Opening the Envelope
The premier gathering of HIV vaccine researchers showcased a healthy dose of progress in understanding HIV’s structure and how it can be utilized to engineer better vaccine candidates
By Yegor Voronin and Noah Sather*
“Three atomic-level structures of HIV envelope were released over the past two years,” said Peter Kwong of the Vaccine Research Center at the National Institute of Allergy and Infectious Diseases (NIAID). “At this meeting, I’ve already seen six new trimer structures, and the meeting is barely half over.”
The meeting was the Keystone Symposium on HIV Vaccines, which took place March 22-27 in Banff, Canada, and, for the record, Peter counted a total of 10 new HIV envelope (Env) structures that were unveiled there, reflecting the rapidly accelerating pace of progress in defining the structure of HIV’s outer surface protein. This sentiment extended to the whole meeting, as several directions of HIV vaccine research, none of which are completely new, showed spectacular progress in the recent months.
Epitope mapping reveals surprises
New structures of Env illuminate how the genetic diversity of HIV results in the diversity of protein structures, which allows the virus to efficiently evade the immune system. They also reveal the conserved motifs that are targeted by antibodies capable of neutralizing a broad swath of HIV isolates (so-called broadly neutralizing antibodies or bNAbs). At Keystone, Pamela Bjorkman from the California Institute of Technology presented data showing that an antibody referred to as 8ANC195 binds to HIV Env in a manner different from all previously identified bNAbs. Although initial analysis showed that 8ANC195 competes with antibodies targeting the CD4 binding site (CD4bs) on Env, crystal structures revealed that it actually binds to a distinct region nearby and that its epitope spans the gp120 and gp41 subunits of Env.
How bNAbs bind is also important. The antibody N6, which was discovered in Mark Connors’ laboratory at NIAID and presented on at the meeting by Jinghe Huang, binds to the well-known CD4bs epitope; however, the way this antibody binds is unique. The heavy chain of N6 interacts with the epitope in a very flexible manner that tolerates viral variability and the light chain moves aside, allowing the heavy chain to bind in a unobstructed way that increases the antibody’s neutralization breadth and potency.
As multiple research groups continue to isolate bNAbs and decipher how they bind to HIV, the view of the Env protein is radically transforming. Previously researchers believed that there were just three or four specific sites on the surface of Env targeted by bNAbs in a very specific manner. But it turns out the vast majority of the Env surface is targeted by one bNAb or another (see Figure, below). This is great news for researchers who are attempting to elicit bNAbs via vaccination because it provides multiple options for designing immunogens. The challenge now is focusing the immune response on these conserved regions and not the variable parts of the virus.
The model of HIV Envelope, showing in various colors footprints of broadly neutralizing antibodies (bNAbs) on one of the monomers in the trimer (left image, side view; right image, top view). As researchers discover more bNAbs, new sites of vulnerability are being discovered. Figure courtesy of Gabriel Ozorowski of The Scripps Research Institute.
From structure to immunogen
A better understanding of Env’s structure dovetails with efforts to create better recombinant immunogens that mimic the trimeric shape in which Env exists on the surface of viral particles. Traditional approaches to create soluble Env trimers resulted in immunogens that had little resemblance to the native structure. Over 15 years ago, John Moore and colleagues began working on an approach to create an immunogen that more faithfully reproduces the native trimer. They preserved the natural cleavage site between the gp120 and gp41 subunits of HIV Env, but introduced mutations that generally stabilized the trimeric structure and linked the monomers with di-sulfide bonds. The resulting protein is referred to as BG505 SOSIP.664 (see CROI: Progress on Prevention and Cure, IAVI Report, Vol. 18, Issue 1, 2014). Because this method of trimerization creates Env trimers that adopt native-like conformation, it enabled elucidation of the many trimer structures seen at this year’s meeting and is leading to unprecedented insight into Env’s structure.
Until recently, only a very small number of Env proteins were sufficiently stabilized using the SOSIP approach, but as reported by Gordon Joyce of NIAID, this is no longer the case. Joyce and colleagues applied a design approach based on the SOSIP trimerization method to approximately 160 different Envs from various clades and identified approximately 40 that formed stable native-like trimers. Most of the new trimers are clade C, but clades A and B were also represented in the panel. These new SOSIP trimers give a more broad representation of the global Env diversity and have the potential to provide valuable insights into the structural differences among the major clades of the virus and how these differences are overcome by bNAbs.
