ABSTRACT
Peptides derived from the C-terminal heptad repeat 2 (HR2) region of the HIV-1 gp41 envelope glycoprotein, so-called C peptides, are very efficient HIV-1 fusion inhibitors. We previously developed innovative gene therapeutic approaches aiming at the direct in vivo production of C peptides from genetically modified host cells and found that T cells expressing membrane-anchored or secreted C peptides are protected from HIV-1 infection. However, an unwanted immune response against such antiviral peptides may significantly impair clinical efficacy and pose safety risks to patients. To overcome this problem, we engineered a novel C peptide, V2o, with greatly reduced immunogenicity and excellent antiviral activity. V2o is based on the chimeric C peptide C46-EHO, which is derived from the HR2 regions of HIV-2EHO and HIV-1HxB2 and has broad anti-HIV and anti-simian immunodeficiency virus activity. Antibody and major histocompatibility complex class I epitopes within the C46-EHO peptide sequence were identified by in silico and in vitro analyses. Using rational design, we removed these epitopes by amino acid substitutions and thus minimized antigenicity and immunogenicity considerably. At the same time, the antiviral activity of the deimmunized peptide V2o was preserved or even enhanced compared to that of the parental C46-EHO peptide. Thus, V2o is an ideal candidate, especially for those novel therapeutic approaches for HIV infection that involve direct in vivo production of antiviral C peptides.
INTRODUCTION
Peptides and proteins have emerged as potent drugs for the therapy of various inherent and acquired human diseases, including human immunodeficiency virus type 1 (HIV-1) infection (1). However, long-term treatment with therapeutic polypeptides often elicits undesirable immune responses that can significantly impair clinical efficacy and pose safety risks to patients (2). A major problem associated with the chronic administration of exogenously produced recombinant proteins is the generation of antidrug antibodies (ADAs) (2). The humoral immune response is mounted upon recognition of antibody epitopes within the protein drug by cognate B cell antigen receptors and subsequent presentation of immunogenic protein fragments on class II HLA molecules on the surface of professional antigen-presenting cells to CD4+ helper T cells (major histocompatibility complex class II [MHC-II] epitopes).
In gene therapeutic approaches aiming at the direct in vivo production of therapeutic proteins from genetically modified host cells, a cellular immune response mediated by CD8+ cytotoxic T lymphocytes (CTLs) is an additional concern. Protein-specific CTLs could rapidly recognize and eliminate gene-modified cells displaying foreign peptide fragments in the context of MHC-I gene products, the class I HLA molecule HLA-A, HLA-B, or HLA-C (CTL epitopes, MHC-I epitopes).
Peptide fragments presented by class I HLA molecules typically comprise between 8 and 11 amino acids (aa), while class II HLA molecules generally display longer peptides (3, 4). In both cases, a core binding motif of approximately 9 amino acids in length binds within the groove of the HLA molecule (3).
However, HLA molecules cannot present all varieties of peptide fragments; instead, each isoform has a unique peptide binding pocket and prefers distinct amino acids at certain positions of the peptide. This peptide preference is mainly determined by the primary and auxiliary anchor residues, where one particular or a closely related amino acid is required for efficient peptide binding (5, 6). Consequently, by mutating the anchor amino acids of an antigenic peptide, which is normally bound by a specific HLA type, the peptide will no longer be presented to the immune system and is thus rendered nonimmunogenic. A similar strategy can be applied to delete antibody epitopes preventing recognition by B cell antigen receptors and the respective antibodies.
Accurate mapping and deletion of all potential antibody and MHC epitopes, while maintaining biotherapeutic activity, may be an elusive goal for larger proteins; however, for small peptide therapeutics, complete deimmunization by epitope removal may be feasible. Accordingly, in the present study, we engineered a small anti-HIV C peptide with reduced immunogenicity while retaining antiviral activity. C peptides are derived from the highly conserved C-terminal heptad repeat 2 region (HR2) of the HIV envelope glycoprotein gp41 and are very potent inhibitors of virus entry (7, 8). Several C-peptide entry inhibitors have been described (8, 9, 11–13), and all of these are composed of viral and/or synthetic sequences and thus are potentially immunogenic. The best-known C peptide, T-20 (enfuvirtide [Fuzeon]), is a highly active 36-amino-acid peptide derived from the HIV-1HxB2 HR2 which has successfully been used in the clinic since 2003. However, T-20 therapy induces local immune reactions, and, moreover, T-20-resistant (T-20r) HIV variants develop rapidly.
According to the HIV Molecular Immunology Database at Los Alamos National Laboratory, C-peptide sequences (LANL; e.g., T-20 or C46, positions 638 to 673 and 628 to 673 of gp160, respectively) do not overlap dominant MHC-I or MHC-II epitopes in gp41. In summary, only a few MHC epitopes in the HR2 region of gp41 have been confirmed experimentally.
However, antibodies against gp41-derived C peptides are found in most HIV-infected patients (15–17). Even though the HR2 region of gp41 is not immunodominant, it contains the antigenic cluster II region (positions 644 to 663 of gp160), bound by several nonneutralizing antibodies (17, 18). Moreover, many C-peptide sequences overlap the membrane-proximal external region of gp41, which is recognized by broadly neutralizing antibodies like 2F5 (ELDKWA at positions 662 to 667 of gp160) (19, 20) and 4E10 (NWFDIT at positions 671 to 676 of gp160) (21).
Consequently, reducing or even eliminating the antigenicity and immunogenicity of C peptides (while retaining full function) may significantly promote safety and the antiviral effect. In this study, we successfully generated a highly active therapeutic C peptide, designated V2o. This peptide has a greatly reduced immunogenic potential due to amino acid substitutions in antibody epitopes and in silico-predicted MHC-I epitopes, while retaining excellent antiviral activity.
