Roads Less Taken

Most HIV researchers accept that, to be truly effective, an HIV vaccine will have to elicit broadly neutralizing antibodies. What they don’t necessarily agree on is how best to elicit that coveted response

By Regina McEnery

Viral vaccines exert their effects mainly by provoking the immune system to make neutralizing antibodies against targeted pathogens (Cold Spring Harb. Perspect. Med. 1, a007278, 2011). But eliciting such responses to HIV is a challenge that has befuddled vaccine designers for decades. This is because the virus has evolved many mechanisms to evade immune targeting. Chief among these is the remarkable mutability, instability, and structural dynamism of the sole antibody target on its surface—the Envelope trimer, or “spike”—which HIV uses to invade the CD4+ T cell.

Yet with the recent discovery of scores of antibodies that appear to neutralize a broad spectrum of HIV’s circulating variants, researchers have become increasingly optimistic about designing broadly effective HIV vaccines to elicit similar antibodies.

To that end, they have parsed the structure of broadly neutralizing antibodies (bNAbs) and modeled, in atomic detail, the means by which they bind to their epitopes on the Envelope. The hope, of course, is that this structural information will inform the design of immunogens that might elicit similar responses. And, indeed, researchers are making swift progress toward that goal. They are harnessing structural information to systematically manipulate Envelope proteins for use as immunogens and even reverse-engineer chimeric immunogens that mimic the shape, context, and spatial orientation of epitopes bound by bNAbs (see Shaping the BattlefieldIAVI Report, Sep.-Oct. 2012). None of these immunogens is anywhere near clinical evaluation. Still, these types of structure-based approaches to vaccine design seem—with fair reason—to be all the rage today.

But some worry that HIV vaccine designers are focusing their efforts on too narrow a range of antibody targets on the Envelope—for instance, the CD4 binding site that is essential to initiating viral entry, or other more accessible epitopes that are known to be targeted by bNAbs. By and large, their objection isn’t so much that structure-based approaches are necessarily wrong-headed, or that currently popular vaccine targets on the Envelope are going to prove disappointing. It is more that with the attention showered on reductionist approaches and currently popular epitopes, other vaccine strategies and targets are not getting the attention they deserve.

Going native

James Binley, for one, thinks all the excitement about a narrow set of target structures might be a bit premature. A researcher at the Torrey Pines Institute for Molecular Studies in California, Binley suspects that reverse-engineered immunogens might not be sufficiently immunogenic, or even capable of inducing appropriate responses. “The evidence to date suggests that you almost inevitably end up with responses that recognize the immunogen, but that they are off target in terms of being able to recognize the native trimer,” he says. “So the [immunogen] is not going to do the job that you want it to do.”

Binley and his colleagues are, ultimately, as interested as any other vaccine designer in devising immunogens that accurately capture the structure the Envelope must assume to elicit neutralizing antibody responses. But they approach the problem in a different way. They are trying to construct immunogens that more closely resemble the whole and natural version of the HIV spike that the virus uses to infect cells. Their strategy entails displaying native Envelope trimers on virus-like particles, and embedding the Envelope immunogens in a lipid membrane, as they would be in nature.

This approach has its own challenges. A heterotrimer of the glycoproteins gp120 and gp41, the Envelope’s instability and structural dynamism distracts the immune response by offering up a constantly changing cast of structural targets. This is, in fact, what prompted some researchers in the first place to focus on developing engineered immunogens that expose or recreate only small sections of the Envelope as neutralizing epitopes. Others, like Quentin Sattentau of the University of Oxford, have addressed this problem by stabilizing Env, treating it with glutaraldehyde to crosslink certain amino acids and so “fix” the protein in a relatively static structure (seeThe Antibody RaceIAVI Report, Spring 2013).

But Binley doesn’t do that. His premise is that since the native trimer is responsible for infection and is also the target of neutralizing antibodies, it may be in this form that the Envelope is best used as an immunogen. His laboratory has succeeded in making virus-like particles (VLPs) bearing the native trimer that, in rabbits at least, can induce high-titer neutralizing antibodies. So far, serum from one of the vaccinated rabbits has been found to potently neutralize a primary tier 2 isolate, an achievement that Binley considers a milestone in his career.

Binley ventured down the path to his current vaccine design strategy more than a decade ago, when he was a post doc in John Moore’s lab at Weill Cornell Medical College in New York. Binley says it was assumed then that viruses only bore cleaved forms of Envelope on their surfaces. That is, with gp41proteins traversing the viral membrane and the gp120 components arrayed on the outside to form a functional, mushroom-like structure, rather than the uncleaved gp160 monomers. However, his team was accumulating data that suggested otherwise, as were other researchers: the surface of HIV, it turns out, is littered with nonfunctional Envelope proteins. “It was kind of hard getting my brain around breaking that [old] assumption,” says Binley. “We had to re-invent the way we think of the virus.”