Somewhat ironically, and indicative of how rapidly the field is progressing, just when the SOSIP approach begins to gain momentum and bear fruit, some researchers are already working on various ways to improve the SOSIP technology. As explained by Rich Wyatt of IAVI’s Neutralizing Antibody Center at The Scripps Research Institute (TSRI), one promising approach is to create native flexibly-linked (NFL) trimers, in which the cleavage site between gp120 and gp41 subunits is replaced by a flexible amino acid linker and the trimer structure is stabilized by modification of a small number of critical amino acid residues. Javier Guenaga, a member of Wyatt’s laboratory, reported on recent progress in optimization of the amino acid sequence of Env to promote formation of stable NFL trimers. The resulting trimer was shown by negative stain electron microscopy (EM) studies to adopt a well-ordered, native-like structure and was found to bind only bNAbs and not the undesirable narrowly neutralizing antibodies. Although they are still in the early developmental phase, NFL trimers are thought to have several desirable attributes—this approach has been successfully applied to Envs from both clades A and B, simplifies production methods, and increases yields during manufacturing.
Regardless of whether they were created by SOSIP or NFL technology, the native-like trimers have a number of potential uses. They can be complexed with bNAbs to study the fine details of Env-antibody interactions using crystallography or EM, or be used as reagents to fish out antibodies that specifically recognize the native Env conformation. Some researchers also expect that they will be vastly improved immunogens, presenting to the immune system a biologically relevant target that is even more stable than the protein on the surface of the virion. So far immunogenicity data have only been presented for the BG505 SOSIP. Joyce Hu from the laboratory of Shane Crotty at the La Jolla Institute for Allergy and Immunology, reported that BG505 was poorly immunogenic in mice, generating antibodies able to neutralize tier-1 viruses (the easiest viruses to neutralize), but not the harder to neutralize tier-2 viruses. This disappointing result may be specific to mice, which have a B-cell receptor (BCR) repertoire that appears to have difficulty targeting highly-glycosylated epitopes.
Data presented by Rogier Sanders of the Weill Medical College of Cornell University showed that immunization of rabbits and macaques with BG505 SOSIP trimers resulted in potent neutralization of tier-2 autologous virus (virus that bears the un-modified BG505 Env that was used to create the SOSIP trimer). Sanders says that previous vaccine approaches have not induced meaningful neutralization against tier-2 viruses, what he refers to as “a level of neutralization that would protect against a robust virus challenge.” He also stressed that the appearance of homologous neutralization is a promising first step, which needs to be followed by additional immunizations designed to increase the breadth of the response.
One of the major findings from the research on the appearance of bNAbs in HIV-infected individuals is that most bNAbs go through an unusually large number of rounds of mutation and selection in germinal centers. As a result of this process, the antibodies become quite different from the so-called unmutated common ancestor (UCA) or germline antibody. Recently several groups showed that previous Env immunogens bound the germline forms of bNAbs very poorly, possibly explaining why vaccine strategies to date haven’t induced neutralizing antibodies with broad specificity. Some Env proteins will bind the germline versions of bNAbs if certain N-linked glycosylation sites are altered (J. Exp. Med. 210 (4), 655, 2013), but this was limited to mostly gp140 fused Env trimers and engineered outer domain (eOD) Env proteins, and was only observed for germline versions of VRC01-class bNAbs.
At Keystone, J.M. Medina-Ramirez, a member of Sanders’ other laboratory at the University of Amsterdam, presented data on the design and characterization of a native-like trimer that is able to engage germline antibodies. Using knowledge gained from antibody function studies and crystal structure analyses, Medina-Ramirez and colleagues selected 18 amino acid changes that were predicted to allow the well-characterized BG505 SOSIP.664 trimer to bind germline antibodies. This modified trimer engaged multiple bNAb germline antibodies, including precursors of four different bNAbs that target the V1/V2 loops and three that target the CD4bs region of Env. Additionally, they confirmed that this germline-optimized BG505 SOSIP.664 was able to stimulate the germline version of the anti-CD4bs bNAb VRC01 in its membrane-bound form (as a BCR) similar to what would be displayed on the surface of a B cell. Stimulating the germline BCR is a critical first step for antibody development. Medina-Ramirez says this is therefore a suitable protein immunogen for trying to initiate the first steps in eliciting bNAbs by vaccination.
Leonidas Stamatatos, one of the meeting’s organizers, also reported progress in the design of germline-optimized immunogens from his work at the Fred Hutchinson Cancer Research Center. Previously his group created a more classical Env construct (a fused trimeric gp140), which despite not being a native-like trimer, was capable of binding the germline version of VRC01. Now, he’s developed highly modified versions of this construct in which a portion of the variable loops (V1-V3) was removed and targeted mutations were introduced to expand its binding to almost all known germline VRC01 antibodies. Stamatatos highlighted that interaction between a BCR created from one such germline precursor and the germline-targeting immunogen that they developed was shown in a transgenic mouse, a more relevant model than in vitro studies.
