Antibody Neutralization of HIV-1 Crossing the Blood-Brain Barrier

HIV-1 can cross the blood-brain barrier (BBB) to penetrate the brain and infect target cells, causing neurocognitive disorders as a result of neuroinflammation and brain damage. The HIV-1 envelope spike gp160 is partially required for viral transcytosis across the BBB endothelium. But do antibodies developing in infected individuals and targeting the HIV-1 gp160 glycoproteins block HIV-1 transcytosis through the BBB? We addressed this issue and discovered that anti-gp160 antibodies do not block HIV-1 transport; instead, free viruses and those in complex with antibodies can transit across BBB endothelial cells. Importantly, we found that only neutralizing antibodies could inhibit posttranscytosis viral infectivity, highlighting their ability to protect susceptible brain cells from HIV-1 infection.

circulating pathogens as well as for controlling brain homeostasis (7). The BBB is made of capillary endothelial cells joined by tight and adherent junctions, covered by a basal membrane, and surrounded by pericytes and astrocytic endfeet (7). Free and cellassociated HIV-1 virions can translocate across the BBB and enter the CNS via mechanisms common to other neuroinvasive pathogens (8). HIV-1 crosses the BBB following a Trojan horse model in which infected cells transmigrate by diapedesis through intact endothelial cells and by a direct paracellular transversal of the BBB damaged as a result of the infection (9,10). HIV-1 also traverses the BBB through a transcytosis pathway (1), which may be initiated by viral envelope gp160 glycoproteins binding to proteoglycans (11) and mannose-6-phoshate receptors on endothelial cells (12). Antibodies specific to the HIV-1 envelope spike are rapidly elicited in infected humans, but only those targeting functional gp160 epitopes are neutralizing (13). Moreover, only rare infected individuals develop broadly reactive antibodies neutralizing most HIV-1 strains (13). These broadly neutralizing antibodies (bNAbs) efficiently protect nonhuman primates from infection and decrease viremia in infected humans (14), and thus, they offer promise for HIV-1 prevention and treatment. Whether bNAbs, and more generally anti-gp160 IgG antibodies, interfere with or block HIV-1 transcytosis across the BBB has remained unknown.
gp160 antibodies do not block HIV-1 transcytosis across the blood-brain barrier. To test the ability of gp160-specific antibodies to block HIV-1 transcytosis across the BBB, we established an in vitro model using the human brain microvascular endothelial cell line hCMEC/D3, which reproduces most characteristics of the BBB endothelium (15). hCMEC/D3 cells grown on Transwell culture membranes for a week assembled in confluent and tight monolayers characterized by a low permeability to 40-kDa DEAE-dextran molecules ( Fig. 1A to C). hCMEC/D3 cell monolayers formed both tight and adherent junctions, with expression of vascular endothelial (VE)-cadherin, junctional adhesion molecule A (JAM-A), and zonula occludens-1 (ZO-1) (Fig. 1D). HIV-1 transcytosis was assayed at day 8 of culture by exposing cells to CCR5-or CXCR4-tropic HIV-1 virions (NLAD8 and NL4.3 strains, respectively), and measuring the HIV-1 p24 protein content in the basal compartment medium of the Transwell (Fig. 1A). We found that on average, ϳ17% of the viral inoculum passing through the porous membrane translocated across the hCMEC/3 endothelium after 4 h (Fig. 1E). In agreement with previous observations (16), envelope-deficient HIV-1 virions (NL-ΔEnv) displayed lower transcytosis capacity (P ϭ 0.0003 versus NLAD8) (Fig. 1E), indicating that HIV-1 transcytosis across the BBB partially depends on envelope glycoproteins. Still, HIV-1 can traverse the BBB endothelium through an Env-independent pathway, for which we cannot completely rule out a paracellular transit of a fraction of HIV-1 virions. However, confocal microscopy analyses of green fluorescent protein (GFP)-labeled HIV-1 showed fluorescent virions located in the cytoplasm and scattered across the depth of endothelial cells (Fig. 1F), arguing in favor of viral entry by endocytosis, as previously shown (17). Whether the endothelial cell transcytosis of HIV-1 involves macropinocytosis, interactions with proteoglycans (11), or receptor-mediated endocytosis through mannose-6-phoshate receptors (12), and possibly its coreceptors CCR5 and CXCR4 expressed on hCMEC/3 cells (11,18), or alternative receptors remains, however, to be precisely determined.
