Characterizing anti-HIV monoclonal antibodies and immune sera by defining the mechanism of neutralization

Abstract

Understanding the nature of neutralization may provide information for crafting improvements in HIV vaccines. Using JR-FL as a prototype primary pseudovirus, we first investigated anti-HIV monoclonal antibodies (mAbs) in several neutralization formats designed to elucidate the timing of neutralization. MAb b12 was most effective before receptor binding, 2G12 neutralized effectively even after CD4 binding, and X5 and a V3 loop mAb (LE311) were inactive in a standard format but were induced by sCD4. Consistent with this latter finding, native PAGE indicated that X5 and V3 mAb binding to Envelope trimers was dependent on sCD4 binding. In contrast, 2F5 and 4E10 were active even post-CD4/CCR5 engagement. We next analyzed the neutralization mechanism of a panel of HIV+ donor plasmas of various potencies. All mediated high levels of post-CD4 neutralization that was not associated with activity in the standard format. None, however, neutralized effectively in the post-CD4/CCR5 format, suggesting that 2F5/4E10-like Abs were absent or at low concentrations. Finally, we analyzed a non-neutralizing plasma spiked with mAbs b12, 2G12 or 2F5, which resulted in increases in neutralization titers consistent with the activities of the mAbs. We conclude that these methods, together with other mapping approaches, may provide a better understanding of neutralization that could be useful in vaccine research.

INTRODUCTION

Despite a massive effort over more than 2 decades, even the most promising HIV-1 vaccine candidates are unable to elicit high titers of neutralizing antibodies (nAbs) that are considered crucial for success [12,57,65]. NAbs block Envelope (Env) binding to receptors and/or fusion [57,65]. The functional Env target of nAbs on the virus surface consists of trimers of non-covalently associated gp120/gp41 heterodimers, in which gp120 is the surface subunit and gp41 is the transmembrane subunit [39]. As depicted in Fig 1, during HIV-1 infection, Env attaches to target cells, binding CD4, then a coreceptor. Subsequently, the gp41 fusion peptide penetrates the target cell membrane, leading to fusion and infection. The difficulty of generating Abs that recognize the trimer target is reflected by the fact that broad neutralization of primary isolates is achieved by only a handful of human mAbs identified to date, including b12, 2G12, 2F5 and 4E10, all isolated from HIV+ human donors. MAb b12 recognizes an epitope overlapping the CD4 binding site (CD4bs) of gp120 [13]; 2G12 recognizes a specific array of high mannose structures on gp120 [51,52]; 2F5 and 4E10 recognize epitopes in the membrane proximal ectodomain region (MPER) of gp41 [15,29,44,45,67]. A vaccine able to induce Abs resembling any of these would be highly desirable.

Fig 1. Depiction of HIV-1 fusion and modified neutralization assays.

The major steps in HIV fusion are shown in cartoon form: (1) pre-attachment, (2) CD4 binding (3) coreceptor binding (4) membrane fusion. The SOS disulfide bond is shown as a red bar between gp120 and gp41 that, after CD4 and CCR5 binding (3), can be broken by treatment with a low concentration of reducing agent. The various neutralization formats are depicted as bars along the top. The neutralizing activities of various mAbs and T-20 (derived from the analysis in Table 1) are depicted by bars along the bottom.

A significant problem with all Env-based vaccine immunogens so far appears to be their tendency to elicit sera that focus on epitopes that are not exposed on native trimers, rather than the intended neutralizing targets. Indeed, tailoring immunogens to exhibit favorable antigenic properties may be insufficient, because the relationship between antigenicity and immunogenicity is poorly understood. As a result, vaccine research is largely a process of informed “trial and error”. To make further progress, a rational approach may be crucial. One way to drive vaccine research might be to unravel the specificities that determine neutralization (or lack thereof) in HIV+ donor and vaccinee sera. This may facilitate informed vaccine improvements in successive rounds of immunization.

In many vaccine studies, the evaluation of immune sera typically involves testing gp120 binding and neutralization. However, attempts to fully profile sera have been infrequent, perhaps in part owing to the challenge of simultaneously evaluating multiple specificities in a single sample. In light of a growing realization of the importance of mapping, some studies have begun to address the challenge [3,4,19,21,27,33,37,41,54,59,64]. These methods can be generally divided into those that address total binding Abs and those that examine the neutralizing fraction.

One method to investigate binding Abs has been to measure reactivity to peptide collections spanning the Env primary sequence. Although this provides some useful information, it overlooks discontinuous epitopes that constitute a dominant fraction of total Env binding [41]. Another method to examine binding Abs is to measure reactivity to intact and denatured forms of an immunogen [59], to determine the extent to which the test protein might have retained its conformation upon adjuvant formulation and immunization. A strong reactivity to denatured Env may indicate a poor retention of conformation. Another method is to examine the ability of a serum to inhibit the binding of conjugated mAbs directed to known epitopes displayed on recombinant Env proteins in competition ELISA [54]. Yet another method is to examine serum reactivity with mutated Env proteins compared to the index immunogen [54].

