Interactions of HIV-1 Inhibitory Peptide T20 with the gp41 N-HR Coiled Coil*

Cellular entry of human immunodeficiency virus type 1 (HIV-1) involves fusion of viral and cellular membranes and is mediated by structural transitions in viral glycoprotein gp41. The antiviral C-peptide T20 targets the gp41 N-terminal heptad repeat region (N-HR), blocking gp41 conformational changes essential for the entry process. To probe the T20 structure-activity relationship, we engineered a molecular mimic of the entire gp41 N-HR coiled coil using the 5-Helix design strategy. T20 bound this artificial protein (denoted 5H-ex) with nanomolar affinity (KD = 30 nm), close to its IC50 concentration (∼3 nm) but much weaker than the affinity of a related inhibitory C-peptide C37 (KD = 0.0007 nm). T20/C37 competitive binding assays confirmed that T20 interacts with the hydrophobic groove on the surface of the N-HR coiled coil outside of a deep pocket region crucial for C37 binding. We used 5H-ex to investigate how the T20 N and C termini contributed to the inhibitor binding activity. Mutating three aromatic residues at the T20 C terminus (WNWF → ANAA) had no effect on affinity, suggesting that these amino acids do not participate in T20 binding to the gp41 N-HR. The results support recent evidence pointing to a different role for these residues in T20 inhibition (Peisajovich, S. G., Gallo, S. A., Blumenthal, R., and Shai, Y. (2003) J. Biol. Chem. 278, 21012–21017; Liu, S., Jing, W., Cheung, B., Lu, H., Sun, J., Yan, X., Niu, J., Farmar, J., Wu, S., and Jiang, S. (2007) J. Biol. Chem. 282, 9612–9620). By contrast, mutations near the T20 N terminus substantially influenced inhibitor binding strength. When Ile was substituted for Thr in the second T20 position, a 40-fold increase in binding affinity was measured (KD = 0.75 nm). The effect of this affinity enhancement on T20 inhibitory potency varied among different viral strains. The original T20 and the higher affinity T20 variant had similar potency against wild type HIV-1. However, the higher affinity T20 variant was significantly more potent against T20-resistant virus. The findings suggest that other factors in addition to binding affinity play a role in limiting T20 potency. As a mimetic of the complete gp41 N-HR coiled coil region, 5H-ex will be a useful tool to further elucidate mechanistic profiles of C-peptide inhibitors.

larger C37 (residues 625-661) (4,6,26,27). These peptides are derived entirely from the C-HR sequence and thus are shifted in the N-terminal direction compared with T20 (Fig. 1B). The interactions of C34 and C37 with gp41 are greatly stabilized by residues Trp 628 , Trp 631 , and Ile 635 near the C-HR N terminus (4). Their bulky hydrophobic side chains pack into a deep hydrophobic pocket on the surface of the N-HR coiled coil. T20 lacks these pocket binding residues and their stabilizing effect. However, T20 does contain bulky hydrophobic residues (Trp 670 , Trp 672 , and Phe 673 ) at its C terminus that might pack into a similar pocket at the other end of the N-HR coiled coil.
High resolution structures of the gp41 trimer-of-hairpins have aided our understanding of the mechanism of C-peptide inhibition. These structures have enabled the design of polypeptides that mimic the gp41 N-HR coiled coil and bind C34/C37, thereby providing a tool to probe the structure-activity relationships of the inhibitors (26, 28 -30). No similar tool is available for investigating T20 inhibition in detail. The structures of the gp41 trimer-of-hairpins do not include the T20 C terminus (9 residues) nor the gp41 N-terminal segments that putatively interact with it. Furthermore, gp41 N-HR-derived peptides predicted to interact with T20 are poorly soluble and difficult to use in solution phase interaction assays (6). Here we describe the design of a soluble protein (denoted 5H-ex) that mimics the putative T20-binding site on the N-HR coiled coil. 5H-ex interacts with T20 with an equilibrium dissociation constant (K D ) of 30 nM, close to the T20 50% inhibitory concentration (IC 50 ) of 3 nM. Using this protein, we explored the extent to which the N and C termini of T20 contribute to its binding activity. First, we showed that the MPER-derived residues at the peptide C terminus do not stabilize the 5H-ex/T20 interaction. Second, we identified an N-terminal substitution that significantly enhanced T20 binding affinity and improved peptide inhibitory activity against T20-resistant HIV-1. The results suggest that T20 binding to the N-HR coiled coil is stabilized primarily by residues derived from the C-HR and not the MPER. 5H-ex is likely to be a useful tool in probing the structure-activity relationship of T20.

