Single-Chain Soluble BG505.SOSIP gp140 Trimers as Structural and Antigenic Mimics of Mature Closed HIV-1 Env

ABSTRACT

Similar to other type I fusion machines, the HIV-1 envelope glycoprotein (Env) requires proteolytic activation; specifically, cleavage of a gp160 precursor into gp120 and gp41 subunits creates an N-terminal gp41 fusion peptide and permits folding from an immature uncleaved state to a mature closed state. While the atomic-level consequences of cleavage for HIV-1 Env are still being determined, the uncleaved state is antigenically distinct from the mature closed state, and cleavage has been reported to be essential for mimicry of the mature viral spike by soluble versions of Env. Here we report the redesign of a current state-of-the-art soluble Env mimic, BG505.SOSIP, to make it cleavage independent. Specifically, we replaced the furin cleavage site between gp120 and gp41 with Gly-Ser linkers of various lengths. The resultant linked gp120-gp41 constructs, termed single-chain gp140 (sc-gp140), exhibited different levels of structural and antigenic mimicry of the parent cleaved BG505.SOSIP. When constructs were subjected to negative selection to remove subspecies recognized by poorly neutralizing antibodies, trimers of high antigenic mimicry of BG505.SOSIP could be obtained; negative-stain electron microscopy indicated these to resemble the mature closed state. Higher proportions of BG505.SOSIP-trimer mimicry were observed in sc-gp140s with linkers of 6 or more residues, with a linker length of 15 residues exhibiting especially promising traits. Overall, flexible linkages between gp120 and gp41 in BG505.SOSIP can thus substitute for cleavage, and sc-gp140s that closely mimicked the vaccine-preferred mature closed state of Env could be obtained.

IMPORTANCE The trimeric HIV-1 envelope glycoprotein (Env) is the sole target of virus-directed neutralizing antibody responses and a primary focus of vaccine design. Soluble mimics of Env have proven challenging to obtain and have been thought to require proteolytic cleavage into two-component subunits, gp120 and gp41, to achieve structural and antigenic mimicry of mature Env spikes on virions. Here we show that replacement of the cleavage site between gp120 and gp41 in a lead soluble gp140 construct, BG505.SOSIP, with flexible linkers can result in molecules that do not require cleavage to fold efficiently into the mature closed state. Our results provide insights into the impact of cleavage on HIV-1 Env folding. In some contexts such as genetic immunization, optimized cleavage-independent soluble gp140 constructs may have utility over the parental BG505.SOSIP, as they would not require furin cleavage to achieve mimicry of mature Env spikes on virions.

INTRODUCTION

Efforts to design an effective vaccine against HIV-1 have so far met with limited success (1, 2). With the discovery and characterization of a multitude of effective antibodies that are capable of neutralizing HIV-1 (3–13) and that have shown substantial promise for immunotherapy and protection (14–17), interest has focused on antibody-based vaccines (18–20). Vaccine strategies have been based on different components or subunits of the Env glycoprotein, which is found on the surface of HIV-1 virions and is the target of broadly neutralizing antibody responses (21–27). Env is a trimer of heterodimers, with each heterodimer consisting of a gp120 molecule and a gp41 molecule. Like other type I fusion proteins, Env requires proteolytic cleavage (specifically at the gp120-gp41 junction) to allow movement of the fusion peptide and possibly to induce rearrangements of the structure of the protein that can allow for interactions with host receptors and virus-host membrane fusion (28). The degree of structural rearrangements varies for different type I fusion proteins, ranging from, for example, rearrangements localized to the region around the cleavage site in the case of influenza virus hemagglutinin (28–30) to major overall structural changes in the case of the fusion glycoprotein of respiratory syncytial virus (31, 32). The precise structural effects of cleavage are unclear in the case of HIV-1 Env; however, it has been shown that uncleaved Env binds to both poorly and broadly neutralizing antibodies, whereas fully cleaved Env preferentially binds to broadly neutralizing antibodies (33, 34). Antigenicity profiling is thus often used for evaluation of native spike mimicry by Env-derived constructs in HIV-1 vaccine design (35, 36).

