A Next-Generation Cleaved, Soluble HIV-1 Env Trimer, BG505 SOSIP.664 gp140, Expresses Multiple Epitopes for Broadly Neutralizing but Not Non-Neutralizing Antibodies

A desirable but as yet unachieved property of a human immunodeficiency virus type 1 (HIV-1) vaccine candidate is the ability to induce broadly neutralizing antibodies (bNAbs). One approach to the problem is to create trimeric mimics of the native envelope glycoprotein (Env) spike that expose as many bNAb epitopes as possible, while occluding those for non-neutralizing antibodies (non-NAbs). Here, we describe the design and properties of soluble, cleaved SOSIP.664 gp140 trimers based on the subtype A transmitted/founder strain, BG505. These trimers are highly stable, more so even than the corresponding gp120 monomer, as judged by differential scanning calorimetry. They are also homogenous and closely resemble native virus spikes when visualized by negative stain electron microscopy (EM). We used several techniques, including ELISA and surface plasmon resonance (SPR), to determine the relationship between the ability of monoclonal antibodies (MAbs) to bind the soluble trimers and neutralize the corresponding virus. In general, the concordance was excellent, in that virtually all bNAbs against multiple neutralizing epitopes on HIV-1 Env were highly reactive with the BG505 SOSIP.664 gp140 trimers, including quaternary epitopes (CH01, PG9, PG16 and PGT145). Conversely, non-NAbs to the CD4-binding site, CD4-induced epitopes or gp41ECTO did not react with the trimers, even when their epitopes were present on simpler forms of Env (e.g. gp120 monomers or dissociated gp41 subunits). Three non-neutralizing MAbs to V3 epitopes did, however, react strongly with the trimers but only by ELISA, and not at all by SPR and to only a limited extent by EM. These new soluble trimers are useful for structural studies and are being assessed for their performance as immunogens.


Introduction
One approach to creating a preventative vaccine against human immunodeficiency virus type 1 (HIV-1) infection is to design an immunogen capable of inducing adequate titers of broadly neutralizing antibodies (bNAbs) [1]. NAbs prevent HIV-1 from infecting target cells by binding to the viral envelope glycoprotein (Env) complex, a trimeric structure comprising three gp120 and three gp41 subunits held together by meta-stable, non-covalent interactions. Induction of NAbs therefore requires the use of an Env-based immunogen. Of these, the most widely tested have been monomeric gp120 subunits, which failed to induce bNAbs and did not prevent infection [2,3,4]. A better mimic of the native, trimeric Env spike may be a superior immunogen for bNAb induction [1,5,6,7,8]. However, creating a true mimic of an Env trimeric spike has proven challenging.
Most approaches to making Env trimers involve truncating the gp41 component to remove the hydrophobic transmembrane region, yielding soluble gp140 proteins containing three gp120 and gp41 ectodomain (gp41 ECTO ) subunits [9]. Soluble gp140 trimers are highly unstable, perhaps because the inherently labile nature of the Env complex is exacerbated by the removal of the transmembrane region. Accordingly, gp140 trimers rapidly disintegrate into individual gp120 and gp41 ECTO subunits unless preventative steps are taken. Two different methods have been used to stabilize gp140 trimers. The most widely used involves eliminating the cleavage site between gp120 and gp41 ECTO and, in some cases, adding an additional trimer-stabilizing motif to the C-terminus of gp41 ECTO , with or without other modifications [9,10,11,12,13,14,15,16,17,18,19]. Trimer-forming constructs such as these are generally referred to as uncleaved gp140s (gp140 UNC ). Our alternative approach involves making fully cleaved trimers but stabilizing them by introducing specific mutations, namely a disulfide bond to covalently link gp120 to gp41 ECTO and an Ile/Pro change at residue 559 to strengthen interactions between the gp41 subunits [5,6]. The resulting trimers are designated SOSIP gp140s. Cleaved and uncleaved trimers are known to be antigenically distinct, in that the latter consistently express the epitopes for various non-neutralizing antibodies (non-NAbs) that are occluded on cleaved trimers, irrespective of whether the Env proteins are soluble or expressed on the cell surface [5,20,21,22].
Here, we describe a new version of SOSIP gp140 trimers based on the subtype A transmitted/founder (T/F) virus sequence BG505, modified to introduce some bNAb epitopes. The membrane-proximal external region (MPER) was also deleted to improve trimer solubility and reduce aggregate formation [23,24]. The BG505 SOSIP.664 gp140 trimers can be produced efficiently and are homogenous and stable. We show here that they express the epitopes for multiple bNAbs, but very few for non-neutralizing antibodies (non-NAbs), when analyzed by ELISA, surface plasmon resonance (SPR), isothermal calorimetry (ITC) and negative stain electron microscopy (EM). Their antigenic properties mimic those of the native Env complexes on the BG505 virus, as judged by the outcome of virusneutralization assays, and they structurally resemble the native complexes when viewed by negative stain EM. These new trimers are the basis for a range of studies of Env structure, alone and as complexes with bNAbs. They have already been used to characterize the epitopes for several bNAbs, including PG9, PGT122 and PGT135 [25,26,27]. The BG505 SOSIP.664 trimers may also be useful as immunogens.

Design of BG505 SOSIP.664 gp140 trimers
Here, we describe the production and properties of stable and homogenous SOSIP gp140 trimers that express multiple bNAb epitopes, based on the BG505 env gene. HIV-1 BG505, a subtype A T/F virus, was isolated from an infant 6-weeks after birth in a mother-infant transmission study [28]. The env sequence was then selected in silico based on its similarity to sequences in the individual from whom the bNAbs PG9 and PG16 were isolated [29]. The monomeric BG505 gp120 protein has the unusual property of binding PG9, although it does not bind efficiently to PG16, and not at all to PGT145; the latter two bNAbs appear to be more dependent on the quaternary structure of the Env trimer [27,29].
Various sequence modifications (see Methods and legend to Fig. 1A) were made to the wild type BG505 sequence to create the protein designated BG505 SOSIP.664 gp140 (Fig. 1A). These alterations included the SOS and I559P changes required for trimer stability, the deletion of the MPER to improve homogeneity and solubility, and the introduction of a T332N substitution to create the epitopes for several bNAbs that depend on the presence of this glycan [5,6,24,30,31]. We also generated two variants with either a D7324-epitope tag or a His-tag located immediately downstream from residue 664, to permit the oriented immobilization of trimers on ELISA plates or SPR chips [32,33,34]. These proteins are designated SOSIP.664-D7324 gp140 and SO-SIP.664-His gp140, respectively (Fig. 1A). For comparison, we expressed and purified a monomeric BG505 gp120 protein containing the same T332N knock-in substitution made to the SOSIP.664 gp140 trimer, as well as the D7324 epitope introduced into the C5 region (see Methods) (Fig. 1A). The BG505 WT.664-His gp140 construct serves as a source of gp41 ECTO -His for ELISA studies (see Methods).

