Functional and Highly Cross-Linkable HIV-1 Envelope Glycoproteins Enriched in a Pretriggered Conformation

ABSTRACT

Binding to the receptor, CD4, drives the pretriggered, “closed” (state-1) conformation of the human immunodeficiency virus type 1 (HIV-1) envelope glycoprotein (Env) trimer into more “open” conformations (states 2 and 3). Broadly neutralizing antibodies, which are elicited inefficiently, mostly recognize the state-1 Env conformation, whereas the more commonly elicited poorly neutralizing antibodies recognize states 2/3. HIV-1 Env metastability has created challenges for defining the state-1 structure and developing immunogens mimicking this labile conformation. The availability of functional state-1 Envs that can be efficiently cross-linked at lysine and/or acidic amino acid residues might assist these endeavors. To that end, we modified HIV-1_AD8 Env, which exhibits an intermediate level of triggerability by CD4. We introduced lysine/acidic residues at positions that exhibit such polymorphisms in natural HIV-1 strains. Env changes that were tolerated with respect to gp120-gp41 processing, subunit association, and virus entry were further combined. Two common polymorphisms, Q114E and Q567K, as well as a known variant, A582T, additively rendered pseudoviruses resistant to cold, soluble CD4, and a CD4-mimetic compound, phenotypes indicative of stabilization of the pretriggered state-1 Env conformation. Combining these changes resulted in two lysine-rich HIV-1_AD8 Env variants (E.2 and AE.2) with neutralization- and cold-resistant phenotypes comparable to those of natural, less triggerable tier 2/3 HIV-1 isolates. Compared with these and the parental Envs, the E.2 and AE.2 Envs were cleaved more efficiently and exhibited stronger gp120-trimer association in detergent lysates. These highly cross-linkable Envs enriched in a pretriggered conformation should assist characterization of the structure and immunogenicity of this labile state.

IMPORTANCE The development of an efficient vaccine is critical for combating HIV-1 infection worldwide. However, the instability of the pretriggered shape (state 1) of the viral envelope glycoprotein (Env) makes it difficult to raise neutralizing antibodies against HIV-1. Here, by introducing multiple changes in Env, we derived two HIV-1 Env variants that are enriched in state 1 and can be efficiently cross-linked to maintain this shape. These Env complexes are more stable in detergent, assisting their purification. Thus, our study provides a path to a better characterization of the native pretriggered Env, which should assist vaccine development.

INTRODUCTION

Human immunodeficiency virus type 1 (HIV-1) entry into target cells is mediated by the viral envelope glycoprotein (Env) trimer (1, 2). The Env trimer is composed of three gp120 exterior subunits and three gp41 transmembrane subunits (2). In infected cells, Env is first synthesized as an uncleaved precursor in the rough endoplasmic reticulum (ER), where signal peptide cleavage, folding, trimerization, and the addition of high-mannose glycans take place (3–6). Exiting the ER, the trimeric gp160 Env precursor follows two pathways to the cell surface (7). In the conventional secretory pathway, the Env precursor transits through the Golgi compartment, where it is cleaved into gp120 and gp41 subunits and is further modified by the addition of complex sugars. These mature Envs are transported to the cell surface and are incorporated into virions (8–11). In the second pathway, the gp160 precursor bypasses processing in the Golgi compartment and traffics directly to the cell surface; these Golgi compartment-bypassed gp160 Envs are excluded from virions (7).

Single-molecule fluorescence resonance energy transfer (smFRET) experiments indicate that, on virus particles, the trimer exists in three conformational states (states 1 to 3) (12). From its pretriggered conformation (state 1), the metastable Env trimer interacts with the receptors, CD4 and CCR5 or CXCR4, and undergoes transitions to lower-energy states (2, 12–23). Initially, the engagement with CD4 induces an asymmetric intermediate Env conformation, with the CD4-bound protomer in state 3 and the unliganded protomers in state 2 (24, 25). Binding of additional CD4 molecules to the Env trimer then induces the full CD4-bound, prehairpin intermediate conformation, with all three Env protomers in state 3 (24–31). An extended coiled coil consisting of the heptad repeat (HR1) region of gp41 is exposed in the prehairpin intermediate (23, 25, 29–31). State-3 Env protomers subsequently interact with the CCR5 or CXCR4 coreceptor to trigger the formation of a gp41 six-helix bundle, a process that results in fusion of the viral and target cell membranes (32–36).

Env strain variability, heavy glycosylation, and conformational flexibility contribute to HIV-1 persistence by avoiding the binding of potentially neutralizing antibodies. Mature Envs from primary HIV-1 strains largely reside in a state-1 conformation, which resists the binding of most antibodies elicited during natural infection (12, 23, 37–39); these high-titer, poorly neutralizing antibodies often recognize state-2/3 Env conformations (40–44). After years of infection, a small percentage of HIV-1-infected individuals generate broadly neutralizing antibodies (bNAbs), most of which recognize the state-1 Env conformation (12, 37, 38, 45–54). Passively administered monoclonal bNAbs have been shown to be protective in animal models of HIV-1 infection, suggesting that the elicitation of bNAbs is an important goal for vaccines (55–60). Unfortunately, bNAbs have not been efficiently and consistently elicited in animals immunized with current HIV-1 vaccine candidates, including stabilized soluble gp140 (sgp140) SOSIP.664 trimers (61–69). Compared with functional membrane Envs, differences in the antigenicity, glycosylation, and conformation of sgp140 SOSIP.664 trimers have been observed (70–77), potentially contributing to the inefficiency of bNAb elicitation. Single-molecule FRET (smFRET) analysis indicates that the sgp140 SOSIP.664 trimers assume a state-2-like conformation (78). These studies imply that the available structures of sgp140 SOSIP.664 and other detergent-solubilized Env trimer preparations (27, 28, 77, 79–91) differ from that of state-1 Env. The extent of the structural differences between the state-1 and state-2 Env conformations and their potential impact on Env immunogenicity are unknown. However, the importance of the state-1 Env as a likely target of vaccine-induced bNAbs provides a rationale for better characterization of this conformation.

HIV-1 is a polymorphic virus with a high mutation rate, allowing escape from host immune responses and antiretroviral drugs (92–99). Env polymorphisms that arise naturally or as a result of tissue culture adaptation can result in altered virus infectivity, receptor binding, or neutralization sensitivity (23, 40, 44–46, 100–119). Specifically, changes in “restraining residues” in gp120 have been shown to destabilize state 1, disrupt the closed pretriggered Env conformation, and lead to increased sampling of downstream conformations (45, 118, 119). These more “triggerable” Env mutants exhibit increased sensitivity to cold, soluble CD4 (sCD4), CD4-mimetic compounds, and poorly neutralizing antibodies (23, 37, 45, 118, 119). Less common Env alterations apparently decrease Env triggerability and stabilize a state-1 conformation (120–125).

Cross-linking of HIV-1 Env amino acid residues, in some cases combined with mass spectrometry, has been used to study Env conformations (37, 73, 126–131). Cross-linking protocols that target lysine or acidic amino acid residues on native proteins have been integrated with mass spectrometry to provide low-resolution structural information (132–135). Here, we introduced lysine and acidic amino acid residues into a primary HIV-1 Env, using natural polymorphisms as a guide. Env changes that were functionally tolerated were combined to create Envs that are potentially able to be conformationally fixed by treatment with specific cross-linking agents. In the process of generating these Env variants, we identified two common polymorphisms that increased virus resistance to cold, sCD4, and a CD4-mimetic compound, phenotypes associated with stabilization of a pretriggered (state-1) Env conformation (120–125). In combination, state-1-stabilizing changes resulted in additive Env phenotypes. Two lysine-rich variants with cold- and sCD4-resistant phenotypes were cleaved more efficiently and exhibited stronger gp120-trimer association in detergent lysates than the parental HIV-1 Env. Such highly cross-linkable and efficiently processed Envs enriched in a pretriggered conformation should assist characterization of state 1.

RESULTS

Env variants with common lysine and acidic residue polymorphisms.

We sought to create functional primary HIV-1 Env variants with an increased number of lysine/acidic residues that could be used to introduce stabilizing cross-links. To identify Env residues that might potentially tolerate such substitutions, we compared Env sequences from 193 representative group M, N, O, and P HIV-1 and SIV_cpz strains (136). We identified Env residues where lysine or acidic substitutions occurred in at least 5% of these natural virus strains from more than one phylogenetic clade. The lysine polymorphisms were grouped by location in Env regions (gp41 and gp120 C terminus, gp120 trimer association domain, and gp120 inner domain) and by the number of substitutions in a set (sets 4 to 7 contain additional lysine substitutions compared with those in sets 1 to 3) (Fig. 1A). The ED2 set contains seven of the most common aspartic acid and glutamic acid polymorphisms in natural HIV-1/SIV_cpz variants (Fig. 1A). None of these polymorphisms affect known sites of N- or O-linked Env glycosylation.

