|Home | About | Journals | Submit | Contact Us | Français|
Despite tremendous efforts, development of an effective vaccine against HIV has proved an elusive goal. Recently, however, numerous antibodies have been identified that are capable of neutralizing the vast majority of circulating HIV strains1–5. These antibodies all exhibit an unusually high level of somatic mutation6, presumably due to extensive affinity maturation over the course of continuous exposure to an evolving antigen7. While substantial effort has focused on the design of immunogens capable of eliciting antibodies de novo that would target similar epitopes8–10, it remains uncertain whether a conventional vaccine will be able to elicit analogs of the existing broadly neutralizing antibodies. As an alternative to immunization, vector-mediated gene transfer could be used to engineer secretion of the existing broadly neutralizing antibodies into the circulation. Here we describe a practical implementation of this approach, vectored immunoprophylaxis (VIP), which in mice induces lifelong expression of these monoclonal antibodies at high concentrations from a single intramuscular injection. This is achieved using a specialized adeno-associated virus (AAV) vector optimized for the production of full-length antibody from muscle tissue. We show that humanized mice receiving VIP appear to be fully protected from HIV infection even when challenged intravenously with very high doses of replication-competent virus. Our results suggest that successful translation of this approach to humans may produce effective prophylaxis against HIV.
Previous efforts to engineer humoral immunity using AAV-based vectors resulted in modest antibody production11 which was subsequently improved through the use of alternative capsids12 and self-complementary AAV (scAAV) vectors13 that increase expression at the expense of carrying capacity. Recently, scAAV vectors were employed to direct expression of SIV-neutralizing immunoadhesins consisting of small, artificially fused antibody fragments14. However, the efficacy of this prophylaxis was limited by an endogenous immune response directed against the immunoadhesin proteins. To ask whether newer capsid serotypes and vector configurations might support long-lived expression of full-length human antibodies from muscle, we produced AAV vectors with the capsid from serotype 815 that expressed either luciferase or 4E10 HIV neutralizing antibody driven from CMV promoters and administered them through a single injection of the gastrocnemius muscle (Fig. 1a). Within one week of vector administration, either luciferase or antibody gene expression was detectable (Supplementary Fig. 1a, left and right respectively). Expression continued to rise, achieving maximum levels after 12–16 weeks and then decreasing two- to three-fold before stabilizing for the duration of the 64-week study. Given the long-lived nature of this expression, it seemed possible that these vectors could be used to engineer lifelong humoral immunity provided by full-length, fully human antibodies. Hence, we carried out a systematic process of vector and transgene optimization to improve the expression characteristics of this system (Supplementary Information). The heavy and light chain variable regions of the HIV-neutralizing b12 antibody were cloned into the vector and AAV stock was produced for intramuscular administration of 1×1011 genome copies (GC) into the gastrocnemius muscle of two immunodeficient and two immunocompetent mouse strains: NOD/SCID/γc (NSG), Rag2/γc (RAG), B6, and Balb/C. Mice produced the encoded antibody at serum concentrations that were 100-fold higher than the levels achieved with the non-optimized vector and this level of expression persisted for at least 52 weeks (Fig. 1b compared to Supplementary Fig. 1a, right). In agreement with previous studies of AAV-induced tolerance in mice16, we detected very limited mouse antibodies raised against human b12-IgG in B6 mice while Balb/C animals generated detectable mouse antibodies against the transgene (data not shown) that did not appear to impact human IgG levels.
To test the ability of VIP to protect mice from challenge in vivo, we adapted a previously described humanized mouse model17 that exhibits CD4 cell depletion following challenge with replication-competent HIV (Supplementary Fig. 5). We administered vector expressing either luciferase or b12 antibody to NSG mice, producing stable serum b12 antibody concentrations of approximately 100µg/mL within six weeks (Fig. 1c). These mice were adoptively populated with expanded human peripheral blood mononuclear cells (huPBMCs), which engrafted over a period of two weeks. Mice were then challenged by intraperitoneal (IP) injection of the NL4-3 strain of HIV. Following HIV challenge, the majority of mice expressing luciferase showed dramatic loss of CD4 cells while mice expressing b12 antibody showed no CD4-cell depletion (Fig 1d).
