Please read an important announcement about the future of the HCV database here.

gp41, A Multifunctional Protein Involved in HIV Entry and Pathogenesis

Eric Hunter

Department of Microbiology and Center for AIDS Research, University of Alabama at Birmingham, Birmingham, AL 35294-2170

Introduction

The entry of HIV into a target cell represents the key initial step in vthe replication cycle of this virus. The entry process involves a coordinated series of molecular interactions between the two components of the virus glycoprotein complex (gp120(SU)/gp41(TM)) and the two components of the receptor complex (CD4 and a chemokine receptor). It results in the fusion of the viral membrane with that of the host cell and the introduction of the viral replicative machinery into the cytoplasm. It has been well established over the past several years that conserved domains of gp120 come together to form the binding site for amino-terminal residues of the CD4 molecule (Moore et al., 1993⁷³). For most HIV and SIV isolates, this SU-CD4 interaction defines the cell type to which the virus binds and ultimately enters (Sattentau and Moore, 1993⁹⁵). Recent data have shown, however, that for the process of membrane fusion to proceed, a second interaction, with a member of the chemokine receptor family of multiple membrane-spanning proteins, must occur. Natural isolates of HIV-1 which primarily infect macrophages and monocytes but lack the ability to infect T-cell lines utilize CCR5 , while T cell-line adapted and syncytium-inducing (SI) viruses primarily utilize CXCR4 (reviewed in Berger, 1997⁴; Doms and Peiper, 1997²⁶; Moore, 1997⁷²). Binding of the receptor complex results in conformational changes in both gp120 and in the TM protein gp41. By analogy with the hemagglutinin of influenza virus, the TM protein is thought to be the engine that drives the membrane fusion process, and it is this multifaceted molecule that is the subject of this review.

Biosynthesis of Env

The SU and TM components of the envelope glycoprotein complex are synthesized, as a single, co-translationally glycosylated, polyprotein precursor (gp160), on ribosomes associated with the endoplasmic reticulum (ER). Individual precursor molecules, shown schematically in Figure 1, initiate folding, presumably in the presence of ER-associated chaperonins, and assemble to form an oligomeric structure (Hunter and Swanstrom, 1990⁵⁷). Whether this is a trimeric or tetrameric molecule remains unresolved, although recent data suggest that a trimer is the biologically relevant form of the oligomer and this would be consistent with the findings for other retroviruses (see below). Mutagenic studies in both HIV and Rous sarcoma virus suggest that it is regions within the TM domain that direct and stabilize the oligomerization process (Earl and Moss, 1993³¹ Einfeld and Hunter, 1994³⁷ Einfeld and Hunter, 1997³⁸). It is likely that this region is structurally conserved between HIV and SIV, since mixed Env oligomers have been described (Doms et al., 1990²⁵). Surprisingly, the membrane-spanning domain is not required for oligomers to form, although it and regions within the cytoplasmic domain may provide stability to the structure (Earl and Moss, 1993³¹ Einfeld and Hunter, 1988³⁸). Formation of an Env oligomer appears to be required for efficient transport out of the ER.

Figure 1. Organization of the HIV-1 env gene product and functional domains of gp41. The env open reading frame is depicted with vertical arrows showing sites for cleavage by the host-encoded signal peptidase and furin-like Golgi proteinase. Oligosaccharide addition sites that are utilized in vivo are shown by vertical lines and CHO. The position and size of critical functional domains as well as T- and B-cell epitopes are shown to scale. The position of the cysteine-bounded loop that is part of epitope Cluster I is shown as two upper case Cs.

Shortly after the stable oligomer is formed, it is transported from the ER to the Golgi complex where the bulk of the mannose-rich oligosaccharide side chains are processed to hybrid/terminally glycosylated forms. In a late Golgi compartment, most likely the trans-Golgi network, the precursor is cleaved to gp120 and gp41 by a cellular proteinase. Furin and PC7 appear to be the most likely candidates for this action (Decroly et al.,1997²³ ; Hallenberger et al., 1997⁵⁰). These enzymes, members of the subtilysin family of proteinases, recognize and cleave at a highly conserved basic sequence (RXXR) that links the C-terminus of gp120 to the N-terminus of gp41. Cleavage is important since it frees the fusion peptide at the amino-terminus of gp41 from the C-terminus of gp120, which is critical for its future role in fusion. It is also probably accompanied by conformational changes in the oligomeric complex since uncleaved gp160 is inefficiently incorporated into virions (Bosch and Pawlita, 1990⁰⁸, Dubay et al., 1995²⁷). A post-cleavage conformational change is consistent with studies on influenza virus, where, in the crystal structure of the cleaved hemagglutinin, the C-terminus of HA1(SU) and the N-terminus of HA2(TM) are separated by 21 (Wilson et al., 1981¹²⁷). In the case of HIV Env, gp120 and gp41 are non-covalently associated following cleavage of the precursor. It can be concluded from an analysis of mutant glycoproteins that residues within conserved regions (C1 and C5) at the N-terminus and at the C-terminus of gp120 play a key role in this association (Helseth et al., 1991⁵³). In gp41, a region (residues 528-562) immediately carboxy-terminal to the fusion peptide is important for this association, although mutations outside this region can also be disruptive (Cao et al., 1993¹⁰).
It is important to emphasize that the stable oligomeric structure, which forms in the ER, is folded from uncleaved polyprotein precursors. This clearly allows the folding of regions of both SU and TM into domains that are structurally constrained in a manner that would be impossible if they were assembled from the mature proteins, gp120 and gp41 (see below). If one considers the folding of this constrained structure as the setting of a spring-loaded trap, cleavage of the precursor might then be analogous to the removal of the safety catch.