Biochemical and biophysical characterization of BG505 SOSIP.664 gp140 trimers
The BG505 SOSIP.664 construct was expressed transiently in HEK293T, or in some experiments CHO-K1, cells together with co-transfected Furin to boost the level of cleavage [5,31]. The two cell substrates yielded trimers of similar quality and antigenicity. The secreted Env proteins were affinity-purified using the 2G12 bNAb, followed by SEC on a Superdex 200 26/60 column to isolate trimers (Fig. 1B). The SEC profile showed that a predominant, trimer-containing peak eluted at 144 ml, while a smaller peak at 164 ml contained SOSIP gp140 monomers. The SEC profile was confirmed by a BN-PAGE analysis followed by Coomassie Blue dye staining; trimers predominated and some gp140 monomers were present, but there were no appreciable amounts of dimers or higher m.wt. aggregates (Fig. 1D). An analytical Superose 6 column assessment of the SEC-purified BG505 SOSIP.664 gp140 trimers showed that they remained trimeric and neither aggregated nor dissociated into gp120 or SOSIP gp140 monomers (Fig. 1C). These results were confirmed by BN-PAGE (Fig. 1E). The lack of aggregates most likely reflects the beneficial effect of deleting the MPER to make the SOSIP.664 variant [23,24]. A single gp140 band was seen on an SDS-PAGE gel performed under non-reducing conditions, with no evidence for the formation of aberrant inter-protomer disulfide-bonds (Fig. 1F). Coomassie blue-or silver-stained SDS-PAGE gels showed that the gp140 band was essentially fully converted (.95%) to gp120 and gp41 ECTO when a reducing agent was present, confirming that the trimers were cleaved efficiently

Author Summary
A protective HIV-1 vaccine is badly needed, but no candidate has yet provided an adequate level of protection against infection. Most existing vaccines provide immune protection by inducing neutralizing antibodies, also a goal of many HIV-1 immunogen design projects. The trimeric envelope protein complex on the HIV-1 surface is the only relevant target for neutralizing antibodies, and is the basis for most strategies aimed at their induction. However, making a soluble, recombinant envelope protein complex that adequately mimics the structure present on the virus has been challenging. Here, we describe a newly designed and engineered Env protein that has the appropriate properties. This protein, termed BG505 SO-SIP.664 gp140, binds most of the known neutralizing antibodies but generally does not bind antibodies that lack neutralization activity. Its appearance in negative stain electron micrographs also resembles native envelope complexes. 664-D7324 gp140 and gp120-D7324 Env proteins. Modifications compared to the original BG505 gp160 sequence are indicated in red and mentioned in the text. The following changes were made to the wild type BG505 amino acid sequence: 1) The tissue plasminogen activator (tPA) signal peptide replaced the natural one; 2) the gp41 transmembrane (TM) and cytoplasmic tail (CT) domains were deleted to create a soluble gp140; 3) the A501C and T605C substitutions were made to form a disulfide bond between gp120 and gp41 ECTO [5]; 4) the I559P substitution was included to promote trimerization [6,80]; 5) an optimal cleavage site (RRRRRR; R6) replaces the natural one, REKR [31]; 6) truncation of the MPER from residue-664 prevents aggregation [23,24]; 7) the T332N substitution facilitates binding of bNAbs dependent on glycan-N332. The D7324-and His-tags are indicated in yellow. Env sub-domains are indicated: 5 conserved domains (C1-C5); 5 variable domains (V1-V5); heptad repeats 1 and 2 (HR1, HR2); the membrane proximal external region (MPER); the transmembrane domain (TM); and the cytoplasmic tail (CT). The glycan assignments in Env are based on previous studies using gp120 [81,82,83], but may be different for trimeric Env [84]. The amino acid sequence of BG505 SOSIP.664 is given in (F) SDS-PAGE analysis using a 4-12% Bis-Tris Nu-PAGE gel of 2G12/SEC-purified BG505 SOSIP.664 gp140, under non-reducing and reducing conditions, followed by Coomassie blue staining. (G) SDS-PAGE analysis using a 10% Tris-Glycine gel of 2G12/SEC-purified BG505 SOSIP.664 gp140, under non-reducing and reducing conditions, followed by silver staining. The conversion of the gp140 band to gp120 and the appearance of a gp41 ECTO band under reducing conditions is indicative of cleavage. doi:10.1371/journal.ppat.1003618.g001 (Fig. 1F,G). Western blotting with anti-gp120 and anti-gp41 MAbs yielded a similar conclusion (data not shown).
We used differential scanning calorimetry (DSC) to assess the thermal stability of the purified, HEK293T cell-expressed BG505 SOSIP.664 gp140 trimers. The DSC profile showed one distinct unfolding peak with a thermal denaturation midpoint (T m ) of 68.1uC ( Fig. 2A). This finding was similar to ones made previously using the same trimers, but produced in HEK293S cells defective for GlcNAc transferase I (GnTI) and, hence, bearing only oligomannose glycans [27]. Of note is that the BG505 SOSIP.664 gp140 trimers are substantially more stable than the corresponding gp120 monomers, which unfold in two phases (T m , 53.5uC and 63.4uC; Fig. 2B), and also than YU2 gp120 (T m , 59.2uC [35,36]) or 92UG031 gp120 (T m , 58.4uC [35]). The BG505 SOSIP.664 trimers are also more stable than the corresponding trimers from KNH1144 (T m of 51.3uC for the first thermal transition [27]), and than JR-FL SOSIP.R6 trimers (which dissociate at ,50uC [6]). We conclude that the BG505 SOSIP.664 gp140 trimers have high thermal stability.
The overall morphology of SEC-purified, BG505 SOSIP.664 gp140 trimers was studied by negative stain EM ( Fig. 2C; Fig. S1). A 3D reconstruction at 24 Å resolution showed that compact and homogeneous trimers were consistently present, as described previously for the same trimers produced in HEK293S cells [27]. We obtained similar results with the SOSIP.664-D7324 and SOSIP.664-His gp140 proteins, indicating that the C-terminal tags did not perturb the overall trimer structure (data not shown). Additional EM images of trimer-bNAb complexes are described below.

Neutralization of the parental BG505.T332N virus
To study the antigenic properties of the BG505 SOSIP.664 gp140 trimers, we first tested a large panel of MAbs for their capacity to neutralize the corresponding Env-pseudotyped virus in a TZM-bl cell-based neutralization assay. The epitope clusters recognized by the MAbs cover most of the surface of the Env trimer [37] except the MPER, which was truncated in the SOSIP.664 construct. Note that the test virus contains the T332N knock-in change to allow comparison with antigenicity data obtained using the BG505 SOSIP.664 gp140 trimers; this change may account for any discrepancies from data described elsewhere using the unmodified BG505 virus [29].
Several test MAbs did not neutralize BG505.T332N (IC 50 .30,000 ng/ml), including b6, 15e, F91 and F105 to the CD4bs; 17b, 412d, X5 and A32 to CD4-induced epitopes; 447-52D, 39F, 19b and 14e to V3; F240 and 7B2 directed to gp41. We confirmed that their epitopes were present on at least one form of BG505 Env protein (e.g., gp120 monomers or gp41 ECTO ), showing that their inability to neutralize the virus was not due to a sequence-dependent lack of the epitope (data not shown, and see below). The CD4i MAbs 17b and 412d also did not neutralize the BG505.T332N virus when sCD4 was also present (data not shown). For the V3 MAbs 19b and 14e, we also performed extended pre-incubation experiments (16 h) before adding the MAb-virus mixtures to the target cells. Even in this assay format, the two V3 MAbs had no measurable neutralization activity, indicating that they do not inactivate the Env spike (data not shown).

