Development and immunological assessment of VLP-based immunogens exposing the membrane-proximal region of the HIV-1 gp41 protein

Software-assisted design of gp41 derivatives

An N- and C-terminally truncated gp41 (“gp41CTM”) was designed according to Lenz et al. [15]. Wild-type sequences for synthesis of viral gene fragments were obtained from GenBank and optimized for human codon usage (Life Technologies). Molecular clones of viral isolates HIV-1 89.6 (subtype B, accession number U39362), HIV-1 96ZM651 (subtype C, AF286224) and SIVmac239 (M33262) served as templates with aa 629–724 of 89.6 and 96ZM651 (HXB2 reference numbering) and aa 639–740 of SIVmac239 (SIVmac239 numbering). Membrane incorporation and orientation of all constructs was checked by the use of the transmembrane prediction TMHMM Server v. 2.0 [16]. A human tissue plasminogen activator (TPA) leader sequence was fused N-terminally. Correct cleavage of the leader sequence was predicted by SignalP 3.0 software [17]. A Kozak consensus sequence (GCCGCCACC) was inserted 5’ of the start codon for enhanced transcription. The sequence for an HA-tag from Influenza virus Hemagglutinin HA1 protein (YPYDVPDYA) was codon-optimized and inserted C-terminally by primer extension PCR for recognition of recombinant proteins (primers 3E8 & 2H4/2H5/2H6, see Additional file 1: Table S1 for primer sequences). The topology of all gene variants was checked with Phobius Prediction Server [18].

Trimer-stabilized constructs and molecular visualization

Two zipper domains were inserted between the TPA leader and gp41-derived sequences by fusion PCR. One motif was derived from the HA2 protein (amino acids 372–417, primers 3H4 & 3H5/3H6/3H7) of human Influenza virus strain H3 followed by a linker composed of amino acids Gly-Ser-Thr [13]. A second domain was derived from the GCN4 protein of Saccharomyces cerevisiae[19] (primers 3B6 & 3B7/3B8/3B9). Sequences for GCN4 and H3 zippers were derived from Genbank (IDs: CY002064, BK006939) and codon-optimized. To confirm the correct gene fusion position of zippers to gp41 fragments, molecular models were built using Modeller and evaluated by calculating the DOPE score (Discrete Optimized Protein Energy [20]. Resulting atomic models were visualized using PyMOL Molecular Graphics System.

SDS-PAGE, Western Blots and Slot Blots

SDS-PAGE was done with acrylamide concentrations of 12.5%. Western Blot transfers of proteins from SDS gels were done in a semi-dry system (Serva, Heidelberg, Germany) according to manufacturer’s instructions. For slot blots, protein solutions were loaded onto a Bio-Dot SF (Biorad) apparatus including 5 layers of pre-wetted Whatman Chromatography paper (Whatman International Ltd, Maidstone, UK). For both methods, nitrocellulose membranes with a pore size of 0.2 μm were applied (Millipore, Bedford, USA) and blocked in TBS (150 mM NaCl, 50 mM Tris/HCl, pH 7.4) containing 5% skim milk powder over night at 4°C. For antibody staining, blots were incubated for 1 h in TTBS (TBS + 0.3% (v/v) Tween-20) with primary antibody, washed four times for 15 min in TTBS, incubated for 1 h in TTBS with secondary antibody, washed again four times for 15 min in TTBS, and subjected to either ECL (2.5 mM luminol, 0.4 mM coumaric acid, 0.1 M TrisHCl pH 8.5 plus equal volume of 0.018% H₂O₂, 0.1 M TrisHCl pH 8.5) or alkaline phosphatase (AP) staining solution (5 mM TrisHCl pH 9.5, 5 mM NaCl, 2.5 mM MgCl₂ plus 1/50 volume of NBT/BCIP stock (Roche)). AP reactions were stopped with excess of dH₂O. ECL reactions were measured in a ChemiluxPro device (Intas, Göttingen, Germany). Intensity of bands was quantified with the aid of Gel-Pro Analyzer software (Media Cybernetics, Bethesda, USA). Human monoclonal antibodies (MAbs) 2F5 and 4E10 (diluted to 5 μg/ml each) served for detection of gp41 derivatives, the HA-tag was recognized by rat MAb 3F10 (“anti-HA High Affinity”, 0.1 μg/ml, Roche). Pr55/p24 was detected with mouse MAb M13/5 (cell culture supernatant, 1:500, [21]). Polyclonal HRP- or AP-coupled anti-human-IgG, anti-rat-IgG, and anti-mouse-IgG antibodies (all 1:2,000, all from Dako, Glostrup, Denmark) served as secondary antibodies.