Immune Responses to gp41

Although the bulk of gp41 might be expected to be sequestered from the humoral immune system within the Env oligomeric complex, several regions of this molecule are recognized by sera from individuals infected with HIV. The first of these regions, often referred to as Cluster I extends from around residue 588 to residue 615 of gp41. The immunodominant epitope is centered around a short disulfide-bonded loop (residues 607-613) that is located within the central region of the gp41 ectodomain (1987b⁴⁸, Gnann et al., 1987c⁴⁹). Almost all sera from infected individuals react with this epitope even though some sequence diversity has been found in viruses of the O-subtype (Eberle et al., 1997³²). Nevertheless, antibodies to this epitope are not neutralizing and do not block infection by HIV. On the contrary, monoclonal antibodies reactive with this region have been reported to enhance infection through a complement-dependent mechanism (Robinson et al., 1990a⁸⁸ Robinson et al., 1990b⁸⁹). The immunologically active form of the seven-residue peptide, CSGKLIC, contains an intramolecular disulfide bond and NMR studies showed that this domain maintains a preference for a folded conformation, including a type I reverse turn about the residues SGKL. The presence of the disulfide bond is integral to the formation of the structure of the loop in solution. Mutations (Cys to Ser) which eliminate the possibility of loop formation result in the failure of antibody binding to this peptide (Oldstone et al., 1991⁸¹).
In addition to this primary immunodominant epitope, a second ectodomain region, located proximal to the membrane-spanning region of gp41, is also recognized by sera from infected individuals (Cluster II; Horal et al., 1991⁵⁵ Xu et al., 1991¹²⁸). While a majority of antibodies reactive with this region do not neutralize HIV infection, one human monoclonal antibody (2F5) has been identified that neutralizes a variety of laboratory strains and clinical isolates of HIV-1. The amino acid sequence ELDKWA (residues 671-676) has been defined as the epitope recognized by this antibody (Muster et al., 1993⁷⁹). Sequence analysis of a variety of primary isolates suggests that the major determinant of MAb 2F5 binding corresponds to the amino-acid sequence LDKW, since naturally occurring and in vitro selected neutralization-resistant viruses contain changes in the D and K positions of the ELDKWA motif (Purtscher et al., 1996⁸⁵). While 2F5 recognizes a putatively linear determinant present on short peptides, it seems likely that the epitope reflects a specific conformation of the contributing amino acids, since attempts to elicit neutralizing antibodies to a synthetic peptide encompassing this region or to the epitope presented on the hepatitis B surface antigen have been unsuccessful (Eckhart et al., 1996³³). However, insertion of this epitope into the loop of antigenic site B of the influenza virus hemagglutinin did elicit ELDKWA-specific, neutralizing immunoglobulins in antisera of mice (Muster et al., 1994⁷⁸). The mechanism of neutralization is not understood, although Neurath and colleagues have reported multiple effects of binding the 2F5 monoclonal antibody to virions that are suggestive of post-binding conformational changes in Env (Neurath et al., 1995⁸⁰). It is also relevant that the 2F5 epitope overlaps the amino acid sequence of a peptide derived from gp41 that is a potent inhibitor of fusion (see below) and a tryptophan-rich region that we have recently shown plays a critical role in fusion and infectivity (Salzwedel and Hunter, unpublished).
An enigma, as far as anti-gp41 neutralizing antibodies are concerned, has been the fact that monoclonal antibodies raised against a hydrophilic peptide corresponding to residues 743-760 (DRPEGIEEEGGERDRDRS) of the cytoplasmic domain could neutralize a variety of HIV-1 isolates. Both monoclonal antibodies and polyclonal antisera raised against this epitope have neutralizing potential (Dalgleish et al., 1988,²¹ Evans et al., 1989³⁹). There is currently no evidence that this region of the cytoplasmic domain can be translocated to the outer surface of the cell and so how such antibodies can neutralize has remained a puzzle. Recent experiments, however, indicate that a monoclonal antibody to this epitope can bind with equal efficiency to infected and uninfected cells (Sattentau et al., 1995⁹⁵), suggesting that the epitope is also present on a cell surface protein. Since it is clear that the lipid envelope of HIV incorporates a large number of cell surface proteins (Arthur et al., 1992⁰³), it is possible that neutralization by antibodies to this epitope is mediated through one of these cell-derived molecules.
The TM protein also elicits activity from the cellular arm of the immune response. One of the CTL epitopes, identified through the use of synthetic peptides and fusion proteins carrying portions of the HIV-1 env gene, overlaps with the immunodominant B-cell epitope (amino acids 591-602 of the most-likely HIV env sequence - AVERYLKDQQLL). Only two positions within this epitope showed variation among North American HIV-1 isolates, and the substitutions were conservative in nature (Johnson et al., 1992⁵⁹). Interestingly, a Lys to Arg substitution at position 597 abolished recognition, probably by interfering with the peptide-MHC interactions (Dai et al., 1992,²⁰ Hammond et al., 1991⁵¹). In additional studies, CTL from four subjects recognized at least three distinct epitopes in this region of gp41 in the context of the MHC class 1 alleles A24, B8, B14, and B27 (Shankar et al., 1996¹⁰¹). The other region of gp41 recognized by CTL is a stretch of 20 amino acids at the C-terminus of the cytoplasmic domain. CTL from seven subjects recognized epitopes within this region that were restricted by A30, B7, B8 and B35 class 1 alleles (Shankar et al., 1996¹⁰¹).