In addition to changes resulting from gp120-gp41 cleavage, the mature HIV-1 Env undergoes a number of conformational and large-scale structural changes upon interaction with its host primary receptor and coreceptor and in transitioning from prefusion to postfusion states (37–39). Since the prefusion “closed” conformation of mature Env, observed before receptor interactions, exposes neutralizing but hides nonneutralizing antibody epitopes, it is a primary target in current vaccine design efforts. Trimeric Env-based immunogens are of special interest due to their potential ability to display antibody epitopes in a structure similar to that observed in functional Env on virions, without exposing additional nonneutralizing decoy epitopes (36, 40). Soluble gp140 molecules in particular have seen a recent surge in interest, particularly with advances in our understanding of Env structure at the atomic level that have enabled rational structure-based immunogen design (41–43). Designing soluble gp140s that can act as structural and antigenic mimics of the closed state of mature prefusion Env, however, has proven difficult. The current best soluble gp140 molecule, named BG505.SOSIP, is a derivative of the clade A HIV-1 strain BG505, with a number of stabilizing mutations that allow for proper structural and antigenic mimicry of the closed state of mature prefusion Env (35, 36, 44, 45). Specifically, BG505.SOSIP is truncated at residue 664 in gp41 and includes the trimer-stabilizing gp120-gp41 disulfide bridge between residues 501 and 605 (termed SOS) and an Ile-to-Pro mutation at gp41 residue 559 (termed IP), as well as a Thr-to-Asn mutation at residue 332 to introduce a glycosylation site within the epitope for broadly neutralizing antibody PGT122 and a modification of the native ₅₀₈REKR₅₁₁ furin cleavage site to six Arg residues for improved cleavage (36). When the cleavage site is mutated to the inactive ₅₀₈SEKS₅₁₁ without the incorporation of the SOSIP mutations, the resulting construct is a poor mimic of the closed state of mature prefusion Env (35, 45). While the addition of the SOSIP mutations partially restores the antigenic and structural properties of the ₅₀₈SEKS₅₁₁ molecule, only the cleaved BG505.SOSIP appears to be fully able to mimic the closed state of mature prefusion Env, with the major differences observed in the proportions of aberrant structures for cleaved versus uncleaved BG505.SOSIP (35, 45).

In contrast to cleavage-dependent constructs such as BG505.SOSIP, cleavage-independent soluble gp140 constructs may present advantages for genetic immunization and antigen multimerization on self-assembling particles (e.g., ferritin [46]) where access to or by furin may be more limited, as well as for engineering soluble gp140 variants in strains where cleavage efficiency may be different from that of BG505.SOSIP. To that end, we set out to design cleavage-independent soluble gp140 constructs capable of folding into the prefusion closed conformation without proteolytic activation. As type I fusion machines store the energy of folding into their prefusion states, to be released upon transition to their low-energy postfusion conformations, it was theoretically unclear if we could achieve stable folding into the mature prefusion state. Nevertheless, with the 3.5-Å structure of the prefusion closed conformation of BG505.SOSIP providing atomic-level information (41), we decided to use BG505.SOSIP as a design platform. Specifically, we sought to replace the furin cleavage site between gp120 and gp41 with flexible peptide linkers and to generate single-chain gp140 (sc-gp140) variants that mimicked cleaved BG505.SOSIP, both structurally and antigenically.

MATERIALS AND METHODS

Design of cleavage-independent gp140 constructs.

The BG505.SOSIP sequence was used as a backbone for design (Fig. 1). The following two design strategies were used: (i) the four-residue furin cleavage site (gp140 residues 508 to 511) as well as surrounding residue segments (specifically, residues 504 to 518, 504 to 521, 505 to 518, and 505 to 521) in BG505.SOSIP were replaced by 1 to 4 repeats of the flexible GGSGG linker (for total linker lengths of 5, 10, 15, and 20) (Table 1), and (ii) only the cleavage site (residues 508 to 511) was replaced by flexible peptide linkers consisting of combinations of Gly and Ser residues, with linker lengths ranging from one to 20 residues, in the context of BG505 with the SOSIP mutations, with only the IP mutation, and with neither the SOS nor IP mutations (Table 2).

FIG 1.

Design of single-chain gp140 constructs. (A) Single-chain gp140 constructs that retain a linkage between gp120 and gp41 in BG505.SOSIP (top) were designed by replacing either both the furin cleavage site (residues 508 to 511) and adjacent residue segments (middle, orange) or only the cleavage site (sc-gp140) (bottom, red) by a flexible peptide linker with variable length. Highlighted are the SOSIP mutations (SOS, a disulfide between residues 501 and 605; IP, an Ile-to-Pro mutation at residue 559) as well as a designed glycosylation site at residue 332 and the truncation of gp140 at residue 664. (B) Mapping of the sc-gp140 design concept onto a BG505.SOSIP structure (PDB ID 4TVP). The three protomers are shown in different shades of gray, with two of the protomers shown as transparent surface and one protomer shown as cartoon ribbon. The residues between 505 and 518 are not observed in the structure and are marked as dotted lines, while the designed flexible linker replacing residues 508 to 511 is shown in red.

TABLE 1.

Qualitative ELISA binding results for gp140 constructs using flexible linkers to replace extended gp140 segments encompassing the native furin cleavage site

Replaced gp140 segment	Bindinga with replacement linker:
	GGSGG		GGSGG × 2		GGSGG × 3		GGSGG × 4
	VRC26.09	F105	VRC26.09	F105	VRC26.09	F105	VRC26.09	F105
504–518	−	+	+	+	+	+	+	+
504–521	−	+	−	+	+	+	+	+
505–518	−	+	+	+	+	+	+	+
505–521	−	++	+	+	+	+	+	+

^aFor each combination of replaced gp140 segment and replacement linker, shown is the measured binding to broadly neutralizing antibody VRC26.09 and poorly neutralizing antibody F105. Different gp140 segments encompassing the cleavage site were replaced by flexible linkers composed of 1 to 4 repeats of the GGSGG sequence. Binding data for each construct are shown relative to binding for BG505.SOSIP. VRC26.09: ++, at most 2-fold lower than BG505.SOSIP; +, between 2- and 4-fold lower than BG505.SOSIP; and − more than 4-fold lower than BG505.SOSIP. F105: −, at most 2-fold higher than BG505.SOSIP; +, between 2- and 4-fold higher than BG505.SOSIP; ++, more than 4-fold higher than BG505.SOSIP.