Antigenic analysis of BG505 SOSIP.664 gp140 trimers by ELISA
We used several methods to quantify the binding of bNAbs and non-NAbs to wild type or epitope-tagged versions of BG505 SOSIP.664 gp140 trimers and, in some cases, the cognate SOSIP gp140 monomers, gp120 monomers, or gp41 ECTO proteins. First, we immobilized 2G12 affinity-and SEC-purified SOSIP.664-D7324 trimers onto ELISA plates and monitored the binding of a large panel of MAbs (Figs. [3][4][5]. The immobilized trimers were recognized efficiently by all of the bNAbs against the CD4bs, the N332-glycan dependent V3 cluster or the N332-glycan dependent outer domain cluster that neutralized the corresponding virus (see above). The 3BC315 and 3BC176 bNAbs bind to an incompletely characterized, but probably glycan-independent, epitope that is induced, to an extent, by CD4 binding [38]. They do not bind to any soluble Env protein tested to date, including uncleaved soluble gp140 trimers [38]. However, both bNAbs interacted efficiently in ELISA with the BG505 SOSIP.664-D7324 gp140 trimers, but not the corresponding gp120 monomers (Fig. 3, Fig. 4C, and data not shown).
The bNAbs that recognize quaternary-preferring epitopes are particularly useful tools for gauging whether soluble Env trimers adopt an appropriate conformation. This bNAb category includes CH01, PG9, PG16 and PGT145 against epitopes that appear to span the V1V2 domains of two gp120s within a single trimer [27]. Although these bNAbs can bind a small subset of monomeric gp120s or uncleaved trimeric gp140s, any such interactions tend to be rare and weak, particularly for PG16 and PGT145 [12,27,29,39]. Here, we show that CH01, PG9, PG16 and PGT145 all bound efficiently to the BG505 SOSIP.664 gp140 trimers in ELISA (Fig. 3, Fig. 4D). In contrast, only PG9 reacted well with monomeric BG505 gp120, while CH01 and PGT145 were completely non-reactive and PG16 bound weakly. We have also found that PG16 and PGT145 do not bind to BG505 SOSIP.664 gp140 trimer mutants that lack the glycans attached to N156 or N160, or to wild-type trimers produced in HEK293T cells treated with the mannosidase inhibitor kifunensine (data not shown). These findings are consistent with the known involvement of hybrid or complex glycans at N156 or N160 in the PG16 and PGT145 epitopes [30,40].
To further study the influence of trimerization on the PG9, PG16 and PGT145 epitopes, we fractionated 2G12 affinitypurified SOSIP.664-D7324 gp140 proteins by SEC and analyzed the column fractions by ELISA (Fig. 4E). Both PG16 and PGT145 bound almost exclusively to the trimer-containing fractions, whereas PG9 bound more strongly to the trimers, but also recognized the SOSIP.664-D7324 gp140 monomers. This reactivity pattern is broadly consistent with previous observations on BG505 gp120 monomers [29]. In contrast, the CD4bs non-NAb b6 bound preferentially to the SOSIP.664-D7324 gp140 monomers, although some binding to trimers was also seen (see below).
For HIV-1 to be neutralized, a NAb must bind to a sufficient number of the native, functional, trimeric spikes present on the virus surface [41,42]. Non-NAbs fail to neutralize because their epitopes are either absent from these trimers, or not accessible at the right time. A soluble gp140 trimer that mimics native, functional spikes should, therefore expose few or, ideally, no epitopes for non-NAbs. Accordingly, we assessed various non-NAbs for their abilities to bind wild type or epitope-tagged versions of BG505 SOSIP.664 gp140 trimers. Several non-NAbs to CD4bs epitopes (F91, F105, b6, 15e) did not bind the SOSIP.664-D7324 trimers or did so only weakly. However, all of them reacted strongly with the corresponding gp120 monomers and/or SOSIP.664 gp140 monomers ( Fig. 3 and Fig. 5A). The diminished or absent reactivity of these non-NAbs with the trimers does not, therefore, reflect the absence of the epitope due to sequence variation, but rather the structural constraints present on ''native-like'' trimers.
We next investigated the binding of CD4i MAbs 17b, 412d, X5 and A32, which were all unable to neutralize the corresponding virus in the presence or absence of sCD4 ( Fig. 3 and data not shown). None of these four non-NAbs bound to the BG505 SOSIP.664-D7324 gp140 trimers. When sCD4 was present, the 17b epitope was induced on the trimers (as were the similar 412d and X5 epitopes, to lesser extents), indicating that CD4-induced conformational changes had taken place ( Fig. 3 and Fig. 5B, and data not shown). The 17b, 412d and X5 MAbs were also gp120 monomer-reactive but only when sCD4 was present ( Fig. 3 and Fig. 5B, and data not shown). In contrast, the A32 MAb to a different category of CD4i epitope failed to bind the trimers even in the presence of sCD4, but bound strongly to the gp120 monomers in the absence of sCD4 and even more so when sCD4 was added ( Fig. 3 and Fig. 5). The induction by sCD4 of conformational changes in the BG505 SOSIP.664-D7324 trimers, measured by ELISA, is consistent with the conformational changes seen in the EM images of sCD4/17b-complexes of the non-tagged trimers (see below).
In contrast to the above observations, the anti-V3 MAbs 39F, 19b and 14e, which did not neutralize the BG505.T332N virus, bound strongly to the BG505 SOSIP.664-D7324 gp140 trimers in ELISA, although less well than to the corresponding gp120 monomer ( Fig. 3 and Fig. 5D). We note, however, that 19b and 14e bound only marginally to the same trimers in SPR-or EMbased assays (see below).
Most non-NAbs did not bind the SOSIP.664-D7324 gp140 trimers in ELISAs, or did so only weakly compared to their reactivity with monomeric proteins. The most striking outliers were the V3 non-NAbs 39F, 14e and 19b, which did bind strongly in this assay (Fig. 5D). The bNAb 2G12 bound more strongly to trimers in ELISA than would be predicted by its neutralization capacity. This outcome might be attributable to the specific enrichment of 2G12reactive soluble trimers during the affinity purification process, given that other, less 2G12-reactive Env spikes will contribute to infection during neutralization assays. When 2G12, 39F, 14e and 19b were excluded from the correlation, the r-value increased to 0.88 (95% confidence interval 0.80-0.94; P,0.0001). , and also CD4-IgG2, to purified BG505 SOSIP.664-D7324 gp140 trimers. (C) Representative binding curves of bNAb 3BC315 with purified BG505 SOSIP.664-D7324 gp140 trimers and gp120-D7324 monomers. (D) Representative binding curves of quaternary structure dependent bNAbs PG9, PG16, CH01 and PGT145 to purified BG505 SOSIP.664-D7324 gp140 trimers and gp120-D7324 monomers. The legend is the same as for panel C. (E) BG505 SOSIP.664-D7324 gp140 trimers were 2G12-affinity purified and fractionated using a Superose 6 10/30 SEC column. The SEC fractions were analyzed for PG9, PG16, PGT145 and b6 binding by D7324-capture ELISA. Note that the scales on the y-axes and x-axes vary from MAb to MAb. doi:10.1371/journal.ppat.1003618.g004 surface plasmon resonance We next investigated the binding of a subset of representative bNAbs and non-NAbs using surface plasmon resonance (SPR). In this assay, binding of MAbs to BG505 SOSIP.664-His gp140 trimers, immobilized via His-Ni 2+ interaction on NTA chips, again generally agreed well with their capacity to neutralize the corresponding Env-pseudovirus ( Fig. 7A-C). Thus, bNAbs 2G12 and PGT135 to glycan-dependent epitopes on the outer domain of gp120, bound to high and intermediate levels, respectively, in the SPR assay. The PGV04 bNAb (CD4bs) bound strongly with markedly slow dissociation, the V1V2-and quaternary-structuredependent bNAbs PG9, PG16, and PGT145 all bound to intermediate levels, while the V3-and N332-dependent bNAbs PGT121, PGT123, and PGT128 bound to intermediate or high extents with distinctive kinetics. In contrast, the CD4bs non-NAb b6 reacted only marginally with the trimers in the SPR assay, while b12 (CD4bs) and F240 (gp41 cluster I) did not bind detectably. The V3-specific non-NAb 14e, which did bind strongly to BG505 SOSIP.664-D7324 gp140 trimers in ELISA, was only marginally reactive with the corresponding His-tagged trimers by SPR; the low signals for 14e contrast markedly with those for the V3-and N332-dependent bNAbs (e.g., the plateau values were 60-70 RU and 750 RU for 14e and PGT128, respectively). Moreover, the plateau signal for 14e at 1,000 nM (150,000 ng/ml) was only twice that for b6 (30 RU).
An even starker contrast between effective binding and complete lack of interaction was observed when Fabs were used instead of IgG molecules: the PGV04 Fab (bNAb) bound to an intermediate level (when the three times lower mass contribution to the resonance is taken into account), whereas there was no detectable binding of the (non-NAb) Fabs b6, b12 or F240 (Fig. 7D). Soluble CD4 (of a similar mass to Fabs) bound to a somewhat higher level than the PGV04 Fab, but with markedly faster association and dissociation kinetics (Fig. 7D).
The converse SPR approach of immobilizing the Env-reactive Abs and allowing the untagged BG505 SOSIP.664 gp140 trimers (at 200 nM; 78,000 ng/ml) to bind from the solution phase yielded broadly similar results for the subset of MAbs tested in this way (Fig. 7E). Thus, strong responses were obtained for trimer binding to immobilized 2G12 or PGT128. The BG505 SOSIP.664 trimers did not bind detectably to the immobilized b12 or F240 IgG (non-NAbs) in this SPR format, but a low level of binding to the b6 IgG (also a non-NAb) was observed. The extent of trimer-b6 binding was greater than in the converse SPR set-up, perhaps because the intrinsically weak paratope-gp120 binding is compensated for by the avidity effect of potentially trivalent interactions with the captured IgG; the 2.7-fold larger mass of the trimer compared to IgG should also be taken into account when assessing the degree of binding.

