Shedding-Resistant HIV-1 Envelope Glycoproteins Adopt Downstream Conformations That Remain Responsive to Conformation-Preferring Ligands

The HIV-1 envelope glycoprotein (Env) opens in response to receptor CD4 binding from a pretriggered (state 1) conformation through a necessary intermediate to the three-CD4-bound conformation. The application of smFRET to test the conformational state of existing Env constructs and ligand complexes used for high-resolution structures recently revealed that they correspond to the downstream conformations. The structure of the pretriggered Env conformation, preferentially recognized by broadly neutralizing antibodies, remains unknown. Here, we identify experimental conditions that stabilize membrane-bound and shedding-resistant virus Env trimers in state 1, potentially facilitating structural characterization of this unknown conformational state.

ABSTRACT

The human immunodeficiency virus type 1 (HIV-1) envelope glycoprotein (Env) trimer of gp120-gp41 heterodimers mediates virus entry into CD4-positive (CD4⁺) cells. Single-molecule fluorescence resonance energy transfer (smFRET) has revealed that native Env on the surface of viruses predominantly exists in a pretriggered conformation (state 1) that is preferentially recognized by many broadly neutralizing antibodies (bNAbs). Env is activated by binding receptor CD4, which drives transitions through a default intermediate conformation (state 2) into the three-CD4-bound open conformation (state 3). The application of smFRET to assess the conformational state of existing Env constructs and ligand complexes recently revealed that all current high-resolution structures correspond to downstream states 2 and 3. The structure of state 1, therefore, remains unknown. We sought to identify conditions whereby HIV-1 Env could be stabilized in the pretriggered state 1 for possible structural characterization. Shedding of gp120, known to severely complicate structural studies, can be prevented by using the uncleaved gp160_JR-FL precursor with alterations in the protease cleavage site (R508S/R511S) or by introducing a disulfide bridge between gp120 and gp41 designated “SOS” (A501C/T605C). smFRET demonstrated that both shedding-preventing modifications shifted the conformational landscape of Env downstream toward states 2 and 3. However, both membrane-bound Env proteins on the surface of intact viruses remained conformationally dynamic, responsive to state-stabilizing ligands, and able to be stabilized in state 1 by specific ligands such as the Bristol-Myers Squibb (BMS) entry inhibitors. The here-described identification of state 1-stabilizing conditions may enable structural characterization of the state 1 conformation of HIV-1 Env.

IMPORTANCE The HIV-1 envelope glycoprotein (Env) opens in response to receptor CD4 binding from a pretriggered (state 1) conformation through a necessary intermediate to the three-CD4-bound conformation. The application of smFRET to test the conformational state of existing Env constructs and ligand complexes used for high-resolution structures recently revealed that they correspond to the downstream conformations. The structure of the pretriggered Env conformation, preferentially recognized by broadly neutralizing antibodies, remains unknown. Here, we identify experimental conditions that stabilize membrane-bound and shedding-resistant virus Env trimers in state 1, potentially facilitating structural characterization of this unknown conformational state.

INTRODUCTION

The envelope (Env) glycoprotein exposed on the surface of the human immunodeficiency virus type 1 (HIV-1) viruses is a trimer of an exterior gp120 subunit that associates noncovalently with a transmembrane gp41 subunit (1). HIV-1 Env is initially synthesized as a gp160 precursor, which is later processed by the Golgi-resident furin protease into the mature Env consisting of gp120 and gp41. The mature cleaved HIV-1 Env is incorporated into virions and mediates virus entry into cells expressing the receptors, CD4, and a chemokine receptor, either CCR5 or CXCR4 (1). CD4-induced conformational changes within gp120 expose chemokine receptor-binding sites; subsequent CCR5 or CXCR4 binding to Env is thought to activate gp41 to fuse the viral and cellular membranes (1–8). The conformational changes within gp120 in response to CD4 have been visualized by single-molecule fluorescence resonance energy transfer (smFRET). These investigations have revealed that Env undergoes transitions from the unliganded, pretriggered conformation (state 1) through a multistep process via at least one obligate intermediate (state 2) to the three-CD4-bound conformation (state 3) (9–11).

