The lipid membrane of HIV-1 stabilizes the viral envelope glycoproteins and modulates their sensitivity to antibody neutralization

The envelope glycoproteins (Envs) of HIV-1 are embedded in the cholesterol-rich lipid membrane of the virus. Chemical depletion of cholesterol from HIV-1 particles inactivates their infectivity. We observed that diverse HIV-1 strains exhibit a range of sensitivities to such treatment. Differences in sensitivity to cholesterol depletion could not be explained by variation in Env components known to interact with cholesterol, including the cholesterol-recognition motif and cytoplasmic tail of gp41. Using antibody-binding assays, measurements of virus infectivity, and analyses of lipid membrane order, we found that depletion of cholesterol from HIV-1 particles decreases the conformational stability of Env. It enhances exposure of partially cryptic epitopes on the trimer and increases sensitivity to structure-perturbing treatments such as antibodies and cold denaturation. Substitutions in the cholesterol-interacting motif of gp41 induced similar effects as depletion of cholesterol. Surface-acting agents, which are incorporated into the virus lipid membrane, caused similar effects as disruption of the Env-cholesterol interaction. Furthermore, substitutions in gp120 that increased structural stability of Env (i.e. induced a “closed” conformation of the trimer) increased virus resistance to cholesterol depletion and to the surface-acting agents. Collectively, these results indicate a critical contribution of the viral membrane to the stability of the Env trimer and to neutralization resistance against antibodies. Our findings suggest that the potency of poorly neutralizing antibodies, which are commonly elicited in vaccinated individuals, may be markedly enhanced by altering the lipid composition of the viral membrane.


elicited in vaccinated individuals, may be markedly enhanced by altering the lipid composition of the viral membrane.
The envelope glycoproteins (Envs) 4 on the surface of HIV-1 particles serve as fusion machinery that mediates entry of the viral genome into host cells (1)(2)(3)(4). Envs are arranged as trimers, each composed of three gp120 surface subunits and three gp41 transmembrane subunits (5)(6)(7). Furin cleavage of the gp160 precursor into gp120 and gp41 allows proper folding of the molecule (8 -11) and increases compactness of the trimer (12). Structural integrity of the cleaved Env trimer is then maintained by a network of interactions within and among the gp120 and gp41 subunits. In addition to intramolecular stabilizing forces, Env also interacts with the lipid membrane in which it is embedded through several contact points. The gp41 membrane-proximal ectodomain region (MPER) contains a conserved cholesterol recognition motif (13)(14)(15)(16)(17)(18). Substitutions in this motif impair virus infectivity (19, 20), suggesting that interaction with cholesterol is required for Env function. The cytoplasmic tail (CT) of Env also associates with the viral lipid membrane (21) and matrix protein (22)(23)(24)(25)(26). The CT maintains Envs in a "closed" conformation; substitutions in (or complete deletion of) this domain increase exposure of multiple epitopes in gp120 and gp41 (27)(28)(29)(30).
On the surface of the infected cell, the Env trimer is contained in organized regions designated lipid rafts from which the virus buds (31)(32)(33)(34)(35). These specialized membrane domains are enriched with cholesterol and sphingolipids and have a unique protein composition (32, 36 -39). Membrane cholesterol is required for HIV-1 particle infectivity; chemical depletion of cholesterol from virions by treatment with ␤-cyclodextrins (BCDs) inactivates them (40 -45). These agents were tested as microbicides to prevent HIV-1 transmission. In a mouse model of HIV-1 transmission, intravaginal application of BCDs effectively blocked infection (46). However, rhesus macaques treated with BCDs were incompletely protected from subsequent challenge with simian immunodeficiency virus (47). The discrepancy between potent in vitro inhibition and inconsistent in vivo efficacy emphasizes the need for a better understanding of how cholesterol contributes to infectivity of HIV-1. Some reports suggest that depletion of cholesterol physically disrupts the virion membrane (40,44), whereas others failed to note effects on virus morphology or protein composition (42,48). That pseudoviruses containing the Envs of HIV-1 are more sensitive to BCDs than those containing the fusion proteins of other viruses (45) suggests that cholesteroldepleting agents may affect Env, in a direct or indirect manner. Indeed, it has been shown that BCDs can disrupt integrity of the Env trimer, manifested by enhanced shedding of the gp120 subunit (48).
We examined the effects of methyl-BCD (MBCD) on diverse HIV-1 strains and observed a wide range of sensitivities that could not be explained by differences in known membraneinteracting domains of Env. Using a panel of primary isolates and closely matched variants, we observed that sensitivity of Envs to cholesterol depletion corresponds with their inherent conformational stability. Low concentrations of MBCD, which cause minimal or no inactivation of the virus, enhanced exposure of otherwise cryptic Env epitopes, and increased sensitivity to structure-perturbing treatments such as cold-induced denaturation and antibody (Ab) neutralization. Substitutions in the cholesterol-interacting motif of Env exerted similar effects on Env conformation as depletion of cholesterol. The surface-acting agent nonoxynol-9 induced similar effects on Env stability and increased HIV-1 sensitivity to neutralization by otherwise low-potency Abs that target gp120 and gp41.

