Structure of a Murine Leukemia Virus Receptor-Binding Glycoprotein at 2.0 Angstrom Resolution

See allHide authors and affiliations

Science  12 Sep 1997:
Vol. 277, Issue 5332, pp. 1662-1666
DOI: 10.1126/science.277.5332.1662


An essential step in retrovirus infection is the binding of the virus to its receptor on a target cell. The structure of the receptor-binding domain of the envelope glycoprotein from Friend murine leukemia virus was determined to 2.0 angstrom resolution by x-ray crystallography. The core of the domain is an antiparallel β sandwich, with two interstrand loops forming a helical subdomain atop the sandwich. The residues in the helical region, but not in the β sandwich, are highly variable among mammalian C-type retroviruses with distinct tropisms, indicating that the helical subdomain determines the receptor specificity of the virus.

Retroviruses are simultaneously a profound human medical problem and a potential medical solution. They can be pathogenic, causing immunodeficiency, leukemia, and neurological disease, but they are also actively studied for their proposed utility as gene therapy vectors. Essential to both roles is the targeting of the virus to the host cell through interactions between viral envelope proteins and cell surface proteins.

Retrovirus envelope glycoproteins (1) are synthesized as single chain precursors that are subsequently cleaved into two subunits, the surface (SU) and the transmembrane (TM) (Fig.1). The SU glycoprotein binds the receptor. The TM subunit contains the hydrophobic fusion peptide and transmembrane segments and is likely to participate directly in the fusion of the viral and cellular membranes after receptor binding.

Figure 1

Domain organization of the Fr-MLV envelope glycoprotein and sequence of Fr-RBD. (A) The Fr-MLV envelope glycoprotein is synthesized as a single polypeptide chain, shown schematically (SS, signal sequence; FP, fusion peptide region; TM, transmembrane segment). Maturation involves removal of the signal sequence, cleavage to generate distinct SU and TM subunits, and removal of a short peptide at the COOH-terminus. Cleavage sites are indicated by dotted lines. The crystallized Fr-RBD region is shaded. Amino acid residues at the junctions of each indicated segment are numbered. (B) Secondary structural elements are indicated above the amino acid sequence of Fr-RBD. β strands are numbered and shown as black bars, with strand 3 dotted because it does not classify as a true strand (9). α helices are lettered and shown as gray bars and 310 helices are dotted in gray. N-linked glycosylation sites are underlined. Regions conserved among MLVs are boxed in gray, whereas regions with lengths and sequences that vary with the tropism of the virus are boxed in blue and labeled VRA, VRB, and VRC (13). N, NH2-terminus; C, COOH-terminus.

Efforts to understand the structural basis of retroviral binding and entry and to develop retrovirus-based gene therapy vectors (2) with engineered tropisms have been hindered by the lack of high-resolution models for the receptor-binding regions of retrovirus envelope glycoproteins. To address this problem, we initiated structural studies on the envelope glycoprotein of the mammalian C-type retrovirus, Friend murine leukemia virus (Fr-MLV). The cellular receptor for Fr-MLV and other ecotropic MLVs (3) is the 14–transmembrane-pass cationic amino acid transporter MCAT-1 (4). The receptor-binding function of Fr-MLV has been localized to the NH2-terminal region of the SU glycoprotein (5,6), which has been defined as a domain by proteolysis (7); thus, we refer to this region as the Friend receptor-binding domain (Fr-RBD) (Fig. 1). Here, we present the x-ray crystal structure (8) of Fr-RBD at 2.0 Å resolution (Table 1).

Table 1

Multiple isomorphous replacement and refinement statistics. Diffraction data were collected by means of a rotating anode source (Rigaku RU200) with an RAXIS IIc detector. All data were collected at −180°C with a Molecular Structure Corporation cryogenic crystal cooler (X stream). Crystals were mounted in a 20-μm rayon fiber loop and flash-frozen in the gaseous nitrogen stream. Reflections [average I/σ(I) of 10.6] were indexed and integrated with the program DENZO (28) and were scaled with SCALEPACK (28). Subsequent data manipulations were carried out with the CCP4 package (29). We accomplished heavy atom derivatization by harvesting crystals into reservoir solution, adding solid heavy atom to 1 mM, and soaking overnight. Heavy atom sites were located for each derivative by Patterson methods. Heavy atom refinement and phasing (overall figure of merit 0.599) were carried out with MLPHARE (30), and the sites for each derivative were verified by calculation of difference Fourier maps with the use of phases from all other derivatives. Density modification was performed on the MIR map with the program DM (31). The 3 Å, density-modified map was displayed with the program O (32) for tracing of the polypeptide chain. The model was refined against data to 2.0 Å with the program X-PLOR (33). Cycles of positional refinement, temperature factor refinement, and rebuilding were continued until the free Rfactor (34) fell below 35%, at which point ordered waters were added and a bulk solvent correction was applied. The quality of the structure was verified by Procheck (29), with no residues falling in disallowed regions of Ramachandran space, and by 1D-3D (35), which gave a score consistent with all side chains being located in acceptable environments.

View this table:

The Fr-RBD is a roughly L-shaped molecule consisting of a lopsided antiparallel β sandwich and a helical subdomain formed from two extended loops of the sandwich (Fig.2, A and B). Six strands (strands 1, 2, 4, 5, 8, and 9) form the large sheet of the Fr-RBD sandwich, whereas three strands (strands 3, 6, and 7) make up the second face (9). The large sheet is curved to form nearly half a barrel, and the small sheet packs against the half-barrel at an angle of about 40°. The first loop of the helical subdomain (between strands 3 and 4) contains two α helices (helices A and D). Also within this loop is an extended coil from Leu50 to Thr100 that passes almost 360° around the end of the helical lobe. The second loop (between strands 6 and 7) contains an α helix (helix G) that packs in an antiparallel orientation against helix D. Fr-RBD contains six disulfide bonds, four of which Linder and co-workers had previously identified biochemically and two of which they had predicted correctly by computer modeling (10). Although the overall tertiary structure of Fr-RBD is unique, a comparison of the Cα coordinates of Fr-RBD against all known folds (11) reveals modest similarity between the Fr-RBD β sandwich and proteins with immunoglobulin (Ig) folds (12) (Fig.3).