EXPERIMENTAL PROCEDURES
Design and Purification of Extended 5-Helix Variants-The T20-binding protein 5H-ex was designed as an extended variant of the engineered HIV-1 entry inhibitor 5-Helix (26). The original 5-Helix contained three N-HR segments of 40 residues (HIV-1 HXB2 residues Gln 543 to Ala 582 ) and two C-HR segments of 38 residues (His 625 to Glu 662 ) alternately connected using 5-amino acid Gly/Ser linkers. In 5H-ex, the N-HR segments were extended at their N terminus by 13 amino acids ( 530 MGAASMTLTVQAR 542 ), and the C-HR segments were extended at their C terminus by 7 amino acids ( 663 LDK-WASL 669 ). To accommodate the disparity in segment lengths, the linkers connecting each C-HR segment to the following N-HR segment were lengthened to 9 amino acids (GSS-GGSGSG). In the slightly truncated 5H-ex variant (denoted 5H-ex⌬5), each N-HR segment was shortened by 5 amino acids at its N terminus. In the mutant 5H-ex variant with altered C37 affinity, W571R and G572D substitutions were introduced into the third N-HR segment to disrupt the hydrophobic pocket at the C terminus of the N-HR coiled coil. All of the proteins contained a C-terminal hexahistidine tag to facilitate purification. Genes encoding each of these proteins were assembled by a combination of overlap extension PCR and QuikChange sitedirected mutagenesis (Stratagene). The constructs were cloned into the pEAD4 vector (31) and confirmed by sequencing the entire open reading frame.
5-Helix was produced as previously described (26). 5H-ex proteins were recombinantly expressed in Escherichia coli (strain RP3098) grown in 2ϫ YT broth at 37°C and induced with isopropyl-␤-D-thiogalactopyranoside (0.4 mM). After 3 h, bacterial pellets were collected by low speed centrifugation and resuspended in TBS (50 mM Tris, pH 8, 100 mM NaCl) supplemented with 4 M guanidine hydrochloride. The bacterial resuspensions were lysed by sonication and clarified by high speed centrifugation. Protein was purified by metal affinity chromatography (nickel-nitrilotriacetic acid-agarose; Qiagen) in solutions of TBS, 4 M guanidine hydrochloride and serially dialyzed, first into TBS with 6 M urea and then into TBS with 500 mM arginine. The final dialyzed sample was further purified on a Superdex 75 column (GE) running TBS, 500 mM Arg. 5H-ex protein eluted at the monomeric molecular weight was assessed to be Ͼ95% pure by SDS-PAGE. The samples were stored diluted (ϳ5 M) in TBS, 500 mM Arg at 4°C and concentrated  (46). In native Env prior to receptor activation, gp41 is held in a metastable conformation by a canopy of gp120 proteins (green). Receptor binding to gp120 stimulates gp41 to extend and insert its fusion peptide segment (red) into the target cell membrane. The N-HR (gray) and C-HR (blue) regions of the gp41 ectodomain are transiently exposed in this prehairpin state. Subsequently, gp41 collapses into the trimer-of-hairpins conformation that brings the gp41 fusion peptides, transmembrane regions (purple), and their associated membranes into the close proximity for membrane fusion. The actual disposition of gp120 in both the prehairpin and trimer-of-hairpins states is uncertain; for clarity, the protein is omitted in the schematic of the trimer-of-hairpins conformation. B, a diagram of HIV-1 gp41 identifying its fusion peptide (FP), N-HR, C-HR, MPER (MP), transmembrane (TM), and cytoplasmic (cyto) domains. Amino acid sequences above and below the diagram are derived from the N-HR and C-HR/MPER regions of Env HXB2 ; all but the MPER sequence WNWF (magenta) were used in the design of 5H-ex. The N-HR and C-HR segments found in the original 5-Helix are boxed in gray and blue, respectively, whereas the sequences of C37 and T20 are denoted by lines. The side chains of the C-HR amino acids marked with ϩ pack into the hydrophobic pocket at the C terminus of the N-HR coiled coil.

T20 Interactions with the gp41 N-HR Coiled Coil
up to 60 M using a Centricon 10 (Millipore) immediately prior to use. Concentration was determined by absorbance at 280 nm using the method of Edelhoch (32).
C-peptide Production-T20 and its variants (T20 YI , T20 NYT , T20 NYI , and f-T20) were synthesized using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry by the peptide synthesis facility at the Kimmel Cancer Center, Thomas Jefferson University. Cleaved, desalted peptides were purified to homogeneity by reverse phase high pressure liquid chromatography (HPLC) using a Vydac C-18 column and a linear gradient of acetonitrile in water containing 0.1% trifluoroacetic acid. Absorbance at 280 nm was used to ascertain the concentration of all T20 variants except f-T20. The f-T20 peptide contained a fluorescein label (Molecular Probes) at its N terminus; its concentration was determined in 50 mM potassium phosphate (pH 9) using an extinction coefficient of 87,500 M Ϫ1 cm Ϫ1 at 490 nm (Molecular Probes).
Production of MBP-T20 and MBP-C37 Fusion Proteins-Genes encoding MBP-T20 and MBP-C37 (27) in the pET-14b vector were kindly provided by Dr. Michael S. Kay (University of Utah). In these fusion proteins, the C-peptide sequence is attached to the C terminus of MBP (maltose-binding protein) through a variable linker (SSSENLYFQGS for MBP-T20 and SSSSSGG for MBP-C37). The proteins also contain a C-terminal hexahistidine tag. Constructs encoding MBP-T20 mutant variants described in the text were generated using QuikChange site-directed mutagenesis (Stratagene) and confirmed by sequencing. These fusion proteins were recombinantly expressed in E. coli (strain RP3098) and purified by metal affinity chromatography and gel filtration (Sephacryl S200 HR, GE). Purified protein (Ͼ95% pure by SDS-PAGE analysis) was dialyzed into water and stored lyophilized at Ϫ20°C. The protein was resuspended in TBS, and its concentration was determined using absorbance at 280 nm.
