Binding of HIV-1 gp41-Directed Neutralizing and Non-Neutralizing Fragment Antibody Binding Domain (Fab) and Single Chain Variable Fragment (ScFv) Antibodies to the Ectodomain of gp41 in the Pre-Hairpin and Six-Helix Bundle Conformations

We previously reported a series of antibodies, in fragment antigen binding domain (Fab) formats, selected from a human non-immune phage library, directed against the internal trimeric coiled-coil of the N-heptad repeat (N-HR) of HIV-1 gp41. Broadly neutralizing antibodies from that series bind to both the fully exposed N-HR trimer, representing the pre-hairpin intermediate state of gp41, and to partially-exposed N-HR helices within the context of the gp41 six-helix bundle. While the affinities of the Fabs for pre-hairpin intermediate mimetics vary by only 2 to 20-fold between neutralizing and non-neutralizing antibodies, differences in inhibition of viral entry exceed three orders of magnitude. Here we compare the binding of neutralizing (8066) and non-neutralizing (8062) antibodies, differing in only four positions within the CDR-H2 binding loop, in Fab and single chain variable fragment (ScFv) formats, to several pre-hairpin intermediate and six-helix bundle constructs of gp41. Residues 56 and 58 of the mini-antibodies are shown to be crucial for neutralization activity. There is a large differential (≥150-fold) in binding affinity between neutralizing and non-neutralizing antibodies to the six-helix bundle of gp41 and binding to the six-helix bundle does not involve displacement of the outer C-terminal helices of the bundle. The binding stoichiometry is one six-helix bundle to one Fab or three ScFvs. We postulate that neutralization by the 8066 antibody is achieved by binding to a continuum of states along the fusion pathway from the pre-hairpin intermediate all the way to the formation of the six-helix bundle, but prior to irreversible fusion between viral and cellular membranes.


Introduction
The surface envelope (Env) glycoproteins of HIV-1, gp120 and gp41, mediate fusion of the viral and cell membranes [1]. The initial events in the fusion process involve the binding of CD4 and the chemokine co-receptor to gp120 triggering a series of conformational changes in both gp120 and gp41 that culminate in fusion of the viral and cell membranes [2,3,4,5,6,7]. Early steps in this process, representing a possible ''activated'' state of gp120/ gp41, have recently been visualized by crystallography and cryoelectron microscopy of a soluble cleaved HIV-1 Env trimer [8,9]. In these Env structures, gp41 is in a pre-fusion state: the trimeric coiled-coil N-heptad repeat (N-HR, residues 542-591) and the C-terminal heptad repeat (C-HR, residues 623-663) do not interact with one another and both structural elements are solvent accessible. This structure approximates the postulated pre-hairpin intermediate in which the viral and cell membranes are bridged via the C-and N-termini of gp41, respectively [4,10,11]. The final conformational rearrangement occurs further along the fusion pathway and involves the formation of a six-helix bundle, the socalled fusogenic/post-fusogenic state, in which the N-HR trimeric helical coiled-coil is surrounded by three C-HR helices [12,13,14,15,16]. The six-helix bundle brings the viral and cell membranes into contact with one another which eventually leads to fusion [11]. Various constructs have been devised to mimic both the pre-hairpin intermediate [17,18,19] and six-helix bundle conformations of gp41 (Figs. 1A and D) [12,16,18].
The N-HR trimer in the pre-hairpin intermediate state of gp41 is transiently accessible (neutralization half-life ,20 min) during the course of fusion [4,10,20] and is the target of several fusion inhibitors, including various monoclonal antibodies [20,21,22,23,24,25,26,27], peptides derived from the C-HR of gp41 [10,28,29,30], and a peptide derived from the N-HR that inhibits trimerization of the N-HR of gp41 by sequestering the N-HR into heterotrimers [31]. Interestingly the latter potentiates the neutralization activity of N-HR targeted antibodies (and even rescues neutralization activity) by prolonging the temporal window for inhibition [32].
In a series of papers [20,21,22] we described a set of monoclonal antibodies selected from the HuCal Gold human non-immune phage library [33,34] by panning against the chimeric construct N CCG -gp41 [17] which presents the N-HR as a stable, helical disulfide-linked trimer that extends in helical phase from the sixhelix bundle of gp41. Panned antibodies that recognized either sixhelix bundle or N-HR trimer constructs were found to be nonneutralizing. Only antibodies that recognized both the six-helix bundle and N-HR trimer constructs were neutralizing. Subsequent affinity maturation by targeted diversification of the CDR-H2 loop resulted in a fragment antigen binding domain (Fab), known as Fab8066, that was highly potent and broadly neutralizing across a wide range of primary HIV-1 isolates and laboratory-adapted HIV-1 strains [22].
