ABSTRACT
The N17 region of gp41 in HIV-1 is the most conserved region in gp160. mRNA selection technologies were used to identify an adnectin that binds to this region and inhibits gp41-induced membrane fusion. Additional selection conditions were used to optimize the adnectin to greater potency (5.4 ± 2.6 nM) against HIV-1 and improved binding affinity for an N17-containing helical trimer (0.8 ± 0.4 nM). Resistance to this adnectin mapped to a single Glu-to-Arg change within the N17 coding region. The optimized adnectin (6200_A08) exhibited high potency and broad-spectrum activity against 123 envelope proteins and multiple clinical virus isolates, although certain envelope proteins did exhibit reduced susceptibility to 6200_A08 alone. The reduced potency could not be correlated with sequence changes in the target region and was thought to be the result of faster kinetics of fusion mediated by these envelope proteins. Optimized linkage of 6200_A08 with a previously characterized adnectin targeting CD4 produced a highly synergistic molecule, with the potency of the tandem molecule measured at 37 ± 1 pM. In addition, these tandem molecules now exhibited few potency differences against the same panel of envelope proteins with reduced susceptibility to 6200_A08 alone, providing evidence that they did not have intrinsic resistance to 6200_A08 and that coupling 6200_A08 with the anti-CD4 adnectin may provide a higher effective on rate for gp41 target engagement.
IMPORTANCE There continue to be significant unmet medical needs for patients with HIV-1 infection. One way to improve adherence and decrease the likelihood of drug-drug interactions in HIV-1-infected patients is through the development of long-acting biologic inhibitors. This study describes the development and properties of an adnectin molecule that targets the most conserved region of the gp41 protein and inhibits HIV-1 with good potency. Moreover, when fused to a similar adnectin targeted to the human CD4 protein, the receptor for HIV-1, significant synergies in potency and efficacy are observed. These inhibitors are part of an effort to develop a larger biologic molecule that functions as a long-acting self-administered regimen for patients with HIV-1 infection.
INTRODUCTION
The gp41 subunit of the HIV envelope glycoprotein (gp160) is the most conserved portion of the protein and is responsible for driving fusion of the viral and target cell plasma membranes during the entry stage of HIV infection (1). As a result of its critical role in HIV infection, gp41 has been targeted by a wide array of therapeutic approaches, including small molecules, peptides, and antibodies (2). Such inhibitors tend to act by impeding the formation of a functional six-helix-bundle structure, which is a prerequisite for virus-cell fusion (3).
The highly conserved 17-amino-acid region in gp41, denoted N17, is alpha helical in structure and has been a key target for therapeutic and vaccine approaches (4, 5). For instance, a monoclonal antibody (MAb) with modest but broad-spectrum antiviral activity against this region has been selected (6). However, additional studies aimed at increasing the potency of the molecule suggested that the full-size MAb was sterically inhibited from reaching the target, as smaller single-chain or Fab versions of the MAb were significantly more potent (7, 8). Smaller peptides, such as the D peptide (4, 9), are also less sterically hindered in accessing the target region of gp41 and can serve as potent inhibitors of fusion.
Adnectins are small proteins (10 to 12 kDa) derived from the 10th type III fibronectin domain of human fibronectin (10–13). Selection from libraries based on this scaffold using mRNA display has previously generated adnectins that bind to CD4 and inhibit infection of CD4-expressing target cells by HIV-1 (14). We hypothesized that, due to their small size, adnectins would be able to access the N17 site of gp41 better than MAbs, thereby inhibiting functional six-helix-bundle formation and membrane fusion with high potency. Here, we report the selection and optimization of an adnectin that binds to the N17 site of gp41 and blocks HIV-1 infection by precluding membrane fusion. While this anti-gp41 adnectin is a potent inhibitor of HIV infection in its own right, linking it to the previously described anti-CD4 adnectin confers significant synergy in antiviral activity compared to each of the individual adnectin inhibitors. This approach may provide a path to develop highly potent, broad-spectrum combination inhibitors of HIV-1 infection.
RESULTS
Selection of gp41-binding adnectins that exhibit anti-HIV activity.As gp41 in the prefusion state is metastable and difficult to maintain in its native conformation as an isolated protein, synthetic peptides that mimic the presentation of specific sequences within gp41 were utilized as selection targets (Table 1). Each target peptide contained an artificial trimerization domain (termed isoleucine zipper [IZ]) at its N terminus (15), followed by a gp41-derived consensus sequence from the N17 region. Some peptides (IZN21 and IZN24) contained additional flanking sequence for context, and the mutant IZN17 peptide contained three of the more common polymorphisms in the N17 region found in clinical isolate databases (16). Targets were validated by confirming binding by the D5 Fab′ (17) and lack of binding by the T20 (18) peptide (data not shown). These targets were alternated during selection both to enhance tolerance of binding to common polymorphisms in the region and to avoid selection of binders to nonnative aspects of the trimer structure, such as the C-terminal end. Additionally, the IZIZ negative-control peptide, which contains only the trimerization domain and no gp41-derived sequence, was used to remove adnectins that bound to the IZ sequence from the selections.
Targets used in adnectin selection and assays
Selections were initiated using a library of adnectins, in which the 3 potential binding loops were randomly mutated, as shown in Fig. 1. Construction of the library utilized a series of overlapping oligonucleotides containing randomized regions synthesized from defined mixtures of phosphoramidite trimers (19–22) in the BC, DE, and FG loops (Fig. 1). Using techniques described previously (14), the library was translated in vitro, and each unique adnectin protein was coupled with the mRNA encoding it via a puromycin linker (Fig. 2A). These coupled mRNA-protein fusion molecules were then exposed to one of the artificial peptide targets containing sequences derived from gp41 (Table 1), allowing specific binders to be enriched in the population. An overview of the entire selection process is given in Fig. 2A to H.
