Suppression of a Fusion Defect by Second Site Mutations in the Ecotropic Murine Leukemia Virus Surface Protein

Abstract

Entry of ecotropic murine leukemia virus initiates when the envelope surface protein recognizes and binds to the virus receptor on host cells. The envelope transmembrane protein then mediates fusion of viral and host cell membranes and penetration into the cytoplasm. Using a genetic selection, we isolated an infectious retrovirus variant containing three changes in the surface protein—histidine 8 to arginine, glutamine 227 to arginine, and aspartate 243 to tyrosine. Single replacement of histidine 8 with arginine (H8R) resulted in almost complete loss of infectivity, even though the mutant envelope proteins were stable and efficiently incorporated into virions. Virions carrying H8R envelope were proficient at binding cells expressing receptor but failed to induce cell-cell fusion of XC cells, indicating that the histidine at position 8 plays an essential role in fusion during penetration of the host cell membrane. Thus, there is at least one domain in SU that is involved in fusion; the fusion functions do not reside exclusively in TM. In contrast, envelope with all three changes induced cell-cell fusion of XC cells and produced virions that were 10,000-fold more infectious than those containing only the H8R substitution, indicating that changes at positions 227 and 243 can suppress a fusion defect caused by loss of histidine 8 function. Moreover, the other two changes acted synergistically, indicating that both compensate for the loss of the same essential function of histidine 8. The ability of these changes to suppress this fusion defect might provide a means for overcoming postbinding defects found in targeted retroviral vectors for use in human gene therapy.

Newly assembled ecotropic murine leukemia virus (MLV) particles begin their life cycle by infecting naive host cells. The host cell plasma membrane presents a barrier to this invasion, and it is the function of the envelope proteins of these retroviruses to overcome that barrier. The envelope surface protein (SU) begins performing this task by binding the virus receptor, securely attaching the virion to the host cell. In addition to anchoring the virion, the interaction between SU and the receptor triggers changes in the structure of the envelope transmembrane protein (TM). These conformational changes activate the fusion potential of TM primarily by inducing the stretch of hydrophobic residues on its amino terminus to flip outward into the host cell membrane (6). Intrusion of this “fusion peptide” in the plasma membrane promotes the formation of a “fusion pore” through which the virus core passes to enter the host cell cytoplasm (6). Thus, virions lacking envelope proteins in their membranes cannot penetrate the host cell membrane and are noninfectious.

Ecotropic MLV envelope protein is the product of the viral env gene. Synthesized as a polyprotein, each molecule of envelope protein (Env) contains a number of glycosylation sites and a short signal peptide that is removed cotranslationally by the signal peptidase. Following glycosylation, the 85-kDa Env precursor oligomerizes, and a second internal cleavage processes it into mature SU and TM molecules. A disulfide bridge between SU and TM retains molecules of SU in the oligomers. Vesicles transport the resulting heterotrimers to the plasma membrane where they are assembled into budding virus particles.

The first 250 residues in SU of ecotropic MLV Env have been identified as a region that is essential for receptor binding. (Note that the nucleotide sequence is numbered according to the work of Shinnick et al. [29] for Moloney MLV [accession no. J02255]. Amino acid residues are numbered from 1 beginning with the alanine on the amino terminus of mature SU after cleavage of the signal peptide and beginning with the glutamate on the amino terminus of TM after cleavage of the precursor protein.) Within this domain, aspartate 84 and a pair of arginines at positions 83 and 95 appear to be critical for receptor recognition, as their mutation abolishes virus binding (3, 16). In Env of PVC-211, a neuropathogenic variant of Friend MLV (FrMLV), the amino acids corresponding to Moloney MLV (MoMLV) residues 115 and 127 are crucial for the expanded host range of that virus, although the changes found at these positions do not appear to affect receptor recognition (16, 19).

Traversing the plasma membrane 14 times, the receptor for the ecotropic MLV normally functions as the principal transporter of cationic amino acids in the cell. Sequences in the third extracellular loop of the receptor-transporter provide the attachment site recognized by SU (1). The sequence of this loop varies widely between species. The versions found in mouse and rat cells function as ecotropic retrovirus binding sites, explaining why only rodent cells are susceptible (2). Replacement of two critical residues in this binding site produced a binding-defective receptor that failed to provide an effective entry for infectious MLV. Moreover, replacement of the corresponding two positions in the homologous human cationic amino acid transporter with the residues found in the mouse protein resulted in acquisition of receptor function (1).

We are seeking to identify key functional residues on Env, particularly those that interact with the critical residues identified in the virus receptor. During virus replication, the viral polymerase reverse transcriptase (RT) frequently introduces nucleotide changes into the genome that produce variants known as quasispecies. The very high particle-to-infectious-unit ratio typically found in ecotropic retrovirus stocks (24, 25) might be due in part to generation of mutant Env. Analysis of these naturally occurring env sequences might lead to identification of important residues in Env. Moreover, it might be possible to influence the env sequences present in quasispecies by passaging the virus on cells expressing a binding site mutant of the receptor. Virus that acquired an envelope mutation compensating for the changes present on the receptor would have a selective advantage. Identification of the positions altered in such an adapted envelope protein would reveal important functional residues in Env.