ELISA

The amount of Pr55 or p24 protein in lysates was quantified with the aid of an Enzyme-linked Immunosorbent Assay (ELISA) using MAb M01 (1:1000, Polymun, Vienna, Austria) as coating antibody [22]. An ELISA with gp41-derived peptides was used to quantify anti-gp41 immunoglobulins in animal sera by the end-point dilution method in duplicates [23]. Peptides spanning the MPER (EQNEKDLLALDSWNNLWNWFDITKWLWYIK) and CHR regions (MQWDREISNYTNTIYRLLEDSQSQQEQNEK, both from Pepscan Presto BV, Lelystad, Netherlands) were used for coating at 100 ng/ml on Nunc Maxisorp plates (Thermo Fisher Scientific, Waltham, USA). HRP-coupled polyclonal anti-rabbit-IgG and anti-human-IgG served as secondary antibodies (1:4,000, Dako). Washing was done with the aid of a high-throughput microplate washing device (MAP-C2 workstation, Titertek Instruments Inc., Huntsville, USA).

Transient transfection

HEK 293 T (ATCC Microbiology Collections) were grown in Gibco Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal calf serum, 100 U/ml penicillin and 0.1 mg/ml streptomycin (Invitrogen). FreeStyle 293 F cells (Invitrogen) were grown in Gibco FreeStyle 293 expression medium (Invitrogen) supplemented with 50 U/ml penicillin and 0.05 mg/ml streptomycin at 125 rpm and handled according to the manufacturer’s instructions. HEK293T cells were seeded in 6-well plates (4x10⁵ cells in 2 ml medium). 24 h after seeding, medium was replaced by 1 ml of DMEM without supplements. Cells were transfected with the respective expression vector(s) according to the poly-ethylenimine (PEI) transfection method using a total of 2 μg DNA and 8 μl of a 1 mg/ml PEI in H₂O solution ad 50 μl DMEM without supplements. Components were mixed, left standing for 10 min at room temperature and added drop-wise to the cell suspension. 4 h later, transfection medium was replaced with 2 ml of supplemented growth medium. For smaller (96-well) or larger (10 cm/15 cm Petri dishes) assays, the number of cells, the amount of DNA and PEI, and the volume of medium was down- or up-scaled, based on the vessel surface area. 293 F cells were adjusted to 1 million cells/ml in fresh FreeStyle medium and 30 ml of cell suspension were transferred to 125 ml flasks. 37.5 μg of DNA and 150 μl of PEI ad 1.8 ml DMEM without supplements were mixed for transfection of one flask. 4 h later, transfection medium was replaced with 30 ml of FreeStyle medium supplemented with 100 U/ml penicillin and 0.1 mg/ml streptomycin.

Cytometry

Cells were detached 48 h p.tr from wells in PBE buffer (PBS + 1% FCS, 1 g/l NaN₃, 2 mM EDTA), pelleted at 300 g for 3 min, washed in PBE, incubated with the indicated fluorochrome-labeled antibodies (3 μg/ml 2F5 or 4E10, 5 μg/ml anti-HA-MAb 3F10) for 1 h at room temperature and washed 2 times with cold PBE. 2F5 and 4E10 antibodies were labeled with AlexaFluor-647 (Protein Labeling Kit, Invitrogen) according to the manufacturer’s instructions. A mean of 6 fluorochromes was attached to each IgG molecule, as measured by photometric analysis (NanoDrop software v3.7.1). anti-HA-antibody was purchased as conjugate with Fluorescein (Roche). Apoptosis staining was done by the use of 7-AAD and Annexin-V-APC antibody (both diluted 1:100, both BD Biosciences). Finally, cells were subjected to cytometric analysis in FACS Canto II (BD, with FACS Diva software v6.0).

