibody-virus complexes with cell surface FcγRs greatly increases virus-cell interactions. On cells that express high levels of attachment factors, it is likely that the virus-cell interactions that are antibody-independent substantially exceeds those that are antibody-dependent and, therefore, the effect of antibodies becomes negligible. Accordingly, the effect of ADE in vitro is most pronounced on monocytes (Kou et al., 2008) and macrophages (Blackley et al., 2007), which exhibit very low susceptibility to antibody-independent infection in vitro (Diamond et al., 2000), and on DCs expressing low levels of attachment factors (Boonnak et al., 2008).
In contrast to the findings that, in vitro, macrophages and monocytes are not easily infected in the absence of antibodies, we have found that, in mice, macrophages of the small intestine lamina propria (as described above), as well as macrophages and DCs of many other tissues throughout the mouse (data not shown), are highly susceptible to antibody-independent infection but not subject to ADE. In contrast, LSECs were the only cells supporting ADE in vivo, and they were, at the viral dose used in this study, only infected to a limited extent in the absence of antibody. It is unclear whether the discrepancy between our results and those obtained with isolated human cells is due to differences between mice and humans or due to differences in experiments performed in vitro versus in vivo. Regardless, our finding that, in mice, ADE occurs with LSECs, but not other cell populations, is in line with the observation that ADE occurs with cells that permit only low levels of antibody-independent infection.
An important point to consider is the influence of potential differences between mouse and human LSECs on our results. Attachment factors on human and mouse LSECs may differ in their expression level and ability to bind dengue virus, and therefore mouse and human LSECs may differ in their susceptibility to ADE. It could be argued that ADE is readily observed on mouse LSECs because they potentially lack attachment factors analogous to the ones binding DENV on human LSECs and, therefore, mouse LSECs could permit ADE even if human LSECs do not. We believe that this is not the case because mouse LSECs, while infected only to low levels in the absence of antibodies at the dose used in the present study, permit much greater levels of antibody-independent infection at higher viral doses (data not shown). This suggests that mouse LSECs, like human LSECs, do express dengue attachment factors, but, unlike macrophages, not at levels high enough to override the effect of antibodies.
The liver has been suggested to be an important site of replication for DENV in humans (Rosen et al., 1999), and the presence of DENV RNA or antigen in the liver of patients who succumbed to DHF/DSS has been reported (Balsitis et al., 2009; Jessie et al., 2004; Rosen et al., 1999). However, results from studies investigating the cellular localization of DENV antigen within the liver have been inconsistent. While Jessie et al. reported the presence of antigen in Kupffer and endothelial cells, Balsitis and colleagues found infection only in hepatocytes. Both experimental and methodological factors could account for differences in these findings. Studies have varied based on which viral proteins were detected, either structural or non-structural, and whether polyclonal or monoclonal Abs were used. In our study, cellular tropism in the liver was assessed using two complementary methods: flow cytometry using a mAb against prM and immunohistochemistry using polyclonal antibodies against NS3. In contrast to previous studies, the cell type permitting DENV infection was determined by both methods to be LSECs based on the expression of cellular markers rather than morphology. Additionally, increased infection of LSECs due to antibodies was observed with four clinical isolates (one of each serotype), further strengthening our results.
Our results confirm in vivo that even neutralizing antibodies have the potential to induce ADE (Mehlhop et al., 2007; Pierson et al., 2007), provided that the occupancy threshold required for neutralization is not reached (Burton, 2002). To be neutralizing, and therefore optimally protective, an antibody must be of high enough affinity for neutralizing epitopes on the surface of the virus, and it must be present in sufficient concentration (Pierson and Diamond, 2008). Failure to fulfill either of these requirements will prevent neutralization. This idea can be applied to sequential DENV infections to explain why antibodies induced by one DENV serotype, although protective against that serotype (Whitehead et al., 2007), may increase the risk of severe disease upon infection with a heterologous serotype due to enhanced infection of cells bearing FcγRs by subneutralized viral particles (Halstead, 2003). It seems likely that due to differences in the surface antigens between serotypes, only some of the antibodies raised against one will react with another, and, of the cross-reactive antibodies, some may have a reduced affinity for the second serotype. Therefore, during a secondary infection by a heterologous serotype, the threshold for neutralization is less likely to be reached and, consequently, ADE-mediated severe disease is more likely to occur. In our study, this is exemplified by the fact that the cross-reactive, neutralizing mAb 4G2 increased mean survival time at high doses, but reduced mean survival time at low doses, demonstrating in vivo that neutralization and enhancement are simply related by the stoichiometry of antibody binding to the surface of viral particles.