We passaged replication-competent MLV on cells expressing binding-defective receptors and then isolated the env genes of representative quasispecies present after this selection, determined their DNA sequence, and deduced the encoded Env sequence. Analysis of the role of residues altered in the envelope proteins from one of the quasispecies revealed an essential role for the histidine at position 8 in virus-cell fusion. Thus, there is at least one domain in SU that is involved in fusion; the fusion functions do not reside exclusively in TM as was previously thought. Replacement of two residues distant from histidine 8 on the peptide chain (glutamine 227 and aspartate 243) suppressed the defect caused by substitution at position 8, suggesting that the three residues might be near each other in the folded protein. Furthermore, the suppression was synergistic, indicating that changes in glutamine 227 and aspartate 243 compensate for the loss of the same essential function of that residue. These results also suggest that in the proper context, one or more of these SU residues might participate in virus penetration of the host cell membrane. Their ability to restore functional loss of the critical histidine residue makes the substitutions at positions 227 and 243 candidates for suppressing postbinding defects found in targeted retroviral vectors for gene therapy.

MATERIALS AND METHODS

Cell lines and viruses.

Mouse NIH 3T3 fibroblasts and nonpermissive human 293 fetal kidney cells (gift of M. Quinlan) were cultured in Dulbecco’s modified Eagle’s medium (DMEM). The 293-derived stable transfectants expressing the binding-defective receptor from cDNAs M-38 and H-63, described elsewhere (17), were maintained in DMEM containing 200 μg of G418 (GIBCO) per ml. XC rat sarcoma cells (ATCC CCL-165) were cultured in DMEM supplemented with 3.5 mg of d-glucose per ml. Recombinant, replication-competent ecotropic DHFR-5HE* virus produced by cells expressing the pMLV DHFR*-5 infectious clone was obtained from H. Stuhlmann. The genome of this virus consists of insertion of a small expression cassette for the mutant dihydrofolate reductase gene (dhfr) that confers resistance to the drug methotrexate into the 3′ long terminal repeat of ecotropic MoMLV (30). Naive NIH 3T3 cells were exposed to DHFR-5HE* virus stock and selected for growth in 150 μM methotrexate (Sigma), and then methotrexate-resistant colonies were pooled to generate producer cell lines. All experiments reported here were performed with virus from the DHFR-5HE*1 producer line.

Plasmids.

pcDNA MoMLV was constructed to provide gag, pol, and wild-type or mutant env genes for virus particle assembly as follows. The 8-kbp BssHII-ClaI fragment from plasmid pPEM-5 (gift of V. Garcia [20]) encoding gag, pol, and the 5′ end of env genes of a MoMLV proviral genome lacking the packaging signal was ligated to the ClaI-BamHI fragment encoding the 3′ end of MoMLV env of pMov3 (gift of H. Stuhlmann [11]) into which a BamHI site was engineered immediately after the env gene termination codon. This genome was then inserted into the HindIII and BamHI sites of the eukaryotic expression vector pcDNA3 (Invitrogen), placing the viral structural and enzymatic genes under the control of the cytomegalovirus promoter. The pBAG plasmid encoding a packageable MoMLV genome lacking gag, pol, or env sequences but containing the Escherichia coli lacZ gene under the control of the retroviral 5′ long terminal repeat and the neomycin resistance gene under the simian virus 40 promoter was the gift of C. Cepko (31). Plasmid encoding wild-type or mutant envelope proteins, pcENV-Mo, was constructed by inserting the NdeI-EcoRI restriction fragment of pcDNA MoMLV containing nucleotide 5403 through the envelope protein stop codon into expression vector pcDNA3.

Sequence analysis of 5HE virus env gene and isolation of env genes from virus quasispecies.

Genomic DNA from the DHFR-5HE*1 virus producer cell line was isolated as previously described (2) and used as a template for PCR. Two overlapping fragments of the env gene were amplified with Pfu thermal polymerase (Stratagene), gel purified, and submitted to sequence analysis with the Exo(−) Pfu Cyclist kit (Stratagene). For quasispecies env genes, genomic DNA was isolated from virus-producing cells precisely as previously described (2) and used as a template for PCR to amplify env genes present after virus passaging. Oligonucleotides used for PCR were 5′-CAAAGTAGACGGCATCGCAGCTTGG-3′ and 5′-GGCGAATTCATCTATGGCTCGTACTCT-3′. Products were subcloned into the pcDNA MoMLV plasmid and used to produce virus particles as described below. Selected plasmids were submitted to DNA sequence analysis.

Virus production.