Role of gp41 in membrane fusion

a) Conformational changes on receptor binding
Several enveloped viruses including influenza virus, vesicular stomatitis virus, and the alpha-viruses require the acidic environment of the endosome to activate the fusogenic potential of their envelope glycoprotein complexes. Thus after binding a receptor molecule on the cell surface, they are taken up into coated pits, undergo endocytosis and then release their replicative machinery into the cell following fusion with the endosomal membrane. In contrast the entry of HIV is not dependent on exposure to an acidic environment (McClure et al., 1988,⁷⁰, Sinangil et al., 1988,¹⁰³ Stein et al., 1987¹⁰⁷).
It is known, for the hemagglutinin of influenza virus and the glycoprotein complex of Semliki Forest virus, that the acid environment of the endosomal compartment induces a conformational change in the proteins that is necessary for activation of the fusion process (reviewed in White et al., 198,¹¹⁸ White, 1992¹¹⁹). In the case of HIV, and SIV analogous conformational changes have been observed following binding of receptor molecules (Sattentau and Moore, 1991,⁹⁴ Sattentau et al., 1993⁹⁵), and recent experiments suggest that these receptor-induced conformation changes may parallel in molecular terms those induced by acid pH in the influenza virus HA.
Initially, experiments aimed at defining the mechanism by which a soluble form of the HIV receptor (sCD4) neutralized the infectivity of HIV demonstrated that treatment of laboratory strains of this virus with sCD4 could induce rapid shedding of gp120 and loss of the characteristic surface spikes seen in EMs of virions (Kirsh et al., 1990,⁶³ Moore et al., 1990⁷⁷). Dissociation was temperature dependent and no significant dissociation of gp120 was observed between 4° and 16° (Hart et al., 1991⁵² Moore and Klasse, 1992,⁷⁴ Moore et al., 1991,⁷⁶ Sattentau and Moore, 1991⁹⁴).
In the case of influenza virus HA, brief treatment of the virus with low pH, which activates fusion, results in the exposure of antigenic epitopes and protease cleavage sites that are normally sequestered within the HA molecule (Daniels et al., 1983,²² Skehel et al., 1982¹⁰⁴ White et al., 1983¹¹⁸). Similarly, dissociation of gp120 by CD4 is accompanied by exposure of epitopes on gp41 that are not reactive in the native molecule (Hart et al., 1991,⁵² Sattentau and Moore, 1991⁹⁴). Epitope exposure is not dependent upon loss of gp120 since similar results were obtained following CD4 binding at 4° (Sattentau and Moore, 1991,⁹⁴ Sattentau et al., 1995⁹⁷). A majority of studies have employed sCD4 treatment of HIV virions or infected cells, but Sattentau and Moore (1993) have demonstrated a similar exposure of gp41 epitopes at the interface of fusing cells. Thus membrane anchored CD4, like sCD4, can induce gp41 epitope exposure following gp120 binding.
The exposure of epitopes in both TM and SU, coupled with the dissociation of gp120 that has been observed with laboratory adapted isolates of HIV-1, appears to represent the most dramatic effect of sCD4 on Env conformation and may reflect the relative instability of the glycoprotein complexes of these isolates. The susceptibility of primary isolates of HIV-1 to sCD4 neutralization is significantly lower than that of cell-line adapted isolates (Daar et al., 1990¹⁹) and sCD4 is very inefficient at inducing gp120 dissociation in the non-adapted strains (Moore et al., 1992⁷⁴), perhaps reflecting the lower affinity for CD4 exhibited by the oligomeric form of Env from natural isolates (Moore et al., 1992⁷⁴).
The glycoproteins of HIV-2 and SIV also have a lower affinity for CD4 and these viruses are similarly resistant to neutralization by sCD4. In contrast to the laboratory adapted strains of HIV-1, both SIV and HIV-2 infection and syncytium formation can be enhanced by sub-inhibitory concentrations of the soluble receptor. Incubation of sCD4 with SIVagm viruses not only accelerates the kinetics of infection but also leads to a 100-fold increase in the titer of the virus (Allan et al., 1990,⁰¹ Clapham et al., 1991,¹⁷ Sekigawa et al., 1990¹⁰⁰ Werner et al., 1990¹¹⁷). A similar enhancement of infectivity has recently been observed with HIV-1 when poorly neutralizable primary isolates are treated with sub-inhibitory concentrations of sCD4 (Sullivan et al., 1995¹⁰⁸). Treatment of SIV or HIV-2 infected cells with the soluble receptor does not result in dissociation of gp120 or increased exposure of TM epitopes - apparently as a consequence of stronger interactions between the subunits in these viruses (Allan et al., 1992,² Sattentau et al., 1993⁹⁶). Nevertheless, it does result in some conformational changes, since an increase was observed in the exposure of the V2 and V3 loops, as detected by antibody binding (Sattentau et al., 1993⁹⁶), and antibodies from SIVagm infected monkeys, which failed to neutralize virus in vitro, could neutralize the sCD4-induced enhancement of SIVagm infection (Allan et al., 1990¹). This is consistent with the suggestion that sCD4 enhancement of SIVagm (and presumably of HIV-1 primary isolates and HIV-2) is a form of receptor mediated activation (Allan et al., 1990¹), analogous to the low-pH activation observed with influenza virus or Semliki Forest virus (Marsh and Helenius, 1989⁶⁹).

