The challenges of developing a vaccine to induce immunity against HIV have become increasingly apparent, prompting efforts to develop an alternative strategy in which patients would be injected with a life-long source of broadly neutralizing anti-HIV antibodies. In a recent paper in Nature, Balazs et al.1 describe the first successful preclinical application of this approach. The authors show that mice injected with a viral vector encoding monoclonal antibodies that neutralize multiple HIV strains are protected to varying degrees when subsequently challenged with a strain of HIV. Previous work has shown that direct administration of anti-HIV monoclonal antibodies could not provide long-term sustained levels of antibodies without repeated injections. In contrast, delivery of the gene encoding the antibodies through the use of the vectors provides a potential means of long-term in situ production of the antibodies.

HIV mutates rapidly, allowing it to escape both antibodies and cellular responses and making vaccines difficult to develop. Moreover, among all HIV proteins, the envelope protein, which is the primary target of antibodies, is the most genetically variable between different virus strains. As a result, most of the anti-HIV antibodies that have been characterized are quite strain specific, and large searches have yielded only a few broadly neutralizing antibodies. These broadly neutralizing antibodies arose, by extensive somatic mutation of the immunoglobulin genes, against viruses that were continuously evolving to escape immune responses2, which may mean that it would be difficult to elicit such antibodies by a vaccine. Moreover, broadly neutralizing antibodies have variable potency, with most effective against no more than 40–50% of viral clades.

An alternative to immunization, passive delivery of broadly neutralizing antibodies, circumvents the need to design an appropriate immunogen and shortens the time needed for the development of these broadly neutralizing antibodies, which appear to require hypermutation of the genes encoding the binding site of the antibodies. Monoclonal antibodies can be delivered directly into the bloodstream by infusion, as is done for cancer therapy, but this is impractical for HIV in a prophylactic setting. Therefore, viral vectors have been developed as vehicles to provide a continuous supply of neutralizing antibodies. These efforts have not succeeded for HIV for various reasons, including insufficient transduction by the vector, low levels of transgene expression and immune responses against the effector protein.

Balazs et al.1 injected mice intramuscularly with a recombinant adeno-associated virus (rAAV) vector encoding one of several different monoclonal antibodies, b12, 2G12, 4E10, 2F5 and VRC01, previously shown to have broadly neutralizing activity against multiple strains of HIV. The vector is injected into muscle which then produces the neutralizing antibody, which is secreted into the bloodstream. This work extends the earlier development of an AAV-8 vector that produced a monoclonal antibody neutralizing VEGFR2 (ref. 3) and an AAVrh.10 vector encoding an anti-HER2 antibody4, both of which could inhibit tumor growth in mice. Balazs et al.1 tested a variety of alterations to the rAAV vector in order to optimize the muscle production of the monoclonal antibodies. They made a new hybrid promoter using portions of three promoters to ensure long-term muscle expression, added a posttranscriptional regulatory element to further increase the production of the transgene, and tested various transcriptional terminators. For the specific production of the antibodies, they utilized a foot-and-mouth disease virus F2A peptide cleavage sequence engineered with a furin cleavage site that had previously been shown3 to greatly increase the yields of antibodies; additionally, they employed a human growth hormone signal sequence, and removed all potential splice donor and acceptor sequences.

Balazs et al.1 humanized the mice by injecting them with human CD4+ cells, which can then be infected by HIV. They then challenged the mice intravenously or intraperitoneally with the NL4-3 strain of HIV and assessed the protective effectiveness of the circulating broadly neutralizing antibodies either by measuring HIV p24, which, because it is an HIV protein, indicates the extent to which the virus has replicated, or by the depletion of CD4 cells. The mice were protected from HIV infection to various degrees depending upon which neutralizing antibody they received (Fig. 1).