Adnectin library design. (A) Depiction of the regions of the 10Fn3 domain that were randomized to create the 3-loop library used for selection, with the BC loop in blue, the DE loop in green, and the FG loop in red. The structure was taken from Protein Data Bank (PDB) entry 1FNF. (B) Sequence of the wild-type 10Fn3 domain. The highlighted sequences in the wild-type 10Fn3 indicate the residues that are randomized in the library, with color coding as in panel A.
Selection pathway to an optimized anti-gp41 adnectin. Each step used in the selection and optimization process is outlined in the flow chart.
From the initial selection (Fig. 2B), which used the IZN17 and mutIZN17 peptides as targets (with clearing of nonspecific binders using peptide IZIZ), 860 unique adnectin sequences were identified as potential binders to the N17 site. These individual adnectins were expressed in bacteria with an attached 6×His tag and isolated with a high-throughput purification system (Fig. 2C) (14). A simple enzyme-linked immunosorbent assay (ELISA)-based binding assay was devised, where the same IZN17 and IZIZ targets used during selection were bound to a neutravidin-coated plate, followed by addition of the adnectins at concentrations between 50 nM and 1 μM. After washing, readout of bound adnectins was accomplished with an anti-6×His antibody coupled with horseradish peroxidase (HRP). Of the 860 adnectins screened in this fashion, 388 (45%) bound strongly to the IZN17 peptide with minimal reactivity to IZIZ, indicating specific binding to the N17 site. These adnectins were further screened for aggregation propensity using size exclusion chromatography. Propensity to aggregate is a property disfavored for the development of a therapeutic biologic agent due to manufacturing (23) and immunogenicity (24, 25) concerns. Fifty adnectins (13% of the N17-specific binders) showed sufficient monomer content (>50%) to warrant antiviral testing.
A cell-cell fusion assay (26, 27) was used as an initial screen for antiviral activity. Adnectins were added at concentrations up to 1 μM to a mixture of two cell lines, one expressing viral envelope protein and the other expressing CD4 and coreceptors, along with a fusion-induced luciferase reporter. Inhibition of cell-cell fusion is detected as a decrease in luciferase activity. In this assay, 23 adnectins were capable of inhibiting fusion by at least 50% at concentrations of less than 1 μM. These adnectins were further assessed in a replicating virus assay, using a modified NL4-3 virus (RepRluc) containing a luciferase reporter gene (14, 28). Of these, the 2428_G03 adnectin was chosen, as it possessed good and consistent inhibitory activity in both assays (Table 2). In the cell-cell fusion assay, 2428_G03 inhibited cell fusion with a 50% effective concentration (EC50) of 150 nM, while against the RepRluc reporter virus, the adnectin possessed an EC50 of 139 ± 25 nM. As a control, the Fab′ portion of the D5 antibody (6) was examined and exhibited an EC50 of 268 ± 30 nM against replicating virus.
Properties of anti-gp41 adnectins
The 2428_G03 adnectin differs from the naturally occurring 10Fn3 domain in the BC (7-amino-acid), DE (4-amino-acid), and FG (6-amino-acid) loops (Table 3, left to right). The FG loop of 2428_G03 is 4 amino acids shorter than the FG loop of naturally occurring 10Fn3.
Protein sequences of selected anti-gp41 adnectins
Optimization selections improve anti-gp41 adnectin properties.In order to enhance the antifusion (and theoretically the antiviral) activity of the 2428_G03 adnectin, a series of optimization selections were conducted to improve the affinity of the adnectin for the N17 site. A new library was constructed in which each of the residues in the BC, DE, and FG loops was diverged from the 2428_G03 sequence at a frequency of approximately 50%. This new library was selected using the IZN17, IZN21, and IZN24 peptides alternately as targets, while the peptide concentration was progressively decreased from 100 nM to 1 nM (Fig. 2D). Alternation of the peptide targets ensured selection of binding to structural features they held in common, while decreasing the target concentration favored selection of higher-affinity binders.
After five rounds of selection in this manner, the enriched variants were cloned and sequenced (Fig. 2E). Seventy-seven unique adnectins were identified, which were expressed, purified, and characterized. A sequence analysis of the recovered clones indicated that the DE loop (SVLS) was nearly invariant and therefore likely to be critical for target binding, while multiple positions within the BC and FG loops tolerated a variety of substitutions (Fig. 3). Functional screening by ELISA and antiviral assays identified one variant, 4058_H08, as having the greatest improvement in antiviral activity (Table 2), with antiviral activity against the NL4-3 reporter virus of 27 ± 6 nM and a similar EC50 (22 nM) in the cell-cell fusion assay. This variant differed from its 2428_G03 parent by 3 residues in the BC loop and 2 residues in the FG loop, while the DE loops remained identical (Table 3).
Sequence analysis of variants recovered from optimization 1. For each randomized position within the BC, DE, and FG loops, the size of the letter representing a given amino acid represents the relative frequency of that amino acid at that position within the 77 clones isolated from optimization selections. The colors indicate amino acid properties as follows: red, negatively charged; blue, positively charged; green, polar; black, hydrophobic. The figure was generated with the WebLogo tool (31).
Further characterization of 4058_H08.The interaction of 4058_H08 with the IZN21 peptide was characterized by isothermal titration calorimetry (ITC). 4058_H08 at a concentration of 25 μM was titrated into a solution of IZN21 peptide at 2.5 μM in 10 mM HEPES, pH 7.4, 150 mM NaCl, 0.25% dimethyl sulfoxide (DMSO) at 25°C. Based on the results (Fig. 4), an affinity of 11 nM was calculated, which was close to the affinity obtained via Biacore surface plasmon resonance (SPR) using the IZN24 target (3.1 nM) (Table 2), especially considering the differences in technique, temperature, and target. Additionally, the ITC experiment indicated a stoichiometry value (n) of 0.38, which is consistent with one adnectin binding per IZN21 trimer.
Isothermal titration calorimetry result for 4058_H08 binding to IZN21 peptide. Representative results are shown.