FIG 1.

HIV-1_AD8 Env modification guided by natural polymorphisms. (A) Natural polymorphisms in HIV-1 Env were used to suggest amino acid residues that might tolerate replacement with a lysine residue or with acidic amino acid residues. The lysine substitutions are grouped according to the Env region in which the residues are located. Compared with sets 1 to 3, sets 4 to 7 contain an increased number of substitutions. (B) A schematic representation of the HIV-1_AD8 Env glycoprotein is shown, with the gp120-gp41 cleavage site depicted as a black triangle. S, signal peptide; V1 to V5, gp120 major hypervariable regions; FP, fusion peptide; HR, heptad repeat region; TM, transmembrane region; CT, cytoplasmic tail. The amino acid changes associated with some of the key Env variants studied here are shown. Red vertical tick marks indicate changes in addition to those found in the 2–4 R Env.

We selected the primary clade B HIV-1_AD8 as the source of the parental “wild-type” Env in this study. Primary HIV-1 Envs differ in triggerability by CD4, a property that influences virus resistance to sCD4, CD4-mimetic compounds, and some antibodies (23). The HIV-1_AD8 Env is efficiently expressed and processed, is well characterized with respect to antibody binding and neutralization sensitivity (Tier 2), and, among primary HIV-1 Envs, exhibits an intermediate level of triggerability by CD4 (7, 23, 37, 73). Single, double, and triple sets of lysine substitutions were introduced into the wild-type HIV-1_AD8 Env. For example, double sets included sets 1 + 2, 1 + 3, 2 + 3, 2 + 4, 3 + 5, etc.; triple sets included sets 1 + 2 + 3, 2 + 3 + 4, 2 + 3 + 6, 3 + 6 + 7, etc. (Fig. 1A). In a preliminary study, a total of 24 Env variants were analyzed for protein expression and processing, ability to support entry of a pseudotyped virus, and sensitivity of the viral pseudotype to neutralization by the 19b antibody. The 19b antibody is a poorly neutralizing antibody that recognizes the gp120 V3 loop and serves as a sensitive indicator of HIV-1 Env transitions to state-2/3 conformations (45, 71–74, 137). With a few exceptions, most of the lysine substitutions were well tolerated with respect to HIV-1_AD8 Env processing, virus infectivity, and sensitivity to 19b neutralization (data not shown). However, Envs with set 3 + 7 and set 3 + 6 + 7 changes were poorly processed and inefficiently supported pseudovirus infection. Viruses with set 3 + 5 changes were more sensitive than the wild-type HIV-1_AD8 to neutralization by the 19b antibody (data not shown). Thus, while most of the introduced lysine substitutions were well tolerated, some specific combinations apparently exert undesirable effects on HIV-1_AD8 Env conformation and function.

Lysine-rich 2–4 R and 2–4 RED2 Envs.

Based on the results of our preliminary analysis, we selected the 2–4 R Env, which contains set 2 + 4 and R315K changes, for more detailed characterization. The ED2 set of acidic substitutions was also added to the 2–4 R Env to create the 2–4 RED2 Env (Fig. 1B). Both 2–4 R and 2–4 RED2 Envs mediated pseudovirus infection as efficiently as the wild-type HIV-1_AD8 Env (data not shown). To evaluate Env expression, proteolytic processing, and gp120-trimer association, HOS cells were transfected with plasmids expressing the wild-type HIV-1_AD8 Env and the 2–4 R and 2–4 RED2 Envs tagged at the C terminus with His₆. Cell lysates were Western blotted directly (input) or were precipitated with nickel-nitrilotriacetic (Ni-NTA) beads in the presence of BMS-806, sCD4, or the dimethyl sulfoxide (DMSO) control. BMS-806 is a small-molecule HIV-1 entry inhibitor that binds gp120 and stabilizes a state-1-like Env conformation (12, 78, 138–140). The uncleaved gp160 Env precursor and mature gp120 and gp41 glycoproteins were detected in lysates of cells expressing the wild-type HIV-1_AD8, 2–4 R, and 2–4 RED2 Envs (Fig. 2A, input lanes). Comparison of the gp120/gp160 ratio in the cell lysates indicates that the 2–4 R and 2–4 RED2 Envs are processed more efficiently than the wild-type HIV-1_AD8 Env (Fig. 2A, input lanes). In the DMSO control sample, although wild-type HIV-1_AD8 gp41 and gp160 were precipitated by the Ni-NTA beads, little gp120 was coprecipitated (Fig. 2A, Ni-NTA lanes). Apparently, under these conditions, gp120 dissociates from the wild-type HIV-1_AD8 Env complex. BMS-806 increased the association of the wild-type HIV-1_AD8 gp120 with the precipitated Env complex, as previously seen (138). In the presence of sCD4, no coprecipitated gp120 was detected, presumably as a result of CD4-induced gp120 shedding (141, 142). Compared with the wild-type HIV-1_AD8 Env, the 2–4 R gp120 was precipitated more efficiently by the Ni-NTA beads in the DMSO control lysates. The coprecipitation of the 2–4 RED2 gp120 from the DMSO-treated cell lysates by the Ni-NTA beads was even more efficient. For both 2–4 R and 2–4 RED2 Envs, the association of gp120 with the Env complex was enhanced by BMS-806 and decreased by sCD4. Thus, the Env changes in 2–4 R and 2–4 RED2 can enhance Env processing and, in detergent lysates, strengthen the association of gp120 with solubilized Env trimers. Both phenotypes were more pronounced for the 2–4 RED2 Env than for the 2–4 R Env.

FIG 2.

Phenotypes of the 2–4 R and 2–4 RED2 Envs. (A) HOS cells were transfected transiently with plasmids expressing His₆-tagged wild-type HIV-1_AD8 Env or the 2–4 R or 2–4 RED2 Env variants. Forty-eight hours later, cells were lysed; the cell lysates were incubated with Ni-NTA beads for 1.5 h at 4°C in the presence of the DMSO control, 10 μM BMS-806, or 10 μg/mL sCD4. Total cell lysates (input) and proteins bound to the Ni-NTA beads were Western blotted with a goat anti-gp120 antibody (upper panels) or the 4E10 anti-gp41 antibody (lower panels). (B) 293T cells were transfected with plasmids encoding the indicated Envs, HIV-1 packaging proteins, and Tat and a luciferase-expressing HIV-1 vector. Forty-eight hours later, cell supernatants were filtered (0.45 μm) and incubated with different antibodies for 1 h at 37°C before the mixture was added to Cf2Th target cells expressing CD4 and CCR5. Forty-eight hours after infection, the target cells were lysed, and the luciferase activity was measured. The concentration of antibody required to inhibit 50% of virus infection (IC₅₀) was calculated using the GraphPad Prism program. (C) Filtered cell supernatants containing recombinant viruses were incubated with sCD4 or the CD4-mimetic compound BNM-III-170 for 1 h at 37°C. The mixture was then added to target cells as described above. In the cold sensitivity assay, viruses were incubated on ice for the indicated times, after which the virus infectivity was measured. The infectivity values were normalized to those obtained without virus incubation in the cold or in the absence of inhibitor. The results shown in panels A and C are representative of those obtained in at least two independent experiments. The means and standard deviations derived from two independent experiments or triplicate measurements are shown in panels B and C, respectively.

The sensitivity of viruses with the wild-type HIV-1_AD8, 2–4 R, and 2–4 RED2 Envs to neutralization by broadly and poorly neutralizing antibodies was examined. The broadly neutralizing antibodies (bNAbs) in our panel included VRC01 and VRC03 against the CD4-binding site of gp120 (143, 144), PG16 against a quaternary V2 epitope (145), PGT121 against a V3-glycan epitope on gp120 (146), and 35O22 against the gp120-gp41 interface (147). The poorly neutralizing antibodies included 17b against a CD4-induced epitope (148), 19b against the gp120 V3 loop (137), 902090 against a V2 gp120 epitope (149), and F105 against the CD4-binding site of gp120 (150). The 2–4 R and 2–4 RED2 viruses were neutralized by bNAbs comparably to the wild-type HIV-1_AD8; like the wild-type HIV-1_AD8, the 2–4 R and 2–4 RED2 viruses were resistant to poorly neutralizing antibodies (Fig. 2B).

The sensitivity of HIV-1 to inactivation by exposure to cold, sCD4, or CD4-mimetic compounds can provide an indication of Env “triggerability,” the tendency to make transitions from state 1 (23, 37, 45, 118, 120–125). Compared with the wild-type HIV-1_AD8, the 2–4 R virus displayed slight but reproducible resistance to cold, sCD4, and BNM-III-170, a CD4-mimetic compound (151) (Fig. 2C). The 2–4 RED2 virus exhibited an even higher level of resistance to cold, sCD4, and BNM-III-170 than either the wild-type or the 2–4 R virus. These phenotypes are consistent with the stability of the state-1 Env conformation exhibiting the following rank order in these variants: 2–4 RED2 > 2–4 R > wild-type HIV-1_AD8.

