|Home | About | Journals | Submit | Contact Us | Français|
Background.Although broadly neutralizing antibodies (bNAbs) have been shown to block a diverse array of cell-free human immunodeficiency type 1 (HIV-1) infections, it remains unclear whether these antibodies exhibit similar potency against mature dendritic cell (mDC)–mediated HIV-1 trans-infection.
Methods.Sensitivity to bNAbs targeting HIV-1 envelope surface unit gp120 (VRCO1, PG16, b12, and 2G12) and transmembrane domain gp41 (4E10 and 2F5) was examined for both cell-free and mDC-mediated infections of TZM-bl and CD4+ T cells.
Results.Compared with cell-free infection, mDC-mediated infection was significantly less susceptible to gp120-directed bNAbs for the majority of virus isolates. A b12 antigen-binding fragment blocked both cell-free and mDC-mediated infection with equal efficiency. In contrast, cell-free and mDC-associated viruses were equally sensitive to gp41-directed bNAbs. Anti-gp41 bNAbs bound to the surface of mDCs and localized at the mDC–T cell synaptic junctions in the absence of virus.
Conclusions.Anti-gp41 bNAbs have the potential to inhibit mDC-mediated HIV-1 infection because they bind plasma membranes prior to the formation of an infectious synapse, positioning them to neutralize subsequent virus transfer. As opposed to gp120-directed antibodies, anti-gp41 bNAbs might prevent HIV-1 infection if transmission or spread at the initial site of invasion occurs from a DC-associated source.
Recently isolated anti–human immunodeficiency virus type 1 (HIV-1) broadly neutralizing antibodies (bNAbs) provide new hope for a preventive HIV-1 vaccine [1, 2]. Both envelope glycoprotein (env) surface unit (gp120)–directed antibodies, such as VRC01, b12, PG16, and 2G12, and anti-transmembrane (gp41) bNAbs, such as 4E10 and 2F5, are effective in blocking diverse variants because they target conserved domains [1–7]. To date, rationally designed immunogens have failed to elicit these types of bNAbs. Given the potency and breadth of the bNAbs, especially VRC01 and PG16, use of novel technologies such as recombinant adeno-associated virus vectors is being contemplated to generate appropriate bNAb titers prior to HIV-1 exposure in vaccinated individuals .
The breadth and potency of the majority of bNAbs have been examined primarily against cell-free viral variants. It remains unclear, however, whether transmission occurs because of cell-associated virus or cell-free virus . During mucosal transmission, which is the most common route of HIV-1 acquisition in the world, cell-associated virus may play a seminal role. Numerous studies have shown that virus associated with a cell, such as dendritic cells (DCs), is more infectious than free virus, and this enhanced infectiousness may allow a virus to establish infection in a new host [10, 11]. Furthermore, in contrast to cell-free virus, cell-associated virus may be protected from antiviral factors, such as antibodies, present in the donor at site of transmission . Given the potential importance of cell-associated virus during mucosal HIV-1 transmission, in this study we examined whether bNAbs have similar efficacy in blocking infection mediated by DC-associated virus, compared with cell-free virus.
HIV-1 clones (NL4-3, 89.6, YU2, JRCSF, REJO, and CH077), Gag-eGFP (which expresses a Gag-eGFP fusion protein), and monoclonal antibodies (mAbs; VRC01, b12, 4E10, 2F5, and 2G12) [1, 5–7, 12, 13] were obtained from the National Institutes of Health AIDS Research and Reference Reagent Program. Lai and Lai/Bal env were obtained from Dr Michael Emerman, and Q23 was obtained from Dr Julie Overbaugh. Human embryonic kidney fibroblast cells (HEK293T) were transfected to prepare infectious virus stocks or Gag-eGFP–containing fluorescent virus-like particles (Gag-eGFP VLPs), and virus titers were determined on TZM-bl cells, as described previously [14–17]. In addition, p24gag content was assessed using a previously described in-house p24gag enzyme-linked immunosorbent assay (ELISA) .