The tolerance of 4058_H08 activity for amino acid substitutions within the N17 site was also examined. Substitutions representing the more common polymorphisms in the N17 sequence found in the Los Alamos National Laboratory (LANL) database (16) were introduced into the NL4-3 reporter virus background as indicated in Table 4, and the resulting viruses were challenged with 4058_H08 or a control inhibitor, 3-[[6-[1-[(4-methoxyphenyl)methyl]indol-6-yl]indol-1-yl]methyl]benzoic acid, which is also known to bind to the N17 site of gp41 (29, 30). Since NL4-3 contains a minor polymorphism at position 580, the major I580V polymorphism (90.2%) was examined, along with the less frequent L565M (39.3%). Additional polymorphisms at Q567R (4.9%) and I573V (1.1%), as well as alanine substitutions not present in the database, L568A and K574A, were also examined.
4058_H08 activity tolerance for amino acid substitutions in the N17 site
The chosen amino acid substitutions were expected to have minimal impact on the potency of the control compound based on structural modeling of the binding site on gp41 (29). As expected, the control compound maintained full potency against all 6 polymorphic viruses, within typical assay variability. 4058_H08 also demonstrated similar EC50s against all 6 polymorphic viruses, indicating tolerance of its antiviral activity against common polymorphisms within the N17 site (Table 4).
Additional optimization by off rate selections and subsequent loop shuffling.In order to explore the sequence space of this adnectin family more fully and to identify more potent variants, another round of optimization (Fig. 2F) was performed. A new library was constructed based on the 4058_H08 sequence, using the same techniques that were applied to produce the 2428_G03 optimization library. This library was subjected to selection against the IZN24 and mutIZN24 peptides, alternating these targets successively, as was done previously. The peptide concentration was maintained at 100 nM for the first three rounds. For rounds 4 to 7, the peptide concentration was reduced to 10 nM to increase the stringency of selection. For round 8, an off rate selection protocol was implemented, in which the population was bound to 10 nM IZN24 for 30 min followed by addition of nonbiotinylated IZN24 in 100-fold excess. Incubation was continued for 1 h before the biotinylated IZN24 and bound adnectins were captured on streptavidin beads. Theoretically, all the adnectins that were not bound to, or fell off, the biotinylated IZN24 would bind to the nonbiotinylated target and not be selected. As variants with higher off rates would be more likely to be released from the biotinylated peptide and bind to the nonbiotinylated peptide, variants with relatively low off rates would be favored to be captured by the streptavidin beads. Two additional rounds of off rate selections were conducted, increasing the incubation time with nonbiotinylated peptide to 2 h and 4 h, respectively. Individual adnectins that survived this selection process were then cloned, sequenced, and characterized (Fig. 2G).
One hundred ninety-four unique adnectins were isolated directly from this 4058_H08 optimization (optimization 2) (see Table S1 in the supplemental material). The sequences shared several conserved features, the strongest of which was the 100% conservation of the SVLS motif in the DE loop. In addition, the BC loops included an interesting binary pattern, where either the 6th or the 7th residue was always an aromatic, but never both. These two classes had different patterns in other locations in both loops. The Weblogos (31) of the BC and FG loops for the full 194 sequences are shown in Fig. 5. However, although many sequences in both classes retained good binding affinity and potency, none showed significant improvement in antiviral activity over 4058_H08. The most potent sequences had a wide variety of BC and FG loops and a range of biophysical properties, so we hypothesized that different combinations of loops might have improved potency and behavior, even if the affinity for our artificial target was not changed. Therefore, in an effort to further improve on 4058_H08, we generated 320 separate adnectins that contained shuffled pairs of the BC and FG loops found in the best 27 of the 194 original adnectins (Fig. 2G). When these recombinant adnectins were assessed for antiviral activity and biophysical properties, 5497_F03 was identified as an improved variant of 4058_H08. Although the antiviral potency was only modestly improved (EC50 = 19 ± 7 nM), the biophysical properties of 5497_F03 showed improved stabilization of the protein, as evidenced by the increase in melting temperature from 46°C to 53.5°C. In addition, the percentage of monomer was increased to >99%, indicative of a well-behaved adnectin protein. This 5497_F03 adnectin contained one reversion in the BC loop to an amino acid found in 2428_G03 (K25E; numbering based on the wild-type [WT] 10Fn3 sequence) and one change in the FG loop from 4058_H08 (N83D), along with two reversions to amino acids found in the wild-type Fn3 sequence (R80Y and L85A) (Table 3).
Sequence analysis of variants recovered from optimization 2. Representations of the diversity of BC and FG loop sequences are presented as in Fig. 3. (A) Diversity of the full set of 192 sequences obtained from oprimization 2. (B) Diversity of the subset of sequences with an aromatic residue in position 6 of the BC loop. (C) Diversity of the subset of sequences with an aromatic residue in position 7 of the BC loop.
As optimization of the canonical loop sequences of this adnectin family did not produce the degree of improvement in potency we desired, we hypothesized that sequences flanking the canonical loops might influence potency, as well. A new library of variants using 5497_F03 as the parent molecule was produced, in which surface-exposed residues within 5 positions from each end of each loop were allowed to vary to every other possible amino acid (Fig. 2H). This library was exposed to the IZN24 target at varying concentrations (optimization 3), and bound variants were sequenced using the Ion Torrent next-generation-sequencing technology (Thermo Fisher Scientific). The amino acid changes that were enriched in frequency compared to the starting library were assessed for their impacts on potency. One such beneficial substitution was T60E (numbering based on 10Fn3), located near the invariant DE loop. This change, along with the E97P substitution previously found to enhance adnectin thermal stability (14), was incorporated into 5497_F03 to produce 6200_A08 (Table 3). 6200_A08 exhibited a nearly 4-fold improvement in potency compared to 5497_F03, along with an improvement in the melting temperature of nearly 11°C (Table 2).