Sensitivity of HIV-1 to cholesterol depletion is strain-specific and affected by factors outside the known cholesterolinteracting motifs
Depletion of cholesterol from many enveloped viruses inactivates their infectivity (33, 44, 49 -51). Pseudoviruses that contain the G protein of vesicular stomatitis virus (VSV-G) are more resistant to cholesterol depletion than those containing HIV-1 Envs (45), suggesting a specific effect on the viral fusion protein. To identify the factors that contribute to HIV-1 inactivation, we examined sensitivity of diverse Envs from primary isolates to the cholesterol-depleting agent MBCD (44,52). Recombinant replication-incompetent HIV-1 pseudoviruses that contain the Envs of nine clade B and C strains were tested, four from chronically infected individuals and five from transmitted/founder strains. Virus particles were attached to protein-binding plates, incubated with MBCD for 1 h, and washed, and then Cf2Th cells that express CD4 and CCR5 were added to measure residual infectivity. A range of sensitivities to inactivation by the cholesterol-depleting agent was observed (Fig. 1A). Infectivity of virus that contains the 89.6 Env was reduced more than 20-fold, whereas infectivity of virus that contains the Env of isolate 1053-07.B15.1648 (designated herein 1053) was modestly increased. We tested the effects of different MBCD concentrations on four of the isolates and on virus that contains the VSV-G protein (45). The Envs showed a range of sensitivities and were significantly more inactivated by this agent than VSV-G-containing virus (Fig. 1B). We examined whether these effects on infectivity are caused by MBCD-induced detachment of virions from the plates or shedding of the gp120 subunit. An ELISA was performed on plate-bound virions that contain the Envs of the MBCD-sensitive 89.6 strain and the MBCD-resistant 1053 strain, using the gp120-targeting mAb 2G12. Limited changes in binding of the Ab to the Envs were observed after MBCD treatment of the virions (Fig. 1C).
Because virus-surface density of VSV-G is higher than that of HIV-1 Envs (53), we hypothesized that differences in MBCD sensitivity between HIV-1 strains may reflect different levels of Env incorporation into virions. To examine this possibility, we measured the amount of Env contained on viruses using the CD4-Ig probe (54). This probe recognizes the highly conserved CD4-binding site (CD4-BS) and reliably quantifies relative expression levels of Env (55,56). CD4-Ig binding to each preparation was compared with virus particle content in the sample, as quantified by p24 antigen. The calculated virus-surface density of the MBCD-sensitive 89.6 Env was higher than that of the MBCD-resistant Envs (Fig. 1D, bottom). Furthermore, differences in cleavage efficiency of the Envs did not correspond with their sensitivity to MBCD (Fig. 1D, top). Therefore, MBCD sensitivity of diverse Envs does not appear to correspond with their level of Env incorporation or processing efficiency. Two regions of Env were shown to interact with the virus lipid membrane: (i) the cholesterol-recognition motif in the gp41 MPER and (ii) the CT. We examined whether isolate sensitivity to cholesterol depletion could be explained by variation in these membrane-interacting domains. The cholesterol recognition motif is composed of amino acids Leu-Trp-Tyr-Ile-Lys at positions 679 -683 of Env (13)(14)(15)(16)(17). Tyr-681 is the primary interacting residue with cholesterol, with lesser contributions of Leu-679 and Lys-683 (14,57). Alignment of the nine isolates showed complete sequence identity at positions 679 -683 with the exception of Arg at position 683 of the MBCD-sensitive 89.6 isolate (Fig. 1E). In addition, 89.6 contains Leu at position 684 of the membrane-spanning domain, whereas other isolates contain Ile. To examine the potential effects of these changes, we introduced R683K and L684I changes (and their combination) into the 89.6 Env. The changes did not significantly alter sensitivity to MBCD (Fig. 1F). Furthermore, introduction of the K683R change into the MBCDresistant 1053 Env did not affect this phenotype.
Env also interacts with the lipid membrane via the CT domain (21,58,59). We tested the effects of CT deletion on the strain-specific pattern of sensitivities to MBCD. As shown previously (60), deletion of the CT increased virus resistance to cholesterol depletion (Fig. 1G). However, this change did not render all Envs similarly resistant to MBCD. The pattern of relative sensitivities among the CT-deleted mutants was similar to that observed among the full-length Envs (Fig. 1B). Interestingly, virions that contain the CT-deleted forms showed a gradual decrease in 2G12 binding after treatment with high concentrations of MBCD (Fig. 1H). This finding suggested that shedding of gp120 may occur under these experimental conditions, as shown previously (48). Disruption of the CT-matrix interaction indeed increases the propensity for gp120 shedding Lipid membrane interactions stabilize HIV-1 Env (29,61,62). Nevertheless, under the unique experimental conditions that we applied, and at MBCD concentrations that induce nearly complete inactivation of the diverse strains, minimal decline in mAb 2G12 binding was observed.
In summary, diverse HIV-1 Envs exhibit a range of sensitivities to inactivation by low concentrations of the cholesteroldepleting agent. Such differences cannot be explained by variation in Env-processing efficiency or incorporation into virus particles; nor can they be explained by variation in Env components known to interact with the membrane. These findings suggested that additional Env motifs or properties, which differ among strains, affect HIV-1 sensitivity to cholesterol depletion.