As the sole HIV-1-specific surface protein, the Env trimer is a main target for adaptive immune responses and antibody recognition (12–15). In response, HIV-1 Env has evolved elaborate immune evasion strategies that include high-sequence variation of the exposed protein surface, a dense glycan shield, and conformationally dynamic epitopes (3, 10, 11, 16–21). The open conformational states of Env are highly immunogenic, leading to the elicitation of high titers of antibodies. However, HIV-1 Env can escape these antibodies by adopting the most immune-evasive conformation, a closed pretriggered state 1 that is conserved among HIV-1 strains. A minority of patients develop antibodies that can recognize conserved features of this conformation, which results in broad neutralization of many HIV-1 isolates (15, 22–27). Using smFRET, we have found that many broadly neutralizing antibodies (bNAbs) exhibit a preference for the state 1 conformation of the native Env trimer (10, 20). Allosteric entry blockers of the Bristol-Myers Squibb (BMS) series (BMS-626529/Temsavir and BMS-378806) also stabilize HIV-1 Env in state 1 (10, 20, 28–31). The preference of bNAbs for state 1 suggests that Env-targeted immunogens intended to elicit such antibodies should resemble the pretriggered state 1 conformation.

The development of immunogens mimicking the HIV-1 Env trimer had to overcome two main hurdles. First, a closed trimer was required that does not expose CD4-induced epitopes, resulting in the elicitation of nonneutralizing antibodies (32). Second, the instability of cleaved HIV-1 Env trimers that easily led to shedding of gp120 had to be addressed (32–36). Over a period of more than 10 years, Moore, Sanders, and Binley developed well-behaved soluble trimers (37–39). They prevented gp120 shedding by the introduction of a disulfide bridge (SOS, A501C, and T605C), stabilized a prefusion conformation by introduction of the I559P (IP) change in gp41, and took advantage of a naturally more stable HIV-1 isolate BG505; in addition, the soluble trimers were truncated at residue 664 to exclude the hydrophobic membrane-proximal external region (MPER), as well as the transmembrane region and cytoplasmic tail (37–39).

The generation of these stabilized soluble trimers has led to major breakthroughs in the structural characterization of the HIV-1 Env trimer (4, 6, 40–43). Importantly, membrane-bound HIV-1_JR-FL Env (including the MPER and the transmembrane anchor) in complex with bNAb PGT151 adopted essentially the same structure (44). To directly compare the conformational states of these Env proteins to those observed by smFRET for Env on native virions, we recently introduced smFRET-compatible fluorophores in the same positions as those used for native Env (20). Surprisingly, we found that the sgp140 SOSIP.664 trimer predominantly resides in a state 2-like conformation (20). The binding of the PGT151 antibody to virion Env induced a similar state 2-like conformation (18). That sgp140 SOSIP.664 proteins exhibit conformations distinct from state 1 agrees with observed differences in the processing of HIV-1 Env glycans (45, 46), in distance constraints of cross-linked lysine residues (47), and in CD4 engagement of more open Env mutants (48). These data indicate that extant HIV-1 Env proteins and Env-ligand complexes currently used for structural characterization likely reside in a downstream state 2-like conformation (20). Correspondingly, the structure of the state 1 Env conformation observed by smFRET on the surface of viruses remains unknown. Given that many bNAbs preferentially bind the state 1 conformation of HIV-1 Env (10, 20), identification of experimental conditions that allow a structural characterization of this conformation will be a prerequisite for the rational design of immunogens presenting the state 1 conformation.