6200_A08 exhibits broad-spectrum activity.6200_A08 was tested in a cell-cell fusion assay against a panel of 123 envelope proteins from clinical HIV-1 isolates spanning 11 subtypes (see Table S2 in the supplemental material). Each of the cloned envelope proteins was transiently expressed (14), and the EC50 for the activity of 6200_A08 against each envelope protein was determined. A fold change (FC) in potency was then calculated for each envelope protein, defined as the ratio of the EC50 against that envelope protein to the EC50 against a reference envelope protein derived from the LAI strain of HIV-1 that was assessed in the same experiment. The result of this analysis is shown in Fig. 6. A range of susceptibilities to 6200_A08 was observed, both across and within subtypes, with the majority of envelope proteins tested showing good susceptibility compared to the reference strain, while some envelope proteins showed FCs of >10, which was thought to be a significant difference in this assay. Sequence analysis of the region encompassing the N17 site was conducted for a panel of 27 envelope proteins spanning this range of susceptibilities in an effort to identify genetic determinants of susceptibility (Table 5). Only 2 viruses had changes in the N17 region, while 6 more had sequence changes outside that region. However, no correlation could be discerned between the sequence of this putative adnectin-binding region and susceptibility to inhibition by 6200_A08, which suggests that determinants other than adnectin binding, such as the fusion kinetics of a particular envelope protein, may affect susceptibility. Similar variability in response has been observed in vitro with the peptide fusion inhibitor enfuvirtide (32). Additionally, 6200_A08 was tested against a panel of 16 clinical isolates of varied subtypes and coreceptor specificities in a replicating virus format (Table 6). EC50s ranged from 4 nM to >3.9 μM, indicating a range of susceptibilities to 6200_A08 in this context similar to what was observed in the cell-cell fusion assay.
Activity of 6200_A08 against clinical envelope proteins. A cell-cell fusion assay was used to examine the ability of 6200_A08 to inhibit fusion mediated by a panel of envelope protein clones. The data are represented as the FC compared to a control envelope protein (from LAI virus) run at the same time. Each point represents a different virus envelope protein, and the data are grouped by virus subtype. The horizontal lines represent median values for each subtype.
Activity of 6200_A08 does not correlate with the sequence of the region surrounding the N17 site
Activity of 6200_A08 against clinical HIV isolates
Resistance selection with 6200_A08.NL4-3 virus was used to infect MT-2 cells in the initial presence of a 2× EC50 concentration (∼11 nM) of 6200_A08. Cell cultures were monitored until indicators of virus breakthrough were observed (cell death and/or syncytium formation), at which time a sample of the supernatant was used to infect new MT-2 cells in the presence of double the 6200_A08 concentration from the previous passage. NL4-3 virus was also passaged concurrently without 6200_A08 selection as a control virus. Virus growth was observed to be much slower in the 6200_A08 selection sample, and resistant virus was identified at passage 17 (64 days in culture) at a final dose of 5.1 μM. Examination of the passage 17 virus population revealed it to have greatly reduced (>929-fold) susceptibility to 6200_A08. Population sequencing of the virus stocks identified a single amino acid change (Q577R by gp160 numbering or Q66R by gp41 numbering) compared to the control virus, which lies within the N17 sequence. A recombinant NL4-3 virus containing just the Q577R substitution exhibited a >482-fold loss of susceptibility to 6200_A08, demonstrating that this single mutation in the N17 sequence can engender high-level resistance. Additionally, the Q577R substitution within the context of the IZN24 target construct rendered adnectin binding undetectable by SPR (data not shown). These results confirm that the target of 6200_A08 is indeed a region encompassed within N17.
Reduced fitness of Q577R virus.The replication fitness of virus with the Q577R mutation was assessed by infecting MT-2 cells with a mixture of the Q577R mutant virus and NL4-3 wild-type virus at a ratio of ~9:1 based on viral genomes, with control samples consisting of 100% WT and 100% Q577R virus. Viruses replicated in the absence of inhibitors for 3 to 7 days postinfection (depending on the extent of syncytium formation) and underwent the next round of infection by passaging 1 to 5% of viral supernatants to infect fresh MT-2 cells. Viral supernatants from each passage were collected, and their genomes were amplified using reverse transcription (RT)-PCR and then quantified by real-time PCR. Sequencing of the envelope protein regions flanking Q577R was also performed (Fig. 7). At the start of passage (P0), 92.8% Q577R virus was present in the mixture. After one passage (P1), the content of Q577R in the mixture was reduced to 30.4%. Q577R was further reduced to 3.6% in P3 and to undetectable levels (0%) in P4. This observation indicated that the fitness of the Q577R mutant virus is suboptimal compared to wild-type NL4-3 virus. This same trend could be seen using population sequencing of the amplified envelope protein genes (data not shown).
Reduced fitness of virus with the Q577R mutation. An approximately 9:1 mixture of Q577R and WT viruses was used to infect MT-2 cells and passaged 4 times. At each passage, the gp41 region was amplified by RT-PCR, and the amounts of Q577 and R577 were quantified by real-time PCR.
Combining the anti-gp41 and anti-CD4 adnectins results in synergistic inhibition.As combinations of inhibitors with independent mechanisms of action are currently required for successful treatment of HIV infection, we assessed whether combining the 6200_A08 anti-gp41 adnectin with the previously identified 6940_B01 anti-CD4 adnectin (14) would provide a benefit over either inhibitor by itself. Simple mixtures of the two adnectins in equimolar concentrations were found to have no potency advantage over either inhibitor alone (Table 7), although as expected, their nonoverlapping resistance profiles maintained potency against viruses with resistance mutations to either adnectin, but not to both.
Activity of tandem inhibitor against resistant virus
We next examined whether physically connecting the two adnectins to form a single molecule via a peptide linker conferred any advantage. A panel of linker types and compositions was assessed, including flexible, rigid, charged, and neutral linkers of various lengths. In each case, the anti-CD4 adnectin was placed in the N-terminal position and the anti-gp41 adnectin in the C-terminal position, with a His6 tag at the N terminus (Fig. 8A). These “tandem” adnectins were expressed and purified in the same manner as the original individual adnectins. Size exclusion chromatography was used to eliminate configurations that encouraged aggregation, and the potency of the remaining tandems was assessed against RepRluc virus.