By the time he had started his own lab in 2004, Binley and other researchers were trying to understand why the human immune system couldn’t mount a stronger antibody response early on in infection. While it was fairly well established that binding antibodies were a major component of the early response, it was unclear why those antibodies are so ineffectual against the virus. Nor was it obvious how these antibodies differed in their binding from the handful of bNAbs known to science at the time. Complicating the picture was the fact that the non-neutralizing antibody b6 competed with the bNAb b12 for the CD4 binding site of gp120. Yet it did not appear to have any effect on the ability of b12 to neutralize HIV.

It was unclear what this meant. But then Binley’s colleague and visiting scientist Penny Moore, who is now a senior scientist at the National Institute of Communicable Diseases in South Africa, led a study that explained the enigma. Her findings, based on studies using VLPs bearing authentic trimers, suggested that non-neutralizing antibodies tend to focus their binding on nonfunctional Env. This implied that nonfunctional forms of Env might serve as decoys, diverting antibody responses away from epitopes on functional Env that are critical to the virus’s ability to bind and infiltrate its target cells (J. Virol. 80, 2515, 2006).

The “junk” Env—which includes gp120-depleted gp41 stumps and uncleaved gp160 (see figure, below)—could thus undermine neutralizing antibody responses against HIV. Binley, meanwhile, suspected that the junk forms of Env were also obscuring the true potential of native trimers as immunogens, and began working with his team to get rid of the stuff on VLPs. He and his colleagues used proteases to strip nonfunctional Env from VLPs, leaving only native trimers intact on the particles—a preparation his team has named “trimer VLPs” (J. Virol. 85, 5825, 2011; J. Virol. 86, 3574, 2012).


HIV-1 virus-like particles (VLPs) and the effect of protease digests  

VLP surfaces bear native Env trimers and “junk” Env, consisting of uncleaved gp160 and gp41 stumps. This “junk” Env appears to be immunodominant and promotes non-neutralizing responses, perhaps at the expense of neutralizing responses. Protease treatments selectively remove this “junk” Env but the native trimer survives (lower panel), thanks to its compact nature, coupled with its virtually impenetrable glycan shield. The lack of antigenic interference by “junk” Env in VLP immunogens (depicted in the lower panel) may allow a refocusing of antibodies’ attention to the native Env trimer, resulting in more effective neutralizing responses. Image courtesy of Tommy Tong / Torrey Pines Institute for Molecular Studies



“This was our Eureka moment,” Binley says. “We had tried many other approaches, but there was a firm logic behind this one: that the compact, glycan-encrusted native trimer might be able to survive protease treatments more effectively than the more floppy non-functional forms of Env.”

Binley’s team then examined the antigenic properties of trimer VLPs. They found that only neutralizing monoclonal antibodies (mAbs) recognized trimer VLPs and that, for the most part, the digests eliminated the binding of all mAbs to the nonfunctional Env consisting of gp160 and gp41 stumps. They also observed that the trimer VLPs retained the ability to infect cells. Their neutralization sensitivity was largely comparable to the undigested, wild-type VLPs, except that they were 100-fold more sensitive to the membrane proximal external region (MPER) bNAbs 4E10 and Z13e1, which suggested an increased exposure of the gp41 base that these antibodies target.

The Binley team later tested these trimer VLPs in rabbits to see what kind of antibody responses they could generate. Binley says serum taken from one rabbit immunized with trimer VLPs showed ID(50) titers of around 1:1,000 against the JR-FL primary tier 2 HIV-1 strain in the TZM-bl assay. “I think that is pretty tough to get,” says Binley, adding that preliminary data shows the specificity of this activity appears to be quaternary and targets a glycan-sensitive region at the base of the V2 loop.

Binley says the vast majority of vaccine studies do not report tier 2 neutralization in the TZM-bl assay, even against the vaccine-matched isolate, which is probably why the more sensitive A3R5 assay is sometimes favored. Binley says they are now working on strategies to elicit this potent activity more consistently and to try to broaden the scope of neutralization.

Rogier Sanders, a University of Amsterdam scientist who has also been trying to generate stable, deglycosylated Env trimers for immunogenicity studies, says removing “junk Env” from Env vaccine preparations is an excellent idea. “It is well-known that most non-functional forms of Env expose highly immunogenic—immunodominant—decoy epitopes that are likely to distract from broad neutralization epitopes,” says Sanders, who is also affiliated with Weill Cornell Medical College in New York.