Sensitivity of Envs to cholesterol depletion corresponds with their sensitivity to surface-acting agents
Apart from depletion of cholesterol, other treatments that alter composition of the lipid membrane can inhibit HIV-1 infectivity. We examined sensitivity of the above strains to two surface-acting agents; nonoxynol-9 (N-9) and glycerol mono- HIV-1 sensitivity to cholesterol depletion is strain-specific and cannot be explained by differences in Env cleavage efficiency, virion incorporation, gp120 shedding, or known membrane-interacting motifs. A and B, effect of MBCD on function of diverse HIV-1 Envs. Luciferase-expressing recombinant viruses that contain the indicated Envs were attached to protein-binding plates and incubated with 1 mM MBCD (A) or a range of concentrations (B) for 1 h. Samples were then washed, and Cf2Th CD4 ϩ CCR5 ϩ target cells were added. Residual infectivity was measured 3 days later. Data are expressed as a percentage of infection measured in samples not treated by MBCD. The color scheme (blue, 1053; green, AD8; purple, 7030; red, 89.6) is maintained throughout the paper. C, antibody binding to HIV-1 particles. Viruses containing Env 89.6 or 1053 were attached to protein-binding plates and treated by MBCD for 1 h. Samples were then washed, and binding of mAb 2G12 was measured by ELISA. RLU, relative light units. D, virion incorporation and cleavage status of Envs. Binding of CD4-Ig (5 g/ml) to viruses is expressed as a fraction of virus particle content in the sample, as measured by p24 antigen. The top panel shows cleavage status of virus-associated Envs. Blots were probed with sheep antiserum raised against HIV-1 gp120. Values represent intensity of the gp120 band relative to the gp160 band. sgp120, soluble gp120; 89.6(U), uncleaved Env mutant. E, sequence alignment of the cholesterol-recognition motif in the gp41 ectodomain. MPER, membrane-proximal ectodomain region; Con_B, clade B consensus. F, effect of mutations in the cholesterol recognition motif of Env on sensitivity to MBCD. G, effect of CT deletion on sensitivity to MBCD. Infectivity after exposure of full-length and CT-deleted mutants to 2 mM MBCD is shown to the right. H, binding of mAb 2G12 to virus particles containing CT-deleted Envs. Measurements were performed in the same experiment as the viruses shown in C. Data are representative of three independent experiments, performed with three replicate samples (in binding measurements) or two replicate samples (in infectivity assays) for each condition. Error bars, S.E.

Lipid membrane interactions stabilize HIV-1 Env
laurate (GML). N-9 is an approved and commercially available spermicide. In addition, it exhibits potent antiviral effects and was tested as a microbicide to inhibit transmission of HIV-1. However, the treatment protocol was associated with formation of genital ulcers, and the trial was halted (63)(64)(65). GML is a nonionic surfactant; it exhibits broad antimicrobial efficacy and was shown to prevent transmission of simian immunodeficiency virus in a rhesus macaque model of HIV-1 infection (66). The mechanism of HIV-1 inactivation by surface-acting agents is incompletely understood (67)(68)(69). We measured sensitivity of the primary Envs to GML and N-9 and observed a pattern that was remarkably similar to that of MBCD: Envs 1053 and AD8 were resistant relative to 89.6 and 7030 ( Fig. 2A). Deletion of the CT increased resistance to both agents. The CT-deleted Envs still showed significant variation in sensitivity to N-9 (Fig. 2B), a pattern similar to that observed for MBCD (Fig. 1F). By contrast, CT deletion abrogated differences in GML sensitivity between isolates.
Therefore, sensitivity to changes in lipid membrane composition (i.e. removal of cholesterol or addition of a surfactant) appears to be an inherent property of each Env. Diverse isolates exhibit different sensitivities to such treatments. The CT can account for variation between isolates in their sensitivity to the surfactant GML. However, sensitivity to the cholesterol-depleting agent MBCD and the surfactant N-9 is also affected by factors outside the CT domain of gp41.

Sensitivity of HIV-1 strains to cholesterol depletion is associated with the inherent stability of their Envs
The 89.6 Env, which is the most sensitive to membrane lipid changes of the isolates we tested, was also shown to be structurally unstable relative to Envs of other primary isolates (70). Indeed, Envs of diverse HIV-1 strains differ considerably in their inherent level of stability (71)(72)(73). Stable Envs maintain a "closed" conformation wherein cryptic epitopes that overlap the coreceptor-binding site (CoR-BS) are not exposed. Additionally, stable Envs are more resistant to the effects of treatments that perturb protein structure, including cold denaturation and inhibitory ligands (70,74). We hypothesized that sensitivity of HIV-1 to membrane lipid changes is affected by stability of Env. To test this hypothesis, we examined a panel of closely matched isolates that vary in their stability.
Two substitutions in the 89.6 Env were shown to increase trimer stability: (i) Met to Ile at position 225 in the second conserved (C2) region of gp120 and (ii) Arg to Glu at position 305 of the third variable (V3) loop (75,76). Both changes increase resistance to cold denaturation (70) and global resistance to Ab neutralization (70,77). We examined the effects of these changes on Env conformation. Envs were expressed in human osteosarcoma (HOS) cells, which are unique for an expression profile of nearly only fully cleaved trimers on their surface (12). Combination of the two changes changed the conformation of Env to a closed form, which does not expose the CoR-BS epitope targeted by mAb 17b (Fig. 3A). Exposure of other epitopes was unaffected by the changes, including those