Among methods to investigate the neutralizing fraction of sera, one approach has been to try to adsorb certain neutralizing Abs in epitope interference assays. For example, peptides or Env protein subunits can be added to neutralization assays to determine their inhibitory effect, if any, on neutralization IC50 [4,27,37,54]. Another approach is to fractionate neutralizing sera using Env proteins immobilized on column matrices, and then determine which fraction(s) contain neutralizing activity. Recently, an approach was reported to evaluate the CD4-induced fraction of Abs in sera [19]. In addition, 2F5 and 4E10-like Abs were detected in some HIV+ donor sera using pseudovirions that bear HIV-2 Env proteins engrafted with the HIV-1 2F5/4E10 epitopes [5].

Here, we describe an analysis of neutralization mechanism to complement existing mapping methods [1,5,6,18,24,28]. We show that altered neutralization assay conditions and Env mutations can differentially affect the accessibility of epitopes. Thus, we analyzed various mAbs, a small panel of HIV+ donor plasmas, and a non-neutralizing plasma spiked with various neutralizing mAbs in a series of assay formats to try to decipher the nature of their activities. To visualize the mechanism of some mAbs, we examine mAb binding to functional trimers derived from viruses in blue native PAGE (BN-PAGE) band shift experiments [53].

MATERIALS AND METHODS

Materials

MAbs, sCD4 and T-20

We assembled a panel of mAbs whose epitopes are described with reference to the concept that gp120 comprises a conserved core consisting of domains C1 through C5, separated by variable loops V1 to V5 [56]. The mAbs included the potently neutralizing mAb b12 and non-neutralizing mAb 15e [13] (directed to the CD4bs), neutralizing mAb 2G12 [14,52], mAb X5 (directed to a CD4-induced epitope of gp120; CD4i) [35], mAb LE311 is a new human anti-V3 mAb, generated according to the methods described in ref [62] and 447-52D (directed to the V3 loop of gp120) [25,65], 2F5 and 4E10 [9,29,44,67]; Monovalent Fabs (for BN-PAGE) were in some cases prepared by digesting the whole IgG with immobilized pepsin (Pierce) [43]. Single chain Fv (scFv) X5 was expressed in bacteria. Four domain soluble CD4 (4D-sCD4) consists of all the 4 outer domains, whereas two domain sCD4 (2D-sCD4) consists of the N-terminal domains 1 and 2. The T-20 peptide recognizes the pre-hairpin intermediate conformation of gp41 [48].

Plasmas

We obtained human HIV-positive plasmas from clade B-infected donors coded as N160, K370, TN11, L503, N308, L909, TN15 and J864 [49], A62, R2, L92 and 739, and #11H. These plasmas came from various clinical cohorts in the US and Europe. A subset of these plasmas N308, A62, #11H, L92, L909, R2 and 739 were selected for their previously known neutralizing activities [49,64]

Cell lines

293T cells were used for transfections. CF2.CD4.CCR5 and CF2.CCR5 cells were used for neutralization assays [34].

Plasmids and mutagenesis

pCAGGS [6] was used to express Env from JR-FL, a clade B primary isolate of moderate neutralization sensitivity. The “SOS” mutation to introduce a gp120-gp41 intermolecular disulfide bond has been described [6,8], as has a gp160 cytoplasmic tail truncation leaving 3 amino acids, gp160ΔCT [6]. The plasmid pNL-LucR-E- [6] expresses an HIV-1 genome truncated in the Env and Nef genes, has a frameshift in Vpr and carries the luciferase gene.

Viruses

Virus-like particles (VLPs) were produced by transient transfection of 293T cells with plasmid pNL-LucR-E-, and a pCAGGS Env plasmid [6]. Transfection supernatants were used directly for neutralization assays. Concentration, as necessary, was achieved by centrifugation, while preserving virus infectivity, as described [43].

Methods

Neutralization mechanism analysis

We developed a series of neutralization assay formats (Fig 1), some of which were described previously [6]. We used VLPs bearing full-length gp160 or gp160ΔCT Envs in WT or SOS form. SOS-VLPs have a disulfide bridge which links gp120 and gp41, preventing conformational changes after CD4/CCR5 binding that are necessary for fusion. This results in an SOS-arrested intermediate that can only proceed to full infection after exposure to 5mM DTT to reduce the disulfide bond [6]. All assays involved a 15 minute ‘spinoculation’ step, followed by a 2h infection at 37°C [6]. Assays were performed at least 4 times and representative data is shown.

Descriptions of the individual formats are as follows:

• i) Standard format. The standard protocol provides a reference for neutralization in other formats. Essentially, Ab and virus were mixed for 1 h before adding to target cells for 2 h. SOS-VLPs were treated exactly as WT-VLPs, except that infection was triggered after the virus-cell incubation using 5mM DTT [6]. Formation of the CD4/CCR5-bound SOS-arrested intermediate allows the steps of infection before and after CD4/CCR5 binding to be separated (Fig 1). An optional wash step before the DTT treatment effectively isolates these two phases of infection [6].