Crystal structures of Fab8066 and a non-neutralizing Fab (Fab8062) from the same affinity matured series differing in only 4 positions in the CDR-H2 loop, complexed to two mimetics of the pre-haipin intermediate (constructs comprising the N-HR timer surrounded by either two or no C-HR helices resulting in 1:1 and 1:3 mimetic:Fab complexes) revealed only subtle structural differences between complexes with the neutralizing and nonneutralizing Fabs [35,36]. Moreover, despite differences in neutralizing activity of three or more orders of magnitude in Env-pseudotyped virus neutralization assays [22], the binding affinities of the neutralizing and non-neutralizing Fabs to various pre-hairpin intermediate mimetics differ by little more than an order of magnitude [35,36]. This suggested to us that the ability to bind to the six-helix bundle, in addition to the exposed N-HR trimer of the pre-hairpin intermediate, may be critical to the neutralization activity of this particular series of Fabs. Interestingly, alanine scanning mutagenesis and Western blot analysis showed that this series of Fabs targets a structurally contiguous epitope on the N-HR that is solvent accessible and located in a shallow groove between two C-HR helices in the six-helix bundle (Fig. 1B) [20,22]. The same residues of the N-HR also comprise part of the binding site in the pre-hairpin intermediate mimetics ( Fig. 2A, B) [35,36]. However, if the Fabs were to bind to the six-helix bundle in the same mode as seen in the crystal structures with the prehairpin intermediate mimetics, there would be steric clash and atomic overlap between the CDR-H1 and CDR-H2 loops of the Fab and one of the C-HR helices of the six-helix bundle (Fig. 2C). The simplest hypothesis for these observations is that binding of this series of Fabs to the six-helix bundle involves displacement of one of the C-HR helices.
Here we investigate the binding of the 8066 and 8062 monoclonal antibodies, as well as of various point mutants within the CDR-H2 loop of 8066, in Fab and single chain-variable fragment (ScFv) formats to mimetics of the pre-hairpin intermediate and six-helix bundle conformation. We show that binding to the six-helix bundle does not involve displacement or fraying of a C-HR helix, indicating that these antibodies must interact somewhat differently with the six-helix bundle and the pre-hairpin intermediate mimetics, although they share a common epitope. Further, we show that there are large differences in binding affinity of neutralizing and non-neutralizing antibodies within this series to the six-helix bundle.

Results and Discussion
Binding affinities of Fab8066 and Fab8062 to pre-hairpin intermediate mimetics Both the neutralizing Fab8066 and the non-neutralizing Fab8062 recognize an epitope at the C-terminal end of the N-HR timer of gp41 comprising a hydrophobic pocket formed at the interface of two N-HR helices (Figs. 2A and B) [35,36]. The relative binding affinities for Fab8066 and Fab8062 to various prehairpin intermediate mimetics, as determined by isothermal titration calorimetry (ITC), differ by factors ranging from 20-fold to only 2-fold. In the case of 5-helix, a single chain pre-hairpin mimetic that presents a single antibody binding site in which the N-HR trimer is surrounded by two C-HR helices, the binding affinities differ by ,20 (#10 vs ,200 nM) [35]; for N35 CCG -N13, an N-HR trimer construct stabilized by disulfide bridges [19] that presents three antibody binding sites, the binding affinities differ by ,8 (,30 vs ,250 nM; Fig. S1 in File S1); and for CCIZN36, an N-HR trimer construct stabilized by a disulfide-linked leucine zipper [37], which also presents three antibody binding sites, the binding affinities differ by a factor of only 2 (,5 versus ,10 nM) [36]. This variation might be attributable to differences in conformational flexibility of the N-HR trimer between the various pre-hairpin mimetics that may facilitate accommodation of the conformational differences within the CDR-H2 binding loops of Fab8066 and Fab8062 [36].

Binding affinities of ScFv variants of Fab8066 and Fab8062 to 5-helix
To further investigate which residues within the CDR-H2 are important for binding target antigen and to ascertain the impact of size on biophysical and biological properties, we engineered ScFv variants (,26 kDa) of Fab8066 and Fab8062 (,49 kDa), denoted as Sc66 and Sc62, respectively, as well as a series of four single substitution mutants of Sc66 (I53L, T56F, T57A and N58V) that match the residue variation in the CDR-H2 region of Sc62 ( Fig. 2B and Fig. S2 in File S1). The binding affinities of these six ScFvs to 5-helix are in the low nM range and cover a range that differs by about an order of magnitude (Table 1 and Fig. S1 in File S1): Sc66 and Sc62 differ by a factor of 7, Sc66 I53L and Sc66 T57A are comparable to the parent Sc66, Sc66 N58V is about three-fold weaker than Sc66, and Sc66 T56F is about 13 and 2-fold weaker than Sc66 and Sc62, respectively. (Note that the N58V and T56F mutations each remove a hydrogen bond to the carboxyamide group of Gln575 of one of the N-HR helices of 5-helix [35]; cf.