Lipid membrane interactions stabilize HIV-1 Env
targeted by mAbs b12 (overlaps the CD4-BS), 10E8 (MPER), and 447-52D (V3 loop) (Fig. 3A). The lower binding of mAb 17b to the 89.6(M225I,R305E) mutant was not caused by disruption of the epitope, because the addition of soluble CD4 resulted in similar binding of 17b to the two Envs (Fig.  3B). Instead, differences likely reflect lower exposure of the 17b epitope, as suggested by its reduced sampling frequency in the double-mutant Env (Fig. 3C) (70,72). The 89.6(M225I,R305E) mutant also sampled less frequently the partially cryptic CD4-binding site (78), whereas sampling of the highly-exposed 2G12 epitope was similar to that measured for WT 89.6. Therefore, the 89.6(M225I,R305E) mutant occupies a more closed conformation than the WT Env.
Primary Envs exhibit a range of sensitivities to cold denaturation. Exposure to 0°C causes gradual structural changes in Env that ultimately lead to loss of function (72,73,79). Sensitivity of Envs to cold denaturation corresponds with their conformational stability. The unstable 89.6(WT) Env was highly sensitive to cold, whereas the 89.6(M225I,R305E) mutant was more resistant (Fig. 3D); their half-lives at 0°C were 2 and 84 h, respectively. We examined sensitivity of viruses that contain the gp120 changes to depletion of cholesterol. Consistent with the above hypothesis, sensitivities of the 89.6 variants to MBCD corresponded with their sensitivities to cold (Fig. 3E). To further establish the relationship between Env stability and resistance to cholesterol depletion, we tested the panel of nine primary isolates. A range of sensitivities to cold inactivation was Envs. HOS cells expressing the indicated 89.6 variants were incubated with mAbs at 0.5 g/ml for 45 min, and binding was measured by ELISA. To normalize for cell-surface expression, values are expressed as a percentage of 2G12 binding to each Env. B, binding of mAb 17b to Envs expressed on the surface of HOS cells in the absence or presence of soluble CD4 (sCD4). C, sampling frequencies of epitopes on 89.6 variants. Env-expressing cells were fixed with 5 mM glutaraldehyde and washed, and mAb binding was measured by ELISA. Binding to unfixed Envs was also measured. Epitope sampling frequency was calculated as the ratio between mAb binding to fixed Envs and its binding to non-fixed Envs. D, sensitivity of 89.6 Env variants to cold inactivation. Viruses were incubated at 0°C for 24 h and re-equilibrated to 37°C, and residual infection was measured. E, sensitivity of 89.6 variants to inactivation by 1 mM MBCD. Values describe the percentage of infection measured in samples not incubated with MBCD. F, relationship between sensitivity of diverse primary Envs to inactivation by cold and MBCD. A correlation is shown between residual infectivity after 4 h at 0°C and after incubation with 1 mM MBCD. r S , Spearman correlation coefficient. p value, two-tailed test. G and H, sensitivity of cholesterol-depleted viruses to cold inactivation. Viruses were incubated with buffer or the indicated concentrations of MBCD for 1 h at 37°C, washed, and further incubated at 23 or 0°C for 2 h. Samples were then equilibrated to 37°C, and target cells were added to measure residual infectivity. p values were calculated by a two-tailed t test: *, p Ͻ 0.05; **, p Ͻ 0.005. Error bars, S.E.
We examined the synergy between effects of MBCD and cold. Synergy was calculated as the level of inactivation induced by the combined treatments relative to the product of the individual treatments. Preincubation of the 89.6 variants with 0.5 mM MBCD modestly enhanced sensitivity to cold, 4.5-fold for the 89.6(WT) Env and 2.6-fold for the stabilized 89.6(M225I,R305E) Env (Fig. 3G). The stable 1053 Env was unaffected by cold or this low MBCD concentration. Higher concentrations of MBCD further enhanced sensitivity of the stabilized 89.6 mutant to cold but minimally affected the stable 1053 isolate (Fig. 3H). These findings suggest that the low concentrations of MBCD we used (Ͻ2 mM) destabilize Env; the loss of function is associated with the inherent stability of each Env.

Substitutions in the cholesterol-recognition motif in the gp41 ectodomain enhance HIV-1 sensitivity to Env-perturbing treatments
We examined whether disruption of the Env-cholesterol interaction by substitutions in the cholesterol-recognition motif of gp41 has similar effects on Env stability as depletion of cholesterol. Alanine mutations were introduced at positions 679 and 681, which were shown to exert the greatest effects on the interaction with cholesterol (14,57). The changes reduced infectivity of viruses that contain the 89.6 and 1053 Envs (Fig.  4A). Cold sensitivity of the virus was significantly enhanced by the Y681A change and more modestly by L679A (Fig. 4B). We compared the effects of MBCD and the gp41 changes on Env conformation. Binding of multiple mAbs that target distinct domains of gp120 and gp41 was measured to fully cleaved trimers expressed on HOS cells. For both strains, MBCD treatment Infectivity of the viruses was measured. Infectivity after incubation for the indicated times at 0°C was also measured. C, Envs of the indicated strains were transiently expressed in HOS cells. Cells were treated by MBCD for 1 h and washed, and binding of the indicated mAbs was measured by ELISA. Values are expressed as percentage of each mAb binding to cells not treated by MBCD. *, no binding of antibody to the Env. D and E, WT and mutant Envs were expressed on HOS cells, and their binding by mAbs was measured. Data are expressed as the ratio of mAb binding to the mutant relative to WT Env. The effect of MBCD and the Y681A mutation on binding of each mAb is compared: p value, two-tailed test. F, effect of the Y681A mutation on HIV-1 neutralization. Viruses containing WT or mutant Envs were incubated with two different pools of serum (PS) from HIV-infected individuals and added to Cf2Th CD4 ϩ CCR5 ϩ cells. Infectivity was measured 3 days later. Error bars, S.E.

