In Vivo Analysis of Infectivity, Fusogenicity, and Incorporation of a Mutagenic Viral Glycoprotein Library Reveals Determinants for Virus Incorporation

ABSTRACT

Enveloped viruses utilize transmembrane surface glycoproteins to gain entry into target cells. Glycoproteins from diverse viral families can be incorporated into nonnative viral particles in a process termed pseudotyping; however, the molecular mechanisms governing acquisition of these glycoproteins are poorly understood. For murine leukemia virus envelope (MLV Env) glycoprotein, incorporation into foreign viral particles has been shown to be an active process, but it does not appear to be caused by direct interactions among viral proteins. In this study, we coupled in vivo selection systems with Illumina next-generation sequencing (NGS) to test hundreds of thousands of MLV Env mutants for the ability to be enriched in viral particles and to perform other glycoprotein functions. NGS analyses on a subset of these mutants predicted that the residues important for incorporation are in the membrane-proximal external region (MPER), particularly W127 and W137, and the residues in the membrane-spanning domain (MSD) and also immediately flanking it (T140 to L163). These predictions were validated by directly measuring the impact of mutations in these regions on fusogenicity, infectivity, and incorporation. We suggest that these two regions dictate pseudotyping through interactions with specific lipid environments formed during viral assembly.

IMPORTANCE Researchers from numerous fields routinely exploit the ability to manipulate viral tropism by swapping viral surface proteins. However, this process, termed pseudotyping, is poorly understood at the molecular level. For murine leukemia virus envelope (MLV Env) glycoprotein, incorporation into foreign viral particles is an active process, but it does not appear to occur through direct viral protein-protein interactions. In this study, we tested hundreds of thousands of MLV Env mutants for the ability to be enriched in viral particles as well as perform other glycoprotein functions. Our analyses on a subset of these mutants predict that the glycoprotein regions embedded in and immediately flanking the viral membrane dictate active incorporation into viral particles. We suggest that pseudotyping occurs through specific lipid-protein interactions at the viral assembly site.

INTRODUCTION

Formation of infectious retroviral particles requires a symphony of viral and host cell proteins to coordinate in a spatial and temporal manner. Typically, this is an exclusive process wherein components from a given virus assemble only with components from the same or closely related viruses, such as HIV-1 and HIV-2 (1–5). However, this exclusivity is not typical for the incorporation of viral glycoproteins. Incorporation of foreign glycoproteins into nonnative viral particles is termed pseudotyping (6–8), and researchers from numerous fields routinely exploit this ability to manipulate viral tropism. Although very little is known about the molecular mechanisms that govern this process, it is clearly not random. For example, we have shown using scanning electron microscopy (SEM) that both vesicular stomatitis virus glycoprotein (VSV-G) and murine leukemia virus envelope (MLV Env) glycoprotein are greatly enriched at assembly sites of foreign viruses, such as HIV-1 (9, 10). In contrast, these glycoproteins displayed no obvious clustering in the absence of viral assembly sites.

Several models have been proposed to describe the classes of interactions that could lead to nonrandom glycoprotein incorporation (10–12). One model proposes a direct interaction between viral glycoproteins and other viral structural proteins. Indeed, both HIV-1 Env and MLV Env appear to utilize specific interactions between the cytoplasmic tail (CT) of Env and the viral Gag protein for incorporation of native glycoproteins (13–19). However, it is unlikely that such direct interactions occur in all examples of pseudotyping, especially when the viral glycoprotein and the viral structural proteins are from unrelated viral families, such as VSV-G (rhabdoviral glycoprotein) and HIV-1 core (lentiviral structural proteins). In addition, many glycoproteins are actively incorporated into virions even when their CT has been removed, as has been demonstrated for MLV Env and HIV-1 Env (9, 10, 20, 21).

