Rational Design of Vaccines to Elicit Broadly Neutralizing Antibodies to HIV-1

ROLE OF ANTIBODIES IN PROTECTION

The design of immunogens able to elicit neutralizing antibodies (NAb) remains a major goal of HIV-1 vaccine development. Most licensed viral vaccines induce antibodies that neutralize the infecting virus, thereby protecting against infection or disease. Although the specific immune responses required to protect humans against HIV-1 infection are not known, studies of lentiviral infection in nonhuman primates (NHPs) have shown that passive infusion of antibodies can prevent infection. Specifically, antibodies that neutralize HIV-1 have been shown to prevent infection by a chimeric simian-human immunodeficiency virus (SHIV) containing the env gene of HIV-1. Early SHIV challenge studies used intravenous inoculation of the challenge virus. In this setting, high concentrations of NAbs were required to block infection, and nonneutralizing antibodies were unable to protect against infection (Mascola et al. 1999; Shibata et al. 1999). Subsequent studies, using a single oral or vaginal inoculation sufficient to infect 100% of control animals, also showed protection by NAb (Baba et al. 2000; Mascola et al. 2000; Hofmann-Lehmann et al. 2001; Parren et al. 2001). The protection against SHIV infection has been most directly associated with the neutralization potency of the infused antibodies (Nishimura et al. 2002; Mascola 2003). However, recent studies have also suggested a role of Fc-mediated antibody effector functions in conferring protection. The neutralizing monoclonal antibody (mAb) IgG1 b12 showed a diminished protective effect if the Fc region of the IgG was altered to knock out complement binding and ADCC activity (Hessell et al. 2007). Importantly, recent passive transfer studies have employed low-dose mucosal inoculation that requires multiple challenges to infect all control animals (Hessell et al. 2009a,b). This model may be more physiologically relevant to the relatively inefficient HIV-1 infection in humans. In the low-dose NHP model, approximately 10-fold less antibody was required to mediate protection against infection compared to prior studies with high-dose inoculations; serum antibody titers sufficient to mediate 90% virus neutralization at 1:5 serum dilution were associated with protection. Hence, vaccines may not need to achieve extraordinarily high levels of HIV-1 NAbs. However, the antibody response will likely need to be durable, and NAbs will have to cross-react with a genetically diverse spectrum of HIV-1 strains.

CLINICAL TRIALS OF CANDIDATE HIV VACCINES

To protect against HIV-1, antibodies must bind the viral surface envelope glycoprotein, a homotrimer composed of the gp120 surface unit and the gp41 trans-membrane domain. The earliest phase I vaccine trials included recombinant envelope products, gp120 or gp160, formulated in various adjuvants. Live recombinant vectors such as vaccinia and canary pox and nucleic acid–based vaccines have also been tested, usually including gene inserts expressing the Env glycoprotein. Recombinant protein vaccines used alone, or as a boost to vaccine vectors, generally elicited high titers of anti-Env antibodies. Initial immunogenicity studies showed that vaccine-elicited antibodies could neutralize HIV-1 in vitro, but it was soon realized that the viral neutralization was limited to prototype laboratory-adapted HIV-1 strains and did not extend to primary HIV-1 isolates (Wrin and Nunberg 1994; Mascola et al. 1996). These neutralization data generated considerable debate regarding the rationale for efficacy testing of protein-based vaccines. Despite this uncertainty, two phase III gp120 vaccine trials were conducted, each with a bivalent formulation of two strains of gp120 formulated in Alum. The VAX004 and VAX003 studies were initiated in 1998 and 1999, respectively, and the results reported in 2003. These gp120 vaccines showed no significant impact on acquisition of HIV-1 infection and had no impact on plasma viremia or peripheral CD4⁺ T-cell counts (Flynn et al. 2005; Gilbert et al. 2005).