Lipid membrane interactions stabilize HIV-1 Env
increased exposure of epitopes that overlap the CoR-BS, CD4-BS, and MPER (Fig. 4C). Binding of the V3 loop-targeting mAb 447-52D was also enhanced. By contrast, binding of the glycantargeting and trimer-dependent mAbs was less affected or not affected by cholesterol depletion. A strong correlation was observed between the effects of MBCD on conformation of the 89.6 and 1053 Envs (see inset in Fig. 4C). We then examined the effects of substitutions in the cholesterol-recognition motif on Env conformation. For both 89.6 and 1053, similar effects of the Y681A change were observed, which include increased exposure of epitopes that overlap the CoR-BS, CD4-BS, V3 loop, and MPER (Fig. 4, D and E). The effects of the Y681A change and MBCD treatment on binding of the mAbs were highly correlated. We also examined the effect of the Y681A change on virus sensitivity to pooled serum from HIV-infected individuals. Consistent with the enhanced binding of multiple mAbs, the substitution increased sensitivity of both 89.6 and 1053 Envs to two different pools of sera (Fig. 4F). The effects of the substitution on the unstable 89.6 Env were significantly greater (up to 100-fold enhancement of neutralization) than its effects on the stable 1053 Env. Therefore, disruption of the cholesterol-interacting motif of gp41 enhances sensitivity to cold and patient sera and increases exposure of otherwise-cryptic epitopes on Env.

Cholesterol depletion increases HIV-1 sensitivity to antibody neutralization
Because cholesterol depletion alters exposure of Env epitopes, we examined the effects of this treatment on HIV-1 neutralization. Based on the effects of MBCD on WT Env conformation (Fig. 4C), we tested three Abs that target partially cryptic epitopes on Env: mAb 17b (CoR-BS), mAb 447-52D (V3 loop), and mAb 10E8 (MPER). Relative to the modest effects of MBCD on cold sensitivity, this agent had a significant effect on Env sensitivity to Abs (Fig. 5, A-C). The concentration of MBCD required to enhance Ab sensitivity depended on the inherent stability of each Env. For example, in the absence of MBCD, mAb 17b did not inhibit the unstable 89.6 strain, whereas at 0.5 and 1 mM MBCD, the potency of this mAb was enhanced 4-and 30-fold, respectively (Fig. 5A). MBCD treatment also enhanced neutralization of 89.6 by mAbs 447-52D and 10E8. Higher concentrations of MBCD were required to sensitize the stabilized 89.6(M225I,R305E) Env to Abs (Fig.  5B). At 0.5 mM, this variant was only neutralized by a high concentration (1 g/ml) of 10E8, whereas treatment with 1 mM modestly sensitized it to 17b and to a lower concentration (0.25 g/ml) of 10E8. Sensitization of the hyperstable 1053 strain to mAb 10E8 required 1 or 2 mM MBCD, which enhanced neutralization by 3-or 26-fold, respectively (Fig.  5C).
Therefore, depletion of cholesterol sensitizes HIV-1 to Ab neutralization. These effects cannot be explained solely by changes in Ab binding. For fully cleaved Envs, the 89.6 and 1053 strains show similar conformational changes after depletion of cholesterol (Fig. 4C). Nevertheless, the inherently unstable 89.6 Env is more readily neutralized, whereas more stable Envs require higher concentrations of MBCD to induce these effects.

Lipid membrane interactions stabilize HIV-1 Env The surface-acting agent nonoxynol-9 sensitizes HIV-1 to antibody neutralization
Sensitivity of diverse HIV-1 isolates to MBCD corresponded with their sensitivity to the surface-acting agent N-9 (Figs. 1B and 2A). We asked whether, similar to MBCD, N-9 treatment can enhance HIV-1 sensitivity to Ab neutralization. We first compared the effects of N-9 on infectivity of the 89.6 variants. The mutations increased Env resistance to N-9; 89.6(WT) was inactivated 160-fold, whereas the mutant only showed a 2.8fold decrease in infection (Fig. 6A). Such differences are significantly greater than those observed for MBCD (Fig. 3E). N-9 induced similar effects on Env neutralization as MBCD. Sensitivity of 89.6(WT) to mAb 10E8 was increased by 360fold, whereas sensitivity to mAbs 17b and 447-52D was increased by ϳ7-fold (Fig. 6B). N-9 enhanced Ab sensitivity of 89.6(M225I,R305E) to a lesser extent than the WT Env; the potency of mAb 10E8 was increased up to 21-fold, with limited effects on neutralization by mAb 17b or 447-52D (Fig. 6C). It could be proposed that the similar effects of cholesterol depletion and the surface-acting agent on Env are mediated by changes in membrane lipid order. Several studies have shown that BCDs reduce order (i.e. increase fluidity) of the HIV-1 lipid membrane (48,80). To our knowledge, effects of the surfaceacting agent N-9 have not been tested. We examined the effects of both agents using the fluorescent dye di-4-ANEPPDHQ. This dye efficiently distributes in the cell membrane from water-based solutions and exhibits different emission maxima based on ordering of the lipid environment; in the disordered lipid (liquid) phase, it is ϳ610 nm, whereas in the ordered (gel) phase, it is ϳ560 nm (81). The dye is highly photostable and sensitive to changes in lipid composition in the two phases (69,(81)(82)(83)(84)(85)(86). We analyzed the changes in membrane order induced by MBCD and N-9 in virus-producing cells by calculating the generalized polarization (GP) value. At 0.25 mM MBCD, the GP value was increased, and then it gradually decreased at higher concentrations, reflecting a gradual decrease in membrane order (Fig. 6D, top). N-9 caused a remarkably different effect; fluorescence intensity at both emission wavelengths was gradually decreased and associated with an increase in GP, reflecting an increase in membrane order (Fig. 6D, bottom). Such effects were observed at both 50 and 80 mM, which enhance neutralization sensitivity of the virus and reduce infectivity of the unstable 89.6 Env but only minimally affect infectivity of the stabilized mutant.