Antigenic analysis of BG505 SOSIP.664 gp140 trimers by isothermal titration calorimetry
To determine the thermodynamic binding characteristics of the BG505 SOSIP.664 gp140 trimers, we performed isothermal titration calorimetry (ITC) experiments using PGT121 Fab, PGT128 Fab and the domain-exchanged 2G12 IgG. All three antibodies have previously been shown to be dependent on the high-mannose glycan at position N332 for Env recognition [30,43,44,45,46]. PGT121, PGT128 and 2G12 all bound the BG505 SOSIP.664 trimer with nanomolar affinities (151 nM = 7550 ng/ml; 5.7 nM = 284 ng/ml; and 16.0 nM = 2400 ng/ml, respectively) and near identical stoichiometries of 2.3-2.4 (Table 1; Fig. 8). These binding stoichiometries are three-fold higher than the value of 0.8 previously reported for PG9 binding (affinity 11 nM = 550 ng/ml) to the same construct (Table 1) [27]. The data therefore imply that three PGT121 Fabs, PGT128 Fabs or 2G12 IgG molecules bind per trimer, which is consistent with their recognition of N332-dependent epitopes on the outer domain of gp120. In contrast, only a single PG9 Fab recognizes the N160dependent epitope at the membrane-distal apex of each trimer [27]. Taken together, the ITC binding data further confirm that the BG505 SOSIP.664 gp140 trimers properly display the highaffinity binding sites for the glycan-dependent bNAbs, PGT121, PGT128, 2G12 and PG9.

Antigenic analysis of BG505 SOSIP.664 gp140 trimers by electron microscopy
We used negative stain electron microscopy (EM) to characterize the binding of the CD4bs bNAb PGV04 to the BG505 SOSIP.664 gp140 trimer ( Fig. 9A; Fig. S3). The reconstruction at 23-Å resolution shows that PGV04 binds the soluble trimers in a manner similar to other CD4bs-directed bNAbs with virionassociated Env, in that it approaches the gp120 protomers from the side [47]. We compared the complex formed between PGV04 and BG505 SOSIP.664 trimers with other such bNAb-trimer complexes. Recent studies with the same trimers, albeit expressed in glycan processing-deficient GnTI 2/2 HEK293S cells and not, as here, HEK293T, have shown how the PGT122 and PGT135 bNAbs bind to their N332 glycan-dependent epitopes. Thus, their angle of approach differs from how PGV04 encounters the CD4bs, but all three bNAbs saturate the three available binding sites on the trimer (Fig. 9C) [25,26]. In contrast, and as noted above, the quaternary preferring, N160 glycan-specific bNAb PG9, only binds to one epitope per trimer (Fig. 9C) [27]. Images of the BG505 SOSIP.664 trimers in complex with sCD4 and 17b show that conformational changes are induced (Fig. 9B) that are consistent with ones described for SOSIP trimers based on the JR-FL and KNH1144 genotypes [23,24,48], and for the full length, virus-associated BaL Env spike [49]. Collectively, the new and  (x-axis). The Pearson's correlation coefficient, r, was calculated using Prism software version 5.0. When accurate midpoint concentrations could not be calculated because of lack of binding or neutralization, the highest concentration tested was included in the correlation analysis (i.e., when the IC 50 of neutralization was .30,000 ng/ml, a value of 30,000 ng/ml was used). The data points for NAbs are indicated in black, those for non-NAbs in gray. Also note that 11 data points are overlapping in the right upper corner; they were derived using MAbs that neither neutralized the virus nor bound the trimer (IC 50 of neutralization .30,000 ng/ml, EC 50 in ELISA .10,000 ng/ml). Only MAbs whose epitope could be shown to be present on at least one form of BG505 Env protein were included in this analysis; MAbs that were non-reactive, presumably because of sequence variation, were excluded. The Pearson's correlation was also calculated without the data points for 2G12, 39F, 14e and 19b, as discussed in the text. The fitted line is based on all data, i.e. including 2G12, 39F, 14e and 19b. doi:10.1371/journal.ppat.1003618.g006 recently published negative stain EM data show that BG505 SOSIP.664 gp140 trimers, derived from HEK293T or HEK293S cells, express multiple different bNAb epitope clusters, and also undergo conformational changes when they bind sCD4 [25,26,27].