MATERIALS AND METHODS
Study population and sample preparation.Frozen sera from 69 HIV-1-infected individuals were obtained from the HIVCENTER, Goethe University Clinic, Frankfurt am Main, Germany, and the University Medical Center Hamburg-Eppendorf, Hamburg, Germany. The sera represented all clinical stages of HIV-1 infection. Sera from 10 long-term nonprogressors (LTNPs) and 10 HIV-2-infected individuals were provided by the Georg-Speyer-Haus, Frankfurt am Main, Germany. Frozen Ficoll-purified peripheral blood mononuclear cells (PBMCs) from HIV-1-infected individuals were received from the University Medical Center Hamburg-Eppendorf. PBMCs of healthy donors (Georg-Speyer-Haus) were isolated by density gradient centrifugation using Pancoll (PAN Biotech, Aidenbach, Germany), cryopreserved with 10% dimethyl sulfoxide in fetal calf serum (PAN Biotech), and stored in liquid nitrogen. Pancoll-purified lymphocytes were immunostained using an HLA-A*02-specific monoclonal antibody (Becton, Dickinson, Heidelberg, Germany) to screen for HLA-A*02-positive donors by flow cytometry (FACSCalibur; Becton, Dickinson). For infection of primary human CD4+ T cells with replication-competent HIV-1, CD8+ T cells within PBMCs from healthy donors were depleted using CD8-specific microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany).
Animals.Sera were obtained from nine chronically simian immunodeficiency virus mac251 (SIVmac251)-infected rhesus macaques (Macaca mulatta). All animals were maintained in accordance with the guidelines of the Committee on the Care and Use of Laboratory Animals under an NIAID-approved animal study protocol, and all studies and procedures were reviewed and approved by the Institutional Animal Care and Use Committees of the NIH and Harvard University.
Cells.The human embryonic kidney cell line 293T was obtained from the American Type Culture Collection and was maintained in Dulbecco's modified Eagle's medium. The T cell line PM1, a subclone of HuT78 expressing CD4, CXCR4, and CCR5, was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, from Marvin Reitz (22) and was cultured in RPMI 1640 medium. All media were supplemented with 10% fetal calf serum, 4 mM l-glutamine, 100 units ml−1 penicillin, and 0.1 mg ml−1 streptomycin (complete media). Human PBMCs were cultured in X-Vivo 15 medium supplemented with 5% human type AB serum, 4 mM l-glutamine, 100 units ml−1 penicillin, 0.1 mg ml−1 streptomycin, and 100 units ml−1 human interleukin 2 (Novartis, Nuremberg, Germany) (X5 medium).
Synthetic peptides and monoclonal antibodies.T-20 (enfuvirtide) was obtained from Roche (Grenzach-Whylen, Germany). All other peptides were custom synthesized (GeneCust, Dudelange, Luxembourg). The amino acid sequences of antiviral C peptides are shown in Fig. 1. Cellular immune responses of HIV-1 PBMCs were investigated either with peptides consisting of 9-mers of C46-EHO or V2 or with 15-mers overlapping for 11 amino acids of C46-EHO or V2o. Using the matrix method, overlapping peptides were arranged in peptide pools consisting of three peptides each, with each peptide being present in two different pools. PBMCs of healthy individuals were incubated with 9-mer peptides derived from human cytomegalovirus (CMV; NLVPMVATV) or Epstein-Barr virus (EBV; GLCTLVAML) or mutated versions of these peptides (CMV-mut; NWVPMVATA) (EBV-mut; GWCTLVAMA). The monoclonal antibody 2F5 (recognizing the epitope ELDKWA within the gp41-derived C-peptide sequence [20]) was from Polymun Scientific, Vienna, Austria.
Schematic overview of HIV-1 gp41 and of HR2-derived C peptides. (A) The functional domains of the HIV-1 gp41 molecule comprise an N-terminal fusion peptide (FP), two leucine zipper-like heptad repeat domains (HR1 and HR2), the membrane-proximal external region (MPER), the transmembrane domain (TM), and the intracellular domain (Intra). HIV entry-inhibitory C peptides C46 and C46-EHO are derived from the HR2 of HIV-1HxB2 and HIV-2EHO/HIV-1HxB2, respectively. Variants V1 to V4 and V2o are based on C46-EHO but contain several amino acid substitutions, as indicated. (B) Flowchart of V2o peptide development based on C34-EHO.
MHC-I epitope predictions.Peptide sequences were analyzed for the presence of epitopes potentially binding class I HLA molecules using the freely online-accessible algorithms SYFPEITHI (Department of Immunology, University of Tübingen, Tübingen, Germany; http://www.syfpeithi.de) (23) and BIMAS (Bioinformatics and Molecular Analysis Section, National Institutes of Health; http://www-bimas.cit.nih.gov) (24). The SYFPEITHI database reflects epitopes present in natural MHC ligands, T cell epitopes, or binding peptides. The BIMAS algorithm is based on experimentally measured β2-microglobulin dissociation half-lives. Both programs provide peptide sequences that are likely to be presented by the selected HLA molecules as well as a score and ranking for each subsequence. Peptides were qualified as MHC-I epitopes, if they scored >20 and >10 by the SYFPEITHI and BIMAS algorithms, respectively.