Here, we describe our initial efforts to use smFRET to identify experimental conditions that stabilize HIV-1 Env in state 1 for eventual structural characterization. We concentrate on membrane-bound virus Env since soluble Env proteins are likely more prone to assume downstream conformations (49). To overcome gp120 shedding that complicates purification of HIV-1 Env, we explored two approaches: (i) using the uncleaved gp160 protein; or (ii) introducing the SOS disulfide bridge (A501C, T605C) that links gp120 and gp41 (38, 50). We introduced the R508S/R511S changes into the cleavage site of the virus-resident HIV-1_JR-FL Env trimer to eliminate proteolytic cleavage between gp120 and gp41. In addition, we introduced the SOS (A501C, T605C) and IP (I559P) alterations into membrane-bound HIV-1_BG505 virus Env to covalently link gp120 and gp41 (38). We demonstrate that while both sets of changes shift the Env conformational landscape toward more open conformations, both membrane-bound virus Env proteins remain dynamic and could still be stabilized in state 1 using specific ligands. These insights could assist structural characterization of the pretriggered state 1 conformation of HIV-1 Env.

RESULTS AND DISCUSSION

We used smFRET to study Env on native virions, comparing the conformational landscapes of the uncleaved HIV-1_JR-FL Env and the cleaved HIV-1_BG505 Env carrying SOS&IP changes (HIV-1_BG505 SOS&IP) with those of their respective mature wild-type Env proteins. Site-specific enzymatic labeling peptides (Q3, GQQQLG; A1, GDSLDMLEWSLM) were introduced into the variable loops V1 and V4 of gp120, respectively. These peptide tags allowed the site-specific incorporation of donor [Cy3B(3S)-cadaverine] and acceptor (LD650-CoA) fluorophores into V1 and V4 using transglutaminase and panthenyl-transferase, respectively (Fig. 1A). As previously described (10), virions bearing one dually labeled Env molecule per particle on average were prepared and immobilized on quartz slides for total internal reflection smFRET imaging. Fluorescence intensities from donor and acceptor were separated by an optical beam splitter, recorded by scientific complementary metal oxide semiconductor (sCMOS) cameras at 25 frames/s, and analyzed using the Spartan software package (51). Figure 1B and C illustrate the approach for the unliganded mature HIV-1_JR-FL Env. Consistent with published results (10, 11), the unliganded HIV-1_JR-FL was observed to be dynamic, transitioning between three distinct conformations: a pretriggered state 1 (low FRET), a default intermediate state 2 (high FRET), and the three-CD4-bound state 3 (intermediate FRET).

FIG 1.

Single-molecule imaging of individual mature HIV-1_JR-FL Env trimers on the surface of intact virions reveals three Env conformational states. (A) Immobilization of intact HIV-1 virions carrying one dually labeled Env molecule per particle in the presence of excess wild-type Env on a quartz slide for total internal reflection smFRET imaging (Cy3, green; Cy5, red). Structures of Env trimers are adapted from PDB access code 4ZMJ (dually labeled gp120; cyan, nonlabeled gp120; blue; gp41, gray). (B, C) Representative fluorescence (B) and resulting FRET (C) traces of unliganded HIV-1_JR-FL Env on the surface of an intact virus carrying fluorophores in V1 and V4 of gp120 (donor Cy3, green; acceptor Cy5, red; resulting FRET, blue; hidden Markov model [HMM] idealization [59, 60], magenta). The anticorrelated feature between the donor fluorophore and acceptor fluorophore, as well as clear transitions among three distinct conformational states (state 1, state 2, and state 3), is highlighted in zoomed windows (B and C, right panels).

Compiling individual FRET traces into population FRET histograms confirmed that the mature HIV-1_JR-FL Env predominantly resides in state 1 (Fig. 2A and B) (10, 11, 20). In contrast, the uncleaved HIV-1_JR-FL Env with the R508S/R511S changes sampled all three conformational states, with a high proportion of the molecules occupying states 2 and 3 (Fig. 2C). This finding is consistent with antigenicity data indicating that the uncleaved Env, compared with mature Env, exhibits a loss of binding of state 1-preferring bNAbs such as VRC01 and increased binding of nonneutralizing antibodies (32, 52–55). We next tested the ability of the uncleaved HIV-1_JR-FL Env to respond to soluble CD4. Primary clinical tiers 2 and 3, such as HIV-1_JR-FL and HIV-1_BG505 isolates, are highly resistant to monomeric soluble CD4 (11, 20). We used the dodecameric soluble CD4 (sCD4_D1D2-Igαtp) as a potent CD4 ligand (11, 20). Interestingly, the uncleaved HIV-1_JR-FL Env precursor remained highly responsive to sCD4_D1D2-Igαtp binding and efficiently opened into state 3 (Fig. 2D and E). In contrast, the allosteric conformational blockers BMS-378806 and BMS-626529/Temsavir, which stabilize the mature HIV-1_JR-FL Env in state 1 (Fig. 2F and H), respectively, shifted the energy landscapes of the uncleaved HIV-1_JR-FL Env toward state 1 (Fig. 2G and I).