Interaction Assays-The reported K D values were measured for the interaction of C-peptides and their binding proteins in solution. The binding parameter was determined by sampling the free (unbound) concentration of R-C37 or f-T20 in reaction mixtures using a Kinexa 3000 flow fluorimeter (Sapidyne Instruments). As reaction mixtures were passed through the instrument's optical flow cell, a minute fraction of the total unbound C-peptide was captured by a small amount (ϳ5 l) of chromatography beads coated with 5-Helix or 5H-ex. The resulting increase in bead fluorescence (⌬f) was directly proportional to the concentration of unbound C-peptide in the reaction mixture.
The beads were prepared as previously described (33). Briefly, 5-Helix or 5H-ex (100 g) was incubated with activated azlactone (50 mg; Pierce) in 50 mM sodium carbonate (pH 9.5, 1 ml) at 4°C overnight. Nonspecific binding sites were subsequently blocked using bovine serum albumin (10 mg) in 1 M Tris (pH 8.0, 1 ml) at 20°C for 1 h. Before use, the bead mixture was diluted into 35 ml of 10TBS (10 mM Tris, pH 8, 100 mM NaCl) containing 0.02% NaN 3 . For each sample, ϳ0.5 ml of the diluted bead mixture was passed through the flow cell capillary, and a small bead pack was retained on a 20-m screen. All of the bead preparations were tested for linear responses to both sample volume and fluorescent C-peptide concentration to ensure that the fluorescent measurement accurately reflected the free C-peptide concentration in reaction mixtures.
Interaction experiments were carried out at 25°C in solutions of 10TBS containing 100 g/ml bovine serum albumin, 0.02% NaN 3 , and 1 mM phenylmethylsulfonyl fluoride. For direct binding measurements, 5-Helix or 5H-ex was titrated into a fixed concentration of R-C37 or f-T20 and incubated for 4 -72 h (empirically determined to ensure that equilibrium had been achieved). Each sample was run through the Kinexa 3000 in duplicate, and concentration dependence to ⌬f was fit to a general bimolecular equilibrium binding model using the manufacturer's software (Sapidyne Instruments).
For C37 displacement assays, MBP-T20 was titrated into solutions containing 1 nM R-C37 N656D and 10 nM 5H-ex (or 5-Helix). Each experiment also included the following two control solutions: 1) 1 nM R-C37 N656D alone and 2) 1 nM R-C37 N656D and 1 M 5-Helix (enough 5-Helix to chelate 99.9% of the R-C37 N656D ). The unbound R-C37 N656D in each reaction mixture was sampled twice in the Kinexa 3000 flow cell using 5-Helix-coated beads. The fraction of unbound R-C37 N656D was determined by normalizing the measured ⌬f by the maximum and minimum ⌬f values obtained from the two control solutions. The dependence of free R-C37 N656D on MBP-T20 concentration was fit to an exact expression for three-state competitive binding (34) using a least squares algorithm implemented in Visual Basic. The output of the algorithm (MBP-T20 K D value) required input of the R-C37 N656D K D value, determined through direct binding experiments to be 1.6 nM. This value is consistent with the ϳ15% free R-C37 N656D observed in the absence of MBP-T20 in each experiment (see Fig. 3). Importantly, the presence of MBP-T20 at high concentrations did not perturb the efficiency of R-C37 N656D capture by 5-Helix-coated beads. This observation was made for each C37 displacement assay by showing that a control solution containing 1 nM R-C37 N656D and MBP-T20 at its highest concentration generated a fluorescence signal within 10% of the maximum ⌬f (data not shown).
For C37 displacement assays using the 5H-ex double mutant variant W572R-G572D, the higher affinity C-peptide R-C37 T639I was substituted for R-C37 N656D . The R-C37 T639I / mutant 5H-ex K D value was determined to be 1.2 nM by direct binding measurements. For C37 displacement assays using MBP-T20 NYI , the higher affinity R-C37 N656S was likewise substituted for R-C37 N656D . The R-C37 N656S /5H-ex K D value was determined to be 0.024 nM. Finally, for C37 displacement assays using T20 peptide, the maximum ⌬f values were determined for each T20 concentration using control solutions containing R-C37 N656D and T20 peptides. This protocol change was implemented because T20 peptide in excess of 125 nM was observed to reduce the maximum fluorescent signal through a nonspecific mechanism.