Lipid membrane interactions stabilize HIV-1 Env
Taken together, our findings suggest that disruption of the interaction between Env and the lipid membrane (by mutations or chemical agents) reduces the conformational stability of Env. Such effects also occur at noninactivating concentrations of the agents and increase virus sensitivity to Abs. The degree of sensitization depends on the inherent stability of each Env. Stable Envs are more resistant to these effects and require greater destabilizing pressure (e.g. higher concentrations of the agent) to increase their sensitivity to Abs. Destabilizing effects are likely not induced by changes in membrane order, because N-9 and MBCD exert similar effects on Env infectivity and sensitivity to Abs, but different effects on membrane lipid order. Instead, the effects of substitutions in the cholesterol-interacting motif of gp41 suggest that disruption of the anchoring interaction with the membrane is responsible for these changes.

Discussion
The HIV-1 Env trimer is a high-potential energy molecular complex. The energy, which is required to drive the fusion process, is retained by multiple interactions that keep the trimer in a functional, compact, and stable conformation. In this form, it masks conserved immunogenic epitopes and is resistant to the effects of Abs and other structure-perturbing agents. Multiple sites on the Env trimer contribute to its conformational stability and prevent it from "springing" into a lower-energy and nonfunctional form (27,72,(87)(88)(89). Partial release of these constraints increases sensitivity of Env to forces that can perturb structure, including denaturing physical forces and ligands such as Abs (70). Here, we show that depletion of cholesterol from the viral membrane likely releases such constraints. The effects of perturbing treatments are cumulative; combined with the inherent stability of each Env, they determine Env functionality and Ab sensitivity. This understanding presents a new strategy to enhance the potency of vaccine-elicited Abs, based on destabilizing the interaction between Env and the viral membrane. The approach allows us to take advantage of the high titers of otherwise nonneutralizing Abs generated in vaccinated and infected individuals, as discussed below.
Previous studies have shown that inactivating concentrations of cholesterol-depleting agents reduce the stability of the virus and cause shedding of gp120 (48). Here we analyze the stability of the Env trimer, both structural and functional. We apply concentrations of the cholesterol-depleting agent and experimental conditions that are not (or minimally) inactivating for most primary strains. Cholesterol depletion is shown to reduce trimer stability, as measured by (i) increased sensitivity to Abs that target different sites on gp120 and gp41 and (ii) increased sensitivity to cold denaturation. Relatively limited shedding of gp120 was observed after MBCD treatment. We believe that the unique experimental conditions we applied for cholesterol depletion and binding measurements likely account for the limited shedding of gp120. First, we attach viruses to protein-binding plates. This allows rapid washing of the viruses after MBCD treatment and immediate measurements of Ab binding and infectivity without a centrifugation step. Second, cholesterol depletion and binding measurements are performed in the presence of high concentrations of protein, which reduce inactivation and increase binding of trimer-specific Abs, suggesting that Env occupies a more "closed" conformation (data not shown). Taken together, our findings suggest that inactivation of Env by cholesterol-depleting agents is preceded by a destabilization phase, which enhances sensitivity to other perturbing treatments such as Abs. The effects of such perturbations on function and neutralization sensitivity are determined by the inherent stability of each Env, which differs among primary strains of HIV-1. At sufficiently high destabilizing pressure, Env is inactivated, likely by induction of gp120 shedding from virions, which is the common path for many inhibitors of HIV-1 (48,62,90).
We also analyzed conformational stability of the trimer (i.e. the propensity to undergo changes after exposure to structureperturbing agents). MBCD treatment exerted similar effects on Env as the substitutions in the cholesterol-interacting motif. An open conformation of the trimer was induced, characterized by increased exposure of epitopes that overlap the CoR-BS, CD4-BS, and MPER. Indeed, different treatments that perturb Env (but do not inactivate function) alter conformation in a similar manner. CD4-mimetic compounds induce a metastable form that exposes the coreceptor-binding site and induces transient formation of the HR coiled-coil (71). At low concentrations of the mimetics, effects on the CoR-BS are reversible (70). Exposure to cold causes gradual denaturation of Env structure and reversibly exposes epitopes that overlap the CoR-BS and (to a lesser extent) the CD4-BS, but reduces exposure of MPER epitopes (70). Furthermore, deletion of the cytoplasmic tail of Env increases exposure of CoR-BS epitopes (28). Here we show that disruption of the interaction between Env and the membrane also causes exposure of epitopes that overlap the CoR-BS and CD4-BS and, in addition, exposes epitopes in the MPER. Sensitivity to neutralization by these Abs is increased. Therefore, induction of an open conformation appears to be a common path that precedes inactivation of function.
The effects of cholesterol depletion and surfactant supplementation are strikingly similar; they induce the same inhibition profile of diverse isolates and increase Ab sensitivity of Env. What then is their mechanism of action on the viral Env? On the surface of the infected cell and virion, Env is contained in lipid rafts (32,33,37). These membrane domains contain high levels of cholesterol and a unique profile of proteins (32, 36 -39). Interactions between Env and the membrane likely occur in this domain (58). The CT-membrane interaction is a known site of contact that stabilizes Env; it increases trimer compactness and resistance to Abs (28,91,92). A second site of Env-membrane contact is the cholesterol-recognition motif in the gp41 ectodomain (93). Changes in lipid content likely destabilize Env conformation by disrupting these membrane-anchoring interactions. Such instability could also be proposed to result from indirect effects of the agents on order of the lipid membrane. Indeed, cholesterol is a major regulator of lipid bilayer fluidity (94,95). Incorporation of surfactants also alters physical properties of biological membranes, which impacts function of transmembrane proteins (96). Our data suggest that effects of the above agents are likely not caused by changes in membrane order. MBCD and N-9 induce similar effects on Env stability but are associated with distinct changes in the GP value.