Q114E and Q576K changes determine state-1-stabilizing phenotypes.

We wished to identify the changes in 2–4 RED2 and 2–4 R responsible for the above-mentioned phenotypes. Because differences among the wild-type HIV-1_AD8, 2–4 R, and 2–4 RED2 Envs were most apparent in the Ni-NTA coprecipitation and virus sensitivity experiments, we used these assays to characterize HIV-1_AD8 Env mutants with single-residue changes corresponding to those in the 2–4 R and 2–4 RED2 Envs. Among the acidic residue substitutions found in the ED2 set, a single change, Q114E, was sufficient to recapitulate the 2–4 RED2 Env phenotypes (Fig. 3A). Similarly, a single lysine substitution originally found in set 4, Q576K, was responsible for most of the 2–4 R Env phenotypes (Fig. 3B). Thus, Q114E or Q567K alone can enhance HIV-1_AD8 Env processing, gp120-trimer association, and virus resistance to cold, sCD4, and a CD4-mimetic compound.

FIG 3.

Major contributions of the Q114E and Q567K changes to the respective 2–4 RED2 and 2–4 R phenotypes. (A) The effects of the Q114E change on gp120-trimer association (left panel) and virus sensitivity to cold, sCD4, and BNM-III-170 (right panels) were measured as described in the legend to Fig. 2. The sensitivities of viruses with the wild-type HIV-1_AD8 Env and the 2–4 RED2 Env are shown for comparison. (B) The effects of the Q567K change on gp120-trimer association (left panel) and virus sensitivity to cold, sCD4, and BNM-III-170 (right panels) were measured. The sensitivities of viruses with the wild-type HIV-1_AD8 Env and 2–4 R Env are shown for comparison. The results shown are typical of those obtained in at least two independent experiments. The right panels report the means and standard deviations derived from triplicate measurements.

Gln 114 is located in the gp120 α1 helix, part of the gp120 inner domain that faces the trimer axis and interacts with gp41 (79–82, 152–155). Gln 567 resides in the N-terminal segment of the gp41 heptad repeat 1 (HR1_N) region, which participates in the formation of the gp41 coiled coil after CD4 binding (32–34). In the available Env trimer structures, which have been suggested to represent a state-2-like conformation (78), the HR1_N region is disordered or structurally heterogeneous (79–89). Although structural information on Gln 114 and Gln 567 in the context of a state-1 Env is currently unavailable, based on their approximate location near the trimer axis and the charge complementarity of the substitutions yielding similar phenotypes, we tested their functional dependence. The phenotypes of a panel of 18 single and double Q114/Q567 Env variants were characterized (Table 1). Only acidic residue substitutions at position 114 resulted in an improvement of the constellation of state-1-associated phenotypes. At position 567, lysine substitution yielded the strongest state-1-associated phenotypes, while arginine substitution exerted a more modest effect. Analysis of the double mutants yielded two insights. First, the phenotypes of the Q114E mutant were not significantly affected by changing Gln 567 to an alanine residue. Likewise, the phenotypes of the Q567K mutant were similar to those of the Q567K Q114A double mutant. Therefore, the state-1-associated phenotypes of the Q114E and Q567K mutants are not dependent on the formation of hydrogen bonds between the side chains of residues 114 and 567. Second, the phenotypic effects of the changes in residues 114 and 567 were additive. Combination of the strongest individual changes yielded the Q114E Q567K variant, with the most pronounced phenotype. Both changes are found in the 2–4 RED2 Env. In summary, the Q114E and Q567K changes independently impart their individual effects on Env function and these effects are additive.

TABLE 1.

Phenotypes of HIV-1_AD8 Env variants with changes in Gln 114 and Gln 567^a

Env	Env processing	Infectivity (%)	Resistance/sensitivity compared to wild type			gp120-trimer association
			Cold	sCD4	BNM-III-170
Wild type	●	●	●	●	●	●
Q114A	●	●	●	●	●	+
Q114D	+	●	RR	RR	RR	++
Q114E	+	+	RR	RR	RR	++
Q114K	None	ND	ND	ND	ND	NA
Q114N	●	ND	ND	ND	ND	●
Q114S	●	ND	ND	ND	ND	●
Q567A	●	+	●	●	●	●
Q567E	●	− − −	●	●	●	●
Q567K	+	●	R	R	R	++
Q567R	+	−	Slight R	●	●	+
Q114A Q567K	●	●	Slight R	R	R	++
Q114A Q567R	●	●	●	●	●	+
Q114D Q567R	+	●	R	RR	RR	++
Q114E Q567A	+	+	RR	RR	RR	++
Q114E Q567K	+	●	RRR	RRR	RRR	++
Q114E Q567E	+	−	R	●	R	++
Q114E Q567R	+	●	RR	RR	RR	++
Q114K Q567E	None	ND	ND	ND	ND	NA

^aThe phenotypes of the wild-type HIV-1_AD8 Env and the indicated Gln 114 and Gln 567 variants were determined as described in the legend to Fig. 2. The values for Env processing efficiency, virus infectivity, and gp120-trimer association, relative to those observed for the wild-type HIV-1_AD8 Env, are shown. The sensitivity or resistance of viruses with the Env variants to cold, sCD4, and BNM-III-170 is reported relative to that of the wild-type HIV-1_AD8 virus. To ensure accurate comparison of the Env variant phenotypes across multiple assays, the wild-type HIV-1_AD8 and key Env mutants (e.g., Q114E or Q567K) were included in all assays. Phenotypes are labeled as follows: •, wild-type level; +, increase; −, decrease; R, resistant; S, sensitive; ND, not determined; NA, not applicable. For virus infectivity: − − −, 0 to 25% of wild-type level; −, 51 to 75%; •, 76 to 125%; +, >125%. An increase in the number of R symbols indicates a greater level of resistancecompared to the wild-type level. The levels of gp120-trimer association exhibit the following order: ++ > + > •, where • indicates the wild-type level. The data shown are representative of results obtained in at least two independent experiments.

We extended our mutagenesis approach to evaluate the potential of other Env residues to influence the Q114E and Q567K phenotypes. A state-1 Env structure would be most relevant to the search for interacting partners but is currently not available. Therefore, we used the available structural models, many of which represent state-2-like Env conformations (78), to suggest candidate amino acid residues. In sgp140 SOSIP.664 trimers, the highly conserved His 72 is located ∼8 Å from Gln 114 (79–81). Replacing His 72 with lysine or glutamine residues resulted in increased sensitivity to sCD4 and BNM-III-170; these phenotypes were partially relieved when these His 72 changes were combined with Q114E (Table 2). Replacing His 72 with an alanine residue resulted in a virus with a neutralization sensitivity similar to that of the wild-type virus. Compared with the Q114E virus, the H72A Q114E virus was less resistant to cold and sCD4. Thus, some changes in His 72 result in an apparent increase in Env triggerability and can influence the Q114E phenotypes.

TABLE 2.

Effects of Env amino acid changes on the phenotypes of the Q114E and Q567K Env variants^a

Env	Env processing	Infectivity (%)	Resistance/sensitivity compared to wild type			gp120-trimer association
			Cold	sCD4	BNM-III-170
Wild type	●	●	●	●	●	●
Q114E	+	+	RR	RR	RR	++
Q567K	●	●	R	R	R	++
Q114E Q567K	+	●	RRR	RRR	RRR	++
H72A	− −	− − −	R	●	●	NA
H72K	− −	− − −	●	SS	SS	NA
H72Q	− −	− − −	●	ND	SS	NA
H72A Q114E	−	− −	R	●	RR	−
H72K Q114E	−	− − −	●	S	SS	−
H72Q Q114E	−	−	R	S	●	−
T202K	●	−	SS	S	SS	●
T202R	●	−	SS	S	SS	●
T202A	●	−	SSS	SSS	SSS	−
T202Q	●	●	SSS	SSS	SSS	−
Q114E T202K	+	+	SS	S	SS	+
Q114E T202R	+	+	SS	S	SS	+
Q114E T202A	●	●	SSS	SS	SSS	●
Q114E T202Q	●	+	SSS	SS	SSS	●
Q203A	●	●	SS	ND	SSS	−
Q114E Q203A	+	●	●	ND	●	●
K117A	●	●	RR	ND	RR	●
K117Q	ND	−	R	ND	RR	ND
K121A	●	− −	●	ND	RR	−
K121Q	ND	−	R	ND	RR	ND
Q114E K117A	●	●	RR	ND	RR	●
Q114E K121A	+	−	Slight R	ND	RR	+
H66N	●	−	R	RRR	RRR	●
A582T	●	●	RRR	RR	RR	+
L587A	−	●	RR	R	RR	−
Q114E H66N	+	●	RR	RRR	RRR	++
Q114E A582T	+	●	RRR	RR	RRR	++
Q567K A582T	●	●	RRR	RR	RRR	+
Q114E L587A	+	●	RR	RR	RR	●
Q114E Q567K A582T	+	+	RRRR	RRRR	RRRR	++
Q114E Q567K L587A	+	−	RR	RR	RRR	+