b) Molecular rearrangements following receptor binding - formation of coiled-coils
Both crystallographic and peptide studies with influenza virus have shown that the low pH-mediated conformational changes that occur in HA involve the structural transition of a region within the HA2 (TM) protein. Thus residues C-terminal to the fusion peptide, that are constrained as an extended polypeptide chain within the inactive HA, form an elongated helical structure following low pH activation (Bullough et al., 1994,⁰⁹ Carr and Kim, 1993¹¹). The newly formed helices of the HA trimer form a triple-helix, coiled-coil structure that is postulated to position the fusion peptide, located at the N-terminus of HA2, for insertion into the target cell membrane. A heptad repeat (leucine-zipper-like) region capable of forming a similar coiled-coil structure is found within the N-terminal 100 amino acids of the TM protein of a majority of retroviruses. Peptides corresponding to this domain form stable coiled-coil structures (Wild et al., 1992¹²³) and fusion of the heptad repeat to the C-terminus of a monomeric form of bacterial protein-A or maltose-binding protein results in assembly of oligomeric chimeric proteins (Bernstein et al., 1995,⁵ Shugars et al., 1996¹⁰²). Thus this region has the capability to both oligomerize and mediate oligomerization of protein molecules. The heptad repeat domain is highly conserved among HIV-1 isolates and was originally postulated to play a role in the assembly oligomerization of the glycoprotein complex that occurs in the ER (Delwart and Mosialos, 1990,²⁴ Gallaher et al., 1989⁴⁶). Mutational analyses of the region have argued against a direct role in oligomer assembly in the ER but showed that even minor amino acid substitutions had dramatic effects on the biological activity of the HIV-1 glycoprotein, by blocking cell-cell fusion and virus infectivity (Chen, 1994,¹⁵ Chen et al., 1993,¹⁶ Dubay et al., 1992²⁸). Peptides corresponding to this region are potent inhibitors of HIV-1 induced cell fusion as well as virus infection, and this property correlates directly with the ability of the peptides to form stable coiled-coil structures; the same mutations which block the biological function of Env also reduce the melting temperature of the peptide's coiled-coil structure as well as abrogate their inhibitory potential (Wild et al., 1994b¹²⁰). We have postulated therefore that binding of CD4 to gp120 might induce equivalent conformational changes to that seen during low pH treatment of the influenza virus HA, with the resulting formation of an extended coiled-coil that can present the N-terminal fusion peptide to the target cell membrane (Figure 2A and 2B) (Bernstein et al., 1995,⁰⁵ Wild et al., 1994b¹²⁰). Such a mechanism would explain why mutations which prevent cleavage of the Env precursor effectively block fusion, since in the absence of a free amino-terminus, the fusion peptide would be unable to undergo such a dramatic displacement.

Figure 2. Ribbon diagram of the neutral and low pH forms of the influenza virus HA. A. Neutral pH form of the HA monomer crystal structure - HA1 is shown in grey, HA2 in colors that correspond to the amino terminal helix, the central extended chain and the carboxy-terminal helix (Wilson et al., 1981). B. Crystal structure of the low pH form of HA2 showing the formation of an extended helix following a major conformational change (Bullough et al., 1994). Note that the region of extended polypeptide chain between the N-terminal (yellow) helix and the C-terminal (red) helix is incorporated into the extended helix following acidification. The fusion peptide can thus be moved approximately 100.

c) Insertion of the fusion peptide
While the formation of a coiled-coil structure might present the fusion peptide to the target cell membrane, the mechanism of action of the fusion peptide itself remains obscure. How, for example, does the amino-terminal fusion peptide of gp41 interact with the membrane and what molecular interactions must occur following insertion? Mutational analyses of the fusion peptides of HIV-1, HIV-2 and SIV, which show sequence homology to the fusion peptides of the paramyxoviruses (Gallaher, 1987⁴⁵), have demonstrated the importance of hydrophobic residues within this region; substitution of hydrophobic with hydrophilic residues reduces syncytium formation and virus infectivity, whereas enhancement of hydrophobicity increases both (Bosch et al., 1989,⁷ Freed and Myers, 1992,⁴² Freed et al., 1990,⁴³ Steffy et al., 1992¹⁰⁶). Recent NMR analyses of the gp41 fusion domain indicated that a peptide corresponding to this region inserts into lipid micelles primarily as a helix (59%), with substantial beta-structure (26.7%). Fluorescence experiments with a tryptophan analogue of the peptide pointed to deep penetration of the peptide into the apolar hydrocarbon core. The region C-terminal to an alanine-15/glycine-16 flexible hinge interacts in a substantial way with the micelle, suggesting that it lies on the surface of the bilayer (Chang et al., 1997¹³). Despite this data, it is still not clear if, in order to function during virus entry, the fusion peptide interacts only with lipid or whether it also interacts with membrane-associated proteins in a manner analogous to the signal peptides of secreted and membrane-spanning proteins.