Sucrose gradients

After purification through a sucrose cushion, VLPs were loaded onto a 10 ml 10-50% sucrose gradient and ultra-centrifuged at 100,000 g for 2.5 h. 20 fractions of 0.5 ml each were collected. For Western Blot analysis, particles were pelleted by trichloroacetic acid precipitation. A quarter of volume of 100% (w/v) TCA was added to the samples, incubated for 10 min at 4°C and spun at 14,000 g and 4°C for 5 min. The pellet was washed two times with cold acetone, dried at 95°C for 5 min and resuspended in SDS sample buffer. For EM experiments, fractions 13–17 were pooled, diluted in PBS ad 12 ml, ultra-centrifuged at 100,000 g for 1.5 h and resuspended in PBS.

Electron microscopy

Sucrose-cushion purified VLPs were incubated with primary antibody 4E10 for 1 h at 37°C and further purified on a sucrose gradient. Pooled and resuspended VLPs were adsorbed to grids after fixation with 2% glutaraldehyde. Grids were washed in TBS (Leica EM IGL), blocked with 3% gelatine in TBS for 1 h at RT, and incubated with an anti-human IgG immuno-gold conjugate (particle size: 10 nm; Aurion, Wageningen, Netherlands) for 1 h at room temperature. After three washes in TBS, grids were contrasted with phosphoric tungstic acid and examined in an electron microscope (Zeiss EM 10C/CR).

VLP production

VLPs for immunization purposes were produced in 293 F cells with the use of a codon-optimized, Rev-independent gene for Gag(IIIB) [24]. For pseudotyped VLPs, plasmids encoding for Gag and Env were mixed in a ratio of 2:1 in a co-transfection assay. VLPs were harvested 72 h p.tr., cleared by centrifugation at 3,000 g for 15 min, loaded onto a 30% sucrose in PBS cushion (5 ml for 30 ml of supernatant) and ultra-centrifuged at 100,000 g for 2 h. The pellet was resuspended overnight in PBS and stored at −80°C, with an aliquot analyzed by SDS gel electrophoresis and Coomassie staining for quality control. Both VLPs and DNA for immunization purposes were checked for low endotoxin levels with aid of Limulus Amebocyte Lysate QCL-1000 assay (Lonza Group Ltd, Basel, Switzerland), with a result of <20 EU/ml.

Rabbit immunizations

The animal study was carried out in strict accordance with the UK Animals (Scientific Procedure) Act 1986, and the protocol was approved by the local Ethical and Welfare Committee of the University of Cambridge and the UK Home Office (Project license no. 80/2238). All efforts were made to minimize suffering. New Zealand rabbits (Harlan UK Ltd, Belton, Leicestershire, UK) were kept under pathogen-free conditions and pre-bled by ear vein lancing at the age of 12 weeks. Animals were grouped into 4 groups of 6 animals each and immunized at weeks 0 and 4 with 500 μg of plasmid DNA by i.m. saline injection in quadriceps muscles, followed by i.m. booster immunizations at weeks 12 and 16 with 100 μg of VLPs (Gag protein, analyzed in Coomassie gel), adjuvanted with 1% Carbopol 974 adjuvant. The integrity of pseudotyped VLPs in Carbopol was checked by sucrose gradient ultracentrifugation and following Western Blot analysis. Rabbits were terminally bled by cardiac puncture under isoflurane anesthesia and then euthanized. Serum was obtained by incubation of blood for 1 h at 37°C, 1 h on ice, centrifugation at 20,000 g for 15 min, and complement inactivation by incubation of supernatant at 56°C for 30 min. Unfortunately 2 animals from group 2 and 3 and one animal from group 1 had to be euthanized prior to finalization of the immunization protocol for reasons not related to the study and were therefore not included in data analysis. Statistical analysis was performed based on the correct number of animals in each group.