Clonal populations of human 293 cells stably expressing the pBAG plasmid were analyzed for β-galactosidase expression by staining with chromogenic substrate (X-Gal [5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside]), and the five clones developing the most intense staining in the shortest period of time, indicative of high levels of transcription of the BAG virus genome, were selected as virus producers. Expression of packageable viral RNA was confirmed by Northern analysis (data not shown). One of these cell clones, H1-BAG, was used in all experiments reported here. To produce lacZ-transducing virus particles, we transiently transfected the H1-BAG cells with pcDNA MoMLV DNA containing wild-type, cloned, or mutated env genes. Transfection was performed by calcium phosphate precipitation as described by Sambrook et al. (28). Virus preparations were freed of producer cells by low-speed centrifugation followed by filtration through a 0.45-μm-pore-size filter. An aliquot of 3 ml was stored at −80°C and later used for virus titration. Virus from the remaining 7 ml was immediately pelleted as described below for immunoblotting. For each virus binding experiment, the entire virus preparations were frozen and concentrated as described below.

Site-directed mutagenesis of env genes.

Nucleotide substitutions in the env gene were generated by the method described by Kunkel (14). For this purpose, we subcloned the 1,300-bp, HpaI-HpaI restriction fragment from the MoMLV env gene on plasmid pMOV3 (11) into vector M13mp18 containing an engineered HpaI site. Site-directed mutagenesis was performed, and the HpaI-HpaI restriction fragment was transferred back to pcDNA MoMLV. The entire 1,300-bp fragment in the resulting plasmid was sequenced with the fmol sequencing kit (Promega) to ensure the absence of unscheduled substitutions and to confirm the presence of desired mutations.

RT assay and virus titers.

RT assays were performed exactly as described by S. Goff and coworkers (10). Cells exposed to stocks of DHFR-5HE* during virus passaging were grown in medium containing 150 μM methotrexate (Sigma) for 2 weeks to select for infected cells. Endpoint dilution titration of all virus stocks was performed as previously described for ecotropic MLV (17).

Western blot analysis.

Virus particles were pelleted from 7 ml of cell-free virus supernatant though a cushion of 25% sucrose in TNE (10 mM Tris [pH 8.0], 1 mM EDTA, 100 mM NaCl) in a Beckman SW-41 rotor (30,000 rpm, 2 h, 4°C). Pellets were taken up in 40 μl of phosphate-buffered saline (PBS) and stored at −80°C. Virus producer cells were lysed immediately after virus harvests in RIPA buffer (20 mM Tris [pH 7.0], 1% Triton X-100, 0.05% sodium dodecyl sulfate [SDS], 0.5% sodium deoxycholate, 150 mM NaCl, 2.5 mM phenylmethylsulfonyl fluoride) by incubation for 30 min on ice. Cell lysates were centrifuged for 10 min at 10,000 rpm in an Eppendorf model 5415C centrifuge to pellet nuclei, and supernatants were frozen at −80°C. Total protein concentration in cell lysates was determined by the Bradford assay (Bio-Rad). Ten microliters of virus pellets or 100 μg of total protein from cell lysates was diluted 1:1 in 2× gel loading buffer (28), boiled for 10 min, chilled on ice, and then subjected to SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes (Protran [Schleicher & Schuell]). Envelope proteins (SU and precursor) were detected with goat anti-Rauscher-gp70 (1:100), structural capsid protein (CA) was detected with goat anti-Rauscher-p30 (1:10,000; Quality Biotech, Inc.), and envelope TM protein was detected with rabbit anti-p15E antiserum (1:1,000; gift of Alan Rein). Subsequent incubation with mouse anti-goat or mouse anti-rabbit antibody conjugated to horseradish peroxidase (1:10,000; Sigma) was performed, and immunoblots were developed by detection of horseradish peroxidase with SuperSignal (Pierce).

Virus binding assays.

Binding assays were performed essentially as described previously (7, 13) with the following modifications. Virus was concentrated 10- to 15-fold on Centriplus-100 concentrators (Millipore) to provide binding under saturating conditions (34a) and to eliminate most of the monomer SU should any be shed from virions or the producer cell surface. To ensure that cells were incubated with equal numbers of particles from each of the virus stocks, the volumes of the concentrated virus were adjusted to achieve comparable particle concentrations based on the RT activity and Western blot quantitation of capsid protein. 293 cells expressing wild-type virus receptor or parental 293 cells were detached from culture plates with PBS containing 0.02% EDTA. Cells (10⁶) were then incubated with 1 ml of adjusted concentrated virus stocks containing equal amounts of virions in the presence of Polybrene (5 μg/ml) for 1 h at 4°C. Cells were then washed with PBA (PBS, 2% fetal bovine serum, and 0.02% sodium azide) and incubated in 500 μl of PBA containing goat anti-gp70 antiserum (1:100) for 30 min at 4°C. After two washes, cells were incubated with 500 μl of secondary antibody, donkey anti-goat antibody conjugated with fluorescein isothiocyanate (Jackson Laboratories) diluted in PBA (1:200) for 30 min at 4°C. Propidium iodide (Sigma) was added to the binding reaction mixture for 5 min at a final concentration of 20 μg/ml. Cells were washed twice and taken up in 500 μl of PBA, and the fluorescence of 5,000 live cells (negative for propidium iodide) was analyzed by flow cytometry (Epics Profile Analyzer; Coulter Cytometry). Experiments were repeated three times.

