ABSTRACT
Long-acting antiretrovirals could provide a useful alternative to daily oral therapy for HIV-1-infected individuals. Building on a bi-specific molecule with adnectins targeting CD4 and gp41, a potential long-acting biologic, GSK3732394, was developed with three independent and synergistic modes of HIV entry inhibition that potentially could be self-administered as a long-acting subcutaneous injection. Starting with the bi-specific inhibitor, an α-helical peptide inhibitor was optimized as a linked molecule to the anti-gp41 adnectin, with each separate inhibitor exhibiting at least single-digit nanomolar (or lower) potency and a broad spectrum. Combination of the two adnectins and peptide activities into a single molecule was shown to have synergistic advantages in potency, the resistance barrier, and the ability to inhibit HIV-1 infections at low levels of CD4 receptor occupancy, showing that GSK3732394 can work in trans on a CD4+ T cell. Addition of a human serum albumin molecule prolongs the half-life in a human CD4 transgenic mouse, suggesting that it may have potential as a long-acting agent. GSK3732394 was shown to be highly effective in a humanized mouse model of infection. GSK3732394 is currently in clinical trials.
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. Building on a bi-specific inhibitor approach targeting CD4 and gp41, a tri-specific molecule was generated with three distinct antiviral activities. The linkage of these three biologic inhibitors creates synergy that offers a series of advantages to the molecule. The addition of human serum albumin to the tri-specific inhibitor could allow it to function as a long-acting self-administered treatment for patients with HIV infection. This molecule is currently in early clinical trials.
INTRODUCTION
Antiretroviral drug discovery has evolved over the past decade. The availability of safe and effective single-pill regimens, first containing three or more antiretroviral agents and now containing two antiretroviral agents (1, 2), has shifted the discovery paradigm for new agents toward longer-acting molecules that can potentially improve compliance, convenience, and prophylaxis. Thus, a once-monthly regimen of injectable cabotegravir/rilpivirine is currently in phase 3 trials (3–5). Aside from small molecules, larger biologic molecules have many of the properties desired for a long-acting agent. The first long-acting biologic for treatment of HIV, ibalizumab, has been approved for biweekly intravenous (i.v.) administration in highly treatment-experienced (HTE) individuals (6), and PRO-140, a monoclonal antibody (MAb) targeted to the CCR5 coreceptor, remains in clinical trials (7).
The isolation and optimization of ever-improving broadly neutralizing antibodies (bnAbs) to HIV-1 have opened up the possibility of their use for treatment and/or preexposure prophylaxis as long-acting agents. However, even with the most improved bnAbs, breadth of activity remains an issue, and models suggest that complete coverage would require multiple bnAbs to different regions of gp160 (8–12). This has led to the development of bi- and tri-specific molecules, whereby 2 or even 3 different bnAb specificities are combined into a single IgG-like molecule (13–15). This may reduce the number of molecules required for complete coverage of circulating HIV-1, although it does not solve the problem of preexisting resistance to portions of these multi-specific molecules. In addition, targeting gp160 in a multi-specific biologic with a cell membrane-targeting moiety can greatly enhance the potency of an anti-HIV biologic. For instance, attaching an HIV-1 fusion peptide inhibitor to a monoclonal antibody targeting CCR5 (16), attaching a cholesterol moiety to the C terminus of an HIV-1 fusion peptide inhibitor (17), and linking an HIV-1 fusion peptide inhibitor to various places on a neutralizing HIV-1 monoclonal antibody (18) are all means to localize the peptide at the surface of the target cell membrane, and all dramatically increase the potency of the combined molecule compared to those of the separate molecules. Also, bi-specific antibodies consisting of anti-HIV-1 neutralizing antibody fragments targeting gp120 fused to ibalizumab or anti-CCR5 showed synergistic increases in potency compared to those of the individual inhibitors (18, 19), while fusion of a CD4-Ig molecule to a coreceptor-mimetic peptide provides greater potency and breadth and a higher resistance barrier than broadly neutralizing antibodies (20, 21). Similarly, linking an adnectin targeting the CD4 molecule with another adnectin targeting the N17 region of gp41 produced a broad-spectrum inhibitor with enhanced potency (>500-fold) compared to the potencies of the individual adnectin components (22). Thus, localization of anti-HIV-1 entry inhibitors to the target cell surface through a variety of methods can significantly increase their local concentration at the site of action, thereby improving potency. GSK3732394 is a single biologic composed of 3 independent inhibitors of HIV-1 entry. One of the inhibitors is an anti-CD4 adnectin that drives synergistic potency of the other two anti-gp41 inhibitors (22) (an adnectin and a helical peptide inhibitor). Each inhibitor by itself has an extremely wide breadth of activity and should be active against the vast majority of circulating HIV-1. Thus, this single biologic molecule should not have the same issue of coverage as mixtures of bnAbs. However, a downside to any biologic modifier is the potential for an immunogenic reaction that effectively decreases the efficacy of the molecule. Although an adnectin is mainly derived from the 10th type III domain of human fibronectin (23–26), the sequence variations needed to allow specific binding to a target may increase immunogenicity of the molecule. Previously, clinical studies with an adnectin (CT-322) targeting VEGRF-2 did induce an immunologic response in a subset of individuals, although these anti-drug antibodies did not affect CT-322 plasma concentrations or VEGF-A biomarker responses (27). Clinical trials of GSK3732394 that are intended to determine the safety, pharmacokinetics, and immunogenicity profile of the molecule have recently been initiated.
(Some of the data included in this report were presented at the Conference for Retroviruses and Opportunistic Infections [CROI], 22 to 25 February 2016, Boston, MA [28].)
RESULTS
Addition of a peptide fusion inhibitor to the bi-specific inhibitor.Previously, we described the creation and development of a potent HIV-1 entry inhibitor containing two independently generated adnectins targeted to either CD4 or the N17 region of gp41 (22, 29). Adnectins are small proteins based on the 10th type III domain of human fibronectin that can be subjected to in vitro selection to identify sequences with specific properties and can be thought of as similar to the VH portion of an antibody (23–26). In an attempt to further improve the virologic properties of this bi-adnectin inhibitor, a third inhibitory domain was added to the end of the anti-gp41 adnectin. This inhibitor is similar to the known fusion inhibitors developed for HIV-1, consisting of an α-helical peptide that binds at the amino terminus of the heptad repeat 1 of gp41 (30–32), upstream of where the anti-gp41 adnectin binds (22). The following considerations were employed in this inhibitor peptide design: optimal length, optimal positioning along gp41 relative to the anti-gp41 adnectin binding site, broad-spectrum activity, potency, low predicted immunogenic risk, and biophysical behavior (minimal tendency to aggregate) in the context of an adnectin-peptide fusion. For a starting molecule we chose T-2635 (31), a sequence that was demonstrated to have stronger helical content, broader spectrum, and a higher barrier to resistance than enfuvirtide. However, T-2635 was designed to have a gp41 binding site shifted several helical turns to the C terminus from that of enfuvirtide, including a significant fraction of the N17 region. Theoretically, this would clash with the binding site of the anti-gp41 adnectin. Therefore, designs with successive turns removed from the N terminus of the peptide (which bind the C-terminal end of the N17 adnectin binding site within gp41) were generated.
Fusions of these peptides with a nonoptimized member of the anti-gp41 adnectin family and a non-HIV-specific adnectin were produced and assayed for potency. It was believed that this approach would best evaluate the potential for antagonism through binding competition and synergy through potency improvements. An initial study was performed and showed that linkage of the fusion inhibitor peptide can act synergistically when the peptide is linked to an anti-gp41 adnectin. Different-length peptides linked identically to either an inert adnectin or the nonoptimized anti-gp41 adnectin 4773_A08 (22) were examined for inhibitory activity (Fig. 1). Peptides of 30, 32, or 37 amino acids in length were linked to the carboxy termini of the two adnectins with identical linkers. The potencies of the peptides joined to the nonspecific adnectin were inversely correlated to the length, with 50% effective concentrations (EC50s) of >200 nM, 141 nM, and 3.2 nM for the 30-, 32-, and 37-amino-acid peptides. Joining the 30- and 32-amino-acid peptides to the anti-gp41 adnectin produced synergistic potencies that were much stronger than the potency of either of the individual components. Fusions to the longest peptide did not significantly increase the potency, as the EC50 for the combination was 1.1 nM, while that of the peptide itself was 3.2 nM. Joining the peptide with the anti-gp41 adnectin has a large synergistic effect on potency when the inhibitors are relatively weak, but the effect may be less pronounced when at least one of the inhibitors is optimized for stronger binding. Therefore, additional optimization work was carried out with shorter, weaker peptides so that improvements in potency and synergy could be more readily seen.