Optimization of linker length between anti-CD4 and anti-gp41. (A) Depiction of the tandem adnectins (amino terminus to carboxy terminus) as follows: anti-CD4 adnectin, linker, anti-gp41 adnectin. (B) Tandem adnectins with different sequence linkers were purified and examined for antiviral potency. The EC50s are plotted compared to the individual adnectins tested at the same time. Averages of duplicate measurements are plotted.
From this screen, charged and relatively rigid linkers based on the ESPEPETPEDE sequence demonstrated the most significant improvement in potency, leading to a >50-fold reduction in the EC50 (Fig. 8B) compared to the individual adnectins. An intermediate linker length (ESPEPETPEDE) was found to be optimal. Thus, while mixing the two adnectins yielded no potency increase, physically connecting them with a peptide linker of specific length and composition resulted in synergistic activity (Table 7). Potency was increased >100-fold, with an EC50 of 37 ± 1 pM.
The antiviral properties of this optimized tandem were further assessed against viruses containing resistance mutations for either the anti-CD4 adnectin (S128N, N197D, T303K, H642Y, and S465P) (14) or the anti-gp41 adnectin (Q577R) or a virus containing both sets of mutations. As expected, the virus that contained both sets of resistance mutations showed greatly reduced susceptibility to the tandem adnectins, as well as to the individual adnectins (Table 7). The potency of the optimized tandem against anti-CD4 adnectin-resistant virus (60 ± 30 pM) was similar to the potency of the tandem against the wild-type virus (37 ± 1 pM) (Table 7). However, against the anti-gp41 adnectin-resistant virus, the potency of the optimized tandem adnectins was reduced by over 100-fold to 6 ± 3 nM, which is equivalent to the potency of just the anti-CD4 adnectin by itself. This result reveals an asymmetry in the synergy conferred by linking the two adnectins into a single molecule. The anti-CD4 adnectin enhances the activity of the anti-gp41 adnectin in a manner independent of the inherent antiviral activity of the anti-CD4 adnectin, but the antiviral activity of the anti-gp41 adnectin is strictly required for the synergistic activity of the tandem adnectins. The hypothesis is that the anti-CD4 adnectin provides its synergistic benefit by anchoring the anti-gp41 adnectin to the target cell surface, thereby greatly enhancing its local concentration at the site of action, in addition to asserting its own inherent antiviral activity. The dependence of the tandem potency on the linker length and composition (Fig. 8) indicates that a particular distance, geometry, or freedom of movement between the two adnectins may be required for optimal synergy. A similar anchoring effect has been observed for bifunctional antibodies composed of modified versions of the anti-CD4 antibody ibalizumab, in combination with a broadly neutralizing antibody, such as PG9 (33) or an enfuvirtide-like peptide (34, 35).
The anti-CD4 adnectin–anti-gp41 adnectin tandem exhibits synergistic potency across clinical envelope proteins.The optimized tandem was also tested against the panel of 123 clinical envelope proteins in a cell-cell fusion assay. As shown in Fig. 9A (see Table S2 in the supplemental material), the synergistic activity of the tandem adnectins extended to all the envelope proteins tested, resulting in 10-fold to >1,000-fold improvements in potency. The geometric mean EC50 for fusion inhibition was 4.4 nM for the anti-CD4 adnectin, 84.1 nM for the anti-gp41 adnectin, and 46.5 pM for the tandem adnectins. An alternate presentation of these data (Fig. 9B) revealed that the tandem is effective for inhibition of fusion of envelope proteins at concentrations well below the EC50 of either adnectin alone for the most susceptible envelope proteins. The tandem adnectins were able to inhibit 90% of envelope proteins with an EC50 of ≤290 pM, but less than 1% of the envelope proteins could be inhibited by either adnectin alone at this low concentration. Furthermore, the high potency of the tandem adnectins was confirmed against a panel of 6 clinical isolates of various subtypes and coreceptor specificities (Table 8), the EC50s for which ranged from 82 pM to 2.4 nM, with a geometric mean EC50 of 236 pM.
(A) Physical linkage of anti-gp41 adnectin, 6200_A08, to anti-CD4 adnectin, 6490_B01, provides picomolar potency against a majority of clinical envelope proteins. The envelope proteins are arranged in order of susceptibility to the anti-gp41 adnectin. (B) The percentage of envelope proteins with an EC50 less than a given concentration is plotted as a function of the inhibitor concentration. The dotted horizontal line indicates coverage of 90% of the panel.
Activities of tandem adnectins against clinical HIV isolates
DISCUSSION
The N17 site of the HIV-1 gp41 protein is an attractive target for therapeutic intervention, as the amino acid sequence of the site is well conserved and molecules of various sizes and compositions that bind to this site can effectively interrupt the conformational changes in gp41 that are required for fusion of the viral and target cell membranes (4). The fact that synthetic peptides can accurately reproduce this trimeric target site outside the context of native gp41 has enabled the use of in vitro selection technologies, such as mRNA display, to aid in the discovery and optimization of novel proteins that bind to this site with high affinity and specificity. The target sequences used are downstream of the binding site for T20 (36). Validation experiments showed that T20 did not bind any of the targets, while the D5 Fab′ did bind, and competition experiments with 2428_G03 correlated with this, as T20 did not compete the binding of the adnectin, while D5 Fab′ did compete (not shown). The adnectin 6200_A08 was identified in this way as a potent inhibitor of membrane fusion mediated by the HIV-1 gp41 protein.
From an initial adnectin library containing trillions of variant sequences (10), one lead adnectin (2428_G03) that exhibited consistent anti-fusion activity (EC50 = 150 nM) was selected. Further optimization of this lead using more stringent selection conditions to drive higher-affinity binding to the N17 sequence of gp41 eventually produced the 6200_A08 adnectin, which displayed greatly enhanced thermal stability and antiviral potency (EC50 = 5.6 ± 2.4 nM). Based on ITC studies conducted with a precursor to 6200_A08 (4058_H08), the stoichiometry of adnectin binding to the IZN21 trimer in solution is 1:1 (one adnectin per trimer). While this solution-based system uses an artificial gp41 target and therefore may not fully represent native gp41 on the virus surface, these results suggest that occupancy of a single N17 site on the gp41 trimer is sufficient to interrupt its capacity to drive membrane fusion.