Multiple factors have been hypothesized to influence the severity of dengue disease, including subneutralizing levels of antibodies (Halstead, 1982), virus serotype and genotype (Rico-Hesse, 2007) and activation of serotype cross-reactive memory T-cells (Rothman, 2003). Here, we demonstrate that the presence of subneutralizing levels of DENV-specific antibodies can be sufficient to modify the course of infection, resulting in higher viral load, and ultimately severe disease characterized by increased cytokine release, vascular permeability, low platelet counts, increased hematocrit, gastrointestinal hemorrhage and early death. Beyond dengue disease, our findings have implications for the fields of humoral immunity and vaccine design, as they suggest a mechanism by which subprotective humoral responses may, under some circumstances, have pathological consequences.
Sv/129 mice deficient in type I (A129) or type I and II (AG129) interferon receptors, obtained from H. Virgin, and all other mice (purchased from The Jackson Laboratory) were housed under SPF conditions. All animal experiments were approved by the Animal Care Committee at LIAI.
2H2 (IgG2a anti-DENV1-4 prM), 4G2 (IgG2a anti-all flavivirus E), 3H5 (IgG1 anti-DENV2 E), 15F3 (IgG1 anti-DENV1 NS1) and C44 (IgG2a anti-colchicine) hybridomas were purchased from ATCC. Hybridomas were grown in PFHM-II (Gibco) with penicillin (100 U/ml), streptomycin (100 μg/ml) and 55 μM β-mercaptoethanol. Clarified supernatants were concentrated using Amicon 50,000 MWCO filter units (Millipore) and purified using protein G-coupled resin according to the manufacturer’s instructions (Pierce). F(ab′)2 fragments were prepared using the F(ab′)2 Preparation kit for mouse IgG2a (Pierce) by digesting purified IgG with immobilized pepsin at 37° C for 4 hours. The digest was incubated with Protein A-coupled resin to remove large Fc fragments and undigested antibody. Unbound protein was dialyzed in 20,000 MWCO Slide-A-Lyzer cassettes (Pierce) to remove any remaining digested fragments. To couple 2H2 F(ab′)2 fragments to the isotype control antibodies, F(ab′)2 fragments were labeled with Sulfo-SMCC (Pierce) and mixed at an equimolar ratio with isotype control antibodies pretreated with SATA (Pierce). All antibodies were dialyzed against PBS, concentrated and sterile filtered prior to use in experiments. The purity of antibody preparations was verified by SDS-PAGE and binding to DENV was assessed by ELISA. Protein content was quantified using a BCA protein assay kit (Pierce).
As previously described (Shresta et al., 2006), to generate D2S10, the C6/36 mosquito cell-adapted DENV2 isolate PL046, obtained from Dr. Huan-Yao Lei (National Cheng Kung University, Taiwan), was passaged 10 times between the serum of AG129 mice and C6/36 mosquito cells. The biological clone S221 was obtained from D2S10 by growing virus from individual plaques on BHK monolayers three times serially. DENV1 strain Julia (Nicaraguan isolate) was obtained from Dr. Eva Harris, UC Berkeley. DENV3 strain UNC3001 (isolate from Sri Lanka) was obtained from Dr. Aravinda de Silva. DENV4 strain H241 (isolate from the Philippines) was purchased from ATCC. Julia, PL046, UNC3001 and H241 were quantified using flow cytometry as described previously (Lambeth et al., 2005) and expressed as FACS infectious units (FIU). Viral stocks were prepared as previously described (Prestwood et al., 2008). For ELISA, virus was purified over a sucrose gradient as previously described (Prestwood et al., 2008).
Immune serum was obtained from AG129 mice infected with 1.5×106 FACS infectious units (FIU, as defined by infection of C6/36 cells) of DENV1 strain Julia, DENV2 strain E128-IC (Prestwood et al., 2008) (used due to its reduced capacity to cause paralysis in AG129 mice compared to the parental strain PL046), DENV3 strain UNC3001, DENV4 strain H241.
If not stated otherwise, AG129 mice were inoculated intravenously with 5×108 GE of S221. Antibody or serum was administered intraperitoneally in 200 μl PBS one day before and one day after infection.
As described previously (Prestwood et al., 2008), tissues were collected into RNAlater (Qiagen) and subsequently homogenized. RNA was isolated and DENV and relative 18S were quantified using real-time qRT-PCR.