TABLE 2.

Flexible linkers used for the different BG505 sc-gp140 constructs for replacing the native furin cleavage site (residues 508 to 511)

Linker length	Sequencea
	Non-SOSIP	IP	SOSIP
1	G	G	G
2	GG	GG	GG
3	GSG	GSG	GSG
4	GGSG	GGSG	GSGG
5	GSGSG	GSGSG	GGSGG
6	GSGGSG	GSGGSG	GSGGSG
7	GGSGGSG	GGSGGSG	GGSGGSG
8	GGSGGGSG	GGSGGGSG
9	GGSGGSGGG	GGSGGSGGG
10	GGSGGGGSGG	GGSGGGGSGG	GGSGGGGSGG
11	GGSGGGSGGSG	GGSGGGSGGSG
12	GGSGGGSGGGSG	GGSGGGSGGGSG
13	GGSGGGGSGGGSG	GGSGGGGSGGGSG
14	GGSGGGGSGGSGGG	GGSGGGGSGGSGGG
15	GGSGGGGSGGGGSGG	GGSGGGGSGGGGSGG	GGSGGGGSGGGGSGG
20	GGSGGGGSGGGGSGGGGSGG	GGSGGGGSGGGGSGGGGSGG	GGSGGGGSGGGGSGGGGSGG

^aFor each linker length, shown are the Gly-Ser combinations used for each of the non-SOSIP, IP-only, and SOSIP cases. Empty fields in the SOSIP column correspond to linker lengths that were not tested.

High-throughput ELISA screen.

Ninety-six-well microplate-formatted transient-transfection expression and enzyme-linked immunosorbent assay (ELISA) binding assay approaches were used for high-throughput screening of various immunogen proteins (32), as briefly described here. Twenty-four hours prior to DNA transient transfection, 100 μl of HEK GnTI^−/− cells was seeded in each well of a 96-well microplate at a density of 2.5 × 10⁵ cells/ml. Two hours prior to transfection, 100 μl of spent medium from each well was replaced with 60 μl of fresh expression medium. DNA-TrueFect-Max complexes were prepared by mixing 0.2 μg plasmid DNA in 10 μl of Opti-MEM transfection medium (Invitrogen, CA) with 0.5 μl of TrueFect-Max (United BioSystems, Herndon, VA) in 10 μl of Opti-MEM and incubating for 15 min prior to transfection. The 96-well plate was incubated at 37°C with 5% CO₂. One day posttransfection, 20 μl of enriched medium (Protein Expression Booster for adherent cells) (ABI, Sterling, VA) was added to each well. After day five posttransfection, the expressed protein in the supernatant in 96-well microplate was characterized for anti-HIV antibody binding within a D7324-coated 96-well ELISA plate. Experiments with all samples were performed in duplicate.

Large-scale expression and purification of BG505.SOSIP sc-gp140 constructs.

Soluble BG505.SOSIP sc-gp140 proteins (containing either a D7324 or a histidine tag) were expressed by transient transfection in GnTI^−/− cells using 293fectin (Invitrogen). The culture supernatant was harvested at 5 days posttransfection and centrifuged at 8,000 × g for 45 min to remove cell debris. The culture supernatants were sterile filtered prior to protein purification. BG505.SOSIP sc-gp140 proteins were purified using lectin (EY Laboratories, San Mateo, CA) affinity chromatography at 4°C, and proteins were eluted using 1 M methyl-α-d-manno-pyranoside–phosphate-buffered saline (PBS), pH 7.4. The eluates were concentrated immediately and subjected to two rounds of size exclusion chromatography (SEC). Relevant fractions corresponding to the trimeric form of the protein were pooled, concentrated, and purified further in batch mode using an F105 affinity column (47).

Differential scanning calorimetry (DSC).

The heat capacity of wild-type and selected sc-gp140 BG505.SOSIP constructs was measured as a function of temperature using a differential scanning VP-DSC microcalorimeter (GE Healthcare/Microcal, Northampton, MA). Protein samples were extensively dialyzed against PBS (pH 7.5) and then degassed to avoid the formation of bubbles in the calorimetric cells. Thermal denaturation scans were conducted from 10 to 100°C at a rate of 1°C/min. The protein concentration was 0.10 mg/ml in all experiments.

Negative-stain electron microscopy (EM).

Samples were diluted to ∼0.03 mg/ml, adsorbed to a freshly glow-discharged carbon-film grid for 15s, and stained with 0.7% uranyl formate. Images were collected semiautomatically using SerialEM (48) on an FEI Tecnai T20 with a 2k × 2k Eagle charge-coupled device (CCD) camera at a pixel size of 0.22 nm/px. Particles were picked automatically, and reference-free two-dimensional (2D) classification was performed in EMAN2 (49). Semiquantitative determination of aberrant conformations was performed by manual counting of unprocessed images using a set of criteria to distinguish trimers from aberrant structures (see Fig. 3A). Between 2 and 5 images were counted per construct.

