Crystal Structure of PG16 Antigen-Binding Portion Reveals Unusual Features
The recent isolation of several new broadly neutralizing antibodies was good news for researchers working to design AIDS vaccine candidates that could induce antibodies capable of neutralizing many of the viral variants currently in circulation (see Adding to the Armamentarium of Broadly Neutralizing Antibodies, Research Briefs, IAVI Report, Jan.-Feb. 2010). Last September, IAVI researchers, in collaboration with researchers at The Scripps Research Institute (TSRI), announced the isolation of two of these antibodies, dubbed PG9 and PG16.
Now, two research teams, one led by Ian A. Wilson and Dennis Burton at TSRI, the other by John Mascola and Peter Kwong at the Vaccine Research Center (VRC) of the National Institute of Allergy and Infectious Diseases (NIAID), have identified the crystal structure of the Fab portion of PG16, which is the part of the antibody involved in antigen binding (1,2). Because PG16 binds best to the native HIV Env trimer, which researchers have so far been unable to crystallize, both teams determined the structure of PG16’s Fab in an unbound state.
One of the most unusual features of PG16, the studies found, is the structure of its extremely long (28 amino acids) CDR H3, which is one of the most variable parts of the antigen-binding region of an antibody and known to often be important for antigen binding. “There has never been a crystal structure of an antibody with a [CDR H3] loop that long,” says Robert Pejchal, a postdoctoral researcher in Wilson’s lab and the first author of the PNAS study.
“[The CDR H3] forms a mini domain that sort of towers above the rest of the antibody,” adds Wilson, Hansen professor of structural biology at TSRI and one of the lead authors of the PNAS study. “We have called it a hammerhead [because it looks] like a hammerhead shark.” This shape suggests that the antibody may use this domain to access an occluded epitope on the Env trimer that would not ordinarily be easy to access, says Pejchal.
PG16 Fab structure with variable (V) and constant (C) parts of the heavy chain in orange and the variable and constant parts of the light chain in yellow. The CDR H3 can be seen on top, with the sulfate group on one of its tyrosine residues shown in yellow (S) and red (O). This is a ribbon diagram that outlines the protein backbone and shows beta strands as arrows and random coils as thin tubes.
Courtesy of Robert Pejchal at The Scripps Research Institute; also published in Proc. Natl. Acad. Sci. 107, 11483, 2010.
Peter Kwong’s group at the VRC found the CDR H3 loop had the same shape, which they call an axe, but they also found that in some cases, the CDR H3 is disordered. “Even though it can form this axe or hammerhead structure, it doesn’t necessarily have a fixed structure. You can deform it relatively easily,” says Kwong, who is chief of the structural biology section at the VRC. “The amount [it] deforms might be important for recognizing its particular epitope. If it was totally fixed it might not be able to get into the little crevice that it might have to get to.”
The CDR H3 loop appears to be necessary for neutralization because mutating parts of it greatly reduced or even eliminated the ability of PG16 to neutralize HIV, the authors of the PNAS study found. “We think that this hammerhead is probably what’s mediating the interaction between PG16 and the HIV trimer,” says Laura Walker, a graduate student in Dennis Burton’s lab, who did the mutation analysis for the PNAS study.
The CDR H3s of PG9 and PG16 differ mostly in a stretch of seven amino acids, and, according to the PNAS study, exchanging the seven amino acid stretch of the two antibodies resulted in PG16 being able to neutralize certain HIV isolates with a similar potency to PG9 and vice versa. Kwong’s group found a similar effect when they swapped the entire CDR H3 between PG9 and PG16.
The authors of the PNAS study found that another unusual feature of PG16 is that one of the tyrosine residues on the CDR H3 contains a sulfate group, the removal of which decreased the neutralization potency of PG16 by about ten-fold, Walker says. “[The sulfate is] not required for neutralization but it enhances neutralization,” says Walker.
Kwong says that PG16 also shows quite an extensive degree of affinity maturation—about 20% of the amino acids in the variable region of PG16 differ from its germ-line version. This is still less than the 30% affinity maturation reported for the variable region of the VRC01 antibody, but more than the 5-10% affinity maturation observed in most other antibodies (see Antibody Fever, IAVI Report, Mar.-Apr. 2010).