Bart Haynes of the Duke University Medical Center discussed another germline targeting approach—the idea of mimicking the natural process of bNAb development. In contrast to rational engineering of Env immunogens with the goal of optimizing binding to germline bNAb precursors, this approach relies on identifying the naturally occurring HIV Envs to stimulate development of bNAbs. This approach is feasible because of the detailed investigation of co-evolution of antibodies and viruses in HIV-infected individuals that develop bNAbs. Haynes and colleagues identified the antibody germline precursor stimulated during natural infection and also identified the stepwise mutational changes that the initial germline version underwent during the affinity maturation process. They also tracked the corresponding changes that occurred in Env sequences as the virus escaped the antibody responses. This detailed examination allows identification of the variants of Env that drive antibody maturation from the initial ancestor, through various intermediates, to the final bNAb. By creating immunogens that correspond to these key Env variants, Haynes and colleagues aim to recapitulate the natural development of bNAbs via vaccination.
Efforts to test these immunization strategies in animal models are already underway and some of the first results in rhesus macaques were presented at the conference. Haynes and Larry Liao from the Duke University Medical Center presented immunization studies with Env variants from a volunteer that eventually developed antibody CH103, which targets the CD4bs. The virus that established infection in this volunteer carried the Env that was able to bind the germline version of the CH103 antibody. In addition to that Env, researchers also created immunogens based on Envs found at weeks 52, 78, and 100, because their analysis suggested that those Envs played an important role in driving the development of the CH103 bNAb. They tested the Env from the infecting virus on its own and in combination with the later versions of Env, which were either combined as a mixture or administered sequentially one after the other. All three approaches resulted in appearance of antibody intermediates at initial stages of CD4bs bNAb lineages, which was an encouraging finding. In macaques immunized sequentially with these Envs, neutralization assays showed appearance of more potent and broad antibodies than in the other two groups, validating the overall approach of attempting to mimic the natural evolution of antibody maturation. However, neutralization titers in plasma were not very high and, therefore, more work is needed to improve upon this initial success.
Joe Jardine, a member of William Schief’s laboratory at TSRI, presented another exciting study involving germline targeting immunization strategies. In an effort to target the germline precursors of VRC01-class antibodies, they have developed the eOD-GT8 60mer—a self-assembling nanoparticle carrying an engineered form of the outer domain of gp120. This protein particle is able to bind germline antibodies that express VH1-2*02, the heavy chain gene that is predominantly used by the human immune system to generate VRC01-like antibodies. To directly explore whether this immunogen could drive the evolution of a VRC01-like antibody by vaccination, their collaborator David Nemazee at TSRI created a transgenic mouse carrying just the heavy chain of the germline version of VRC01. The light chain of the germline antibody is not provided, so the antibodies produced in these animals consist of the germline VRC01 heavy chain and various light chains coming from the natural mouse repertoire. After immunization with the eOD-GT8 60mer, two important observations were made. First, a large number of antibodies appeared with light chains that have just five amino acids in the CDR-L3 loop, despite the fact that such short light chains are extremely rare in mice. VRC01-like antibodies require a light chain that has such a short CDR-L3 loop, so this result bodes well for plans to elicit such antibodies in humans. Second, in response to vaccination the heavy chain of the germline antibody was undergoing somatic hypermutation and many of the mutations that appeared were identical to mutations found in VRC01, indicating that antibody maturation followed the same pattern that led to the appearance of VRC01 in the HIV-infected individuals from which it was isolated. While the transgenic mouse system is quite artificial, these results indicate that immunization with the eOD-GT8 60mer can lead to selection of the necessary heavy and light chains and drive antibody evolution. Together with Liao’s findings, these studies strongly suggest that engaging germline versions of antibodies and guiding them towards a desired specificity is a feasible approach.
As immunization studies with native-like trimers and germline-optimized immunogens get underway, investigators are focusing on closely monitoring humoral immune responses to vaccines using recent advances in high-throughput technologies, such as next generation sequencing and B-cell cloning. These approaches allow almost real-time evaluation of vaccines at the sequence population level. The goal of such studies is to gather information that may guide development of better immunogens or immunization strategies.
Gunilla Karlsson-Hedestam of the Karolinska Institutet reported on her laboratory’s progress in defining anti-Env vaccine-specific immune responses in nonhuman primate studies. The researchers begin by using cell sorting techniques to separate B cells expressing anti-Env antibodies. The heavy and the light chain genes from single B cells are cloned and used to produce antibodies that are tested for binding to and neutralization of HIV. Epitope mapping techniques provide information on the regions of Env trimers that were targeted by these antibodies, thus giving the researchers a clear picture of the immunogenic properties of the tested Env. In parallel, they leverage high throughput Illumina next generation sequencing to study heavy chain genes of the entire population of anti-Env B cells. This approach provides minute details on the types of B-cell receptors being stimulated by the vaccine being tested and on the levels of antibody maturation that they are able to achieve, allowing researchers to monitor and evaluate vaccine-induced antibody responses in unprecedented detail.
Yegor Voronin is senior Noah Sather is an assistant
science officer at the professor at the Seattle
Global HIV Vaccine Biomedical Research