A Comparative Immunogenicity Study of HIV-1 Virus-Like Particles Bearing Various Forms of Envelope Proteins, Particles Bearing no Envelope and soluble monomeric gp120

Production of VLP immunogens

In this pilot study, we opted for moderate Env expression by plasmid pCAGGS, together with pNL4-3.Luc.R-E- to express VLPs bearing functional Env trimers (Binley et al., 2003). To stabilize gp120-gp41 association and eliminate the potential problem of gp120 shedding that might occur with “WT-VLPs”, we generated “SOS-VLPs” that introduce a gp120-gp41 disulfide bond (Abrahamyan et al., 2003; Binley et al., 2003; Binley et al., 2000), and UNC-VLPs that abolish gp120-gp41 processing (Fig. 1). Collectively, we refer to these as “Env-VLPs, distinct from “naked” VLPs, lacking Env.

We compared full-length and truncated gp160 forms of Env on VLPs. Gp160ΔCT_WT was completely dependent on the expression of structural proteins by pNL4-3.Luc.R-E- for secretion into supernatant (Fig. 2A, lanes 1–3), but full-length gp160 was secreted into the supernatant, even when pNL4-3.Luc.R-E- was not co-expressed (not shown), perhaps indicating spontaneous vesicle formation or an increase in cell lysis. In addition, gp160ΔCT exhibited higher expression and gp120/gp41 cleavage than the full-length form, as observed previously (Binley et al., 2003; Yamshchikov et al., 1995; Yao et al., 2000). Gp160ΔCT_SOS was also fully cleaved and reducible by DTT (Fig. 2A, lanes 4 and 5). Since gp160ΔCT is also fusogenic (Abrahamyan et al., 2005; Binley et al., 2003; Crooks et al., 2005; Edwards et al., 2002), we reasoned that this is an authentic form of Env suitable for testing as an immunogen. In contrast to WT- and SOS-VLPs, the Env in UNC-VLPs was uncleaved, as expected (data not shown and ref. (Moore et al., 2006)).

Electron microscopic analysis of 293T cells co-transfected with pCAGGS gp160ΔCT_WT and pNL4-3.Luc.R-E- revealed efficient particle budding (Fig. 2B). SDS-PAGE and Western blotting of VLP preparations provided further evidence that gp160ΔCT was associated with particles (Fig. 2C). The Gag to Env ratio appeared to be similar in WT-, SOS- and UNC-VLPs and live activated HIV preparations ADA and MN in Western blots (data not shown). Concentrated VLPs preparations (1,000x) typically contain 5μg gp120 per ml (determined by comparison to a known JR-FL gp120 reference in Western blot) and 500ug total protein per ml, measured spectrophotometrically. Again, the 1000x concentrated live inactivated virus preparations were very similar. High-resolution negative stain electron microscopy revealed trimers on the surfaces of WT-VLPs (Fig. 2D), similar to those observed on inactivated SIV (Zanetti et al., 2006; Zhu et al., 2003; Zhu et al., 2006). The average spike density on several VLP electron micrographs was 27 (range 15–37), approximately twice the number previously counted on live inactivated HIV (range 7–14) (Chertova et al., 2002; Zhu et al., 2003; Zhu et al., 2006).

Analysis of Env oligomers

Analysis of Env liberated from WT- and SOS-VLPs in BN-PAGE revealed a major band of approximately 400kDa, corresponding to authentic trimers (Zanetti et al., 2006; Zhu et al., 2006), as well as a gp120/gp41 monomer band (Fig. 3A; (Moore et al., 2006)). This contrasts with soluble gp140_SOS that readily dissociates into monomers (Schulke et al., 2002), suggesting that expression in membranes lends some stability to Env trimers, as we initially hypothesized. The dominance of trimer band in both WT and SOS-VLPs suggests that gp120 shedding from WT-VLPs is not a major problem (Chertova et al., 2002; Zhu et al., 2003; Zhu et al., 2006). However, probing similar native PAGE blots with gp41 antibodies reveals gp41 stumps in WT that are not seen with SOS (Moore et al., 2006), suggesting that some gp120 shedding may occur (data not shown). The consistent production of gp41 immunodominant loop antibodies in natural infection suggests that gp41 stumps are a common antigen that might be best eliminated from immunogens, perhaps by the SOS mutation.

In contrast to WT- and SOS-VLPs, UNC-VLP-derived Env separated as several species that appear to be monomers, dimers, trimers and tetramers (Fig. 3A) (Earl, Doms, and Moss, 1990; Moore et al., 2006; Staropoli et al., 2000; Zhang et al., 2001). Denaturing but not reducing conditions caused WT and SOS trimers to dissociate into monomers (Fig. 3B), but UNC oligomers remained largely unchanged (Fig. 3B). Denaturing and reducing conditions eliminated oligomers in all preparations (Fig. 3C). SDS-PAGE gels run in parallel indicated that little, if any uncleaved Env existed in WT-VLPs (Fig. 3D) and confirmed that reducing conditions disrupted the gp120-gp41 SOS bond (compare Fig. 3D and E). Here, the gp120 from WT- and SOS-VLP appeared as a doublet. We previously proposed that the two forms of gp120 arise from trimers and putative monomers of gp120/gp41 present on virion surfaces (Moore et al., 2006).