Primary human immature and mature DCs were derived from peripheral blood CD14+ monocytes, as described previously . Primary human CD4+ T cells were isolated from CD14-depleted peripheral blood mononuclear cells by use of CD4+ T-cell–conjugated magnetic beads (Miltenyi Biotech), activated using 2% phytohemagglutinin for 2 days, and cultured in complete Roswell Park Memorial Institute 1640 (RPMI) medium supplemented with 50 U/mL of recombinant human interleukin 2.
The PG16 antibody was generated from HEK293T transfections of pFUSEss-CHIg-hG1 and pFUSE2ss-CLIg-hk (InvivoGen) containing synthetically constructed variable heavy (VH) and light (VL) chains (GenBank accession numbers GU272043.1 and GU272044.1, respectively) . Supernatant immunoglobulin G (IgG) was concentrated through Amicon filters, passed through a Protein A column (Sigma-Aldrich), and quantified by ELISA (Immunology Consultants). The b12 antigen-binding fragment (Fab) was generated by ficin protease digestion (Pierce). PG16 IgG and b12 Fab sizes were confirmed by sodium dodecyl sulfate polyacrylamide gel electrophoresis analysis. Mature DCs (1 × 106 cells) were exposed to 500 ng p24gag of cell-free virus stock (multiplicity of infection, 0.2) for 2–3 hours at 37°C and then were washed 4 times with complete RPMI medium to remove unbound HIV-1 particles. Antibody Fc-binding receptors (FcRs) on mDCs were blocked by incubating cells with FcR blocking reagent (Miltenyi Biotec) for 1 hour at room temperature prior to addition of bNAbs. Two-fold serial dilutions of the inhibitor were incubated with either 500 cell-free infectious viruses (approximately equivalent to 10 ng of p24gag) or 1 × 104 virus-exposed mDCs (containing 5 ng p24gag virus equivalents) for 1 hour at 37°C prior to adding 1 × 104 target cells in each well. Inhibition was assessed by measuring relative light units (RLUs) either from β-galactosidase expression using Galacton-Light Plus (Applied Biosystems) at 48 hours after infection in TZM-bl cells or from firefly luciferase production using Bright Glo Luciferase Assay Systems (Promega) at 72 hours in autologous CD4+ T cells .
Mature DCs (2 × 106) were incubated with 20 μg/mL antibody in phosphate-buffered saline (PBS) for 1 hour at 4°C and washed with cold PBS 3 times, and antibodies and nucleus were visualized by staining with Alexa 594–conjugated goat antihuman IgG (Invitrogen) and DAPI (Sigma), respectively. Antibody localization at mDC–T cell synaptic junctions was determined by incubating 1 × 105 Gag-eGFP VLP–exposed or Gag-eGFP VLP–unexposed mDCs with 50 μg/mL antibody for an 1 hour at 37°C, followed by washing and then incubating for an additional 2 hours with 2 × 105 CellTracker Blue (Invitrogen)–prelabeled autologous activated CD4+ T cells. Cocultures were washed twice with PBS, fixed with 4% paraformaldehdye, permeabilized, and stained with CD81 mAb (a known marker of mDC–T cell synaptic junctions; clone JS-81, BD Biosciences) and Alexa 488–conjugated goat antimouse IgG and/or Alexa 594 IgG. Images were acquired with an Axiovert 200 microscope (Zeiss) or an Olympus IX70 microscope equipped with Delta Vision deconvolution software (Applied Precision), and images were analyzed with ImageJ and Photoshop (Adobe) software. The number of mDCs with cell surface–bound antibodies and colocalization of bNAbs with Gag-eGFP VLPs at mDC–T cell infectious synapses was determined on 100 different cells and mDC–T cell synaptic junctions, respectively, in 50 independent fields from each experiment.
Differences between mDC-associated and cell-free HIV-1 were assessed by the t test, the Wilcoxon rank-sum test, or the matched-pairs Wilcoxon rank-sum test. All P values are based on a 2-sided test. All statistical analyses were done with Intercooled Stata, version 8.0 (StataCorp).
Prior to testing the susceptibility of mDC-laden HIV-1 versus that of cell-free HIV-1, we assessed whether target cells were exposed to similar virus amounts during both modes of infection. For each isolate, RLUs generated in the absence of antibody were not significantly different between mDC-mediated and cell-free infections (P > .05, Wilcoxon rank-sum test) (Figure 1). Among all isolates, there was no significant difference in the median RLUs between mDC-mediated and cell-free infections (P > .05, Wilcoxon matched-pairs rank-sum test). This suggests that the amounts of infectious virus in the assay were similar in mDC-mediated and cell-free infections.