Lentiviral packaging.Lentiviral vector supernatants were produced by transient transfection of 293T cells using the calcium phosphate precipitation method as described previously (25). The lentiviral transfer vector plasmid pHR′SIN-cPPT-SEW (26), encoding enhanced green fluorescent protein (eGFP), was packaged using a lentiviral Gag-Pol construct (pCMV-dR8.91 [27]) and one of several envelope proteins, either HIV-1HxB2 (kindly provided by G. Melikyan, Atlanta, GA), HIV-1HxB2_T-20r(SIM) (28), HIV-1HxB2_T-20r(DTV) (28), HIV-1Ba-L (29), HIV-1Ba-L_C46r (25), SIVmac251 (30), or the glycoprotein of vesicular stomatitis virus (VSV-G) (31). Lentiviral supernatants were filtered (0.45-μm pore size), concentrated by low-speed centrifugation (approximately 16 h, 4,000 × g, 4°C), and stored at −80°C until use. All supernatants were titrated on PM-1 cells.
Neutralization assay.A total of 2 × 104 PM-1 cells per well were seeded in 96-well tissue culture plates. PM-1 cells were transduced at a low multiplicity of infection (MOI; ∼0.2) with lentiviral vector particles encoding eGFP and packaged with various envelope glycoproteins. Transduction was performed in the presence of increasing concentrations of synthetic antiviral C peptides in duplicate. The 96-well plate was centrifuged (1,000 × g) for 1 h at 31°C. Subsequently, cells were incubated in a humidified CO2 incubator at 37°C with 5% CO2, and transduction efficacy (eGFP-positive cells) was determined by flow cytometry 5 days later. Calculation of half-maximal (50%) inhibitory concentrations (IC50s) was performed using GraphPad Prism (version 5.00) software (GraphPad Software, San Diego, CA); sigmoid dose-response curves were calculated by nonlinear regression [log(inhibitor) versus normalized response-variable slope] after logarithmic transformation of the data.
Infection of primary human CD4+ T cells with replication-competent HIV-1.A total of 2 × 105 primary human CD4+ T cells were seeded as triplicate samples in 48-well plates. Synthetic antiviral peptides were added at different concentrations, and cells were infected with different strains of replicating HIV-1. After incubation at 37°C for 24 h, cells were washed three times with phosphate-buffered saline (PBS), resuspended in medium without peptides, and reseeded in 48-well plates. Cells were cultured at 37°C. For quantitative detection of p24 antigen expression, cell culture supernatants were analyzed by p24 enzyme-linked immunosorbent assay (ELISA) at day 6.
p24 ELISA.ELISA plates (Costar; Corning, Corning, NY) were coated with anti-p24 M01 antibody (Polymun) at a concentration of 200 ng/well overnight at 4°C or at 37°C for 2 h. The plates were washed three times with 0.1% Tween 20–PBS. Samples and the p24 protein standard (Polymun) were added at different dilutions in a total volume of 50 μl per well. p24 was detected by a secondary biotin-labeled antibody (37G12-biotin; Polymun) diluted in 0.1% Tween 20–1% bovine serum albumin (BSA)–PBS (1 h at room temperature). ELISA plates were washed again three times and incubated with streptavidin–β-galactosidase (Roche) in 0.1% Tween 20–1% BSA–PBS with 2 mM MgCl2 for 20 min. Again, the ELISA plates were washed three times and resorufin (Sigma-Aldrich, Steinheim, Germany) was added. The optical density was measured at 550 and 655 nm using an ELISA microplate reader (Bio-Rad, Vienna, Austria). The detection limit of the assay was ≤60 pg p24/ml.
Dot blotting.Nitrocellulose membranes (VWR International, Darmstadt, Germany) were hydrated in 10× phosphate-buffered saline, and 0.5 μg of synthetic peptides was spotted onto membranes. Nonspecific sites were blocked with 5% fat-free milk powder (Sigma-Aldrich) in phosphate-buffered saline containing 0.1% Tween 20 (MPBST). Membranes were stained with primary control antibodies (1.25 μg ml−1 in MPBST) or patient sera (5 μl ml−1 in MPBST) and secondary horseradish peroxidase-conjugated human/mouse IgG-specific antibodies from goat (diluted 1:10,000 in MPBST; Dianova, Hamburg, Germany). Detection was performed using enhanced chemiluminescence (GE Healthcare, Munich, Germany) according to the manufacturer's instructions.
Epitope mapping.Nitrocellulose membranes spotted with 15-mer peptides overlapping for 13 amino acids were obtained from JPT Peptide Technologies GmbH (Berlin, Germany). Membranes were hydrated with methanol, washed with Tris-buffered saline (TBS), and blocked overnight at 4°C in 1% blocking solution (Roche). Membranes were incubated with either the primary control antibody 2F5 (2 to 5 ng ml−1 in 0.5% blocking solution) or HIV-1 patient serum (2 to 5 μl ml−1 in 0.5% blocking solution) for 3 h with rocking and washed with TBS containing 0.05% Tween 20. Subsequently, membranes were incubated for 2 h with a secondary horseradish peroxidase-conjugated human IgG-specific antibody from goat (diluted 1:10,000 in 0.5% blocking solution; Dianova) and washed three times with TBS. Detection was performed using enhanced chemiluminescence (GE Healthcare) according to the manufacturer's instructions.