Immunogenicity of VLPs in guinea pigs

In this pilot study, we wished to identify an immunization algorithm suitable for further development. Thus, we evaluated multiple concepts relating to immunogens and the end stage serum analysis. We immunized a total of 17 guinea pigs with recombinant JR-FL gp120, UNC-VLPs, SOS-VLPs, and naked-VLPs. Animals were immunized as outlined in Fig. 4A. Each was immunized 3 times at 6 week intervals with bleeds collected 1 or 2 weeks thereafter. Although VLPs can induce potent Ab responses without adjuvants (Lorin et al., 2004; Wagner et al., 1998), to ensure maximal Ab titers, we included adjuvants in our regimes (Hammonds et al., 2005). One consideration in selecting adjuvants was that those that are emulsion-based could disrupt membranes. We therefore selected immunostimulatory DNA CpG as our main adjuvant, which is unlikely to affect VLP membranes. However, the CpG formulations used were optimized for rabbits and mice, and it was not known if they would be effective in guinea pigs, for which a commercially available CpG product was not available. Therefore, in a few animals (P1, P13 and P14), CpG was supplemented with Ribi Ras3c or QS-21, as indicated (Fig. 4).

Binding titers to gp120 and a gp41 peptide

We measured serum gp120 titers over the course of immunizations. Of the 3 gp120-immunized control animals, P1 had the highest titer (Fig. 4A), perhaps owing to the QS21 adjuvant. In the Env-VLP-immunized animals, we observed high anti-Env titers, comparable to the A62 human plasma and gp120-immune sera (Fig. 4A). The kinetics of the responses to Env-VLPs was rapid: high titers were generated even after a single inoculation (compare animals S2 and S11 in Fig. 4B). The animals that received VLPs formulated in Ribi or QS21 as well as CpG generated titers similar to others, suggesting that CpG was sufficient for maximum titers. As expected, naked-VLPs did not elicit significant gp120 titers (Fig. 4A, B).

To evaluate the conformation dependency of antibody binding, we measured serum titers against denatured and reduced gp120 (dgp120). In most Env-VLP sera, dgp120 titers were <10% of those against intact gp120 (exceptions were S8 and S11), but in gp120-immunized animals, they were even lower. In contrast, a HIV-1 seropositive donor plasma exhibited similar titers to dgp120 and intact gp120. Taken together, this suggests that Env conformation was well preserved during adjuvant formulation and delivery and that Ribi or QS-21 formulation had similar effects on antigens compared to CpG alone.

To determine the effect of the SOS mutation on gp41 exposure, we measured antibody responses against a peptide derived from the immunodominant epitope of gp41 (the “AVERY” peptide). Most Env-VLP sera had very low titers. Exceptions were animals S7 and S8 that had been immunized with UNC-VLPs. As anticipated, the naked-VLP- and gp120 sera did not react with the peptide. In contrast, the HIV-1 seropositive plasma exhibited a high titer, in line with the use of similar peptides for HIV-1 diagnosis (Fig. 4A).

Mapping the specificity of immune sera

To map the immune sera, we used virus capture competitions and peptide ELISAs.

Analysis of neutralization activity

Reports on particle immunogens for HIV-1 have variably described neutralization or enhancement of infection by immune sera (Arthur et al., 1992; Buonaguro et al., 2005; Haffar et al., 1991; Hammonds et al., 2005; McBurney, Young, and Ross, 2007; Poon et al., 2005a; Poon et al., 2005b; Verrier et al., 2000). This may be in part related to the generation of antibodies against non-Env membrane proteins that can have unpredictable effects on infection. Indeed, it is often unclear whether “neutralization” is mediated by Abs directed to Env or to other membrane proteins (“anti-cell” antibodies). With this in mind, we evaluated neutralization in several formats, incorporating controls to help distinguish anti-Env activity from nonspecific effects.

We first measured neutralization against the index primary virus, JR-FL, in a luciferase assay using canine CF2 target cells (Fig. 6A). Some SOS- and UNC-VLP sera (S6, S11, S12, P14 and P15) enhanced infection by 2–3 fold (Fig. 6A, column 1), even at high dilutions (exemplified by serum P15 in Fig. 7A), while other Env-VLP sera and the gp120 sera had no effect. In contrast, the naked VLP sera exhibited apparent “neutralization”. In the following two sections, we separately address the activity of gp120/Env-VLP sera and naked VLP sera in the CF2 neutralization assay.

Neutralization activity of the gp120 and Env-VLP sera

Enhancement could arise either from antibodies directed to Env that somehow activate fusion or from antibodies directed to membrane components that somehow facilitate virus attachment. To investigate, we tested the effect of sera on SIVmac316 infection (Fig. 6A, column 2). This virus was made by transfection of the same 293T cells as the HIV pseudoviruses and, therefore, would be expected to bear the same set of cellular membrane proteins. Similar enhancement of this virus was noted with many of the sera, as it was for JR-FL, suggesting that the enhancement stemmed from Abs directed to non-Env membrane components of VLPs. In contrast, the control HIV+ plasmas A62 and N308 neutralized JR-FL, but not SIVmac316.