109 GE per well of sucrose-purified S221 was used to coat 96-well plates overnight at 4°C in 50 μl 0.1M NaHCO3. Next, virus on plates was UV-inactivated and plates were washed of unbound virus using 0.05% (v/v) Tween 20 (Sigma) in PBS (Gibco). After blocking with 2% (w/v) BSA in PBS (1 hour, room temperature), purified monoclonal antibodies or F(ab′)2 fragments were titrated on virus and unbound antibody was removed by washing. Bound antibody was detected using HRP-conjugated goat anti-mouse IgG antibodies specific for either Fc or F(ab′)2 portions (Jackson Immunoresearch) and TMB (eBioscience).
Three-fold serial dilutions of serum were incubated with 1.5×1010 GE of S221 for 1 hour at room temperature in a total volume of 250 μl PBS. Subsequently, 2×106 C6/36 cells were infected with 200 μl of the virus-antibody mix for one hour at 28°C. Cells were washed two times with 1 ml of PBS, and the cells were incubated at 28°C in 500 μl L-15 medium (Gibco) containing 5% FCS, penicillin (100 U/ml) and streptomycin (100 μg/ml) for 23 hours. For each antibody dilution, the percent of infected cells was determined by flow cytometry as previously described (Lambeth et al., 2005) using 2H2-biotin and streptavidin-APC (Biolegend). The percent of infected cells was normalized to 100% (infection without antibody) and the average of two experiments is shown.
Serum samples were analyzed by the Immune Reconstitution Facility, Duke University, Durham, NC, USA, using the Bio-Plex Multiplex Cytokine Assay (Bio-Rad). TNF in serum was measured using the Ready-SET-Go! TNF ELISA kit (eBioscience).
Platelet counts and hematocrit were measured on a Hemavet 950FS (Drew) according to manufacturer’s instructions.
Functional grade rat anti-TNF (clone MP6-XT22; eBioscience) was used to neutralize TNF as previously described (Shresta et al., 2006), except that 100 μg was injected i.p. each day for four days following infection, starting on day 1. As previously described, 500 μg of functional grade rat anti-IL-6 (Starnes et al., 1990) (clone MP5-20F3; BioXCell) or anti-IL10R (Brooks et al., 2006) (clone 1B1.3A; BioXCell) were administered i.p. on day 2 and 3. To block FcγRs II and III, 500 μg of mAb 2.4G2 (BioXCell) was administered i.v. on day 0 and 2.
Evans Blue (Sigma), which binds avidly to albumin, is used to assess vascular permeability, as the extravasation of albumin into tissues indicates relative integrity of the vascular endothelium. Evans Blue (0.2 ml, 0.5% (w/v) in PBS) was administered intravenously and, after 15 minutes, animals were sacrificed and extensively perfused with PBS. Livers were collected into formamide and, after overnight extraction at 37° C, the dye was quantified by measuring absorbance at 610 nm.
Tissues were collected into RPMI (Gibco) containing 10% fetal calf serum (Gemini Bio-Products) and stored on ice. To isolate cells from the small intestine lamina propria, Peyer’s patches were excised and the tissue was filleted longitudinally and washed in cold PBS. Epithelial cells were removed from segmented tissue by agitation for 30 min at 37° C in PBS containing 10% (v/v) FCS, EDTA (10 mM), HEPES (20 mM), penicillin (100 U/ml), streptomycin (100 μg/ml) and polymyxin B (10 μg/ml; Calbiochem), followed by extensive washing with PBS. The small intestine and liver were minced and digested for 20 min at 37° C with agitation in RPMI containing 10% fetal calf serum and 1.5 mg/ml collagenase VIII (Sigma) in approximately 25 ml. Tissues were immediately washed in RPMI and digests were collected via centrifugation. Resuspended tissues were pressed through 70 μm strainers (BD) using the plunger of 3 ml syringes (BD). Liberated cells were collected, and pelleted by centrifugation. Cells were resuspended in a 14.7% Optiprep (Sigma) solution diluted in 150 mM NaCl with 10 mM HEPES at pH 7.4 and layered over 22.2% Optiprep solution diluted in the same buffer. Cells were spun for 20 minutes at 750 × g at 25° C and washed immediately after with RPMI. 2×105 cells per well were plated in 96-well plates, and FcγRs were blocked with 1 μg of mAb 2.4G2. Cells from the small intestine were stained using FITC anti-Ly-6C (Pharmingen), PE anti-CD103 (Pharmingen), PE-Cy7 anti-CD11c (Pharmingen), APC anti-CD11b (Biolegend), Pacific Blue anti-MHC II (Biolegend), Alexa-Fluor 700 anti-Gr-1 (Pharmingen) and PerCP-Cy5.5 anti-F4/80 (Biolegend) or with appropriately labeled isotype controls. Cells from the liver were stained with PE anti-CD31 (Pharmingen) and PE-Cy7 anti-CD45 (Pharmingen). Cells were fixed and permeabilized using the Cytofix/Cytoperm Kit (BD). For intracellular staining, all steps were performed in 1x BD Perm/Wash. Cells were blocked briefly and stained for 30 minutes with biotinylated 2H2 or isotype control. After washing, cells were stained with streptavidin conjugated to either APC (eBioscience) for the liver cells or APC-Alexa Fluor 750 (Caltag Laboratories) for the small intestine cells. The gate for DENV antigen-positive cells was set at 0.1% of the isotype control staining. Data were collected using an LSR II (BD) and analyzed with FlowJo software (Tree Star).