Perhaps because the target was an artificial construct rather than the full natural protein structure, it was found that optimizing for affinity alone did not produce optimal antiviral activity. From the set of active sequences, BC and FG loops that were associated with higher potency were identified, and after shuffling these loops, new combinations that were significantly more potent than any of the parents were isolated. These results suggest that in a situation where affinity for an artificial target is not directly aligned with potency, generating and screening a wide variety of variants may be required for optimization.
Testing 6200_A08 against a panel of HIV-1 clinical envelope proteins from a variety of subtypes revealed that the majority of the envelope proteins were susceptible to inhibition by the adnectin. The potency of 6200_A08 against susceptible envelope proteins was generally in the nanomolar range, which is similar to or less potent than other inhibitors targeting similar regions in gp41, such as C34, T20, or D5 (17, 36, 37). A sequence analysis of the N17 site and surrounding region of gp41 from the envelope proteins with reduced susceptibility demonstrated no obviously correlated polymorphisms (such as Q577R), which suggests that the mechanism involved may not be related to altered binding of the adnectin to its target sequence. A similar phenomenon has been observed for another fusion inhibitor, enfuvirtide (32). In that instance, mutations outside the enfuvirtide binding site that affect the kinetics of membrane fusion through various means also affect enfuvirtide susceptibility. Conformational changes in gp160 are required to expose the enfuvirtide binding site, resulting in a transient window of access that begins when gp120 engages CD4 and ends with 6-helix-bundle formation. Envelope proteins that progress through this window more rapidly are less susceptible to inhibition, as enfuvirtide has less time in which to find its binding site and engage gp41. Since 6200_A08 binds an adjacent but not overlapping region of gp41, a similar effect of fusion kinetics on inhibitory activity could be expected.
Resistance profiling of this inhibitor produced a single point mutation, Q577R, located within the putative N17 binding site, that was sufficient to abrogate adnectin binding and thereby confer resistance. While this mutation clearly prevented adnectin binding in SPR studies, it may have additional effects on gp41 structure or function that contribute to the resistance phenotype. This mutation has been found to cause resistance to other fusion inhibitors, as well (9, 38–42), some of which bind to regions of gp41 outside the N17 site (43). In thermal denaturation studies, the Q577R mutation was found to increase the stability of the postfusion 6-helix-bundle complex (39–42), which could in turn affect the kinetics of the conformational changes in gp41 that drive fusion, as well as access to the N17 site. The Q577R mutation, however, also appears to cause a reduction in replication fitness (Fig. 7). In confirmation of our observations, others have also found Q577R to reduce the infectivity of various strains by up to 90% (38, 39, 43, 44). Thus, while this mutation confers resistance to the 6200_A08 adnectin, the virus pays a significant fitness price. The Q577R mutation is present at low levels (1.8% of 4,556 independent isolates) in sequences from the LANL HIV Sequence Database (16).
Previous studies of other fusion inhibitors, such as enfuvirtide-like peptides, have shown significant improvements in potency by direct linkage of the fusion inhibitor to an antibody targeted to a cell surface receptor, including CCR5 (45) and CD4 (34, 35). In addition to the advantages conferred by combining two mechanisms of antiviral activity, these bifunctional molecules benefit from an anchoring effect, whereby the CCR5- or CD4-binding component localizes the linked fusion inhibitor to the target cell surface, concentrating it at the site of action. The enhanced local concentration of the fusion inhibitor should in turn enhance the effective on rate for binding to its target site in gp41, which would be expected to counteract the potency-reducing effects of more rapidly fusing envelope proteins described above.
To explore whether the potency and spectrum of the 6200_A08 adnectin could be enhanced in a similar fashion, we connected it to the previously described anti-CD4 adnectin, 6940_B01 (14). Based on previous work in which two adnectins with different specificities were successfully linked in a bispecific tandem format (46), a panel of linker types and lengths was defined and tested for compatibility with monomeric expression and enhancement of potency. A relatively rigid and charged linker of a specific length, ESPEPETPEDE, was found to provide optimal potency. This physical linkage of the anti-CD4 and anti-gp41 adnectins conferred up to a 1,000-fold potency advantage against a broad spectrum of envelope proteins and maintained potency against viruses resistant to either adnectin alone. Importantly, this tandem format maintained subnanomolar potency against clinical envelope proteins that initially showed reduced susceptibility to the 6200_A08 adnectin. The improvement in potency of this dual adnectin is mostly in the low picomolar range, while the breadth of activity is very high, with picomolar potency obtained against 120 of 123 envelope proteins, with the other 3 envelope proteins exhibiting EC50s in the single-digit nanomolar range (Fig. 9; see Table S2 in the supplemental material). This potency compares favorably to those of other bifunctional constructs (34, 35, 45). Thus, the tandem molecule appears to overcome the likely altered fusion kinetics that reduce susceptibility of these envelope proteins to 6200_A08 through the enhanced local concentration or optimized presentation of 6200_A08 afforded by binding to CD4 via the anti-CD4 adnectin, which may provide a higher effective on rate for gp41 target engagement.
We have previously shown the utility of the adnectin scaffold for developing inhibitors of HIV entry that bind to CD4 (14), and in this report, we extend the utility of the scaffold to target other proteins involved in HIV entry, namely, gp41. Through various selection techniques, an adnectin that binds tightly to the N17 sequence of gp41 was isolated and optimized. This adnectin exhibits broad-spectrum, potent inhibitory activity. Moreover, coupling of this adnectin to a previously optimized anti-CD4 adnectin (14) improved potency to <100 pM and provided enhanced coverage of a variety of HIV-1 subtypes and clinical isolates. Thus, the physical connection of two potent inhibitors of HIV-1 entry with distinct mechanisms of action can produce a third inhibitor with greatly enhanced potency and spectrum; novel bispecific molecules of this type may serve as the next generation of potent antiviral agents.