Antigenic analysis of BG505 SOSIP.664 gp140 trimers: Summary
Taken together, the various antigenicity assays show that every MAb that neutralizes the BG505.T332N virus efficiently also bound strongly to the soluble BG505 SOSIP.664 gp140 trimers (wild type and/or epitope tagged), except for MAbs to MPER epitopes that were not present in the trimer construct (data not shown) (Fig. 11). We conclude that the trimers display a range of bNAb epitopes from multiple different clusters. In contrast, non-NAb epitopes are generally structurally occluded and not displayed on the SOSIP trimer (Fig. 11).

Discussion
We describe here the design and properties of a nextgeneration, fully cleaved and highly stable soluble gp140 trimer based on the BG505 subtype A sequence. EM imaging shows that the BG505 SOSIP.664 gp140 trimers are homogeneous and that their architecture is very similar to that of native Env spikes on virions. We used a range of techniques to assess the antigenic properties of the soluble trimers, particularly their abilities to bind bNAbs and non-NAbs and the relationship between trimer binding and virus neutralization. The various techniques were generally concordant, with one notable exception relating to V3 MAbs. Overall, there were few discrepancies between the antigenicity of the trimers and the neutralization sensitivity of the corresponding BG505.T332N virus. Thus, in general, all the bNAbs that neutralized the virus also bound to the soluble trimers, with the obvious exception of bNAbs to the MPER, a region that was eliminated from the trimers to improve their biophysical properties. The presence of so many bNAb epitopes on a soluble, The reported values are averages from at least two independent measurements. The associated errors are approximately 10% of the average. Representative isotherms are shown in Fig. 8. b The change in Gibbs free energy (DG) was determined using the relationship: DG binding = RTlnK d [87]. c The stoichiometry of binding (N) is directly affected by errors in protein concentration measurements, sample impurity and heterogeneity of gp140 glycans. d Dissociation constant associated with a second (low affinity) binding event.
e The binding isotherms do not allow the stoichiometry and enthalpy associated with the second binding event to be determined accurately. f Data previously described elsewhere [27]. doi:10.1371/journal.ppat.1003618.t001  Table 1. doi:10.1371/journal.ppat.1003618.g008 generally homogenous and highly stable trimer is highly beneficial for structural studies, and may also be valuable for their immunogenicity properties. Whether the favorable antigenic profile translates into the induction of bNAbs will be determined experimentally.
In contrast to bNAbs, non-NAbs were rarely strongly reactive with the BG505 SOSIP.664 gp140 trimers, even when their epitopes were present on less complex forms of BG505 Env (e.g., gp120 or gp140 monomers, or gp41 ECTO ). This finding was particularly striking for non-NAbs against the CD4bs, such as F91 and F105, and implies that the steric constraints on MAb access to this region of virion-associated Env also applies to the soluble trimers. The same argument applies to the various CD4i epitopes, which were inaccessible on the soluble trimers unless sCD4 was also bound, and to non-NAb epitopes in gp41 ECTO . We note that the A32 epitope, while present and further induced by sCD4 on BG505 gp120 monomers, was absent from the trimers whether sCD4 was present or not. This observation may be relevant to arguments that the A32 epitope is an important target for ADCCmediated killing of infected cells [50]. The only discordance between the trimer-binding and virus-neutralization assays involved the V3 region of gp120. Thus, the V3 non-NAbs 19b, 14e and 39F bound efficiently to the D7324-tagged trimers in the capture ELISA. However, the outcomes of the SPR and negative stain EM assays were quite different, in that, in these assays, the V3 MAbs were only minimally reactive with His-tagged, D7324tagged or non-tagged trimers. One explanation may be that the capture onto the ELISA plate via the D7324 antibody might induce some local unfolding of the trimers. As we think it unlikely that the low level of binding of 19b and 14e seen by EM can explain the rather strong binding in ELISA, we favor this explanation but acknowledge that it is speculative in nature. We do, however, note that, at high concentrations, some binding of a subset of other non-NAbs can be observed in ELISA (Fig. 5). It is therefore also possible there is some conformational heterogeneity in the trimer population, with a minor subset displaying some non-NAb epitopes. Negative stain EM does show that some trimers can bind one or (very rarely) two non-NAb Fabs (Fig. 10). Another possibility is that the trimers are flexible, allowing different conformations to be sampled over time in a way that registers more strongly in an ELISA than in other binding assays [51,52].   Table 1 and Fig. 8 for details. . The following scoring was used for EM analyses: +: N = 3 (except for PG9, PG16 and PGT145 where N = 1) for .50% of the trimers; +/2: N = 3 for ,50% of the trimers; 2: N = 1,2 or 3 for ,10% of the trimers, where N is the number of Fabs bound per trimer. See Table 2, Fig. 9 and Fig. 10  We note that an absolute stoichiometry of ,3 (i.e., 2.4) was found for 2G12 IgG in the ITC binding experiments, whereas a value of 3 might have been expected for trimers that had been affinity-purified on a 2G12-IgG column. The discrepancy might arise from errors in glycoprotein concentration measurements. Sample impurity can also contribute to lower apparent binding stoichiometries. However, another possibility is that some gp120 protomers on a trimer do not express the 2G12 epitope due to variation in the glycosylation process. The presence of one or two 2G12 epitopes per trimer is probably sufficient for binding to the 2G12 affinity column. Of note is that ITC also yielded trimerbinding stoichiometries of 2.3 to 2.4 for the glycan-dependent Fabs PGT121 and PGT128, whereas Fab PGT122, which is very similar to PGT121, saturated all three binding sites on the trimer as visualized by EM (Fig. 9B). The explanation(s) might be similar to those suggested for 2G12.
In the ITC experiments, a weak secondary binding event could be seen for the PGT128 and 2G12 bNAbs, in addition to the saturating high-affinity event. PGT128 and 2G12 have high affinities for mimetic (i.e., non-Env) oligomannose glycan substrates, and might therefore interact weakly with other oligomannose glycans on the SOSIP.664 gp140 trimers, in addition to their high-affinity epitopes. Whether such low affinity binding events imply that high concentrations of these antibodies would react with secondary binding sites on the virion-associated Env trimer, and hence contribute to neutralization, is not yet known.