^aThe phenotypes of the wild-type and mutant HIV-1_AD8 Envs were determined as described in the legend to Fig. 2. The values, relative to those of the wild-type HIV-1_AD8 Env, are reported as described in Table 1, footnote a. The phenotypes are labeled as follows: •, wild-type level; +, increase; −, decrease; R, resistant; S, sensitive; ND, not determined; NA, not applicable. Relative to the wild-type level (•), the levels exhibit the following order: ++ > + >•> − > − − > − − −. For virus infectivity: − − −, 0 to 25% of wild-type level; − −, 26 to 50%; −, 51 to 75%; •, 76 to 125%; +, >125%. An increase in the number of R or S symbols indicates a greater level of resistance or sensitivity, respectively, compared to the wild-type level. The data shown are representative of results obtained in at least two independent experiments.

In HIV-1/SIV_cpz Envs, Thr/Lys polymorphism in residue 202 often exhibits covariance with Gln/Glu polymorphism in residue 114 (136). Compared with the wild-type HIV-1_AD8, viruses with Thr 202 replaced by alanine, lysine, arginine, or glutamine residues were more sensitive to cold, BNM-III-170, and the 19b anti-V3 antibody (Table 2 and data not shown). These phenotypes, which are indicative of increased Env triggerability and state-1 destabilization, were minimally compensated by the addition of the Q114E change. Replacing the conserved Gln 203 residue with an alanine residue (Q203A in Table 2) also resulted in a state-1-destabilized phenotype, but in this case, the Q114E Q203A mutant exhibited phenotypes close to that of the wild-type HIV-1_AD8. Thus, the Q114E change can compensate for some but not all state-1-destabilizing changes.

In the unliganded sgp140 SOSIP.664 and PGT151-bound EnvΔCT structures (PDB ID 4ZMJ and 5FUU, respectively) (82, 86), the side chains of Gln 114, Lys 117, and Lys 121 from each Env protomer point toward the trimer axis, stacking in three layers. Interprotomer Lys 117-Lys 117 and Lys 121-Lys 121 cross-links were formed in a cross-linking/mass spectrometry study of the sgp140 SOSIP.664 trimer, confirming the location of these residues in the trimer core in these Env structures (73). Substitution of Lys 117 or Lys 121 with an alanine or glutamine residue resulted in viruses that were more resistant to cold and BNM-III-170 than the wild-type virus (Table 2). No additive or synergistic effect was observed when the Q114E change was combined with the K117A or K121A changes. In fact, the double mutants exhibited less stable association of gp120 with solubilized Env trimers (Table 2). Thus, the effects of the Gln 114, Lys 117, and Lys 121 changes on the viral phenotypes are redundant, whereas in the detergent-solubilized Envs, the K117A and K121A changes nullify the trimer-stabilizing effects of the Q114E change. Similar phenotypic effects of the K117A and K121A changes were observed in the context of the E.2 and AE.2 HIV-1_AD8 constructs discussed below (Table 3).

TABLE 3.

Effects of Env amino acid changes on the phenotypes of the E.2, AE.2, and 2–4 RM6 AE Env variants^a

Env	Env processing	Infectivity (%)	Resistance/sensitivity compared to wild type			gp120-trimer association
			Cold	sCD4	BNM-III-170
Wild type	●	●	●	●	●	●
E.2	+	+	RRR	RRR	RRR	++
AE.2	+	−	RRRR	RRRR	RRRR	++
E.2 K117A	+	●	RRR	RRR	RRR	−
AE.2 K117A	+	− −	RRRR	RRRR	RRRR	+
K59A	●	●	R	R	RR	●
V255I	●	●	RR	NA*	NA*	●
E.2 K59A	+	−	RRR	RRR	RRR	++
AE.2 K59A	+	− −	RRRR	RRRR	RRRR	++
E.2 K59A V255I	+	− − −	RRRR	NA*	NA*	++
E.2 V255I	+	−	RRRR	NA*	NA*	++
AE.2 V255I	+	− − −	RRRR	NA*	NA*	++
AE.2 K59A V255I	+	− − −	RRRR	NA*	NA*	++
K59A Q114E V255I	+	− −	RRRR	NA*	NA*	++
R542A	−	− − −	S	SSS	SSS	−
I595F	−	−	R	SSS	SSS	−
L602H	−	−	S	SSS	SSS	−
2−4 RM6 AE	+	−	RRR	RRR	RRR	++
2−4 RM6 AE R542A	+	− − −	RR	RRR	RR	++
2−4 RM6 AE I595F	+	− − −	R	●	●	++
2−4 RM6 AE L602H	+	− −	RRR	RRR	RR	++

^aThe indicated amino acid changes were introduced into the HIV-1_AD8 Env or into the E.2, AE.2, or 2–4 RM6 AE Envs. The phenotypes of these Env variants were determined as described in the legend to Fig. 2. The phenotypes are labeled as follows: •, wild-type level; +, increase; −, decrease; R, resistant; S, sensitive. Relative to the wild-type level (•), the levels exhibit the following order: ++ > + >•> − > − − > − − −. For virus infectivity: − − −, 0 to 25% of wild-type level; − −, 26 to 50%; -, 51 to 75%; •, 76 to 125%; +, >125%. An increase in the number of R or S symbols indicates a greater level of resistance or sensitivity, respectively, compared to the wild-type level. The values, relative to those of the wild-type HIV-1_AD8 Env, are reported as described in Table 1, footnote a. The data shown are representative of results obtained in at least two independent experiments. *NA (not applicable); because Val 255 is near the binding site for sCD4 and the CD4-mimetic compounds, the V255I change may directly decrease the binding of these Env ligands. Therefore, the sensitivity of Env mutants containing the V255I change to sCD4 and BNM-III-170 does not allow conclusions about the conformational state of these Env variants.

As Gln 567 is disordered in most Env trimer structures, we used a low-resolution model of the uncleaved HIV-1_JR-FL Env (156) to suggest potential interaction partners. However, alanine substitutions in these potentially interacting HIV-1_AD8 residues (Glu 47, Glu 83, Glu 87, Glu 91, Asp230, Glu 492, and Glu 560) did not affect the phenotypes of the Q567K mutant virus (data not shown).

Q114E and Q567K synergize with other state-1-stabilizing Env changes.

Previous studies suggested that changes in His 66, Ala 582, and Leu 587 could enrich the state-1 HIV-1_YU2 Env conformation through different proposed mechanisms: H66N destabilizes the CD4-bound conformation, A582T directly stabilizes the pretriggered conformation, and L587A destabilizes the gp41 3-helix bundle (121, 122, 125). We confirmed that individually these changes increased HIV-1_AD8 resistance to cold, sCD4, and BNM-III-170 (Table 2). Of the three changes, only A582T enhanced gp120-trimer association in cell lysates. Both the H66N and A582T changes synergized with the Q114E and Q567K changes in producing viral phenotypes associated with state-1 stabilization (Table 2). A combination of three changes in the Q114E Q567K A582T Env resulted in the most robust phenotypes.

Cross-linkable E.2 and AE.2 Envs with enhanced state-1 stability.

To generate HIV-1 Envs enriched in a pretriggered conformation and containing multiple lysine residues for cross-linking, we added two lysine substitutions (R252K, A667K) and Q114E to the lysine-rich 2–4 R Env to create the E.2 Env construct (Fig. 1). The AE.2 Env contains, in addition, the A582T change. The A582T change was chosen because it not only resulted in viral phenotypes additive with those of Q114E and Q567K but also increased gp120 association with the detergent-solubilized Env, a property that K117A, K121A, H66N, and L587A lacked (Table 2). Both E.2 and AE.2 Env were cleaved more efficiently than the wild-type HIV-1_AD8 Env and resisted gp120 dissociation from the solubilized Env trimer (Fig. 4A). By comparison, the wild-type Env from another primary strain, HIV-1_JR-FL, was poorly processed and highly unstable in detergent.

FIG 4.