It has been suggested that enhancing Abs may mask neutralization (Giannecchini et al., 2001; Langlois et al., 1992). Indeed, in a previous study, removal of anti-cell antibodies against FIV particle vaccine by adsorption against producer cells was reported to uncover neutralizing activity (Giannecchini et al., 2001). To test this concept, we adsorbed SOS-VLP serum P15 against 293T cells and removed ~90% of the anti-cell reactivity, as assessed by FACS (data not shown), while retaining strong gp120 titers (diluted only 2–3 fold during adsorption). However, this did not uncover any neutralization (not shown).

To further determine whether enhancing Abs mask neutralization, we spiked the JR-FL-enhancing serum P15 with neutralizing mAb b12. Comparing neutralization by b12 alone to a b12/serum mixture (Fig. 7B) showed that the serum did not reverse b12 neutralization. Thus, infection by a virus that has been neutralized was not rescued by factors that otherwise enhance infection. In further neutralization assays using a VSV-G pseudotyped virus, the Env-VLP sera showed no enhancement (Fig. 6A, column 3), perhaps because VSV-G entry is amphotropic and therefore influenced by different factors compared to HIV and SIV.

Though we did not see convincing neutralization against the index virus, JR-FL, ELISAs had suggested that sera contained potent anti-Env reactivity (Fig. 4). To assess neutralization in alternative conditions, we measured “post-CD4” activity (Crooks et al., 2005) by pre-incubating JR-FL particles with sCD4 and graded dilutions of sera and then measuring infection of CF2 cells expressing only CCR5. We found that all Env-VLP and gp120 sera neutralized effectively in this format (Fig. 6A, column 4), as did both HIV+ plasmas. However, as expected, naked-VLP sera did not. Enhancement was not observed in the post-CD4 format, providing further evidence that factors that enhancing effects are negated when the virus is neutralized. With reference to our previous analysis (Crooks et al., 2005), post-CD4 neutralization suggested the presence of Abs that recognize the V3 loop and/or CD4-induced epitope(s), consistent with our mapping analysis (Fig. 5). The contribution of V3 Abs to neutralization was investigated in “peptide interference” neutralization assays (Beddows et al., 2005; Li et al., 2006; Selvarajah et al., 2005). Here, V3 peptides strongly inhibited post-CD4 neutralization by serum P13 (Fig. 7C), suggesting a major role for V3 Abs in post-CD4 neutralization.

Although we were encouraged by the strong serum activity in the post-CD4-format, it is important to point out that this assay dramatically expands access to epitopes that are normally cryptic in native trimers and become exposed only fleetingly during infection, as evidenced by the generally weak to modest activity of these types of antibodies in traditional neutralization assays (Bou-Habib et al., 1994; Crooks et al., 2005; Decker et al., 2005). Indeed, most HIV+ plasmas contain high concentrations of CD4i Abs without strong neutralizing titers. Therefore, the significance of post-CD4 activity should not be over-interpreted.

Given the lack of neutralization against the JR-FL index virus in the standard format, we tested some more sensitive isolates. The P1 serum strongly neutralized HXB2, a highly sensitive X4 isolate (Fig. 6A, column 5). Perhaps underlying this activity, mapping had indicated high levels of CD4bs-overlapping Abs in this serum (Fig. 5A). However, no Env-VLP other serum neutralized HXB2, probably because the predominant V3 Abs (Fig. 5) do not cross-react with this isolate.

The gp120 serum P1 weakly neutralized SF162, as did SOS-VLP sera S12 and P15, but no others (not shown). We also observed weak neutralization against ADA and BaL (not shown), but moderate neutralization against the 1196 isolate (Fig. 6A, column 6, and Fig. 7D). Epitope interference assays suggested that the 1196 neutralization was V3-mediated (Fig. 7E), consistent with the greater susceptibility of this virus to V3-directed neutralization (Binley et al., 2004).

Collectively, our neutralization data so far suggested that Env-VLP sera neutralize via anti-V3 Abs and the gp120 serum P1 neutralizes via CD4 binding site overlapping Abs. The lack of SF162 neutralization via the Env-VLP sera may stem from differences in the V3 loop between JR-FL and SF162. When we substituted a residue in SF162 to match the N-terminus of the V3 to JR-FL, the mutant virus was significantly more sensitive to Env-VLP sera (Fig. 6A, column 7). The SF162 variant was slightly less sensitive to the neutralizing N308 HIV+ plasma, revealing that the mutation did not globally increase neutralization sensitivity.

Analysis of neutralization by naked VLP sera

In light of the absence of JR-FL neutralization by gp120 and Env-VLP sera, it was not immediately clear why naked VLP sera potently “neutralized” (Fig. 6A, column 1). Evidence that the neutralization was non-specific came from the observation that the sera also neutralized SIVmac316 (Fig. 6A, column 2). In addition, interference of P18 neutralization of the 1196 isolate using V3 peptides had no effect (not shown), consistent with nonspecific neutralization.

Another method to help distinguish nonspecific activity from genuine neutralization is that the former may not embody the sigmoidal dose-response typical of neutralization. We and others have observed that IC90 titers typically occur at approximately a 10 fold higher Ab concentration than the IC50 (Binley et al., 2004). In contrast, the “neutralization” naked VLP sera took the form of “shallow” curves, with IC50 titers of ~1:500 in many cases, that never reached an IC90 (at a dilution of 1:20 or greater), suggesting nonspecific neutralization (Fig. 8A). Different isolates were affected to varying degrees by these non-specific antibodies (Fig. 6A, column 5) in a way we do not yet understand.