Inhibition Experiments-Antiviral activities of peptides and proteins were measured in cell-cell fusion and viral infectivity assays that have been described in detail previously (4,33). For cell-cell fusion assays, Env HXB2 and TAT-expressing Chinese hamster ovary cells (clone 7d2, kindly provided by M. Krieger, Massachusetts Institute of Technology) were cocultured with HeLa-CD4-LTR-␤-Gal cells (M. Emerman, obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, National Institutes of Health) in the presence of varying concentrations of inhibitor for 15 h at 37°C. Multinucleated syncytia were identified by staining with 5-bromo-4chloro-3-indolyl-␤-D-galactopyranoside (X-gal; Sigma) and manually scored. For the infectivity assay, virions pseudotyped with Env HXB2 were generated by cotransfection of the Env-deficient HIV-1 NL4-3 genome (pNL4-3E Ϫ V Ϫ Luc ϩ ) and an Envexpressing plasmid (pEBB_HXB2) into 293T cells. HIV-1 harvested 48 h post-transfection was used to infect HOS-LES cells (N. Landau, obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, National Institutes of Health) in the presence of varying inhibitor concentrations. The level of viral infectivity was measured 48 h later by assaying for luciferase production in infected cells (luciferase assay system; Promega).
IC 50 values were determined by fitting the normalized syncytia number or luciferase activity (F) to inhibitor concentration ([Inh]) using the Langmuir equation: F ϭ (1 ϩ [Inh]/IC 50 ) Ϫ1 . For viral infectivity experiments performed in the presence of T20 and either 5-helix or 5H-ex, combination indices (CI) were determined to asses synergistic, additive, or antagonistic behavior of inhibitor pairs (35,36). Using mixtures of T20 and 5-Helix as an example, the CI is defined for inhibitors that act in a mutually exclusive manner as shown in the following equation, Here x is the percentage of inhibition (0 Ͻ x Ͻ 100) observed in a mixture of T20 at concentration [T20] and 5-Helix at concentration . IC x, T20 is the T20 concentration required to achieve x percent inhibition in the absence of 5-Helix, and IC x, 5-Helix is the 5-Helix concentration required to achieve x percent inhibition in the absence of T20. Synergistic activity leads to CI values less than 1, whereas antagonistic behavior leads to CI values greater than 1. A CI value equal to 1 indicates that the activities are additive. The average CI values shown in Fig. 6D were calculated from various levels of inhibition observed using different mixtures of T20 and 5-Helix or 5H-ex.

Design of an Extended 5-Helix Variant-
The design of the T20-binding protein 5H-ex followed the strategy utilized to develop the HIV-1 entry inhibitor 5-Helix (26). 5-Helix contains three peptide segments derived from the N-HR and two peptide segments derived from the C-HR alternately connected (N-C-N-C-N) through short Gly-Ser linkers to form a single polypeptide. The alternating connectivity enables the N-HR and C-HR segments to pack in the antiparallel fashion observed in the trimer-of-hairpins structure. 5-Helix adopts a conformation that resembles the helical core of the gp41 trimer-of-hairpins except that one C-HR segment is absent (5,37). This vacancy creates a single C-peptide-binding site that interacts fully with the C37 sequence but only partially with the shifted T20 sequence (Fig. 2A). Consequently, 5-Helix binds strongly to C37 (K D ϭ 0.00065 nM) but relatively poorly to T20 (K D ϳ 850 nM; Fig. 2, B and C). T20 lacks the 13 N-terminal residues of C37 that include the amino acids Trp 628 , Trp 631 , and Ile 635 , the side chains of which pack into the deep hydrophobic pocket at the C terminus of the N-HR coiled coil. The data suggest that The instrument flow cell contained beads that specifically captured free (unbound) f-T20. The difference in fluorescence signals (⌬f) measured before sample load (0 ml) and after the sample washout (2 ml) was proportional to the free f-T20 concentration in reaction mixtures (see "Experimental Procedures"). C and F, titrations of fluorescent C37 (2.5 pM) and T20 (20 pM) with either 5-Helix or 5H-ex. In C, R-C37 (squares) and f-T20 (circles) were titrated by 5-Helix, whereas in F, f-T20 was titrated by either 5-Helix (circles) or 5H-ex (triangles). The fraction of unbound C-peptide was calculated by normalizing each ⌬f to a minimum and maximum ⌬f obtained from samples in which the C-peptide is fully bound or fully unbound. The K D values were determined by fitting these data to a bimolecular equilibrium binding model (solid lines). this sequence accounts for approximately half of the free energy that stabilizes C37 binding. 3 Three properties of the original 5-Helix protein made it an attractive scaffold for further engineering. First, 5-Helix is extremely stable to thermal and chemical denaturation (T M Ͼ 100°C, C1 ⁄ 2 Ͼ 7 M guanidine hydrochloride at 20°C) (26). We hypothesized that such rigidity would assist in stabilizing extensions of N-HR and C-HR segments that otherwise may be highly flexible in the context of unlinked peptides. Second, 5-Helix is much more soluble than N-HR peptides of similar length because of the presence of two attached C-HR segments (6). We reasoned that this enhanced solubility would reduce the tendency of the hydrophobic extensions to promote aggregation. Third, C-peptide binding to 5-Helix is a simple bimolecular interaction that is much less complicated to analyze than the assembly of a hexameric complex formed by N-HR and C-HR peptides.