Across the entire bNAb and non-NAb test panel, there was an excellent concordance between the outcomes of trimer-reactivity (by ELISA) and virus-neutralization assays. Thus, a formal correlation plot yielded a highly significant r-value of 0.65 (P,0.0001). Such an outcome would not be the case for Envbinding assays using gp120 monomers or uncleaved gp140 trimers, because of their strong reactivity with multiple non-NAbs [5,20,21,22]. The V3 non-NAbs were the predominant outliers in the correlation analysis and, as noted above, their strong ELISA reactivity is not supported by the SPR and EM studies. One other outlier was the 2G12 bNAb. Here, the discrepancy is likely to be rooted in the use of a 2G12-affinity column for purifying the BG505 SOSIP.664 gp140 trimers (D7324-tagged or not). Thus, the column is likely to select for a trimer sub-population that has a high affinity for 2G12. In contrast, the virions used in neutralization assays undergo no such 2G12-selection procedure and would have a more ''average'' affinity for this bNAb. When 2G12 and the V3 MAbs were excluded, the r-value for the trimerreactivity and neutralization correlation increased to 0.88 (P,0.0001). Overall, we are encouraged by the general occlusion and/or absence of non-NAb epitopes on the BG505 SOSIP.664 gp140 trimers, not only because of their antigenic, and arguably structural, fidelity with respect to virion-associated trimers, but also for immunogenicity studies. Thus, when the goal is to induce bNAbs, non-NAb epitopes represent a distraction to the immune system that is best avoided. Less sophisticated Env immunogens that do not mimic the native spike efficiently induce non-NAbs, if and when this is a desired outcome [3,53]. It is possible that the exposure of some V3-associated non-NAb epitopes on the BG505 SOSIP.664 trimers under certain experimental conditions in vitro (e.g., in the ELISA) might also occur when they are used as immunogens. While V3 is not an important neutralization site for primary viruses, some V3 Abs are active against a subset of viruses in vitro. Hence, any induction of V3 Abs in vivo might be useful under some circumstances. A converse argument, however, is that V3 is an immunodominant epitope cluster that may distract the immune system from focusing on more worthwhile targets elsewhere on the trimer. If so, it would be best to mask or stabilize the V3 region on a new variant of BG505 SOSIP.664 gp140 trimers, for example by introducing a glycan site(s) at an appropriate position [54]. Care would need to be taken, however, to occlude only undesirable regions of V3 (e.g., the 19b/ 14e sites) without affecting nearby areas that contribute to genuine bNAb epitopes, such as those for PGT121-123 and PGT125-130. We also note that glycan-masking might introduce unwanted neo-epitopes. Additional structural information that would help guide future SOSIP gp140 trimer re-designs is being actively sought.
It may never be possible to make a soluble gp140 trimer that precisely mimics the native Env spike, because deleting the transmembrane and cytoplasmic domains (and, in the case of SOSIP.664 trimers, also the MPER) will have at least some impact on trimer structure. Thus, point substitutions in gp41 HR1, HR2, MPER and the intracytoplasmic tail, as well as the length of the cytoplasmic tail, can influence the interaction of antibodies with the gp120 moieties of trimers on virions and infected or transfected cells [55,56,57,58,59]. Nonetheless, the BG505 SOSIP.664 gp140 trimers, as assessed by a variety of different antigenicity assays, do seem to come very close to being a faithful Env-spike mimetic.
Ongoing structural studies will help confirm or refute this assessment. Whether the properties of the present trimers can be further improved by a targeted mutagenesis approach remains to be determined. We will show elsewhere just how critical cleavage of the gp120-gp41 linkage is for making soluble trimers that resemble native spikes.
The BG505 SOSIP.664 gp140 trimers have already been already useful for structure-based studies aimed at defining bNAb-Env interactions [25,26,27], as were the corresponding trimers based on the KNH1144 env gene [46,48,60]. The BG505 trimers are also substrates for ongoing efforts aimed at defining crystal or high-resolution EM structures of the Env trimer. In addition, we are assessing their immunogenicity in rabbits, guinea pigs and macaques, as well as determining how to make them in the much larger quantities, and of the appropriate quality, required for any future testing in humans. A high priority will also be to identify additional genotypes that yield SOSIP.664 trimers with the same favorable properties as the ones described here. Among hypotheses to explore is that the T/F and/or pediatric status of the BG505 isolate is relevant to the trimers' homogeneity and stability; a second relates to the presence of certain trimer-stabilizing residues associated with thermal stability. Thus, the M535, N543 and K567 residues that are present in BG505 SOSIP.664 have been reported to contribute to the trimerization efficiency of soluble gp140 and the thermal stability of Env trimers on virus particles [61,62].
Variants of the BG505 SOSIP.664 gp140 trimers bearing either a His-tag or a D7324 epitope-tag sequence at the C-terminus of gp41 ECTO were also made by adding the amino acid sequences GSGSGGSGHHHHHHHH or GSAPTKAKRRVVQREKR, respectively, after residue 664 in gp41 ECTO and preceding the stop codon. These proteins are designated SOSIP.664-His gp140 and SOSIP.664-D7324 gp140. We also made a His-tagged gp140 with the C501 and C605 cysteines replaced by their original residues, and with P559 similarly reverted to the original isoleucine (BG505 WT.664-His gp140). When expressed in the presence of excess furin to ensure efficient precursor cleavage, the absence of the SOS disulfide bond means the gp140 trimer is unstable and dissociates to gp120 and a trimeric form of His-tagged gp41 ECTO (BG505 gp41 ECTO -His); the latter can be used in a NiNTAcapture enzyme-linked immunosorbent assay (ELISA; see below).
A monomeric BG505 gp120 with a similar sequence to the gp120 components of the gp140 trimers was designed by: introducing a stop codon into the SOSIP.664 construct at residue 512; reverting the optimized cleavage site to wild type (RRRRRRRREKR at residues 508-511); reverting the A501C change; introducing the D7324 epitope into the C5 region (R500K+G507Q); and making a L111A substitution to decrease gp120 dimer formation [29,63]. A slightly modified version of BG505 gp120 that has been described previously [25] was used in DSC experiments. For this modification, the BG505 gp120 gene was cloned downstream of an IgK secretion signal in a phCMV3 plasmid and upstream of a His-tag. The cleavage site was mutated to prevent the His-tag from being cleaved off, leading to the following C-terminal sequence: RAKRRVVGSEKSGHHHH HH.
The BG505 gp160 clone for generating Env-pseudoviruses for neutralization assays has been described elsewhere [29]. We modified this clone by inserting the same T332N substitution that is present in the BG505 SOSIP.664 trimers, and refer to the resulting virus as BG505.T332N.

Env protein expression
The Env proteins from various env genes described above were expressed in wild type, adherent HEK293T cells or the 293F variant that is adapted for suspension cultures, or in CHO-K1 cells, essentially as described [25,27,46,64]. HEK293T and CHO-K1 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS), penicillin (100 U/ml), streptomycin (100 mg/ml), Glutamax (Invitrogen), non-essential amino acids (0.1 mM), sodium pyruvate (0.1 mM) and HEPES (0.1 mM). For gp140 trimer production, HEK293T or CHO-K1 cells were seeded at a density of 5.5610 4 /ml in a Corning Hyperflask. After 3 days, when the cells had reached a density of 1.0610 6 /ml, they were transfected using polyethyleneimine (PEI) as described elsewhere [65]. Briefly, PEI-MAX (1.0 mg/ml) in water was mixed with expression plasmids for Env and Furin [5] in OPTI-MEM. For one Corning Hyperflask, 600 mg of Env plasmid, 150 mg of Furin plasmid and 3 mg of PEI-MAX were added in 550 ml of growth media. Culture supernatants were harvested 72 h after transfection. BG505 gp120 used in differential scanning calorimetry (DSC) experiments was produced in HEK293F cells using a protocol similar to that previously described [25].