To determine whether non-specific neutralization influences genuine neutralization, we evaluated the effect of spiking a naked VLP serum (P18) with the b12 mAb at 2.5μg/ml. Here, b12 neutralization could be distinguished only when its concentration was sufficient to reach an IC90 (compare Figs. 8A, B).

One possible explanation as to why naked VLP sera nonspecifically neutralized while Env-VLP sera generally enhanced might be that the specificity or titer of anti-cell Abs in Env-VLP sera may differ from those in naked VLP sera. FACS analysis revealed that naked VLP serum P18 bound more potently to 293T cells than serum P12 (median fluorescence intensities at a 1:2,000 serum dilution were 717 and 181, respectively). These differences could underlie the differential effects of these sera in neutralization assays, where nonspecific neutralization might occur above a threshold of anti-cell Abs.

It is possible that the cells in which virus was produced and/or the target cells in which virus infection is measured might influence neutralization. We therefore assessed serum binding to various cell lines by FACS. We found that SOS-VLP serum P12 reacted extremely strongly with 293T cells, the same cell line used to produce VLPs (Fig. 8C) and almost as strongly against U87 cells (a human glioma cell line). However, its reactivity against canine CF2 cells used in the neutralization assays described above was much lower, though still significantly positive. This suggests that a fraction of the anti-cell activity in sera is cell line-specific.

Alternative neutralization assays

To investigate the effect of differential cell reactivity on neutralization, alternative neutralization assays were performed at a separate site (D.C.M.). In an assay measuring infection of 293T cell-produced virus in TZM-BL target cells (Fig. 6B columns 1–5), the nonspecific neutralizing activity of naked VLP sera was much lower and reactivity against the MuLV control virus was also low (Fig. 6B, column 5). Serum titers against SF162, JR-FL, 1196, or BaL pseudotyped viruses were, in general, higher (Fig. 6B, columns 1–4), than the corresponding titers using CF2 target cells (Fig. 6A and data not shown). Serum P1 was very potent against SF162 (Fig. 6B, column 1), consistent with the presence of CD4 binding site-overlapping antibodies. Neutralizing activity against JR-FL, though higher than before, remained relatively modest (Fig. 6B, column 2). Neutralization of 1196 and BaL was more impressive. Titers against 1196 were, in many cases, more than ten fold higher than in the initial CF2 assay (compare Fig. 6A, column 6 to Fig. 6B, column 3). It is possible that Ribi and QS21 adjuvants given to animals P13 and P14 contributed to especially high titers.

The above results were surprising, considering the greater overall anti-cell reactivity that would be expected against the human HeLa-based TZM-BL target cells. In addition to the cell line, differences in protocol may contribute to this effect. In the TZM-BL assay, but not the CF2 assay, sera, virus and cells are co-incubated during the entire infection period. We have found that this tends to increase neutralization IC50 titers (not shown). Another possibility may relate to coreceptor density. The CF2 cells in the initial assay express high levels of CCR5, to assist in efficient infection. This may decrease the window of opportunity for antibodies that neutralize via binding the V3 or coreceptor binding sites (Crooks et al., 2005). The lower CCR5 density on the TZM-BL cells could therefore contribute to the greater neutralization potencies. Finally, in the CF2 assay, we used a short spinoculation step to increase overall infection levels that was not used in the second assay. However, spinoculation did not significantly affect the activity of neutralizing mAbs (not shown).

In further neutralization assays employing PBMC-produced JR-FL primary viruses, also assayed on PBMC, we observed complete neutralization with most VLP sera at a 1:18 dilution, but little or no neutralization by the gp120 sera (Fig. 6B, column 6). We also observed strong activity against SIVmac251 in this format (Fig. 6B, column 7). It appears that, as in the CF2 assay toxicity adversely affected neutralization measurement. Taken together, these results highlight the unpredictable effects of anti-cell antibodies. Given that the rules governing these effects are far from clear, our results provide a cautionary note, suggesting that particular care should be taken when examining sera generated against membranous antigens.

Measurement of trimer binding in blue native PAGE

In circumstances when neutralization assays are compromised by high background activity, BN-PAGE offers an independent way examine trimer binding. We recently reported that a shift in the migration of a VLP-derived trimer band in BN-PAGE, associated with a depletion of unliganded trimer correlates with mAb neutralization (Crooks et al., 2005; Moore et al., 2006). Here, we adapted the method to measure neutralization by polyclonal sera and plasmas. We first tested the ability of the guinea pig sera to bind to JR-FL Env trimers in the presence or absence of additional neutralizing mAb b12 (Fig. 9A). We found that Env-VLP sera depleted the unliganded trimer to 43–96% of the control density, suggesting weak neutralization. Tellingly, the naked-VLP serum, P16 did not deplete unliganded trimers significantly, suggesting that nonspecific effects are inconsequential in this native PAGE assay. In experiments in which VLP sera and b12 were mixed together, the effect of b12 to deplete unliganded trimers was not affected (Fig. 9A). The shifted trimers were somewhat diffuse compared to when b12 was used alone, perhaps indicating multiple forms of trimer-Ab complexes.