We next measured HIV-1 transcytosis in the presence of nonneutralizing antibodies (NnAbs) and bNAbs targeting various gp160 epitopes. Unexpectedly, none of the 13 tested single NnAbs and bNAbs showed inhibitory activities against the transport of NLAD8 and NL4.3 virions across hCMEC/D3 cell monolayers ( Fig. 2A). The neonatal Fc receptor (FcRn) is expressed on BBB endothelial cells (19), but its role in antibody transcytosis across the BBB is still debated (20). FcRn expression on cultured hCMEC/D3 cells is also greatly reduced compared to that on human primary BBB endothelial cells (21). Thus, as HIV-1 virions alone and bound by antibodies exhibited comparable transcytosis rates, the uptake of antibody-virus complexes by hCMEC/D3 cells is unlikely to occur through the FcRn, as previously shown for mucosal epithelial cell transcytosis (22). Since HIV-1 entry into the CNS is thought to result mainly from HIV-1-infected cells migrating through the BBB (the Trojan horse hypothesis) (1), we then used human T lymphoblastic cells chronically infected with HIV-1 NL4.3 viral strain as a source of virions and tested the same panel of antibodies in the in vitro BBB model. As observed with cell-free virus, none of the NnAbs and bNAbs tested as IgG antibodies inhibited the transendothelial transit of cell-associated HIV-1 ( Fig. 2A). However, whether HIV-1 gp160 antibodies have the capacity to limit the translocation of HIV-1-infected cells into the brain is a key point that still remains to be resolved. As CNS-invading HIV-1 subpopulations are mainly transmitted/founder (T/F) viruses (23), we measured the transcytosis of CH058 T/F virions and also found the viral endocytic transport across the BBB endothelium to be unaffected by HIV-1 gp160 NnAbs and bNAbs (Fig. 2B). In this regard, it would be interesting to reproduce these experiments with T/F viruses isolated from the cerebrospinal fluid of HIV-1-infected subjects and for which the neutralization sensitivity to some of the bNAbs tested here has been previously measured (24). Finally, HIV-1 Neutralization at the Blood-Brain Barrier ® we evaluated whether combinations of antibodies recognizing nonoverlapping epitopes would be more effective than single molecules against HIV-1 transcytosis. Neither of the two antibody mixtures composed of 4 bNAbs with and without 2 NnAbs decreased NLAD8 viral transcytosis across hCMEC/D3 endothelial monolayers (Fig. 2C). Similarly, elite neutralizers' serum IgGs immunopurified against trimeric YU2 gp140 glycoproteins to deplete anti-p24 antibodies blocking p24 HIV-1 detection posttranscytosis had no impact on the intra-endothelial cell migration of NLAD8 virions ( Fig. 2D and E).
HIV-1-bNAb complexes can cross the BBB endothelium but lack infectious potential. To investigate the mechanisms of HIV-1 transport in the BBB endothelium in the presence of anti-gp160 antibodies, we performed confocal microscopy experiments using the in vitro BBB model with GFP-labeled HIV-1 viruses alone or incubated with the bNAb 3BNC117, the NnAb 5-25, and the non-HIV-1-targeting control IgG mGO53 (Fig. 3A). As expected, viruses and antibodies were observed inside adherent hCMEC/ D3 cells and found as antibody-bound HIV-1 virions only with anti-gp160 antibodies regardless of their neutralizing potential ( Fig. 2A). Pearson coefficient analyses of the fluorescent objects revealed a significant trend for HIV-1 colocalization with gp160specific antibodies, predominantly with the bNAb 3BNC117 (26% [P Ͻ 0.0001] versus 17% [P ϭ 0.0045] for 5-25), but not with the isotype control (Fig. 3B). In agreement with this, overlapping virus-antibody fluorescent signals were more frequently detected with 3BNC117 than with 5-25 (45% versus 16% [P ϭ 0.0077]) and in less than 1% with the IgG control (Fig. 2B). As for HIV-1 and antibodies alone, virions in complex with 5-25 and 3BNC117 were located in the cytoplasm and distributed across the endothelial cell depth ( Fig. 3A and C).