Kwong’s group found that the extensive affinity maturation was important for the neutralization potency and breadth of PG9 and PG16. When the researchers made less affinity matured versions of these antibodies that had some, or all, of their variable gene portion reverted to germ line, they found that increased affinity maturation correlated with increased potency and breadth of neutralization of a panel of HIV isolates.
Still, somewhat surprisingly, a PG9 version that had its variable gene portion completely reverted to germ line could still neutralize one HIV isolate. This led Kwong and colleagues to suggest that a possible vaccine strategy to kick start the affinity maturation process would be to use Env proteins from the HIV isolate that can still be neutralized by the less affinity-matured version of the antibodies. This could then be followed by boosts with a cocktail of Env proteins from HIV variants that are neutralized by increasingly more affinity-matured versions of the antibodies. However, it’s still unclear how the CDR H3 part of the antibody may be generated, and how it could be induced by a vaccine.
Image adapted from a cryoelectron tomographic structure of the native HIV Envelope trimer (outer shell), filled with the crystal structure of the b12-bound monomeric gp120 core. PG9 and PG16 are thought to bind to the V1/V2 and V3 loops (shown as dark blue and teal ovals, respectively), whose structure is unknown. Carbohydrate chains are shown in purple, and the oligomannose cluster targeted by 2G12 is shown in yellow.
Robert Doms, who chairs the department of microbiology at the University of Pennsylvania and was not connected to the study, says the findings fit a trend among broadly neutralizing antibodies. “Virtually all broadly neutralizing antibodies to date are structurally ‘odd’, in that they have one or more features that have either never been seen before, or that have been seen in other antibodies only rarely. Will unusual structural features be required for potent and broad neutralizing activity? If so, we know essentially nothing about how to elicit such antibodies.”
Next, it will be important to solve the structure of the HIV Env trimer and PG16 bound to it, says Pejchal. “That will really reveal how these antibodies work,” he says.
It won’t be easy, however, because PG16 binds best to the native trimeric form of Env, which researchers have so far been unable to crystallize. “That’s an enormously challenging undertaking,” Wilson says. “The Envelope [trimer] is not very stable, it’s very hard to produce and you cannot produce large quantities of it for structural studies. It’s probably one of the most challenging things that still have to be done.” —Andreas von Bubnoff
Possible Explanation for HLA-class I Associated Control of HIV Infection
Researchers have known for some time that elite controllers—HIV-infected individuals who, without antiretroviral therapy, can keep their viral load at a level that is undetectable by currently available commercial viral load assays—are more likely to have certain versions, or alleles, of genes that encode major histocompatibility complex (MHC) class I molecules, including one called HLA B57 (in humans, MHC is called human leukocyte antigen, or HLA). This suggests that genes like B57 are involved in the control of viral load in elite controllers.
Now, a research team led by Arup Chakraborty, a professor of chemistry, chemical engineering, and biological engineering at the Massachusetts Institute of Technology, and Bruce Walker, a professor of medicine at Harvard Medical School, have developed a model that suggests that people with HLA B57 control viral load better in part because their T cells go through a less rigorous selection process in the thymus (1).
In the thymus, CD8+ T cells encounter cells that present self peptides—small pieces of the body’s own proteins—bound to MHC class I receptors. CD8+ T cells whose T-cell receptors bind strongly to these self peptides die, which is important for avoiding autoimmune disorders.
In their model, Chakraborty and colleagues suggest that more protective versions of the MHC class I receptor, such as B57, bind and therefore present a smaller diversity of the body’s own peptides to the CD8+ T cells in the thymus than less protective versions. As a result, immature CD8+ T cells in the thymus of people with B57 have fewer opportunities to die as a result of binding strongly to self peptides, making it more likely that CD8+ T cells survive that are more cross-reactive to point mutations of viral peptides. “[In] B57 [positive] people, the T cells have to pass a less rigorous test because they have to avoid binding strongly to only a smaller diversity of self peptides,” Chakraborty says. “Their T cells were educated in a thymus with fewer types of self peptides.”