For reference purposes, we analyzed the ability of the neutralizing plasma N308 and the non-neutralizing plasma L503 to shift JR-FL trimers. Like b12, N308 effectively depleted unliganded trimers (Fig. 9B). At low dilution, this was accompanied by a smear, that resolved into distinct shifted bands at higher dilutions, suggesting differing numbers of neutralizing Abs attached to trimers. In contrast, the non-neutralizing plasma L503 did not detectably deplete unliganded trimers, even at the lowest dilution, confirming that neutralization can be measured unequivocally by BN-PAGE, as defined by a loss of unliganded trimer, accompanied by a gain in higher molecular weight liganded forms.

We next determined whether Env-VLP sera could mediate more effective trimer shifts of isolates they effectively neutralized. In Fig. 9C, mAb b12 and the Env-VLP sera P13, P14 and S12 all shifted SF162 308T/H trimers to varying degrees. P13 was more effective than P14 and S12, as reflected in neutralization assays (Fig. 6A, column 7). On the other hand, the naked VLP serum P16 did not shift trimers, consistent with its lack of neutralization activity.

Strong reactivity of VLP sera and HIV plasmas to a non-Env component of VLPs

We examined the overall antibody reactivity with VLPs by comparing neutralizing mAbs, VLP sera, HIV patient sera and HIV negative sera as probes for VLPs in native PAGE/Western blot, revealing some remarkable findings (Fig. 9D). The neutralizing mAbs bound to the strong trimer band, and the weaker monomer band ((Moore et al., 2006). In contrast, the SOS-VLP serum bound extremely strongly to a ~150 kDa band and much less to the trimer, consistent with the idea that gp120/gp41 monomers are a prime target of B cells. Surprisingly, the naked VLP serum also bound extremely well to a band of the same size as the monomer. Since the naked VLP sera do not react with Env (Fig. 4), this indicates reactivity to a non-Env component of VLPs. A HIV positive patient serum gave similar results to the SOS-VLP sera: weak trimer binding, but strong reactivity to a band around the size of the monomer. In contrast, a HIV negative serum was essentially unreactive. Probing SIV-VLPs, we also observed HIV plasma reactivity to the ~150kDa protein, suggesting that VLPs and HIV infection elicit strong responses to a non-Env protein (Gag or host protein) component of VLPs. Overall, this further suggests that non-trimer components of particles are preferred targets for B cells both in immunizations and natural infection.

Nonfunctional gp120/gp41 monomers may be a prime Ab target on VLPs

In spite of very high binding titers to monomeric gp120 and neutralization activity against sensitive primary isolates, SOS-VLP sera did not neutralize the JR-FL parent virus or most other neutralization-refractive isolates. This mirrors the majority of vaccine studies in which immune sera neutralize only V3-sensitive primary isolates like BaL, SF162 and 1196 (Koff, Kahn, and Gust, 2007). Indeed, the neutralizing activity of VLP sera appeared to stem at least in part from V3 antibodies. The fact that the V3 loop is cryptic on trimers of neutralization-resistant viruses like JR-FL (Bou-Habib et al., 1994) raises the question: where do these anti-V3 Abs come from? Since our control immunogens, UNC-VLPs also elicited strong responses to the V3 loop, we suggest that, in each case, B cells responded to forms of Env other than functional trimers, perhaps gp120/gp41 monomers (SOS-VLPs) or unprocessed gp120-gp41 (UNC-VLPs). Thus, although SOS-VLPs eliminate gp120 shedding, the SOS mutation is only a partial solution toward presenting trimeric Env on VLPs for vaccine purposes.

Previous analyses have suggested that gp120/gp41 monomers, unlike trimers, these are highly accessible to V3 Abs (Burrer et al., 2005; Moore et al., 2006; Nyambi et al., 1998; Poignard et al., 2003) and may be a primary target for B cells. We hypothesize that monomers act as decoys, distracting the attention of B cells from the more compact trimers, in what might be considered an “immunogenic hierarchy”. Evidence of such hierarchies in studies of immune responses to HIV-1 have in fact been widely reported (Cleveland et al., 2000; Garrity et al., 1997; Jelonek et al., 1996; Liu et al., 2006). Perhaps at odds with this concept, affinity selection in germinal centers dictates that higher affinity B cell blasts proliferate at the expense of lower affinity ones. Thus, antibody responses to multiple separate proteins might be expected to develop independently and equivalently, as if they were given alone, as for the trivalent measles, mumps, and rubella vaccine (Usonis et al., 2005). However, the situation with VLPs is in fact quite different, because trimers and monomers are linked on the same particle, and therefore may be considered as a single target. Thus, B cells would be selected against only the most accessible foreign target(s), the V3 loop, of the most accessible structure (the monomer), at the expense of other epitopes, in this case, the trimer.

We previously showed that monomer is only a minor component of particles, compared to the predominant trimer (Fig. 3 and ref (Moore et al., 2006)). This raises the question why a minor species would dominate the B cell response. The answer may simply be that form rather than quantity dictates the antibody response. We previously showed that non-neutralizing antibodies capture HIV with high efficiency, even though they do not bind efficiently to trimers, suggesting that monomer binding mediates the capture. BN-PAGE further confirmed the more promiscuous binding properties of monomers (Moore et al., 2006). Thus, the monomer may dominate as an antigen, simply because it is a particularly accessible target.