Effect of joining the peptide inhibitor to the carboxy terminus of the anti-gp41 adnectin. Potencies of individual fusion peptide inhibitors fused to an inactive adnectin were compared to potencies of the same peptides linked to an anti-gp41 adnectin (4773_A08) (22). The names of the proteins are shown above the diagram, and the sequences of each of the anti-fusion peptides are shown below the diagram.
The sequence of the peptide was further optimized by using structural models that identified nine amino acids whose side chains are likely to be solvent accessible when the peptide is bound to its target. Starting with a protein which consists of an inert adnectin fused to a shortened peptide fusion inhibitor (PRD-1022 [see Table S1 in the supplemental material), a small library was generated from oligonucleotides with degenerate positions, such that in each member of the library, one of the nine positions was randomized. The peptide, with the nine randomized positions underlined, is SRIEALIRAAQEQQEKNEAALRELDKWAS. One hundred sixty-three separate sequences were expressed and tested for potency. Most of the positions did not show any improvement upon mutation (data not shown), but the Asp residue (DKWAS) was profoundly sensitive to changes, showing improvements of up to 50-fold when mutated to hydrophobic residues larger than valine (Fig. 2) (33, 34). A tyrosine was chosen in the final construct based upon the high potency and the biophysical properties of the adnectin-peptide molecule. Adding the optimal N-terminal sequence element into this peptide gave the sequence ultimately used in our clinical candidate, GSK3732394: TIAEYAARIEALIRAAQEQQEKNEAALRELYKWAS.
Potency of peptides with single amino acid substitutions of the Asp. The short peptide in the PRD-1022 adnectin-peptide fusion was mutated to replace the Asp (SRIEALIRAAQEQQEKNEAALRELDKWAS) residue with 1 of 15 other amino acids and examined for antiviral activity. Absolute EC50s for each adnectin-peptide proteins are shown.
Antiviral properties of the peptide fusion inhibitor.The 35-amino-acid sequence was made into a standalone peptide (named 203613-24) in order to measure its potency and binding characteristics. Because a glycine-based linker would ultimately be used to connect this peptide to an anti-gp41 adnectin, the 203613-24 synthetic peptide included an additional glycine residue in the N-terminal position to provide a similar context. The antiviral potency of this isolated 36-amino-acid peptide was measured in a multiple-cycle experiment against a luciferase-expressing NL4-3 virus. 203613-24 exhibited subnanomolar inhibition, with an EC50 of 0.40 ± 0.27 nM (Table 1). The molecule also showed no cytotoxicity in cell culture against MT-2 cells, with a 50% cytotoxic concentration (CC50) of >10,000 nM. In order to assess the binding of the isolated peptide component to gp41, we designed a 5-helical bundle protein (PRD-828 [see the supplemental material]) using the sequences from gp41 that are thought to contact the peptide. The single-chain molecule has all three inner helices and two of the outside helices with connecting linkers, leaving one outer slot open for the peptide to bind. Biotinylated PRD-828 was captured onto a neutravidin-coupled surface of a CM5 Biacore T200 SPR chip, and the 203613-24 peptide was flowed over in solution at various concentrations. The peptide exhibited binding to the artificial gp41-like trimer substrate, with a ka (association rate constant) of 2.3 × 106 (1/Ms), a Kd (dissociation rate constant) of 2.5 × 10−4 (1/s), and a KD (equilibrium dissociation constant) of 0.1 nM (Table 1). Experiments were conducted under physiological buffer conditions and temperature (37°C).
Antiviral potency and binding affinity of peptide 203613-24
Resistance to the isolated peptide inhibitor.In order to select viruses resistant to 203613-24 in cell culture, MT-2 cells were infected with NL4-3 virus in the initial presence of a 2× EC50 concentration (∼0.8 nM) of the peptide and passaged in increasing concentrations of inhibitor. Virus with decreased susceptibility to the peptide was identified at passage 11 (33 days in culture) at a final peptide concentration of 0.51 μM. The virus population at passage 11 was examined and found to have an ∼18-fold reduction in susceptibility to 203613-24. Population sequencing of the virus stocks identified a single amino acid change of V549A (V38A when amino acid numbering is initiated at gp41) compared to the control virus. This amino acid is within the proposed binding site for the peptide, thus confirming the target of the inhibitor. Even though the sequences of 203613-24 and enfuvirtide are different, it has been reported that the V549A substitution is selected by enfuvirtide and is a common clinical resistance mutation (35). Given this overlap with a known enfuvirtide resistance mutation, we examined 5 additional known enfuvirtide resistance mutations for their effect on peptide susceptibility. Table 2 shows that the peptide retained good activity against most of these enfuvirtide resistance mutations, exhibiting a reduced susceptibility only to V549A (V38A), and even in this case, enfuvirtide has a fold change (FC) greater than an order of magnitude in response to the V38A mutation. Thus, although the peptide exhibits some cross-resistance to enfuvirtide resistance mutations, it tends to exhibit a more restrictive and potent profile.
Activity of 203613-24 against envelope proteins in the RepRLucNL with enfuvirtide resistance mutations
Creation of a tri-specific inhibitor.The length of the G4S-based linker used to connect the peptide to the carboxy end of the anti-gp41 adnectin was optimized as shown in Table 3. As done previously, a shortened, weaker form of the peptide was used to more readily observe changes in synergistic potency. Adnectin-peptide fusion constructs were made with G4S-based linkers. We hypothesized that linker length may affect the synergy between the adnectin and peptide and that the C-terminal residues of the adnectin could also affect the synergy. Therefore, a series of molecules was made with different carboxy termini on the adnectin (4773_A08) and with different linker sequences. Table 3 shows the sequences of the adnectin C terminus, linker, and peptide used. Based upon these data, the C terminus of the anti-gp41 adnectin in the final molecule was altered to NYRTP and the linker sequence used was GGGGSGGGGSGGGGSGGGG ([G4S]3G4).
Optimization of the linker between the anti-gp41 adnectin and peptidea
The potencies of the individual inhibitors, combinations of two inhibitors, and the tri-specific inhibitor (all are the made from the final optimized adnectins and peptide) are shown in Table 4. A non-HIV-specific adnectin was used to substitute for either the anti-CD4 or anti-gp41 adnectins in some of the constructs in order to retain the molecular geometry and context, while allowing the dissection of the relative contributions of the individual inhibitors to potency. The linker sequences used were also the optimized linkers from the final tri-specific molecule (full sequences of the molecules are shown in Table S3). Addition of an inert adnectin in tandem with an optimized, active adnectin (X_41_ and C_X_, where X is the nonspecific adnectin, C the anti-CD4 adnectin, and 41 is the anti-gp41 adnectin) decreased the potency of the active moiety (∼5.6-fold for the anti-CD4 adnectin and ∼21.6-fold for the anti-gp41 adnectin), while having the two active moieties (C_41) synergistically increased potency to 0.02 ± 0.01 nM (22). Interestingly, addition of the peptide to the bi-specific adnectin molecule dropped the EC50 ∼4-fold, to 0.09 ± 0.01 nM. Good activity was also observed when the nonspecific adnectin was swapped for the anti-gp41 adnectin in the presence of the two other inhibitors (C_X_P, where P is peptide 203613-24; EC50 = 0.21 ± 0.03 nM), but the potency dropped ∼10-fold compared to that of the C_41_P optimized inhibitor.