FIG 3.

Negative-stain EM of BG505 sc-gp140 constructs. (A) Example of counting trimer structures in EM images. Representative images (e.g., left) were classified (right) according to the presence of trimer versus aberrant structures. The criteria for identifying an aberrant structure were that (i) the negative stain has to be strong enough to clearly identify it as a particle, (ii) it should not have 3-fold symmetry, and (iii) it has to be larger than the native trimer in at least one dimension. We note that whereas the top views of trimer structures are structurally well defined, a number of factors such as staining blotches, side views of native trimers, etc., could cause a structure to be inadvertently categorized as aberrant. Trimer structures mimicking closed mature Env are shown in small blue circles, whereas aberrant structures are shown in larger white circles. Large black circles show examples of particles that were not classified as native trimers but were not counted as aberrant structures because they did not fulfill all of the criteria. (B) EM analysis was performed on proteins before and after F105 negative selection (see Materials and Methods). Shown are images for three of the sc-gp140 constructs (with 1-, 6-, and 15-residue linkers), BG505.SOSIP, and gp140-foldon after F105 negative selection. Representative structures are shown as magnified insets. The scale bar represents 50 nm in the main panels. (C) Classification of negative-stain EM for selected sc-gp140 constructs. Structures were analyzed before and after F105 negative selection (F105-selection). Constructs were categorized according to the fraction of “good” trimer structures observed: −, <10%; +, 10 to 50%; ++, 50 to 70%; +++, 70 to 85%; n/a, experiment not performed.

Antibodies for antigenic characterization.

Different subsets of a panel of antibodies were used for the antigenic characterization of the sc-gp140 and cleaved BG505.SOSIP constructs in the different binding assays performed in this study. The panel included the broadly neutralizing antibodies VRC26.09, VRC01, 35O22, 8ANC195, PGT145, PGT151, 2G12, b12, PGT121, and PGT128 and the poorly neutralizing antibodies F105, 17b, 447-52D, and 48D (3, 4, 6, 7, 9, 13, 50–56).

Antigenic characterization of BG505.SOSIP sc-gp140 constructs using biolayer interferometry.

A fortéBio Octet Red384 instrument was used to measure binding of BG505.SOSIP sc-gp140 and YU2 full-length gp120 and YU2 foldon-stabilized gp140 (gp140-foldon) (63) proteins to a panel of antibodies. All assays were performed with agitation set to 1,000 rpm in PBS (pH 7.4) supplemented with 1% bovine serum albumin (BSA) in order to minimize nonspecific interactions. The final volume for all the solutions was 50 μl/well. Human antibodies (50 μg/ml) in PBS were used to load anti-human IgG Fc (AHC) biosensors for 300 s; typical capture levels were ∼1.5 nm. Biosensor tips loaded with human antibodies were equilibrated for 180 s in 1% BSA–PBS (pH 7.4), followed by binding with 250 nM BG505.SOSIP sc-gp140 and BG505.SOSIP proteins in 1% BSA–PBS for 300 s; interacting proteins were allowed to dissociate for 300s in 1% BSA–PBS. In the case of antibody 8ANC195, the assay was performed with an association time of 600s. Baseline drift was corrected by subtracting the measurements recorded for a sensor with monoclonal antibody incubated in 1% BSA–PBS. Kinetics measurements to determine binding constants were conducted with VRC26.09 IgG (50 μg/ml) and BG505.SOSIP sc-gp140 and BG505.SOSIP proteins in the 500 to 1 nM range. An assay similar to that described above was used, except interacting proteins were allowed to dissociate for 6,000 s in 1% BSA–PBS. Data analyses were carried out using Octet software, version 8.0, and displayed using GraphPad Prism.

Antigenic characterization of BG505.SOSIP sc-gp140 constructs using electrochemical luminescent immunoassay (ECLIA).

Standard 96-well bare Multi-Array Meso Scale Discovery (MSD) plates (MSD; catalog number L15XA-3) were coated with a panel of HIV-1 antibodies in duplicates (30 μl/well) at a concentration of 10 μg/ml and diluted in 1× PBS, and the plates were incubated overnight at 4°C. The following day, plates were washed (wash buffer, 0.05% Tween 20 plus 1× PBS), blocked with 150 μl of blocking buffer (5% [wt/vol] MSD blocker A [MSD; catalog number R93BA-4]), and incubated for 1 h on a vibrational shaker (Heidolph Titramax 100; catalog number P/N:544-11200-00) at 650 rpm. All the incubations except the coating step were performed at room temperature. During the incubation, BG505.SOSIP trimer was titrated down in serial 2-fold dilutions starting at 4 μg/ml of the trimer in assay diluent (1% [wt/vol] MSD blocker A plus 0.05% Tween 20). After the incubation with blocking buffer was complete, the plates were washed and the diluted trimer was transferred (25 μl/well) to the MSD plates and incubated for 2 h on the vibrational shaker at 650 rpm. After the 2-h incubation with trimer, the plates were washed again and secondary detection MSD Sulfotag-labeled 2G12 antibody (prior to running the assay, 2G12 was labeled with MSD Sulfotag [MSD; catalog number R91AN-1] at a conjugation ratio of 1:15 [2G12/Sulfotag]), which was diluted in assay diluent at 2 μg/ml and was added to the plates (25 μl/well) and incubated for 1 h on the vibrational shaker at 650 rpm. The plates were washed and read using the 1× read buffer (MSD read buffer T [4×]; catalog number R92TC-2) on an MSD Sector Imager 2400.