Env protein purification
Env proteins were purified from the supernatants by affinity chromatography using either a 2G12 column or a Galanthus nivalis (GN)-lectin column [25,27,46,64]. Briefly, transfection supernatants were vacuum filtered through 0.2-mm filters and then passed (0.5-1 ml/min flow rate) over the column. The 2G12 column was made from CNBr-activated Sepharose 4B beads (GE Healthcare) coupled to the bNAb 2G12 (Polymun Sciences, Klosterneuburg, Austria). Purification using this column was performed as follows: the beads were washed with 2 column volumes of buffer (0.5 M NaCl, 20 mM Tris, pH 8.0) before eluting bound Env proteins using 1 column volume of 3 M MgCl 2 . The eluted proteins were immediately buffer exchanged into 75 mM NaCl, 10 mM Tris, pH 8.0, using Snakeskin dialysis tubing (10K WCMO) (Thermo Scientific). The buffer-exchanged proteins were further concentrated using Vivaspin columns with a 30-kDa cut off (GE Healthcare). For GN-lectin affinity purification, the wash buffer was Dulbecco's phosphate buffer saline (DPBS) supplemented with 0.5 M NaCl was used, and elution was carried out using DPBS supplemented with 1 M methyl mannopyranoside.
In both cases, the affinity-purified Env proteins were further purified to size homogeneity using size exclusion chromatography (SEC) on a Superdex 200 26/60 column (GE Healthcare). A Superose 6 column was sometimes used for analytical or preparative purposes. The trimer fractions and, occasionally also the SOSIP gp140 monomer fractions, were collected and pooled. Protein concentrations were determined using either a bicinchonic acid-based assay (BCA assay; Thermo Scientific, Rockford, IL) or UV 280 absorbance using theoretical extinction coefficients [66].

SDS-PAGE and Blue Native-PAGE
Env proteins were analyzed using SDS-PAGE and BN-PAGE [6,67] and stained using Coomassie blue or silver stain. The input material was mixed with loading dye and directly loaded onto a 4-12% Bis-Tris NuPAGE gel or a 10% Tris-Glycine gel (Invitrogen). The gels were run for 1.5 h at 200 V (0.07 A) using 50 mM MOPS, 50 mM Tris, pH 7.7 as the running buffer (Invitrogen).

Differential scanning calorimetry (DSC)
Thermal denaturation was probed with a VP-DSC calorimeter (GE Healthcare). Before carrying out the experiments, all samples were extensively dialyzed against phosphate-buffered saline (PBS). The protein concentration was subsequently adjusted to 0.1-0.3 mg/ml, as described above. After loading the protein sample into the cell, thermal denaturation was probed at a scan rate of 90uC/h. Buffer correction, normalization and baseline subtraction procedures were applied before the data were analyzed using Origin 7.0 software. The data were fitted using a non-two-state model, as the asymmetry of some of the peaks suggested the presence of unfolding intermediates.

Antibodies and Fabs
Antibody concentrations are generally recorded in ng/ml for neutralization assays and trimer binding ELISAs, but in nM for ITC and SPR experiments. Since the molecular mass of an average IgG molecule is approximately 150,000 Da, the conversion factors for IgG are: 1000 ng/ml = 6. Fab PGT121, PGT128 and PGV04, as well as IgG 2G12 used in isothermal titration calorimetry (ITC) and electron microscopy (EM) experiments, were produced following a protocol similar to that previously described [25,27]. Briefly, heavy and light chain genes were transfected in HEK293F cells using 293Fectin (Invitrogen). Secreted Fab or IgG were harvested 6-7 days posttransfection. The supernatant was directly loaded on either an anti-human l light chain affinity matrix (CaptureSelect Fab l; BAC) for PGT121 and PGT128 Fabs, an anti-human k light chain affinity matrix (CaptureSelect Fab k; BAC) for PGV04 Fab or on a Protein A column for 2G12 IgG. Elution was performed using a buffer containing 100 mM glycine, pH 2.7. The PGT121, PGT128 and PGV04 Fabs were subjected to MonoS (GE Healthcare) cation exchange chromatography to eliminate light chain dimers. All antibodies were subsequently purified to size homogeneity by gel filtration chromatography using a Superdex 200 column (GE Healthcare) in a buffer containing 150 mM NaCl, 20 mM Tris, pH 8.0.
The b6, F240, 14e and 19b Fab9 fragments used in EM and SPR experiments were produced by IgG digestion at 37uC for 1 h with the enzyme IdeS in a buffer containing 150 mM NaCl, 20 mM Bis-tris, pH 6.0. A reduction and alkylation reaction involving the addition of 10 mM dithiothreitol for 1 h, followed by 5 mM iodoacetamide, produced Fab9 from (Fab9) 2 . Subsequently, the Fab9 was purified away from the Fc fragment and undigested IgG using a Protein A affinity column.

Neutralization assays
For Env-pseudovirus production, HEK293T cells (2610 5 ) were seeded at 2 ml per well in a 6-well tissue culture plate (Corning). After 1 d, the cells reached a confluence of 90-95%. Prior to transfection, the culture medium was refreshed using 2 ml of supplemented medium and the cells were transfected using Lipofectamine 2000 (Invitrogen). For one well, 1.6 mg of BG505.T332N plasmid and 2.4 mg of pSG3DEnv plasmid (obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH from Drs. John C. Kappes and Xiaoyun Wu) were mixed in 250 ml of OPTI-MEM. A 10-ml aliquot of lipofectamine 2000 was mixed with 240 ml of OPTI-MEM immediately before addition to the solution containing the expression plasmids. After incubation for 20 min at room temperature, the mixture was added to the cells to initiate transfection. Culture supernatants were harvested 48 h later.
The TZM-bl reporter cell line, which stably expresses high levels of CD4 and the co-receptors CCR5 and CXCR4 and contains the luciferase and b-galactosidase genes under the control of the HIV-1 long-terminal-repeat promoter, was obtained through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH (John C. Kappes, Xiaoyun Wu, and Tranzyme Inc. Durham, NC) [68,69]. One day prior to infection, 1.7610 4 TZM-bl cells per well were plated on a 96-well plate in DMEM containing 10% FCS, 16 MEM nonessential amino acids, penicillin and streptomycin (both at 100 U/ml), and incubated at 37uC in an atmosphere containing 5% CO 2 for 48 h. A fixed amount of virus (5 ng/ml of p24antigen equivalent) was incubated for 30 min at room temperature with serial 1 in 3 dilutions of each test MAb [70,71]. This mixture was added to the cells and 40 mg/ml DEAE, in a total volume of 200 ml. Two days later, the medium was removed. The cells were washed once with PBS (150 mM NaCl, 50 mM sodium phosphate, pH 7.0) and lysed in Reporter Lysis Buffer (Promega, Madison, WI). Luciferase activity was measured using a Luciferase Assay kit (Promega, Madison, WI) and a Glomax Luminometer according to the manufacturer's instructions (Turner BioSystems, Sunnyvale, CA). All infections were performed in duplicate. Uninfected cells were used to correct for background luciferase activity. The infectivity of each mutant without inhibitor was set at 100%. Nonlinear regression curves were determined and 50% inhibitory concentrations (IC 50 ) were calculated using a sigmoid function in Prism software version 5.0.