HIV-1 neutralization by the ScFv variants of Fab8066 and 8062
The neutralization properties of the ScFv variants were compared to those of the parent Fabs using an Env-pseudotyped neutralization assay [20,38,39] (Table 1). The neutralization potencies (IC 50 ) of Sc66 and Fab8066 are comparable for several HIV-1 strains (Table 1). Although Sc62, unlike Fab8062 [22], displays weak neutralization activity, it is more than 2 orders of magnitude less potent than Sc66. The Sc66 T56F and Sc66 N58V  [19]. N-HR, C-HR and linker residues are shown in green, orange and black (underlined), respectively. Numbering of N-HR and C-HR regions is according to their location in Env from HIV-1 (strain HXB2). Positions in the helical wheel (blue italic) of N-HR residues that are solvent accessible in the six-helix bundle conformation are indicated. (B) Interactions of Fab8066 with N-HR residues in the context of the six-helix bundle construct core S mapped by Ala scanning mutagenesis and immunoblotting [22]. The core S trimer is shown as a surface representation with N-HR and C-HR regions of gp41 in white and light orange, respectively. N-HR surface accessible residues (H564, W571, K574 and Q575) identified as sites of interaction with Fab8066 are shown in distinct colors. (C) Relative migration on SDS-PAGE of the gp41 mimetics used in this study. Molecular weights of constructs and markers (M) are indicated in kDa. N-HR and C-HR denote peptides which assemble to form the six-helix bundle conformation of core SP . (D) Ribbon representations of the gp41 constructs. 5-helix, CCIZN36 and N35 CCG -N13 are pre-hairpin intermediate mimetics with one or more exposed N-HR helices, that are otherwise partially shielded in the six-helix bundle. The three N-HR peptide chains in N35 CCG -N13 and CCIZN36 are stabilized as disulfide-linked trimers by fusion with either a 13-residue repeat of the N-HR [19] or an N-terminal isoleucine zipper segment [37], respectively. 6-helix and 5-helix are single chain polypeptides with the N-HR (N) and C-HR (C) regions connected by a sixresidue linker in the order N-C-N-C-N-C and N-C-N-C-N, respectively [18]. Core S and core SP also form a six-helix bundle but as a trimer consisting of 3 hairpin (N-linker-C) peptides and a hexamer consisting of 3 N-HR and 3 C-HR peptides, respectively (see panel A). doi:10.1371/journal.pone.0104683.g001 mutations result in a reduction in neutralization potency by more than two orders of magnitude. Interestingly at the highest concentration employed in the assay (20 mM), Sc62, Sc66 N58V and Sc62 T56F , inhibit fusion by close to 100%, while Fab8062 displays no inhibitory effect even at this high concentration [22]. In contrast the Sc66 T57A mutation only reduces neutralization activity by a factor of 2, while the Sc66 I53L mutation increases neutralization activity 2-3 fold. Thus, while there is an approx- with the two exposed N-HR helices (green) of 5-helix, and (C) interaction of the CDR-H1 and CDR-H2 loops of Fab8066 (yellow) with two N-HR helices (white) and one C-HR helix (orange) of 5-helix. The addition of a third C-HR helix (transparent orange) to 5-helix to complete the six-helix bundle would result in steric clash with the CDR-H1 and CDR-H2 loops. Color coding in panels A and B is as follows: N-HR and C-HR helices of 5-helix are shown in green and orange, respectively; the CDR heavy and light chain loops of Fab8066 are shown in yellow and white, respectively; the remainder of the heavy and light variable domains are shown in dark red and blue, respectively; the light and heavy constant domains of Fab8066 are shown in pink and light blue, respectively. In panel C, the three N-HR helices are shown in white and residues mapped by alanine scanning mutagenesis of a six-helix bundle construct as the epitope for binding Fab8066 [20,21,22,23,24,25,26,27], are indicated on one of the N-HR helices (helix Na). Also shown in panel A is the design employed to construct the corresponding ScFv by linking the C-terminus of the light chain variable domain (blue) to the N-terminus of the heavy-chain variable domain (dark red) via a 15-amino acid linker (36GGGGS). The coordinates are taken from PDB IDs 3MA9 (Fab8066/5-helix complex [35]) and 1SZT (core S trimer [16] imate correlation between binding affinity to the pre-hairpin mimetic 5-helix and neutralization activity, the differential is far greater (more than two orders of magnitude) for neutralization than binding (approximately an order of magnitude).