FIG 2.

The uncleaved HIV-1_JR-FL Env precursor samples multiple conformations but remains responsive to ligands, including state 1-stabilizing ligands. (A) Schematic comparison of the mature HIV-1_JR-FL Env and the uncleaved HIV-1_JR-FL Env precursor. (B, C) Wild-type HIV-1_JR-FL Env predominantly resides in state 1, whereas the uncleaved Env precursor exhibits relatively higher occupancy of states 2 and 3. FRET histograms of unliganded wild-type HIV-1_JR-FL Env (B) and the uncleaved HIV-1_JR-FL Env precursor (C) were compiled from 99 and 113 FRET traces, respectively. Fitted curves (red) for three Gaussian distributions (black) centered at FRET values of ∼0.1 (state 1), ∼0.3 (state 3), and ∼0.65 (state 2) were overlaid on the histograms, consistent with three different conformational states (indicated by the state nomenclature in white). The corresponding state occupancies are shown in the pie charts. (D, E, F, G, H, and I) Shifting of the conformational landscape of the wild-type HIV-1_JR-FL and uncleaved Env upon binding of different ligands. FRET histograms with state occupancies for the wild-type HIV-1_JR-FL Env and uncleaved Env in the presence of sCD4_D1D2-Igαtp (10 μg/ml) (D, E), BMS-378806 (100 μM) (F, G), and BMS-626529 (100 μM) (H, I). The uncleaved HIV-1_JR-FL Env is shifted into state 3 by sCD4_D1D2-Igαtp and into state 1 by both BMS-378806 and BMS-626529. Error bars of the FRET histograms are standard errors determined from three independent sets of FRET traces.

We next evaluated the conformational sampling of HIV-1_BG505 Env bearing the SOS&IP changes on the surfaces of intact virions (Fig. 3A). We performed these studies with the BG505 isolate in which the SOS&IP mutations have been extensively studied (37–39). In agreement with previous observations, the introduction of the SOS&IP changes into the full-length HIV-1_BG505 Env resulted in an ensemble of conformations with a preferential stabilization of state 2 (20), such that the state 1 conformation, predominant in wild-type Env, was disfavored (Fig. 3B and C). However, in contrast to the conformationally restricted, soluble HIV-1_BG505 gp140 SOSIP.664 trimer (20), the HIV-1_BG505 virus SOS&IP Env remained responsive to ligands, opening into state 3 upon addition of sCD4_D1D2-Igαtp (Fig. 3D) and closing into state 1 upon addition of BMS-626529 (Fig. 3E). Recent X-ray structures of complexes confirm that BMS-626529 does not induce globally distinct conformations upon binding the soluble gp140 SOSIP.664 trimer (29). The observation that BMS-626529 stabilizes state 1 in a virus Env carrying SOS&IP changes may open an additional avenue toward enriching a state 1 conformation.

FIG 3.

HIV-1_BG505 virus Env carrying the SOS&IP changes remains conformationally dynamic. (A) Schematic comparison of the HIV-1_BG505 wild-type Env and HIV-1_BG505 SOS&IP Env. (B, C) FRET population histograms of unliganded HIV-1_BG505 Env (B) and HIV-1_BG505 SOS&IP Env (C) on the surface of intact virions. (D, E) HIV-1_BG505 SOS&IP opens in response to 10 μg/ml sCD4_D1D2-Igαtp (D) and closes in the presence of 100 μM BMS-626529 (E).