Synergistic and antagonistic properties of joining the anti-HIV-1 adnectins and peptide
Addition of a PK-enhancing element to create the final GSK3732394 molecule.Although the tri-specific molecule is highly potent, for it to be useful as a long-acting antiviral agent, it needs to have a long intrinsic half-life in vivo. Probably as a consequence of their small size, adnectins by themselves are known to have short half-lives in vivo, likely due to renal clearance (36). Thus, a pharmacokinetic enhancer (PKE) was required to improve the intrinsic in vivo half-lives of these molecules. After examining the biophysical, antiviral, and PK effects of inserting several different PKE elements at different sites in the molecule, a human serum albumin (HSA) molecule was added to the amino terminus of the anti-CD4 adnectin via a 25-amino-acid linker. The resulting molecule is GSK3732394 (formerly BMS-986197) (Table 4), whose sequence is shown in Fig. 3.
Amino acid sequence of GSK3732394/BMS-986197. The human serum albumin component is highlighted in red, the anti-CD4 and anti-gp41 adnectins are highlighted in green and blue, respectively, and the peptide is highlighted in purple. The linker sequences are not highlighted. In addition, the human serum albumin component contains a C34A mutation (yellow) to remove the only free sulfhydryl group in the molecule.
The effect of adding the human serum albumin on antiviral activity was addressed using two different molecular constructs. One molecule (C_41_P) is the exact match to GSK3732394 except that it is missing the HSA molecule and linker sequence connecting it to the anti-CD4 adnectin. It exhibited an EC50 of 0.09 ± 0.01 nM. The final tri-specific molecule, GSK3732394, exhibited an EC50 of 0.27 ± 0.17 nM (Table 4). Thus, the addition of HSA to the amino terminus of a 3-component molecule maintains good potency but does decrease it ∼3-fold in the context of C_41_P. In addition, the cytotoxicity of the molecule was examined using an XTT [2,3-bis-(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide] method. There was no cytotoxicity observed up to the highest concentration of GSK3732394 tested (>2.9 μM). Given that the assay measures the metabolism of XTT by mitochondrial enzymes, this suggests that at the concentrations tested, GSK3732394 is neither cytotoxic nor cytostatic. Finally, serum binding effects on the potency of GSK3732394 were examined with the addition of 40% human serum. The EC50 and EC90 in the presence of human serum were within 2-fold their values without human serum (EC50 fold change, 1.14 ± 0.64; EC90 fold change, 1.29 ± 0.28). Thus, the presence of human serum does not have a significant effect on the potency of GSK3732394.
Binding affinities of inhibitors in the context of GSK3732394 compared to isolated inhibitors.In order to examine whether the binding of GSK3732394 to its targets differed from that of its individual components, the affinity of GSK3732394 was measured by surface plasmon resonance (SPR) against the 3 targets used for analysis against the individual components (22, 29). Thus, binding of GSK3732394 was measured against soluble human CD4 protein, the N17-containing peptide trimer IZN24 (22), and the 5-helical bundle reagent containing the target for the fusion peptide inhibitor (PRD-828). All experiments were conducted under physiological buffer conditions and temperature (37°C). The results are shown in Table 5. The KD of binding to CD4 was 27-fold weaker with GSK3732394 than with 6940_B01 (the anti-CD4 adnectin), due primarily to a lower on-rate. This result correlates with data showing that addition of HSA appears to reduce potency (Table 4). However, CD4 binding could also be affected by the linkage of the anti-gp41adnectin and/or the fusion peptide inhibitor.
Binding affinity of GSK3732394 to inhibitor targets and comparison to the individual inhibitors
The KD for binding of GSK3732394 to IZN24 was 4-fold weaker than the binding of the isolated anti-gp41adnectin (6200_A08) to IZN24. It is possible that steric hindrance resulting from attaching the HSA and anti-CD4 adnectin to the anti-gp41 adnectin may adversely affect interaction with gp41. Similarly, the binding affinity of GSK3732394 for PRD-828 was 3- to -4-fold weaker than the affinity of the isolated peptide component for the same target. Thus, linking the components together into a single molecule does result in decreased binding to their specific targets. However, that is compensated for by the synergies associated with the linkages, which result in increased potencies.
High potency of the GSK3732394 at low receptor occupancies.From the antiviral potency and the SPR affinity data (Tables 4 and 5), it is clear that fusing the anti-CD4 adnectin (6940_B01) to HSA and to the anti-gp41 adnectin (6200_A08) plus the peptide in the context of GSK3732394 has a detrimental impact on the binding of 6940_B01 to CD4, manifested primarily as a lower on-rate. To study this effect further, the isolated anti-CD4 adnectin (6940_B01) and GSK3732394 were assessed for the ability to compete with fluorescently labeled 4945_G06 for binding to CD4 on the surface of MT-2 cells. 4945_G06 is a progenitor of 6940_B01 that differs slightly from it, but the differences do not affect the potency of the molecule or its ability to compete with 6940_B01 for binding to CD4 (29). MT-2 cells are used for antiviral EC50 determinations, so the binding affinity can be directly compared with antiviral potency. A dose-response curve for binding to human MT-2 cells was generated with the 6940_B01 molecule and compared with the dose-response curve for antiviral potency of this molecule (Fig. 4). The EC50 for binding to MT-2 cells was 7.0 nM; the EC50 for antiviral activity in MT-2 cells was similar (4.9 nM). More importantly, the dose-response curves for the two activities with the isolated anti-CD4 adnectin were similar and almost superimposable. This indicates that antiviral potency of the individual adnectin is directly related to binding (in a 1:1 fashion) and suggests that saturation of binding to CD4 with the anti-CD4 adnectin as an individual inhibitor must be accomplished in order to obtain complete inhibition of infection.
Comparison of binding affinity of GSK3732394 to MT-2 cells with antiviral activity observed against RepRLucNL virus. The concentration of GSK3732394 is platted against maximal activity (binding to CD4 or antiviral activity). The antiviral activity of the anti-CD4 adnectin alone superimposes with its cell binding activity, while the antiviral activity dose response of GSK3732394 is much stronger than that of its cell binding activity.
When GSK3732394 was used to generate a dose-response curve for binding to MT-2 cells, the binding to cells was weaker than for 6940_B01. The binding was 100-fold weaker, with an EC50 of 200 nM. However, the dose-response curve of antiviral activity of GSK3732394 in MT-2 cells is ∼4 log10 stronger than for CD4 binding and >2 log10 better than the antiviral activity of 6940_B01 in the MT-2 cells. This suggests that the antiviral activity of GSK3732394 is potent at relatively low receptor occupancy (RO). Further experiments at lower concentrations showed that at an EC50 of GSK3732394 of 0.27 nM in cell culture, only ∼0.2% of CD4 molecules on MT-2 cells are bound to GSK3732394, while at an EC90 of 2 nM, ∼1.5% of CD4 receptors are bound to GSK3732394 (data not shown). Thus, potent antiviral activity is observed at relatively low receptor occupancy of the inhibitor on CD4 on the surface of cells. Similar values for cell binding were generated for both the individual adnectin and GSK3732394 using human peripheral blood mononuclear cells (PBMCs) (data not shown).