Binding of Fabs and ScFvs to the six-helix bundle mimetic core s
Given that all the neutralizing Fabs in our series were shown to bind to both pre-hairpin intermediate and six-helix bundle mimetics of gp41 by Western blotting, we decided to investigate in more detail the binding of our Fabs and ScFv variants to three six-helix bundle mimetics (Figs. 1C, D). These include core S , a trimer in which each subunit comprises a single N-HR and C-HR connected by a six-residue linker [16]; core SP , a hexamer assembled from three N-HR and three C-HR peptides [12]; and finally 6-helix, a single chain in which the N-HR and C-HR helices are sequentially joined together by five six-residue linkers (in the order N-C-N-C-N-C, where N and C represent the N-HR and C-HR helices, respectively) [18].
ITC measurements on the interactions of Fabs and ScFvs with the three six-helix bundle mimetics displayed negligible thermal responses. We therefore resorted to optimize a method to monitor antigen-antibody binding for these various constructs by native polyacrylamide gel electrophoresis (native-PAGE) and analytical size-exclusion chromatography with in-line multiangle light scattering, refractive index and UV detectors (SEC-MALS) (Fig. 3).
On native-PAGE core S migrates between Fab8066 and Sc66 (Fig. 3, left panels), and elutes as a trimer on SEC-MALS (Fig. 3A, right panel, black trace). When mixed at a 1:1 ratio of 10 mM core S trimer and 10 mM Fab8066, nearly all the core S and Fab8066 shift in position to a slower migrating complex (marked 1+1) on native-PAGE above the Fab8066 band (Fig. 3A, left with the six-helix bundle core S antigen (Ag). Left panels: 10 mM core S trimer mixed with Fab or ScFv (shown above the lanes) in molar ratios of 1:1, 1:2 and 1:3 were subjected to 20% homogeneous native-PAGE. Core S and antibody are color coded orange and blue, respectively. Calculated molecular weights of Fab8066, Sc66, Sc62, core S and their complexes are indicated in kDa. Right panels: Protein mixtures (total of 200 mg) at a trimer (core S ) to antibody ratio of 1:1 were subjected to SEC-MALS. Experimental average masses and compositions are indicated. Also shown in panel A (right) are the SEC-MALS traces for core S alone (black), corresponding to a trimer (of calculated mass 368284 g/mol), and a 1:1 mixture of core S and Fab8062 (dashed gray) which shows no evidence of complex formation. The black arrows in panels B and C (right) indicate the retention volume of Sc66 (see Figure S2C in File S1). In panel C (right), the major peak corresponds to uncomplexed ScFv and core S which co-elute, while the minor peak represents a barely detectable amount of 1:1 Sc62:core S complex. Observed masses (g/mol) for free Fab8066, Fab8062 and Sc66 are 470606559, 485606629 and 276306774, respectively. doi:10.1371/journal.pone.0104683.g003 panel). The same 1:1 mixture elutes as a single peak on SEC-MALS corresponding to a combined mass of 1 core S timer+1 Fab8066 (,73 kDa), with a very small shoulder, corresponding to free Fab8066 (,53 kDa). Addition of more Fab8066 at 1:2 and 1:3 core S :Fab ratios does not result in the binding of additional Fab8066 molecules to core S , as evidenced by the presence of an excess Fab8066 band in the native-PAGE lanes (Fig. 3A, left panel). A similar 1:1 mixture of core S and Fab8062 barely shows any shift on native-PAGE and elutes on SEC-MALS mainly in free form with masses of ,52 and 26 kDa, corresponding to Fab8062 and core S , respectively (Fig. 3A, right panel, dashed trace).