Cell-cell fusion assays.

The fusion-from-without assay was performed as described by Rein and Bassin (24) with modification as follows. XC cells were exposed to virus stock containing 20 μg of Polybrene per ml or to medium containing 20 μg of Polybrene per ml and incubated with continuous exposure at 37°C for 16 h; then they were fixed and stained with 0.5% methylene blue–0.16% basic fuchsin in methanol (26). The fusion-from-within assay was performed as described by Jones and Risser (12) with the following modifications. Briefly, 60-mm-diameter dishes of 70% confluent 293 cells were transfected with 20 μg of wild-type or mutant pcENV by calcium phosphate precipitation, at 24 h posttransfection XC cells (10⁶ per dish) were added, and cultures were cocultivated for an additional 24 h. Then cells were stained as described above and observed by light microscopy.

RESULTS

We attempted to utilize a genetic selection to generate viruses containing envelope variants whose analysis would lead to the identification of critical residues in the virus envelope proteins, particularly those involved in virus-receptor interactions. For this selection, we passaged a replication-competent ecotropic virus on human cells expressing a binding-defective receptor as the only ecotropic retrovirus receptor present. We chose DHFR-5HE* as the replication-competent ecotropic virus constructed from proviral clones of MoMLV for use in these studies. Our choice was based primarily on the design of the genome of this recombinant virus. It encodes all the necessary functions for efficient retroviral replication, as well as a mutant dihydrofolate reductase gene conferring resistance to at least a 30 μM concentration of the drug methotrexate (30) (Fig. 1A). Since infection by laboratory strains of ecotropic retroviruses does not result in any detectable morphological changes in most cultured mammalian cells, the presence of the selectable marker gene was particularly useful because cells replicating virus could be selected by growth in medium containing methotrexate.

FIG. 1.

(A) Diagram of the proviral genome of the DHFR-5HE* virus used in the genetic selection on cells expressing defective ecotropic MLV receptors. In addition to viral gag, pol, and env genes, the genome encodes a mutant dihydrofolate reductase (dhfr*) gene, initially introduced into the unique 3′ (U3) region and duplicated into the U3 of the 5′ long terminal repeat (LTR) during virus replication (30). (B) Defective receptors M-38 and H-63 mediate DHFR-5HE* virus entry as inefficiently as they mediate ecotropic MoMLV entry. Titers were determined by endpoint dilution on human 293 cells stably expressing the receptor cDNA indicated (n = 4). Values are the averages from three independent experiments.

The two defective mutants of the ecotropic receptor used for these studies were M-38 and H-63. M-38 contains a tyrosine-to-proline alteration at residue 235 (Y235P) and an E237V change in the third extracellular domain of the ecotropic receptor, while H-63 contains a V244E change in the third extracellular domain of the human homolog of the ecotropic receptor. They were chosen because each is capable of participating in virus entry but with much less efficiency than that of the wild-type receptor (1), giving a possible advantage of several orders of magnitude for an adapted virus. We determined the titers of our DHFR-5HE*1 (5HE) virus stock on human 293 cells expressing either the wild-type receptor or one of the defective receptors. Cells expressing the defective receptors were at least 10,000-fold less susceptible to 5HE virus than were cells expressing wild-type receptor (Fig. 1B), the same pattern of susceptibility that we had previously found for replication-defective viruses (17).

The parental 5HE stock did not contain variants adapted to defective receptors M-38 and H-63.

Two possibilities could account for the low level of entry routinely observed in the cells expressing M-38 and H-63. First, infection might represent the inefficient use of the defective receptor by wild-type virus. Second, it might involve infection by a minor variant present in the parental 5HE stock that was capable of using the defective receptor efficiently. To determine if the latter was the case, we transfected the M-38 or the H-63 cDNAs into naive human 293 cells and exposed each transfected population to 5HE virus 3 days later, at which time transient expression of the receptors would be waning. The exposed cells were then grown in medium containing methotrexate but not G418 to select for infected cells in the absence of selection for receptor expression, preventing virus spread so that each colony would represent only the initial infection event. The 9 methotrexate-resistant colonies and the 15 colonies isolated from exposed M-38 cells and H-63 cells, respectively, were propagated under methotrexate selection, and virus-containing supernatants were harvested from each culture. A comparison of endpoint dilution titers revealed that none of the viruses produced by the methotrexate-resistant colonies used the defective receptors better than did parental 5HE virus stock (data not shown), indicating that the low level of infection by 5HE stock likely represents inefficient use of the defective receptors by wild-type virus.

Genetic selection of quasispecies.