Selection of GSK3732394-resistant virus in cell culture.In order to select viruses resistant to GSK3732394 in cell culture, MT-2 cells were infected with NL4-3 virus in the initial presence of a 2× EC50 concentration (∼0.5 nM) of GSK3732394. The drug concentration was progressively increased until it reached 300 nM, and then it was kept constant at that level through multiple passages. NL4-3 virus was also passaged concurrently without GSK3732394 selection as a control. Virus growth was observed to be slower in the GSK3732394 selection sample, and 37 passages (175 days in culture) were required for the virus to grow well enough to warrant harvesting. At that time, the virus population was examined and found to have an 18-fold-reduced susceptibility to GSK3732394. Population sequencing of the virus stocks identified 7 amino acid changes in gp160 compared to the control virus (Fig. 5). Five of these changes were in the gp120 region, with two (T138I and N301K) destroying potential N-linked glycosylation sites (PNGS). The N301K change was also observed during selection with 6940_B01 (29). There is an additional PNGS located between amino acids 396 and 401; the F396S and S401T mutations may have affected the glycosylation occupancy of this site as well. Previous resistance selection using 6940_B01 showed that resistance mapped to the loss of glycosylation sites in gp120, similar to that observed with ibalizumab (37). In addition to the changes in gp120, a Q577R substitution was observed in the N17 region. This change was selected by the anti-gp41 adnectin alone and should render the virus resistant to the individual 6200_A08 component (22). Finally, an L544S substitution was observed in the proposed region targeted by the peptide inhibitor. No other changes were observed in gp160. Recombinant virus containing a gp160 gene with all 7 of these substitutions exhibited a 60-fold loss in susceptibility to GSK3732394. The virus was growth impaired, similar to the case with the earlier Q577R virus alone (22).
Selection of GSK3732394-resistant virus. The table at the top lists all the changes observed during selection by passage 37 and the fold change compared to wild-type NL4-3 virus. Recombinant virus represents a pure RepRLucNL virus containing all the changes shown. For the selection, population sequencing and susceptibility analysis of virus were performed every 3 passages (p) and are graphed as fold change (FC) versus wild-type virus (WT). GSK3732394 concentrations are shown on the right. Identified mutations are listed at specific passages in the boxes, with estimated frequencies. If no frequency is listed, mutation was fixed at ∼100%.
Virus populations from approximately every third passage from the selection were collected and viral RNA was purified from a portion of the viral supernatants. The genomes were then population sequenced, while the remaining supernatants were examined for susceptibility to GSK3732394. As can be seen from Fig. 5, the N301K mutation occurred first at ∼30% in passage 9, and it was fixed at ∼100% by passage 12. At this passage, the Q577R mutation was first observed in a small percentage of genes (∼20%) and was fixed at ∼100% by passage 24. Also, at passage 24, all of the other present gp120 substitutions were fixed, but the FC associated with this virus was still low. The appearance of the L544S substitution (∼50%) at passage 33, which was fixed by passage 37, corresponded to the first significant FC of ∼18. Thus, a larger FC was observed only when resistance to both anti-gp41 inhibitors emerged.
GSK3732394 retains activity against virus resistant to components of the GSK3732394 molecule.A potential advantage of joining the individual inhibitor components into a single molecule may be an effect on the resistance barrier compared to that with the individual components. This was examined using recombinant viruses that contain mutations inducing resistance to one or more of the individual inhibitors. Recombinant envelope proteins resistant to each of the individual adnectins have been described previously and were selected as described above (22, 29), and additional recombinant viruses with mutations for resistance to two components (against either the anti-CD4 adnectin plus anti-gp41 adnectin [anti-CD4r + anti-gp41r viruses], anti-CD4 adnectin plus peptide [anti-CD4r + peptider viruses], or anti-gp41 adnectin plus peptide [anti-gp41r + peptider viruses]) were constructed and tested. The results of these in vitro studies are shown in Table 6. As expected, the individual components exhibited higher fold changes against the viruses containing their selected substitutions but were fully active against recombinant viruses containing the substitutions inducing resistance to the other components. Importantly though, the full-length GSK3732394 did not exhibit a noteworthy FC against any of the 3 viruses resistant solely to one of the components. The only viruses where a large FC was observed included both anti-gp41r + peptider substitutions. Interestingly, in separate experiments, the tri-specific molecule (C_41_P [Table 4]) that is missing the human serum albumin molecule and linker was examined against additional recombinant viruses (Table 6) that contained mutations for resistance to the anti-CD4 adnectin and either the anti-gp41 adnectin or the peptide. As with GSK3732394, a high fold change was observed against virus with resistance to the two gp41 targets (FC = 125), but when examined against either the anti-CD4r + peptider or anti-CD4r + anti-gp41r viruses, full susceptibility was observed (FCs of 0.3 and 0.4, respectively). This suggests that the enhanced potency in the molecule is driven mainly by the activity of the gp41 inhibitors, presumably as a result of targeting to the cell membrane through CD4 binding. This demonstrates that joining the components in a single molecule could overcome resistance to any one component, as well as to some of the dually resistant combinations, which suggests that another property of joining the inhibitors into one molecule may be to raise the resistance barrier of the GSK3732394 molecule compared to those of the individual components.
Activities of GSK3732394, individual inhibitors, and C_41_P against RepRLucNL viruses with selected mutations that encode resistance to one or multiple components
Breadth of antiviral activity.Previously, a cohort of 124 functional envelope gene populations was used to examine the spectrum of activity for the individual adnectin inhibitors in a cell-cell fusion assay (22, 29). This same cohort was used to analyze the breadth of activity of the GSK3732394 biologic molecule. This assay usually shows greater run-to-run variability than an infectious virus assay, so the results are normalized against an LAI envelope clone and expressed as fold change.
The 124 envelope gene populations were derived mainly from clinical samples and span 11 different HIV-1 subtypes (22). Figure 6 shows the FC of the cohort compared to the LAI envelope. GSK3732394 was active against 100% of these envelopes. Of the 124 gp160 populations, only 2 envelope proteins exhibited an FC of >10 in this assay. All 64 subtype B envelope proteins and 26 subtype C envelope proteins exhibited little to no fold change in the assay. The only two envelope proteins to exhibit an FC of >10 were both from subtype D (FC, 11.5 and 17.3), while two other subtype D envelope proteins showed no significant FC. When the population of envelope genes from one of these subtype D viruses was cloned into recombinant virus, it exhibited an FC of 13.3. When the envelope gene from this subtype D virus was sequenced, Q577K and L544V substitutions were observed, which could impact the anti-gp41 adnectin and peptide inhibition and account for the FC seen in this virus. Cloning of the other subtype D envelope did not produce an infectious virus, and sequencing of the gp160 gene did not identify mutations that were likely to impact the anti-gp41 inhibitors. When envelope gene populations exhibiting the next highest FCs, 6 and 5.8, were cloned into recombinant viruses and examined against GSK3732394, the FCs were 1.7 and 0.7, respectively. This suggests that all other envelopes are highly susceptible to GSK3732394. Thus, this cell-cell fusion data demonstrate that the GSK3732394 molecule is highly active against the vast majority of virus envelope proteins, including all 90 of the envelope proteins examined from the major subtypes B and C.
Observed fold change of cloned envelope populations against GSK3732394 in a cell-cell fusion assay. Envelope populations were divided into subtypes as shown. Subtypes with one or a few envelopes are grouped into “Others,” and the numbers of isolates and color codes are shown to the right. Significant changes in susceptibility were estimated to be FCs of 10 or above, as described in the text.
To confirm the breadth of activity of GSK3732394, the molecule was further examined against a series of primary clinical isolates. A panel of 19 HIV-1 clinical isolates from various group M subtypes (including subtypes A, B, C, D, F, G, and CRF01_AE) and 1 virus each from group N and group O were evaluated in dose-response experiments. GSK3732394 was active against all the clinical isolates, with EC50s ranging from 0.10 nM to 2.1 nM (Table 7), confirming the wide spectrum of activity of GSK3732394.