The failure of these gp120 vaccines was generally viewed as evidence that a successful vaccine would need to induce more potent NAbs that can neutralize circulating strains of HIV-1. Prior to the release of the VAX003 and VAX004 results, an additional phase III trial was planned. The RV144 study included priming immunizations with an avipox vector (ALVAC) and boosting with the same bivalent gp120s used in the VAX003 study. The results of this study, released in 2009, showed that volunteers in the vaccine arm of the study acquired 31% fewer HIV-1 infections than those in the placebo arm (Rerks-Ngarm et al. 2009). This modest efficacy, although not deemed adequate for licensure, was the first indication that a vaccine could protect against HIV-1 infection. The immunologic data from the RV144 vaccine trial are still being analyzed. Immune correlates analysis will include a set of in vitro antibody assays, including traditional virus neutralization, and assays of antibody binding, ADCC, and additional measures of Fc-mediated antibody effector functions. Whether any of these measures of vaccine-elicited antibodies will correlate with protection is yet to be determined. It is also unclear why this ALVAC/gp120 trial produced a modest protective effect whereas the gp120-only VAX003 and VAX004 studies did not. Differences in the studies include the ALVAC prime in the RV144 study and substantial differences in the risk factors and routes of HIV-1 infection for the populations studied. In summary, human efficacy trials indicate that vaccine-elicited protection against HIV-1 infection is achievable, but the specific antibody responses that may contribute to protection have not been elucidated. Despite these limitations in our knowledge of immune correlates, the modest 31% protection observed in the RV144 study suggests much room for improvement. One means of achieving improved vaccine efficacy may be through the elicitation of more potent and cross-reactive NAbs.

HUMORAL IMMUNE RESPONSE DURING HIV-1 INFECTION

Most vaccines seek to mimic the immune response generated during natural infection. HIV-1, however, shows extensive genetic variability owing to error-prone reverse transcription of the viral genome and a high tolerance for mutations that mediate viral immune evasion to antibody and CD8 T-cell responses. As a consequence of chronic persistent replication in the face of immune pressure, the HIV-1 epidemic is comprised of diverse genetic variants and a poorly understood level of antigenic diversity. A critical feature of HIV-1 appears to be its general resistance to NAbs. This resistance is manifest during the early phase of HIV-1 infection, in which NAb responses to the infecting virus may not appear until several months after infection. Once such autologous NAbs arise, the virus quickly escapes, leading to an ongoing cycle of adaptive antibody responses and further viral escape (Albert et al. 1990; Montefiori et al. 1991; Richman et al. 2003; Wei et al. 2003; Flynn et al. 2005; Gilbert et al. 2005). Despite this impressive immune evasion capability, recent evidence suggests that there are some vulnerabilities in the protective armor of HIV-1 that could be exploited by rationally designed antibody-based vaccines.