We exposed 5 × 10⁵ cells of human cell lines stably expressing either M-38 or H-63 receptors to the 5HE virus stock. A typical exposure led to the growth of 10 to 20 methotrexate-resistant colonies. Colonies were pooled and propagated until confluent monolayers were obtained, at which time their virus-containing supernatant was harvested (first-passage virus stock) and used to expose fresh cultures of the same cell line. The virus supernatant exhibited a decline in RT activity after the first passage on both cell lines. RT activity then gradually increased to 11- and 7-fold greater than that in the parental 5HE stock after the third passage. Although the RT activity suggested abundant virus particle production, the titer of the third-passage stocks on cells expressing the wild-type receptor was low compared to that of the parental 5HE stock—2,000 and 200 CFU/ml, respectively, for M-38 and H-63 selected stocks compared to >10⁶ for 5HE stock. Further passaging on M-38 and H-63 cells did not improve the apparent number of infectious particles in the virus stocks. Continued cultivation of the third-passage producer cells in methotrexate-containing medium resulted in a severalfold increase in RT activity and increased infectious particle production to greater than 10⁴ CFU/ml. Since cells infected with 5HE virus acquire resistance to superinfection slowly (30), one explanation for these observations is that virus production increased upon continued cultivation of cell populations because cell-to-cell virus spread increased the number of integrated virus genomes. From each of these two populations of producer cells, 12 representative env genes of the quasispecies present after the selection were isolated, subcloned into the pcDNA MoMLV plasmid, and then transfected into H1-BAG cells to produce particles that transduce lacZ and are pseudotyped by the cloned env genes.

Characterization of representative env sequences from quasispecies present after selection.

Virus stocks generated with six of the env genes from selection on cells bearing M-38 receptors and four env genes from selection on H-63 receptors were as infectious as wild-type MoMLV, suggesting that these Env proteins either were unaltered or contained innocuous changes (data not shown). The remaining 14 stocks were at least 100-fold less infectious than MoMLV. The decrease in infectivity might be the result of poor transfection efficiency during virus production, poor incorporation of Env into virions, or changes in Env residues critical to interaction with the receptor. To distinguish between these possibilities, virus from equal volumes of transfected H1-BAG cell supernatant was pelleted through a sucrose cushion and Western blot analysis was performed with anti-SU and anticapsid antiserum. For comparison, we also analyzed supernatant from cells transfected with the pcDNA MoMLV plasmid containing wild-type env and with plasmid containing env clone 832 that produced highly infectious particles. Aliquots of each supernatant were removed prior to centrifugation and used for endpoint dilution titration on naive NIH 3T3 cells and human cells expressing the wild-type receptor (Fig. 2A).

FIG. 2.

The reduction in infectivity of virus variants is predominantly due to lack of envelope protein incorporation into virions. (A) Naive NIH 3T3 cells (stippled bars) or 293 cells stably expressing the wild-type ecotropic MLV receptor (black bars) were exposed to 10-fold serial dilutions of virus stocks containing virions pseudotyped with envelope proteins encoded on the 14 cloned env genes that showed reduced infectivity. Titers were calculated from the endpoint dilution (n = 4). Values are the averages from four independent experiments. Titers of viruses with Env from clones 871, 875, and 877 were less than 1 on NIH 3T3 cells. MoMLV, wild-type ecotropic MoMLV; mock, mock-transfected cells. (B) Western blot analysis of virions containing cloned env genes. Viral proteins were separated by SDS-polyacrylamide gel electrophoresis on 8% polyacrylamide gels. Membranes were cut in two parts at the position indicated by the black line, roughly that of the 45-kDa molecular mass standard. The top portion was probed with anti-SU antisera, and the bottom was probed with anticapsid (CA) antisera. The immunoblot shown is representative of four made from independent virus preparations. (C) Western blot analysis of virus producer cell lysates. Proteins were separated on 8% polyacrylamide gels, and membranes were probed with anti-SU antisera.

Upon short exposure of immunoblots, the intensities of the bands corresponding to capsid protein were within twofold of each other in all pellets, except those from mock-transfected cells (data not shown), indicating that transfection efficiencies were comparable. The majority of cloned env sequences encoded proteins that were very poorly incorporated into virions (Fig. 2B). Steady-state levels of Env in lysates of the producer cells transfected with clones 837, 857, 871, 879, 891, and 895 were too low to be detected (Fig. 2C), indicating that low infectivity resulted from poor incorporation of unstable envelope proteins. Env from clones 829, 865, and 877 showed evidence of either truncated forms or degradation products of SU in cell lysates. Notably, Env from clones 855 and 875 showed evidence of less-than-normal cleavage of precursor protein into mature SU (Fig. 2C), and virions contained a slightly larger species that was the size of envelope precursor protein. All virus stocks, including wild-type MoMLV, consistently exhibited a 4- to 20-fold-greater titer on 293 cells expressing receptor than on NIH 3T3 cells, suggesting that the difference in infection results from a property of the host cell line rather than of the virus. For instance, the 293 receptor cell line has a greater number of virus binding sites than do NIH 3T3 cells (1a), a feature that might account for this difference.