In engineering a 5-Helix-like T20-binding protein, we assumed that the helical structures of N-HR and C-HR segments extend through the T20-binding site, although no structural data exists to support this conjecture. In 5H-ex, each N-HR segment was lengthened by 13 amino acid residues at its N terminus (Fig. 2D). Initially, we attempted to stabilize this N-HR extension by adding 11 amino acid residues to the C terminus of each C-HR segment so that its sequence would extend as far as the T20 sequence. However, this engineered protein was insoluble in aqueous solutions. The aggregation problem was linked to the last four residues of each extended C-HR segment: Trp 670 , Asn 671 , Trp 672 , and Phe 673 , collectively denoted WNWF. Substituting the WNWF sequence with Gly-Ser-Gly-Ser produced a soluble, well behaved protein with the stable helical structure expected from the design (supplemental Fig. S1). In the final version of 5H-ex, each N-HR segment extended from Met 530 to Ala 581 , and each C-HR segment extended from His 625 to Leu 669 .
Binding Properties of 5H-ex-When correctly folded, 5H-ex should contain an intact C37-binding site as well as the putative T20 interaction sequence. 5H-ex had the same binding affinities for C37 and two C37 variants (N656S (K D ϭ 0.024 nM) and N656D (K D ϭ 1.6 nM)) as the original 5-Helix (not shown), confirming the proper fold of the extended molecule. In a solution phase interaction assay, 5H-ex bound fluoresceintagged T20 (f-T20) with a K D value of 32 nM, an ϳ30-fold improvement over the 5-Helix/T20 affinity (Fig. 2, E and F). The K D value is significantly closer to the observed T20 IC 50 (ϳ3 nM), but the T20 affinity is still more than 4 orders of magnitude weaker than the C37 affinity (see "Discussion"). T20 binding 5H-ex is sufficient to displace the lower affinity C37 variant N656D from its binding site (Fig. 3, A-C). Titrating either free T20 peptide or a maltose-binding protein-T20 fusion (MBP-T20, kindly provided by Dr. Michael Kay, University of Utah) into a solution of prebound 5H-ex/C37 N656D complexes led to increasing amounts of free (unbound) C37 N656D . By comparison, MBP-T20 did not significantly displace C37 N656D bound to 5-Helix, consistent with the much lower affinity of T20 for this original construct (Fig. 3C). Quantitative characterization of C37 N656D displacement from 5H-ex using a three state competitive binding model yielded K D values for T20 and MBP-T20 of ϳ30 nM, in close agreement with the affinity obtained using f-T20 in the direct-binding assay (Fig. 3D).
To confirm the predicted binding site for T20, we generated a 5H-ex variant with the double mutation W571R/G572D. Because Trp 571 and Gly 572 line the hydrophobic pocket at the C terminus of the N-HR coiled coil, the dual substitution should significantly destabilize C37 interaction but have only a minor effect on T20 binding. Consistent with this prediction, C37 affinity for the mutant 5H-ex variant was reduced by 4 orders of FIGURE 3. Properties of the 5H-ex/T20 interaction. A, overview of the C37 displacement assay. T20 and R-C37 compete for the same binding site on 5H-ex, and therefore titration of T20 (or MBP-T20) leads to an increase in the free (unbound) concentration of R-C37. The circled R denotes the rhodamine label linked to C37. B, Kinexa 3000 fluorescence response as solutions of 5H-ex (10 nM), R-C37 N656D (1 nM), and MBP-T20 (indicated concentration) were passed over beads that specifically captured unbound R-C37 N656D . The dashed and dotted lines were generated by control samples in which R-C37 N656D is fully bound or fully unbound, respectively (see "Experimental Procedures"). The fraction of unbound R-C37 N656D in each sample was calculated by normalizing the difference fluorescence signal ⌬f to the minimum and maximum ⌬f obtained from the control solutions. C, C37 N656D displacement from 5H-ex (triangles) or 5-Helix (circles) as T20 (open symbols) or MBP-T20 (filled symbols) was titrated into reaction mixtures. The data points represent the means Ϯ S.E. of at least three separate measurements and have been fit to a three-state competitive binding model (solid lines). D, tabulated K D values for the interaction of T20 with either 5-Helix or 5H-ex as determined using the direct binding assay (f-T20) or the C37 displacement assay (T20, MBP-T20). The values represent the means and S.E. of three or more separate measurements. E, effect of the W571R/G572D substitution in the N-HR hydrophobic pocket on C-peptide binding affinities. The data are represented as fold change in K D value relative to the wild type 5H-ex/Cpeptide K D . The K D values were determined using the direct binding assay for C37 values or the C37 displacement assays for MBP-T20 values. FEBRUARY 6, 2009 • VOLUME 284 • NUMBER 6 magnitude (K D ϭ 7 nM). By contrast, the dual substitution did not significantly alter the K D value for T20 binding to 5H-ex (K D ϭ 49 nM) (Fig. 3E). Together, the binding and displacement data imply that T20 adopts the predicted mode of interaction with the N-HR coiled coil, packing into the same hydrophobic groove as C37 but outside of the pocket region.