• ii) Pre-attachment format. Neutralization was assessed by incubating virus with Ab, followed by centrifugal pelleting in a microfuge (20,000 × g for 30 min) and washing to remove any Ab that was unable to bind trimers in their native state, i.e. before adding virus to target cells [6].

• iii) Post-CD4 format. To measure neutralizing activity between the CD4 and CCR5 binding steps, we used a post-CD4 format, similar to that described previously [19]. VLPs were premixed with sCD4 at 3 μg/ml for 15 min at 37 °C, then for 1 h with Ab, before adding to cells that express only coreceptor (CF2.CCR5 cells) [34].

• iv) Post-CD4/CCR5 format (SOS only). Post CD4/CCR5 binding neutralization was measured using SOS-VLPs, as previously described [6]. Essentially, SOS-VLPs were allowed to attach to target cells for 2 h. Unbound VLPs were washed away and graded concentrations of mAbs were added. Following a 1 h incubation, infection was activated using 5mM DTT for 10 minutes [6].

• v) Temperature-arrested format. Full-length gp160 WT-VLPs were allowed to attach to target cells at 14°C for 2 h. The medium was then replaced, and the cells were incubated with Ab for a further 1 h at 14°C before being washed PBS washing, replacing the medium and warming to 37°C to allow fusion to proceed.

BN-PAGE

BN-PAGE band shifts [43,46,53], were performed by incubating VLPs with a Fab, scFv and/or sCD4 for 10 minutes [53], then adding 2x solubilization buffer (1% Triton X-100/1% NP40 in 1mM EDTA) in 1.5M aminocaproic acid [43]. An equal volume of sample buffer (100mM morpholine sulfonic acid, 100mM Tris-HCl, pH 7.7, 40% glycerol, 0.1% Coomassie blue) was then added. Final concentrations of 5μg/ml sCD4, 30μg/ml Fab or scFv were present in the sample at loading onto a 4-12% Bis-Tris NuPAGE gel (Invitrogen). Ferritin (Amersham) was used as a size standard, producing two bands of 220kDa and 439kDa. Electrophoresis was performed at 4°C for 3 h at 100 V as described [43], followed by Western blotting. Excess Coomassie blue dye was removed by methanol washing. The blot was probed using 1μg/ml each of mAbs 2G12, b12, X5 and LE311 in 4% nonfat milk/PBS for 30 min then a goat anti-human Fc alkaline phosphatase conjugate (Jackson).

RESULTS

MAb neutralization mechanism

We, and others previously analyzed the breadth and potency of a mAb panel in a VLP-based neutralization assay [9]. Here, we determined the neutralization mechanism of a similar panel of mAbs. To facilitate comparisons, VLPs bearing JR-FL Env from a primary clade B virus of moderate neutralization sensitivity were used throughout.

The standard neutralization format [6] measures activity over the entire infection period (Fig 1), serving as a reference for other formats that emphasize various stages of infection (Fig 1). In some cases, VLPs bearing mutated Env were used [6]. Cytoplasmic tail truncation (gp160ΔCT) has been reported to increase exposure of certain epitopes and increase fusion kinetics [2,23,61]. The SOS mutation introduces a disulfide bond between gp120 and gp41, which arrests infection after CD4/CCR5 binding, pending the addition of a low concentration of reducing agent to allow infection to progress to completion [1,6]. This allows separation of the pre- and post-CD4/CCR5 phases of infection. The mAb IC50 neutralization activities in each format are shown in Table 1 and summarized in Fig 1.

TABLE 1. Analysis of mAb neutralization mechanism.

Neutralization activity against JR-FL-based VLPs was analyzed in various formats, organized vertically, as indicated in the leftmost column. The Env gene used (full-length gp160 or gp160ΔCT, and WT or SOS) and the use of a wash step (before DTT addition in SOS mutant viruses) is indicated (columns 2-4). IC50 data are given in μg/ml. To assist in comprehension, neutralization has been color-coded, as shown. N.D.=not done.

	column	1	2	3	4	5	6	7	8	9	10
	format	standard					pre-attach.	post-CD4	post-CD4/CCR5		temp.-arrest.
	160/160ΔCT	160	160ΔCT	160	160ΔCT	160ΔCT	160ΔCT	160ΔCT	160ΔCT	160ΔCT	160
	WT/SOS	WT	WT	SOS	SOS	SOS	WT	WT	SOS	SOS	WT
IC50 titer:	wash	−	−	+	−	+	−	−	−	+	−
>30μg/ml	b12 (CD4bs)	0.02	0.04	0.09	N.D.	0.02	0.1	0.08	>30	N.D.	>30
10-30μg/ml	2G12 (high mann.)	0.75	0.2	N.D.	N.D.	0.14	N.D.	0.05	>30	N.D.	N.D.
1-10μg/ml	X5 (CD4i)	>30	>30	N.D.	N.D.	>30	N.D.	0.0003	>30	N.D.	N.D.
0.1-1μg/ml	LE311 (V3)	>30	>30	N.D.	N.D.	12	N.D.	0.0003	>30	N.D.	N.D.
<0.1μg/ml	2F5 (MPER)	2.16	0.2	0.9	0.02	0.03	0.8	0.02	0.3	0.4	2
	4E10 (MPER)	>30	2	4	N.D.	0.8	N.D.	0.12	3	N.D.	N.D.
	T-20	0.03	0.03	N.D.	0.01	0.2	2	0.02	0.03	0.3	0.005