Activity of GSK3732394 against clinical isolates
Activity of GSK3732394 in a humanized mouse model of infection.In order to examine the potential of GSK3732394 to inhibit virus in vivo, a human immune system was reconstituted in NOG mice with hematopoietic stem cells isolated from human cord blood (38). After 14 weeks of engraftment, mice were infected with the YU2 strain (R5 tropic) via IP injection. Table 8 shows the activity of the various individual components and GSK3732394 against YU2 virus in cell culture. All molecules were active against this virus, although the potency was slightly decreased compared to that of NL4-3 (Table 4). At day 37 postinfection, mice were analyzed for viral load and distributed into 5 groups of 8 mice each. Viral loads of most mice at the time of GSK3732394 administration were in the range of 105 to 106 copies/ml. One group was treated subcutaneously (s.c.) with a vehicle only, while one group was given a regimen of raltegravir (RAL) plus tenofovir disoproxil fumarate (TDF) plus lamivudine (3TC) daily incorporated in the food pellet (ART group). Based on an average consumption of 4 g of food per day, each animal received daily 2.4 mg of TDF, 2.35 mg of 3TC, and 19.2 mg of RAL. The other 3 groups were treated with GSK3732394, injected s.c. every 3 days at a dose of 4, 12.5, or 32 mg/kg of body weight. Every 9 days (every third dose) prior to dosing, blood was taken for viral load and other analyses. The study lasted approximately 2 months (63 days), after which treatment was stopped. At this point, most animals were exsanguinated, with the exception of three animals each in the ART-treated group and the group treated with 32 mg/kg of GSK3732394. These animals were left untreated for an additional 21 days (with blood taken 9 days after treatment termination and at end of study) to probe virus rebound. Treatment with GSK3732394 was generally safe and well tolerated. Although some mice died during the 2-month course of the study, more died in the vehicle and ART groups (3 and 2 mice, respectively) than in the GSK3732394 treatment groups (1 each of the 4- and 32-mg/kg treatment groups).
Activities of inhibitors against YU2 virus
The receptor occupancy of GSK3732394 on CD4 was measured every 9 days (Fig. 7A). Variability was observed within each cohort, but a clear dose-dependent increase in RO was observed, with the RO among animals in each cohort falling within an ∼20% range. The 4-mg/kg dose produced the lowest RO, between a few percent and ∼25% at each time point, while the 32-mg/kg dose exhibited ROs between ∼40 and 60% at the various time points. The 12.5-mg/kg dose was intermediate with respect to RO (∼20 to 40%).
(A) CD4 receptor occupancy with GSK3732394 in YU2-infected humanized mice; (B) GSK3732394 concentrations at trough on the respective days of the study. The doses are indicated by line type. Days indicate time after first dose (not including the 37 days of YU2 infection prior to first dose).
As expected, the plasma concentrations of GSK3732394 correlated with the RO values (Fig. 7B). A dose-dependent increase in plasma concentrations was observed and remained consistent throughout the study. There was greater variability in the plasma concentrations at the highest dose than at the lower doses, especially the 4-mg/kg dose.
A summary of the average viral loads for each group at specific time points is shown in Fig. 8. Interestingly, the vehicle cohort saw approximately a 1-log increase in viral titers over the 2 months of the study, while all the inhibitor-treated samples exhibited viral load declines. The data show that at all doses tested, GSK3732394 treatment resulted in significant reductions in viral loads. This even includes the lowest-dose cohort (4 mg/kg), which averaged relatively low CD4 receptor occupancy at trough (Fig. 7) throughout the study. In the GSK3732394-treated cohorts, a dose-dependent decline in viral load was observed, with the highest dose seeing almost a 4-log10 drop over the course of the study compared to titers at the start of the experiment. However, all three GSK3732394-treated dose cohorts tended to show an increase in viral load beginning from days 36 to 45 until the end of the study. Thus, by day 63, the average viral load decline was over 3 log10 for the 32-mg/kg group and slightly lower and higher than 1 log10 for the 4- and 12.5-mg/kg groups, respectively. In addition, there were mice in certain groups whose viral loads were below quantitative levels at certain times. These are indicated by the numbers shown in Fig. 8. Thus, 6/8 animals had undetectable viral titers by day 36 in the 32-mg/kg group, while by day 63 it was 4/7 animals, while 2/8 animals in the 12.5 mg/kg group possessed viral loads below detectable levels at day 63. By that time, all animals in the ART group (6/6) had undetectable viral loads.
Efficacy of GSK3732394 in a mouse model of infection. Lines represent viral titers in dose cohorts, the identities of which are shown on the right. The numbers in fractions indicate the number of samples with undetectable viral load at this time point. Since the lower limit of quantitation of the quantitative PCR assay was 100 copies/ml, for graphing purposes, undetectable samples were arbitrarily given a viral load of 100 copies/ml.
At day 63, the study was terminated for most mice and plasma was collected from all mice but 3 each in the ART-treated and 32-mg/kg cohorts. Plasma samples were taken from all mice in the 4-mg/kg dose cohort and 7 mice in the 12.5-mg/kg dose cohort that exhibited measurable viral titers at day 63, along with 4 mice from the highest (32-mg/kg) dose cohort with viral titers. The gp160 genes were amplified from plasma virus by reverse transcription-PCR (RT-PCR), and the gene products were population sequenced. Amplification was successful in most cases, except for one animal in each of the dose groups. In all samples at day 63 except one, only one amino acid substitution was observed within the gp160 gene. This was a Q577R change within the N17 region of gp41. The only sample without this change came from the one mouse in the 32-mg/kg cohort whose titer at day 63 was below the level of quantitation but had an amplifiable gp160. This mouse contained the wild-type Q577 and no other changes. The Q577R mutation was previously shown to elicit resistance to the anti-gp41 adnectin that is part of GSK3732394 (22).
Thus, in a humanized mouse model of infection, GSK3732394 was able to significantly reduce virus titers, even at relatively low CD4 receptor occupancies. Over time, breakthrough of viruses occurred, and the durability of response was dose dependent over the 2-month-long duration of the experiment. Breakthrough viruses contained a selected mutation that engendered resistance to the anti-gp41 adnectin portion of GSK3732394.
Three mice without measurable viral titers in the highest-dose group (32 mg/kg of GSK3732394) and 3 mice in the ART treatment group were kept in the study without drug for an additional 21 days after day 63, with plasma taken 9 days after treatment termination (day 72) and at the end of the study. Viral titers were measured at both time points, although there was only enough plasma available for analysis at the day 72 time point. In the ART treatment group, one sample (either day 72 or day 84) of each of the animals showed a measurable viral titer, while all three animals in the 32-mg/kg treatment group rebounded to titers similar to that observed at start of GSK3732394 treatment. Plasma samples of these 3 mice from day 72 were amplified, and the gp160 gene was population sequenced. These samples showed that two mice contained virus with wild-type Q577, while one mouse contained virus with a Q577Q/R mixture.
DISCUSSION
Previously, we described a bispecific inhibitor comprised of an anti-CD4 adnectin (29) linked to an anti-gp41 adnectin targeting the N17 region of gp41 (22). The linkage was optimized so while each individual inhibitor possessed single-digit nanomolar EC50 potency, the connected molecule was >100-fold more potent than the mixture of the two (22). However, studies suggested that the anti-gp41 adnectin suffered from a relatively low resistance barrier. To address that deficiency and potentially make the molecule effective enough that it could become a long-acting antiretroviral agent, an anti-gp41 antiviral peptide was linked downstream of the anti-gp41 adnectin targeting a region in gp41 similar to enfuvirtide. Addition of the peptide did decrease the potency of the molecule (from 0.02 to 0.09 nM) but improved the resistance barrier of the combined molecule by making the molecule fully active against viruses resistant to one of the components.
Once a tri-specific inhibitor was constructed, the next stepping stone was to improve the half-life of the molecule to enable it to be administered less frequently than once daily. Adnectins by themselves tend to have short half-lives in humans (36, 39), so a pharmacokinetic-enhancing (PKE) molecule needed to be included in the final molecule. After analyzing multiple formats, a human serum albumin molecule placed at the amino terminus was chosen as the PKE element. The final molecule, GSK3732394, exhibited an EC50 of 0.27 ± 0.17 nM, which is ∼3-fold weaker than the molecule without the PKE element, perhaps due to weaker affinity for CD4 when the HSA molecule is present. GSK3732394 was active against all viruses and viral envelope proteins in the experimental panel. There was one subtype D virus envelope protein that showed an FC of ∼17.3 in potency in a cell-cell fusion assay compared to the control. Upon sequence analysis, this envelope protein was found to possess mutations (Q557K and L544V) that could result in decreased susceptibility to both anti-gp41 components, which could explain the result. Table 6 shows that viruses with resistance mutations targeted to both gp41 inhibitors do show an enhanced fold change. These mutations in envelope are relatively rare within the Los Alamos National Laboratory (LANL) database, with the Q577R mutation found in 1.9% of 5,454 sequences (2017 release), while Q577K was found in only 8 isolates (0.15%). The L544V mutation was found in 2.8% of the isolates in the LANL database.