Removal of anti-cell antibodies

HEK293F cells were stacked by centrifugation at 300 g for 3 min, supernatant was aspirated and cells were stored at 4°C for up to 5 days until usage. Stacked cells were dissolved in an equal volume of animal serum and shaken 3 times for 4 h at 800 rpm at room temperature, until the signal of anti-cell antibodies was below two times the background signal in cytometric analysis.

Neutralization assays

Virus neutralization assays were performed in the laboratory of D.C. Montefiori (Duke University School of Medicine, Durham, USA), according to the published protocol [25]. Assays were done with Env-pseudotyped viruses in TZM-bl cells for isolate MW965.26 and with Env.IMC.LucR viruses in A3R5 cells (CEM human lymphoblastoid cell origin) for isolate SF162. Dilutions mediating inhibition of infection by 50% (ID50) were subtracted by ID50 of pre-bleed sera or in any case subtracted by a minimum of 20 (detection limit).

Statistical analysis

Significance of anti-gp41 responses in ELISA and neutralizing activity in neutralization assays was determined using the Wilcoxon Two Sample Test. Each group was compared to its respective pre-bleeds to assess a potential neutralizing activity, and all groups were compared to each other to detect inter-group differences in neutralizing activity.

Design of gp41 constructs with enhanced MPER exposure

Based on published gp41 constructs (33), novel gp41 immunogens were designed based on the sequence of three molecular clones: 96ZM651 (HIV-1 clade C) was selected for antigen development, because anti-MPER bnAbs have been originally isolated in clade C-infected patients [26] and clade C strains represent the majority of global HIV infections. 89.6 (HIV-1 clade B) and SIVmac239 (“SIV”) derived antigens served as controls. A signal peptide from the human tissue plasminogen activator gene (TPA) was N-terminally fused to the gp41 moiety to ensure that the protein is efficiently directed towards the cell membrane in mammalian cells. A hemagglutinin (HA) tag of nine amino acids was added to the protein’s C-terminus for uniform detection of the corresponding proteins (Figure 1). As described earlier, the C-terminal tail partially loops back to the extracellular and extraviral space, enabling access to antibodies [27]. As it is unknown whether the CHRs adopt a trimeric or monomeric conformation in the pre-fusion state, we designed trimer-stabilized variants in addition to the non-stabilized structure. Trimerization is generally thought to be beneficial for proper display of immunologically relevant sites within both gp41 and gp120, which - at the same time - assists the occlusion of irrelevant sites believed to distract B cells from vulnerable sites. Two heterologous trimerization domains were introduced between the signal peptide and the gp41-derived part: GCN4 and H3 zipper domains have been used previously to stabilize a gp41 post-fusion [19] or intermediate structure [13],[28], respectively. For the H3 zipper domain, the orientation was adopted from an existing fusion protein by Hinz et al. [13], including a three amino acid spacer. Concerning the GCN4 domain, instead of fusing it to the NHR building the inner core of 6HB [6], it was fused N-terminally to the CHR, anticipating the stabilization of an exposed pre-hairpin conformation [28]. The correct fusion site was deduced by combining existing crystal structures of both the GCN4 domain and gp41. A molecular model of the ectodomain of a GCN4-gp41CTM fusion protein (comprising the GCN4 zipper domain and residues 1–50 of gp41 CHR) was built based on a solved structure of a GCN4-NHR fusion protein (1ENV, residues 1 to 80, [19]). Another three models were built with insertion of one, two, or three flexible aa (G, S, T) as linkers using the Modeller server, and the minimized energy was calculated by the aid of DOPE score (Figure 1C). The model without insertions showed the minimal energy and was chosen for cloning as it likely represents the structure with the highest stability. Its calculated energy (−29.7 kJ/mol) was close to that calculated for the modeling template GCN4-NHR of −30.8 kJ/mol. Two further crucial steps, the cleavage of the signal peptide and the membrane orientation, were controlled with appropriate prediction software. Proper signal peptide cleavage was predicted for all constructs. However, a wrong membrane orientation was predicted for the SIVmac239-based gp41CTM construct. For this reason, the cytoplasmic tail was elongated for 9 aa compared to 96ZM651- and 89.6-based templates. Correct membrane topology was predicted for all constructs after this modification (Figure 1D exemplarily for basal construct gp41CTM based on isolate 96ZM651).