Evaluation of mAbs in the standard format using full-length gp160 WT-VLPs revealed that b12, 2G12, and 2F5 all effectively neutralized. However, mAbs directed to the V3 loop (LE311) and CD4-induced epitopes (CD4i; X5) did not neutralize at <30μg/ml, suggesting that these epitopes are cryptic in JR-FL, perhaps due to V1V2 loop masking [10,35,47,62]. The lack of 4E10 neutralization at <30μg/ml, is also consistent with previous data [9]. MAb 15e (CD4bs) also did not neutralize, as expected (not shown).

To measure neutralizing activity between CD4 and CCR5 engagement, we used a “post-CD4” assay, that involved complexing VLPs with sCD4, adding graded concentrations of mAb and then adding this mixture to cells that express CCR5 [19] (Fig 1). To evaluate the method, we compared the ability of virus that had or had not been pre-incubated with sCD4 to infect CF2.CCR5 cells. We observed post-CD4 infection levels equivalent to ∼10% of those obtained in the standard format. When sCD4 was omitted, infection on CF2.CCR5 cells was reduced to background levels (data not shown). Thus, sCD4 appeared to effectively prime the virus for infection. Infection of CF2.CD4.CCR5 cells by sCD4-engaged virus was also observed (not shown), but to ensure separation of the CD4 and CCR5 binding steps we used CF2.CCR5 cells.

We compared mAb activities in each modified format, as follows:

b12 (CD4bs)

As might be expected, mAb b12 was slightly less potent in the post-CD4 format using gp160ΔCT WT-VLPs (Fig 2, Table 1, column 7), compared to the standard format (Table 1, column 2), consistent with sCD4 overlapping b12's epitope on gp120. It is perhaps surprising that blocking of b12 was not more pronounced (Fig 2), considering how sCD4 efficiently blocks b12 binding to monomeric gp120 [42]. It may be that sCD4's monovalency and perhaps its lower affinity allows it to be displaced by b12 IgG. In contrast, the pre-attachment assay favored b12 compared to 2F5 and T-20 (Table 1, compare columns 2 and 6), whereas the post-CD4/CCR5 formats strongly disfavored b12 neutralization (Table 1, column 8). Together, this data suggests that the optimum target of b12 is the native, unliganded trimer.

Fig 2. Comparison of standard and post-CD4 neutralization.

MAb activity in standard (filled circles) and post-CD4 (open circles) against JR-FL 160ΔCT WT-VLPs. A) b12, B) 2F5, C) LE311, and D) X5.

2G12 (high mannose gp120 epitope)

The activity of 2G12 was modestly induced in the post-CD4 format compared to the standard format (Table 1, compare columns 1,2 and 5 to column 7). 2G12 did not, however, neutralize after CD4/CCR5-engagement (Table 1, column 8), consistent with neutralization by blocking Env-CCR5 binding.

X5 (CD4i)

Although X5 can neutralize certain isolates [35,64], it did not neutralize JR-FL in the standard format (Table 1, columns 1-5), but showed dramatic neutralization in the post-CD4 format (Fig 2, Table 1, column 7), consistent with previous studies [19,20,50] and with a similar sCD4 induction of its epitope on monomeric JR-FL gp120, observed in ELISA [35]. This cooperative X5 and sCD4 binding to functional Env trimers is visualized in further experiments described below (Fig 3). X5 had no activity, however, in the post-CD4/CCR5 format, reflecting the contribution of this epitope to CCR5 binding [58,60,63]. That X5 did not neutralize in the standard format (Table 1, columns 1 and 2) suggests that the post-CD4, pre-CCR5 phase of infection is a very narrow window of opportunity for neutralization of the JR-FL virus. Truncation of the gp160 cytoplasmic tail or SOS mutation did not appear to increase X5's activity. This is perhaps surprising, considering that cytoplasmic tail deletion was previously found to better expose the CD4i epitope in an HXB2 Env [22,23,31]. The extent of the tail truncation and the particular properties of the Env clone may account for this difference [61].

Fig 3. BN-PAGE shift assay to visualize mAb binding to Env proteins.