T20 Interactions with the gp41 N-HR Coiled Coil
Probing the T20 Structure-Activity Relationship: Contributions of the C-terminal CHR/MPER Sequence to T20 Binding-The increased stability of the T20/5H-ex complex (compared with T20/5-Helix) must be due to additional interactions of the C-peptide C terminus with the N-HR N terminus. We initially focused on the last four residues of T20: the WNWF sequence known to play a critical role in HIV-1 inhibition. Substituting Ala residues for the three aromatic amino acids (WNWF 3 ANAA) functionally ablates the potent antiviral activity of the peptide (1,24,25). However, recent studies suggest that the WNWF sequence may not directly bind the gp41 N-HR (38). To test this possibility, we investigated how the WNWF sequence contributes to the interaction between T20 and 5H-ex in two ways. First, we generated the mutant MBP-T20 ANAA variant with Ala residues in place of the aromatic amino acids. Second, we produced a slightly truncated 5H-ex variant (5H-ex⌬5) that lacked the residues modeled to interact with the WNWF sequence (Met 530 , Gly 531 , Ala 532 , Ala 533 , and Ser 534 in each of the three N-HR segments; Fig. 4A). C37 N656D displacement assays were used to measure the binding affinities of the following two interactions: 1) MBP-T20 ANAA and fulllength 5H-ex and 2) MBP-T20 and the truncated 5H-ex⌬5. In both cases, the C37 N656D displacement curves were indistin-guishable from results obtained using MBP-T20 and the fulllength 5H-ex (Fig. 4B). The data indicated that the three interacting molecular pairs had the same K D value, implying that the WNWF sequence does not contribute to the stability of the T20/5H-ex complex. The results corroborated recent findings that demonstrated minimal destabilization of T20/N-HR peptide complexes when amino acids near the T20 C terminus were replaced with their enantiomers (38). These observations suggest that the WNWF sequence of T20 is involved in interactions outside of the gp41 N-HR region during inhibition of HIV-1 entry (see "Discussion").
Because the WNWF segment did not appear to interact with the N-HR, we refocused our attention on the adjacent eight residues at the C-HR/MPER junction ( 662 ELDKWASL). Using scanning mutagenesis, we confirmed a role for this segment in stabilizing the T20/5H-ex interaction. Alanine substitution for Leu 663 , Asp 664 , and Trp 666 and glutamine substitution for Ala 667 significantly decreased free energy of binding (Fig. 4C). The greatest effect was seen for the mutation at Leu 663 , a residue previously predicted to lie within the hydrophobic N-HR/ C-HR interface (8). Three and four residues down from Leu 663 , Trp 666 and Ala 667 would be expected to make contact with the N-HR if the C-HR helical register were maintained. Interestingly, glycine substitution at Ala 667 had almost no impact on T20 binding affinity (K D ϭ 35 nM). Taken together, the consequences of mutations at Ala 667 suggest that the residue packs within the N-HR/C-HR interface in a manner that cannot accommodate the added bulk of a glutamine side chain. Finally, a putative helical structure would also position Asp 664 in close proximity to Arg 542 of the N-HR; the small destabilizing effect of the D664A mutation might reflect disruption of this potential electrostatic interaction. Together, the pattern of destabilization strongly suggests that the C-HR ␣-helix extends into this region at least through Ala 667 .
Probing the T20 Structure-Activity Relationship: Affinity Enhancement through the T639I Substitution-We have previously shown that C37 binding to 5-Helix is exquisitely sensitive to mutations in the C-HR locus formed by amino acids Asn 637 , Tyr 638 , and Thr 639 . 4 Specifically, a C37 peptide variant with the T639I substitution was found to bind 5-Helix with ϳ9-fold higher affinity (K D ϳ 70 fM). The T20 sequence begins at Tyr 638 and thus does not include Asn 637 at the N terminus of the tripeptide segment. To investigate the effect of this C-HR locus on T20 affinity, we first elongated the peptide sequence to include Asn 637 . This new MBP fusion variant, denoted MBP-T20 NYT , possessed the same activity in the C37 displacement assay as the original MBP-T20, indicating that the additional Asn residue did not significantly change binding affinity (Fig. 5A). However, substituting Ile for Thr 639 created an MBP construct (denoted MBP-T20 NYI ) that was able to displace C37 N656D fully from 5H-ex at extremely low concentrations (Fig. 5A). The T639I variant also showed significantly greater activity when the higher affinity C37 N656S peptide (K D ϭ 0.024 nM) was used in place of C37 N656D in the displacement assay (Fig. 5B). The K D value of T20 NYI is 0.75 nM, indicating that the T639I substitution increased T20 peptide affinity 40-fold.  . Role of the C-terminal C-HR/MPER sequence in stabilizing T20 binding to 5H-ex. A, schematic representations of 5-Helix, 5H-ex, and the slightly truncated 5H-ex⌬5. The extent of the C-peptide binding site of each protein is indicated relative to T20 and T20 ANAA peptides. B, C37 N656D displacement measured using 5H-ex and MBP-T20 ANAA (open circles), 5H-ex and MBP-T20 (filled circles), 5H-ex⌬5 and MBP-T20 (open squares), and 5-Helix and MBP-T20 (filled diamonds). The data represent the means and S.E. of three or more independent experiments and were analyzed as described in the legend to Fig. 3. C, effect of C-peptide point mutations on the stability of the T20/5H-ex complex. Binding free energies were calculated using the formula ⌬G o ϭ ϪRTln(K D ), where the K D values were obtained using the C37 N656D displacement assay with MBP-T20 constructs as titrants. The horizontal lines indicate the binding energies of wild type MBP-T20 with either 5H-ex (solid) or 5-Helix (dotted).