ELISpot assay.The release of gamma interferon (IFN-γ) from cytotoxic CD8+ T lymphocytes was measured by enzyme-linked immunospot (ELISpot) assay using an ELISpotPlus kit from Mabtech (Hamburg, Germany) according to the manufacturer's instructions. In short, frozen PBMCs were thawed in X5 medium containing DNase (2 U ml−1) to prevent cell clumping and were incubated in a humidified CO2 incubator at 37°C and 5% CO2 for 3 h. The cells were then centrifuged, resuspended in fresh X5 medium, and counted using trypan blue staining. One hundred microliters of the cell suspension (5 × 104 to 2 × 105 cells per well) and 50 μl peptide (antigen) solution (final concentration, 5 μg ml−1) were added in duplicate to ELISpotPlus plates precoated with capture antibodies specific for IFN-γ. PBMCs of HIV-1 patients were stimulated with either the CEF peptide pool (Mabtech) (32), HIV-1-derived 9-mer peptides (Gag [77SLYNTVATL85], reverse transcriptase [RT; 309ILKEPVHGV317], Vpr [59AIIRILQQL67]), synthetic 9-mer peptides (C46-EHO, V2), or peptide pools generated by mixing three different 15-mer peptides of C46-EHO and V2o on the basis of the matrix method. PBMCs of healthy donors were incubated with 9-mer peptides of CMV, CMV-mut, EBV, or EBV-mut. After incubation in a humidified CO2 incubator at 37°C and 5% CO2 for 20 h, cells were removed and plates were developed using a biotinylated IFN-γ-specific antibody and streptavidin-alkaline phosphatase. Blue spots displayed on the plate membranes were determined with a Zeiss microscope and normalized to a cell number of 1 × 105 PBMCs. The number of IFN-γ-secreting cells was obtained by calculating the mean value of the duplicates and subtracting the number of spots in unstimulated control wells.
RESULTS
Substitution of primary anchor amino acids reduced the number and scores of in silico-predicted MHC-I epitopes.HIV-1 gp41-derived C peptides, e.g., T-20 or C34, are very efficient inhibitors of HIV-1 entry. The peptide C34-EHO, originating from the HIV-2 strain EHO, however, is highly active not only against HIV-1 but also against SIV (11). This is a useful and desirable characteristic, as it permits preclinical efficacy testing in the rhesus macaque model. However, at least for small HIV-1-derived antiviral C peptides like T-20 (36 aa), C34 (34 aa), or a membrane-anchored version of T-20, rapid development of resistant virus variants has been observed (17, 29, 33, 34). On the other hand, we previously found that a membrane-anchored version of the elongated peptide C46 (46 aa derived from HIV-1HxB2) was highly active but less prone to induce the generation of resistant virus strains (25).
Thus, as a lead structure for the development of a nonimmunogenic antiviral C peptide, we generated the 46-amino-acid peptide C46-EHO, consisting of the HIV-2EHO-derived C34-EHO elongated at the C terminus by 12 amino acids derived from HIV-1HxB2 (Fig. 1). The HIV-1HxB2-derived peptide C46 (Fig. 1), which has already been used in preclinical and clinical gene therapy studies (15, 35–37), was used as a control peptide in this study.
Putative MHC-I epitopes in C46 and C46-EHO were determined in silico using two matrix-based bioinformatic algorithms, SYFPEITHI and BIMAS (23, 24). In both programs, a given polypeptide sequence is first dissected into overlapping peptide fragments between 8 and 11 amino acids in length. Then, the capacities of the individual peptides to bind to different HLA types are evaluated and the peptides are ranked according to their score in silico. Potential MHC-I epitopes in C46 and C46-EHO predicted by SYFPEITHI and BIMAS concentrated mainly in the N- and C-terminal parts of the peptides (Fig. 2). For both therapeutic peptides, SYFPEITHI issued fewer potential MHC-I epitopes than the BIMAS software. C46-EHO epitopes 10 (FLDANITKL) and 34 (QELDKWASL) were identified as high-score binders for different HLA alleles by both algorithms (Table 1). BIMAS predicted three additional major MHC-I epitopes binding several HLA alleles with high scores: epitope 8 (VRFLDANIT) overlapping the epitope FLDANITKL, also predicted by SYFPEITHI, as well as epitopes 2 (QQWERQVRF) and 25 (QQEKNMYEL) (Table 1).
Substitution of primary anchor amino acids reduced the number and scores of in silico-predicted MHC-I epitopes. Using the SYFPEITHI (A) and BIMAS (B) software, potential MHC-I epitopes in C peptides C46, C46-EHO, V2, and V2o were predicted for different HLA types in silico. Each 8/9/10- or 11-mer that was predicted for any HLA type is indicated with its respective score. Scores below 20 (SYFPEITHI; A) and 10 (BIMAS; B) are not shown.
Major MHC-I epitopes in C46-EHO for different HLA types according to SYFPEITHI and BIMAS predictions
We aimed to eliminate as many of these in silico-predicted MHC-I epitopes in C46-EHO as possible, while preserving antiviral activity. Therefore, the primary anchor amino acids at positions P-2 and P-9 (epitopes 10 and 34) or merely at position P-2 (epitopes 2, 8, and 25) of the five major MHC-I epitopes in C46-EHO were replaced by less preferred residues. In order to maintain antiviral activity, amino acids were substituted with ones having similar physical properties or with residues that can be found at the same position in other antiviral C peptides. Using this approach, four variants of C46-EHO, V1 to V4, were generated (Fig. 1). In silico analysis of V1 to V4 using SYFPEITHI and BIMAS predicted a reduced number of MHC-I epitopes and lower scores compared to those for the original C46-EHO for all peptide variants (an example for V2 is shown in Fig. 2).
The optimized peptide V2o was designed on the basis of the amino acid sequence of V2 (Fig. 1). We selected variant V2 for further optimization, as only a few HIV-1 patients had preexisting antibodies to this peptide (see below; Table 2) and its antiviral activity was very high (see below; Table 3). To generate V2o, the C-terminal membrane-proximal external region (662ELDKWASLWNWF673) originating from HIV-1HxB2 was replaced by the corresponding amino acid sequence of HIV-2EHO (662KLNQWDIFSNWF673) to further reduce preexisting humoral immunity (see below). Moreover, several amino acids in the N-terminal region were modified, thus further decreasing immunogenicity according to SYFPEITHI and BIMAS predictions (Fig. 2; see Fig. S1 in the supplemental material). Taken together, mutation of primary anchor residues of MHC-I epitopes considerably decreased the in silico-predicted immunogenicity of C peptides.