d) Structural analyses of gp41: implications for the fusion process
Although the three-dimensional crystal structure of the influenza virus HA was established over 15 years ago (Wilson et al., 1981¹²⁷), similar attempts to determine the structure of the intact Env complex of HIV have been unsuccessful. Using the known structure of HA2 and structure-prediction algorithms, Gallaher (1989) proposed a secondary structure for the HIV TM that included a number of structural motifs common to the TMs of diverse retroviral isolates. While this model has been conceptually valuable, protein-structure prediction remains unreliable, particularly for membrane proteins, and, more importantly, it fails to take into account the potential for formation of intermediate, quasi-stable structures such as those observed in HA. Within the past two years, however, significant progress has been made in obtaining high-resolution structure information on the fusion-active form of the HIV TM.
Following expression of the ectodomain of gp41 in bacteria, the insoluble product could be converted to a soluble form through limited proteolysis. This generated two peptide fragments N51 and C43, derived from the N- and C-regions of the gp41 ectodomain respectively (Lu et al., 1995⁶⁸). The N51 peptide corresponds to the 4-3 heptad repeat region (residues 549-599) discussed previously, while the C43 peptide is derived from a C-terminal potentially helical region (residues 633-675). Biophysical studies showed that these two peptides associate to form a highly thermostable, helical, trimeric complex of heterotrimers in which the N51 and C34 helices are arranged in an anti-parallel fashion (Figure 3A; Lu et al., 1995). An almost identical, antiparallel, coiled-coil bundle could be isolated using the same approach with the SIV TM ectodomain (Blacklow et al., 1995⁰⁶). This provided the initial indication that the N- and C-terminal regions of TM can fold back onto themselves to form a stable, predominantly helical, complex. While these helical bundles lacked the interconnecting 33 amino acid region containing the immunodominant epitope (residues 600-632), expression of the intact gp41 ectodomain (minus the fusion peptide) in insect cells yielded soluble oligomeric (trimer) proteins that appear in the electron microscope to form similar rod-like structures and which by circular dichroism have high alpha-helical content (~80%) (Weissenhorn et al., 1996¹¹⁶). Thus the presence of the cysteine-loop region does not appear to interfere with the anti-parallel packing of the N- and C-terminal helices.

Figure 3. Coiled-coil bundle formed by the N-terminal and C-terminal helical domains of gp41. Panel A shows proposed packing of helices in protease-resistant core from bacterially expressed gp41. The 33 amino-acid connecting loop with cysteine-bonded loop that missing from the structure determined by Lu et al., (1996) is included schematically in this figure. Panel B shows a ribbon diagram of the coiled coil bundle formed by N36(555-590) and C34(637-670) peptides from gp41 and Panel C shows a side-view of the space-filling model of the same structure (from the co-ordinates of Chan et al., 1997). White spheres are water molecules.

Figure 4. Hypothetical two-stage model for the conformational changes resulting in membrane proximity during fusion. A. Model for trimeric HIV Env complex. One SU molecule is cut away to show N-terminal heptad repeat constrained in an unstructured form. B. Following binding of the receptor complex, the extended-chain, heptad repeat region (shown in cyan) can assemble into a trimeric coiled-coil, thereby allowing the fusion peptide to insert into the target cell membrane. C. A second (rate limiting?) step in the conformational change is the formation of the coiled-coil bundle which results in apposition of the viral and target cell membranes. Peptide inhibitors might act by competing for the C-terminal heptad repeat domain during this latter step.