Of particular significance to the present study, a previous report compared the immunogenicity of UNC and “SOSIP” mutants of a soluble gp140 glycoprotein and gp120 monomer in rabbits (Beddows et al., 2006). Sera generated against SOSIP stabilized trimers potently neutralized the parental JR-FL isolate. This activity was not due to V3 Abs, but its precise specificity remains unclear. Although the neutralization was type-specific, because the JR-FL isolate is refractive to neutralization, the results with SOSIP suggest a possible step in the right direction for HIV vaccine research.

Both SOSIP gp140 and SOS-VLPs vaccine approaches emerged from the early observation that SOS gp140 readily dissociates into monomers and a need to stabilize them as trimers (Schulke et al., 2002). The relative success of SOSIP gp140 raises questions about what might underlie the quite different findings of the present study. At odds with the idea that particles present native forms of Env, and therefore may be particularly effective immunogens, the results of these two studies might be interpreted to suggest that the reverse is, in fact, true. More likely, however, the differences probably arise from the fact that monomeric forms of Env were completely eliminated in SOSIP gp140 immunizations, but not in SOS-VLPs (Moore et al., 2006). However, since VLP trimers are recognized exclusively by neutralizing Abs (Moore et al., 2006), unlike even the most advanced soluble trimers (Dey et al., 2006), they may still have an advantage for presenting truly native trimers, if the right formulation can be found.

Factors affecting the quality of the immune response

We were cautious that commonly used adjuvants such as alum and Ribi might disrupt particle membranes, perhaps affecting Env conformation. Therefore, we selected immunostimulatory CpG as our primary adjuvant. Our mapping data indicate, however, that sera from the few animals that received VLPs in emulsion adjuvants differed little from those who received CpG alone. There is a suggestion, however, that the additional adjuvants lead to higher neutralization titers against certain viruses (exemplified by the activities of sera P13 and P14 in Fig. 6). Therefore, adjuvants may play a significant role in getting the most out of immunogens. In contrast to Env-VLP sera, naked VLP sera all showed potent nonspecific neutralization in certain neutralization assays. The lack of Env on these particles may have driven a potent anti-cell Ab response. Further studies should help us to fully understand these effects.

The route of immunization could affect the stability of VLPs after immunization, perhaps also affecting the quality of the antibody response. Intranasal administration, for example, has been shown to be effective (McBurney, Young, and Ross, 2007). However, since the nAb titers in other studies have not been dramatically improved, the more fundamental problem of non-functional forms of Env may need to be solved before the extent of the advantages offered by varying route and adjuvant can be fully appreciated. Another, perhaps more significant factor is the choice of model species for vaccation trials, based partly on the notion that some species may simply not have the repertoire to generate nAbs (Koff, Kahn, and Gust, 2007). Indeed, we have recently begun VLP immunizations in macaques that suggest a greater focus on the CD4bs and improved neutralization activity. Developments in mapping technology, we attempted here, should in future help us to make more informed adjustments to our vaccines.

The effects of “anti-cell” antibodies

A drawback of membranous immunogens is that they induce Abs against non-Env components, exemplified by the strong reactivity of VLP and HIV patient sera to a non-Env VLP band in BN-PAGE (Fig. 9D). Antibodies against non-Env VLP surface components can enhance or non-specifically “neutralize” infection in a largely unpredictable manner (Arthur et al., 1992). We incorporated controls at various levels in neutralization assays to identify any nonspecific activity. However, the cleanest method to assess neutralization was the native trimer shift assay (Fig. 9A–C) that should help in assessing the progress of any membranous HIV vaccine.

The differences between IC50s in assays of the two laboratories that tested VLP sera were somewhat surprising. Perhaps related to this, a recent proficiency test of the TZM-BL neutralization assay in multiple laboratories (D.C.M., unpublished) has illuminated unexpected discrepancies that appear to implicate pseudovirus stock preparation as an important variable. Further proficiency testing is now underway to understand this issue toward the ultimate goal of standardizing neutralization assays.

The future of HIV-VLP immunogens

In natural infection, the delay in development of nAbs (Deeks et al., 2006), despite the early development of gp120 binding antibodies suggests that nonfunctional forms of Env provide the virus with a valuable fitness advantage by diverting antibody responses from functional targets. To make progress, we may need to get beyond these defenses (Burton and Parren, 2000). It may be possible to refocus Ab responses against trimers (Moore et al., 2006) by modifying or blocking epitopes on nonfunctional Env (Cole et al., 2004; Garrity et al., 1997; Keller and Arora, 1999; Selvarajah et al., 2005; Srivastava et al., 2003). However, it could be argued that trimers are inherently poor immunogens, and that removal of irrelevant forms of Env may not change that fact. On the other hand, the development of cross-neutralizing activity in a subset of infected HIV patient sera, however, suggests that, eventually, B cells can in fact respond to trimers and the neutralizing activity of SOSIP gp140 sera provides evidence that trimers are not inherently poor antigens. Thus, if we can find a way to steer B cells away from irrelevant targets, we may be able to entice responses against native trimers.