Tissues were embedded in O.C.T. compound (Sakura). Sections (6 μm) were cut and stored at −80° C. Frozen sections were thawed and fixed for 10 minutes in acetone at 25° C, followed by 8 minutes in 1% paraformaldehyde (EMS) in 100 mM dibasic sodium phosphate containing 60 mM lysine and 7 mM sodium periodate at pH 7.4 on ice. Sections were blocked first using the Avidin/Biotin Blocking Kit (Vector Labs) followed by 5% normal goat serum (Caltag Laboratories) and 1% BSA (Sigma) in PBS. Sections were stained overnight with purified rabbit polyclonal anti-DENV NS3 (a generous gift from the Novartis Institute for Tropical Diseases, Singapore) and either rat anti-mouse CD31 (Pharmingen) for liver sections or PE-labeled rat anti-mouse F4/80 (Caltag Laboratories) for small intestine sections. Sections were then washed and stained with DyLight 649-labeled goat anti-rabbit IgG (Jackson Immunoresearch) and biotinylated goat anti-rat IgG (Pharmingen) and then with PE-labeled streptavidin (eBioscience). Sections were then blocked again with Avidin/Biotin Blocking Kit followed by 10% rat serum in PBS, and finally stained with biotin anti-mouse CD68 (Serotec) and FITC-labeled streptavidin (Pharmingen). Images were recorded using a Marianas deconvolution fluorescence microscope (3i) and prepared using Adobe Photoshop.
Supplementary figure S1 (related to figure 3)
Cytokine levels in the serum of DENV-infected mice in the presence or absence of anti-DENV antibody 48, 72 and 90 hours after infection (n=3–9).
P-values from two-tailed unpaired t-test with Welch’s correction, c.i. 95%: * P≤0.05, ** P≤0.01, *** P≤0.001, n is the number of mice per group.
Supplementary figure S2 (related to figure 4)
The presence of virus in tissue homogenates was assessed by plaque assay on BHK cells 90 hours post-infection in the presence (black symbols) or absence (white symbols) of anti-DENV antibody.
P-values from two-tailed unpaired t-test with Welch’s correction, c.i. 95%: * P≤0.05, ** P≤0.01, *** P≤0.001, n is the number of mice per group, MLN: mesenteric lymph node, PLN: peripheral lymph nodes (axillary, brachial and inguinal), BM: bone marrow.
Supplementary Figure S3 (related to figure 4)
ELISA of isotype control (white square), anti-DENV antibody (2H2, IgG2a against prM/M, black square), F(ab′)2 of anti-DENV antibody (black triangle), and F(ab′)2 of anti-DENV antibody coupled to the isotype control (black diamond) on S221-coated plates. Antibody against either the Fc fragment (left panel) or the F(ab′)2 fragment of IgG (right panel) were used for detection.
Supplementary figure S4 (related to figure 6)
Percentage of DENV antigen-positive cells in MHC-II− lamina propria cell populations. Gating is indicated on FACS plots (the grey histogram depicts cells stained with an isotype control antibody) and population statistics are indicated in the bar graphs.
The project described was supported by NIH grants AI060989 and AI077099-01 to S.S. and NIH grant U54 AI057157 from Southeast Regional Center of Excellence for Emerging Infections and Biodefense to P.F. Sparling, a fellowship from the Swiss National Science Foundation to R.M.Z. (PBEZB-118877) and a fellowship from the Novartis Foundation to R.M.Z.
We thank Drs. Chris Benedict, Carl Ware, Shane Crotty, Lauren Yauch and Stuart Perry at LIAI for critical reading of the manuscript. We thank Robyn Miller, Dr. Olga Turovskaya, Sarala Joshi, Dr. Satoshi Fukuyama, Steven Lada and Burton Barnett for technical help.
The authors declare that they have no competing financial interests.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.