Reactivity of sera from HIV-1- or HIV-2-infected individuals and SIVmac251-infected rhesus macaques with therapeutic C peptides
IC50s for entry inhibition of replication-incompetent lentiviral vectors packaged with various envelope glycoproteins by therapeutic C peptides
Replacement of primary anchor amino acids according to in silico predictions prevented activation of peptide-specific CTLs in vitro.Since in vivo no more than a small amount of predicted MHC-I epitopes is actually processed, bound, and presented by HLA molecules, we intended to (i) verify in silico-identified MHC-I epitopes experimentally and (ii) determine the impact of mutating the primary anchor amino acids of predicted MHC-I epitopes on the stimulation of cellular immunity.
Strong T cell responses against the HR2 domain of gp41 have never been measured experimentally (LANL HIV Molecular Immunology Database), and the scores of the in silico-predicted MHC-I epitopes in C46-EHO were rather low. Therefore, we first aimed to prove the feasibility of our concept using two stronger MHC-I epitopes derived from human CMV and EBV. We mutated the primary anchor amino acids of the viral MHC-I epitopes, both known to be presented by HLA-A*0201, and analyzed the impact on activation of peptide-specific CTLs in vitro using IFN-γ ELISpot assays. The peptide 495NLVPMVATV503 derived from the CMV matrix protein pp65 (SYFPEITHI and BIMAS scores, 30 and 160, respectively) was mutated to NWVPMVATA (CMV-mut; where the mutated amino acids are underlined; SYFPEITHI and BIMAS scores, 14 and 0.002, respectively), while the EBV early lytic protein-derived MHC-I epitope 259GLCTLVAML267 (SYFPEITHI and BIMAS scores, 28 and 149, respectively) was changed to GWCTLVAMA (EBV-mut; SYFPEITHI and BIMAS scores, 12 and 0.002, respectively). The wild-type and mutated peptides or control antigens were incubated with HLA-A*02-expressing PBMCs from healthy individuals for 20 h, and IFN-γ secretion was quantified by ELISpot assay. The CEF peptide pool consisting of 23 highly immunogenic peptides from CMV, EBV, and influenza virus was used as a positive control for CTL stimulation and caused a specific IFN-γ response in the range of 7 to 194 spot-forming units (SFUs) (see Fig. S2A in the supplemental material). C46-EHO-derived MHC-I epitopes 2 (not predicted for HLA-A*0201) and 10 (predicted for HLA-A*0201) were used as negative controls and, as expected, did not stimulate CTLs from healthy donors (<2.5 SFUs; see Fig. S2A in the supplemental material). Incubation of PBMCs with wild-type CMV or EBV peptides resulted in low to moderate IFN-γ levels (0.25 to 111 SFUs and 0.5 to 74 SFUs, respectively); however, no IFN-γ secretion was detected for the mutated peptides CMV-mut and EBV-mut (<1 SFU for both; see Fig. S2A in the supplemental material). Hence, modification of primary anchor amino acids of MHC-I epitopes according to in silico predictions impaired activation of peptide-specific CTLs in vitro, indicating successful elimination of the MHC-I epitopes.
Absence of preexisting C-peptide-specific CTL response in HIV-1-infected individuals.In the next step, ELISpot assays were performed with PBMCs of 23 HIV-1-infected individuals incubated with peptides corresponding to the five predicted MHC-I epitopes of C46-EHO or the mutated versions of these epitopes that can be found in the V2 variant of C46-EHO (Fig. 3). The CEF peptide pool used as a positive control caused a specific IFN-γ response in the majority of patient samples (89.5%); however, incubation of the patient cells with the HIV-1-derived control peptide Vpr (59AIIRILQQL67) induced only marginal IFN-γ levels (1 to 9 SFUs), while for the HIV-1 peptides Gag (77SLYNTVATL85) and RT (309ILKEPVHGV317), IFN-γ responses were near background levels (<4 SFUs), indicating a poor general immune status of the patients. Correspondingly, neither the five predicted MHC-I epitopes of C46-EHO nor the corresponding mutated epitopes of V2 induced IFN-γ secretion (Fig. 3). Consequently, the HIV-1 patients did not have a preexisting CTL response against the in silico-predicted MHC-I epitopes of C46-EHO or against the potentially less immunogenic C-peptide variant V2.
Absence of preexisting C-peptide-specific CTL response in HIV-1-infected individuals. PBMCs from HIV-1-infected patients (Pt.) were incubated with peptides representing the five major MHC-I epitopes predicted in silico for C46-EHO (2-, 8-, 10-, 25-, and 34-wt [wild type]) or with the respective peptides from V2 with mutated primary anchor residues (2-, 8-, 10-, 25-, and 34-mut). After 20 h, IFN-γ secretion from activated peptide-specific CTLs was quantified as the number of SFUs by ELISpot assay. A CEF peptide pool and the HIV-1-derived peptides Gag, RT, and Vpr were used as positive controls for CTL activation. The dotted line represents the detection limit of the ELISpot assay. Data represent the means of duplicate samples reduced by the number of spots detected for PBMCs incubated with medium.