Figure 1: Vectored immunoprophylaxis against HIV.
figure 1

Delivery of a rAAV vector encoding a broadly neutralizing antibody circumvents the difficult tasks of discovering and producing the appropriate antigen and speeding up the immmune system's production of these antibodies. Recombinant AAV vectors (rAAV) encoding one of several monoclonal neutralizing antibodies are injected into the muscle of immunodeficient mice. The mice are then 'humanized' by injection of human peripheral blood mononuclear cells, followed by challenge with the NL4-3 strain of HIV. Different antibodies provide varying degrees of protection as indicated by less loss of CD4+ T cells.

Passive admininstration of broadly neutralizing antibodies has two distinct advantages compared to a vaccine. First, it circumvents the need to discover and formulate antigens capable of inducing the right (i.e., broadly-neutralizing) antibodies. Second, it shortens the time required by the immune system to produce the antibodies, which in the case of the broadly neutralizing anti-HIV antibodies discovered to date, is quite a significant benefit because of the extensive somatic hypermutation of the immunoglobulin genes required.

But several open questions must be investigated before the promising results of Balazs et al.1 can be applied in humans. The relevance to humans of all preclinical models for testing HIV vaccines and immune interventions is unknown. Results based on protection against intravenous or intraperitoneal virus challenge may not predict efficacy for human-to-human transmission, which, except for intravenous drug users, is through mucosal surfaces. Challenge viruses used in experimental laboratory systems, even those considered to be primary isolates of HIV, likely differ from the wild-type viruses, particularly founder transmitting viruses, that infect humans. Finally, some infection occurs not by free virus, but by virus inside of cells5,6 and thus antibodies alone might not protect against infection that occurs via cell-to-cell virological synapses. It would be interesting to know what would happen upon challenge with other strains of virus or upon repeat challenge of the mice with either the same or a different strain of HIV, as these are the normal patterns for human exposure.

Broadly neutralizing antibodies themselves also require further study. Although they have been shown to neutralize various isolates, they arose in individuals who had been infected for years, so it is not clear whether their in vivo potency is sufficient for prophylaxis of HIV. They may not be effective against strains of virus that escape initial neutralization and mutate; knowing if and how HIV mutates in the presence of these broadly-neutralizing antibodies will help predict their utility as therapeutic agents for individuals already infected with HIV. Given that the efficacies of the monoclonal antibodies tested did not necessarily correlate with their in vitro potency, it is clear that we do not yet understand all the factors that determine the in vivo effectiveness of antibodies.

Despite the recent successful use of a rAAV vector for therapy of hemophilia7, the safety of rAAV vectors for prophylaxis in healthy individuals requires a closer examination of some issues. For example, despite the tropism of rAAV vectors for hepatocytes and muscle, they may transduce other tissues, even when injected into muscle. AAV induces innate8 as well as vector-specific immune responses9 (even for an AAV capsid that does not exhibit heparin-binding activity10), which means that side effects from inflammatory responses and possibly decreased immunoglobulin production over time will need to be fully evaluated. Furthermore, immune responses can be produced against the antibody (anti-idiotype antibodies), or against the peptide fragment from the foot-and-mouth disease virus F2A self-processing sequence, which although ultimately cleaved from the antibody, is expressed as an initial part of the antibody construct. Lastly, although preclinical tests have shown no evidence of any integration of rAAV vectors into the germline11, further studies may be needed prior to a prophylactic rather than a therapeutic use.

Whether or not rAAV vectors are able to protect against HIV infection, they could be useful for limiting the viral loads in infected individuals. They represent a new tool for studying the effects of one or more broadly neutralizing antibodies on viral mutation. They may help us understand the differences between in vitro neutralization capability and in vivo effectiveness. The vectors developed by Balazs et al.1 may also be useful for studying the differences between the founder viruses that initiate an HIV infection and subsequent mutated virus. The insights provided by such studies could help guide the development of effective antigens for HIV vaccine development. Finally, the results of Balazs et al.1 support the hope of developing a passive prophylactic or therapeutic approach for other infectious diseases, such as cytomegalovirus infection in organ transplant patients.