T20 Interactions with the gp41 N-HR Coiled Coil
The observed enhancement in binding affinity caused by the T639I substitution did not translate into increased potency against wild type HIV-1 HXB2 (Fig. 5, C and D). The IC 50 values for T20 and the slightly elongated T20 NYT (with the additional N-terminal Asn residue) were ϳ3 nM, indistinguishable from the IC 50 values of the corresponding peptides with the T639I substitution (T20 YI and T20 NYI , respectively). The absence of a correlation between IC 50 and K D is similar to results obtained in studies of 5-Helix inhibition (33). The finding implies that T20 inhibitory potency against wild type Env is limited by factors independent of equilibrium binding affinity (e.g. association kinetics, see "Discussion"). However, T20 YI and T20 NYI were significantly more potent against T20-resistant HIV-1 (Fig. 5,  C-E). The dual substitution G547D/I548T in the gp41 N-HR region led to a 100-fold reduction in T20 inhibitory potency that was partially overcome by the T639I substitution (39). Similar results were also obtained with a different T20-resistant Env variant, V549E (data not shown) (40). The data suggest that T20 inhibition of these resistant Env glycoproteins is at least partially dependent on peptide binding affinity.
Inhibitory Properties of 5H-ex-Like T20, 5-Helix and 5H-ex potently inhibit gp41-mediated membrane fusion in both cellcell fusion and HIV-1 infectivity assays (Fig. 6). 5-Helix was previously shown to target the gp41 C-HR region and disrupt trimer-of-hairpins formation in a manner complementary to C-peptide inhibition (26). The inhibitory mechanism was established, in part, by demonstrating that 5-Helix and C37 mutually antagonized each other's inhibitory activity; the residues that mediate gp41 binding for each inhibitor were sequestered in the interface of associated 5-Helix/C37 complexes, thereby reducing the concentrations of active inhibitor in the fusion assays. To confirm that the extended 5H-ex also inhibited by targeting the C-terminal region of the gp41 ectodomain, we similarly tested for antagonism using MBP-C37 and MBP-T20 constructs. Because of the bulkiness of the attached MBP, neither C-peptide fusion protein significantly blocked viral entry at 100 nM (27). However, titration of 5H-ex into assays containing 100 nM of either MBP construct resulted in a significant reduction in 5H-ex potency (Fig. 6A). By contrast, 5-Helix inhibitory activity was only reduced in the presence of MBP-C37 (Fig. 6B). The presence of 100 nM MBP-T20 led to no significant change in 5-Helix potency, consistent with the low affinity of the 5-Helix/T20 interaction. Thus, we conclude that, despite its extra bulk (ϳ30% increase in mass and estimated length), 5H-ex can still target the C-terminal region of the gp41 FIGURE 5. Effect of the T639I substitution on T20 binding and inhibitory activities. A, C37 N656D displacement from 5H-ex by MBP-T20 (squares), MBP-T20 NYT (triangles), and MBP-T20 NYI (circles). The data were obtained and analyzed as described in the legend to Fig. 3. B, as in A, except that the higher affinity C37 N656S was used in place of C37 N656D . Note that only the MBP-T20 NYT (triangles) and MBP-T20 NYI (circles) titrations are shown. C, titration of viral infectivity by T20 (squares) and T20 YI containing the T639I substitution (circles). HIV-1 was pseudotyped with either wild type Env HXB2 (filled symbols) or its mutant variant containing T20 escape mutations G547D/I548T (open symbols). The data represent the means Ϯ S.E. of three to seven separate experiments and have been fit to a Langmuir function (solid lines) to determine IC 50 (see "Experimental Procedures"). D, C-peptide IC 50 values determined for the inhibition of HIV-1 psuedotyped with wild type Env (gray) or the G547D/I548T mutant variant (black). Inset, IC 50 values for inhibition of wild type HIV-1 are plotted on an expanded scale. E, table of C-peptide IC 50 and K D values. Res refers to the T20-resistant Env containing the G547D/I548T substitutions. ND, not determined. and HIV-1 infectivity (Virus) assays and represent the means and S.E. of three or more independent experiments. The CI determined for multiple mixtures of T20 and 5-Helix or 5H-ex was calculated as described under "Experimental Procedures." T20 Interactions with the gp41 N-HR Coiled Coil FEBRUARY 6, 2009 • VOLUME 284 • NUMBER 6 ectodomain. In fact, the similarity in IC 50 values for 5-Helix and 5H-ex suggests that these inhibitors encounter little steric hindrance when binding gp41 (see "Discussion").