SOS-VLPs (A-C) were incubated with various Fabs or a scFv (final concentration=30μg/ml) and/or two- (2D) or four- (4D) domain sCD4 (final concentration 10μg/ml) then processed for BN-PAGE/Western blot. The symbol (−) indicates no Fab was added. A) SOS-VLPs were incubated with 2D-sCD4 and Fab X5 alone and in combination. Bands are identified in cartoon form as gp120/gp41 monomers and trimers, where red circles depict gp120, green boxes represent gp41, and a blue triangle represents the SOS bond [43] (lane 1), and as complexes with 2D-CD4 (string of green circles in lane 2). In lane 4, 3 molecules each of sCD4 and Fab are illustrated binding to trimers, assuming a 3:1 ligand-trimer ratio; B) Similar to A), two domain sCD4 is substituted with 4 domain sCD4 (lanes 1-4) or Fab X5 substituted with scFv X5. The 4D-sCD4 shift of gp120/gp41 monomers is shown in cartoon form here, since the 2D-sCD4 (lane 5) was too small to give a perceptible shift; C) The binding of four domain sCD4 and Fab 447-52D were analyzed.

LE311 (V3)

Despite a lack of activity in the standard format, like X5, LE311 dramatically neutralized in the post-CD4 format (Fig 2, Table 1, column 7), providing further evidence that the window between the CD4 and CCR5 binding phases is very narrow for JR-FL. Thus, while the V3 loop was inaccessible on the native Env trimer [10], it became fully exposed after CD4 engagement, consistent with the role of the V3 loop in CCR5 engagement [16,32,55] and the previously observed synergy between V3 and CD4-based ligands [38,40]. The cooperative binding of CD4 and a V3 loop mAb to Env trimers is visualized in further experiments below (Fig 3).

Although directed to very different epitopes, it is interesting that X5 and LE311 mAbs were similarly induced by sCD4 (Table 1). LE311 can, however, be distinguished from X5 by its behavior in the standard format using gp160ΔCT SOS-VLPs. Here LE311 activity is detectable, but X5 is not (Table 1, column 5). To put this into context, b12, 2G12, 2F5 and 4E10 also neutralize gp160ΔCT SOS-VLPs somewhat more effectively than gp160ΔCT WT-VLPs [6] (Table 1, compare columns 2 and 5). This may be because SOS becomes arrested midway in infection (and needs to be induced by DTT), perhaps increasing Env's accessibility during the earlier stages of fusion. The sCD4-induction of the V3 epitope observed was not recapitulated by monomeric gp120 in ELISA, probably because the V3 loop is already well exposed. This stands in contrast to CD4i mAbs like X5, whose epitopes are induced on monomeric gp120 by sCD4 binding, just as they are on trimers (data not shown; [40]).

2F5 and 4E10 (gp41 MPER)

Despite modest (2F5) or undetectable (4E10) neutralization in the standard format against full-length gp160 WT (Table 1, column 1) [9], certain modified assay conditions favored 2F5 and 4E10. Thus, gp41 tail truncation and SOS mutation each increased 2F5 and 4E10 neutralization (Table 1, compare columns 2 and 3 to column 1) and together, these mutations increased 2F5 activity still further (Table 1, column 4). The more effective neutralization of SOS-VLPs in the standard format (Table 1 compare columns 3, 4 and 5 to columns 1 and 2) is reminiscent of the activity observed for LE311 (Table 1, compare columns 1 and 2 to column 5). The hallmark of 2F5 and 4E10 was their ability to neutralize SOS-VLPs post-CD4/CCR5 (columns 8 and 9) with similar efficiency as in the standard format (Table 1, column 2). 2F5 performed modestly less effectively than b12 in the pre-attachment format, compared to the standard format (Table 1, compare columns 2 and 6) [6], but was relatively effective in the post-CD4 format (Fig 2, Table 1, column 7). Collectively, this suggests that 2F5 and 4E10 bind to Env trimers at various stages of infection, from native trimers to the post-CD4/CCR5 intermediate [6,18] (Fig 1). Neutralization therefore may occur by interfering with gp41 refolding, which occurs after receptor binding.

T-20

Consistent with its recognition of an intermediate that appears very late during fusion, T-20 was insensitive to assay modifications that affect the early phases of infection (Fig 1, steps 1-3, Table 1, columns 1, 2, and 4). Unlike b12 or 2F5, SOS neutralization formats with an extra wash step appear to cause T-20 to be removed before its target becomes exposed (Table 1, compare columns 4 and 8 to columns 5 and 9). T-20's activity was, however, lower in the pre-attachment assay (Table 1, column 6), but was particularly potent against full-length gp160-VLPs arrested at a lower temperature [6] (Table 1, column 1), suggesting that this protocol captures a very late intermediate. Although T-20 does not reflect any activity by known mAbs, its unique properties make it a useful tool for evaluating neutralization assays.

