Replication-Competent Variants of Human Immunodeficiency Virus Type 2 Lacking the V3 Loop Exhibit Resistance to Chemokine Receptor Antagonists

Abstract

Entry of human immunodeficiency virus type 1 (HIV-1) and HIV-2 requires interactions between the envelope glycoprotein (Env) on the virus and CD4 and a chemokine receptor, either CCR5 or CXCR4, on the cell surface. The V3 loop of the HIV gp120 glycoprotein plays a critical role in this process, determining tropism for CCR5- or CXCR4-expressing cells, but details of how V3 interacts with these receptors have not been defined. Using an iterative process of deletion mutagenesis and in vitro adaptation of infectious viruses, variants of HIV-2 were derived that could replicate without V3, either with or without a deletion of the V1/V2 variable loops. The generation of these functional but markedly minimized Envs required adaptive changes on the gp120 core and gp41 transmembrane glycoprotein. V3-deleted Envs exhibited tropism for both CCR5- and CXCR4-expressing cells, suggesting that domains on the gp120 core were mediating interactions with determinants shared by both coreceptors. Remarkably, HIV-2 Envs with V3 deletions became resistant to small-molecule inhibitors of CCR5 and CXCR4, suggesting that these drugs inhibit wild-type viruses by disrupting a specific V3 interaction with the coreceptor. This study represents a proof of concept that HIV Envs lacking V3 alone or in combination with V1/V2 that retain functional domains required for viral entry can be derived. Such minimized Envs may be useful in understanding Env function, screening for new inhibitors of gp120 core interactions with chemokine receptors, and designing novel immunogens for vaccines.

During viral entry, the human immunodeficiency virus (HIV) envelope glycoprotein (Env) mediates complex and highly coordinated steps that include binding of gp120 to CD4, a subsequent interaction with a chemokine receptor (either CCR5 or CXCR4), and the release of the transmembrane protein (TM) to interact and ultimately fuse with the target cell membrane (11, 41). These events continue to occur in the face of strong host humoral immune responses owing to a number of structural attributes of Env, particularly its ability to tolerate extensive genetic variation (40, 66). The sites for this variation are located predominantly on gp120 variable loops, V1/V2, V3, and V4, which face outward on the trimeric gp120/TM oligomer (3, 18, 28, 30, 71). Variation is greatest in the V1/V2 and V4 loops, while for V3 variation is most prominent among isolates that utilize CXCR4 (16, 18, 22, 28, 63). In addition, the V1/V2 and V3 loops may protect critical domains on the gp120 core that include, respectively, the recessed CD4 binding site and the bridging sheet, a four-stranded antiparallel beta sheet, formed from amino acids in the V1/V2 stem and the C4 domain, that likely binds to the chemokine receptor amino terminus (15, 29, 47, 48, 57, 65). The V3 loop also plays a key role in interacting with chemokine receptors and determines tropism for CCR5- or CXCR4-expressing cells (8, 9, 12, 18, 20, 23, 39, 55, 69). The recently solved V3 structure on a CD4-bound gp120 core shows that its base is contiguous with the surface formed by the bridging sheet while its more distal region projects toward the cell membrane, where it has been proposed to contact the coreceptor's extracellular loops (ECLs) (22). However, despite extensive data from mutagenesis, the precise nature of these interactions is unknown, as are their contributions to Env function (18).

The structure, function, and immunogenicity of the HIV (or simian immunodeficiency virus [SIV]) Env have been explored by deriving replication-competent viruses with functional Envs that lack variable loops (25, 50, 56, 65, 67, 70). Envs with partial or complete deletions of V1/V2 are more neutralization sensitive (25, 50, 56, 70) and in the case of SIV are less dependent on CD4 (43). Not surprisingly, given the importance of V3 for interacting with CCR5 and CXCR4, Envs lacking V3 function poorly in fusion assays and infectious viruses without V3 have not been described (50). Although V3 shows extensive amino acid diversity across HIV and SIV phylogeny, unlike V1/V2 and V4, which can tolerate insertions and deletions (4, 10, 22, 46), the V3 length is highly conserved and is typically 34 or 35 amino acids (8, 16, 18, 22). This conservation is consistent with the view that a critical V3 length is required to contact the chemokine receptor (22).

Small-molecule inhibitors of CCR5 and CXCR4 have been described, which bind to a pocket defined by transmembrane helices (53) or membrane-proximal regions of the ECLs (17, 19, 58), respectively, and inhibit viral entry. Although their mechanism of action is not fully understood, rather than blocking virus binding directly, they are thought to act through an allosteric mechanism, altering the repertoire of conformational states available to chemokine receptors and rendering them nonpermissive for a functional gp120 interaction (53, 62). Resistance to these compounds in vitro appears to result from an altered use of the chemokine receptor, although the underlying mechanisms are unclear (27, 36, 51, 52, 59). As these compounds are now in clinical trials, a more complete understanding of how they interfere with the HIV Env as well as what mechanisms are responsible for viral resistance is needed.