We focused further analyses on clones 838 and 839 encoding Env proteins that were processed normally and efficiently incorporated into virions but showed a 100-fold reduction in infectivity (Fig. 2). In addition, we continued analysis of clone 881 because its steady-state levels of expression and processing appeared normal but its incorporation into virions was very poor (Fig. 2). The DNA sequence of these three clones was determined, and the sequence of the encoded protein was deduced from it. The sequence of the envelope gene from DHFR-5HE* was determined for comparison. It was identical to the published sequence of MoMLV (29) from which it was derived. env clone 881 encodes a single substitution of proline for serine (T→C at nucleotide 7361) at position 60 in the putative leucine zipper domain of TM.

Replacement of glutamine 227 and aspartate 243 can suppress a defect in infection caused by replacement of histidine 8 of SU.

Clones 838 and 839 were identical in sequence; they encoded the same three alterations of histidine 8 to arginine (H8R; A→G at nucleotide 5898), glutamine 227 to arginine (Q227R; A→G at nucleotide 6555), and aspartate 243 to tyrosine (D243Y; G→T at nucleotide 6602). We placed these changes in the wild-type env sequences as single mutations or as all combinations of substitutions to determine which is responsible for the decrease in infectivity. Figure 3 shows the results of the characterization of the encoded proteins. A single H8R substitution almost completely abolished infection, while single and double mutations at the other two positions did not affect infection (Fig. 3A). These results indicate that the change in histidine 8 is responsible for the reduced infection found with clones 838 and 839. Interestingly, the triple mutation corresponding to the original 838 and 839 clones was only 100-fold less infectious than wild-type virus, suggesting that one of the other two mutations can suppress the defect introduced by the change of histidine 8. However, neither a Q227R nor a D243Y mutation in combination with an H8R mutation improved infectivity appreciably, indicating that both changes are required and that they exert a synergistic effect on the H8R mutation.

FIG. 3.

A histidine-8-to-arginine mutation introduces a defect in virus entry that is overcome by the synergistic action of amino acid substitutions at positions 227 and 243. (A) Infectious titers on NIH 3T3 cells (stippled bars) and 293 cells expressing exogenous wild-type receptor (black bars). Titers were calculated from the endpoint dilution (n = 4) after exposure to virions pseudotyped with envelope proteins containing the indicated substitutions. Each value is the average from five independent experiments. The titer of H8R virus on NIH 3T3 cells was less than 1 in all experiments. (B) Western blot analysis of virions containing mutant envelope proteins. Proteins were separated on a 12% polyacrylamide gel. The membrane was cut into three parts at the positions indicated by the black lines, roughly that of the 45- and 25-kDa molecular mass standards. The top portion was incubated with anti-SU antisera, the middle portion was incubated with anti-CA antisera, and the bottom part was incubated with anti-TM antisera. The immunoblot shown is representative of five made from independent virus preparations. (C) Western blot analysis of virus producer cell lysates. Proteins were separated on 8% polyacrylamide gels and then blotted to anti-SU antisera. Numbers at left of panels B and C show molecular mass in kilodaltons.

The defect in virus entry is at the level of fusion and can be suppressed by the changes in residues 227 and 243.

All of the mutant envelope proteins were efficiently incorporated into virions (Fig. 3B) and processed normally (Fig. 3C), indicating that the reduction in infectivity of H8R viruses was not due to lack of expression on producer cells or to failure to incorporate into virions. Therefore, it was likely that mutation of histidine 8 abolished a function of the envelope involved in virus entry. We performed binding and fusion assays to determine if virus binding or virus penetration of the host cell membrane was affected.

Viruses pseudotyped in H8R Env bound human cells expressing receptor as proficiently as did wild-type MoMLV (Fig. 4, left panels), indicating that histidine 8 is not involved in virus attachment and suggesting that it is involved in a downstream entry step. Following virus binding, the envelope protein induces fusion of viral and cellular membranes to release the nucleocapsid into the cytoplasm. Because direct measurement of the fusogenic potential of an envelope protein during virus-cell fusion is difficult, indirect cell-cell fusion assays using XC cells have been widely used (3, 12, 23). XC cells express the ecotropic retrovirus receptor on their surface and are susceptible to MLV infection. In addition, they exhibit an extraordinary propensity to undergo cell-cell fusion when exposed to wild-type MLV (fusion from without) or when cocultivated with cells expressing virus envelope (fusion from within) (12, 24). This ability to induce syncytium on XC cells has been utilized to evaluate the fusion properties of the envelope proteins of ecotropic viruses (3, 12, 23, 33). We performed fusion assays with XC cells and found that viruses pseudotyped with H8R Env failed to induce XC cell fusion in the same assay in which wild-type virus induced extensive fusion (Fig. 4, middle panels). Cells expressing this Env mutant also failed to participate in fusion with XC cells (Fig. 4, right panels). In contrast, the presence of the Q227R and D243Y mutations suppressed the H8R mutation, partially restoring syncytium-inducing capability in both the fusion-from-without and the fusion-from-within assays. These results suggest that histidine 8 of SU plays an essential role in virus-cell fusion and imply that its role can be replaced by substitution of glutamine 227 and aspartate 243. It is noteworthy that in the absence of any change at residue 8, the ability of Q227R D243Y Env to induce cell-cell fusion was compromised. XC cells exposed to Q227R D243Y virions and XC cells cocultivated with Q227R D243Y Env-expressing cells exhibited numerous small syncytia in contrast to the extensive syncytia induced by wild-type protein, even though the Q227R D243Y particles were as infectious as MoMLV. This reduction in the size of syncytium might explain why the suppressing mutations do not completely restore fusion function and infectivity to an H8R mutant.