Lipid membrane interactions stabilize HIV-1 Env
Diverse HIV-1 strains exhibit different levels of inherent conformational stability. The primary strain 89.6 is inherently unstable; it samples open conformations of Env more readily and is sensitive to structure-perturbing treatments such as cold denaturation, soluble forms of the CD4 receptor, and structureperturbing Abs (70). As shown here, such Envs are also more sensitive to disruption of the interaction between Env and the membrane. Substitutions in gp120 that increase conformational stability of Env also increase resistance to disruption of the Env-membrane interaction. Conversely, disruption of the Env-membrane interaction destabilizes Envs and renders them more sensitive to Abs. We have previously shown that HIV-1 sensitivity to neutralization is determined by (i) binding efficiency of the Ab, (ii) inherent conformational stability of the Env, and (iii) structure-perturbing properties of the Ab (72,73).
Here we analyzed the changes in Env recognition by Abs using the fully cleaved form expressed on the surface of cells. Such forms, which reflect the conformation of fusion-competent Envs, undergo similar changes after depletion of cholesterol and due to disruption of the cholesterol-interacting motif of gp41 (Fig. 4, D and E). MBCD-induced changes in Ab binding are similar for the unstable 89.6 Env and the hyperstable 1053 Env (Fig. 4C), despite significant differences in sensitization to Abs (Fig. 5). These findings suggest that inherent stability of the protein (rather than solely effects on exposure of the epitopes) determines the level of Env sensitization to Abs by disruption of the Env-membrane interaction. Inherently unstable Envs are destabilized more readily. Application of sufficient destabilizing pressure, by higher concentrations of MBCD or surfactants or by combination treatments with mild denaturants (70) can overcome the inherent stability of each Env.
We describe here a novel strategy to sensitize HIV-1 to otherwise less-neutralizing or nonneutralizing Abs. Such effects also occur at low noninactivating concentrations of the agents. Low-potency Abs that target cryptic epitopes are normally elicited in vaccinated and infected individuals (97)(98)(99). The approach enhances the potency of such Abs. Several surfaceacting agents and other lipids have been developed over the past years to replace earlier generations and reduce their toxicity (80,100,101). The low concentrations of such agents that are required to sensitize the virus improves their safety profiles for application at the mucosal sites of virus transmission. The similar lipid composition of the viral lipid raft and spermatozoal acrosome (102,103) suggests that such agents could serve as dual spermicides-microbicides. Indeed, the three agents tested here exert specific effects on spermatozoa; GML and N-9 inhibit their function and motility (66), whereas MBCD induces premature capacitation (103,104). The unique properties and composition of lipid rafts render them an attractive site for a broad range of agents that can perturb them and thus inactivate virus fusion potential.

Cells, antibodies, and reagents
Human embryonic kidney 293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum and 100 g/ml penicillin and strepto-mycin (DMEM/10% FCS). Cf2Th cells expressing CD4 and CCR5 (Cf2Th CD4 ϩ CCR5 ϩ ) were cultured in DMEM/10% FCS containing 400 g/ml G418 and 200 g/ml hygromycin. The mAbs indicated below were obtained through the NIH AIDS Reagent Program. The mAb 447-52D was contributed by Susan Zolla-Pazner (105). James Robinson provided mAb 17b (106). The mAbs 10E8 and 4E10 were contributed by Mark Connors (107) and Polymun Scientific (108), respectively. Hermann Katinger kindly provided mAb 2G12 (109). Recombinant soluble CD4 was contributed by Progenics Pharmaceuticals, Inc. The CD4-Ig fusion protein, which is composed of the Fc region of human IgG1 linked to two copies of the two N-terminal domains of the CD4 molecule, was produced and purified as described previously (54).
For Western blots of virus-associated Envs we used polyclonal serum from sheep immunized with HIV-1 gp120 (110) (kindly provided by Michael Phelan) and a secondary HRPconjugated goat anti-sheep IgG preparation. To determine virus particle content, we measured HIV-1 p24 antigen in samples by Western blot analysis using primary mouse anti-p24 Ag contributed by Michael Malim (111) and a secondary HRPconjugated goat anti-mouse IgG preparation. MBCD (Sigma-Aldrich) and N-9 (LKT Labs) were dissolved in PBS to stock solutions of 30 mM and 160 mM, respectively. GML (Colonial Company) was dissolved in 75% methanol to a stock solution of 18 mM (5 mg/ml) and then diluted in RPMI/10% FCS before use. The fluorescent dye di-4-ANEPPDHQ (Invitrogen) was dissolved in ethanol to a stock solution of 7.5 mM.

Preparation of recombinant luciferase-expressing viruses and infectivity measurements
Single-round, recombinant HIV-1 viruses that express the luciferase gene were generated by transfection of 293T cells using JetPrime transfection reagent (Polyplus), as described previously (56). Virus particles were concentrated by ultracentrifugation at 100,000 ϫ g for 2 h at 12°C through a cushion of 30% (w/w) sucrose in PBS. Virus pellets were then reconstituted in PBS, snap-frozen on dry ice immersed in ethanol for 15 min, and stored at Ϫ80°C until use. Infectivity tests were performed using virus particles attached to protein-binding plates. The

Lipid membrane interactions stabilize HIV-1 Env
approach maintains infectivity of the viruses and allows accurate control over their exposure time to the different treatments (70). Viruses were attached to 96-well plates by centrifugation at 2,000 ϫ g for 2 h at 10°C, and the samples were then blocked overnight using DMEM/10% FCS. Plate-bound viruses were incubated with MBCD, N-9, or GML in RPMI/10% FCS for 1 h at 37°C in the absence or presence of mAbs. Samples were subsequently washed twice with RPMI/10% FCS. In some experiments, plate-bound viruses were further incubated at 37 or 0°C to measure the effect of cholesterol depletion on cold inactivation. Cf2Th CD4 ϩ CCR5 ϩ cells (3 ϫ 10 4 cells/well) were then added to quantify luciferase activity as a measure of infection, as described previously (70).