To explore the role of HIV variable loops in coreceptor interactions and to probe the mechanism of action of antiviral chemokine receptor antagonists, we derived replication-competent variants of HIV that lacked these loops. Using an iterative process of deletion mutagenesis and viral adaptation in an HIV type 2 (HIV-2) model, we describe the first derivation of an infectious HIV lacking V3 alone or in combination with a deletion of V1/V2. V3-deleted Envs were CD4 dependent but could utilize both CCR5 and CXCR4, indicating that domains on the gp120 core could recognize both receptors. Remarkably, Envs and viruses with partial or complete deletions of V3 became resistant to small-molecule antagonists of CCR5 and CXCR4, suggesting that these compounds likely act by disrupting an interaction of V3 with the chemokine receptor.

MATERIALS AND METHODS

PCR mutagenesis, Env cloning, and virus construction.

The HIV-2/VCP Env expression plasmid in pCR3.1 for cell-cell fusion assays and the proviral molecular clone based on pACR23 expressing HIV-2/VCP Env (pACR23.VCP.Env) have been described previously (32-34). The QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) was used to construct variable-loop deletion mutants. For ΔV1/V2, residues 110 to 194 were deleted and replaced with a glycine-alanine-glycine linker (GAG) using primers 5′-CAAATTAACACCCTTATGTGTAGGTGCCGGCCATTGCAATACATCAGTC-3′ and 5′-GACTGATGTATTGCAATGGCCGGCACCTACACATAAGGGTGTTAATTTG-3′; for the ΔV3(6,6) deletion in V3, residues 303 to 323 were deleted and replaced with GAG using primers 5′-GTAAGAGGCCGGGAAATAAGGGTGCCGGCAAACCCAGGCAAGCATGG-3′ and 5′-CCATGCTTGCCTGGGTTTGCCGGCACCCTTATTTCCCGGCCTCTTAC-3′; and for the ΔV3(1,1) deletion, residues 298 to 326 were deleted and replaced with GAG using primers 5′-CTCACTATGCATTGTAAGGGTGCCGGCTGGTGTTGGTTCAAAGGC-3′ and 5′-GCCTTTGAACCAACACCAGCCGGCACCCTTACAATGCATAGTGAG-3′. To derive HIV-2/VCP env clones from infected cells, genomic DNA was prepared with the QIAamp DNA minikit according to the manufacturer's instructions (QIAGEN, Inc., Valencia, CA) and Env sequences PCR amplified (5′ primer, 5′-CGGGATATGTTATGAACGAAAGGGC-3′; 3′ primer, 5′-GCACGCGCCCGCAAGAGTCTCTCTTGTAGCCCTCGC-3′) from genomic DNA using HotStar Taq (QIAGEN) and TA cloned into pCRII or pCR2.1 (Invitrogen, Carlsbad, CA). Human CD4, CXCR4, and CCR5 in pcDNA3 and the reporter plasmid encoding luciferase under the control of a T7 promoter (T7.luciferase), used in cell-cell fusion assays, have been described previously (34). To generate proviral clones with loop deletions or with adapted Envs derived from infected cells, the env genes were cloned into pACR23 with BsmBI and BamHI as described previously (32).

Derivation and characterization of mutagenized viruses and Envs.

Viral stocks were generated by electroporating (250 V; 950 μF) SupT1 cells (5 × 10⁶ cells in 4-mm cuvettes) with 20 μg of pACR23 containing the parental VCP Env or its mutagenized derivatives. Viral growth was monitored by immunofluorescence microscopy of p27^Gag-expressing cells, and cell-free passage of virus onto uninfected SupT1 cells was performed when cultures were >80% for p27 expression. Viral replication was quantified by determining reverse transcriptase (RT) activity in a [³H]thymidine-based assay on culture supernatants pelleted by ultracentrifugation (14, 31). To analyze Env proteins, viral supernatants were precipitated with Galanthus nivalis-agarose and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Western blotting with the anti-gp120 monoclonal antibody DA6 (14).

Immunofluorescence microcscopy.

Cells were spun onto glass slides (450 rpm for 5 min), fixed with a 1:1 solution of methanol and acetone (30 min, 22°C), and stained with anti-p27 monoclonal antibody 3A8 (kindly provided by Jan McClure, Bristol-Myers-Squibb, Seattle, WA) and a fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin antiserum.

Effects of CXCR4 and CCR5 inhibition on fusion and viral growth.

CXCR4 antagonists included AMD3100 (Langley, British Columbia, Canada) and T140 peptide, a generous gift from Stephen Peiper (Medical College of Georgia). CCR5 antagonists included Sch-D from the NIH AIDS Repository, Sch-351867 from Schering Plough Inc., Cmpd-167 from Merck Research Laboratories, and Tak-779 from Takeda Chemical Industries (Osaka, Japan). Cell-cell fusion assays (49) were performed in the presence of various concentrations of inhibitors, as previously described (44, 45). Briefly, quail QT6 target cells were cotransfected with a luciferase reporter construct under the control of a T7 promoter (T7-luc; Promega), CD4, and either CCR5, CXCR4, or pcDNA3.1(control) expression plasmid and were preincubated with different concentrations of drugs for 15 min at 37°C. QT6 effector cells transfected with Env expression plasmids and infected with vaccinia virus encoding T7 polymerase (vTF1.1) were incubated with the same concentrations of drugs and added to target cells. After incubation for 7 h at 37°C, fusion was quantified by assaying luciferase activity.