FIG. 4.

Replacement of histidine 8 with arginine results in a defect in envelope fusion function that is suppressed by the glutamine-227-to-arginine and aspartate-243-to-tyrosine mutations. (Left panels) Virus binding. Dashed line, binding of wild-type virus to 293 cells lacking receptor; dotted line, nonspecific binding of antisera in the absence of virus to 293 cells stably expressing receptor. Shaded area represents binding of wild-type or mutant viruses to the 293 cells expressing receptor. (Middle panels) Fusion from without. XC rat sarcoma cells were exposed to virions pseudotyped with the envelope protein indicated. Induced syncytia were stained with methylene blue and basic fuchsin. (Right panels) Fusion from within. XC cells were cocultivated with 293 cells expressing the indicated Env. Mock, mock-infected XC cells; pcDNA3, XC cells cocultivated with 293 cells expressing vector pcDNA3. Arrows in H8R Q227R D243Y panels point to syncytium. H8R, histidine 8 to arginine; Q227R, glutamine 227 to arginine; D243Y, aspartate 243 to tyrosine.

DISCUSSION

Using a genetic selection similar to that used by other groups to obtain variants adapted to mutant receptors, soluble receptors, or inhibitory drugs (18, 21, 22), we did not obtain a virus variant that utilized the defective receptors more efficiently than did wild-type virus (34a). It may not be possible to complement the precise changes present in these defective receptors either because no amino acid in SU can reestablish the required interaction or because any change that can reestablish that interaction is deleterious to envelope folding, processing, or stability. Alternately, the desired complementing changes might be somewhat deleterious, allowing only a low level of expression of Env on the cell surface that leads to poor incorporation into virions. Such an outcome might well counteract any benefit obtained from a stronger binding to the defective receptor and interfere with enrichment for viruses bearing that envelope variant. We did not analyze enough envelope sequences to draw conclusions about either possibility.

Although we did not isolate a variant adapted to any of the mutant receptors, we did isolate a number of mutant env genes encoding proteins with interesting properties. Analysis of these mutants provided new insights into regions of Env that are essential for its assembly and productive virus-receptor interactions. It is unclear what role the defective receptors played in selection of these particular env genes, as it might also be possible to isolate the same variants by virus passaging on cells expressing a functional receptor. Envelope clone 881 contained a point mutation resulting in a serine-to-proline mutation at position 60 of TM. Residues 46 through 78 of TM form an alpha-helical structure that is the putative coiled-coil oligomerization domain for the envelope trimer (9). In the Saccharomyces cerevisiae two-hybrid system, this segment acts as a dimerization domain (15). When prolines were placed in the analogous domain of human immunodeficiency virus TM, the contact site between SU and TM was disrupted and virions did not incorporate envelope proteins (4). However, when amino acid substitutions other than proline were placed in this domain of MoMLV or human immunodeficiency virus, Env incorporation and virus binding were normal but virions were fusion defective, suggesting that this is a domain involved in fusion and not an oligomerization domain (5, 23, 34). Our data show that a serine 60-to-proline change in this domain markedly decreases the amount of Env on virions, supporting the former results. The strong helix-breaking properties of proline might induce changes in structure that interfere with or weaken envelope trimerization and/or association of SU and TM, resulting in poor assembly of the encoded protein into virions.

SU contains a domain essential to the fusion function of envelope protein.

Changes in glutamine 227 and aspartate 243 acted synergistically on the mutation at histidine 8, suggesting that they suppress the H8R mutation by compensating for the same essential function of that residue. It is striking that residues so distant on the peptide chain of SU can influence each other’s function, suggesting that they might be adjacent in the folded protein. Moreover, it was previously thought that the functional domains involved in virus binding reside in SU while virus-cell fusion domains reside in TM. However, recent deletion analysis of SU revealed that removing residues 2 through 8 almost completely abolished infectivity (3). Replacement of histidine 8 with alanine, glutamate, or lysine, as well as deletion of that position, dramatically reduced infection and completely abolished Env-mediated XC cell fusion but did not affect virus binding (3). These results suggested that the aromatic side chain of histidine 8 plays a critical role in Env-mediated fusion. Bae and coworkers (3) suggested that histidine 8 is a critical residue in an amino-terminal fusion domain consisting of the first eight residues of SU. Our independent identification of the role of histidine 8 in virus-cell fusion confirms the importance of that residue in fusion. However, it is not clear that residues 2 to 7 are also members of the histidine 8 fusion domain. The 20-fold reduction in fusion found by Bae and colleagues upon deletion of amino acids 2 to 7 might be due to structural changes in the proximity of histidine 8 that occur when the adjacent residues are missing, not to loss of function of these residues. Instead of belonging to a contiguous domain, histidine 8 might be an essential member of a noncontiguous, three-dimensional fusion domain. Some insights into the composition of such a domain may be found in the phenotype of the Q227R D243Y Env mutant. Q227R D243Y Env induced markedly reduced levels of fusion from within and fusion from without, indicating that these two residues can profoundly influence Env fusion properties possibly by producing subtle conformational changes in adjacent residues that are constituents of the histidine 8 fusion domain.