Figure 2

Structure of Fr-RBD. (Left) A ribbon representation is shown with disulfide bonds and carbohydrates indicated as balls and sticks. Strands are numbered and helices are lettered to correspond to Fig. 1. Variable regions are indicated (13). VRA includes the extended coil in the helical subdomain (residues 50 to 100) and helix D, and VRB corresponds to helix G. VRC is located in the loop between strands 4 and 5 and is separated from VRA and VRB in the Fr-RBD structure by conserved elements such as strands 4, 5, 8, and 9 (16). The figure was generated with Ribbons (36). (Right) A stereo diagram of Fr-RBD is numbered every 10 residues. The figure was generated with Molscript (37).

Figure 3

Topological relation between Fr-RBD and the Ig fold. Side views of the topologies of a v-type Ig fold and of Fr-RBD are shown with the β sandwiches splayed open between strands c" and d in the Ig fold and between strands 5 and 6 in Fr-RBD. Strand c" of the v-type Ig fold can be replaced by a loop (38), as is the case in Fr-RBD. Strand a of the v-type Ig fold can be associated with either sheet (38); only one alternative is shown. Strands 1 and 2 of Fr-RBD are shown in light gray to indicate that they are a modification of the v-type Ig-fold topology (38).

Amino acid sequence alignments of receptor-binding domains from C-type MLVs that use distinct receptors have highlighted regions that vary with tropism (Fig. 1). These variable regions, called VRA, VRB, and VRC (13), are mapped onto the Fr-RBD structure in Fig. 2A. The sequence throughout most of the β sandwich is conserved, whereas the sequence corresponding to the helical lobe is highly divergent. VRA, VRB, and, to a lesser extent, VRC vary in length as well as in sequence between MLVs with different tropisms. For example, VRA in amphotropic MLVs is considerably shorter than VRA of ecotropic MLVs (61 residues and one putative disulfide bond compared with 98 residues and three disulfides bonds), whereas amphotropic VRB is longer by 19 residues, including two additional cysteines. The pattern of sequence conservation in the receptor-binding domains of MLVs, when viewed in light of the Fr-RBD structure, indicates that the Ig-like core is used as a scaffold for displaying the receptor-binding regions of the envelope glycoprotein.

Chimeric envelope proteins have been constructed by others in attempts to localize the receptor-choice determinants in MLV SU glycoproteins. Although functional chimeras can be constructed readily between various nonecotropic viruses (14), it has not been possible to construct functional chimeras containing both ecotropic and nonecotropic variable regions (15). Upon examination of the Fr-RBD structure, this observation is not surprising. VRA and VRB are intimately associated with one another in the helical lobe, and a significant alteration of the length and sequence of either VRA or VRB alone would not easily be accommodated by the rest of the structure (16).

Despite the absence of useful chimeras, mutagenesis experiments have identified amino acids in the VRA region of the Moloney murine leukemia virus (Mo-MLV) SU that, when altered, inhibit MCAT-1 binding without affecting envelope processing and assembly into virions (17). Substitution of Lys for Asp84 in Mo-MLV SU (corresponding to Asp86 in the closely related Fr-MLV) blocks binding and infection. When residues Arg83, Glu86, and Glu87 in Mo-MLV (homologous to Arg85, Asp88, and Glu89 in Fr-MLV) are changed individually to Glu, Lys, and Lys, respectively, binding is abrogated, although virus titer is not significantly affected (18). The binding surface defined by the positions of these mutants is a textured landscape that includes a hydrophobic pocket and a charged ridge (Fig. 4).

Figure 4

The putative receptor-binding surface. A molecular surface is superimposed on the structure of the region containing likely receptor-contacting residues. Residues Arg85, Glu89, Pro90, and Thr92 form a ridge across the top of this region. The face just below the ridge contains the critical Asp86, which lies adjacent to a hydrophobic pocket. This pocket is lined with the residues Pro90, Leu94, and Trp102, as well as Asp86, and its base is formed by Leu91. The relation of the putative receptor-binding surface to the intact domain can be derived from the inset, in which the expanded region is boxed in red, and the helices are lettered as in Figs. 1 and 2A. The orientation is roughly a 180° rotation about they axis from the views in Fig. 2. The figure was generated with Grasp (39).

The Fr-RBD structure demonstrates that, in the SU glycoproteins of MLVs, the polypeptide chain is partitioned structurally into a conserved Ig-like antiparallel β-sheet framework and a variable subdomain. The Fr-RBD architecture will therefore facilitate the development of viruses with novel receptor specificities for the purpose of gene therapy. Current approaches have included tethering large binding domains such as single-chain antibodies to the NH2-terminus of the envelope protein (19) or substituting erythropoietin in place of SU sequences (20). The resultant recombinant viruses have very low transduction efficiencies and in some cases (20) require the wild-type glycoprotein to be co-incorporated into virions. With the use of the Fr-RBD structure as a model, modifications can now be directed to the receptor-binding lobe alone, with a detailed understanding of how the components of this subdomain interact with one another and with the core of the protein. The feasibility of replacing the variable subdomain sequences to generate new receptor specificities and tropisms has already been demonstrated by the success of viral evolution in generating the receptor diversity of this subgenus.


View Abstract

Navigate This Article