The advent of high-throughput neutralization assays using recombinant Env pseudoviruses has permitted the screening of sera from relatively large cohorts of HIV-1-infected donors. These data reveal that between 10% and 25% of HIV-1-infected subjects make NAb that cross-react with a substantial portion of diverse HIV-1 strains (Dhillon et al. 2007; Li et al. 2007; Doria-Rose et al. 2009; Sather et al. 2009; Stamatatos et al. 2009). A smaller subset of these sera is able to neutralize the large majority of circulating HIV-1 isolates (Li et al. 2007; Simek et al. 2009). The observation that some individuals make broadly cross-reactive neutralizing antibodies has spurred a renewed effort to isolate neutralizing mAbs to define vulnerable epitopes on the viral Env that could serve as targets of vaccine design. Until 2009, only a handful of known neutralizing mAbs were able to neutralize primary isolates of HIV-1, and even these antibodies displayed limitations in overall potency or breadth of reactivity (reviewed in Zolla-Pazner 2004; Pantophlet and Burton 2006; Mascola and Montefiori 2010). The effort to isolate new mAbs has been bolstered by several recent technological advances. Multiparameter flow cytometry can be used to identify and sort individual HIV-1 Env-specific memory B cells (Scheid et al. 2009). The antibody heavy and light chain can then be genetically recovered from the cDNA of single B cells and the full IgG expressed. An alternate methodology involves the screening of thousands of unselected individual memory B cells, each stimulated to secrete IgG. The supernatants of these B-cell cultures are screened for HIV-1 neutralization using a high-throughput microneutralization assay (Walker et al. 2009). The application of these new B-cell technologies has resulted in the isolation of several new HIV-1 mAbs that are more potent and broadly reactive than were prior antibodies. For example, the PG9 and PG16 mAbs are directed to a previously undefined site within the V2-V3 region of the HIV-1 viral spike (Walker et al. 2009). Another example is the VRC01, VRC02, and VRC03 mAbs that bind to a functionally conserved region of gp120 that interacts with the host cell receptor CD4 (Wu et al. 2010; Zhou et al. 2010). The fairly rapid and efficient isolation of these new mAbs suggests that numerous additional neutralizing mAbs will be forthcoming. Because each category of new broadly neutralizing mAbs potentially identifies a highly conserved target of HIV-1 neutralization, the structural analysis of these antibodies bound to HIV-1 Env can provide valuable insights for vaccine design.

STRUCTURAL VIROLOGY

Structural biology provides atomic-level details about the three-dimensional organization and chemical structure of proteins. Such details facilitate an understanding of mechanism and—combined with recent advances of in silico protein design—provide a means by which to manipulate and to optimize Env-based immunogens (Nabel et al. 2011). Overall, the host-derived envelope that surrounds HIV-1 sequesters most of its proteins from antibody-mediated recognition. Only two HIV-1 proteins protrude through this protective membrane: the HIV-1 gp120 envelope glycoprotein and its gp41 trans-membrane partner. Three gp120s associate noncovalently with three gp41s to make up the viral spike (reviewed in Wyatt and Sodroski 1998).

A combination of X-ray crystallography and electron microscopy has served to illuminate many of the details of the viral spike (Fig. 1). HIV-1 uses a two-receptor mechanism involving the host receptor CD4 and a coreceptor (generally CCR5 or CXCR4), along with considerable Env-based conformational change, to fuse viral and target cell membranes and to enter host cells (Wyatt and Sodroski 1998). In the first or “unliganded” state, the HIV-1 viral spike covers most of its surface with host-derived N-linked glycan (Wyatt et al. 1998). These glycans appear as “self” to the humoral immune system and therefore are mostly immunologically silent. In the unliganded state, only a few nonglycosylated sites are available for antibody-based recognition. After CD4 engages gp120 at the cell surface, the HIV-1 viral spike undergoes substantial structural rearrangements: in the CD4-bound conformation of gp120, a four-stranded bridging sheet forms and the V3 loop is sprung, and in gp41, the previously occluded N-heptad repeat becomes available for T-20 recognition (reviewed in Wyatt and Sodroski 1998). In the CD4-bound state, much more of the glycan-free viral spike surface is uncovered and potentially available for recognition by neutralizing antibody, such as V3-directed antibodies or CD4-induced (CD4i) antibodies that bind to the bridging sheet. However, the close proximity of viral and target cell membrane sterically occludes access of immunoglobulins to these epitopes (Labrijn et al. 2003). CD4 engagement induces the formation of a high-affinity site for coreceptor binding (Wu et al. 1996). Although details remain to be deciphered, coreceptor binding facilitates additional conformational changes. A transition intermediate has been hypothesized to form, in which gp41 is fully extended, with a fusion peptide thrown into the target cell membrane and the carboxy-terminal trans-membrane region embedded in the viral membrane. This transient intermediate resolves itself by forming a highly stable six-helix bundle, the postfusion state, in which fusion peptide and trans-membrane regions of gp41 are juxtaposed.