Antigenic characterization of BG505.SOSIP sc-gp140 constructs using ELISA.

Ninety-six-well MaxiSorp plates (Thermo Fisher Scientific) were coated overnight at 4°C with 100 μl/well of snowdrop lectin from Galanthus nivalis (Sigma-Aldrich) at 2 μg/ml, diluted in 1× PBS. The plates were then blocked at room temperature for 1 h using 200 μl/well of 5% skim milk–1.5% bovine serum albumin (BSA) in 0.05% Tween 20 plus 1× PBS. After the plates were washed (wash buffer, 0.05% Tween 20 plus 1× PBS), they were incubated for 2 h at room temperature with BG505.SOSIP constructs at 2 μg/ml, diluted in 10% fetal bovine serum (FBS)–1× PBS. During this incubation step, HIV-1 antibodies were 5-fold serially diluted in 0.2% Tween 20 plus 1× PBS starting at 5 μg/ml. Upon completion of trimer incubation, the plates were washed, and the antibody dilutions were transferred to the plate and incubated for 1 h. Following another washing step, the plates were then incubated for another hour with horseradish peroxidase (HRP)-conjugated anti-human IgG (1:5000) diluted in 0.2% Tween 20 plus 1× PBS. After a final wash, the plates were then developed using SureBlue TMB peroxidase substrate (KPL) for 10 min. The reaction was stopped with 1 N H₂SO₄, and then the absorbance was measured at 450 nm. All incubations were at 100 μl/well at room temperature, except where noted otherwise.

Structure modeling.

The BG505.SOSIP sc-gp140 construct with a one-residue Gly linker replacing residues 508 to 511 (i.e., connecting residues 507 and 512) was modeled using the BG505.SOSIP crystal structure (PDB ID, 4TVP). Potential loop conformations for the segment encompassing residues 502 to 521 were generated using the default parameters in Loopy (57), with the rest of the trimer structure remaining in a fixed conformation.

RESULTS

Soluble gp140s with substituted residue segments encompassing the native gp120-gp41 cleavage site.

The precise cleavage-induced structural changes in Env are currently not known; however, high-resolution structures of cleaved BG505.SOSIP have revealed the approximate positions of the gp120 C terminus and gp41 N terminus after cleavage (41–43). The structure of BG505.SOSIP captured in the mature closed state by antibodies PGT122 and 35O22 reveals the location of residues 505 (of gp120) and 518 (of gp41) to be separated by a C_α-C_α distance of ∼37 Å. While it is structurally possible for a fully extended peptide to connect these residues, it was not known if this could occur, as the intervening residues of the cleaved soluble structure were not sufficiently ordered to be visualized in the electron density. To provide an initial assessment of the importance of the connecting residues, we evaluated BG505.SOSIP gp140 constructs where flexible linkers of 1 to 4 repeats of GGSGG were used to substitute for the cleavage site and its neighboring residues (Fig. 1). Thus, these linked (or single-chain) gp140 constructs retain an uninterrupted peptide link between gp120 and gp41, similar to uncleaved gp140 molecules. Unlike uncleaved gp140, however, these single-chain gp140 constructs aimed to achieve a structure similar to the mature closed prefusion state observed for the cleaved molecules. For these single-chain gp140 constructs, however, substantial levels of binding to the poorly neutralizing antibody F105 were observed, suggesting that the residues surrounding the gp120-gp41 cleavage site may be important for stability in the context of BG505.SOSIP (Table 1).

Screening of linker insertions to replace the gp120-gp41 cleavage site.

We next elected to keep the residues surrounding the gp120-gp41 cleavage site but to replace only the site itself with flexible linkers of various lengths (Fig. 1 and Table 2). We named these single-chain constructs sc-gp140 and evaluated them in the context of BG505 with the SOSIP mutations, with only the IP mutation, or with neither the SOS or IP mutations. To identify promising candidates for further evaluation, we initially performed qualitative ELISAs (see Materials and Methods). Specifically, we screened for binding to the trimer-specific broadly neutralizing antibody VRC26.09 as well as the poorly neutralizing antibody F105 and compared the various sc-gp140 constructs to cleaved BG505.SOSIP (Table 3). Non-SOSIP sc-gp140 constructs performed the worst, with more than 2-fold-higher binding to F105 than cleaved BG505.SOSIP, although some residual VRC26.09 binding was also observed for most linker lengths. The majority of the IP and SOSIP sc-gp140 constructs generally showed little difference in their VRC26.09 and F105 binding profiles compared to cleaved BG505.SOSIP. In particular, for the BG505.SOSIP sc-gp140 constructs, VRC26.09 binding was similar to or better than, while F105 binding was similar to or at most 2-fold worse than, that of cleaved BG505.SOSIP, with longer linkers exhibiting generally better binding profiles (Table 3).