Sanders says the induction of potent neutralization against tier 2 JR-FL is a remarkable achievement. “Sporadic neutralization of autologous tier 2 viruses has been seen before, but not with this potency,” says Sanders. “On a cautionary note, the induction of this activity remains sporadic—only one of the rabbits had this neutralization activity. The next bottleneck to tackle is the induction of consistent tier 2 neutralization, even if only autologous.”

Binley’s group is also collaborating with NIAID’s Vaccine Research Center to test the trimer VLPs as baits to isolate new neutralizing mAbs from memory B cells of HIV-infected donors. And they are also in the process of immunizing monkeys with the trimer VLPs to see if neutralizing responses might be easier to induce in this model.

But Binley’s approach is not exactly easy to execute. For one thing, the trimers are extremely difficult to make in the laboratory, says Richard Wyatt, director of viral immunology in IAVI’s Neutralizing Antibody Center in California, who is collaborating with Binley to test his trimers in nonhuman primates.

“You have to use liters [of virus suspensions] to get hundreds of milligrams of the Env on there. It is a lot of work,” he says. “And if they work, then you have to figure that it will be a challenge just to get the GMP material to go into a small Phase 1 trial. Then, even if that works, the question becomes, can you scale it up for millions of doses.”

But Wyatt says the goal right now should be finding the best immunogen for a vaccine candidate. “If you get something that works,” he says, “you may not necessarily be married to that platform. Right now, we don’t have a positive control where we elicit broad neutralization. Once we do, there are probably many more ways to make the production of these [proteins] more efficient.”

The MPER’s domain

If you think trimeric Env is tough, try eliciting viable bNAbs against the MPER region of gp41, a highly conserved site at the stem of the HIV spike. Many HIV vaccine researchers once eyed this region as an attractive target for a universal vaccine because it contains some of the most highly conserved sequences of the HIV genome and plays a crucial role in the fusion of the viral and cellular membranes, a critical step in HIV’s invasion of the cell.

But cryo-electron microscopy (cryo-EM) studies to determine the three-dimensional structure of the pre-fusion state of the Env spike suggest this stretch of the viral spike is exposed only transiently during infection (see IAVI Report blog, Structure of pre-fusion state of the HIV Env trimer determined, Aug. 20, 2012). Further, the MPER domain probably assumes varying conformations, depending on the state of the rest of the envelope. Thus researchers only have a fuzzy notion of the structure of the MPER in the state in which it is available for antibody binding and neutralization.

Still, there’s no escaping the fact that a handful of NAbs, including the bNAbs 2F5, 4E10 and 10E8, are known to target epitopes within the MPER. Trouble is, using this region as bait for antibody responses poses some serious problems. Most troublingly, perhaps, antibodies that target the MPER may derive from autoreactive B-cell clones that would be deleted or made tolerant to self-antigen during B-cell maturation. This hypothesis was proposed about a decade ago in a study led by Duke University scientist Barton Haynes, who is today director of one of the two CHAVI-ID virtual centers that are deeply involved in efforts to elicit bNAbs through vaccination. Haynes and his colleagues found that 4E10 and 2F5 cross-react with cardiolipin, a component of the mitochondrial membrane and a target for antibodies implicated in autoimmune disease (Science 308, 1906, 2005).

Other studies have cast some doubt on this finding (AIDS 21, 2131, 2007 and AIDS 25, 1247, 2011), and the recently discovered bNAb 10E8, one of the most potent isolated so far, does not appear to bind self-antigens. There are, however, other problems. Structural studies suggest, for example, that 4E10 and 2F5 epitopes would be relatively inaccessible to most antibodies because they lie at the base of the HIV spike and are partly embedded in the lipid bilayer.

Researchers tried unsuccessfully to produce AIDS vaccine candidates that mimic the MPER domain using synthetic peptides, or by grafting the MPER sequences onto a protein scaffold, or onto proteins displayed on VLPs. Frustrated by failed experiments, many turned their attention to other possible targets on the Envelope, like elements of its glycan shield, its exposed variable loops and, perhaps most avidly, the CD4 binding site.

Yet some recent evidence suggests that strategies to elicit antibodies to the MPER are not necessarily blind alleys on the journey to an AIDS vaccine. Last year, the lab of Mark Connors, chief of the HIV-specific immunity section at NIAID, isolated 10E8, which binds the MPER stalk in an unusual way. Unlike previously identified MPER bNAbs, it doesn’t bind phospholipid and doesn’t appear to be autoreactive (see Tapping the Sanguine HumorIAVI Report, Mar.-Apr. 2012). Its higher potency and breadth of neutralization compared to other MPER bNAbs too has improved the reputation of gp41 as a vaccine target of choice.