Tests of the susceptibility of mDC-mediated HIV-1 trans-infection to mAbs have provided conflicting data, and susceptibility has not been assessed against the newly identified bNAbs PG16 and VRC01 [11, 20–22]. The ability of the bNAbs to inhibit mDC-mediated versus cell-free spread was examined for CCR5-dependent HIV-1 (YU-2, Q23, JRCSF, Lai/Balenv), CXCR4-using virus isolates (Lai and NL4-3), and 1 dually tropic variant (89.6) (Figure 2A–2J and Table 1). Because previous studies suggest that successfully transmitted viruses in newly infected individuals have env with unique genotypic and phenotypic features, we also assessed sensitivity of 2 full-length transmitted/founder strains (REJO and CH077) [23–25]. In all cases except for 89.6, the VRC01 concentration required to inhibit infection by 50% (IC50) was significantly lower for cell-free infection as compared with mDC-associated trans-infection (Figure 2A–E and Table 1). For 89.6, VRC01 did not demonstrate <50% inhibition of either cell-free or mDC-associated HIV-1 at the highest tested doses (2 μg/mL). For bNAb PG16, 3 of the 7 viruses (Lai/Balenv, Lai, and 89.6) demonstrated <50% inhibition at the highest tested concentration (2 μg/mL) (Table 1). In the remaining cases, the IC50 for 3 of the viruses (Q23, REJO, and CH077) was significantly lower for cell-free infection as compared with mDC-associated trans-infection; for JRCSF, YU-2, and NL4-3, the PG16 IC50 was not significantly different between infections initiated with cell-free virus and those initiated with mDC-associated virus (Figure 2H–K and Table 1).
Similar to VRC01 and PG16, the IC50 and IC90 for 2 other anti-gp120–directed bNAbs, 2G12 and b12, were significantly higher (2.5–10-fold) for almost all mDC-mediated virus transmission, compared with cell-free HIV-1 infection (Table 1). Only cell-free and mDC-mediated infection of 89.6 virus particles demonstrated no significant IC50 difference against 2G12. These results suggest that gp120-specific bNAbs are inefficient at neutralizing mDC-mediated HIV-1 trans-infection. In contrast, mDCs transferred significantly less virus to target cells when exposed to Lai virus particles in the presence, as opposed to the absence, of b12 (data not shown). This suggests that mDC-mediated virus transfer can be inhibited by bNAbs such as b12 if the antibody is present at the time of virus capture by mDCs.
We hypothesized that the close physical proximity between the virus-bearing cell and the susceptible target cell may prevent relatively large bNAbs from efficiently inhibiting HIV-1 spread from mDCs to target cells . To examine whether bNAb size influences the inhibition efficiency, we examined the susceptibility of mDC-mediated trans-infection to the b12 Fab. Unlike inhibition by bNAb b12 (Table 1), both Lai and Lai/Balenv were suppressed equivalently by the b12 Fab irrespective of whether target cells were challenged with cell-free or mDC-associated virus particles (Figure 3).
We next determined whether these results also extended to 4E10 and 2F5, which bind gp41 transmembrane domain and lipid determinants in the virus particle membrane [27, 28]. Surprisingly, for viruses that were inhibited by >50% at the highest tested concentration, 4E10 and 2F5 suppressed both cell-free and mDC-mediated virus infection with relatively equal efficiency (Figure 4 and Table 1). Neutralizing potencies of 4E10 and 2F5 are often enhanced when target cells express certain FcRs . Blocking the FcRs expressed on mDCs prior to antibody exposure had negligible impact on the ability of 4E10 and 2F5 to inhibit mDC-mediated trans-infection (Figure 4).