Plasmids and mutagenesis

The plasmid pCAGGS was used to express membrane-bound forms of the primary R5 isolate, JR-FL (Binley et al., 2003; Moore et al., 2006). We expressed full-length gp160 and gp160ΔCT, truncated at amino acid 709, leaving 3 amino acids of the gp41 cytoplasmic tail. The SOS and UNC mutants to stabilize gp120-gp41 association have been described previously (Binley et al., 2000). Gp160-expressing plasmids of isolates SF162, BaL, HXB2, ADA, and 1196 were obtained from Drs. Leo Stamatatos, David Montefiori, and the NIH AIDS Repository. A mutant of SF162 Env in which the threonine residue at 308 of the V3 loop was exchanged for histidine (SF162 308T/H) was created by Quikchange mutagenesis. Truncated gp160ΔCT versions of 1196, SF162, SF162 308T/H, and SIVmac316 Envs were cloned into pCAGGS in a similar manner to that described for JR-FL (Binley et al., 2003). The plasmid pVSV-G was obtained from Dr. Nathaniel Landau. Sub-genomic plasmids pNL4-3.Luc.R-E- and pSG3ΔEnv have been described previously (Binley et al., 2002; Li et al., 2005a).

MAbs, soluble CD4, recombinant gp120 and HIV+ donor plasmas

Anti-gp120 monoclonal antibodies (mAbs) included b12 and 15e, directed to epitopes that overlap the CD4 binding site (CD4bs) (Burton et al., 1994); 2G12, directed to a unique glycan-dependent epitope on gp120 (Sanders et al., 2002a; Scanlan et al., 2002); X5 directed to a CD4-inducible (CD4i) epitope (Darbha et al., 2004; Labrijn et al., 2003); LE311, 39F and PA1 directed to the V3 loop (Crooks et al., 2005; Schulke et al., 2002); and B12, directed to an epitope in the C2 domain of gp120 that is preferentially exposed on denatured forms of the molecule (Moore et al., 2006). Anti-gp41 mAb 7B2 is directed to the cluster I region (Binley et al., 2000).

MAb 2G12 was provided by Dr. H. Katinger (Polymun Scientific Inc., Vienna, Austria). MAbs 39F, LE311, and 7B2 were provided by J. Robinson (Tulane University). MAb B12 was provided by Dr. George Lewis (Institute of Human Virology, Baltimore, MD). MAb PA1, four-domain soluble CD4 (sCD4) and JR-FL gp120 were provided by Progenics Pharmaceuticals Inc (Tarrytown, NY). MAbs b12 and 2G12 are broadly neutralizing (Binley et al., 2004); mAb X5 neutralizes primary isolates in the presence of soluble CD4 (Crooks et al., 2005; Moulard et al., 2002); the V3 loop-specific Abs neutralize a subset of primary isolates (Binley et al., 2004); mAb 7B2 is nonneutralizing. The HIV-1-infected donor plasmas A62, L503 and N308 were described previously (Crooks et al., 2005). The anti-p24 serum was obtained from the NIH AIDS Reagent Program.

Peptides

Eleven residue-overlapping 15-mer peptides of the V3 loop of JR-FL were obtained from the NIH AIDS Reagent Program. Five residue-overlapping 15-mer peptides of the V1/V2 loop of JR-FL were synthesized at The Torrey Pines Institute for Molecular Studies and purified to >90%. A peptide of the gp41 immunodominant region, sequence RVLAVERYLKDQQLLGIWGCSGKLIC, termed the “AVERY” peptide was purchased from AnaSpec (San Jose, CA).

Production of VLPs

VLPs were produced by transient transfection of 293T cells with plasmids pNL4-3.Luc.R-E- and a pCAGGS-based Env-expressing plasmid by calcium phosphate precipitation. Two days later, supernatants were collected, precleared by low speed centrifugation, filtered through a 0.45 μM filter and pelleted at 50,000 xg in a Sorvall SS34 rotor. To remove residual medium, pellets were diluted with 1ml of PBS, then recentrifuged in a microfuge at 15,000 rpm. VLPs were then resuspended in PBS at 1000 x the original concentration. VLPs were referred to as SOS-VLPs, UNC-VLPs, WT-VLPs, depending on the form of Env on their surfaces, or as “naked-VLPs” bearing no Env (produced by transfecting pNL4-3.Luc.R-E- alone). Particles were inactivated using aldrithiol (AT-2) (Rossio et al., 1998), after which they were re-centrifuged and washed with PBS.

Immunoprecipitation

Supernatants of 293T cells were metabolically labeled and analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), as previously described (Binley et al., 2000). Reduced and nonreduced samples were prepared by boiling for 5 min in Laemmli sample buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 25% glycerol, 0.01% bromophenol blue) in the presence or absence of 20mM dithiothreitol (DTT).

SDS-PAGE, Native PAGE and Western blot

VLP preparations were separated by SDS-PAGE, and Western blots were probed with mAbs PA1 and B12 diluted to 1 μg/ml in PBS containing 2% nonfat milk, detected with a goat anti-mouse alkaline phosphatase conjugate (Jackson) and developed using BCIP/NBT colorimetric reagents (Sigma-Aldrich). VLP-derived Env was also analyzed by Blue Native PAGE (BN-PAGE), as described previously (Moore et al., 2006).

Electron Microscopy

Transfected 293T cells producing VLPs were fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.3) for 2 h. Samples were then washed in 0.1 M sodium cacodylate, fixed in 1% osmium tetroxide and incubated in 0.5% tannic acid in 0.05 M cacodylate for 30 mins. Following a 10 min wash with 1% sodium sulfate/0.1 M cacodylate buffer and further rinsing in 0.1 M cacodylate, the cells were dehydrated in a graded ethanol series, cleared in propylene oxide and embedded in EMbed-812/Araldite (Electron Microscopy Sciences, Fort Washington PA). Ultra thin sections were cut and mounted on parlodion-coated copper grids, stained with uranyl acetate and lead citrate. The sections were then examined on a Philips CM-100 electron microscope (FEI, Hillsborough, OR) and documented on Kodak SO-163 photographic film.