We next explored the antiviral activity in mixtures of T20 peptide and either 5-Helix or 5H-ex. Because 5-Helix poorly interacts with T20, we specifically asked whether these two inhibitors might display additive or synergistic antiviral activity. In the experiments shown in Fig. 6C, viral infectivity was measured in the presence of 3 nM T20 and 8, 40, or 200 nM 5-Helix. The level of inhibition correlated well with that expected if the two inhibitors acted additively by binding gp41 in a mutually exclusive manner (solid line). By comparison, considerably less inhibitory activity was observed for the same levels of T20 and 5H-ex, which can interact over this concentration range. Synergistic, additive, and antagonistic behaviors can be discriminated quantitatively using the combination index (CI, see "Experimental Procedures" for details) (35,36). The average CI value determined for multiple T20 and 5-Helix concentrations was approximately 1, expected for additive inhibitory activities. The CI value greater than 2 for T20 and 5H-ex mixtures indicated antagonistic inhibitory activities and was consistent with lower active inhibitor concentrations due to 5H-ex/T20 binding.

DISCUSSION
In this study, we have utilized the 5-Helix design strategy to create a molecule that mimics the gp41 N-HR coiled coil region that interacts with the HIV-1 inhibitory peptide T20. The engineered 5H-ex protein binds T20 with a K D value of 30 nM, substantially larger than the C37 K D value of 650 fM. The comparatively low T20 affinity might signify that 5H-ex does not fully recapitulate the gp41 structure that binds T20. Recent studies have suggested that the gp41 MPER and fusion peptide regions can form a stable complex (41)(42)(43). Hence, T20 binding may be stabilized through an interaction of its WNWF sequence with residues from the gp41 fusion peptide. 5H-ex variants with N-HR segments extending further toward the gp41 N terminus are currently under development to test this possibility. Alternatively, other factors independent of the T20/gp41 interaction might stabilize peptide binding during HIV-1 membrane fusion. T20 has been shown to interact with lipid membranes much more avidly than peptides similar to C37 (1,44). This enhanced membrane interaction appears to be linked to aromatic residues in the WNWF sequence that critically determines T20 inhibitory potency (1,2,44). The results suggest that T20 binding to gp41 might be additionally stabilized through interactions of the peptide tail with the target cell membrane. In this manner, T20 could inhibit at lower concentrations than it binds to 5H-ex, as observed in these studies.
Although T20 variants with the T639I substitution bound 5H-ex with significantly higher affinity, they did not inhibit wild type HIV-1 with higher potency. The poor correlation between binding affinity and inhibitory potency is similar to results obtained in a study of 5-Helix inhibition (33). Both T20 and 5-Helix target transient intermediate gp41 conformations during the fusion process. As a consequence, their inhibitory potencies are influenced by nonequilibrium parameters such as the lifetime of the sensitive intermediate state and the rate of inhibitor association (20,21,33,45). These kinetic factors are the main determinants of 5-Helix antiviral activity (33), and, as a result, 5-Helix binding affinity can be altered by 4 orders of magnitude without significantly affecting the inhibitory potency. The similarity in IC 50 values (2-3 nM) for T20, T20 NYI , and even C37 despite the wide range of K D values (650 fM to 30 nM) is consistent with C-peptide inhibitory activity being restricted by kinetic factors.
The higher affinity T20 variants were found to be 4-fold more potent against T20-resistant Env. The results imply that resistance emerges at least partially through a mechanism of binding destabilization. However, the 4-fold improvement in inhibitory activity was small compared with the observed 40-fold enhancement in binding affinity. In fact, even C37, with its subpicomolar affinity, was unable to fully overcome the resistance phenotype (Fig. 6, D and E). These findings suggest that the mechanism of escape from C-peptide inhibition is more complex than mere affinity disruption. One possibility is that the N-HR mutations may reduce the rate of C-peptide association, thereby decreasing the amount of inhibitor able to bind during the lifetime of the sensitive intermediate state. Future interaction experiments using mutant 5H-ex variants might help to elucidate the full mechanism of C-peptide resistance.
The inhibitory activity of 5H-ex illuminates an important property of C-HR exposure. Compared with 5-Helix, 5H-ex targets a larger portion of gp41 with potentially higher affinity. However, the IC 50 values for 5-Helix and 5H-ex are equal, implying that inhibition by 5H-ex is limited by the same kinetic factors that restrict 5-Helix potency. These kinetic factors lead to an inverse relationship between the IC 50 value and the rate of inhibitor association (k on ) (33). Any steric obstruction hindering access to the C-HR region would necessarily reduce the k on value for the bulkier inhibitor, thereby increasing its IC 50 . Hence, the identical potencies of 5-Helix and 5H-ex imply that the two inhibitors possess similar k on values. These findings support an access pathway to the C-terminal region of the gp41 ectodomain largely free of steric obstacles.