In the past year, two groups have succeeded in obtaining high resolution crystallographic data on the helical bundle formed by the N- and C-terminal domains of gp41. Chan et al. (1997) determined the structure of a slightly smaller complex, N36 (residues 555-590)/C34 (residues 637-670), to 2.0 Å resolution. This complex is a six-stranded helical bundle (fig 3B). The center of the bundle consists of a trimeric coiled-coil of three N36 helices arranged in parallel in a left-handed super helix. Three C34 helices are wrapped, antiparallel to the N36 helices, in a left-handed direction around the central coiled-coil trimer (Fig 3C). Overall the complex comprises a cylinder that is 35 Å in diameter and 55 Å in height. Using a somewhat different approach, Weissenhorn et al. (1997) determined the structure of the protease-resistant part of the gp41 ectodomain solubilized with a trimeric GCN4 coiled-coil in place of the amino-terminal fusion peptide. This structure, in which gp41 residues 550-597 form the triple-stranded alpha-helical coiled-coil core and residues 637-674 form the alpha-helices packed in the reverse direction against the outside of the coiled coil, is essentially identical to that of Chan et al. (1997). The C-terminal helices interact with the central coiled-coil core mainly through hydrophobic residues that are located in three grooves on its surface and the residues lining these grooves are highly conserved between HIV and SIV. It is important to note that peptides corresponding to the C-terminal helices do not have helical character in solution (Wild et al., 1995,¹²² Wild et al., 1994a¹²⁴) and so the formation and stability of these helices must depend to a large extent on their interaction with the triple-stranded central core. Indeed, it is quite possible that within the native Env complex on the virion, both the N-terminal heptad repeat region and the C-terminal domain are constrained as extended polypeptide chains, and that their transition to helical structure occurs only after receptor binding.
The trimeric coiled-coil complex has implications for the valency of the Env complex itself, for the structural rearrangements of Env involved in membrane fusion and for the development of novel antiretroviral drugs. The HIV Env complex has been described as a tetramer and as a dimer of dimers (Earl et al., 1990,³⁰ Pinter et al., 1989,⁸⁴ Schawaller et al., 1989⁹⁹) as well as a trimer (Gelderblom et al., 1987,⁴⁷ Weiss et al., 1990¹¹³). The trimeric nature of the gp41 coiled-coil complex, even though it is formed in the absence of gp120, clearly suggests that the native Env complex is also trimeric. This would be consistent with other retroviral Env complexes (Einfeld and Hunter, 1988,³⁵ Einfeld and Hunter, 1990,³⁶ Kamps et al., 1991⁶¹) and with the influenza virus HA (Wilson et al., 1981¹²⁷).
The remarkable structural similarity of the HIV gp41 coiled-coil bundle to the fusion pH-induced conformation of influenza virus HA2 (Bullough et al., 1994⁹) and to the transmembrane subunit of Moloney murine leukemia virus (Fass et al., 1997⁴⁰) led both Chan et al. (1997) and Weissenhorn et al. (1997) to propose that this structure represents the fusion-active form of the gp41 ectodomain. A number of factors support this conclusion. First the coiled-coil bundle is very thermostable, in contrast to the native Env complex, indicative of a terminal low energy conformation. Second, mutations that block fusion and infectivity map to residues that stabilize the coiled-coil bundle. For example the isoleucine at position 582, which we have shown is critical for fusion and for stability of the core triple helix (Wild et al., 1994b¹²⁰), makes a homotrimeric contact at the center of the coiled coil but also interacts with Trp640 in the outer helix . Similarly, Asn665 is in an a position (relative to the a-g positions of a heptad repeat) on the outer helix and packs into the groove on the central core. Mutation of this residue would be predicted to disrupt an extensive hydrogen bonding network that stabilizes the outer helix and, as might be expected if this were the case, abrogates fusion (Cao et al., 1993¹⁰). Finally peptides corresponding to the N- and C-terminal helical domains that make up the coiled-coil bundle have potent anti-viral activity and block HIV fusion (Jiang et al., 1993,⁵⁸ Lu et al., 1995,⁶⁸ 1994b; Wild et al., 1993,¹²¹ 1994a¹²⁴). Given the stability of the coiled-coil structure, these peptides would not be expected to exert their antiviral effect once it has formed. In contrast, the structure supports the concept that the inhibitory peptides function by interacting with gp41 during the conformational changes to the fusion-active state. Thus the N-terminal peptides might block the formation of the internal coiled-coil or interact (as oligomeric peptides) with the C-terminal region, preventing its interaction with the N-terminal domain. C-terminal peptides might act competitively to prevent binding of this region of gp41 into the grooves of the central core. Indeed, a peptide comprising residues 647-682 (DP178) can bind to a bacterially produced gp41 ectodomain that has been deleted in this region. Moreover, the formation of this ``molecular clasp'' reconstituted an epitope recognized by a monoclonal antibody prepared against the intact ectodomain of gp41 (Chen et al., 1995¹⁴).
Taken together with the high resolution structural information, the data from mutations and inhibitory peptides points to a model for Env-induced fusion which involves two interdependent conformational changes in gp41 (Fig 4). In the first stage, following binding of the receptor complex, the constrained N-terminal heptad-repeat region of gp41 would be freed to assemble into the central coiled-coil, moving the fusion peptide in the process to a position where it could interact with the target cell membrane. The second stage would then be the formation of the coiled-coil bundle. Since this latter structure is arranged such that the N-terminal fusion peptide and the C-terminal membrane-spanning domain of gp41 are at the same end of the bundle, its formation would force the target cell membrane and the viral membrane in close proximity. This mechanical process thus might facilitate the initiation of the process where lipid mixing occurs. Neither the source of energy for the second conformational change and membrane apposition, which would be expected to have a high DG, nor the time frame over which such a process might occur is obvious. However, the fact that peptides such as DP178 can inhibit the fusion process at very low concentrations (pM), suggests that the formation of the coiled-coil bundle might be a rate limiting step in fusion.