In high magnification negative stain electron microscopy, aliquots (25 μl) of virus particle preparations were washed in 140 μl of PBS and pelleted by an Airfuge centrifuge (Beckman Coulter) at 100,000 xg for 15 min. The virus particles were then resuspended and fixed for 30 min in 20 μl of 2.5% glutaraldehyde/PBS at 4°C. The fixed virus particles were washed in PBS and pelleted in Airfuge centrifuge again, and finally resuspended in 15 μl of saline. 4.5 μl of resuspended sample was placed on a carbon-covered grid for 30 sec. Excess virus particles and buffer were removed by filter paper blotting, and 5 μl of 1% uranyl formate stain was immediately added and incubated for 1 min. Excess stain was then removed, and the samples were allowed to air dry. Sample grids were screened at x 30,000, and electron micrographs were recorded at a nominal magnification of x 100,000 at 100 kV on a JEOL JEM 1200EX electron microscope.

Immunogenicity of VLPs in guinea pigs

Dunkin Hartley guinea pigs were immunized with SOS-VLPs, UNC-VLPs, “naked” VLPs or gp120, at the Scripps animal facility or Pocono Rabbit Farm (denoted by the animal identifier prefixes “S” and “P”). Five microgram Env equivalents of VLPs were used for each dose, as determined by Western blot, comparing to the known monomeric JR-FL gp120 standard. Naked VLPs were normalized for equivalent p24 to the other VLP preparations in Western blot. Recombinant monomeric gp120 was initially used at 5 μg dose (immunizations 1 and 2) consistent with the dose of Env in Env-VLP vaccinations. However, due to weak responses, the third and final dose was increased to 50 μg. Most immunogens were formulated in an equal mixture with the CpG-based adjuvant ImmunEASY (Qiagen) optimized for rabbits and mice. In some immunizations, the saponin-based QS21 or Ribi Ras3c adjuvants were used. All immunizations were administered by a combination intradermal and intramuscular routes in two locations each, at days 0, 43 and 97. Sera were collected 1 or 2 weeks after each immunization.

Neutralization assays

Sera were heat-inactivated at 56°C for 30 min, then analyzed for neutralization. Each assay was performed in duplicate and repeated at least 2 times. Representative data are shown. HIV+ plasmas A62 and N308 were used as reference controls. Neutralization assays were performed using either Env-pseudotyped viruses produced by transfection of 293T cells, or PBMC-grown viruses.

Adsorption of sera against 293T cells

Sera were adsorbed against 293T cells to remove reactivity against membrane-associated proteins (Giannecchini et al., 2001; Langlois et al., 1992). Briefly, pellets containing 10⁸ 293T cells were mixed with 200 μl of serum and mixed overnight at 4 °C. Cells were then pelleted and the serum retrieved. The process was repeated 2 more times. The supernatant serum was completely cleared of cells, and then assayed by ELISA to detect anti-gp120 Abs compared to unabsorbed sera, to determine whether anti-gp120 titers were affected by adsorption.

Flow cytometry

Guinea pig serum binding to 293T, CF2 and HOS cells was assessed. Cells were incubated with sera at graded dilutions for 1 h at RT. A FITC conjugate (Jackson, West Grove, PA) was used to detect guinea pig IgG binding by FACS.

Analysis of serum binding titer and specificity by ELISA

We measured anti-gp120 titers of guinea pig sera by ELISA, as described previously (Binley et al., 1997). A HIV+ donor plasma, A62, was included as a control (Crooks et al., 2005) A goat anti-guinea pig or human alkaline phosphatase conjugate (Accurate) was used to detect bound Ab using the AMPAK system (Dako). Reactivity against epitopes exposed on denatured gp120 was measured against captured gp120 that had been treated with 1% SDS and 50mM DTT and boiled for 2 min and then diluted in PBS before capture on D7324-coated plates. Peptide ELISAs were performed in a similar manner.

Serum mapping by inhibition of mAb-virus capture

To determine the specificity of guinea pig sera, we examined their ability to inhibit JR-FL WT-VLP capture by a panel of mAbs, as described previously (Derby et al., 2006). VSV-G was co-expressed on particle surfaces to enhance detection of captured virus and also to provide a read-out of capture that is essentially unaffected by possible serum neutralization (Moore et al., 2006). Briefly, mAbs were coated on ELISA microwells overnight at 5 μg/ml. Wells were then washed and blocked with 3% bovine serum albumin (BSA) in PBS. Graded dilutions of sera were incubated with JR-FL WT-VLPs bearing VSV-G for 30 min. For X5 mAb capture, a fixed concentration of 5μg/ml sCD4 was added throughout. Mixtures were then added to mAb-coated ELISA wells for 3 h followed by washing with PBS (Poignard et al., 2003). CF2.CD4.CCR5 cells (Kolchinsky, Kiprilov, and Sodroski, 2001) were overlaid and two days later, infection was measured by assaying luciferase.