Phenotypes of the E.2 and AE.2 Envs. (A) HOS cells transiently expressing His₆-tagged Envs (wild-type HIV-1_AD8 Env, the E.2 Env, the AE.2 Env, or the HIV-1_JR-FL Env) were lysed. Cell lysates were incubated with Ni-NTA beads for 1.5 h at 4°C in the presence of the DMSO control or 10 μM BMS-806. Total cell lysates (input) and Ni-NTA-bound proteins were Western blotted with a goat anti-gp120 antibody (upper panels) and the 4E10 anti-gp41 antibody (lower panels). (B) Recombinant luciferase-expressing viruses with the indicated Envs were incubated with antibodies for 1 h at 37°C, after which the mixture was added to Cf2Th-CD4/CCR5 target cells. Forty-eight hours later, the target cells were lysed and the luciferase activity was measured. The IC₅₀ values were calculated using the GraphPad Prism program. (C) Recombinant luciferase-expressing viruses with the indicated Envs were incubated with sCD4 or BNM-III-170 for 1 h at 37°C before the mixture was added to Cf2Th-CD4/CCR5 target cells. Cold sensitivity was assessed by incubation of the viruses on ice for the indicated times, after which virus infectivity was measured on Cf2Th-CD4/CCR5 cells as described above. The results are representative of those obtained in at least two independent experiments. The values reported in panels B and C represent the means and standard deviations from at least two independent experiments or triplicate measurements, respectively.

To evaluate the functional E.2 and AE.2 Envs in more detail, we tested virus sensitivity to a panel of broadly and poorly neutralizing antibodies. In addition to the antibodies used in Fig. 2, we included two bNAbs, PGT151 against the gp120-gp41 interface (157) and PGT145 against a quaternary V2 epitope (158), and the poorly neutralizing F240 antibody against gp41 (159). Envs from the clade B HIV-1_JR-FL and clade A HIV-1_BG505 tier 2/3 strains were included for comparison. All Env variants resisted neutralization by poorly neutralizing antibodies, as expected (Fig. 4B). Compared to the wild-type HIV-1_AD8, HIV-1_JR-FL, and HIV-1_BG505, the E.2 and AE.2 viruses were just as sensitive, and even more sensitive in some cases, to neutralization by broadly neutralizing antibodies.

The sensitivity of the viruses to cold inactivation, sCD4, and the CD4-mimetic compound BNM-III-170 is shown in Fig. 4C. Compared with the wild-type HIV-1_AD8, the E.2 virus exhibited increased resistance to cold, sCD4, and BNM-III-170. Alteration of Glu 114 in the E.2 Env to glutamine largely reverted these phenotypes, suggesting that the Q114E change is a critical determinant of the stabilized pretriggered conformation in the E.2 Env (data not shown). The inclusion of the A582T change in the AE.2 Env further increased cold, sCD4, and BNM-III-170 resistance to the levels of the tier 2/3 HIV-1_JR-FL and HIV-1_BG505 strains. In addition, the E.2 and AE.2 viruses were more sensitive than the wild-type HIV-1_AD8 to the state-1-preferring entry inhibitors, BMS-806 and 484 (45, 118); the AE.2 virus was more sensitive to these small-molecule inhibitors than the E.2 virus (data not shown).

In an attempt to improve the E.2 and AE.2 Envs further, we added the K59A and/or V255I changes. Lysine 59 is a highly conserved residue in the gp120 inner domain, within the disulfide loop (layer 1) that includes His 66, discussed above. Valine 255 packs against the critical Trp 112 and Trp 427 residues in the CD4-binding Phe 43 cavity of gp120 (152); the V255I change was associated with resistance to AAR029b, a cyclic peptide triazole inhibitor of CD4 binding (160). The K59A and V255I changes alone rendered HIV-1_AD8 more cold resistant, and the K59A virus was also relatively resistant to sCD4 and BNM-III-170 (Table 3). However, the K59A and V255I changes had only modest effects in the E.2 and AE.2 background on state-1-associated phenotypes but led to significant reductions in infectivity (Table 3). These observations hint that further stabilization of state-1-associated phenotypes in the AE.2 context may be accompanied by decreases in Env function.

Effects of state-1-destabilizing changes in different Env contexts.

In the above-described studies, the Q114E change could revert the viral phenotypes associated with state-1 destabilization by the Q203A change but not by changes in the adjacent Thr 202 residue (Table 2). We evaluated whether an Env with multiple state-1-stabilizing changes, 2–4 RM6 AE, would better tolerate state-1 destabilization. The 2–4 RM6 AE and AE.2 Envs are identical except for the R252K change in the latter (Fig. 1). The 2–4 RM6 AE virus is resistant to cold, sCD4, and BNM-III-170 and exhibits strong gp120-trimer association in detergent (Table 3). We individually introduced the R542A, I595F, and L602H changes into the wild-type HIV-1_AD8 Env or the 2–4 RM6 AE Env. These gp41 changes rendered HIV-1 more sensitive to the nonpeptidic inhibitory compound RPR103611, which suggested that they might destabilize the pretriggered (state-1) Env conformation (161). In agreement with this hypothesis, the R542A and L602H HIV-1_AD8 variants exhibited increased sensitivity to cold, sCD4, and BNM-III-170 relative to HIV-1_AD8 (Table 3). The I595F virus was sensitive to sCD4 and BNM-III-170 as well as to the 19b anti-V3 antibody but was slightly more resistant to cold inactivation than HIV-1_AD8. Interestingly, increases in sensitivity to cold, sCD4, BNM-III-170, and 19b associated with these gp41 changes were not evident in the 2–4 RM6 AE background (Table 3). Thus, the state-1-stabilizing changes in 2–4 RM6 AE apparently resist the state-1-destabilizing effects of the R542A, I595F, and L602H changes in the gp41 ectodomain.

Correlations among key Env phenotypes.

To understand the relationships among key Env phenotypes and to visualize the effects of specific amino acid changes on the progression of successive generations of Env mutants, we plotted the relative levels of resistance to cold, BNM-III-170, and gp120-trimer dissociation for all characterized Env variants (Fig. 5). Virus resistance to cold inactivation reflects the stability of the functional Env trimer on virions and is independent of the binding of an Env ligand. Virus resistance to the CD4-mimetic compound generally correlates with resistance to sCD4 (122, 154, 162). Of interest, there exists a strong correlation between virus resistance to the CD4-mimetic compound and to cold (Fig. 5). Beginning with the wild-type HIV-1_AD8 Env, Envs incorporating additive state-1-stabilizing changes displayed upward shifts toward highly resistant phenotypes, comparable to those of the HIV-1_JR-FL and HIV-1_BG505 Envs. Envs with state-1-destabilizing changes grouped together in the lower left quadrant.

FIG 5.

Correlations among key Env phenotypes. The plot shows the relative level of resistance to the CD4-mimetic compound BNM-III-170 versus the relative level of cold resistance for the HIV-1 Env variants tested in this study. The levels of resistance are scored as described in the footnotes to Tables 1 to 3: •, wild-type level; R, resistant; S, sensitive. Key Env variants are designated with stars. Envs are colored according to their relative gp120-trimer association level reported in Tables 1 to 3, as measured by Ni-NTA coprecipitation of gp120 with the His₆-tagged gp41 glycoprotein: black, wild-type level; blue, not determined or not applicable; light green, +; green, ++; red, −. The V255I change is located near the binding site for CD4-mimetic compounds and may directly affect Env interaction with BNM-III-170. Note that the E.2 and AE.2 Envs exhibit resistance to cold and BNM-III-170 comparable to those of the HIV-1_JR-FL and HIV-1_BG505 Envs but also display significantly better gp160 processing. Relative to the HIV-1_JR-FL Env, the E.2 and AE.2 Envs exhibit a tighter association of gp120 with the Env trimer solubilized in detergent.

Env variants that exhibited a higher level of gp120-trimer association in detergent, relative to that of the wild-type HIV-1_AD8 Env, are colored green in Fig. 5. The skewed distribution of these Env variants in the upper right quadrant indicates that a tighter association of gp120 with the solubilized Env trimer is related to virus resistance to cold and BNM-III-170, phenotypes associated with state-1 stabilization. Note that several HIV-1_AD8 Env variants and the natural HIV-1_JR-FL Env achieve virus resistance to cold and BNM-III-170 without increasing gp120-trimer association in detergent-solubilized Envs (Fig. 5, black symbols in the upper right quadrant). Therefore, increasing gp120-trimer association is not the only means of achieving a more stable pretriggered (state-1) Env conformation.

Cross-linking efficiency of the wild-type AD8, E.2, and AE.2 Envs.

The lysine-rich E.2 and AE.2 Envs are expected to cross-link more efficiently than the wild-type HIV-1_AD8 Env with lysine-reactive cross-linkers like 3,3′-dithiobis (sulfosuccinimidyl propionate) [DTSSP] and glutaraldehyde. DTSSP has a spacer arm of 12 Å, whereas, because of its tendency to polymerize, glutaraldehyde forms cross-links of more variable lengths (163). Both DTSSP and glutaraldehyde cross-linked the E.2 and AE.2 Envs more efficiently than the wild-type AD8 Env (Fig. 6A). For example, after treatment with 5 mM glutaraldehyde, the E.2 and AE.2 Envs cross-linked into gel-stable trimers, whereas the wild-type HIV-1_AD8 Env mostly formed monomers and dimers. Apparently, a greater number of lysine residues accessible to the cross-linkers exist on the E.2 and AE.2 Env trimers than on the wild-type HIV-1_AD8 Env.