Analyzing plasma neutralization mechanisms

Since modified neutralization formats can distinguish mAb activities (Table 1), we tested whether the neutralizing activities of HIV+ donor plasma in various formats might reveal information about the Ab specificities they contain. Thus, we investigated a collection of 13 HIV+ donor plasmas (Table 2). We determined ELISA titers against JR-FL gp120 to allow us to distinguish any specificity differences between samples in the context of any differences in overall binding titer (Table 2). All the HIV+ donor plasma ELISA IC50 titers ranged from 1:80,000 for plasma 739 to 1:1,500,000 for L909. The plasmas were arranged according to their neutralizing IC50's against gp160ΔCT WT-VLPs (Table 2, column 1), with the most potent at the top. Thus, plasmas N308, A62 and #11H were strongly neutralizing, L92, L909, R2 and 739 were moderately neutralizing, and TN15, J864, N160, K370, TN11, L503 and a HIV-seronegative control plasma were weakly or non-neutralizing. The most potent format with all samples was post-CD4 attachment, except for the HIV-seronegative control, which had negligible activity (Table 2, column 2). Based on the analysis in Table 1, the potent post-CD4 plasma neutralization may be explained by CD4i or V3 loop-specific Abs, or both, neither of which were able to neutralize effectively in the standard format. Accordingly, neutralization in the standard and post-CD4 formats did not correlate (Table 2, columns 1 and 2). Although neutralization by 2F5- or 2G12-like Abs is also favored in the post-CD4 format (Table 1), the dramatic post-CD4 activity compared to the standard format can not be ascribed to these mAbs, because the difference is too great in magnitude. With reference to X5 and LE311 post-CD4 IC50s (Table 1, column 7), we estimate that plasma N308 contains ∼70μg/ml LE311 or X5 (240,000x0.0003) equivalents or a proportion of each. These estimates are not unreasonable, considering the total IgG concentration in plasma is ∼10mg/ml, of which it has been estimated that 3701,600μg/ml may be gp120-specific [7].

TABLE 2. Analysis of serum neutralization mechanism.

Neutralization activity of various plasma against JR-FL-based VLPs was analyzed in various formats. As in Table 1, neutralization formats are organized in columns. IC50 data are given as reciprocal plasma dilutions. Neutralization has been color-coded in a similar manner to Table 1. N.D.=not done. JR-FL gp120 ELISA titers for each serum against gp120 are shown. IC50 data not shown in the table for plasmas N308, A62, L909, and TN15 were 4,000, 300, 170 and 67, respectively against gp160WT-VLPs in the standard format and 15,000, 3,000, 150 and 99 against gp160ΔCT WT-VLPs in the standard format.

		column	1	2	3
		format	standard	post-CD4	post-CD4/CCR5
		WT/SOS	WT		SOS
IC50 titer:	human plasma	gp120 titer
<50	N308	250,000	9,000	240,000	<40
50-300	A62	100,000	2,000	89,600	<40
300-1000	#11H	500,000	1,000	90,000	<40
1000-3000	L92	1,000,000	400	500,000	<40
>3000	L909	1,500,000	320	309,000	<40
	R2	600,000	150	500,000	<40
	739	80,000	100	200,000	<40
	TN15	603,000	64	143,000	40
	J864	248,000	50	84,000	<40
	N160	500,000	<40	160,000	<40
	K370	381,000	<40	100,000	<40
	TN11	119,000	<40	19,000	<40
	L503	612,000	<40	162,000	<40
	HIV-	<300	<40	250	<40
	spiked human plasma
	J864	248,000	50	84,000	<40
	J864+b12	N.D.	950	90,000	<10
	J864+2G12	N.D.	120	80,000	<10
	J864+2F5	N.D.	110	82,000	71

Among known neutralizing specificities that might be responsible for the neutralizing activity of N308, A62, L92, L909, R2 and 739, those of b12, 2G12 and 2F5/4E10-like Abs remain. Some clues as to the nature of plasma activities emerged from evaluating neutralization in the post-CD4/CCR5 format (Table 2). None of the plasmas exhibited appreciable activity (IC50>1:40) in this format (Table 2, column 3). Considering that 2F5 and 4E10 neutralize JR-FL equivalently in the post-CD4/CCR5 and standard formats (Table 1), this suggests that 2F5/4E10-like activities were effectively absent in our collection of plasmas. Although post-CD4/CCR5 activity was previously detected in other plasmas by this method, titers were also <1:40 [6], suggesting that 2F5-like Abs are extremely rare. Indeed, based on 2F5's post-CD4/CCR5 IC50 of 0.4μg/ml, a plasma titer of <1:40 titer equates to <12μg/ml 2F5 equivalents, consistent with the rare nature of this specificity [6].

Neutralization in the standard format using SOS instead of WT (Table 2, legend) gave slightly higher titers for N308 and A62 but not L909. Conversely, titers against full-length gp160 WT-VLPs were generally slightly lower. These differences reflect similarly mild differences in certain mAb activities in these formats (Table 1), though it is difficult to ascribe these specificities to the neutralization activities of plasmas. Comparing several plasmas in the pre-attachment and standard formats, similar titers were noted (not shown). Although this may indicate the presence of mAbs such as b12 that function relatively effectively in the pre-attachment format, the narrow distinction of b12 and 2F5 IC50s in these formats (Table 1) precludes any categorical inference. Overall, considering the absence of post-CD4/CCR5 titers in particular, it appears that neutralization observed in our samples is best approximated by b12 or 2G12, or perhaps other unknown specificities that act during the earlier stages of fusion.