As we observed intracellular fluorescent clusters of HIV-1 with anti-gp160 IgGs by confocal microscopy, we next determined whether HIV-1 transcytosed alone and in the presence of antibodies was still infectious using TZM-bl reporter cells. Infectivity data normalized for virus input by p24 content revealed that the viral transit through hCMEC/D3 endothelial monolayers substantially decreased HIV-1 infectivity compared to virions that only crossed the membrane without hCMEC/D3 cells, with an average drop of 1.2 to 1.4 log 10 (Fig. 3D). Comparable low but detectable infectivity levels were measured with transcytosed viruses preincubated with anti-gp160 NnAbs (Fig. 3D). However, as opposed to HIV-1 alone and mixed with NnAbs, posttranscytosis infectivity of virions bound by bNAbs reached basal signal levels (Fig. 3D). HIV-1 residual infectivity following BBB intracellular migration was completely inhibited by bNAbs regardless of the targeted neutralizing epitope (Fig. 3D). To estimate the amount of antibodies required to neutralize transcytosed HIV-1, we tested the IgG bNAb 3BNC117 across a broad concentration range. Even if virions were poorly infectious after crossing endothelial cells, we observed a dose-dependent effect of 3BNC117 on viral neutralization posttranscytosis, which was less effective below 1 g/ml (Fig. 3E). Collectively, our data show that although the envelope spike of HIV-1 is partially needed for its transcytosis across the BBB, gp160-specific antibodies, singly or in combination, had no measurable effects on the intracellular transport of the virus in vitro. HIV-1 virions migrating by endocytosis through the BBB endothelium alone or in the presence of NnAbs remained infectious, but at much lower levels than nontranscytosed viruses. Thus, it is not clear whether transcytosed HIV-1 would support the productive infection of susceptible CNS cells in vivo. Alternatively, HIV-1 opsonized by NnAbs could be captured by brain phagocytic cells, such as microglia, and then eliminated by antibody-dependent cellular phagocytosis. Deprived of neutralization capacity, HIV-1-binding antibodies could still promote killing of infected cells crossing the BBB via Fc effector functions. In this regard, HIV-1 NnAbs in the cerebrospinal fluid of infected individuals have been shown to induce the destruction of cells infected by a laboratory-adapted strain by antibody-dependent cellular cytotoxicity (ADCC) (25). However, while lab-adapted strains are commonly sensitive to NnAb-mediated ADCC, primary and T/F viruses are generally resistant (26,27). On the other hand, we demonstrate in this study that gp160 bNAbs are able to fully neutralize cell-free HIV-1 crossing the BBB. We and others have previously found that the neutralizing activity of anti-gp160 antibodies is essential in protecting target cells from being infected by HIV-1 translocating across mucosal barriers (22,28,29). Hence, we propose that the neutralization ability of gp160 antibodies is also a key factor to protect the CNS from HIV-1 infection and spread. Since bNAbs are also potent inducers of Fc-dependent antiviral activities such as ADCC (30,31), they are the most suitable HIV-1 antibodies for preventing viral brain invasion.
HIV-1 endothelial cell transcytosis assay. hCMEC/D3 cell monolayer transwells with TEER values from 150 to 300 ⍀/cm 2 8 days postculture were used for the transcytosis assay. HIV-1 viruses (5 ng of p24 HIV-1 cell-free NLAD8, NL4.3, or NL-ΔEnv virions in 160 l [final volume]) were incubated for 1 h at 37°C with purified recombinant antibodies (66.67 nM final concentration, unless specified otherwise), antibody cocktails, or gp140-immunopurified IgG fractions (6.67 nM final concentration), and mixtures were added to the cell monolayers. In each experiment, HIV-1 virions were also added to inserts with and without hCMEC/D3 cells. After 4 h of incubation at 37°C (or 16 h prior to infectivity testing), media were collected from both upper and lower compartments. Each condition was tested in triplicate or quadruplicate. HIV-1 p24 amount was determined for each transwell with an HIV-1 p24 antigen capture assay (Advanced Bioscience Laboratories, Rockville, MD), using two different dilutions of the basal medium. The percentage of transcytosis as normalized percentage of input was calculated following the formula ([p24] sample /[p24] Ctl ) ϫ 100 (where Ctl is control).
In vitro HIV-1 infectivity assay. Infectivity of transcytosed HIV-1 virions in the presence or absence of anti-gp160 or isotypic control IgG antibodies was measured using the TZM-bl cell assay. TZM-bl reporter cells (1 ϫ 10 4 per well) in full DMEM containing 8 g/ml of DEAE-dextran were incubated with 200 l of transcytosed viruses recovered from the basal medium for 48 h at 37°C. In each experiment, wells from transcytosis experiments were tested in triplicate and in parallel with 12.5 mM nevirapine and NLAD8 virus alone (0.05 ng) as controls and incubated with 20 g/ml of 10-1074 IgG1. After 48 h, cells were lysed, and the assay was developed with the Bright-Glo luciferase assay reagent (Promega). Luminescence signal was measured as relative light units (RLU) using the Enspire microplate luminometer (Perkin Elmer). Normalized infectivity per well was calculated by dividing the mean number of RLU by the p24 concentration (in nanograms per milliliter).
Statistics. Percentages of viral transcytosis and posttranscytosis infectivity levels were compared between antibody groups to the "no antibody" control group using Dunn's multiple-comparison test. Percentages of colocalization and overlapping fluorescent events measured by confocal microscopy were compared across groups of antibodies using the Kruskal-Wallis test and the post hoc Dunn's multiple-comparison test and using the Mann-Whitney test, respectively. Statistical analyses were performed using GraphPad Prism software (v8.1.2, GraphPad Prism Inc.).