The effect of connecting the 3 independent entry inhibitors together has several advantages over a mixture of individual inhibitors. An obvious advantage is the need to progress only one clinical candidate through the development pipeline rather than multiple separate molecules, although the added complexity of the macromolecule may make development more problematic. In addition, as observed when the two adnectins are joined or when the two anti-gp41 inhibitors are joined with the correct linkage, improved potency is achieved through multiple synergies. The improvement in potencies is the probable result of an avidity effect of placing the inhibitors near their site of action, driven by the binding of the anti-CD4 adnectin to its target. The peak synergy probably results from the greatly increased concentration of the two gp41 inhibitor components at the cell surface compared to the concentration if the gp41 inhibitors were floating in plasma. The higher local concentration of inhibitors should greatly increase their binding on-rate to gp41, and hence their potency, which is what was seen experimentally even when only the anti-gp41 adnectin was linked to an anti-CD4 adnectin (Table 4) (22). Studies on linker length connecting the two adnectins suggest that an optimal distance between the two inhibitors is needed, and if that distance is increased, the potency begins to decrease (22). Thus, GSK3732394 contains linker lengths and compositions that are optimal for the gp41 inhibitors to engage the 3-helix trimer and block it from converting to the 6-helix state. This is similar to the potency increase observed when certain bnAbs are linked in a heterologous MAb to ibalizumab (19).
Another advantage to linking the three inhibitors can be observed when the molecule is tested against viruses containing mutations known to induce resistance to the various components. Viruses containing mutations for resistance to one, two, or all three inhibitor components were recombinantly generated and examined against the individual inhibitors or GSK3732394 (Table 6). Against the individual inhibitors, significant fold changes compared to wild-type virus were observed, as expected, against the homologous virus-inhibitor pair. However, viruses containing mutations for resistance to only one of the inhibitors did not exhibit a significant fold change (FC, 1.1 to 2.1) against GSK3732394. This compares to an FC of >660 against the anti-gp41 adnectin or of 7.0 against the peptide inhibitor. A significant fold change against a recombinant virus was observed only if it contained mutations for resistance to all 3 component inhibitors (FC, 98) or to the two anti-gp41inhibitors (FC, 89) in the protein. This suggests that linking the inhibitors into a single molecule should improve the resistance barrier compared to that with the individual components, since a virus would need resistance to multiple inhibitors to see a phenotypic change. A lack of a significant fold change was also observed in recombinant viruses with resistance to the anti-CD4 adnectin and either of the two anti-gp41 inhibitors (Table 6). Resistance to the anti-CD4 adnectin maps to loss of potential N-linked glycosylation sites in gp120 and does not affect the binding of the adnectin to CD4 (29). Thus, even in these two doubly resistant viruses, GSK3732394 was bound to CD4, although not producing an optimal antiviral effect. Therefore, this strongly suggests that the high potency observed with GSK3732394 is driven by the anti-gp41 inhibitors, while the synergy is driven by binding of the molecule to CD4 through the action of the anti-CD4 adnectin.
This enhanced resistance barrier is further exemplified by the length of time it took to select for a fully resistant virus in cell culture (Fig. 5). It took approximately 6 months and 33 cell culture passages for virus with a significant fold change to be selected against GSK3732394. Interestingly, initial changes were observed as early as passages 9 to 12, where one PNGS site was lost and the key resistance mutation against the anti-gp41 adnectin (Q577R) appeared, although the fold change was still low. Loss of PNGS is the mechanism of development of resistance to the anti-CD4 adnectin, but it took the loss of multiple sites to induce resistance to the individual adnectin (29). By passage 24, the Q577R mutation was fixed as well as additional mutations at 396/401, these mutations potentially affecting the occupancy of another PNGS between those amino acids. An additional mutation of T138I was fixed at this time point, and a mutation at T63I was observed at a frequency of ∼30%. The T63I change does not affect a PNGS, but the T138I mutation does delete another PNGS in the envelope (perhaps reducing PGNS by 3 or 4), which could affect the susceptibility to the anti-CD4 adnectin. Interestingly, this virus still retained a relatively low fold change. This is not surprising, as even if it has reduced susceptibility to both adnectins, data from Table 6 show that it should still be susceptible to GSK3732394. Peptide resistance due to the L544S change was borne out in the next two sets of sequenced passages (33 and 37), where the increase in L544S to 50% and then 100% resulted in a large jump to 18.7- and 19.7-fold changes. At this point, the virus had selected mutations that reduced susceptibility to all three inhibitors and thus induced a significant, but not overly high, fold change. This could suggest that the synergistic potential of the combined molecule allows it to retain a fair amount of activity even when resistance is developed against all individual components.
Perhaps the most advantageous property associated with GSK3732394 is its ability to produce high potency at low levels of CD4 receptor occupancy (Fig. 4). At the EC50 of GSK3732394, only ∼0.2% of CD4 on the cell surface is bound, while at the EC90, only ∼1.5% of CD4 is bound by inhibitor. How, then, is inhibitor bound to such a small percentage of CD4 protein able to inhibit virus at an EC90 level? The increased local concentration of the anti-gp41 inhibitors clearly helps, but the very high percentage (≫95%) of free CD4 molecules on the cell still provides unencumbered target for virus to bind to, while any virus bound to a CD4 protein also bound by GSK3732394 will be inhibited by the anti-CD4 adnectin. The most logical explanation is that GSK3732394 can inhibit in trans by having the anti-gp41 inhibitors reach out and inhibit fusion of virus bound to a different CD4 molecule on the cell. Based upon the antiviral and biophysical data obtained with GSK3732394, the individual components and the partial progenitor constructs, a structural model to explain the high degree of synergy can be proposed. The model takes into account what is known about the stoichiometry of gp160 and the fusion process. For instance, electron microscopy studies show that each HIV-1 virion contains between 7 and 14 trimer spikes (40, 41). In addition, multiple trimer spikes need to undergo a series of conformational changes to form the fusion pore, which initiates virus-cell fusion (41, 42). In the model, GSK3732394, when attached to CD4 via the anti-CD4 adnectin domain, becomes anchored on the surface of the CD4+ T cell at an optimal distance from the cell surface to interact with the long-lived 3-helix gp41 trimer spike bound to a separate CD4 molecule (Fig. 9).
Model of inhibition by GSK3732394. The CD4 molecule (A) is bound to the anti-CD4 component of GSK3732394 (B and C). (D) A gp41 molecule in the semistable 3-helix conformation, as part of a gp160 trimer bound to a different CD4 within the fusion pore. The GSK3732394-CD4 complex is in the correct orientation and at an optimal distance for the anti-gp41 adnectin and peptide inhibitors to allow them to bind to the gp41 (E), thus disarming 2 distinct epitopes on gp41. Color coding for the 3-helix HIV-1 gp41: teal, N-terminal fusion peptide domain; purple, helical domain binding to peptide inhibitor; green, N-terminal N17 region; and dark blue, C-terminal domain.