In the case of Sc66, both 1:1 and 1:3 core S :Sc66 complexes are apparent on both native-PAGE (Fig. 3B, left panel) and SEC-MALS (Fig. 3B, right panel). That the peaks seen in SEC-MALS correspond to the same compositions as observed by native-PAGE was verified by subjecting peak fractions from the SEC-MALS column to native and SDS-PAGE (Fig. S3 in File S1). In contrast, Sc62 binds only very weakly to core S and only a very small amount of 1:1 complex is apparent by either native-PAGE (Fig. 3C, left) or SEC-MALS (Fig. 3C, right). Similar results are seen with the ScFv mutants: the behavior of the neutralizing mutants Sc66 I53L and Sc66 T57A is very similar to Sc66, while the Top, core SP alone; middle, core SP + Fab8066 (H and L denote heavy and light chains, respectively); bottom, core SP +Sc66. (C) CD of core SP alone (black); a 1:1 mixture of (core SP +Fab8066) minus Fab8066 alone (blue); a 1:1 mixture of (core SP +Sc66) minus Sc66 alone (orange); and the C34 peptide (green, corresponding to the C-HR) alone. The CD data indicate that there is no change in helicity of core SP when complexed to either Fab8066 or Sc66. doi:10.1371/journal.pone.0104683.g004 non-neutralizing mutants Sc66 T56F and Sc66 N58V behave like Sc62 (Fig. S4 in File S1).
Binding of Fabs and ScFvs to the six-helix bundle mimetics core sp and 6-helix The binding of Fab8066 and Sc66 to core S does not allow one to distinguish whether binding occurs directly to the surface accessible region of the N-HR helices between the surrounding C-HR helices in the six-helix bundle conformation, or whether binding involves displacement (or fraying) of one or more C-HR helices in the six-helix bundle permitting interaction with the resulting fully exposed epitope on the trimeric N-HR coiled coil. To this end we investigated the binding of Fab8066 and Sc66 to core SP (Fig. 4) and 6-helix (Fig. 5). In the case of core SP , displacement of the C-HR would result in dissociation of the C-HR helix into free solution (where the free C-HR peptide would adopt a largely random coil structure with a very weak CD helical signature; cf. Fig. 4C and Fig. S4C in File S1), while for 6-helix only a single C-HR could be displaced without unraveling the protein.
The SEC-MALS profiles observed for 1:1 mixtures (10 mM of each component) of Fab8066 and Sc66 with core SP are similar to those obtained with core S . Specifically, a 1:1 complex is observed for the interaction of Fab8066 with core SP (Fig. 4A, middle panel) while 1:1 and 1:3 antigen:ScFv complexes are observed for the interaction of Sc66 with core SP (Fig. 4A, bottom panel). SDS-PAGE of the elution fractions indicate that the 1:1 and 1:3 antigen:antibody complexes contain N-HR and C-HR peptides together with heavy and light chains for Fab8066 (Fig. 4B, middle panel) and a single chain for Sc66 (Fig. 4B, bottom panel). In addition, circular dichroism (CD) of 1:1 mixtures of core SP and Fab8066 or Sc66 indicates that the helicity of core SP remains unchanged upon complexation with antibodies ( Fig. 4C and Fig.  S5A and S5B in File S1). Likewise, native-PAGE (Fig. 5A) and SEC-MALS (Fig. 5B) of a 1:1 mixture (10 mM of each component) of Sc66 and 6-helix reveal the presence of 1:1 and 1:3 antigen:ScFv complexes with no change in helicity of 6-helix upon complexation with Sc66 as monitored by CD (Fig. 5C and Fig. S5D in File S1). Thus, one can conclude unambiguously that no displacement of the C-HR is involved upon binding of Fab8066 Average compositions and masses are indicated next to the peaks. (C) CD spectra of 6-helix alone (black) and a 1:2 mixture of (6 helix+Sc66) minus Sc66 alone (orange). The CD data indicate that there is no change in helicity of 6-helix upon complexation with Sc66. doi:10.1371/journal.pone.0104683.g005 and Sc66 to 6-helix bundle mimetics, and therefore the mode of interaction of these antibodies with the 6-helix bundle mimetics must be slightly different from that observed by crystallography with pre-hairpin intermediate mimetics [35,36] to avoid steric clash between the CDR-H1 and CDR-H2 loops of the antibodies and one of the C-HR helices (Fig. 2C). Since we were unable to successfully use ITC to quantitatively determine the binding affinities of our neutralizing antibodies in either Fab or ScFv formats to the six-helix bundle mimetics, we used native-PAGE band-shift analysis to obtain a semi-quantitative upper limit for the strength of the interaction of Fab8066 with core S . Fab8066 rather than Sc66 was used in this assay since the larger size of Fab8066 improves detectability on gel staining and Fab8066 exclusively forms a 1:1 complex with core S rather than the mixture of 1:1 and 1:3 antigen:Fab complexes observed with Sc66. In the first set of native-PAGE band-shift experiments we prepared 1:2 mixtures of core S trimer with Fab8066 in 10 mM Tris-HCl, pH 7.6, 150 mM NaCl at Fab concentrations ranging from 10 to 1.25 mM. As the limit of detectability with Coomassie staining was reached at 1.25 mM of the core S -Fab8066 1:1 complex (Fig. 6A, top), silver staining was used to provide improved detection at lower concentrations. A set of dilutions (Fig. 6A, bottom) was made in which Fab8066 at a constant concentration of 1 mM was titrated by addition of core S to give final concentrations ranging from 1 mM to 0.25 mM core S . The increase in band intensity of the 1:1 complex, accompanied by a gradual decrease in intensity of free Fab8066, in going from 0.25 to 1 mM core S shows that complex formation is accompanied by depletion of Fab8066. At the lowest concentration of core S only a small amount of the 1:1 complex can be detected. This serves to set an upper limit of K D #250 nM for the binding of Fab8066 to core S . By way of contrast, a lower limit of K D ..10 mM for the binding of Fab8062 and Sc62 to core S can be estimated from the SEC-MALS and native-PAGE data shown in Figs. 3 A and C, respectively.