TABLE 3.

Qualitative ELISA data for binding of BG505 sc-gp140 constructs (non-SOSIP, IP, and SOSIP) to broadly neutralizing antibody VRC26.09 and poorly neutralizing antibody F105

sc-gp140 construct	Linker length	Bindinga to:
		VRC26.09	F105
BG505.non-SOSIP	1	−	+
	2	−	+
	3	+	++
	4	+	+
	5	+	+
	6	++	+
	7	++	+
	8	+	+
	9	++	+
	10	++	+
	11	++	+
	13	++	+
	12	++	+
	14	+	+
	15	+	+
	20	+	−
BG505.IP	1	++	+
	2	++	−
	3	++	+
	4	++	+
	5	++	−
	6	+++	−
	7	+++	−
	8	+++	+
	9	+++	+
	10	++	−
	11	+++	−
	12	+++	−
	13	+++	−
	14	+++	−
	15	+++	−
	20	++	−
BG505.SOSIP	1	++	−
	2	++	−
	3	++	−
	4	+++	−
	5	+++	−
	6	+++	−
	7	+++	−
	10	+++	−
	15	+++	−
	20	+++	−

^aBinding data for each sc-gp140 are shown relative to that for BG505.SOSIP. VRC26.09: +++, better than BG505.SOSIP; ++, at most 2-fold lower than BG505.SOSIP; +, between 2- and 4-fold lower than BG505.SOSIP; −, more than 4-fold lower than BG505.SOSIP. F105: −, at most 2-fold higher than BG505.SOSIP; +, between 2- and 4-fold higher than BG505.SOSIP; ++, more than 4-fold higher than BG505.SOSIP.

Purification and oligomeric characterization of select sc-gp140s.

Based on the initial qualitative screening, we selected specific BG505.SOSIP sc-gp140 variants for further characterization. Constructs with short (1- and 2-residue), medium (6- and 7-residue), and long (10- and 15-residue) linkers were expressed in GnTI^−/− cells and purified over a lectin column (see Materials and Methods). Initial gel filtration profiles after lectin purification were heterogeneous (Fig. 2A), possibly due to promiscuous reactivity by lectin; however, a second round of purification of the putative trimer fractions for each sc-gp140 construct and BG505.SOSIP resulted in single peaks that were selected for further characterization (Fig. 2A). Protein yields after the two rounds of size exclusion chromatography for most constructs were within 2-fold of the yield for cleaved BG505.SOSIP using the same purification scheme (Table 4). Additionally, F105 negative selection was performed for each of the constructs to remove residual F105 binding (see Materials and Methods). Further characterization, including DSC, negative-stain EM, and antigenicity, of the sc-gp140 constructs and BG505.SOSIP was based on the protein fractions after F105 negative selection. As expected, the sc-gp140 constructs showed no difference on SDS-PAGE under reducing or nonreducing conditions, indicating lack of gp120-gp41 cleavage, unlike the BG505.SOSIP control (Fig. 2B). DSC measurements indicated increased stability for the sc-gp140 constructs: compared to BG505.SOSIP, the sc-gp140 constructs with linker lengths of 6, 10, and 15 residues had substantially increased T_ms (by >2°C) and denaturation enthalpies (by ∼20 to 40 kcal/mol), compared to those of cleaved BG505.SOSIP (Fig. 2C). The sc-gp140 construct with a 15-residue linker had the highest T_m, 69°C, compared to 66.3°C for cleaved BG505.SOSIP.

FIG 2.

Characterization of transiently expressed and purified BG505.SOSIP sc-gp140. (A) Gel filtration profiles for selected BG505.SOSIP sc-gp140 constructs and wild-type BG505.SOSIP. Two rounds of size exclusion chromatography (SEC) were performed for each of the constructs, as described in Materials and Methods. (B) SDS-PAGE analysis of purified sc-gp140 under nonreducing (left) and reducing (right) conditions. Lanes M, molecular weight markers. Bands corresponding to gp41 can be seen under reducing conditions in the BG505.SOSIP lane and are highlighted with a bracket and an arrow. (C) DSC measurements for selected constructs (colored curves) after F105 negative selection are shown along with T_m temperatures and ΔH denaturation enthalpies.

TABLE 4.

Protein yields of BG505.SOSIP sc-gp140 constructs

BG505.SOSIP construct	Yielda (mg/liter)
	Initial	SEC round 1	SEC round 2
1-sc gp140	3.78	1.37	0.53
2-sc gp140	1.62	0.61	0.55
3-sc gp140	0.73	0.28	0.16
4-sc gp140	2.38	1.05	0.43
6-sc gp140	1.77	0.82	0.70
7-sc gp140	2.39	1.04	0.70
10-sc gp140	2.36	0.76	0.59
15-sc gp140	2.12	0.94	0.79
20-sc gp140	2.60	1.47	0.56
BG505.SOSIP	2.76	1.05	0.93

^aThe reported yields were obtained from 1 liter of culture before SEC (initial) and after two SEC rounds.