Membrane exposure and selective recognition by bnMAbs

Expression and membrane location of gp41 variants is a prerequisite for incorporation into VLPs, display of relevant epitopes and usability in immunization studies. In a first experiment, in vivo properties of gp41CTM were evaluated in comparison to gp145, which is a codon-optimized version of gp160 with mutated cleavage site and truncated C-terminal tail which increases surface presentation [29]. The clade C molecular clone 96ZM651 was intended to serve as source for cloning the Env/gp41 derivatives. The clade B isolate 89.6 and SIVmac239 were chosen as additional controls for proper display, as they possessed either both or none of the epitopes for 2F5 and 4E10, respectively. HEK293T cells were transfected with expression plasmids encoding gp145 and gp41CTM and stained with the mentioned bnMAbs. gp41CTM exhibited a slightly, yet non-significantly, enhanced mean fluorescence intensity (MFI) in cytometric analysis compared to gp145 (Figure 2A). The staining with 2F5 and 4E10 showed a selective binding of 2F5 to 89.6-based constructs only, whereas 4E10 as expected bound to both 89.6 and 96ZM651-based constructs. Following staining of the HA-tag with an anti-HA-antibody, all variants showed proper membrane localization and similar expression levels (Figure 2B). For linking bnMAb binding with the level of Env expression, MFIs of 2F5 and 4E10 were normalized to the anti-HA intensity. The SIVmac239-based variants bound to none of the tested bnMAbs and were only detectable by the HA-tag, as expected. Signal intensities were similar for the basal and trimer-stabilized constructs, which indicated a similar accessibility of bnMAb epitopes in these constructs, irrespective of trimerization domains.

Incorporation of oligomeric gp41 derivatives into VLPs

Further analysis of gp41 variants was restricted to the constructs based on the molecular clone 96ZM651, as these were chosen for following immunization studies. As it is known that HIV-1-derived VLPs band at a density of 1.14-1.18 g/cm³ corresponding to ~35% glucose in PBS [30], the incorporation of novel gp41 immunogens into VLPs was addressed by sucrose gradients and Western Blot analysis. VLPs were produced by co-transfection of HEK293T cells with plasmids encoding Gag and gp41CTM by RNA- and codon-optimized genes, respectively. Gag as well as gp41 derivatives that were incorporated into VLPs prior to or during budding were detected in Western Blot analysis using polypeptide specific antibodies. All variants showed a co-banding of both proteins in fractions 12–14, suggesting successful incorporation into VLPs (Figure 3A). Higher molecular weight bands for GCN4-gp41CTM and H3-gp41CTM indicated the formation of stable dimers and trimers. For a relative quantification of incorporated gp41 proteins, VLPs were purified, normalized to Gag content, lysed and loaded onto a slot blot, followed by detection with 4E10 (Figure 3B). All gp41 variants exhibited similar signal intensities and are therefore supposed to be incorporated at similar amounts into VLPs. Since the recognition by 4E10 of a similar soluble standard protein (GCN4-gp41CTM-FD) was substantially lower than that of gp41CTM and its derivatives (data not shown), it was not possible to determine the absolute number of gp41CTM molecules on the surface of VLPs. The morphology of VLPs pseudotyped with gp41 derivatives and the antigenic properties of incorporated proteins were further evaluated by electron microscopy. VLPs were labeled with 4E10 and secondary immuno-gold antibody. gp41 derivatives had no apparent impact on particle morphology and size of 100–150 nm in diameter (Figure 3C). Staining with 4E10 showed specific labeling of VLPs pseudotyped with all three types of gp41 derivatives, whereas non-pseudotyped VLPs were not labeled.