If this model is correct, it would suggest that people with B57 can better control viral load in part because their CD8+ T cells are more likely to recognize a diverse array of HIV peptides presented by MHC class I molecules on the surface of HIV-infected cells, including peptides that come from mutants of HIV that arise during infection. As a result of this recognition, these CD8+ T cells will turn into cytotoxic T lymphocytes that kill the infected cell. Therefore, people with genes like HLA B57 would be more likely to generate cytotoxic T cells that can kill cells infected with a wider array of HIV strains, particularly HIV mutants that arise during infection.
The first hint that led Chakraborty and colleagues to their model came when they calculated how many self peptides could be bound by the different MHC receptor types encoded by different HLA alleles. Using published data from binding experiments, they first tested whether computer algorithms developed to predict which peptides can bind to different versions of MHC class I receptors were accurate. They found that the most accurate algorithm predicted that B57 would bind fewer self peptides than other versions of the MHC class I receptor that are less protective to HIV. “[This] was striking since HLA B57 is the allele most associated with HIV control,” says Walker.
Using a computer model they had developed to simulate the thymic selection of T cells, they also found that T cells that survive after having been presented fewer self peptides in the thymus are more likely to bind to a diverse set of HIV peptides, particularly ones that come from HIV mutants that arise during infection. Using yet another computer model, they then showed that this would be expected to lead to better control of viral load.
Finally, the researchers tested these computer modeling predictions with viral load data from cohorts of about 1,100 controllers and 600 progressors. They found that HIV-infected people with the MHC class I versions B57 and B27, which are predicted to bind to fewer self peptides, are more likely to be HIV controllers, whereas HIV-infected people with B07 and B35, versions that are predicted to bind more self peptides, are more likely to be HIV progressors. “We have identified one contributing factor to why people with this [B57] allele may be more likely to control HIV,” Chakraborty says.
The model is consistent with observations that people with protective HLA alleles are more likely to have autoimmune and hypersensitivity reactions, for example the HLA B57-associated hypersensitivity reaction to the antiretroviral drug Abacavir, Chakraborty says.
Another prediction of the model is that the better cross-reactivity of CD8+ T cells in people with B57 should also enable them to better control infections with other pathogens. “This [CD8+ T cell] cross reactivity would not be specific only for HIV,” Chakraborty says. Unpublished data from Walker and colleagues indicate that HLA alleles such as B57 are also good at controlling hepatitis C virus (HCV), another rapidly mutating RNA virus.
But Mark Connors, chief of NIAID’s HIV-specific immunity section who was not involved in the study, says this doesn’t necessarily fit with his observations. He found that B57 positive HIV non-progressors who are also infected with cytomegalovirus (CMV) or HCV have only a few CMV and HCV peptides presented by B57 (2). “If this [mechanism] was operating by the way that the authors claim, you would expect that B57 would dominate all responses to all viruses because you are going to have more broadly reactive T cells,” Connors says. “But it doesn’t.”
Walker says that the different observations in Connors’ study might be because Connors mostly studied people who are co-infected with HIV and CMV or HCV, whereas Walker’s unpublished data mostly comes from people who were not co-infected. “We find that the strong correlation of B57 with protective HCV responses is diminished for individuals co-infected with HCV and HIV,” Walker says, “suggesting that co-infection may have unique aspects to it that impact immune control.”
David Baltimore, a professor of biology at the California Institute of Technology who was not connected to the study, called it “a fascinating new interpretation of the effect of an MHC locus on a person’s ability to fight off HIV.”
However, Connors says, the study’s authors could have experimentally validated their prediction that protective MHC class I receptor variants really bind fewer self peptides than non-protective variants. “You can do things like elute peptides and determine whether in fact B57 does bind more or fewer [different self peptides],” Connors says, adding that not all protective alleles seem to bind fewer self peptides.
But Chakraborty says that even though the effect might not be as strong for all alleles, that doesn’t necessarily mean the model is wrong. “The new factor we have identified is not the only factor that contributes to virus control,” he says, adding that this new mechanism should contribute the most for alleles that bind either very few or very many types of self peptides, according to his experimental data. “All effects that are identified should not be thought to be mutually exclusive.” —Andreas von Bubnoff