Jamie Scott, a molecular immunology professor at Simon Fraser University in Canada, has put together a team to develop DNA, liposome, and VLP vaccine candidates to elicit antibodies similar to 2F5, 4E10 and 10E8, all MPER-targeting bNAbs. “Vaccine research around the MPER has been impeded,” says Scott. “People think, ‘Oh, those are all auto-reactive antibodies’ and so they don’t want to work on it. There is huge controversy around this. There are people so tuned into pushing a particular agenda that they may be blind to other possibilities.”

Yet researchers have been unable to make immunogens that mimic the structural conformation of the MPER that is targeted by bNAbs. Scott likened this structure to a “speed bump on the surface of a viral membrane” that literally obstructs antibody targeting. To overcome this impediment, she guesses, you’d need a really flexible paratope—the part of the antibody that recognizes antigen.

Recently, Scott shed some light on an important role the transmembrane domain of gp41 plays in exposing the epitopes of three MPER bNAbs—2F5, 4E10 and Z13e1. The study showed how DNA constructs encoding the MPER, the transmembrane region and 27-amino acid residues of the cytoplasmic tail produced optimal antibody binding to the MPER (J. Virol. 86, 2930, 2012). Importantly, mutants of the 2F5 and 4E10 bNAbs that bind to MPER peptides but do not neutralize the virus also do not bind to the MPER in the context of the cell membrane.

Scott thinks the gp41 transmembrane region helps to fully expose the MPER to neutralization-competent binding by the 4E10 bNAb. She and her colleagues also developed molecular models explaining the difference in exposure of the bNAb epitopes in constructs where the MPER was fused to the transmembrane domain of gp41 and its short cytoplasmic tail, as compared to a longer stretch of the MPER fused to the transmembrane domain of the platelet-derived growth factor receptor.

On the heels of those findings, Scott and her collaborators secured $2.7 million in NIH grants and $330,000 from the Canadian government to develop DNA vaccine candidates targeting the MPER region.

Progress has been uneven. In their initial attempt to design a DNA vaccine candidate that could be expressed on the surface of the plasma membrane, Scott’s team realized that not enough of the 4E10 epitope was exposed. “So we turned around and made another vaccine [candidate],” says Scott. The second attempt improved 4E10 epitope exposure, but did not elicit antibodies against the chosen epitope. They are now tweaking the transmembrane region to improve responses in collaboration with William DeGrado’s lab at the University of California-San Francisco.

Meanwhile, her attempts to elicit responses to the 2F5 epitope—in collaboration with Shan Lu of the University of Massachusetts—have proved equally vexing. Antibody responses in rabbits immunized with DNA vaccines hit the MPER region targeted by 2F5, but only weakly. To try to improve those responses, Scott and her colleagues decided to pair a DNA prime with a liposome-associated peptide boost developed by José Nieva-Escandón’s lab at the University of the Basque Country in Spain. Dr. Nieva-Escandón’s lab is developing two liposome-based MPER vaccines: one that covers the 2F5 epitope and another covering the 4E10/10E8 epitopes. They hope that association with the liposome will expose the MPER in a structure that is amenable to binding by 4E10 and 2F5 but not by non-neutralizing antibodies.

Lu is involved in the design of the immunization strategies, and is testing the vaccine candidates in rabbits. While supportive of Scott’s hypothesis, he is not quite sure it will ever be possible to get the immune system to produce potent antibodies against the MPER. “We have some [animal] data from previous studies that showed we can generate antibody, at least against [the 2F5 epitope],” says Lu. “Of course, neutralization is not impressive or potent, but at least we hope that we can somehow improve the avidity.”

Scott certainly feels it’s worth a try. In her view, too much time and money have been spent assessing the putative autoreactivity of MPER antibodies at the expense of addressing larger questions in AIDS vaccine research. “We need to understand how to drive a B-cell response to make bNAbs,” she says. “There are lots of hypotheses about the crucial features of the bNAbs and the B-cells that produce them, and how we can reproduce them with a vaccine. So why are we worrying about deletion of clones? None of the HIV vaccines under development elicits high-titer bNAbs yet—it’s not just the MPER vaccines. All the money that has gone into whether [2F5 or 4E10] are autoantibodies could have been used to better understand how bNAb responses are made in the first place.”