To further refine our K D estimate for the interaction of Fab8066 with core S we made use of SEC-MALS, injecting very small amounts of core S (3 mg) and Fab8066 (6 mg) (Fig. 6B). Deconvolution of the SEC-MALS profile together with concentration estimations from the known volumes of the peaks yielded an approximate K D of 70 nM.

Concluding remarks
The HIV-1 broadly neutralizing Fabs selected from the HuCal Gold human non-immune phage library by panning and subsequent affinity maturation (by targeted diversification of the CDR-H2 loop) against N CCG -gp41, a chimera comprising the fully exposed N-HR trimer linked in helical phase to a disulfide-linked six-helix bundle of gp41, are characterized by binding to both the fully exposed N-HR trimer and the six-helix bundle [20,22]. Fabs from this series that only bound to either the N-HR trimer or the six-helix bundle were found to be non-neutralizing. In this light we sought to characterize in more detail the binding of two monoclonal antibodies from this series (in Fab and ScFv formats) to mimetics of the pre-hairpin intermediate and six-helix bundle conformations of gp41. One of the antibodies, Fab8066 (Sc66), is broadly neutralizing, while the second, Fab8062 (Sc62), differing in only four positions within the CDR-H2 loop, is non-neutralizing [22]. The difference in neutralization potency between these two antibodies ranges from two to over three orders of magnitude, depending on format (ScFv versus Fab), yet the difference in binding affinity to three different pre-hairpin intermediate mimetics ranges from only 2 to 20-fold. Further single point mutations in the CDR-H2 of Sc66, at positions where the CDR-H2s of Sc66 and Sc62 differ, show that the residues at positions 56 and 58 are critical for neutralization activity, but have only a relatively small effect on binding affinity to the pre-hairpin intermediate mimetic 5-helix (Table 1). SEC-MALS and native-PAGE analysis indicate that a significantly larger difference ($ 150-fold) in binding affinity of neutralizing (Fab8066) and nonneutralizing (Fab8062) Fabs is obtained with core S (and expected to be the same for core SP and 6-helix). Moreover, SEC-MALS, CD and SDS-PAGE demonstrate conclusively that no displacement (or local fraying) of the C-HR helix (helices) occurs upon Fab binding to the six-helix bundle mimetics. Previous scanning mutagenesis and Western blotting analysis showed that the epitope for the binding of Fab8066 to the six-helix bundle involves surfaceexposed residues at the C-terminus of the N-HR located in a groove between two adjacent C-HR helices [20,22]. This epitope overlaps with that observed in crystal structures of Fab8066 and Fab8062 with pre-hairpin intermediate mimetics [35,36]. Since the introduction of a C-HR helix would result in steric clash with the CDR-H1 and CDR-H2 loops in the latter complexes (Fig. 2C), these results imply that the details of the interaction of the Fabs with solvent accessible N-HR residues in the six-helix bundle must be sufficiently different from those observed when the N-HR trimer is fully exposed (as in the pre-hairpin intermediate mimetics) to circumvent steric clash with a C-HR helix.