MATERIALS AND METHODS
Peptides and proteins.Peptides containing gp41-derived sequences were synthesized using standard solid-phase chemistry (GenScript) and included C-terminal amidation. Biotinylated peptides were synthesized with an N-terminal biotin-6-aminohexanoic acid moiety, and nonbiotinylated versions included N-terminal acetylation. Peptide mass was confirmed by liquid chromatography-mass spectrometry (LC-MS), and purity (>95%) was confirmed by high-performance liquid chromatography (HPLC). Lyophilized peptides were reconstituted in water at 1 mg/ml prior to use. The D5 antibody Fab′ fragment was generated by in-house expression of full-length antibody based on published sequence information (17), followed by digestion with pepsin and isolation of F(ab)2. Reduction with l-cysteine followed by capping with N-ethyl maleimide produced Fab′.
Adnectin library design and construction.Adnectin libraries based on the 10Fn3 domain of human fibronectin were constructed via PCR using overlapping oligonucleotides synthesized with phosphoramidite trimers encoding defined mixtures of amino acids in specific locations, similar to methods described previously (14, 19–22).
Primary selection of adnectins binding gp41 by mRNA display.The first round of mRNA display selection was carried out essentially as described previously (14, 19, 47, 48) against a target consisting of a biotinylated synthetic helical peptide homotrimer (biotin-IZN17) (15), which contains the N17 region of gp41 fused to an IZ trimerization domain (Table 1). Two nanomoles of library mRNA was translated in 2 ml of rabbit reticulolysate (Ambion Life Technologies, Foster City, CA) and then reverse transcribed to yield 35 pmol of mRNA-protein fusions. Prior to selection, 0.4 ml of streptavidin-coated magnetic beads (Invitrogen, Carlsbad, CA) were washed with buffer comprised of phosphate-buffered saline (PBS) and 0.025% Tween 20 and blocked overnight using a buffer containing PBS, 0.025% Tween 20, 1 mg/ml bovine serum albumin (BSA), and 0.1 mg/ml sheared salmon sperm DNA (Ambion). The library of mRNA-protein fusions was allowed to equilibrate with 100 nM biotin-IZN17 for 45 min. This was followed by capture with streptavidin-coated magnetic beads for 10 min. The beads were washed with PBS containing 0.025% Tween 20, and cDNAs encoding bound adnectins were eluted with 100 mM KOH. The eluted cDNA was then amplified by PCR. The resulting post-round 1 PCR product was then quantitated using quantitative PCR (qPCR).
Rounds 2 to 7 were carried out on a smaller scale. The protocol was similar to that described above except that 150 pmol of library mRNA was translated in 0.2 ml of rabbit reticulolysate. Also, to deplete the library of members that bound nonspecifically to the IZ sequence or to the beads, a negative selection was performed prior to each round by exposing the library of mRNA-protein fusions to streptavidin-coated magnetic beads charged with biotin-IZIZ peptide, which contains only the IZ trimerization domain and no gp41-derived sequence (Table 1), for 30 min to remove any streptavidin or IZ domain binders. This was repeated 3 times. The portion of the library that did not bind to the biotin-IZIZ/streptavidin-coated magnetic beads was then allowed to proceed into the positive selection using biotin-IZN17. The selection conditions were identical to those described above for round 1 except that equilibration with 100 nM target occurred for 30 min during these rounds. Additionally, for rounds 5 to 7, biotin-mutantIZN17 (a version of the biotin-IZN17 target with 3 point mutations identified from a survey of clinical envelope protein sequences) was used as the selection target instead of biotin-IZN17 (Table 1). The post-round 7 PCR product was cloned, sequenced, and expressed as described previously (14). Single adnectins were screened for binding to the N17 targets and for antiviral activity as described below. A single adnectin clone with antiviral activity was identified for further optimization.
Optimization of initial hits.An optimization library using doped oligonucleotides was designed to survey the sequence space of the lead clone and to optimize the binding interaction involving these loops. BC, DE, and FG loop sequences were doped 30% at the nucleotide level (14). This amount of doping at the nucleotide level led to ∼50% doping at the amino acid level. The optimization library was constructed via PCR, as described previously (14). This optimization library was subjected to selection via mRNA display in a manner similar to that described above for the primary selections. The first round of selection was carried out using 20 pmol of DNA as the input (approximately 1.2 × 1013 individual members) in 80 μl of in vitro transcription/translation volume. The resulting mRNA-protein fusions were selected against 100 nM biotin-IZN17, captured, and eluted as described above. In rounds 2 to 4, the selection used variants that extend beyond the N17 site of gp41 (biotin-IZN21 and biotin-IZN24). Additionally, selection stringency was increased through the successive decrease of the target concentration: 100 nM biotin-IZN24 in round 2, 10 nM biotin-IZN21 in round 3, and 1 nM biotin-IZN24 in round 4.
Stringency was further increased in round 5 with an on rate selection. mRNA-protein fusions were incubated with 10 nM biotin-IZN24 and allowed to bind for 30 s at room temperature before 10 μM unbiotinylated IZN24 was added to the mixture. Bound adnectins were captured with streptavidin-coated magnetic beads. These populations were washed with buffer containing PBS and 0.025% Tween 20, eluted from the beads with 100 mM KOH, and measured by qPCR. The post-round 5 PCR product was cloned, sequenced, and expressed as described previously (14). From the sequences that emerged from the optimization, 25 were chosen to enter the final loop-shuffling steps. All 25 sequences had identical DE loops. The vectors containing the 25 sequences were pooled, and the 5′ portions of the genes, containing the BC loop and some fixed sequence, were amplified by PCR. The 3′ portions, containing the FG loop and fixed sequence overlapping the 5′ amplicons, were likewise amplified. The two amplicons were mixed and amplified with outside primers to generate a loop-shuffled library containing all the possible combinations of the BC and FG loops found in the 25 sequences. This library was cloned, sequenced, and expressed as described previously (14). The resulting adnectins were evaluated for favorable biophysical properties and potency. Further polishing was achieved by constructing scanning libraries with NNK nucleotides (N is any base, and K is guanine or thymidine) at selective positions that were (i) on the surface of the WT 10Fn3 domain, (ii) near in space to the amino acids in the loops that were most conserved after selection, and (iii) not a member of the set of amino acids that were diversified in the original or optimization library. The entire scheme for selection and optimization of an adnectin inhibitor is outlined in Fig. 2A to H.