All bNAbs tested were of IgG1 isotype, and thus they were of relatively similar molecular mass. However, compared with the gp120-directed bNAbs, gp41-directed bNAbs can also bind several host cell–derived lipids in the virion membrane bilayer [30, 31]. Because gp41-targeting bNAbs were equally effective at inhibiting both cell-free and mDC-associated virus infections, we hypothesized that they overcame size limitations in the infectious synapse by binding to mDC plasma membrane prior to formation of synapses with target cells. Immunofluorescence microscopy showed that, in contrast to the gp120-directed bNAb 2G12, both gp41-directed bNAbs (ie, 4E10 and 2F5) readily bound mDCs in the absence of virus (Figure 5). Binding of 4E10 and 2F5 was mostly seen at the cell periphery as distinct puncta in the mDC membrane (Figure 5A). Counting of the number of cells with distinct antibody puncta at the cell surface revealed that 4E10 (median, 27.6%; range, 21.2%–37.1%) and 2F5 (median, 26.2%; range, 16.5%–38.3%) bound a significantly greater percentage of mDCs, compared with VRC01 (median, 5.4%; range, 5.2%–6.4%), PG16 (median, 11.7%; range, 6.5%–21.4%), b12 (median, 9.1%; range, 6.2%–10.4%), and 2G12 (median, 3.6%; range, 1.8%–7.3%) (P < .05, Wilcoxon rank-sum test) (Figure 5C). All antibodies bound a significantly greater percentage of mDCs, compared with the secondary antibody alone (median, 1.1%; range 0%–5.6%) (P < .05, Wilcoxon rank-sum test).
This mDC-membrane attachment that occurs independently from virus could potentially poise a bNAb to bind virus particles released from antibody-inaccessible HIV-containing compartments within mDCs prior to the formation of DC–T cell conjugates. To investigate this possibility, localization of bNAbs 2G12 and 4E10 at the mDC–T cell synaptic junction was examined in the presence or absence of virus particles. Because mDCs can both capture and localize HIV-1 particles at virological synapses in an env-independent manner, mDCs were exposed to env-deficient Gag-eGFP VLPs in a manner similar to that for the mDC trans-infection assay . Around 72% (range, 55%–85%) of the mDC–T cell synaptic junctions displayed colocalization of Gag-eGFP VLPs with 4E10 as compared with only 18% (range, 7%–25%) for 2G12 (Figure 6). Furthermore, 4E10 but not 2G12 localized at DC–T cell synaptic junctions (marked by CD81 staining) in the absence of Gag-eGFP VLPs.
To confirm whether these findings extend to inhibition of virus spread from mDCs to T cells, we compared the ability of b12 and 4E10 to inhibit mDC-mediated trans-infection and cell-free virus infection to activated autologous CD4+ T cells. Two different luciferase-expressing replication competent viruses, Lai and Lai/Balenv, required significantly higher b12 but relatively similar 4E10 concentrations to block mDC-mediated versus cell-free infection of autologous T cells; the b12 and 4E10 IC50 and IC90 in mDC–TZM-bl and mDC–autologous T cell cocultures were comparable (Table 1).
In this study, we report that higher gp120-bNAb concentrations, but not gp41-directed bNAb concentrations, are required to inhibit mDC-mediated virus spread, compared with cell-free transmission. Our data suggest that relative resistance of mDC-mediated HIV-1 trans-infection to anti-gp120 neutralizing antibodies can partly be explained by occlusion of the antibodies from the DC–T cell virological synapse. Furthermore, the localization of polyreactive anti-gp41 bNAbs to DC–T cell synaptic junctions independent of virus particles can circumvent infectious synapse size restrictions and lead to effective neutralizing antibody concentrations at the site of mDC-mediated HIV-1 transfer.
In contrast to effective neutralization with gp41 bNAbs, the mDC–T cell infectious synapse enables high-efficiency transmission of HIV-1 particles even in the presence of neutralizing anti-gp120 antibodies (Table 1). Interestingly, a Fab of b12 bNAb could potently inhibit mDC-mediated trans-infection (Figure 3). Recent confocal microscopy and cryo–electron microscopy reconstructions demonstrate that HIV-1 virions are localized within deep invaginations in the mDC plasma membrane and that transfer occurs after T-cell finger-like extensions query the mDC surface [26, 32]. Alternatively, captured HIV-1 particles translocate from mDCs to T cells after localizing within presumably immunoglobulin-inaccessible CD81+ nonlysosomal endocytic compartments [16, 33]. An estimated diameter of 15–40 nm for the mDC–T cell synaptic cleft  likely provides steric hindrance to anti-gp120–directed immunoglobulins, thereby preventing them from accessing the mDC-trapped HIV-1 or from accessing virus particles during transfer to target cells.