What is the relevance of the ability of neutralizing Fabs within our series to bind to both pre-hairpin intermediate mimetics and six-helix bundles of gp41? The time window at which the neutralizing Fabs act is very similar to that of peptides derived from the C-HR of gp41, as well as a peptide (N36 Mut(eg) ) engineered from the N-HR to not interact with the C-HR, thereby inhibiting gp41 trimerization by forming heterotrimers with one or two gp41 N-HR helices [20,22,32]. The C-HR derived inhibitory peptides can only target the fully exposed N-HR trimer of the pre-hairpin intermediate, and thus both the Fabs and the N36 Mut(e,g) peptide must also target this state. In addition, the N36 Mut(e,g) peptide could also potentially act at subsequent stages in the fusion process in which the six-helix bundle is either partially formed (the pre-bundle state [40]) or fully formed but before irreversible membrane fusion actually takes place. The observation that our series of neutralizing Fabs can also bind to the six-helix bundle suggest that the neutralizing Fabs can target six-helix bundle conformations prior to actual fusion in addition to the prehairpin intermediate, and that the ability to target a continuum of states from the pre-hairpin intermediate to the six-helix bundle prolongs the time window over which these Fabs can inhibit fusion, thereby increasing their potency and neutralization properties.

E. coli plasmid constructs
Previously reported plasmid constructs used in this study are the six-helix bundle constructs 6-helix [18] and core S [16,21], and the pre-hairpin mimetics 5-helix [18], and N35 CCG -N13 [19]. Constructs used for the first time are the ScFvs Sc66, Sc62, Sc66 I53L , Sc66 T56F , Sc66 T57A , Sc66 N58V , and the six-helix bundle construct core SP . The strategy for linking the light and heavy variable regions as a single chain to generate Sc66 from Fab8066 is shown in Fig. 2A [41] and its amino acid sequence in Figure S1A. DNA inserts were cloned either in pET11a or pET15 vectors between Nde1/BamH1 and Nco1/BamH1 sites, respectively. Site-directed mutagenesis was carried out using the Quik-Change mutagenesis kit (Agilent Technologies) on Sc66 DNA (template) to derive Sc62, Sc66 I53L , Sc66 T56F , Sc66 T57A and Sc66 N58V constructs. DNA sequences and purified proteins were verified by DNA sequencing and electrospray ionization mass spectrometry (ESI-MS), respectively.
Protein purification and folding E. coli BL21(DE3) bearing the appropriate plasmid vector were grown in Luria-Bertani medium and induced for expression at an A 600 ,0.7 for 4 hrs. All of the expressed proteins used in this study accumulated in the insoluble fraction [19]. After isolating the insoluble fraction (,80% purity), the various gp41 constructs were further purified on size-exclusion Superdex-200 or 75 columns under denaturing conditions. They were finally subjected to reverse-phase HPLC using a water/TFA and acetonitrile/TFA gradient. The desired amount of protein was dialyzed against 50 mM sodium formate, pH 3, concentrated (8-10 mg/ml) and stored. When needed, an aliquot of the protein was diluted in 10 mM Tris-HCl, pH 7.6, 150 mM NaCl (buffer A), dialyzed against the same buffer and concentrated.
Core SP was derived from a construct comprising the N-HR (residues 546-582) and C-HR (residues 625-662) of gp41 separated by a spacer SGLVPRGSGG (Fig. 1A). Thrombin cleavage was carried out resulting in the separation of the N-HR and C-HR regions prior to purification. These peptides were assembled by dialysis from 6 M guanidine hydrochloride into 10 mM Tris-HCl, pH 7.6, 150 mM NaCl (buffer A) to form a sixhelix bundle complex but without a spacer connecting the N-HR and C-HR regions as in core S , and fractionated as a complex on a Superdex-75 column in buffer A. The pre-hairpin intermediate mimetic N35 CCG -N13 trimer was prepared as described [21]. ESI-MS of N35 CCG -N13 showed a trimer mass of 22634 Da close to the calculated value of 22638 Da (367546). Experiments with N35 CCG -N13 were carried out in 50 mM sodium acetate, pH 5.2 because of its insolubility above this pH.
ScFvs were folded using previously described protocols [41,42,43]. Upon folding from denaturing conditions, Sc66 exhibited nearly complete disulfide bridge formation (Fig. S1B in File S1). The ScFvs were finally subjected to size-exclusion chromatography on Superdex-75 in buffer A. The concentrated ScFvs (1-2 mg/ml) are stable upon storage for several months at 4uC, and retain their monomeric status, binding affinity as determined by ITC, and neutralization activity (see below). All of the purified and folded ScFvs used for various characterizations are shown in Fig. S1C in File S1. By SEC-MALS (Fig. S1D in File S1) Sc66 was shown to elute with a monomer mass of 276306774 g/mol (calculated 25540). Sc66 mutants and Sc62 also exhibit similar elution profiles.