Surface plasmon resonance
MAb binding to trimers at 20uC was detected by two SPRbased methods using a Biacore 3000 instrument (GE Healthcare). In the first approach, His-tagged trimers were immobilized on Ni-NTA chips and the binding of solution-phase MAbs was recorded. After removing metallic contaminants via a pulse of EDTA (350 mM) in running buffer (150 mM NaCl, 10 mM Hepes, pH 7.4 plus 0.005% Tween20) for 1 min at a flow rate of 30 ml/ min, the chip was loaded with Ni 2+ by injecting NiCl 2 (2.5 mM) for 1 min at a flow rate of 10 ml/min, resulting in a response of ,50 RU. For all steps between the high EDTA pulses, the running buffer was supplemented with 50 mM EDTA. Purified SOSIP.664-His gp140 trimers (10,000 ng/ml) were injected at 10 ml/min for 2-3 min to capture the equivalent of ,500 RU ( = R L ). Control channels received neither trimer nor NiCl 2 . However, control cycles were performed by flowing the analyte over Ni 2+ -loaded NTA in the absence of trimer; there were no indications of non-specific binding. The analyte (IgG at 1,000 nM (150,000 ng/ml) or Fab at 500 nM (25,000 ng/ml)) was injected into the trimer sample and control channels at a flow rate of 50 ml/min. Association was recorded for 300 s, and dissociation for 600 s. After each cycle of interaction, the NTA-chip surface was regenerated with a pulse of EDTA (350 mM) for 1 min at a flow rate of 30 ml/min, followed by 3 washes with running buffer. A high flow rate of analyte solution (50 ml/min) was used to minimize mass-transport limitation; ln(dY/dX) plots for the association phase were linear with negative slopes, indicating that the binding was largely kinetically limited. Both control-channel and zero-analyte responses were subtracted.
In the second approach, Env-reactive MAbs were captured onto the chip by an immobilized anti-Fc Ab and the binding of solutionphase, untagged BG505 SOSIP.664 gp140 trimers was recorded. Affinity-purified goat anti-human IgG Fc (A80-104A, Bethyl Laboratories, Inc.) was diluted to 50 mg/ml in sodium acetate (pH 4.5) and then amine-coupled to dextran, reaching levels ,10 4 RU, in all four channels of CM5 chips. Env-reactive Abs were added (1 mg/ml in sodium acetate, pH 4.5) to three channels on each chip, at a flow rate of 5 ml/ml, and captured to response levels of 800-900 RU; the fourth channel served as a control surface. BG505 SOSIP.664 gp140 trimers at 200 nM (78,000 ng/ ml) in running buffer (150 mM NaCl, 10 mM Hepes, pH 7.4, 3 mM EDTA plus 0.005% Tween20; note the higher EDTA concentration) were injected at a flow rate of 30 ml/min. Association was recorded for 300 s, and dissociation for 600 s.

Isothermal titration calorimetry
ITC was performed using an Auto-iTC 200 instrument (GE Healthcare) using a protocol similar to one previously described [25,27]. Briefly, prior to conducting the titrations, proteins were dialyzed against Tris-saline buffer (150 mM NaCl, 20 mM Tris, pH 8.0). Absorbance at 280 nm using calculated extinction coefficients served to determine and adjust protein concentrations [66]. The ligand present in the syringe was PGT121 Fab, PGT128 Fab or 2G12 IgG at concentrations ranging between 113 mM and 167 mM, while the BG505 SOSIP.664 trimer was present in the cell at a concentration of 4.3 mM. In each binding experiment, a 5 mcal reference power determination preceded the first injection of 0.5 ml, which was followed by 15 injections of 2.5 ml each at intervals of 180 s. Origin 7.0 software was used to derive the affinity constants (K d ), the molar reaction enthalpy (DH) and the stoichiometry of binding (N), by fitting the integrated titration peaks via a single-site binding model (PGT121) or a two-site binding model (PGT128 and 2G12). All measured and derived thermodynamic parameters of binding are reported in Table 1.

Electron microscopy
SEC-purified BG505 SOSIP.664 gp140 trimers, either alone or as Fab complexes (with b6, F240, 14e, 19b, PGV04, sCD4/17b), were analyzed by negative stain EM. A 3 mL aliquot containing ,0.03 mg/mL of the trimer or Fab-trimer complex was applied for 5 s onto a carbon-coated 400 Cu mesh grid that had been glow discharged at 20 mA for 30 s, then negatively stained with Uranyl formate or Nano-W (Nanoprobes) for 30 s. Data were collected using a FEI Tecnai F20 or T12 electron microscope operating at 120 keV, with an electron dose of ,55 e 2 /Å 2 and a magnification of 52,0006 that resulted in a pixel size of 2.05 Å at the specimen plane. Images were acquired with a Gatan US4000 CCD or Tietz TemCam-F416 CMOS camera using a nominal defocus range of 900 to 1300 nm.

Image processing and 3D reconstruction
Particles were picked automatically using DoG Picker and put into a particle stack using the Appion software package [74,75]. Initial, reference-free, two-dimensional (2D) class averages were calculated using particles binned by five via the Xmipp Clustering 2D Alignment [76] and sorted into classes. Particles corresponding to trimers or complexes were selected into a substack and binned by four before another round of reference-free alignment was carried out using the Xmipp Clustering and 2D alignment and IMAGIC software systems [77]. To analyze the interactions of the non-neutralizing Fabs (b6, F240, 14e, 19b) with BG505 SO-SIP.664 gp140 trimers, the reference-free 2D class averages were examined. The Fabs were clearly visualized if they are bound to the trimer, allowing the percentage of bound trimers relative to unbound trimers to be tabulated.
For Fab-containing complexes, the unliganded trimer (EMDB 5019; [49]) was used as the initial model and refined against reference-free 2D class averages for 89 iterations without imposing symmetry. Fab densities were visible after 3 iterations. This model was then refined against raw particles for an additional 89 cycles with C3 symmetry imposed. For the unliganded BG505 SOSIP.664 gp140 trimer, an ab initio common lines model was calculated from reference-free 2D class averages in EMAN2 [78]. The final volumes for the EMDB 5019 trimer and BG505 SOSIP.664 gp140 trimer reconstructions were nearly identical. EMAN [79] was used for all 3D reconstructions. For the 3D average of BG505 SOSIP.664 with PGV04, 32,867 particles were included in the final reconstruction. For the 3D average of BG505 SOSIP.664 in complex with sCD4 and 17b, 22,145 particles were included in the final reconstruction. For the 3D average of unliganded BG505, 15,352 particles were included in the final reconstruction.