FIG 6.

Cross-linking of membrane and soluble Envs with state-1-stabilizing changes. (A) VLPs consisting of the HIV-1 Gag-mCherry fusion protein and the wild-type HIV-1_AD8 Env, the E.2 Env, or the AE.2 Env were incubated with different concentrations of the DTSSP or glutaraldehyde cross-linkers. After the reactions were quenched, VLPs were pelleted and lysed. The VLP proteins cross-linked with DTSSP or glutaraldehyde were analyzed by nonreducing or reducing PAGE, respectively, followed by Western blotting. The ratio of gel-stable (dimers + trimers) to (gp120 + gp160) provides an indication of interprotomer cross-linking by DTSSP (upper right panel). (B) Luciferase-expressing viruses pseudotyped with the wild-type HIV-1_AD8, E.2, or AE.2 Envs were incubated with the BS3 cross-linker. After quenching the reaction, the viruses were added to Cf2Th-CD4/CCR5 cells. Luciferase activity in the target cells was measured 48 h later. (C) The HIV-1_AD8 sgp140 SOSIP.664 and sgp140 SOSIP.664 Q114E/Q567K/A582T glycoproteins in the supernatants of transfected 293T cells were incubated with the indicated concentrations of the BS3 cross-linker and then analyzed by Western blotting. The percentage of the sgp140 SOSIP.664 glycoproteins that cross-link into trimers is plotted in the graph on the right. The results are representative of those obtained in at least two independent experiments. The values reported in panel B represent the means and standard deviations derived from triplicate experiments.

We also examined the relative sensitivity of the functional wild-type HIV-1_AD8, E.2, and AE.2 Envs to bissulfosuccinimidyl suberate (BS3), another lysine-specific cross-linker with spacer arms of 12 Å. The infectivity of viruses pseudotyped with the E.2 and AE.2 Envs was inhibited by BS3 at 3- to 4-fold lower concentrations than those required for inhibition of viruses with the wild-type HIV-1_AD8 Env (Fig. 6B). These results suggest that BS3 cross-links occur more efficiently on the E.2 and AE.2 Envs than on the wild-type HIV-1_AD8 Env, leading to a loss of infectivity at lower BS3 concentrations.

We tested the potential effect of the three amino acid changes that stabilize state 1 in the HIV-1_AD8 membrane Env on the integrity of a soluble Env trimer. Envs from different HIV-1 strains form sgp140 SOSIP.664 trimers with various levels of efficiency (63, 164) and could potentially benefit from additional stabilizing modifications. The effect of the three key state-1-stabilizing Env changes (Q114E/Q567K/A582T) on the oligomerization of the HIV-1_AD8 sgp140 SOSIP.664 glycoprotein was examined by cross-linking with saturating levels of BS3. Compared with the HIV-1_AD8 sgp140 SOSIP.664 glycoprotein, sgp140 SOSIP.664 with the Q114E/Q567K/A582T changes exhibited 2- to 4-fold more trimers (Fig. 6C). This result indicates that the Q114E, Q567K, and A582T changes can enhance the trimerization of the sgp140 SOSIP.664 glycoproteins.

DISCUSSION

Despite more than 3 decades of intense research, an effective HIV-1 vaccine remains elusive. The metastability and multiple conformational states of the HIV-1 Env create challenges for the generation of broadly neutralizing antibodies, either following vaccination or during natural HIV-1 infection. In Env-expressing cells, both uncleaved and cleaved (mature) Envs are present on the cell surface. A significant fraction of the uncleaved Env bypasses the conventional Golgi secretory pathway to traffic to the cell surface; these Envs differ from mature Envs in glycan processing, conformation, and recognition by antibodies (7). Uncleaved Envs may function as a decoy to the host immune system and divert antibody responses away from the mature Envs. The pretriggered (state-1) conformation of the mature virion Env of primary HIV-1 strains is the target for most broadly neutralizing antibodies (12, 37, 38, 45). This native conformation, however, is unstable and can transition into more open state-2/3 conformations that are able to be recognized by poorly neutralizing antibodies. Therefore, it is of significant interest to devise methods to lock Env in its native state-1 conformation by means that resist perturbation during Env purification, characterization, and immunization.

Here, we tackled the challenges posed by HIV-1 Env conformational flexibility in two ways. First, we used polymorphisms in naturally occurring HIV-1 strains to guide the introduction of extra lysine and acidic amino acid residues in the HIV-1_AD8 Env. Chemical cross-linkers that couple lysine or acidic residues on proteins under physiological conditions are available (132–135). During the iterative process employed to identify HIV-1_AD8 Envs that are potentially more susceptible to cross-linking, we required that the Env variants exhibit efficient processing, subunit association, and the ability to support virus entry. Some of the functional HIV-1_AD8 Env variants developed by this approach contain up to 11 extra lysine residues (33 per Env trimer) and up to 7 extra acidic residues (21 per Env trimer). Using DTSSP or glutaraldehyde as cross-linking agents, two Env variants, E.2 and AE.2, were shown to form interprotomer cross-links more efficiently than the wild-type HIV-1_AD8 Env. The infectivity of viruses with these Envs was inactivated more efficiently than that of viruses with the wild-type HIV-1_AD8 Env by another lysine-specific cross-linker, BS3. These assays document the accessibility of some of the additional lysine residues introduced into the E.2 and AE.2 Envs. Chemical cross-linking can enrich the representation of otherwise labile native conformations in Env preparations for structural analysis or immunogenicity studies. Cross-linking/mass spectrometry can provide distance constraints between Env residues that can be used to validate available structural models or to derive new models (132–135). A previous study utilized cross-linking/mass spectrometry to detect differences between soluble and membrane-bound Envs (73). The inclusion of the 2–4 RED2, E.2, and AE.2 Envs in future cross-linking/mass spectrometry studies should increase the number of distance constraints and thereby improve our ability to discriminate among alternative structural models.

The second strategy employed in our approach was to screen the Env variants for function and viral phenotypes associated with stabilization of a state-1 Env conformation. For this purpose, we evaluated viral resistance to cold, sCD4, and the CD4-mimetic compound BNM-III-170. Cold inactivation reflects the resistance of the functional HIV-1 Env trimer to the detrimental effects of ice formation at near-freezing temperatures (165–167). The sensitivity of HIV-1 variants to cold inactivation is related to the intrinsic reactivity or triggerability of Env; Envs that more readily make the transition from state 1 to downstream conformations are invariably cold sensitive (23, 121, 122). HIV-1 sensitivity to sCD4 and BNM-III-170 inhibition is a function of Env triggerability (23, 45, 118, 119, 125); because we generally avoided changes to the highly conserved and well-defined BNM-III-170 binding site on gp120 (151, 162), most of the observed differences in virus sensitivity to this CD4-mimetic compound reflect changes in the ability of Env to negotiate transitions from a state-1 conformation. Our study documents the strong correlation between HIV-1 resistance to cold and resistance to BNM-III-170. This screening strategy identified two changes, Q114E in gp120 and Q567K in gp41, that individually increased the resistance of the HIV-1_AD8 Env to inactivation by cold, sCD4, and BNM-III-170. These viral phenotypes were additively enhanced by combining the Q114E and Q567K changes in Env variants, such as the lysine-rich E.2 and AE.2 Envs. Cold, sCD4, and BNM-III-170 resistance were further increased by the inclusion in the AE.2 Env of the A582T gp41 change, which previously was shown to stabilize a pretriggered Env conformation (123, 125). Viruses with the E.2 and AE.2 Envs were inhibited efficiently by BMS-806, a small molecule that exhibits some preference for a state-1 Env conformation (12, 78, 138–140). The functional E.2 and AE.2 Envs exhibited an antigenic profile consistent with a state-1 conformation, conferring virus sensitivity to bNAbs that target quaternary epitopes (PG16, PGT145, PGT151, 35O22) and resistance to poorly neutralizing antibodies (17b, 19b, 902090, F105, F240). Of note, compared with the wild-type HIV-1_AD8, viruses with the E.2 and AE.2 Envs were more sensitive to neutralization by the 10E8.v4 bNAb directed against the gp41 membrane-proximal external region (MPER) (168) (Fig. 4B). Although 10E8.v4 can recognize Env prior to receptor engagement (169), in contrast to most MPER-directed bNAbs (170), stabilization of the state-1 Env conformation would not necessarily be expected to increase virus sensitivity to 10E8.v4 neutralization. Indeed, upon further investigation, we found that an HIV-1_AD8 mutant with the three main state-1-stabilizing changes (Q114E/Q567K/A582T) was inhibited by 10E8.v4 comparably to wild-type HIV-1_AD8 (171). Therefore, the relatively increased sensitivity of the E.2 and AE.2 viruses to 10E8.v4 neutralization is apparently due to changes in these Envs other than those related to state-1 stabilization. We note that both E.2 and AE.2 Envs have lysine substitutions within and near the gp41 MPER that could potentially affect 10E8.v4 bNAb recognition.