Structural and antigenic characterization of select sc-gp140s.

To characterize the structures of the sc-gp140 constructs, we performed low-resolution negative-stain EM both before and after F105 negative selection (see Materials and Methods). Images of the analyzed sc-gp140 constructs revealed different ratios of (i) “good” trimer structures mimicking closed mature Env versus (ii) other aberrant structures (Fig. 3). The ratio of “good” versus “aberrant” structures for a given sc-gp140 construct was typically not substantially different before and after F105 negative selection, while linker lengths of six residues or more appeared to more closely match the ratio of “good” versus “aberrant” structures that was observed for cleaved BG505.SOSIP. In particular, both the 15-residue BG505.SOSIP sc-gp140 and the cleaved BG505.SOSIP exhibited on average >80% “good” trimer structures (84.3% and 82.9%, respectively) after F105 negative selection (Fig. 3C).

To determine the antigenic specificity of the BG505.SOSIP sc-gp140 constructs after F105 negative selection, we initially used biolayer interferometry to test binding to two broadly neutralizing antibodies, VRC26.09 and PGT145, as well as to the poorly neutralizing antibodies F105 and 17b. All of the sc-gp140 constructs showed binding to the broadly neutralizing but not to the poorly neutralizing antibodies, at levels similar to those for cleaved BG505.SOSIP (Fig. 4A). In particular, the affinities of the 15-residue BG505.SOSIP sc-gp140 to the quaternary-specific antibodies VRC26.09 and PGT145 were also measured, resulting in K_Ds of ∼36 nM and ∼4 nM, respectively, nearly identical to the K_Ds of ∼37 nM and ∼2 nM for cleaved BG505.SOSIP (Fig. 4B). To determine whether the introduction of a flexible linker in place of the gp120-gp41 cleavage site may lead to substantial structural changes in the cleavage site region, we tested binding of the BG505.SOSIP sc-gp140 constructs to three antibodies targeting hybrid gp120-gp41 epitopes: 35O22, 8ANC195, and the cleavage-specific PGT151. The sc-gp140 constructs had abolished binding to PGT151 but showed levels of binding to 35O22 and 8ANC195 that were similar to those for cleaved BG505.SOSIP (Fig. 4A), likely indicating structural changes in the sc-gp140 constructs that are localized to the cleavage site region. Since the 15-residue BG505.SOSIP sc-gp140 showed the highest fraction of “good” trimer structures after F105 negative selection, we selected this construct for further antigenic characterization on an expanded panel of antibodies using two different methods: electrochemical luminescent immunoassay (ECLIA) using 2G12 detection and a lectin ELISA (Fig. 5 and 6). The lectin-based ELISA format was used to capture saturating concentrations of the HIV-1 antigens. The ECLIA on the MSD platform was carried out using capture of the HIV-1 antigens with a panel of neutralizing and nonneutralizing antibodies, applied at the same concentration, and detection of the HIV-1 antigen binding by electrochemical luminescence (ECL) with labeled 2G12. The multiple assay systems allowed us to assess the antigenic characterization of the constructs without their being subject to inherent differences in the assay systems. In both assays, the 15-residue sc-gp140 and the cleaved BG505.SOSIP exhibited similar binding profiles for each antibody, with the exception of PGT151, where the 15-residue sc-gp140 had abolished binding, similar to what was observed in the biolayer interferometry experiments.

FIG 4.

Antigenic characterization of purified BG505.SOSIP sc-gp140 constructs by biolayer interferometry. (A) The panel of antibodies included V1V2-targeting broadly neutralizing antibodies VRC26.09 and PGT145, CD4-binding-site poorly neutralizing antibody F105, CD4i poorly neutralizing antibody 17b, and gp120-gp41 antibodies PGT151, 35O22, and 8ANC195, as well as CD4-binding-site antibody VRC01 as a control. The heat map (white to red) is colored according to the maximum biolayer interferometry response observed in an Octet binding assay; values for the gp120(full-length) control are not directly comparable to the values for the gp140 molecules due to its substantially lower mass. (B) Binding kinetics for BG505.SOSIP and 15-sc gp140 with VRC26.09 and PGT145.

FIG 5.

Antigenicity characterization by ECLIA. An extended panel of both broadly and poorly neutralizing antibodies was assessed using ECLIA for cleaved (black open circles) and 15-sc gp140 (red filled circles) BG505.SOSIP constructs.

FIG 6.

Antigenicity characterization by ELISA. An extended panel of both broadly and poorly neutralizing antibodies was assessed using lectin capture ELISA for cleaved (black open circles) and 15-sc gp140 (red filled circles) BG505.SOSIP constructs.