e) gp41/fusion as an anti-viral target
It is clear from the work of several groups that peptides corresponding to the N-terminal heptad repeat (DP107) and the C-terminal helical domain (DP178, C34) represent potent inhibitors of HIV-1 entry and replication (Jiang et al., 1993,⁵⁸ Lu et al., 1995,⁶⁸ 1994b; Wild et al., 1993,¹²¹ 1994a¹²⁴). The possible mechanism of action of these peptides is discussed above and suggests that the conformational changes in gp41 are a novel target for anti-HIV therapeutics. A particularly important development of this concept, therefore, has been the initial clinical trial of the C-terminal (DP178) peptide in HIV-1 infected patients. In this study, 5 of 5 patients, receiving 100mg of the peptide twice daily for a period of 17 days, showed greater than 100-fold reductions in viral load (Saag et al., 1997⁹²) - an anti-viral effect greater than several RT inhibitors. Thus it appears that the fusion/entry process is a viable in vivo therapeutic target. The presence of the highly conserved, deep cavities on the N-terminal coiled-coil trimer with which conserved C-terminal residues interact, raises the possibility that novel small molecule or peptidomimetic inhibitors of HIV infection might be developed (Chan et al., 1997,¹² Weissenhorn et al., 1997¹¹⁵).

f) Requirement for a membrane anchor
The membrane-spanning anchor region of the TM protein is an important structural domain, as might be expected if the initial result of fusion peptide insertion is an oligomeric structure that acts as a bridge between the viral and cell membranes. The region of gp41 that spans the membrane to yield a fusion-active form of the HIV-1 Env remains ill-defined, but C-terminal truncations indicate residues 692-716 can function as a membrane anchor (Dubay et al., 1992,²⁸ Helseth et al., 1990,⁵⁴ Johnston et al., 1993⁶⁰). Point mutations and small deletion mutations within this region can abrogate fusion without affecting the membrane-anchor function (Helseth et al., 1990,⁵⁴ Owens et al., 1994⁶⁰). Nevertheless, because the membrane-spanning and cytoplasmic domains of the HIV-1 glycoprotein can be replaced by those of CD4 or CD28 without loss of fusogenicity (Vincent et al., 1993,¹¹² Wilk et al., 1996¹²⁶), there appears to be no requirement for a specific amino acid sequence within the transmembrane domain for Env to mediate the fusion reaction. In contrast, substitution of a glycosylphosphatidylinositol (GPI) anchor for the HIV-1 protein anchor abrogates fusion and infectivity, even though this GPI-linked protein is efficiently transported to the plasma membrane and incorporated into virions (Salzwedel et al., 1993,⁹³ Weiss and White, 1993¹¹⁴). A similar result has been reported for a GPI-linked form of the influenza virus HA protein, although in this case lipid mixing of the outer leaflet of the bilayer (hemi-fusion) was seen in the absence of complete fusion and mixing of cell contents (Kemble et al., 1994⁶²). Thus a protein component that spans both leaflets of the lipid bilayer appears to be critical for complete mixing of the viral and target membranes. Weissenhorn et al. (1997) have suggested that the rod-shaped gp41 coiled-coil bundles might cluster at their hydrophobic tips (which includes the membrane-spanning domain), thereby forming asters with centers that could be the points of initial membrane fusion.

Role of the HIV/SIV Cytoplasmic Domain

The TM proteins of lentiviruses have unusually long cytoplasmic domains. In HIV and SIV they exceed 150 amino acids and in the case of Visna virus it is more than 200 amino acids in length. This contrasts with most retroviral TMs where the cytoplasmic domain ranges from 20-40 amino acids (Hunter and Swanstrom, 1990⁵⁷). In an attempt to define the biological role of the additional residues in the cytoplasmic domain, several investigators have introduced premature termination codons in both the HIV-1 and SIV Env coding sequence to progressively truncate the cytoplasmic domain. The results of these experiments indicate that the cytoplasmic domain is dispensable for HIV or SIV Env-mediated cell-cell fusion. In the case of HIV-1, however, truncation of the cytoplasmic domain by as few as 20 amino acids results in significantly decreased virus replication in most cell types, although the virus is still viable in the highly permissive MT4 cell line (Dubay et al., 1992,²⁷ Freed and Martin, 1995,⁴¹ Gabuzda et al., 1992,⁴⁴ Wilk et al., 1992¹²⁵). SIV replication, in contrast, appears to have no requirement for the cytoplasmic domain of Env (Johnston et al., 1993,⁶⁰ Zingler and Littman, 1993¹²⁹). Indeed growth of SIV in human cells results in the selection for a viral Env with a spontaneous truncation which broadens the host range of the virus, probably by increasing the number of Env molecules incorporated into virions (Johnston et al., 1993⁶⁰) (Zingler and Littman, 1993¹²⁹). Nevertheless, reversion back to wild-type virus has been observed following inoculation of virus encoding the Env truncation into macaques, suggesting that in vivo this region plays an important role (P. Luciw, personal communication; Kodama et al., 1989⁶⁴). A number of structural elements within the cytoplasmic domain may contribute to this in vivo function.