A model for the mechanism of suppression of the fusion defect.

The recent solution of the crystal structure of an amino-terminal fragment of ecotropic FrMLV (8), an ecotropic MLV closely related to MoMLV, provides insights into our findings. The three residues are conserved between the Moloney and Friend viruses: a histidine occupies the eighth position in FrMLV SU, while glutamine 229 of FrMLV corresponds to MoMLV 227, and aspartate 245 of FrMLV corresponds to MoMLV 243. Since the first eight residues were not in the structure of FrMLV SU, presumably because their conformation is variable, and the crystallized fragment ended at residue 236, we must speculate as to the positions of two of the three critical residues. Figure 5 shows a speculative model of the mechanism of suppression of the fusion defect. The space where the imidazole ring of histidine 8 would be expected to lie adjacent to glutamine 9 is directly next to the charged side group of the residue corresponding to arginine 232 (Fig. 5, left panel). In the mutant, repulsion between the guanadino groups of arginines 8 and 232 might displace one of the two side chains, presumably that of position 8 because this region of the peptide chain appears to be flexible, leaving an opening into which another aromatic side chain might slip (Fig. 5, right panel). The residue corresponding to glutamine 227 lies at the carboxy-terminal end of a β-strand consisting of alternating hydrophobic residues and polar or charged amino acids (residues 218 to 227 [Fig. 5]). This β-strand changes from LTFGIRLRYQ to LTFGIRLRYR in the mutant, adding another repeat of hydrophobic (italics) and charged or polar (underlined) residues. If this change extended the β-strand by one residue, then the peptide chain might twist slightly, repositioning the downstream residues including residue 243. Thus, placing arginines at positions 8 and 227 might cause structural changes that bring the aromatic ring of the tyrosine at position 243 close enough to the position normally occupied by the aromatic ring of histidine 8 to provide the required contribution to fusion. This model is consistent with our results in that it predicts the synergistic action of the suppressing mutations. It also predicts that the double H8R Q227R and H8R D243Y mutants should be almost as defective as the H8R mutant and that fusion might be compromised in a Q227R D243Y mutant even though there is a histidine in position 8.

FIG. 5.

A speculative model of the mechanism of suppression of the fusion defect. (Left panel) The diagram of the structure of residues 9 to 234 (bright colors) of the wild-type MoMLV SU was generated with the crystal structure coordinates of the closely related FrMLV surface protein (8) in the RasMol visualization program, based on the assumption that the structure is conserved between MoMLV and FrMLV SU because they have 89% amino acid identity and 97% conservation in this region. Residues 8, 12, 84, 227, 232, and 243 are represented as space-filled amino acids; the remaining residues are represented in the ribbon model. The structure of residues 1 to 8 and 235 to 245 (muted colors) is hypothetical. The amino-terminal eight residues were presumed to form a random coil. Residues 235 to 239 (IGPNP) were modeled as a turn and residues 240 to 245 (VLADQQ) were modeled as a β-strand based on the structures predicted by the Chou-Fasman and Garnier-Robson algorithms, respectively, of the Protean program of DNASTAR sequence analysis. (Right panel) Diagram of the hypothetical structure of the H8R Q227R D243Y mutant. Arrows indicate the direction of side chain repositioning resulting from placing arginines at positions 8 and 227. The perspective is from the side opposite from that in the left panel to show putative side chain interactions between mutated residues 8 and 243.

MoMLV particles are readily pseudotyped with virtually all other retroviral envelope proteins so that they utilize the receptor specified by the heterologous SU. This promiscuity with respect to envelope incorporation has been exploited to develop retroviral vectors for use in targeted human gene therapy. However, the modified viruses developed for this use are very poorly infectious or completely noninfectious and thus not practical for use in gene delivery. Others have suggested that a major source of their poor infectivity is a defect in the ability of these modified envelope proteins to perform postbinding events, particularly fusion (7, 27, 32, 35). It is noteworthy that in these targeted envelope proteins, the sequences to direct binding of the virus to a new receptor were either inserted within the first 20 residues of SU or used to replace the amino terminus. The latter design deletes the essential histidine side chain while the insertions likely change its position, producing essentially the same defect as that caused by the H8R mutation. Thus, it might be possible to dramatically increase the infectivity of these targeted retroviral vectors by constructing them with an envelope with the suppressing arginine and tyrosine substitutions at positions 227 and 243, respectively.