Steric constraints, however, did not prevent anti-gp41 bNAbs from restricting the spread of both mDC-mediated and cell-free virus equally, suggesting that the antibody epitope also influences inhibition efficiency. Previous studies have demonstrated that the neutralizing efficiency of 4E10 and 2F5 is directly linked to its ability to recognize both host cell lipids in the virus particle membrane and the gp41 membrane proximal external region (MPER) epitope [27, 28, 31, 35]. These 2 bNAbs initially bind to virus membrane phospholipids such as phospatidylserine and phosphatidylethanolamine prior to recognition of the gp41 MPER epitope. Because the lipid raft-like virus particle lipidome is derived upon virus particle budding from specialized phosphatidylserine, cholesterol, and glycosphingolipid-enriched plasma membrane sites [36, 37], it is probable that 4E10 and 2F5 bNAbs recognize similar lipid raft–enriched mDC or T-cell membrane microdomains present at the mDC–T cell synaptic junctions. In our results, we show that 4E10 binds to mDC cell membranes in the presence and absence of viral particles, which suggests that this surface lipid attachment allows 4E10 to recognize the MPER epitope exposed after virus env encounters the CD4 receptor in the mDC–T cell junctions.
Previous studies have demonstrated that both gp120-directed and gp41-directed bNAbs can block virus transmission to Rhesus macaques upon intravenous or topically applied challenge [38–43]. It should be emphasized, however, that in all these studies the challenge stock consisted of cell-free virus. It remains unclear whether transmission occurs because of cell-free virus or cell-associated virus present in the infected source . Infected specimens such as breast milk, semen, and vaginal secretions contain both cell-free virus and infected cells. In vitro, HIV-1 spread occurs more efficiently between cells than via fluid-phase diffusion, suggesting that cell-to-cell contact may be important during in vivo transmission . Indeed, studies of mother-to-infant transmission via breast milk suggest that the risk of virus acquisition is more closely correlated with cellular virus loads as opposed to free-floating virus RNA levels . Furthermore, recent studies show that an antibody with defective ability to engage Fc-binding receptors was worse at preventing virus acquisition, suggesting that optimal protection requires either effector cell–mediated neutralization of cell-free virus or inhibition of virus dissemination from early infected target cells at the site of invasion . Our studies suggest that anti-gp120 bNAbs may be less effective in preventing HIV-1 acquisition if antibodies need to block cell-to-cell transfer to achieve sterilization.
There have been conflicting reports in the literature on the ability of neutralizing antibodies to inhibit DC-mediated spread of HIV-1 particles to T cells [20–22], possibly because differences in the assay systems and infection assessments have complicated interpretation of the results. Although we used replication-competent viruses, the short assay duration prevented multiple rounds of replication in TZM-bl and CD4+ T cells, and thus we unequivocally addressed the effects of antibodies on entry inhibition. Furthermore, in contrast to previous studies that determined neutralizing antibody inhibitory effects on HIV-1 transfer from productively infected DCs [20, 21], we assessed the effect of bNAbs on mDC-mediated trans-infection only, and there was no contribution of virus transmission from productively infected DCs in our assay system. Furthermore, the target cells were exposed to relatively similar numbers of cell-free or mDC-associated infectious virus particles (Figure 1), suggesting that inhibition differences observed between mDC-associated and cell-free infection cannot be attributed to variation in the number of infectious virions in the challenge.