Isothermal titration calorimetry
ITC measurements were made using a Microcal high-precision iTC200 titration calorimeter (GE Healthcare). Titrations were performed at 28uC. Protein in the sample cell was titrated either with nineteen 2-mL injections or fifteen 2.45-mL injections of titrant. ITC of Fab8066 with core S or 6-helix (as titrant) and of Sc66 with 6-helix (as titrant) when performed at 28uC or 37uC in 10 mM Tris-HCl, pH 7.6, 150 mM NaCl (buffer A) yielded no thermal response. Data were analyzed using the Origin software provided with the instrument. from 100-150 ml. Typically 200 mg of total protein was mixed in a molar ratio of 1:1 antigen to antibody. Individual proteins were injected as controls. The sample was centrifuged at 12,800 rpm for 4 min in an Eppendorf 5415 centrifuge and the supernatant applied to a pre-equilibrated Superdex-75 column (1.0630 cm, GE Healthcare) at a flow rate of 0.5 ml/min at room temperature and eluted in buffer A, unless stated otherwise. Under these conditions, the eluting concentrations are expected to be 6-10 mM, near the concentrations used for native-PAGE. Molecular masses were calculated using the Astra software (version 6) provided with the instrument.

Viral fusion assays
Assays to quantify viral entry in the presence of antibodies employed Env-pseudotyped HIV strains with host cells genetically modified to express CD4 and co-receptors CCR5 and CXCR4, as described previously [20,38,39]. Envs were derived from four HIV-1 subtype B laboratory-adapted strains, HXB2, SF162, JR CSF and 89.6. The IC 50 represents the midpoint of the sigmoid dose-response curve obtained from a fit to a simple dose-response relationship (percent infection = S6100/(1+[Ab]/IC 50 ) where S is a scale factor.

Native-PAGE
Samples were mixed to achieve a final concentration of 10 mM core S (as trimer, unless stated otherwise) and appropriate molar ratio of antibody as indicated. This concentration was chosen because it represents a good detection limit to differentiate between the binding of neutralizing and non-neutralizing Fabs and ScFvs. They were incubated at room temperature for 30 min and 2 ml of sample was layered for electrophoresis on a 20% homogeneous Phastgel using 1 ml/eight-lane applicators (unless stated otherwise) and native buffer strips (GE Healthcare). Gels were stained in PhastGel Blue R, destained and digitized. For detection at lower concentrations, 4 ml/six-lane applicators and PlusOne silver staining kit (GE Healthcare) were used.

Circular dichroism
CD spectra were recorded in 10 mM Tris-HCl, pH 7.6, 150 mM NaCl (buffer A) at 20uC on a JASCO J-810 spectropolarimeter using Spectra Manager software and a 0.1 cm path length flat cell. Compositions similar to those used for band-shift and SEC-MALS analysis were used. Mean residue ellipticity was calculated using the instrument's software.

Supporting Information
File SI Five supporting figures. Figure S1. ITC analysis of Fab and ScFv binding to two pre-hairpin intermediate mimetics.
Binding of (A) Fab8066 and (B) Fab8062 to the disulfide-linked trimer N35 CCG -N13 in 50 mM sodium acetate pH 5.2 buffer (required to maintain the solubility of N35 CCG -N13). 10 mM Fab was in the cell and N35 CCG -N13 (,100 mM in monomer) was the titrant. (C to F) Binding of ScFvs to 5-helix in 10 mM Tris-HCl, pH 7.6, 150 mM NaCl. 5-helix (5-7 mM) in the cell was titrated with the designated ScFv at a concentration of 50-70 mM. HIV-1 neutralizing antibodies are shown in the left-hand panels and nonneutralizing on the right-hand panels. Values of the equilibrium association constant (K A ) are shown in each panel. Figure S2.  Figure S3. Native-and SDS-PAGE of peak fractions from SEC-MALS analysis of 1:1 mixtures of core S -Fab8066 and core S -Sc66 complexes. Gels shown in (A) and (B) correspond to elution profiles and masses shown for core S -Fab8066 and core S -Sc66 in Figs. 3A and B, respectively, in the main text. Numbering of lanes corresponds to elution volume. Figure S4. Native-PAGE of core S -Sc66 single mutant complexes.
No binding is observed with Sc66 T56F , and relatively weak (1:1) binding is seen with Sc66 N58V . Shifts in the presence of mutants Sc66 I53L and Sc66 T57A closely resemble those for Sc66. Core S and ScFv are color coded orange and blue, respectively. Calculated molecular weights of core S , ScFv and their corresponding complexes are indicated in kDa. Figure S5. Raw CD data for Fab8066 and Sc66 complexed to the six-helix bundles core SP and 6-helix. (A) core SP +Fab8066; (B) core SP +Sc66; (C) the C-HR peptide C34; (D) 6 helix+Sc66. (PDF)