In this short report, we describe the conformational behaviors of the uncleaved HIV-1_JR-FL Env and HIV-1_BG505 Env carrying the SOS&IP changes on the surface of intact virions, two Env constructs potentially useful for high-resolution structural studies and evaluation of immunogenicity. We observed that these modifications shift virus Env away from the pretriggered state 1 conformation toward the downstream conformational states 2 and 3. Specifically, the uncleaved HIV-1_JR-FL Env sampled states 1, 2, and 3, while HIV-1_BG505 Env SOS&IP were stabilized in state 2. These propensities reflect significant complications for the use of these Env variants to investigate the pretriggered state 1 conformation. However, the entry inhibitors BMS-626529/Temsavir and/or BMS-378806 were capable of re-equilibrating the conformational landscape of both Env proteins on the virus toward state 1. These observations open a possible path toward detailed structural characterizations and immunogenicity studies of the state 1 conformation using these Env variants.

The stabilization of membrane-embedded Env carrying SOS&IP changes in state 1 by the allosteric BMS entry inhibitors is in contrast to the behavior of soluble gp140 SOSIP.664, which fails to be significantly stabilized in state 1 by these compounds (20). This finding likely indicates a role for the virus membrane in stabilizing state 1. Indeed, it has recently been shown that membrane cholesterol plays an important role in regulating HIV-1 Env conformational stability (49). Further work will be required to understand how the virus membrane, likely by interacting with the Env transmembrane domain and possibly the MPER region, affects Env conformation.

Finally, it is interesting to speculate why the uncleaved Env samples all three conformational states, a finding that is consistent with the previously reported increased binding of nonneutralizing antibodies (52–55). A high level of conformational flexibility of the Env precursor may facilitate the processing of Env carbohydrates into complex glycans in the Golgi. This is in agreement with the observation that the rigid sgp140 SOSIP.664 proteins are impaired in their glycosylation maturation, exhibiting an unusually large amount of high-mannose glycans (46). During HIV-1 infection, uncleaved Env may present nonneutralizing antibody epitopes, distracting the host immune system from the elicitation of antibodies targeting the pretriggered state 1. Thus, the conformational properties of the uncleaved Env, compared to the mature Env incorporated into the virus, differentially serve to divert and evade, respectively, the host antibody response.

MATERIALS AND METHODS

Preparation of Env-tagged virus.

Preparation of HIV-1_JR-FL with tagged wild-type or R508S/R511S mutant Envs and HIV-1_BG505 with wild-type or SOS&IP mutant Envs was essentially as described (10, 11, 20). Briefly, two enzymatic labeling peptides (Q3, GQQQLG; A1, GDSLDMLEWSLM) were introduced into variable loops V1 and V4 of the gp120 subunit, respectively (designated V1-Q3/V4-A1). Tagged Env of primary clade B HIV-1_JR-FL and clade A HIV-1_BG505 isolates were validated in terms of infectivity, Env processing and incorporation, and neutralization sensitivity to trimer-specific antibodies (10, 11). To make enzymatically tagged HIV-1_BG505 for smFRET imaging, a 40:1 ratio of wild-type full-length HIV-1 Q23 BG505 ΔRT plasmid to V1-Q3/V4-A1 tagged HIV-1 Q23 BG505 ΔRT plasmid was used in cotransfections of HEK293 cells. HIV-1 Q23 BG505 is an infectious clade A HIV-1 clone carrying the BG505 Env with Q23 as the backbone (56). Similarly, tagged HIV-1_BG505 Envs carrying A501C/T605C and I559P changes (SOS&IP) were generated by cotransfecting HEK293 cells with a 40:1 ratio of full-length HIV-1 Q23 BG505 SOS&IP ΔRT plasmid to V1-Q3/V4-A1 tagged HIV-1 Q23 BG505 SOS&IP ΔRT plasmid.