Generation of HIV-2 homology model, computational calculation of mutational substitutions, and electrostatic calculations.

The protein sequence of HIV-2/VCP gp120 was aligned with that of HIV-1/JR-FL using Clustal W. Conserved disulfide bonds, residues interacting with CD4, and conserved secondary structure were used for further alignment. The HIV-2/VCP sequences were then submitted to SWISS-MODEL using HIV-1/JR-FL (core+V3; protein data bank accession number 2B4C) as a template. The generated model was subjected to energetic minimization (crystallographic and nuclear magnetic resonance) and the minimized model examined visually in O and used to explore the effects of adaptive mutations and to generate electrostatic potential surfaces (GRASP software).

RESULTS

Derivation of functional Envs and viruses with V3 and V1/V2 deletions.

The HIV-2/VCP Env was selected to derive viruses with deletions of V3 and V1/V2. As previously described, this Env utilizes both human CXCR4 and CCR5 and was obtained from the vitro passaging of HIV-2/NIHz in SupT1 cells (14). Its use of CXCR4 is CD4 independent, while on human CCR5 it is strictly CD4 dependent (14, 33). HIV-2/VCP gp120 also binds to CXCR4 with high affinity, a feature that permitted the identification of residues on CXCR4 that are important for gp120 binding (32).

Mutagenesis strategies for deletions of the V1/V2 and V3 loops are shown in Fig. 1. A mutation was first introduced into the parental VCP Env that removed all but the N- and C-terminal cysteines and the first and last amino acids of V1/V2 and that contained a Gly-Ala-Gly (GAG) linker. This ΔV1/V2 Env retained approximately 80% of its fusion activity for CXCR4 and CCR5 and, like parental VCP, exhibited CD4-independent fusion on CXCR4 (Fig. 2A). When ΔV1/V2 was inserted into the full-length HIV-2 provirus, a replication-competent, highly fusogenic virus that exhibited rapid growth kinetics in SupT1 cells, comparable to that of wild-type VCP, was obtained (not shown), indicating that the loss of 85 amino acids from this region was easily tolerated. In contrast, removal of V3 except for the first and last amino acids (plus a GAG linker) resulted in an Env that was poorly functional in cell-cell fusion assays and unable to generate an infectious virus (not shown). However, a deletion that left the first and last six amino acids, not including the cysteines (Fig. 1), designated ΔV3(6,6), retained 8 and 60% of the fusion activity of parental VCP on CXCR4 and CCR5, respectively (Fig. 2A). When inserted into the HIV-2 provirus and electroporated into SupT1 cells, a spreading infection occurred, and cell-free virus obtained after one passage could replicate on SupT1 cells, albeit with slower kinetics than wild-type VCP (Fig. 2B). However, after this early ΔV3(6,6) virus (designated p1 in Fig. 2B) was serially passaged on SupT1 cells 16 times, it acquired rapid growth kinetics (Fig. 2B) and was able to induce cell killing and syncytium formation (not shown). An Env clone from this adapted virus, designated ΔV3(6,6)⁺a-p16, exhibited nearly a fourfold increase in fusion activity on CXCR4 and CCR5 relative to the unadapted ΔV3(6,6) Env (Fig. 2A). An HIV-2 containing the ΔV3(6,6)⁺a-p16 Env also showed rapid growth kinetics comparable to that of wild-type VCP (Fig. 2C). The sequence of the ΔV3(6,6)⁺a-p16 clone is described below and shown in Fig. 5, but the ΔV3(6,6) mutation remained intact during this adaptation, and no additional mutations were observed in the V3 remnant.

FIG. 1.

Deletion mutations introduced into HIV-2/VCP gp120. The left diagram shows the ΔV1/V2 deletion, and the right diagram shows ΔV3(6,6) and ΔV3(1,1) deletions, used to generate replication-competent V1/V2- and V3-deleted viruses. A Gly-Ala-Gly (GAG) linker (bold) was introduced with each deletion. Env clones containing both the ΔV3(1,1) and ΔV1/V2 deletion mutations were designated ΔV1/V2/V3.

FIG. 2.