Two unanticipated beneficial phenotypes associated with the E.2 and AE.2 Envs are more efficient Env processing and greater stability of solubilized Env trimers. HIV-1 Env cleavage has been suggested to contribute to the stability of the state-1 conformation (138, 139, 172–177). As uncleaved HIV-1 Envs sample multiple conformations, including those reactive with poorly neutralizing antibodies, achieving a high level of gp120-gp41 processing may be important for an effective vaccine immunogen. The E.2 and AE.2 Envs achieve levels of state-1-associated phenotypes comparable to those of the tier 2/3 HIV-1_JR-FL and HIV-1_BG505 but. notably, are processed much more efficiently. In addition, relative to the wild-type HIV-1_AD8 and HIV-1_JR-FL Envs, the E.2 and AE.2 Envs solubilized in detergent exhibit much greater gp120 association with the Env trimer. The Q114E, Q567K, and A582T changes individually strengthen the noncovalent association of gp120 with the solubilized Env trimers, a property that will assist purification and characterization. Of interest, the Q567K change was included in a combination of Env changes that were reported to stabilize HIV-1 Env trimers in different contexts (178–180). In our panel of HIV-1 Env variants, enhancement of Env trimer stability in detergent was strongly correlated with virus resistance to cold and BNM-III-170, state-1-associated phenotypes. We note that the binding of the state-1-preferring compound, BMS-806, also stabilizes gp120-trimer association (138). In future studies, the ability of Q114E, Q567K, A582T, and other state-1-stabilizing changes to enhance Env cleavage efficiency, strengthen gp120-trimer association, and increase the sampling of pretriggered Env conformations in other HIV-1 strains will be explored. That some of the individual Env changes identified in our study demonstrate similar phenotypes in the Envs of other HIV-1 strains (121, 122, 125, 178–180) is encouraging in this regard.

In addition to Q114E and Q567K, we identified other changes (K59A, K117A, K121A) that individually yielded Env phenotypes consistent with state-1 stabilization. These and previously identified state-1-stabilizing changes (H66N, L587A) (121, 122, 125) were tested in combination with the Q114E and/or Q567K changes in various Env backgrounds. In no case did we observe an additive improvement in viral phenotypes associated with state-1 stabilization, and several of these combinations resulted in attenuated virus replication or gp120-trimer dissociation in detergent. It is not surprising that as state-1 stability is increased, virus replication diminishes as the activation barriers governing state-1-to-state-2 transitions increase. However, the validation of state-1 stabilization would be less straightforward for replication-incompetent Envs, given current uncertainties about a state-1 Env structure. Therefore, we deferred investigation of these potentially state-1-stabilizing changes until better assays to characterize the conformations of nonfunctional Envs are established.

The amino acid changes identified in our study that stabilize a state-1 conformation could potentially be applied to membrane Env immunogens presented in proteoliposomes or virus-like particles, or expressed in an mRNA vaccine (181, 182). Soluble forms of HIV-1 Env trimers have attracted significant interest as possible vaccine immunogens (61–69). Although differences exist between the conformations of membrane and soluble Env trimers (70–78), we evaluated the reciprocity of trimer-stabilizing changes in these two contexts. The changes that stabilize state 1 in the membrane HIV-1_AD8 Env (Q114E/Q567K/A582T) modestly increased the levels of trimerization of HIV-1_AD8 sgp140 SOSIP.664 glycoproteins.

We also evaluated the effect of changes that stabilize soluble gp140 trimers on the native HIV-1_AD8 Env. Changes in gp41 (I559P, L555P) that are intended to prevent the formation of the HR1 coiled coil have been used to stabilize sgp140 SOSIP.664 trimers (70, 88, 183). However, introduction of these changes in combination with the major state-1-stabilizing changes (Q114E/I559P, Q114E/Q567K/I559P, and Q114E/Q567K/L555P) resulted in membrane Envs that were not processed (data not shown). We also considered another gp41 change, Q658E, that has been shown to stabilize sgp140 SOSIP.664 trimers (184). Introduction of the Q658E change into the wild-type HIV-1_AD8 Env resulted in increased virus sensitivity to cold, sCD4, BNM-III-170, and the 19b antibody (data not shown). These phenotypes are consistent with those reported in other HIV-1 strains (184), and as they suggest a lower occupancy of state 1, we did not evaluate the Q658E change in combination with the Q114E and Q567K changes. These results indicate that not all conformationally stabilizing changes identified in soluble Env trimers are applicable to membrane Env glycoprotein trimers.

Although a state-1 Env structure is currently unknown, mapping the Env residues identified in this study on available Env trimer models can provide some insights. Figure 7 uses a PGT151-bound HIV-1_JR-FL EnvΔCT trimer structure (PDB ID 5FUU) (82) to show the locations of Env residues in which changes resulted in increases or decreases in state-1-associated phenotypes. The binding of the PGT151 antibody induces a state-2-like conformation that is asymmetric, with two antibody Fabs bound to the Env trimer (78, 82). We chose this structure because, unlike most HIV-1 trimer structures, the HR1_N region containing Gln 567 is resolved; however, in keeping with the asymmetry of the PGT151-bound Env trimer, the positions of the Gln 567 residues differ among the three Env protomers. Gln 567, Gln 114, and Ala 582 are close to the trimer axis in the EnvΔCT structure (Fig. 7A). The C_α-C_α distances between Gln 114 and Gln 567 residues vary from 11.6 to 15.2 Å, and the side chains of these residues do not apparently interact in this Env conformation. Gln 114 is stacked above Lys 117 and Lys 121, the side chains of which project toward the Env trimer axis (Fig. 7B, right panel). Although a precise structural explanation for the observed state-1-stabilizing phenotypes will require more data, the implicated residues are positioned near intersubunit or interprotomer junctions and therefore could potentially modulate trimer opening. For example, electrostatic repulsion among Lys 117 residues that destabilizes the Env trimer could be mitigated by their conversion to alanine residues or by replacing Gln114 with acidic residues.

FIG 7.

Location of Env residues in a structural model of an HIV-1 Env trimer. Env residues studied herein are depicted as CPK spheres in a PGT151-bound HIV-1_JR-FL Env ΔCT trimer (PDB ID 5FUU) (82). The binding of two PGT151 Fabs introduces asymmetry into the Env trimer. In this depiction, the PGT151 Fabs have been removed from the structure. The individual Env protomers are colored pink, light blue, and gray. In this orientation, the gp120 subunits are at the bottom and gp41 subunits are at the top of the figures. (A) Env residues (Gln 114 [magenta], Gln 567 [orange], and Ala 582 [blue]) associated with state-1-stabilizing changes are shown, with a closeup image in the panel on the right. The distances between the C_α atoms of Gln 114 and Gln 567 residues in this asymmetric trimer structure are 11.6, 13.1, and 15.2 Å. The HR1_N regions of the three Env protomers differ in conformation. (B) Env residues (Lys 59, His 66, Gln 114, Gln 567, and Ala 582) associated with state-1-stabilizing changes are colored green. Env residues (His 72, Thr 202, Gln 203, Arg 542, Ile 595, Leu 602, and Gln 658) associated with state-1-destabilizing changes are colored red. Changes in the residues (Lys 117, Lys 121, and Leu 587) colored yellow resulted in Envs that were resistant to cold and a CD4-mimetic compound but were subject to gp120 dissociation from the Env trimer solubilized in detergent. The right panel shows the side chain stacking of residues Gln 114, Lys 117, and Lys 121 near the Env trimer axis.

The state-1-destabilizing changes identified in this study (red residues in Fig. 7B, left panel) are less localized than the state-1-stabilizing changes (green and yellow residues in Fig. 7B, left panel). This is consistent with the expectation that a metastable structure can be disrupted by a diverse set of changes, whereas a more limited and strategically placed set of changes is required to strengthen the structure. In this study, we provide an example of how state-1-stabilizing changes in Env can counter the phenotypic effects of state-1-destabilizing alterations, even when these changes involve amino acid residues very distant on current Env trimer structures.

In a related paper (171), we report the ability of the state-1-stabilizing changes identified herein to counter the phenotypic consequences of disruption of the gp41 MPER. Although further work will be required to understand fully the mechanisms underlying these observations, the ability of the Q114E, Q567K, and A582T changes to counteract the disruptive effects of distant changes suggests that they may have significant utility in preserving pretriggered Env conformations in multiple circumstances.