The overall effect of the binding of GSK3732394 has a number of aspects. First, it reduces the number of CD4 molecules on a T cell available for functional attachment of the virus and formation of the fusion pore. For instance, if one has ∼50% receptor occupancy at trough, this suggests that for most of the dosing period, more than half of the molecules on the cell are engaged with GSK3732394, which may make it difficult to form a functional fusion pore with multiple proteins if the anti-gp41 components work in trans. Second, the higher local concentration of inhibitors should also greatly increase their effective binding on-rate to gp41 and, hence, their potency. Studies on linker length connecting the two adnectins suggest that an optimal distance between the two inhibitors is needed, and if that distance is increased, the potency begins to decrease (22). Thus, GSK3732394 contains a linker length that is optimal for the gp41 inhibitors to engage the 3-helix trimer and block it from converting to the 6-helix state, and since the gp41 inhibitors probably work in trans, the mobility of the drug-bound CD4 molecules, and their ability to deliver the gp41 inhibitors to the sites of the fusion intermediates during the brief window in which they are accessible, is likely a key to the mechanism of action. The overall result is that inhibition by the complete GSK3732394 is much more potent than the sum of its parts.
In order to probe its ability to inhibit virus infection in vivo as a long-acting agent, GSK3732394 was examined in a humanized mouse model of infection. Humanized mice were infected with virus for 37 days and then treated every 3 days subcutaneously with either 4, 12.5, or 32 mg/kg of GSK3732394 for 63 days. The animals showed a dose-dependent increase in CD4 receptor occupancy and a dose-dependent decrease in viral load, with 4/7 animals at the 32-mg/kg dose exhibiting viral titers below the level of detection by day 63 of dosing. GSK3732394 was effective at all 3 doses in a dose-dependent fashion, with ≥1-log10 decreases in viral titers after 9 days in the 12.5- and 32-mg/kg dose groups, and with the 4-mg/kg dose group showing viral titers decreased by 1 log10 by day 27. Thus, GSK3732394 is effective as an antiviral agent in vivo. The 32-mg/kg dose produced viral load decreases of >3 log10 by day 27, and the profile was similar to that seen when mice were treated with a daily regimen of RAL, 3TC, and TDF. In the 4- and 12.5-mg/kg dose groups, viral titers began to rebound between days 36 and 45. Receptor occupancies in the highest-dose group were between 40 and 60% at trough, with lower ROs at the lower doses. At day 63, plasma from all but one mouse contained the single Q577R mutation (the only change seen in any sample) reflective of resistance selection to the anti-gp41 adnectin. Thus, Q577R was probably selected over the dosing term by GSK3732394. Although in vitro a virus with a Q577R substitution should retain susceptibility to GSK3732394 (Table 6), the potentially suboptimal concentrations of GSK3732394, illustrated by the receptor occupancies in Fig. 7, probably allows for this selection and virus breakthrough. One can envision a scenario whereby at the lower ROs, occasionally there will be a situation where a GSK3732394-bound CD4 is not close enough to a gp160-bound CD4 to allow for inhibition by the anti-gp41 components in trans, so the virus can infect the cell. Once infected, GSK3732394, as an entry inhibitor, has no effect on the infected cell, which can produce new virions. From these and resistance selection studies, it appears that the anti-gp41 adnectin has the lowest resistance barrier of the 3 inhibitors, as it is the first mutation selected in vivo. Even, at the highest dose (32 mg/kg), the RO of 40 to 60% may be enough to suppress some, but not all, mice from selecting Q577R. In this animal model, a higher RO may be needed for GSK3732394 dosed as monotherapy to completely suppress virus and avoid the selection of Q577R during longer-term treatment. All in all, these data describe a novel tri-specific inhibitor of HIV-1 virus-cell fusion that has the potential to be a long-acting inhibitor of virus infection. Although GSK3732394 could be used as a part of a long-acting highly active antiretroviral therapy (HAART) regimen, achievement of high receptor occupancy levels may allow it to be used as a standalone therapy under certain conditions. As with all biologic proteins, a current unknown is the potential immunogenicity of the molecule and the effect this might have on its potency and pharmacokinetic properties. Subcutaneously dosed GSK3732394 is now in phase 1 studies in order to answer these questions and to study its potential as a long-acting antiretroviral agent.
MATERIALS AND METHODS
Expression and purification of adnectin-peptide molecules.Adnectins with different linkers and/or peptides at the carboxy end of the anti-gp41 adnectin were expressed with His tags at the amino terminus of the adnectin and purified via cobalt affinity chromatography as described previously (29).
Expression and purification of GSK3732394.GSK3732394 was expressed via transient transfection of HEK 293-6E cells. Briefly, HEK 293-6E cells were seeded into a 1-liter shake flask containing 350 ml of F17 medium (Invitrogen) supplemented with Glutamax (0.3 mM; Invitrogen) and Pluronic F68 (0.1%; Invitrogen) at a density of 7 × 105 cells/ml. Cells were grown overnight at 37°C, 5% CO2, and 80% humidity with shaking at 110 rpm. The following day, cells were transfected with expression plasmid DNA using the Polyplus transfection reagent PEIpro (VWR) according to the manufacturer’s recommendations. The day after transfection, cells were fed with 18 ml/flask of 20% tryptone N1 (Fisher Scientific). Cells were cultured for an additional 5 days. Harvest of conditioned medium was accomplished by centrifugation to pellet cells.
GSK3732394 was purified using three chromatography steps followed by a final ultrafiltration/diafiltration (UFDF) step for concentration and formulation. Initially, hydrophobic interaction chromatography (HIC) purification was accomplished with a Toyopearl butyl 650 M column (Tosoh). The eluate was then subjected to additional chromatography using 40 μM type I ceramic hydroxyapatite resin (Bio-Rad). Anion-exchange chromatography using a Poros HQ50 column was employed as a final polishing step. The final eluate was concentrated to 10 mg/ml using a Millipore Pellicon 2 50-kDa 0.1M2 membrane and then buffer exchanged with 6 volumes of 25 mM phosphate–150 mM trehalose (pH 6.8). Post concentration, the product was spiked to 0.1% Pluronic F68 and sterile filtered.
Kinetics of binding to defined targets.Determination of the binding kinetics of the two individual adnectins has been described previously (22, 29). Binding activity of the peptide component was determined using His-tagged PRD-828 (Table S2). This peptide contains three identical sequence segments from the HR1, and two segments from the HR2 regions of gp41, and spontaneously forms a five-helix bundle, displaying a single open peptide binding site, analogous to that described elsewhere (43). By design, PRD-828 contains only the stretch of gp41 involved in peptide binding and does not include the N17 region. 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 5-helix bundle peptide PRD-828 was diluted in running buffer (HBS-P+; GE Healthcare) to 10 nM and flowed over the neutravidin surface until 122 RU had accumulated. GSK3732394 diluted in running buffer was flowed over the captured PRD-828 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 min or 10 min. A surface consisting of nonbinding peptide biotin-IZIZ (Table S2) captured onto neutravidin was used for reference subtraction, and buffer-only samples were included for background subtraction. The PRD-828 surface was regenerated between cycles with two injections of 0.1% SDS. A 1:1 Langmuir binding model was fit to the double-referenced sensorgrams to determine kinetic parameters using Biacore T100 evaluation software, version 2.0.1 (GE Healthcare).
Cells, viruses, and antiviral assays.MT-2, HEK 293T, and CEM-NKR-CCR5-Luc cells, the proviral DNA clone of NL4-3, and primary clinical isolates were obtained from the NIH AIDS Research and Reference Reagent Program. B6 cells were also used and contain an integrated copy of a luciferase marker gene driven by the HIV long terminal repeat (LTR) (44). The replication-competent NL4-3 variant RepRLucNL virus expresses Renilla luciferase and was used for most antiviral assays as described previously (45). Populations of envelope clones were obtained and used as described elsewhere (29). Cytotoxicity was determined in MT-2 cells after 4 days of incubation using XTT [2,3-bis-(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide] to measure cell viability using an established protocol (46).