Fusion activity and infectivity for variable-loop-deleted HIV-2/VCP Envs. (A) The fusion activities of the indicated Envs were evaluated in a cell-cell fusion assay on QT6 cells expressing CCR5 or CXCR4 with or without CD4 (34). Envs included parental VCP and VCP with the indicated deletions in V3 and/or V1/V2. A superscript + indicates Envs that contained adaptive mutations acquired during serial passaging of virus in SupT1 cells (shown in Fig. 5 and summarized in Table 1). (B) Viruses containing Envs with the ΔV3(6,6) deletion from an early passage (p1) and after 16 passages (p16) were added to SupT1 cells at the indicated p27 concentrations and RT activity followed over time. (C) Viruses containing the indicated Envs were inoculated onto SupT1 cells (20 ng/ml p27) and RT activity followed over time. (D) pACR23 plasmids containing HIV-2 proviruses with the indicated Envs were electroporated into SupT1 cells and RT activity measured. The ΔV1/V2/V3⁺ Env, which contained adaptive mutations acquired during passaging of viruses with ΔV3(6,6) and ΔV3(1,1) deletions but lacked adaptive changes acquired during serial passaging of a virus with the full V1/V2/V3 deletion, did not generate a spreading infection during this time, whereas ΔV1/V2/V3⁺a-p16A and -16B Envs cloned after serial passaging in SupT1 cells replicated comparably to parental VCP.

FIG. 5.

Sequence analysis of Env clones from variable-loop-deleted HIV-2/VCP. Amino acid sequences of adapted Env clones containing the ΔV3(6,6), ΔV3(1,1), and ΔV1/V2/V3 deletions are shown. Each of these Envs could generate replication-competent viruses (see growth curves in Fig. 2C and D). Deletions in V3 and V1/V2, with the Gly-Ala-Gly linker introduced, are indicated. Shown are the locations of gp120 conserved domains (C1 through C4) and variable loops (V1/V2, V3, and V4), the gp120/gp41 cleavage site, the predicted MSD, and predicted glycosylation sites (•). A summary of the observed mutations is presented in Table 1. The VCP parental clone contained a premature stop codon in the gp41 cytoplasmic tail. However, when this mutation was corrected in VCP and in the ΔV1/V2/V3⁺a-p16B Env clone, viruses remained replication competent, indicating that premature termination was not required for the replication competence of viruses lacking these variable loops (not shown).

Although the initial ΔV3(1,1) mutation was poorly functional when introduced on parental VCP Env, when introduced into ΔV3(6,6)⁺a-p16, this Env, termed ΔV3(1,1)⁺, which also contained mutations acquired during the adaptation of the ΔV3(6,6) virus, retained low but detectable fusion activity (Fig. 2A). When ΔV3(1,1)⁺ was inserted into the HIV-2 provirus and electroporated into SupT1 cells, a spreading infection occurred, albeit one with delayed kinetics relative to wild-type VCP, and produced a virus that could be serially passaged on SupT1 cells (Fig. 2C). Sequencing of genomic DNA from this culture after 36 passages showed that the ΔV3(1,1) mutation was retained but that additional mutations in gp120 and TM were acquired (see below). An Env clone from this culture, designated ΔV3(1,1)⁺a-p36, also retained a small but detectable amount fusion activity in the cell-cell fusion assay (Fig. 2A). When a ΔV1/V2 mutation was introduced into ΔV3(1,1)⁺a-p36, producing a clone termed ΔV1/V2/V3⁺ that also contained mutations acquired during previous adaptations (see below), this Env retained detectable fusion activity (Fig. 2A). Virus containing the ΔV1/V2/V3⁺ Env initially grew slowly but again acquired rapid growth kinetics after 16 passages in SupT1 cells (Fig. 2C, ΔV1/V2/V3⁺a-p16). Envs from this culture, designated ΔV1/V2/V3⁺a-p16A and -p16B, when introduced into the HIV-2 provirus exhibited rapid growth kinetics following electroporation into SupT1 cells (Fig. 2D), retained the ΔV3(1,1) and ΔV1/V2 mutations, and acquired additional mutations described below. In contrast, an HIV-2 proviral clone containing the ΔV1/V2/V3⁺ Env that lacked these additional mutations was unable to generate a spreading infection during this time period. Interestingly, while parental VCP and ΔV1/V2 Envs could utilize CXCR4 in the absence of CD4, all Envs with V3 deletions became strictly CD4 dependent (Fig. 2A), and viruses containing these mutations, although replication competent on SupT1 cells, were unable to infect BC7, a CD4-negative variant of SupT1 (not shown) (14).

Thus, with sequential rounds of mutagenesis and virus adaptation, replication-competent variants of HIV-2 lacking V3 and/or V1/V2 were derived. A summary of this stepwise derivation of HIV-2s lacking V1/V2 and V3 is shown in Fig. 3. PCR of an Env-containing region from SupT1 cells infected by viruses containing the ΔV1/V2, ΔV3(6,6)⁺a-p16, ΔV3(1,1)⁺a-p36, or ΔV1/V2/V3⁺a-p16A Env generated DNA fragments of the expected sizes, and Western blots of Envs from pelleted virions using an anti-gp120 antibody yielded proteins of 75, 115, 105, and 65 kDa, respectively, compared to 120 kDa for parental VCP (Fig. 4).

FIG. 3.