Other recent findings, meanwhile, bolster the notion that MPER antibodies will be difficult to elicit with a vaccine. A study led by Duke University scientist Garnett Kelsoe identified two self-antigens bound by 2F5 and 4E10 antibodies—human kynureninase (KYNU) and splicing factor 3b subunit 3 (SF3B3). These findings buttress the hypothesis that the autoreactivity of conserved MPER epitopes represents yet another mechanism HIV has evolved to evade immune attack (J. Exp. Med. 210, 241, 2013). Such mimicry of host antigens, they concluded, is the prime reason for the poor immune responses to the MPER region of Env. Mapping studies of the endogenous antigens targeted by autoreactive antibodies reinforced Kelsoe’s conclusion. It turns out that opossums carry a KYNU gene that abolishes 2F5 binding. When immunized with gp140 and then boosted with a peptide immunogen containing the 2F5 and 4E10 linear epitopes, opossums produced high titers of antibody to the 2F5 epitope but little or nothing in response to the 4E10 epitope. For comparison, mice were primed and boosted with the same regimen at two week intervals for 12 weeks. Serum antibody to both the 2F5 and 4E10 epitopes was significantly delayed and, at their peak, antibody responses to both were significantly weaker than those observed in opossums.

Kelsoe and his colleagues argue that this suggests that some humans too might carry KYNU polymorphisms that abolish the 2F5 epitope, suggesting that some people might be able to produce such antibodies. They note, however, that this is for now just a hypothesis.

A recent study led by Haynes might further cheer fans of the MPER. Researchers used a knockin mouse strain expressing 2F5 to show that their immune system eliminates most, though not all, of the B cells that make 2F5 because they are autoreative. But Haynes’ group also showed that if apoptosis of bone marrow B cells is circumvented by genetic manipulation, it is possible to rescue B cells bearing self-reactive 2F5 heavy chain/long chain pairs (J. Immunol. 187, 3785, 2011). “If deletion is complete then it would be hopeless to induce these antibodies,” says Haynes. “The good news is that the deletion is not complete. A group of anergic B cells can maintain modified heavy and light chains of broadly neutralizing antibodies. The body has not modified them and, with the right immunogen, we can wake them up.”

The hunt will not be easy. Just ask Michael Zwick. An associate professor of immunology and microbial science at The Scripps Research Institute (TSRI) in California, Zwick, who worked initially in Scott’s lab, actually started out studying the MPER antibodies in Dennis Burton’s lab at TSRI. He led the effort years ago to characterize the MPER-binding antibodies 4E10, 2F5 and a third antibody he identified from a phage library (J. Virol. 75, 10892, 2001).

“From there, it became this sort of journey,” says Zwick. “The linear peptide epitopes corresponding to these antibodies are relatively straightforward. Antibody binding can be knocked out by a few different residues. We did some early immunizations in mouse models, but the antibody responses were poor. It has been like that for over a decade.”

Zwick says some labs, including his own, dug deeper to try and answer some essential questions about the MPER, such as how it becomes exposed on a functional Envelope glycoprotein spike and how much of a window antibodies have to access relevant epitopes on the MPER. But the MPER constructs that they generated induced neutralizing responses that were weak at best.

Such difficulties have sucked the life out of efforts to elicit MPER-specific responses through vaccination, says Zwick. More recently, as potent and broadly effective antibodies that bind choice epitopes have been discovered in scads, interest in the MPER domain has declined precipitously among vaccine designers. Zwick himself has broadened his research efforts to find ways to overcome the instability of the HIV Envelope, the source of much of his current NIH funding. His lab recently combined seven mutations that increased trimer stability, primarily in the gp41 region but also in the V1 region of gp120, to generate a native Env trimer with superior homogeneity and stability (PLoS Pathog. 9, e1003184, 2013).

“One problem with HIV Env,” he says, “is if you produce it as a soluble glycoprotein, it does not readily fold in the same way as a native Envelope. It looks different when probing it using non-neutralizing antibodies. Often it doesn’t hold together or properly mimic native features of gp41. We thought by selecting mutations it might help get more reliable structures.”

Immunizing with VLPs or killed inactivated viruses, says Zwick, induces antibody responses. But often the neutralizing antibody titers are low, despite the generally vigorous antibody responses to Env. “We don’t know why this is exactly, and we don’t really know how antibody responses are made to Env in vivo,” says Zwick. “But we are getting much better tools to address this problem, with better mAbs and more ways to study B-cell responses.”

Zwick says that he, at least, hasn’t abandoned the MPER. “It’s too early to tell what might work as a vaccine,” says Zwick. “I’m a fan of trying things though.”

Even strategies and targets, presumably, that aren’t all the rage today.