HIV-1_JR-FL viruses were produced by pseudotyping a pNL4-3 ΔRT ΔEnv backbone with HIV-1_JR-FL wild-type or precursor (R508S/R511S) Envs. Specifically, a tagged HIV-1_JR-FL virus was generated by cotransfecting HEK293 cells using a 40:1 plasmid ratio of plasmid pCAGGS expressing wild-type JR-FL gp160 and V1-Q3/V4-A1 tagged JR-FL gp160 with the same total amount of the plasmid expressing the pNL4-3 ΔRT ΔEnv backbone. In the case of the HIV-1_JR-FL Env precursor, plasmids expressing JR-FL gp160 with R508S/R511S changes and V1-Q3/V4-A1 tagged JR-FL gp160 with R508S/R511S were used instead. Env-tagged viruses were harvested 40-h posttransfection, filtered, and concentrated by centrifugation on top of a 15% sucrose cushion at 25,000 rpm for 2 h. The viruses were then resuspended in 50 mM HEPES buffer (pH 7.5, 10 mM MgCl₂, and 10 mM CaCl₂).

Preparation of fluorescently labeled Env on virus.

For labeling, Cy3B(3S)-cadaverine (0.5 μM), transglutaminase (0.65 μM; Sigma-Aldrich), LD650-CoA (0.5 μM; Lumidyne Technologies), and AcpS (5 μM) were added to the above virus suspensions and incubated at room temperature overnight (57, 58). The next day, polyethylene glycol 2000 (PEG 2000)-biotin (0.02 mg/ml; Avanti Polar Lipids) was added to the labeling reaction and incubated for 30 min. Labeled virions were then purified by ultracentrifugation at 40,000 rpm over a 6% to 18% OptiPrep (Sigma-Aldrich) gradient to remove free dyes.

smFRET imaging acquisition.

All smFRET imaging experiments were performed on a home-built total internal reflection fluorescence (TIRF) microscope (51). Env-tagged fluorescent viruses were immobilized on passivated, streptavidin-coated quartz slides prior to imaging. A single-frequency 532-nm laser was used to excite donor fluorophores. Both donor and acceptor fluorescence signals were collected through a 60× (1.27-numerical-aperture [NA]) water-immersion objective (Nikon) and optically split using a 650DCXR dichroic filter (Chroma) mounted on a MultiCam LS image splitter (Cairn Research). Intensities/images were simultaneously recorded on two synchronized ORCA-Flash4.0v2 sCMOS cameras (Hamamatsu) at 40 frames/s for 80 s. Each recorded movie capturing the conformational motions of HIV-1 Env had a total of 2,000 frames. During data acquisition, virus samples were imaged in buffer containing Tris (pH 7.4, 50 mM) with NaCl (50 mM), a cocktail of triplet-state quenchers, and protocatechuic acid (PCA; 2 mM) with protocatechuic 3,4-dioxygenase (PCD; 8 nM) to remove molecular oxygen. The conformational effects of ligands were tested by preincubating fluorescently labeled viruses with 10 μg/ml sCD4_D1D2-Igαtp (James Arthos Laboratory) or 100 μM BMS-378806 (Amos B. Smith III Laboratory) and BMS-626529 (ApexBio Technology) for 30 min at room temperature prior to imaging in the continued presence of the ligands.

smFRET data analysis.

smFRET data were analyzed by using the Spartan software package (51). From the raw movies containing fluorescence information in each image frame, the background signal was first identified based on the fluorophore bleaching point and subtracted. Donor and acceptor fluorescence intensity trajectories/traces were then extracted from movies. FRET efficiency was calculated according to FRET= I_A/(γI_D + I_A), where I_D and I_A are the fluorescence intensities of donor and acceptor, respectively, and γ is the correlation coefficient, which corrects for the difference in quantum yields and detection efficiencies of donor and acceptor. FRET trajectories/traces identified based upon the criteria of sufficient signal-to-noise (S/N) ratio and anticorrelated features between donor and acceptor intensity were then compiled into FRET histograms. Based on the observations of clear state-to-state transitions in each of the FRET traces, the FRET histograms were fit to the sum of three Gaussian distributions in MATLAB-MathWorks, where the area under each Gaussian curve was used to estimate the occupancy of each FRET state, displayed as pie charts in each figure panel. The idealization of representative FRET traces into 3-state hidden Markov models (Fig. 1C) was performed in the Spartan software environment using a segmental K-means algorithm (59, 60).