Scheme summarizing the stepwise derivation of env clones and viruses with variable-loop deletions. Mutagenesis steps in which the indicated deletions were introduced are in italic. Viruses were derived by electroporating full-length viral DNA into SupT1 cells and adapted by serial passage in SupT1 cells for the indicated number of times. env clones PCR amplified from the adapted viruses and described in the text are shown (boxed) and were used in cell-cell fusion assays (Fig. 2A) and to generate replication-competent viruses (Fig. 2B and C). env clones with a superscript + after the indicated deletion mutation contained additional mutations acquired during serial passaging in SupT1 cells.

FIG. 4.

Analysis of env DNA and gp120 from V1/V2- and V3-deleted viruses. (A) Genomic DNAs from cultures infected with viruses containing the indicated variable-loop deletions were amplified by PCR using primers that spanned the entire env gene. (B) Viral supernatants from these cultures were pelleted, solubilized, and analyzed by Western blotting with monoclonal antibody DA6, which recognizes a linear epitope in the C1 domain of gp120 (14). DNA fragments and Env proteins from variable-loop-deleted viruses exhibited the expected size reduction compared to parental HIV-2/VCP.

Env mutations acquired during adaptation of ΔV3- and ΔV1/V2-deleted viruses.

Aligned amino acid sequences of parental VCP and adapted Env clones containing the ΔV3(6,6), ΔV3(1,1), and ΔV1/V2/V3 mutations are shown in Fig. 5 and summarized in Table 1. Compared to parental VCP, the ΔV3(6,6)⁺a-p16 Env contained three mutations in gp120: V13I in the signal peptide, T391A at the base of the V4 loop, and V427I in the conserved C4 domain. V427I is located in a region corresponding to the HIV-1 bridging sheet domain (see below), while T391A eliminates a predicted N-linked glycosylation site that flanks this region (Fig. 6). Three mutations were observed in TM: L524V distal to the amino-terminal fusion peptide, A567T in the heptad repeat 1 domain, and A679T in the predicted membrane-spanning domain (MSD). Interestingly, in three independent adaptations of viruses with the ΔV3(6,6) mutation, the loss of the glycosylation site at position 391 and the L524V mutation in TM were observed, indicating that there were strong selection pressures for these changes when V3 was truncated (not shown). Adaptive changes in the ΔV3(1,1)⁺a-p36 Env clone, included the following: in gp120 E203K in the V1/V2 stem, N279D in the C2 domain, E333K and E336K distal to the V3 remnant, and E435V and L436F in C4, and in TM A580E distal to the heptad repeat 1 domain and V678I in the predicted MSD. Of note, all mutations acquired during adaptation of the ΔV3(6,6) virus were conserved when V3 was further truncated and viruses with the ΔV3(1,1) mutation were derived. Lastly, following introduction of the ΔV1/V2 mutation into the ΔV3(1,1)⁺a-p36 Env and adaptation, changes in the ΔV1/V2/V3⁺a-p16A and -p16B clones included, in gp120, I55T in C1, G402D in V4, and I416K and E433K (for ΔV1/V2/V3⁺a-p16B only) in C4. No additional TM mutations were seen in the ectodomain or the MSD in the ΔV1/V2/V3⁺a-p16A and -p16B adapted clones. Notably, no mutations in parental VCP were seen in bulk sequencing of genomic DNA after eight serial passages in SupT1 cells, indicating that the acquired changes occurred as a result of the V3 deletions with or without an associated deletion of V1/V2.

TABLE 1.

Amino acid mutations in Env clones obtained during the adaptation of V3- and V1/V2-deleted HIV-2/VCP viruses

Mutation	Env clone	Amino acid mutation ina:
		gp120												gp41 TM
		C1 (72, 55)	C2		C3		V4		C4					Ecto			MSD
			203, 203	278, 279	336, 333	339, 336	388, 391	410, 402	423, 416	434, 427	440, 433	442, 435	443, 436	534, 524	578, 567	591, 580	689, 688	690, 689
None	VCP	I	E	N	E	E	T	G	I	V	E	E	L	L	A	A	V	A
ΔV3(6,6)	ΔV3(6,6)+a-p16						A			I				V	T			T
ΔV3(1,1)	ΔV3(1,1)+a-p36		K	D	K	K	A			I		V	F	V	T	E	I	T
ΔV1/V2/V3	ΔV1/ΔV2/ΔV3+a-p16A	T	K	D	K	K	A	D	K	I		V	F	V	T	E	I	T
ΔV1/V2/V3	ΔV1/ΔV2/ΔV3+a-p16B	T	K	D	K	K	A	D	K	I	K	V	F	V	T	E	I	T

^aAmino acid changes relative to parental HIV-2/VCP are shown for Envs used to derive replication-competent viruses with deletions of V3 and/or V1/V2. Regions on gp120 and gp41 where mutations occurred are shown: conserved (C) and variable (V) regions in gp120 and the ectodomain (Ecto) and predicted MSD in gp41. Numbers indicate the amino acid number for HIV-2/VCP followed by the corresponding amino acid number for HIV-1/HXB.

FIG. 6.