Measurements of antibody binding to virus particles and cells
Binding of mAbs to Envs on virus particles was measured using a modified protocol of the cell-based ELISA system described previously (71,73). Purified viruses were attached to protein-binding 96-well plates by spinoculation at 2,000 ϫ g for 2 h at 10°C. To determine background binding to the viruses, in all experiments, we used a negative control virus that was generated using an Env construct that contains a stop mutation at position 46 (according to standard HXBc2 numbering (113)). Binding in the negative control wells was subtracted from the binding value in the Env-containing samples. Ab binding to the plate-bound viruses or HOS cells transfected with Envs was measured by ELISA in 96-well plates (70). To examine the effects of cholesterol depletion on Env conformation, platebound viruses or cells were treated with MBCD (0 -10 mM for viruses, 0 and 10 mM for cells) in RPMI/10% FCS for 45 min at 37°C. Samples were then washed twice with RPMI/10% FCS and incubated with Abs in blocking buffer (20 mg/ml BSA, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 25 mM Tris, pH 7.5, and 140 mM NaCl) for 45 min at 37°C (72). In some experiments, soluble CD4 was added with Abs to measure changes in formation/exposure of epitopes. All mAbs were added at 0.5 g/ml (except CD4-Ig, which was used at 5 g/ml). Binding of Abs was then quantified by luminescence using an HRP-conjugated secondary Ab, as described previously (72). In some experiments, binding of mAbs is expressed as a fraction of the binding of mAb 2G12 measured in the same experiment. The epitope of mAb 2G12 is highly exposed on the trimer (109,114) and thus minimally affected by Env conformational changes. To examine exposure of Env epitopes, we measured their sampling frequencies. Envs expressed on the surface of HOS cells were fixed with glutaraldehyde or left untreated, and binding of mAbs to the fixed and unfixed cells was measured, as described previously (70,72,115). Sampling frequency of the epitope on each Env is calculated by binding efficiency of the Ab to fixed Env relative to unfixed Env.

Electrophoresis and Western blotting of HIV-1 proteins to determine Env incorporation and cleavage status
Cleavage status of Env on virus particles was determined using virus stocks generated in 293T cells and purified by ultracentrifugation at 100,000 ϫ g for 2 h at 10°C, over a 30% (w/w) cushion of sucrose in PBS. Virus pellets were lysed with Nonidet P-40 buffer (0.5% Nonidet P-40, 0.5 M NaCl, and 10 mM Tris, pH 7.5) containing protease inhibitor mixture (Roche Diagnostics) for 30 min at 4°C with constant agitation. As controls, we used virus containing an Env with mutations in the furin cleavage site (12) and soluble gp120, which was shed and collected from the supernatant of cells transfected with gp160expressing constructs. Cell supernatant was centrifuged to remove debris, and gp120 was immunoprecipitated using Protein A beads and a combination of mAbs 2G12, VRC03, and PGT121 (all at 1 g/ml). Samples were then analyzed by SDS-PAGE, transferred to polyvinylidene difluoride membranes, and probed with sheep anti-gp120 antiserum followed by HRPconjugated goat anti-mouse IgG. Band intensity was quantified by densitometry.
To measure Env incorporation into virus particles, we calculated the ratio between binding of saturating concentrations of CD4-Ig (5 g/ml) to each sample and the amount of p24 antigen it contains. Quantification of Env content in viruses by CD4-Ig is less affected by the varying antigenicities of diverse strains than other methods, such as polyclonal serum (55). The p24 antigen content in each sample was quantified by densitometry of p24 bands on Western blots using polyclonal mouse anti-HIV-1 p24 antigen and secondary (HRP-conjugated) goat anti-mouse IgG.

Measurements of changes in lipid membrane order
Virus-producing 293T cells (8.5 ϫ 10 5 cells/sample) were detached from cell culture plates by Accutase and incubated in 400 l of RPMI/10% FCS supplemented with MBCD or N-9 for 1 h at 37°C. Cells were then washed twice with PBS (pH 7.4) and resuspended in 500 l of PBS. The phase-sensitive dye di-4-ANEPPDHQ (Invitrogen) was then added to the cells (final concentration 1.5 M) and incubated at 37°C for 45 min. Cells were then washed twice with PBS and analyzed by flow cytometry. Cells were gated to exclude debris, and fluorescence intensities at 575 and 660 nm were measured. As an indicator of lipid membrane order, we calculated the GP value: GP ϭ (I 575 Ϫ I 660 )/(I 575 ϩ I 660 ), where I 575 and I 660 are the median fluorescence intensities emitted at the corresponding wavelengths.

Statistical analyses
Measurements of Ab binding to Envs were performed with three replicate samples for all conditions in each experiment. Measurements of infectivity were performed with two replicate samples for all conditions. Two-tailed Spearman rank-order tests were used to evaluate the correlation between data variables. p values of Ͻ0.05 were considered statistically significant. S.E. values were used to describe experimental variation within or between experiments, as indicated.