Induction of anti-gp41 antibodies

The immunogens were evaluated for immunogenicity in a rabbit immunization study. A DNA prime and VLP boost regimen was chosen, because heterologous immunization schedules using other Env constructs had induced higher antibody titers compared to DNA or VLPs alone [31]. After attesting low endotoxin levels in the immunogens formulation, VLPs were formulated with 1% 974 carbopol as adjuvant. Carbopol was chosen because of its optimal formulation properties with lipid membranes, compared to Freund’s adjuvant or oil in water formulations. The integrity of VLPs and incorporated proteins after mixing with Carbopol was verified by sucrose gradient and Western Blot analysis (data not shown). Rabbits were first primed twice with 500 μg of DNA at weeks 0 and 4, then boosted two times with 100 μg of VLPs formulated in Carbopol, at weeks 12 and 16 (Figure 4A). Bleeds were taken 2 weeks before the first immunization and 2 weeks after the last immunization. The induction of anti-gp41 Abs is a prerequisite for effective neutralization, and was analyzed by IgG ELISA. Reactivity of sera was dissected using an MPER (Figure 4C) and a CHR peptide immobilized to the solid phase (Figure 4D). Whereas animals that received gp145 (positive control) developed significant titers against both peptides, only low titers were obtained for the groups that were immunized with gp41 derivatives. Immune responses in non-reacting animals might have been too low to be detected in unfractionated sera.

Neutralizing capacity of unfractionated immune sera

Immunization of rabbits with VLPs derived from heterologous production systems (HEK293T cells) can induce antibodies directed against HEK293T cell derived membrane proteins, and these antibodies interfere with neutralization assays. As anti-cell antibodies were detected in some immunized animals (Figure 5A), sera were incubated with HEK293T cells three times to absorb and deplete anti-cell antibodies [32]. Antibodies directed against a 30-mer MPER-peptide, that were present in animal sera, were not detectably affected by this procedure (Figure 5B, upper panel), while anti-cell antibodies were successfully removed (Figure 5A). As a control, 2F5 and 4E10 recombinant antibodies were incubated with HEK293 cells as well, and were depleted partially in this assay. 4E10 was depleted more pronouncedly than 2F5, as 4E10 has a higher membrane-binding activity than 2F5 (Figure 5B, lower panel). Subsequently, all sera were tested in standardized assays for neutralizing activity in a TZM-bl assay. A tier-1 molecular clone of clade C (MW965.26) and a tier-1 clone of clade B (SF162) were chosen for readout. Clade C isolate MW965.26 contains the 4E10 epitope but not the 2F5 epitope, like 96ZM651, whereas clade B isolate SF162 contains the epitopes for both 4E10 and 2F5, like isolate 89.6 [33]. In general, results from neutralization tests tended to reflect titers of anti-MPER-peptide antibodies as measured in ELISA. Effective neutralization was observed only for the rabbit group receiving gp145 DNA and homologous VLPs (VLP-gp145), with two rabbits showing neutralization at a 50% inhibitory dilution (ID50) of >1,000 against MW965.26 (Figure 6A). For all three rabbit groups receiving gp41-derived DNA prime and VLP boost, neutralizing activities were weak. However, statistically significant neutralization at low level was observed for group 1 against MW965.26 and group 2 against SF162. An ID50 of up to ~100 could be detected in several further animals against both isolates. Interestingly, neutralizing activity of sera of group 1–3 after gp41 immunization appeared to be more pronounced against the clade B isolate than against the clade C isolate, indicating better cross-clade neutralizing activity than obtained with group 4 after gp145 immunization.