Structural modeling and electrostatic analysis of HIV-2/VCP and ΔV3 viruses. (A) Structure of HIV-2/VCP gp120 (core with V3). The structure of HIV-1/JRFL (core with V3) in its CD4-bound conformation was used as a template to generate a homology model of HIV-2/VCP. A Cα-worm diagram of the HIV-2/VCP homology model is shown, with the inner domain in yellow, the outer domain in gray, the bridging sheet in orange, and V3 in red. In this orientation, CD4 would bind to front face of gp120, and the viral membrane would be positioned above and the target cell membrane below. All atoms are shown for residues altered during adaptation of viruses containing V3 deletions alone or in combination with V1/V2. These mutations are colored pink for ΔV3(6,6) (T391A and V427I), blue for ΔV3(1,1) (E203K, N279D, E333K, E336K, E435V, and L436F), and purple for ΔV1/V2/V3 (G402D, I416K, and E433K). All are in VCP numbering as shown in Fig. 5. (B) The electrostatic potential is depicted at the molecular surface (blue, electropositive; red, electronegative; white, apolar surfaces). The first column shows core plus V3, and the next three columns show the VCP core for parental Env, ΔV3(1,1)⁺a-p36 [containing the ΔV3(1,1) mutation], and ΔV1/V2/V3⁺a-p16B (containing the ΔV1/V2/V3 mutation). The middle row shows the same orientation as in panel A, the top row shows a view of the “bottom” (middle row orientation rotated about a horizontal axis), and the bottom row shows the view from the “right side” (middle row orientation rotated about a vertical axis). The figure was made using PyMol.

Modeling of adaptive mutations and their effects on gp120 core electrostatic potential.

The adaptive changes observed in the derivation of an HIV lacking V1/V2 and V3 loops revealed a striking acquisition in gp120 of positively charged residues and/or the loss of negatively charged residues at positions predicted to be within the bridging sheet domain (E203K, I416K, E433K, and E435V) and distal to the V3 remnant (E333K and E336K). To assess the impact of these changes in a structural context, we used homology modeling on the HIV-1/JRFL (core plus V3) structure to construct models of the VCP core gp120, with or without V3, as well as the core structures of the ΔV3(1,1)⁺a-p36 and ΔV1/V2/V3⁺a-p16B Envs. These homology models allowed us to carry out electrostatic calculations to assess the effects of the various deletions and adapted mutations on the electrostatic potential of the predicted solvent-accessible surfaces (Fig. 6).

Relative to parental VCP Env, the ΔV3(6,6)⁺a-p16 and ΔV1/V2/V3⁺a-p16B Envs acquired a surface of high positive electrostatic potential that included the bridging sheet domain and was contiguous with an outer face of the gp120 core as bounded by the C3 region distal to the V3 cysteines. Although the contribution of individual mutations to the replication competence of V3-deleted viruses awaits further study, these findings suggest that within gp120 loss of the positively charged V3 loop was compensated for by the acquisition of a positively charged surface on the gp120 core in a region flanking the previous site of the V3 loop.

Effects of V3 loop deletions on sensitivity to CCR5 and CXCR4 antagonists.

Due to the importance of V3 in mediating coreceptor interactions, we determined the impact of V3 deletions on the susceptibility of the loop-deleted Envs to coreceptor inhibitors (44, 45). Inhibitors first were evaluated for the ability to block cell-cell fusion (44, 45) of parental VCP Env and Envs from in vitro-adapted viruses containing the ΔV3(6,6), ΔV3(1,1), and ΔV1/V2/V3 mutations. The CCR5 inhibitors Sch-D and Cmpd-167 blocked fusion of parental VCP Env by more than 60% when target cells expressed CD4 and CCR5, with 50% inhibitory concentrations (IC₅₀s) of approximately 10 nM (Fig. 7 A and B). These compounds inhibited an R5-tropic HIV-1 Env with IC₅₀s of 8 and 5 nM, respectively. Remarkably, all Envs bearing a deletion or truncation in V3 were completely resistant to these R5 inhibitors. Resistance was also observed for Tak-779 and Sch-351867, two other small-molecule CCR5 antagonists (not shown). When target cells expressed CD4 and CXCR4, VCP Env fusion was blocked by CXCR4 inhibitors AMD3100 and T140 with IC₅₀s of 60 and 75 nM, respectively (Fig. 7C and D), while an X4-tropic HIV-1 Env was inhibited with IC₅₀s of 20 and 90 nM. As with the R5 inhibitors, V3-deleted or truncated Envs were completely resistant to these compounds, with IC₅₀s not being achieved for concentrations of up to 10,000 nM (Fig. 7C and D).

FIG. 7.