Furthermore, we analyzed whether any additional MHC-I epitopes that had not been predicted in silico were present in the C46-EHO and V2o sequences. Overlapping peptides (15-mers overlapping by 11 amino acids) covering the whole amino acid sequence of C46-EHO or V2o were arranged in peptide pools by the matrix method (three peptides per pool, each peptide present twice) and used to stimulate PBMCs of 20 HIV-1-infected individuals. However, no significant stimulation of IFN-γ secretion was detected by ELISpot assay (see Fig. S2B in the supplemental material). These results were further confirmed by stimulating PBMCs of one HIV-2-infected individual as well as PBMCs of seven SIVmac251-infected rhesus macaques with overlapping peptides of V2o (data not shown). In conclusion, neither the original peptide C46-EHO nor modified variant V2 or V2o induced any MHC-I epitope-dependent IFN-γ T cell response in vitro.
Preexisting antibodies against C46, but not C46-EHO, are detectable in most HIV-1-infected patients.Although broadly neutralizing antibodies are rarely found in HIV patient sera, antibodies binding the viral envelope proteins are frequently detected in infected individuals (15–17). Since C peptides are derived from the viral gp41 sequences, preexisting IgG antibody responses in serum samples of 89 HIV-infected patients and nine SIV-infected rhesus macaques were investigated in dot blot assays. The HIV group comprised 10 HIV-2-infected individuals as well as 79 HIV-1-infected patients, which were subdivided into 35 patients who obtained antiretroviral therapy, 34 therapy-naïve patients, and 10 long-term nonprogressors. The C peptides were spotted on nitrocellulose membranes and incubated with the serum samples, and C-peptide-specific antibodies were detected using a peroxidase-conjugated human IgG-specific secondary antibody.
As expected, the majority of the HIV-1-infected patient sera reacted with C46; however, neither the HIV-2-infected patients nor the SIV-infected rhesus macaques had preexisting antibodies to this peptide (Table 2). Within the HIV-1 group, most of the untreated patient samples contained C46-specific antibodies (88%), whereas in patients on antiretroviral therapy, this fraction was considerably smaller (60%), probably due to the decreased viral antigenic load. C46-EHO, which is mainly derived from HIV-2 sequences, was recognized only by about 10% of all HIV-1-infected patient sera, while in most of the HIV-2 and SIV samples, C46-EHO-specific antibodies could be detected (Table 2). Hence, C46-specific antibodies were mainly preexisting in HIV-1-infected individuals, while antibodies to C46-EHO were primarily found in HIV-2- and SIV-infected sera.
Preexisting antibodies in HIV-1-infected patient sera bind the HIV-1-derived C terminus of C46 and C46-EHO.In the next step, we aimed to map the epitopes within the C46 and C46-EHO sequences recognized by preexisting antibodies of HIV-1-infected patients. Cellulose membranes spotted with overlapping peptide fragments (15-mers, overlapping for 13 amino acids) spanning the whole length of C46 and C46-EHO were used for epitope mappings. Upon incubation with patient sera containing C-peptide-specific antibodies, a positive reaction occurred at one or more peptide spots and the relevant epitope was determined. The well-known monoclonal antibody 2F5 that binds the linear epitope 662ELDKWA667 (20, 21, 38) present in C46 as well as in C46-EHO was used as a positive control. In accordance with published results, 2F5 recognized peptide fragments 15, 16, and 17 of C46- and C46-EHO-spotted membranes, corresponding to the amino acid sequence 659ELLELDKWASLW670 of C46 and 659ELQELDKWASLW670 of C46-EHO (see Fig. S3 and Table S1 in the supplemental material).
We performed epitope mappings for C46 using serum samples from 14 randomly chosen HIV-1-infected patients, all of which reacted with the C-terminal part of C46, with 659ELLELDKW666 being the core epitope (see Fig. S3 and Table S1 in the supplemental material). In addition, two of the patient serum samples recognized the amino acid sequence 628WMEWDREINNYT639 at the N terminus of C46. Out of seven HIV-1-infected patient serum samples that contained C46-EHO-specific antibodies, three were used for epitope mapping of C46-EHO. None of these serum samples reacted with the HIV-2-derived part of C46-EHO, but instead, the C terminus, which is almost completely derived from HIV-1 and thus is identical to the C terminus of C46, except for a single amino acid substitution (L661Q), was recognized. In conclusion, the major epitope for antibody binding in C46 and C46-EHO is located in the HIV-1-derived C-terminal part of the peptides.
Absence of preexisting humoral immunity against V2o in HIV-1-infected patients.The HIV-1-infected patient sera containing C46-EHO-specific antibodies as well as 15 to 20 anti-C46-EHO antibody-negative HIV-1-infected patient sera were used in a dot blot to screen for preexisting antibodies binding the C46-EHO-derived peptide variants V1 to V4 and V2o. Peptides V1, V3, and V4 were recognized by most of the anti-C46-EHO-positive sera and, in addition, by some sera that lacked C46-EHO-specific antibodies (Table 2). However, antibody reactivity for V2 was clearly reduced within the anti-C46-EHO-positive sera, and only 1 out of 20 anti-C46-EHO-negative HIV-1-infected patient serum samples was able to detect V2. In peptide variant V2o, the HIV-1-derived C terminus had been replaced by HIV-2-derived sequences, and thus, preexisting antibodies to V2o were not found in any of the HIV-1-infected patient sera tested. However, all of the HIV-2 and most of the SIV samples reacted with V2o (Table 2). Consequently, antibodies to the C46-EHO peptide variants were primarily present in HIV-2- and SIV-infected sera but virtually absent in the HIV-1-infected patient sera tested. In particular, no preexisting humoral immunity against V2o was observed in any of the HIV-1-infected patients analyzed.