Non-neutralizing plasma spiked with mAbs b12, 2G12 or 2F5

It is possible that interpretations about neutralization mechanism may be complicated by interplay between neutralizing and non-neutralizing Abs in polyclonal plasmas. To determine if the mAb activities were identifiable amid a background of non-neutralizing Ab (Table 2) and if mechanism analysis can distinguish b12, 2G12 and 2F5 specificities in plasma, we investigated samples of the essentially non-neutralizing plasma J864 that had been spiked with a final concentration of 20μg/ml b12, 40μg/ml 2G12 or 40μg/ml 2F5. Considering that plasmas contain an estimated 370-1,600μg/ml anti-gp120 Ab [7], these concentrations of mAbs might be expected to approximate the concentrations of nAbs present in strongly neutralizing plasmas. In the standard format (Table 2, column 1), b12, 2G12 and 2F5 spiked J864 gave IC50 titers of 1:950, 1:120 and 1:110, respectively. Back calculation of the concentration of parent mAb required to generate these titers (using IC50 titers in Table 1, column 2) gives estimates of 38μg/ml b12, 24μg/ml 2G12,and 22μg/ml 2F5, closely matching the actual concentrations of mAbs added. This implies that the weakly or non-neutralizing Abs present in the J864 have little or no enhancing or blocking effect on the added nAbs. In the post-CD4/CCR5 format, the 2F5-spiked sample was the only one to effectively neutralize and it did so with a titer (1:71) similar to that observed in the standard format (Table 2, column 3). Again, the estimated 2F5 concentration based on the IC50 of 21μg/ml closely matched the amount added. Thus, we can discriminate 2F5/4E10-like and b12/2G12-like activities in plasmas with high levels of non-neutralizing antibodies. This data adds further weight to our earlier suggestion that the neutralizing activity observed in some HIV+ samples does not stem to any great extent from 2F5 or 4E10-like Abs, because post-CD4/CCR5 activity was very low in all plasmas [11].

Direct visualization of Env-mAb binding by BN-PAGE band shifts

To complement the methods to evaluate neutralization mechanism, blue native PAGE (BN-PAGE) provides a way visualize Ab binding to Env proteins. BN-PAGE preserves oligomeric proteins by exploiting the tendency of hydrophobic surfaces to associate with Coomassie dye that imparts a uniform negative charge, generally proportional to the Stoke's radius, allowing proteins to separate under non-denaturing conditions [43,46,53]. BN-PAGE previously revealed functional trimers and putative non-functional gp120/gp41 monomers and gp41 stumps on particle surfaces [43]. Considering that BN-PAGE preserve protein-protein interactions, we adapted the method to measure Ab or ligand binding, visualized as band shifts [43,53]. We used monovalent Fabs, to avoid interference with detection in Western blots via anti-Fc conjugates.

Previous studies revealed that most mAbs recognized the VLP-derived monomers, while only neutralizing Fabs bound to trimers [43]. To further examine the correlation between trimer binding and neutralization, we investigated whether the sCD4-dependent neutralization of X5 and V3 mAbs observed above (Table 1, Fig 2) could be visualized as BN-PAGE band shifts. SOS-VLP derived Env separated as gp120/gp41 trimers and monomers (Fig 3A). Two-domain (2D) sCD4 shifted both the monomer and trimer bands (Fig 3A). In contrast, the X5 Fab shifted neither the monomer nor trimer, consistent with its lack of strong neutralizing activity against this isolate in the standard format. However, in the presence of 2D-sCD4, X5 mediated “super-shifts” of both monomer and trimer bands, reflecting the simultaneous binding of both sCD4 and mAb (Fig 3A). Importantly, the trimer super-shifts reflect the potent post-CD4 neutralization activity of X5 noted above (Fig 2D). Further experiments using four-domain (4D) sCD4 or scFv X5, gave shifts that were commensurate with the size of the respective ligands (Fig 3B), further confirming cooperative CD4-X5 binding to trimers. To investigate the post-CD4 neutralization of the V3 mAb LE311 (Fig 2C), we conducted similar band shift experiments with V3-specific Fab 447-52D and 4D-sCD4. In preliminary experiments, mAb 447-52D exhibited a similar dependency on sCD4 for neutralization, as did LE311 and several other V3 mAbs (not shown). In BN-PAGE, 447-52D binding to trimer was completely dependent on sCD4, consistent with neutralization (Fig 3C). One notable difference between the behavior of 447-52D and X5 was that the V3 mAb alone was able to partially bind to the lower, gp120/gp41 monomer band (Fig 3C).