Similar to some of the previously published results, we observed virus isolate–specific differences in susceptibility to neutralizing antibodies [20–22]. For instance, PG16 equally inhibited both fluid-phase and mDC-associated spread of YU-2 and JRCSF, but it was less efficient at blocking mDC-mediated transfer than at blocking cell-free infection involving Q23 and transmitted/founder strains CHO77 and REJO (Figure 2 and Table 1). Virus isolate–specific differences in susceptibility to anti-gp120–specific neutralizing antibodies could potentially reflect variations in the number of env spikes per capsid, although no published studies to date have quantitatively defined numbers of gp120 trimers per virus capsids for each of these different virus isolates. Because our previous studies have argued that HIV-1 capture and trafficking to the mDC–T cell virological synapse can occur in an env-independent, glycosphingolipid-dependent manner [14, 16], variable incorporation of glycosphingolipid among different virus isolates could potentially also result in differential ability to access the mDC trans-infection pathway [16, 32, 33]. It will be interesting to examine whether other primary isolates, especially other transmitter founder strains, also display relative insensitivity to gp120-directed bNAb blocking of mDC-associated infections [24, 25]. Furthermore, future studies should aim to decipher the viral determinants that influence bNAb susceptibility for cell-mediated transfer.
Different gp120-directed bNAbs also displayed varying abilities to stop mDC-linked spread, compared with fluid-phase diffusion. For instance, compared with cell-free transmission, mDC-mediated transmission was blocked less efficiently by VRC01 for all viruses examined, while PG16 inhibited around 50% of the isolates equivalently, regardless of the mode of virus spread (Figure 2 and Table 1). These differences may potentially stem from variation in antibody structure, which may allow one but not the other antibody to penetrate mDC–target cell synapses. For instance, difference in the complementarity-determining region 3 hypervariable loop length, charge, and hydrophobicity of PG16 versus VRC01 might allow PG16 to more efficiently access mDC-associated virus particles present in deeply recessed compartments . Alternatively, the mechanism with which a bNAb inhibits viruses (ie, envelope trimer occupancy, ligation of multiple trimers on the virus particle surface, induction of gp120 shedding, or induction of β-chemokines) may also determine its potency against mDC trans-infection [46, 47]. Interestingly, the data presented in this report suggest that PG16, along with 4E10 and 2F5, might fall into the category of the recently described class of polyreactive anti-HIV neutralizing antibodies that are postulated to bind to 2 different determinants: a high-affinity interaction with HIV-1 env and low-affinity binding to a poorly characterized secondary epitope (presumably host cell-derived lipids) on the surface of HIV-1 [35, 48]. Interestingly, PG16, which displayed relatively large variations in its ability to inhibit mDC-mediated trans-infection at different dilutions for different virus strains (Figure 2F–J), also showed greater variation in the percentage of bound mDCs, compared with b12, 2G12, and VRC01 (Figure 5), suggesting donor-dependent differences in the expression of the plasma membrane lipid epitope. Our data suggest that use of such polyreactive antibodies in vaccine regimens might significantly improve neutralization ability, especially in the context of cell-associated HIV-1 transmission.
While previous studies have described multiple ways in which HIV-1 avoids neutralization, including gp120 masking by glycosylation and a high mutation rate in the gene encoding env [15, 49, 50], our studies argue for an additional DC-dependent mechanism the virus can use to avoid neutralization. This DC-mediated escape may be especially pertinent for antibody-dependent preventive strategies if DCs play a seminal role in either the acquisition or the spread of virus from the initial site of invasion. To overcome this potential pitfall, blocking HIV-1 acquisition will require the elicitation of relatively high bNAb concentrations, such as those recently achieved with recombinant adeno-associated viruses . Furthermore, future in vitro studies that assess the efficacy of antibody neutralizing should evaluate the effects of the putative inhibitors on DC-associated (and, in general, cell-associated) virus transmission to target cells.
We thank Phyu Hninn Nyein for help with the assays; Boston University Medical Center FACS Core for technical assistance; Dennis Burton for b12 antibody; and the National Institutes of Health (NIH) AIDS Research and Reference Reagent Program for 4E10, 2F5, 2G12, and VRCO1 antibodies and for the following plasmids: YU2, NL4-3, 89.6, CHO77, and REJO, and pGag-eGFP. We thank Greg Viglianti, Andy Henderson, Wendy Puryear, and members of the Sagar and Gummuluru laboratories for their comments.
This study was supported by the NIH (grants AI1077473 to M. S., AI064099 to S. G., and AI081596 to S. G.) and a Doris Duke Charitable Foundation Early Career Development Award (to M. S.).
All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.