SITES OF HIV-1 VULNERABILITY

The assembled viral spike is highly protected from antibody-mediated recognition. Nonetheless, as described above, a number of broadly neutralizing antibodies have recently been discovered. The breadth of these antibodies indicates that they recognize conserved Env regions of functional importance; conversely, the susceptibility of these regions to antibody-mediated recognition indicates that they form sites of HIV-1 vulnerability to the humoral immune response.

One such site of Env vulnerability is the initial site of attachment for the CD4 receptor (Fig. 2A,B) (Zhou et al. 2007; Chen et al. 2009). This site, on the outer domain of gp120, consists of roughly two-thirds of the surface HIV-1 uses to fully engage the CD4 receptor. Although the site is itself glycan-free, N-linked glycosylation and variable loops (e.g., V5) surround much of it. Moreover, most of the surrounding surface that is glycan-free is highly susceptible to conformational change; antibodies that bind to these regions induce conformations in the neighboring V1/V2 variable regions that are incompatible with the functional viral spike. Precise targeting by antibodies onto the initial site of CD4 attachment thus appears to be required for effective neutralization at this site (Chen et al. 2009).

The first monoclonal antibody found to bind effectively to this site, the b12 antibody, was identified from a phage library (Burton et al. 1994) and found to use a heavy-chain-only means of recognition (Zhou et al. 2007), which is generally not observed in naturally elicited antibodies. The b12 antibody nevertheless displays reasonable recognition of the site (Zhou et al. 2007), although extension outside the target site, especially around the CD4-binding loop, allowed for antigenic variation, especially with non-B clade isolates (Wu et al. 2009b). As a consequence, the b12 antibody only neutralizes ∼35% of circulating isolates.

Other antibodies have been isolated more recently that also target this site. Antibody HJ16 is a natural human antibody isolated by direct assessment of neutralization coupled to single B-cell antibody sequencing (Corti et al. 2010). The atomic-level structure of this antibody in complex with gp120 has not yet been determined, but initial mapping suggests that antibody HJ16 strays into the “loop D” region, outside of the site of vulnerability (Corti et al. 2010); overall HJ16 neutralizes ∼30% of circulating isolates (Corti et al. 2010). Antibody VRC01 and related antibodies VRC02 and VRC03 appear to target the site of vulnerability more precisely. These antibodies show neutralization breadths of up to ∼90% of circulating isolates (Wu et al. 2010). The structure of VRC01 in complex with gp120 reveals an extraordinary mimicry between VRC01 heavy chain and CD4 receptor (Zhou et al. 2010).

Another site of vulnerability on the viral spike shows quaternary structural constraints, and maps to the second and third variable regions of gp120 (variable loops V2 and V3) (Fig. 2C). The structure of this portion of the Env has not yet been determined, and its functional importance is also not clear, but may relate to HIV-1 recognition of the α4β7-integrin, the gut-homing receptor for HIV-1 used by primary isolates during early stages of infection (Arthos et al. 2008). A number of antibodies have been identified that recognize this region, including the human monoclonal antibodies 2909, PG9, and PG16 and a number of rhesus antibodies (Gorny et al. 2005; Walker et al. 2009; Robinson et al. 2010). Structural analysis indicates that all of these antibodies use extended heavy chain third-complementarity determining regions, which are anionic and tyrosine sulfated (Pancera et al. 2010b; Pejchal et al. 2010; Changela et al. 2011). Despite this similarity, these antibodies vary dramatically in their neutralization breadth, with 2909 being extremely strain-specific and PG9/PG16 able to neutralize 70%–80% of current circulating isolates (Honnen et al. 2007; Walker et al. 2009). This divergence in breadth appears related to the specific immunotype of the quaternary site of vulnerability recognized, with strain-specific variants recognizing rare variants of the site, and more broadly neutralizing antibodies recognizing more common variants of the site (Wu et al. 2011).