MATERIALS AND METHODS

Env glycoprotein constructs.

The HIV-1_AD8 and HIV-1_JR-FL Envs were coexpressed with the Rev protein in the pSVIIIenv expression vector, using the natural HIV-1 env and rev sequences (23). The Asp718 (Kpn I)-BamHI fragment of HIV-1_AD8 env was cloned into the corresponding sites of the pSVIIIenv plasmid expressing the HIV-1_HXBc2 Env and Rev. Thus, the “wild-type” HIV-1_AD8 Env used in this study contains a signal peptide and part of the cytoplasmic tail from the HIV-1_HXBc2 Env. Viruses with this chimeric Env exhibit a degree of CD4 dependence and an antibody neutralization profile consistent with those of a primary, tier 2 HIV-1 (23, 37). The initial single, double, and triple sets of lysine substitutions shown in Fig. 1A were introduced into the HIV-1_AD8 Env lacking an epitope tag. A carboxy-terminal GGHHHHHH (His₆) epitope tag was added to the Env variants shown in Fig. 1B and derivatives thereof. The mutations were introduced by site-directed PCR mutagenesis using Pfu Ultra II polymerase (Agilent Technologies), according to the manufacturer’s protocol. The plasmid expressing the HIV-1_BG505 Env (BG505.W6M.ENV.C2) was obtained through the NIH HIV Reagent Program and was contributed by Julie Overbaugh.

To express sgp140 SOSIP.664 glycoproteins, the pSVIIIenv plasmid containing the wild-type HIV-1_AD8 env gene was mutagenized using Q5 high-fidelity DNA polymerase (New England Biolabs) according to the manufacturer’s instructions. The following changes were introduced into the wild-type HIV-1_AD8 Env: A501C, modification of the gp120-gp41 cleavage site (REKR) to RRRRRR, I559P, T605C, and a C-terminal GSGHHHHHH tag; the sgp140 SOSIP.664 glycoprotein is truncated due to a stop codon replacing the codon for Asp664 (73, 183). The Q114E, Q567K, and A582T changes were introduced into this sgp140 SOSIP.664 construct.

Cell lines.

293T cells (ATCC) and HOS cells (ATCC) were grown in Dulbecco’s modified Eagle’s medium/nutrient mixture F12 (DMEM-F12) supplemented with 10% fetal bovine serum (FBS) and 100 μg/mL of penicillin-streptomycin. A549 cells expressing HIV-1 Envs with Gag-mCherry fusion proteins were grown in DMEM-F12 medium supplemented with 10% FBS, 1× penicillin-streptomycin, 1× l-glutamine, and 0.2% amphotericin B. Cf2Th cells stably expressing the human CD4 and CCR5 coreceptors for HIV-1 were grown in the same medium supplemented with 0.4 mg/mL of G418 and 0.2 mg/mL of hygromycin. All cell culture reagents are from Life Technologies.

Env processing and gp120-trimer association in Ni-NTA precipitation assay.

HOS cells were cotransfected with a Rev/Env-encoding pSVIIIenv plasmid and a Tat-encoding plasmid at a 1:0.125 ratio using the Effectene transfection reagent (Qiagen). At 48 h after transfection, HOS cells were washed with 1× phosphate-buffered saline (PBS) and lysed in 100 mM (NH₄)₂SO₄, 20 mM Tris-HCl, pH 8, 300 mM NaCl, and 1.5% Cymal-5 (Anatrace) containing DMSO, 10 μM BMS-806, or 10 μg/mL soluble CD4-Ig. Lysates were clarified, and aliquots were saved as the input samples. The remaining lysates were incubated with nickel-nitriloacetic acid (Ni-NTA) beads (Qiagen) for 1.5 h at 4°C. The beads were gently pelleted and washed 3 times with room temperature washing buffer [100 mM (NH₄)₂SO₄, 20 mM Tris-HCl, pH 8, 1 M NaCl, and 0.5% Cymal-5]. The beads were then boiled in LDS sample buffer (Invitrogen), and the proteins were analyzed by Western blotting using 1:2,000 goat anti-gp120 polyclonal antibody (Thermo Fisher Scientific) and 1:2,000 horseradish peroxidase (HRP)-conjugated rabbit anti-goat IgG (Thermo Fisher Scientific) or 4E10 anti-gp41 antibody (Polymun) and 1:2,000 HRP-conjugated goat anti-human IgG (Santa Cruz).

Virus infectivity, neutralization, and cold sensitivity.

Single-round virus infection assays were used to measure the ability of the Env variants to support virus entry, as described previously (23). Briefly, 293T cells were cotransfected with the Rev/Env-encoding pSVIIIenv plasmid, a Tat-encoding plasmid, the pCMV HIV-1 Gag-Pol packaging construct, and a plasmid containing the luciferase-expressing HIV-1 vector at a weight ratio of 1:0.125:1:3 using a standard calcium phosphate transfection protocol. At 48 h after transfection, virus-containing supernatants were collected, filtered through a 0.45-μm membrane, and incubated with soluble CD4, BNM-III-170, or antibody for 1 h at 37°C. The mixture was then added to Cf2Th-CD4/CCR5 cells, which were cultured at 37°C and 5% CO₂. To enhance infection by recombinant viruses with the HIV-1_BG505 Env, virus-antibody mixtures were spinoculated with target cells at 1,800 rpm for 1 h at room temperature and then incubated for 1 more hour before additional medium was added. Luciferase activity in the Cf2Th-CD4/CCR5 target cells was measured 48 h later. To measure cold sensitivity, the viruses were incubated on ice for various lengths of time prior to measuring their infectivity. To measure the sensitivity of virus infectivity to cross-linking, the viruses were incubated with BS3 (Thermo Fisher Scientific) for 15 min at room temperature; the reaction was quenched with 15 mM Tris-HCl, pH 8.0, for 10 min, and the mixture was then added to the target cells.

Cross-linking of Envs on VLPs.

A549 cells inducibly expressing virus-like particles (VLPs) consisting of the HIV-1 Gag-mCherry fusion protein and the wild-type HIV-1_AD8 Env have been previously described (7, 138). The D1253 A549-Gag/Env cell line expressing VLPs with wild-type HIV-1_AD8 Env was selected by fluorescence-activated cell sorting (FACS) for Gag-positive and PGT145-positive cells. The D1555.042321.sort A549-Gag/E.2 Env cells and the D1553.042321.sort A549-Gag/AE.2 Env cells inducibly expressing VLPs with the E.2 and AE.2 Envs, respectively, were established similarly. After FACS sorting, these cells were >90% dual positive for Gag expression (KC567 antibody positive) and Env expression (PGT145 antibody positive).

Equivalent numbers of cells from the three cell lines described above were seeded, and the expression of Gag-mCherry/Env VLPs was induced with 2 μg/mL doxycycline. Forty-eight to 72 h later, supernatants containing VLPs were centrifuged at low speed to remove cell debris and then filtered (0.45 μm). Clarified supernatants were centrifuged at 100,000 × g for 1 h at 4°C. VLP pellets were resuspended in 1× PBS, aliquoted, and incubated with different concentrations of either DTSSP (Thermo Fisher Scientific) or glutaraldehyde cross-linkers. The cross-linking reaction with DTSSP was carried out for 30 min at room temperature, after which the reaction was quenched with 100 mM Tris-HCl, pH 8.0, for 10 min at room temperature. Glutaraldehyde cross-linking was carried out for 5 min at room temperature, after which the reaction was quenched with 50 mM glycine for 10 min at room temperature. VLPs were then pelleted at 20,000 × g for 30 min at 4°C. VLP pellets were resuspended in 1× PBS/LDS, boiled, and Western blotted with a goat anti-gp120 antibody, as described above. The intensity of the gp120, gp160, dimer, and trimer bands was quantified using the Bio-Rad Image Lab program.

Cross-linking of sgp140 SOSIP.664 glycoproteins.

293T cells were cotransfected with a 1:8 ratio of a Tat-expressing plasmid and a pSVIIIenv plasmid expressing either the HIV-1_AD8 sgp140 SOSIP.664 glycoprotein or a sgp140 SOSIP.664 glycoprotein with Q114E/Q567K/A582T changes, using polyethyleneimine (PEI) according to the manufacturer’s protocol. Six hours later, the medium was replaced. Forty-eight hours after transfection, cell supernatants containing the sgp140 SOSIP.664 glycoproteins were collected, filtered (0.45 μm), and incubated with BS3 (Thermo Fisher Scientific) at different concentrations for 30 min at room temperature. The cross-linking reaction was quenched with 100 mM Tris-HCl, pH 8.0, for 10 min. LDS and dithiothreitol (DTT) were added to the samples, which were analyzed by reducing SDS-PAGE and Western blotting, as described above.