To compare the protective abilities of the historically available broadly neutralizing antibodies, vectors expressing b12, 2G12, 4E10, and 2F5 were produced and administered to NSG mice. Seven weeks after administration, NSG mice produced 20–250µg/mL of the indicated antibodies (Fig. 2a). Interestingly, in vivo serum concentrations of 4E10 and 2F5 were somewhat lower than b12 and 2G12 despite comparable expression in vitro (Supplementary Figure 3b), perhaps resulting from the previously described self-reactivity of these clones18. Transduced mice were adoptively populated with huPBMCs, challenged by intravenous (IV) injection with HIV, and sampled weekly to quantify CD4 cell depletion over time (Fig. 2b). Animals expressing b12 were completely protected from infection, while those expressing 2G12, 4E10 and 2F5 were partially protected. Groups demonstrating partial protection consisted of animals with delayed CD4 cell depletion as well as animals that maintained high CD4 cell levels throughout the course of the experiment. Interestingly, mice expressing 250µg/mL of the 2G12 antibody were only partially protected despite antibody levels being over 300-fold higher than previously established IC50 values for this antibody-strain combination in vitro19. Eight weeks post-challenge, mice were sacrificed and paraffin-embedded spleen sections underwent immunohistochemical staining for the HIV-expressed p24 antigen to quantify the extent of infection (Fig. 2c). Remarkably, mice expressing b12 had no detectable p24 expressing cells while those expressing other antibodies exhibited significant positive staining (Fig. 2d).
To determine the robustness of protection mediated by VIP, a large cohort of mice expressing b12 antibody were adoptively populated with huPBMCs. Prior to challenge, all mice expressed high levels of human IgG, presumably due to engrafted human B-cells (Supplementary Fig. 6a), but only those receiving the b12-expressing vector produced IgG specific for gp120, which reached 100µg/mL (Supplementary Fig. 6b). Mock-infected mice expressing either luciferase or b12 demonstrated consistent high-level CD4 cell engraftment throughout the course of the experiment, showing that transgene toxicity was not contributing to CD4 cell loss (Fig. 3). In contrast, mice expressing luciferase that received 1ng of HIV experienced rapid and extensive CD4 cell depletion. At higher doses, infection in luciferase-expressing mice became more consistent and resulted in depletion of CD4 cells below the level of detection in some cases (25, 125ng doses). Remarkably, all mice expressing b12 demonstrated protection from CD4 cell loss, despite receiving HIV doses over 100-fold higher than necessary to deplete seven out of eight control animals (Fig. 3).
As newer anti-HIV antibodies have become available, we have compared the relative efficacy of b12 to VRC01 antibody. VRC01 neutralizes over 90% of circulating HIV strains in vitro1, making it an excellent candidate for human trials. We administered decreasing doses of vector expressing either b12 or VRC01 to NSG mice and monitored expression of the antibodies over time. For both antibodies, we observed clear dose-dependent expression at all time points analyzed (Supplementary Fig. 7a and Fig. 4 (left)). Mice expressing luciferase or antibodies at various levels were adoptively populated with huPBMCs. Just prior to challenge, a gp120-specific ELISA confirmed the effective antibody concentration in each group (Supplementary Fig. 7b and Fig. 4 (middle)). Following IV challenge with 10ng of HIV, CD4 cells were monitored to determine the impact of antibody concentration. An average b12 concentration of 34µg/mL and VRC01 concentration of 8.3µg/mL protected mice from infection (Supplementary Fig. 7c and Fig. 4 (right)). Groups expressing lower concentrations of b12 and VRC01 were only partially protected, with several animals showing no detectable loss of CD4 cells and others exhibiting delayed CD4 cell depletion.
Here we demonstrate that broadly neutralizing human antibodies expressed by VIP are capable of protecting animals from even high-dose HIV exposure in vivo. Human-to-human HIV transmission rates vary with behavior but do not generally exceed one per hundred heterosexual exposures20, and recent studies have demonstrated that infections are generally initiated by a single founder strain21. Humanized mice with b12 serum concentrations of 100µg/mL were resistant to HIV infection at challenge doses 100-fold higher than necessary to infect the vast majority of animals, suggesting that this level of protection may far exceed what would be necessary to provide protection against HIV infection in humans. In contrast to previous approaches, VIP produces full-length antibodies that are identical in sequence to those produced by the immune system22. Recent experiments have suggested that full-length antibody structures possess superior in vitro neutralization activity as compared to modified architectures such as immunoadhesins23. Utilization of such naturally occurring antibody architectures should also reduce immune responses against the transgene, which were previously shown to reduce the effectiveness of prophylaxis against SIV14.