Another critical function the HIV-1 envelope performs relates to fusion of viral and target cell membranes, which is required for virus entry. Virtually all of the functionally conserved surfaces required for fusion are occluded in the functional viral spike and available for antibody-mediated neutralization only as transient intermediates in the entry process. Even then, access is limited. For example, entry requires the N-heptad repeats of gp41 to snap back on themselves, and small molecule mimics of the carboxy-terminal heptad repeat (e.g., T-20 or Fuzeon) are effective therapeutics (Baldwin et al. 2003). Nonetheless, antibodies that recognize this potential site of vulnerability such as D5 or HK20 have weak potency and limited breadths of neutralization, properties attributed to steric occlusion at the viral membrane-target cell membrane interface (Luftig et al. 2006; Gustchina et al. 2010; Sabin et al. 2010). One area of vulnerability that appears less sterically occluded is the membrane-proximal external region (MPER) (Fig. 2C). The precise role that this region plays in viral entry is unclear, but alteration of hydrophobic residues in the MPER leads to loss of fusion capabilities (Munoz-Barroso et al. 1999; Salzwedel et al. 1999). Human antibodies 2F5, Z13e, and 4E10 have been found to recognize the MPER and to show reasonable neutralization breadths and potencies (Muster et al. 1993, 1994; Trkola et al. 1995; Stiegler et al. 2001; Zwick et al. 2001; Binley et al. 2004). Interestingly, this site of HIV-1 vulnerability includes not only the HIV-1 Env (i.e., specific amino acids in the MPER), but also the neighboring or surrounding lipid membrane. Antibodies that recognize the MPER thus require a hydrophobic membrane-binding component to neutralize virus (Ofek et al. 2004; Sun et al. 2008; Alam et al. 2009; Julien et al. 2010; Ofek et al. 2010b). Such membrane “corecognition” appears to lead to self-recognition, and the MPER-directed antibodies are generally self-reactive (Haynes et al. 2005; Alam et al. 2007). Such self-reactivity may impede antibody development (B-cell deletion or anergy), and few MPER-directed neutralizing antibodies are observed in sera from HIV-1-infected individuals (Walker et al. 2009).

A number of other sites of vulnerability can be inferred from antibodies that neutralize with different specificities. A conserved cluster of high-mannose glycans (around residues 295, 332, and 392 of gp120) is recognized by the 2G12 antibody (Sanders et al. 2002). This antibody has a highly unusual structure that involves variable-domain swapping (Calarese et al. 2003), and most sera do not competitively inhibit 2G12 binding, suggesting that antibodies against this site are very rare. A recent analysis of the elite neutralizers from the protocol G screen of sera from almost 2000 HIV-1-infected individuals, however, indicates that some of the best neutralizers recognize this face of the Env trimer (Simek et al. 2009; Walker et al. 2009, 2010a). Other antibodies directed at the V3 loop (up to 10%–30% breadth) or of CD4-induced specificity (up to 10% breadth) indicate that these regions form partial sites of vulnerability (Xiang et al. 2003; Zolla-Pazner and Cardozo 2010).

IMMUNOGEN DESIGN

CONCLUDING REMARKS

HIV-1 has evolved multiple mechanisms to evade the neutralizing antibody response. Specifically, Env evades host recognition by virtue of its sequence diversity, limited exposure to the immune system because of carbohydrate masking, and conformational flexibility. HIV-1 has been resistant to classification by traditional serotyping, suggesting that standard approaches to vaccine development are unlikely to succeed. Recent progress in the definition of broadly neutralizing antibodies, the elucidation of the structures of these antibodies complexed to HIV-1, and the utilization of structural biology to define the relationship between antigenicity and immunogenicity in vaccine candidates has catalyzed a resurgence in this area of research. An understanding of the biology of HIV-1, the human immune response to the virus, and the application of structural biology to immunogen design have provided new opportunities to advance the goal of identifying vaccines that elicit broadly neutralizing antibodies that prevent or contain infection and/or progression to AIDS.