Expression and purification of adnectins.Adnectins were expressed in bacterial systems with attached 6×His tags and purified via standard nickel or cobalt affinity chromatography, as described previously (14).
ELISA for identifying gp41-binding adnectins.Nunc MaxiSorp 384-well plates were coated overnight at 4°C with 20 μl per well of 3 μg/ml neutravidin (Pierce) in PBS. The coated plates were washed 3 times in PBS with Tween 20 (PBST), and then 20 μl of 10 nM biotinylated gp41 peptides (based on the trimeric-bundle concentration) (Table 1) in 5% BSA-PBST was added. The plates were incubated for 1.5 h at room temperature and then washed 3 times in PBST. Individual cloned adnectins were diluted to the desired concentration in 5% BSA-PBST and added to the plate (15 μl/well) in duplicate. After incubation for 1 h at room temperature, the plates were washed 3 times in PBST. Detection of bound adnectins, which contained a 6×His tag, was accomplished by addition of anti-6×His MAb 050-HRP conjugate (R&D Systems; 15 μl/well diluted 1:3,000 in 1% BSA-PBST). After incubation at room temperature for 1 to 2 h, the plates were washed 3 times in PBST, and TMB substrate (Invitrogen) was added (15 μl/well). The reaction was stopped at 7 min or less with 15 μl/well 2 N sulfuric acid. The plates were read (optical density at 450 nm [OD450]) in a SpectraMax plate reader (Molecular Devices).
Cells and viruses.MT-2 cells, HEK 293T cells, CEM-NKR-CCR5-Luc cells, and the proviral DNA clone of NL4-3 were obtained from the NIH AIDS Research and Reference Reagent Program. B6 cells, which contain a long terminal repeat (LTR)-driven luciferase reporter, were generated at the DuPont Pharmaceutical Company (49). RepRluc, a derivative of NL4-3 that contains the Renilla luciferase gene in place of nef, was generated at Bristol-Myers Squibb (28). HeLa C14 cells expressing CD4, CXCR4, and CCR5 and containing an integrated copy of a tetracycline (Tet)-responsive inducible luciferase reporter gene were constructed at Bristol-Myers Squibb (50). Growth conditions for cells and viruses were described previously (14).
Cell-cell fusion inhibition and antiviral activity assays.Assays to assess the activity of adnectins with respect to inhibiting cell-cell fusion and viral replication were conducted as described previously (14).
Resistance selection.Resistance selections were carried out as described previously (14), using MT-2 cells infected with NL4-3 virus in the presence of a 2× EC50 inhibitor concentration at a multiplicity of infection (MOI) of 0.005 to 0.05. Upon viral breakthrough, as evidenced by >10% syncytium formation, the viral supernatant was passaged with a 2-fold increase of the inhibitor concentration. The selection process continued until a potency shift of more than 10-fold was observed in the B6 antiviral assay (14). At this point, viral genomes from the resistant population were amplified by RT-PCR and sequenced. The identified substitution, Q577R, was then placed in the RepRluc and WT NL4-3 viruses for additional molecular analyses. The RepRluc virus previously generated to be resistant to the anti-CD4 adnectin (anti-CD4r) (14) was also used for molecular analyses, as was a recombinant virus containing substitutions that would engender resistance to both adnectins (Dualr).
Determination of kinetics of adnectin binding to gp41 by SPR.Neutravidin (Pierce) was diluted to 10 μg/ml in 10 mM acetate, pH 4.5, and immobilized on a T-series CM5 Biacore chip (GE Healthcare) via a standard amine coupling kit (GE Healthcare) to a level of 6,200 response units (RU). The neutravidin surface was conditioned with three injections of 1 M NaCl, 40 mM NaOH. Biotinylated peptides, IZN17 or IZN24 (Table 1), were diluted in running buffer (HBSP+; GE Healthcare) to 10 nM and allowed to flow over the neutravidin surface until 130 RU had accumulated. Analyte (adnectin or D5 Fab′) diluted in running buffer was then allowed to flow over the captured IZN17 or IZN24 surface at 37°C at various concentrations at a flow rate of 50 μl/min with a contact time of 3 min. Dissociation was measured for 2 to 10 min. A surface consisting of the nonbinding peptide biotin-IZIZ (Table 1) captured on neutravidin was used for reference subtraction, and buffer-only samples were included for background subtraction. The IZN17 or IZN24 surface was regenerated between cycles with two injections of 10 mM glycine, pH 2. A 1:1 Langmuir binding model was fitted to the double-referenced sensorgrams to determine kinetic parameters using Biacore T100 evaluation software, version 2.0.1 (GE Healthcare).
ACKNOWLEDGMENTS
We thank Eric Furfine and Sharon Cload for support of this project and Caryn Picarillo for her help.
D.W., Z.L., S.Z., M.C., and M.K. were employees of Bristol-Myers Squibb at the time of data generation and are currently employees of ViiV Healthcare; all employees own stock/stock options in GlaxoSmithKline, the majority owner of ViiV Healthcare. Y.S., T.M., and J.D. are currently employees of Bristol-Myers Squibb; all employees own stock/stock options in Bristol-Myers Squibb. D.F. was an employee of Bristol-Myers Squibb at the time of data generation.
FOOTNOTES
- Received 10 March 2018.
- Accepted 26 April 2018.
- Accepted manuscript posted online 9 May 2018.
Supplemental material for this article may be found at https://doi.org/10.1128/JVI.00421-18.
- Copyright © 2018 American Society for Microbiology.