DISCUSSION

In the context of the virus, and in the case of HIV-1 in particular, the cleavage process may play a dual role: (i) to allow for structural rearrangements necessary for spike maturation and (ii) to free the fusion peptide in anticipation of interactions and subsequent fusion with a host cell. Both of these roles are essential for virus infectivity. With soluble HIV-1 Env variants, however, the second role is no longer required, so any cleavage-independent candidate constructs may need only to fulfill the first role to mimic the effects of cleavage. Here, we showed that simply replacing the gp120-gp41 cleavage site in the lead soluble gp140 molecule BG505.SOSIP with a long enough peptide linker should be sufficient to obtain constructs that can generally mimic the cleaved BG505.SOSIP both structurally and antigenically. The approach for the design of single-chain gp140 constructs in this study is similar to those described previously (58, 59), but it benefits from the use of antibodies that specifically select for trimeric Env forms (13), the structural characterization of construct variants using negative-stain EM, and the evaluation of the effects of linker length on the structure and antigenicity of the single-chain constructs. While the profiles of binding of the sc-gp140 constructs to broadly versus poorly neutralizing antibodies were somewhat independent of linker length, the negative-stain EM results suggested that longer linkers (consisting of six or more residues) may be better. In particular, the characterization of the 15-residue linker sc-gp140 indicated this construct to be virtually identical to cleaved BG505.SOSIP, both antigenically and structurally.

The observed relatively good binding to the trimer-specific broadly neutralizing antibodies by BG505.SOSIP sc-gp140 constructs with short linkers (at most four residues) was somewhat surprising, considering that the native gp120-gp41 cleavage site is four residues long. The EM results, however, indicate that a possible reason for this observation could lie in the presence of a mixture of “good” trimers and “aberrant” structures, so the trimer-specific antibody binding could be a result of the “good” trimers. The observed somewhat lower frequency of “good” trimer structures for the short-linker constructs is also in agreement with previous EM analysis of BG505.SOSIP where the wild-type ₅₀₈REKR₅₁₁ cleavage site was mutated to the inactive ₅₀₈SEKS₅₁₁ (45). Structural modeling of the one-residue linker sc-gp140 suggests that while it may be possible to use such a short linker in place of the cleavage site, the structurally feasible conformations may be more constrained (data not shown).

The fact that binding to the trimer-specific antibodies is generally retained for the BG505.SOSIP sc-gp140 constructs indicates that the apex of the molecule is likely unchanged structurally compared to that of the cleaved molecule. In contrast, the observation that binding for the BG505.SOSIP sc-gp140 constructs to cleavage-specific antibody PGT151 was abrogated could indicate either that the inserted linker masks the direct epitope of PGT151 or that more substantial structural changes occur in the cleavage site region (such as possible displacement of the linker-connected termini, especially in the case of shorter linkers). The retained binding to the other two gp120-gp41 antibodies (35O22 and 8ANC195) analyzed here suggests that any structural changes induced by the introduction of the flexible linkers are more likely to be localized to the cleavage site region than to represent more substantial rearrangements in the gp120-gp41 region. Further, the distance between residues 505 and 518 in the BG505.SOSIP structure stabilized by antibodies 35O22 and PGT122 (41) is ∼37 Å within a protomer but only ∼25 Å between adjacent protomers, indicating that it is possible that, at least for a fraction of the cases, the flexible linkers replacing the cleavage site may be linking gp120 with gp41 from adjacent, rather than the same, protomers. The precise structural effects of introducing these flexible linkers, however, can be addressed definitively only through higher-resolution structural information on the sc-gp140 molecules. While to our knowledge the immunogenicity of such flexible Gly-Ser linkers has not been explicitly evaluated, such linkers are commonly used in a variety of protein design scenarios (such as the construction of single-chain antibodies [60]) and are not expected to be major immunogenic targets (61, 62).

One potential advantage of the single-chain soluble gp140 concept described here is the lack of a requirement for furin cleavage, which could be a useful property for studies involving genetic immunization. In our hands, both our more extensively characterized 15-residue sc-gp140 construct and the cleaved BG505.SOSIP underwent a step of negative selection against molecule subspecies binding to the poorly neutralizing antibody F105, and both the 15-residue sc-gp140 and the cleaved BG505.SOSIP exhibited similar yields after F105 negative selection (at around one-half to one-third of the yields before negative selection). This need for negative selection to remove F105-reactive species can be a challenge for using these molecules as immunogens, but it may also be a result of the particular lectin purification scheme used here, rather than an intrinsic property of the molecules.

A major requirement—and a formidable barrier—for structural and antigenic mimicry of the mature closed Env conformation by soluble gp140 molecules is the presence of efficient cleavage. In this paper, we have shown that linked gp120-gp41 single-chain constructs can mirror the properties of the current best soluble gp140 molecules epitomized by the BG505.SOSIP construct. The eliminated requirement for furin cleavage may make such single-chain gp140 constructs especially useful for multimerization onto protein nanoparticles (such as ferritin [46]) where access to the cleavage site may be more obstructed, as well as potentially for developing immunogens from diverse viral strains with variable cleavage efficiency. The single-chain soluble gp140 concept may thus be of importance for HIV-1 vaccine development.