Primary clinical isolates were examined for susceptibility to GSK3732394 using the CEM-NKR-CCR5-Luc cells as a reporter line. The human T-cell line CEM-NKR-CCR5-Luc expresses CD4, CXCR4, and CCR5 receptors on its cell surface and carries the luciferase reporter gene under transcriptional control of the HIV-2 LTR (47). For susceptibility analyses, the virus was used to infect CEM-NKR-CCR5-Luc cells in the presence or absence of serial dilutions of compound. On the assay setup day, 5 × 106 cells were prepared per 96-well plate, concentrated via low-speed centrifugation (1,000 rpm), and resuspended in 0.5 ml of selection medium (47). HIV-1 was incubated with the cells at 37°C and 5% CO2 for 1 h within the range of multiplicities of infection (MOI) of 0.005 to 0.01. Serial 4- or 5-fold dilutions of GSK3732394 or other inhibitors were diluted in a 10-μl volume on a 96-well black- or clear-bottom plate. The cell-virus mixture was diluted to the proper volume using assay media (RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum [FBS], 100 U/ml of penicillin–100 μg/ml of streptomycin, 10 μg/ml of Polybrene) and 190 μl was added to each well culture plate. The cultures were incubated at 37°C and 5% CO2 for 5 to 8 days and the assay was processed and quantitated for virus growth by the amount of expressed luciferase using the Bright-glo luciferase kit (Promega). Susceptibility of viruses to inhibitors was determined by XL-Fit analysis of luciferase signals. The results from 2 experiments were averaged to establish the EC50s.
Resistance selection.Two million MT-2 cells were infected with NL4-3 virus in the presence of a 2× EC50 inhibitor concentration at an MOI of 0.005 to 0.05. Syncytium formation as a marker for viral infection was monitored. When syncytium formation reached approximately >10%, a 1/1,000 to 1/100 volume of the infectious supernatant was transferred to fresh MT-2 cells in the presence of the inhibitor at an increased concentration of 2× stepwise. When consistent virus breakthrough was observed, the viral supernatant was evaluated against the inhibitor in the B6 antiviral assay. A potency shift more than 10-fold usually indicates the appearance of resistant virus. The infected cells or the viral supernatants were used to obtain the viral genomes by PCR or RT-PCR, followed by sequencing to identify the amino acid changes. Amino acid changes were then introduced into the viral genome using site-directed mutagenesis and cloning. The recombinant viruses were evaluated in a replicating virus assay for potency shift versus wild-type virus.
Mouse model and efficacy studies.GSK3732394 was examined for efficacy in a humanized mouse model of HIV infection established and running at TransCure bioServices SAS (Archamps, France). A NOD/Shi-scid/IL-2Rγ nonspecific immunodeficient mouse strain (NOG) was humanized with hematopoietic stem cells isolated from cord blood. After reconstitution of the human immune system and confirmation of the presence of CD4+ cells (about 14 weeks posttransplant), the humanized mice were infected with HIV-1 strain YU2. A total of 40 mice with >25% circulating human CD45+ were used in the study. Selected humanized mice were inoculated with the HIV YU2 strain by intraperitoneal injection. Infection proceeded for 31 days, after which plasma was obtained to assess viral loads by quantitative RT-PCR and the level of human CD4+ cells by flow cytometry. Mice were placed into 5 groups, normalizing HIV loads and human CD4 cells across the different groups. There were 3 groups treated with GSK37323794 at 4, 12.5, and 32 mg/kg, along with a vehicle treatment group and a control HAART-treated group given raltegravir, lamivudine, and tenofovir disproxil fumarate in their food. The GSK3732394 and vehicle treatment groups were dosed s.c. starting on day 37 postinfection every 3 days, while the HAART treatment group was dosed daily via food. Dosing continued for an additional 60 days (20 doses). At every third dose (9 days) during the dosing schedule, plasma was obtained prior to the s.c. injection for analysis of viral loads and receptor occupancy. At day 63 postdosing (day 100 after virus infection), most mice were sacrificed, except for 3 mice in the HAART-treated group and 3 mice in the 32-mg/kg dose group. Those mice were kept for an additional 21 days without drug treatment to look at virus rebound. Plasma was obtained from these animals after 9 days and at the end of the experiment.
Viral loads were determined using 20 μl of extracted plasma. HIV loads were determined using the generic HIV Charge Virale quantitation kit (Biocentric, France). HIV loads were considered not detectable when threshold cycle (CT) values were lower than 37. The limit of sensitivity was estimated to be 100 copies/ml. Receptor occupancy of GSK3732394 was determined via flow cytometry on an Attune NxT flow cytometer (Life Technologies). Human immune hematopoietic subpopulations were monitored using fluorescein isothiocyanate (FITC) anti-human CD3 (Miltenyi Biotec), anti-CD4 MAb OKT4 (BioLegend), brilliant violet 510 anti-human CD8 (BD Biosciences), and an Alexa Fluor 647-labeled anti-CD4 adnectin, 4945_G06-107 (29).
Receptor occupancy of GSK3732394 was determined via flow cytometry on an Attune NxT flow cytometer (Life Technologies). OKT4 does not compete with GSK3732394 for binding to CD4 and therefore can be used as a marker for CD4+ cells regardless of the presence of GSK3732394. The AF647-4945_G06 adnectin does compete with GSK3732394 for binding to CD4 and is used as a probe for occupancy. For each time point, the AF647 median fluorescence intensity (MFI) of the CD4+ CD8− CD3+ cell population from vehicle control animals was indicative of 0% occupancy, and the AF647 MFI of the CD4− CD8+ CD3+ population from GSK3732394-dosed animals was used as a surrogate for 100% occupancy. The RO for a given sample was calculated using this equation:
Plasma levels of GSK3732394 were quantitated via an enzyme-lined immunosorbent assay (ELISA). Streptavidin-coated black 96-well plates were coated with biotinylated PRD-828 at 1 μg/ml in phosphate-buffered saline–Tween (PBST; Invitrogen) for 1 h at room temperature. The coated plates were washed 3 times with PBST and then incubated with SuperBlocker T20 (Thermo Scientific) blocking buffer for 1 h. The plates were washed 3 times with PBST, and mouse plasma samples were serially diluted 5-fold and added to the assay-ready plate. The plates were incubated at room temperature with shaking for 1 h and washed 3 times with PBST before 1:20,000-diluted goat polyclonal horseradish peroxidase (HRP)-conjugated anti-HSA antibody (AbCam) was applied for 30 min. Plates were washed 3 times with PBST and developed with Supersignal ELISA pico substrate (Thermo Scientific). Signals were read on an EnVision plate reader. Concentrations were determined through comparison with a standard curve, which was created through 3-fold dilutions of GSK3732394 in PBST with added normal mouse plasma (VWR).
For PCR amplification of gp160 genes in mouse plasma, HIV-1 RNA was extracted from 200 μl of mouse plasma samples using a QIAamp MinElutevirus kit and eluted in 40 μl of RNase-free water. cDNA was generated from 34 μl of viral RNA using a SuperScript III first-strand synthesis kit (Invitrogen) per the manufacturer’s protocol. Eighty microliters of cDNA solution was run over the MinElute PCR purification kit (Qiagen) and eluted in 10 μl of EB buffer (10 mM Tris-Cl [pH 8.5]) per the manufacturer’s protocol. Amplification used YU2-specific primers outside of the envelope gene using a Platinum Taq polymerase high-fidelity kit (Invitrogen) The first-round PCR product was purified using a MinElutePCR purification kit, eluted in 10 μl of EB buffer, and used as template for the second round of PCR. Amplified envelope PCR products were subjected to population-based sequencing using a library of envelope-specific primers. The HIV-1 env sequences from the mice were aligned to the YU2 sequence obtained from TransCure using AlignX software in the Vector NTI package (Invitrogen).
ACKNOWLEDGMENTS
We acknowledge Mei Sun, Yulia Benitez, Dieter Drexler, and many other present and former employees of Bristol-Myers Squibb for their support of this program.
ViiV Healthcare has assumed responsibility for clinical development of GSK3732394. Most authors were employees of Bristol-Myers Squibb at the time of this work, and many of those are currently employed by ViiV Healthcare.
FOOTNOTES
- Received 29 May 2019.
- Accepted 19 July 2019.
- Accepted manuscript posted online 2 August 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/JVI.00907-19.
- Copyright © 2019 American Society for Microbiology.