Our results demonstrate that VIP administration results in long-lived production of human antibodies at super-prophylactic levels in immunocompetent animals. Clinical trials involving AAV have demonstrated remarkable success when targeting immunoprivileged sites such as retinal tissue24, but transduction of liver resulted in an adaptive immune response against vector capsid25. Studies in non-human primates have shown that the elicitation of capsid-specific cytotoxic T-lymphocytes is limited to AAV capsids that exhibit heparin-binding activity26. Interestingly, serotypes lacking heparin-binding activity, including AAV8, did not induce CTL responses, suggesting that AAV8-based vectors, like the one we have used, may circumvent previously observed immunological barriers to long-term transduction. Additionally, in contrast to liver transduction, administration of AAV via intramuscular injection has been shown to result in very long-lived, albeit low-level, Factor IX expression27, suggesting that the route of administration can significantly impact the duration of expression. While the expression level attainable in humans remains to be determined, it is worth noting that the significantly longer serum half-life of a human antibody in humans28, as opposed to mice29, may result in higher steady-state levels than those observed in the present study. Regardless of this, our results suggest that even if VIP administration in humans results in serum antibody concentrations 100-fold lower than those observed in mice, it may still confer protection against HIV infection.
Given the urgency that exists in combating the ongoing global HIV pandemic and the incremental progress towards a vaccine, novel paradigms of prophylaxis must be explored towards solving this global health crisis. Our work demonstrates the feasibility of directly translating the existing repertoire of broadly neutralizing antibodies into functional immunoprophylaxis with robust protective abilities in vivo. As more potent broadly neutralizing HIV antibodies are isolated30, VIP can deliver these in concert with existing antibodies to provide increased potency, broader coverage, and greater resistance to escape mutations. This approach may find broad utility in the rapid development of effective prophylaxis against any existing or future infectious disease for which broadly neutralizing antibodies can be isolated. Beyond infectious diseases, VIP can be applied to therapeutic regimens in which continuous production of monoclonal antibodies in vivo is desirable. Given the level of protection that VIP has demonstrated in vivo, we believe that highly effective prophylaxis through expression of existing monoclonal antibodies against HIV in humans is achievable.
AAV2/8 was produced by transient transfection and purification from culture supernatant by PEG precipitation and cesium chloride ultracentrifugation. Virus was quantified by qPCR against CMV sequences and functionally validated in vitro to confirm gene expression prior to use in vivo. Mice were given single injections with purified vector in the gastrocnemius muscle. Antibody concentration in the serum was determined using an ELISA specific for either total human IgG or human IgG against HIV-gp120. Humanized mice expressing antibodies were produced by adoptive transfer of expanded huPBMCs into mice previously transduced with AAV vectors. HIV challenge was carried out via IP or IV injection and blood was sampled weekly to determine the ratio of CD4 to CD8 cells by flow cytometry.
We thank James Wilson (University of Pennsylvania) for AAV8-related plasmids and assistance, Dennis Burton (Scripps Research Institute) for b12 and 2G12 expression plasmids, Gary Nabel (Vaccine Research Center) for 4E10, 2F5 and VRC01 expression plasmids, and the Caltech Protein Expression Center for providing purified antibodies. The following reagents were obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH: pNL4-3 from Dr. Malcolm Martin and TZM-bl cells from Dr. John Kappes and Dr. Xiaoyun Wu. The authors wish to thank Jesse Bloom, Darrell Kotton, Devdoot Majumdar, Gustavo Mostoslavsky, Ryan O’Connell and Alex Sigal for their helpful comments, and other members of the Baltimore lab for their assistance in carrying out this work. This project was supported by the Bill and Melinda Gates Foundation (Seattle, WA) through the Grand Challenges in Global Health Initiative (awarded to D.B.) Grand Challenge Grant #37866 and by the National Institutes of Health (HHSN266200500035C) through a contract from the National Institute of Allergy and Infectious Disease (NIAID). A.B.B. is supported by the amfAR postdoctoral research fellowship #107756-47-RFVA. D.S.R. is supported by career development award 1K08CA133521 from the NIH.
Author ContributionsA.B.B. and D.B. conceived the study with assistance from L.Y., A.B.B. designed the experiments. A.B.B., J.C., and C.M.H. carried out experiments. A.B.B., J.C., and C.M.H. analyzed the data. D.S.R. performed immunohistochemistry and analysis. A.B.B. and D.B. wrote the paper with contributions from all authors.
Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this article at www.nature.com/nature.