Resistance of V3-deleted Envs to R5 and X4 antagonists. (A, B, C, and D) The sensitivity of Envs containing variable-loop deletions was determined in cell-cell fusion assays run in the presence of the indicated concentrations of CCR5 inhibitors (Sch-D and Cmpd-167) and CXCR4 inhibitors (AMD3100 and T140). Results are expressed as a percentage of parental VCP Env fusion in the absence of inhibitor and represent the average ± standard error of the mean from three independent experiments. In each panel an R5-tropic (R3A) or X4-tropic (HXB) HIV-1 Env was used as a control. (E) AMD3100 sensitivity of parental HIV-2/VCP and viruses containing ΔV3(6,6) or ΔV1/V2/V3 deletions on infection of SupT1 cells. Cultures were inoculated with 10 ng/ml of p27 and RT activity monitored. While VCP could be completely inhibited, viruses with V3-deleted Envs were resistant.

Inhibition by AMD3100 was also assessed in viral infection assays on SupT1 cells for viruses containing the ΔV3(6,6)⁺a-p16 and ΔV1/V2/V3⁺a-p16A Envs (Fig. 7E). Although AMD3100 could completely inhibit VCP replication in a dose-dependent manner, with >50% inhibition at 100 nM and >95% inhibition at 1,000 nM, no inhibition of viruses with the ΔV3(6,6) or ΔV1/V2/V3 deletions was observed at concentrations up to 10,000 nM. Thus, in contrast to parental VCP Env, in cell-cell fusion assays and/or viral infection assays, Envs containing V3 deletions were resistant to CCR5 and CXCR4 antagonists.

DISCUSSION

The HIV Env variable loops V1/V2, V3, and V4 negotiate a critical balance in viral pathogenesis, protecting functional domains on the gp120 core from immune attack while permitting Env to bind to CD4 and chemokine receptors during entry. Although these loops face outward and elicit potent humoral immune responses, they can circumvent this attack through genetic variation. The V1/V2 and V4 loops vary markedly in length, amino acid content, and glycosylation site distribution, while V3 is highly conserved in length and varies principally in amino acid composition (18, 28, 63), although for HIV-1, V3 diversity is far greater among X4 isolates than among R5 isolates (22). Unlike V1/V2 and V4, the V3 loop also plays a key functional role in interacting with chemokine receptors, where it determines tropism for CCR5- or CXCR4-expressing cells (8, 9, 12, 20, 23, 39, 55, 69). While a direct interaction of V3 with a chemokine receptor has not been demonstrated (12, 18), compelling evidence from mutagenesis and peptide inhibition experiments (12) and from the recently determined V3 structure on a gp120 core (22) is consistent with a model in which the base of V3, contiguous with the bridging sheet, contacts the chemokine receptor N terminus, while more distal regions of V3 interact with the ECLs (2, 12, 18) (Fig. 8). How these events are coordinated in the context of conformational changes following CD4 binding, how cooperative interactions among variable loops contribute to entry and/or trimer stability, and how gp120 binding to the chemokine receptor leads to the release of TM and the initiation of membrane fusion remain crucial questions in understanding viral entry and the role of Env in immune evasion.

FIG. 8.

Diagram depicting gp120/chemokine receptor interactions and a proposed model for resistance of V3-deleted Envs to coreceptor antagonists. (A) Following or associated with CD4 binding, coreceptor interactions include (i) the bridging sheet and base of V3 with the CCR5 or CXCR4 amino terminus and (ii) more-distal regions of V3 with the chemokine receptor's ECLs (5, 12, 22). These events lead to the release of gp41 to mediate membrane fusion. (B) R5 and X4 antagonists bind to residues within or flanking the MSDs of the ECLs (17, 19, 53, 58) and are proposed to inhibit HIV entry principally by blocking V3/ECL interactions. (C) For Envs adapted to function with a truncated or absent V3, the amino terminus/bridging sheet interaction becomes sufficient for entry. With less dependence on V3, these Envs and viruses become resistant to R5 and X4 entry inhibitors.

In this report we show in an HIV-2 model that HIV can be adapted in vitro to retain infectivity in the absence of V3 alone or in combination with a loss of V1/V2. In our most minimized Env, a virus lacking V3 and V1/V2 was derived that expressed a “gp120” of only 65 kDa and that mediated rapid and highly cytopathic infection of SupT1 cells. Having HIV Envs that can tolerate complete or partial deletions of V3 while remaining competent for entry is remarkable given the central role played by V3 during chemokine receptor engagement (12, 18, 35). Derivation of ΔV3 viruses required repetitive rounds of deletion mutagenesis, viral adaptation, and Env cloning and depended on an accumulation of compensatory mutations on the gp120 core and TM. While an investigation of the contribution of these mutations to the function of ΔV3 Envs is under way, it is intriguing that mutations on the gp120 core included a striking accumulation of positively charged residues within and around the bridging sheet as well as the loss of a flanking N-linked glycan at position 391 (Fig. 5 and 6). Because this region likely interacts with the Tyr-sulfated, negatively charged N termini of CCR5 and CXCR4 (1, 5, 6, 15, 61, 68), these mutations may create an exposed, positively charged surface that mediates a core interaction with the N terminus and that in the absence of V3 becomes sufficient for subsequent entry events to proceed. Interestingly, adaptive mutations were also noted in TM, including an L524V mutation that was observed in independent derivations of ΔV3 viruses. In this context, TM mutations could facilitate poorly understood signaling events between gp120 and gp41 that ultimately free TM to form the six-helix bundle and drive membrane fusion (11, 37, 41).