DISCUSSION

HIV vaccine research may be accelerated by a full understanding of the differences between neutralizing from non-neutralizing sera. Profiling sera specificities should help us to better understand Env-Ab interactions following vaccination. In light of the growing appreciation of the importance of this area to vaccine development, new mapping methods are being generated by researchers [5,19,37,54]. Here we described methods to examine neutralization mechanism. First we measured the activities of various mAbs, to provide a frame of reference for evaluating plasmas that may contain similar Abs. We also investigated a non-neutralizing plasma sample spiked with neutralizing mAbs to investigate whether the mAbs could be detected amid a background of non-neutralizing plasma Abs. We also showed that the neutralization mechanism of V3 loop and CD4i mAbs post-CD4 binding corresponded with similar CD4-dependent Env trimer binding in native PAGE.

The activities of each of our small panel of plasmas in the post-CD4/CCR5 neutralization assay was too weak to ascribe 2F5 and 4E10-like Abs to the potent neutralizing activity of some plasmas in the standard format. Low titers of 2F5/4E10-like Abs are consistent with our previous analysis of several other plasma samples [6]. This inference was verified by the observation that spiking a non-neutralizing plasma with 2F5 caused a perceptible increase post-CD4/CCR5 neutralization. Furthermore, despite their activity in the post-CD4 format, CD4i and V3 specificities could not explain the neutralization activity in plasma, the latter being consistent with previous studies indicating that V3 peptides are unable to absorb neutralizing activity of sera against the primary isolate JR-FL [4,37].

Our data indicated that the neutralization activity in the plasmas was confined to the earlier stages and Abs more akin to b12 and 2G12. Cross-referencing plasma and mAb neutralization activities provides insight into the concentrations of mAbs (in μg/ml equivalents) that would be required to mediate the plasma IC50 titers observed. To generate a titer of 1:9,000 for plasma N308 (Table 2) would require 360μg/ml b12, 1,800 μg/ml 2G12 or 2F5 or 18,000 μg/ml 4E10. Considering estimates of gp120-specific Abs in HIV+ plasma of 370-1,600μg/ml, realistically, only b12 has remotely sufficient potency to explain this neutralization titer.

In neutralizing plasmas, nAbs exist amid a background of mostly non-neutralizing noise. However, our analysis of a non-neutralizing plasma spiked with various nAbs revealed that the non-neutralizing anti-Env Abs did not interfere with neutralization, suggesting that our methods are sufficient to identify neutralizing Abs at reasonable concentrations, even against a tapestry of non-neutralizing activities. At this stage, however, we can not make any conclusive inferences regarding the nature of neutralization in our plasmas. Further mapping by other methods such as virus capture competition assays may reveal further information regarding the nature of this activity (E.Crooks, P.Moore and J.Binley, unpublished). So far, it appears that, in many cases, neutralization could be mediated by potent nAbs that defy categorization based on known specificities, a notion supported by the recent isolation of mAbs directed to novel quaternary epitopes [26,30,66]. Although broadly neutralizing plasmas might exhibit patterns distinguishable from those that do not neutralize, a further complication is that different plasmas may achieve potent neutralization via different nAb specificities and possibly different mechanisms. However, evidence suggests that the complexity of sera is not unlimited. Spectrotyping analysis has shown that mature anti-Env Ab responses are better described as oligoclonal than polyclonal [17], suggesting that comprehensive mapping methodologies might one day be feasible.

Major steps have recently been taken to standardize neutralization assays using a centralized reference panel of Envs to help unequivocally identify promising vaccine concepts [36]. Similarly, standard methods for mapping serum neutralization activity are likely to benefit vaccine research in helping to delineate a rational path forward. In the future, it may be possible to combine mapping approaches to formulate a multi-assay mapping algorithm [5,19,37,59]. Here we focused on mapping with JR-FL as a prototype primary isolate. However, mapping methodologies could easily be adapted to match the Envs of the centralized standard Env reference panel [36], as well as to strains matching the vaccine, to help evaluate type-specific Abs.

Detailed mapping of HIV+ plasmas and vaccinee sera may provide useful information for improving future generations of vaccine immunogens. An ideal immunogen would bind solely to nAbs. We showed that BN-PAGE band shifts provide a unique method to examine Ab binding to Env proteins. Because protein-protein binding is preserved, Ab-Env binding can be directly visualized to assess the Env's topological properties. BN-PAGE can be applied to both soluble and membrane bound Env-based immunogens and divergent Envs [43]. Env vaccines can be modified to dampen non-neutralizing epitopes and/or enhance the exposure of neutralizing ones [54], the efficacy of which could be visualized in BN-PAGE band shifts. Since binding to VLP trimers (Fig 3) is associated with neutralization [43], BN-PAGE band shifts may also be adaptable as a surrogate neutralization assay. “BN-PAGE neutralization assays” using inactivated virus particles could be performed at sites where bio-containment facilities necessary for conventional neutralization assays are not available.

In summary, we suggest that an algorithm for serum mapping may provide a better understanding of what underlies neutralization in HIV+ human plasmas and vaccinee sera. This may provide insights into how antigenicity relates to immunogenicity and an impetus for fresh directions in vaccine design.