Endocytosis/Basolateral Targeting Signals

Cytoplasmic amino-acid motifs can function to direct endocytosis of a protein from the plasma membrane (reviewed in Trowbridge et al., 1993¹¹¹). Over the past few years, the sequences of these internalization/targeting signals have been identified for several constitutively reycling cell surface receptors and have been shown to involve tyrosine residues within a tight turn of the peptide backbone (Robinson et al., 1996,⁸⁷ Trowbridge et al., 1993¹¹¹). For the transferrin receptor (TR) this sequence is Tyr-X-Arg-Phe and a general consensus for this type of internalization motif is Tyr -X-Arg/Lys-Y, where Y is usually a hydrophobic amino acid (Collawn et al., 1990¹⁸). We noted previously (Hunter, 1994) that a majority of retroviral cytoplasmic domains contained a similar motif. In HIV and SIV, a Tyr-X-Pro-Leu/Val sequence, located membrane-proximally (residues 721-724) within the cytoplasmic domain, is highly conserved. Hoxie and colleagues have described a spontaneous mutant of an SIV Env protein in which the tyrosine in this motif was substituted by a cysteine. This mutant was poorly endocytosed, and was expressed on the cell surface and incorporated into virions at high levels (LaBranche et al., 1995,⁶⁵ Sauter et al., 1996⁹⁸). Similarly, Siliciano and colleagues have shown that this tyrosine containing motif is important for efficient endocytosis of the HIV-1 Env molecule (Rowell et al., 1995⁹¹). Additionally, Sauter et al. (1996) were able to demonstrate directly the role of the SIV motif in endocytosis by substituting the cytoplasmic domain of Env for that of CD4.
What is the rationale for conserving signals that result in the internalization and degradation of a protein that must be incorporated into virions for infectivity? It has been proposed that this process might reduce immunological recognition of infected cells; alternatively, the motifs might play a role in correct intracellular targeting. In epithelial cells, where the apical and basolateral membranes have distinct protein populations, retroviral glycoproteins are specifically targeted to the basolateral surface (Roth et al., 1983⁹⁰). When HIV-1 Gag and Env are co-expressed in epithelial cells, it is at this surface that virus assembly and release occurs (Lodge et al., 1994,⁶⁶ Owens et al., 1991⁸³). Mutation of the same tyrosine-containing motif that directs endocytosis abrogates this basolateral targeting of HIV-1 Env and basolateral assembly of virions (Lodge et al., 1997⁶⁷).
Interestingly, the rate of internalization of Env protein from the surfaces of HIV-1-infected cells is slow, and the presence of the Pr55gag precursor protein appears to be necessary and sufficient to inhibit Env internalization (Egan et al., 1996³⁴). Thus an interaction of the matrix domain of the Gag protein with Env during assembly may preclude the interaction of Env with host adaptin molecules that recruit plasma membrane molecules into clathrin-coated pits and thereby effect incorporation into virions.

Calmodulin-binding domain

The cytoplasmic domains of HIV-1, HIV-2 and SIV contain two regions that would be predicted to fold into amphipathic helical segments which closely resemble those found in calmodulin-activated enzymes. Synthetic peptides corresponding to these regions have been shown to bind to calmodulin (CaM) with high affinity in the presence of calcium (Miller et al., 1993,⁷¹ Srinivas et al., 1993¹⁰⁵). These amphipathic peptides inhibit CaM-regulated activation of bovine brain phosphodiesterase in vitro, and have also been shown to inhibit mitogen-induced lymphocyte activation, a property shared by CaM antagonists. These properties are not an artifact of peptides since purified HIV-1 gp160 as well as full-length gp41 binds to CaM and expression of HIV Env in cells results in CaM co-localization (Radding et al., 1996⁸⁶). Moreover, while CaM binding by the C-terminal amphipathic helix is very sensitive to amino acid changes, the natural variation found in this region of Env leaves CaM-binding intact (Tencza et al., 1997,¹⁰⁹ Tencza et al., 1995¹¹⁰). Since calmodulin is intimately involved in the metabolism and functioning of T-cells, these studies point to a novel mechanism of viral cytopathogenesis mediated by the interaction of the HIV TM protein with cellular CaM, that could in turn uncouple critical cellular signal transduction pathways (Miller et al., 1993,⁷¹ Srinivas et al., 1993¹⁰⁵).

Conclusions

Despite the wealth of new information summarized above, much remains to be learned about the role of gp41 in HIV/SIV replication and pathogenesis. The similarities between the HIV Env complex and the HA oligomer of influenza virus has allowed functional extrapolations to be drawn from the limited structural information obtained from bacterially expressed proteins. Nevertheless, even moderate resolution structural data on the native Env complex would provide information critical to understanding the extent of the conformational changes that have only been inferred to-date. High resolution three-dimensional structures of the native molecule would clearly assist in the development of small molecule inhibitors that could block the postulated conformational changes in gp41 following receptor binding and would thus interfere with the process of membrane fusion. It seems likely that the long cytoplasmic domain of the lentiviruses plays a key role in viral pathogenesis. In SIV the cytoplasmic domain can be truncated without affecting virus replication in vitro, yet the conservation of endocytosis signals, CaM binding domains, as well as structural elements that interact with the assembling capsid precursors, suggests that this region might play multiple roles in both virus replication and pathogenesis in vivo. Experiments aimed at defining the structure of the cytoplasmic domain and the cell-encoded molecules with which it likely interacts could yield important information in this regard. The HIV/SIV TM protein provides an excellent example of how multiple, complex functions can be packaged in a small molecule. Although much has been learned in the last few years about the function of gp41, even more remains to be uncovered. The next few years promise to yield exciting new information both on it and the molecular events involved in virus-cell membrane fusion.