C46-EHO and its variants are highly active HIV-1 and SIVmac251 entry inhibitors.The antiviral activity of the peptides C46 and C46-EHO and the C46-EHO variants was analyzed by single-round infection assays using replication-incompetent lentiviral vectors encoding eGFP and PM-1 cells as target cells. The lentiviral vectors were packaged with various HIV-1 envelope glycoproteins, the SIVmac251 envelope glycoprotein, or VSV-G. The 36-amino-acid C peptide T-20 (enfuvirtide), the only HIV-1 entry inhibitor approved for clinical application, was used as a control in these assays.
All C peptides inhibited entry of wild-type HIV-1 envelope glycoproteins HxB2 (X4 tropic) and Ba-L (R5 tropic) in a dose-dependent manner, with IC50s being in the low-nanomolar range (Fig. 4 and Table 3). In addition, lentiviral vectors packaged either with HxB2-derived envelope glycoproteins bearing mutations in the gp41 HR1 547GIV549 motif that confer resistance to T-20 [HIV-1HxB2_T-20r(SIM) and HIV-1HxB2_T-20r(DTV)] or with a Ba-L-derived envelope glycoprotein (Ba-L_C46r [where C46r indicates reduced sensitivity to C46] carrying three mutations in gp120 [I184T, N302Y, E351K] and two mutations in gp41 [A582T, N637K]) (25) with reduced sensitivity to C46 were used in the single-round infection assays. As expected, T-20 was not able to prevent entry of the T-20-resistant pseudoviruses at the highest concentration tested (100 nM) and had a clearly increased IC50 for the inhibition of Ba-L_C46r-mediated entry (Table 3). C46 was active against T-20-resistant HIV-1HxB2 strains but showed an increased IC50 for the inhibition of Ba-L_C46r, in accordance with results previously published by our group (25). However, neither T-20 nor C46 was able to inhibit SIVmac251-mediated entry. In contrast, C46-EHO and the less immunogenic peptide variants V1 to V4 and V2o were highly efficient inhibitors of all HIV-1 and SIV strains analyzed. Interestingly, the mutations in peptide variants V1, V2, V4, and V2o that were introduced to reduce immunogenicity resulted in even improved antiviral activity compared to that of the original C46-EHO peptide (Table 3). The optimized peptide V2o especially demonstrated superior entry-inhibitory properties, with IC50s in the picomolar range for most envelope glycoproteins tested (Fig. 4). Entry of VSV-G-pseudotyped virus particles was not affected by any of the C peptides, indicating HIV/SIV-specific entry inhibition.
Mutated peptide V2o has improved antiviral activity in single-round infection assays. PM1 cells were infected with replication-incompetent lentiviral vector particles (packaged with various HIV-1 or SIV envelope glycoproteins and encoding eGFP) in the presence of increasing concentrations of C peptides. eGFP-positive cells were determined by flow cytometry. (A) Data are means from duplicates of two independent experiments; error bars show SDs. (B) Data are means from 2 to 11 independent experiments; error bars show SDs.
Taken together, the HIV-1-derived C peptide C46 was able to prevent entry of pseudoviruses with HIV-1 envelope glycoproteins efficiently, while HIV-2EHO-derived peptides and variants thereof had superior antiviral activity not only against various HIV-1 strains (including T-20-resistant strains) but also against SIVmac251.
The deimmunized peptide V2o efficiently inhibits HIV-1 replication.In addition to the single-round infection experiments using replication-incompetent pseudoviruses, we performed neutralization assays using primary human CD4+ T cells infected with various strains of replication-competent HIV-1 in the presence of different concentrations of V2o. T-20 peptide was used as a control in these experiments. We found that V2o inhibited HIV-1 replication at least as efficiently as T-20 for all virus strains tested, including dual-tropic D117II, X4-tropic NL4-3, and the two R5-tropic strains Ba-L and Ba-L_C46r (Fig. 5).
The deimmunized C peptide V2o is an efficient HIV-1 entry inhibitor. Primary human CD4+ T cells were infected with different HIV-1 strains in the presence of increasing concentrations of C peptide V2o or T-20. Virus replication was analyzed by p24 ELISA at day 6 postinfection. Data are means from triplicate samples; error bars show SEMs. The dotted lines represent the detection limit of the p24 ELISA.
DISCUSSION
In the present work, we engineered a highly active anti-HIV peptide fusion inhibitor with greatly reduced immunogenicity. The novel peptide V2o was developed by identification and removal of antibody and MHC-I epitopes in the prototypic peptide C46-EHO, which is derived from HR2 sequences of HIV-2EHO and HIV-1HxB2. Using rational design, we have introduced amino acid substitutions into this C peptide to minimize antigenicity and immunogenicity, while preserving or even enhancing antiviral activity.
Over the past years, several C peptides with improved pharmaceutical properties, such as enhanced solubility, helical structure, six-helix bundle stability, and antiviral activity, have been engineered (9, 12, 13, 39). However, to our knowledge this is the first description of a deimmunized anti-HIV C peptide. Unwanted clinical immunogenicity can reduce the biotherapeutic activity of protein drugs and pose safety risks to patients. Epitope removal by amino acid substitution has been successfully used in the past to deimmunize therapeutic proteins, even though complete deimmunization of larger protein drugs probably remains out of reach. However, the intrinsic antigenicity and immunogenicity of recombinant therapeutic proteins such as erythropoietin (40), clotting factor VIII (41), and staphylokinase (43, 44) have been significantly reduced by identification and removal of a limited number of immunodominant epitopes.