Envs with either the ΔV3(6,6) or the ΔV3(1,1) deletion became highly resistant to CCR5 and CXCR4 antagonists that inhibit HIV-1. CCR5 inhibitors (Sch-D, Sch-351867, Tak-779, and Cmpd-167) all likely interact with residues lining a cavity formed by transmembrane helices of the ECLs (53, 60), while CXCR4 antagonists (AMD3100 and T140) likely interact with one or more negatively charged residues in membrane-proximal domains of the ECLs (17, 19, 58). Rather than directly inhibiting gp120 binding, these compounds are thought to alter conformational states of the ECLs, resulting in noncompetitive, allosteric inhibition (62). Because more-distal regions of V3 have been implicated to interact with the CXCR4 and CCR5 ECLs, particularly ECL2 (2, 5, 12), resistance of ΔV3 Envs to these compounds could be explained by the loss of a requirement for this interaction (Fig. 8). Although Envs containing a complete V3 deletion accumulated positively charged residues on the gp120 core, the ΔV3(6,6)⁺a-p16 Env, which lacked the distal half of V3 but did not contain these changes (Fig. 6), was also fully resistant to R5 and X4 antagonists. Therefore, the acquisition of positively charged residues on the gp120 core was not required for resistance to these inhibitors.

It is intriguing that Envs lacking V3 were capable of mediating fusion on CCR5 and CXCR4. For HIV-1, while both V3 and the bridging sheet are involved in coreceptor interactions, it is the V3 loop that determines coreceptor specificity (8, 9, 18, 20, 23, 55, 69). Although tropism determinants for HIV-2 are less well studied, V3 likely plays a similar role for this virus (24, 54). The bridging sheet and neighboring regions around the V3 base are comprised of amino acids that are well conserved among HIV-1 and HIV-2 isolates (7, 48, 64), suggesting that structural determinants that underlie its interaction with coreceptors are highly conserved. Thus, if the functional V3-deleted Envs derived in this study are dependent solely on a bridging sheet interaction, their ability to use both CXCR4 and CCR5 further suggests that this domain recognizes a motif in the N terminus that is shared by both receptors.

In contrast to parental VCP, which could utilize CXCR4 in the absence of CD4, all V3-truncated or -deleted Envs and viruses that we derived became strictly dependent on CD4 for infection. This finding is consistent with the view that without V3, CD4 is required to stabilize the bridging sheet and/or its interaction with the N terminus that leads to fusion. In this context, CD4 could enhance bridging sheet binding to the N terminus by decreasing its off rate and/or facilitating conformational changes required for gp41 triggering (38). Interestingly, although our studies show that for some viruses the V3 interaction can be dispensable for entry, Platt et al. have shown that HIV-1 can also be adapted to utilize chemokine receptors that lack an N terminus (42) and, presumably, the binding site for the bridging sheet. Given that CD4-independent viruses have been well described for X4 and R5 isolates (13, 14, 21, 26, 31), it appears that among the three interactions involved in HIV entry, i.e., (i) CD4 binding, (ii) the coreceptor N terminus binding to the bridging sheet, and (iii) the ECLs binding to V3, any one of them can be dispensable but at least two are required for fusion to proceed.

This study represents a proof of concept that at least some HIVs can be adapted to replicate without V3, with or without an associated V1/V2 deletion. What attributes enabled HIV-2/VCP to tolerate a loss of V3 are unclear, although this Env is remarkable in a number of ways, including its CD4 independence, promiscuous use of chemokine receptors, and high affinity for CXCR4 (14, 32-34). Although our findings have yet to be fully extended to HIV-1, we recently derived a replication-competent variant of a primary HIV-1 isolate with a deletion of approximately half of its V3 loop, alone and in combination with a V1/V2 deletion, and found that these Envs are competent for entry and resistant to CCR5 antagonists (30a). Thus, our findings in the current study are not unique to HIV-2.

Variable-loop-deleted viruses could have several practical applications. By isolating functions of the HIV gp120 core that are required for entry, they could be useful to screen for novel inhibitors of gp120/chemokine receptor interactions. Indeed, our findings suggest that the current generation of CCR5 or CXCR4 inhibitors all target a V3 interaction (Fig. 8) and raise the possibility that functional, V3-deleted Envs could be useful to identify compounds that inhibit interactions between the gp120 core and the chemokine receptor N terminus. Theoretically, such compounds might have the ability to target both R5- and X4-tropic viruses. In addition, because variable loops are immunodominant and may protect other conserved domains on the Env trimer, it is possible that Envs that retain entry functions in the absence of these loops will reveal novel and/or cryptic epitopes on gp120 and/or TM and elicit qualitatively different and more effective humoral immune responses. The ability to generate minimized HIV Envs that isolate functional core domains along with studies that explore the role of compensatory mutations that enable these Envs to mediate entry provide novel approaches to probe Env structure and function.