Both MHC-I epitope prediction programs used in this study are known to predict many more epitopes than are relevant in vivo (23, 45, 46). Therefore, we concentrated on epitopes predicted by both algorithms for several HLA alleles and with a high score in order to engineer the deimmunized C peptide. It is well-known that the immunogenicity of peptides strongly correlates with the ability to induce a stable MHC-peptide complex (47). Amino acid substitutions to less preferred residues are expected to destabilize the MHC-peptide complex due to enhanced dissociation and/or decreased binding affinity of the modified peptide (47). Consequently, anchor amino acids of the predicted major MHC-I epitopes were exchanged for residues which fit less well into the binding pockets of the respective HLA molecules in order to prevent efficient presentation of the peptide to the immune system.
Although the cellular immunogenicity of the optimized peptide V2o could be significantly reduced compared to that of the parental peptide C46-EHO in silico, successful removal of the MHC-I epitopes could not be confirmed experimentally, as not even the parental peptide C46-EHO induced a strong CTL response in ELISpot assays with PBMCs from HIV-1-infected individuals. Nevertheless, we clearly demonstrated the feasibility of our strategy using two dominant viral class I MHC epitopes. Here, amino acid substitutions at the peptide anchor positions P-2 and P-9 led to dramatically reduced CTL activation in vitro, consistent with the in silico calculations.
The removal of antibody epitopes, however, had a striking effect on the seroreactivity of V2o. While about 10% of HIV-1-infected patients had preexisting antibodies to C46-EHO, the engineered peptide V2o was no longer recognized by any of the HIV-1-infected patient sera tested, making it a much safer protein drug. Even though it contains several amino acid exchanges, the V2o peptide sequence is still closely related to HIV-2 sequences. Consequently, all HIV-2-infected patients and most of the SIVmac251-infected rhesus macaques had preexisting antibodies to V2o.
Mapping and destruction of antibody epitopes are a sensible approach to deimmunize protein therapeutics. However, they will not necessarily prevent the generation of novel antidrug antibodies. The estimated human antibody repertoire diversity is in the range of 1010 to 1011 (48, 49), making it impossible to mutate all potential antibody epitopes in a therapeutic protein. Here, removal of MHC-II epitopes is a more reasonable strategy, as the diversity of HLA molecules comprises only several thousand alleles (51). However, bioinformatic calculation of MHC-II epitopes is not as straightforward as prediction of MHC-I epitopes, as peptides extending the 9-amino-acid core binding motif can also be presented by class II HLA molecules. As a result, currently available algorithms for prediction of MHC-II binding peptides perform much worse than those for MHC-I epitope prediction and are rather unreliable (52–54).
The optimized peptide V2o contains a substantial number of amino acid substitutions compared with the sequence of parental peptide C46-EHO. Nevertheless, using rational design, we were able not only to preserve its antiviral activity but also even to enhance its therapeutic potential: in particular, V2o is active against different HIV-1 strains (including T-20-resistant strains and strains with reduced sensitivity to C46) but also inhibits entry of SIVmac251 with an IC50 in the low-nanomolar range. The same efficacy profile was observed for the C46-EHO peptide, however, with 2- to 6-fold lower antiviral activity. In contrast, neither of the HIV-1-derived peptides T-20 (data not shown) and C46 prevented cell entry mediated by the SIVmac251 envelope glycoprotein at the highest concentration tested. Yet, efficient inhibition of SIV entry is also a highly desirable feature of viral fusion inhibitors, as this enables preclinical efficacy studies in the rhesus macaque model system.
The failure of HIV-1-derived C peptides to inhibit HIV-2 or SIV entry has previously been observed by others (55) and is due to mechanistic differences in the fusion process of HIV-1 and HIV-2/SIV. Gallo and colleagues (55) have shown that C peptides have a much smaller window of opportunity for binding their target structure during the HIV-2 and SIV fusion process compared with that with HIV-1 entry. Thus, inhibition of HIV-2 or SIV fusion by C peptides is generally much harder to achieve (55).
The C46-EHO and V2o peptides are based on the amino acid sequence of the HIV-2EHO strain, which is a highly pathogenic isolate (11, 56). It has previously been shown that the gp41 HR2 domain is responsible for enhanced fusion kinetics and the pathogenicity of a highly pathogenic HIV-1 envelope glycoprotein (58). Consequently, the C peptides derived from HIV-2EHO might bind the prehairpin structure more rapidly and thus have the chance to efficiently interfere even with six-helix bundle formation of SIV and potentially also HIV-2 glycoproteins. In addition, enhanced C-peptide binding could impede the generation of resistant virus strains.
In summary, we generated a highly and broadly active, deimmunized C peptide for therapy of HIV infection. Its superior antiviral activity along with the decreased immunogenicity makes V2o an attractive antiviral peptide for further clinical development.
ACKNOWLEDGMENTS
This study was sponsored by the project iGene funded by the German Federal Ministry for Research (BMBF). F.B. was supported by a scholarship from the Deutsche Forschungsgemeinschaft (DFG; GRK1172).
We thank Janine Kimpel, Catherine Dold, Felix G. Hermann, and Christoph Leder for support and critical discussions, Jan van Lunzen for providing HIV-1 patient sera and PBMC samples, as well as Ursula Dietrich and Schlomo Staszewski for providing HIV-1 and HIV-2 patient sera.
D.V.L. is inventor of membrane-anchored C peptides and participator in the biotech company Vision7 GmbH, which holds intellectual property on membrane-anchored anti-HIV C peptides. L.E. and D.V.L. are listed as inventors on a patent application related to secreted C peptides. For the other authors, no competing financial interests exist.
FOOTNOTES
- Received 1 June 2012.
- Returned for modification 27 June 2012.
- Accepted 16 October 2012.
- Accepted manuscript posted online 12 November 2012.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AAC.01152-12.
- Copyright © 2013, American Society for Microbiology. All Rights Reserved.