Indirect interactions between viral components could also lead to the observed enrichment at viral assembly sites. Several variations on this model have been proposed. In principle, one or more cellular proteins could serve as adapters to bridge the various viral components together (12, 22–24), although no specific cellular proteins have been definitively identified to fill this role. Alternatively, interactions with lipid microdomains, such as lipid rafts, could serve as the common element that interacts with both viral components. Even though glycoproteins are not observed to cluster in large preexisting domains in the absence of viral assembly, viral assembly could trigger the coalescence of smaller microdomains that contain the viral glycoproteins. It has been demonstrated that the cholesterol recognition amino acid consensus (CRAC) domain in the membrane-proximal external region (MPER) of HIV-1 Env has a propensity to interact with cholesterol-rich domains in the plasma membrane (25–29) and is necessary for remodeling of the HIV-1 lipid membrane to facilitate fusion (30–32). Additionally, the tryptophan-rich region immediately upstream of the HIV-1 Env CRAC domain has been implicated in HIV-1 Env acquisition into viral particles (30). Thus, the MPER and membrane-spanning domain (MSD) of MLV Env may play roles similar to those of the HIV-1 CRAC domain.

An important challenge that remains is to delineate the molecular interactions, glycoprotein amino acids, and protein domains that dictate active translocation of viral glycoproteins to viral assembly sites. The formation of infectious viral particles requires not only Env incorporation but also Env-mediated recognition of the target cell through protein-protein interactions and initiation of virus-to-cell membrane fusion to form the viral entry pore. These functional requirements generate overlapping selection pressures that confound identification of the protein components that dictate these processes. In this study, we sought to define the amino acids in MLV Env that modulate its active incorporation into viral assembly sites, focusing specifically on the MPER and MSD. To this end, we coupled large-scale in vivo selections with next-generation sequencing (NGS) to characterize a combinatorial mutagenic library of MLV Env glycoproteins with mutations in their MPERs and MSDs. Our screen sought to identify mutants that could (i) be incorporated into viral particles, (ii) promote viral infectivity, and (iii) retain fusogenic activity.

MATERIALS AND METHODS

Generation of the mutagenic MLV Env library in Escherichia coli.

All subsequent mutagenesis and experimentation were done using ecotropic Friend MLV Env sequence (GenBank accession number X02794.1). Forward and reverse randomly mutagenized oligonucleotide primers were annealed and extended using Taq polymerase to create a double-stranded DNA insert. Unique restriction sites were engineered at the 5′ (BstxI) and 3′ (MfeI) ends of the forward and reverse primers which were used for cloning the double-stranded cassettes into the modified MLV packaging construct depicted in Fig. 1B. Following restriction digestion and ligation of the MLV packaging construct (Fig. 1B) and mutagenized cassettes, clones were transformed into E. coli (DH5α) via electroporation (approximately 1.2 × 10⁵ to 1.5 × 10⁵ colonies from transformation). Colonies were pooled and plasmid was isolated using a plasmid extraction kit (Sigma). The purified DNA plasmid library containing the randomly mutagenized MLV Env target region was used for generating the stable cell library described below.

FIG 1.

Target region of mutagenesis and selection strategy. (A) Diagram of the region in MLV Env targeted by random mutagenesis. (B) Strategy for creating the stable mutant cell population and selection procedures. Virus containing the mutant Env library (the red asterisk highlights the location of random mutagenesis) was generated using the three-plasmid system depicted. Virus produced in 293FT cells was used to transduce 293T mCAT-1 cells at a low MOI. Cells were then treated with puromycin to generate a pure cell population wherein a majority of the cells were stably expressing a single unique MLV Env mutant.

Plasmids and cell culture.

The amino acid substitutions were introduced using oligonucleotide-mediated mutagenesis. Constructs expressing the truncated version of MLV Env were created by introducing a stop codon after the sequence encoding RLVQFVK, which removes 25 residues from the cytoplasmic tail. Assays involving HIV infectivity utilized an NL4-3-derived HIV-cytomegalovirus (CMV)-green fluorescent protein (GFP) proviral vector defective for Vif, Vpr, Vpu, Nef, and Env (Vineet Kewal-Ramani, National Cancer Institute). This construct has a CMV immediate early promoter driving a GFP reporter in place of Nef. Generation of the tTA/tetracycline response element (TRE)-Gaussia luciferase (Gluc) system has been described previously (20). Briefly, the gene encoding the tTA (tet-off) protein was cloned into the pQCXIP vector downstream of the CMV immediate early promoter. The TRE-driven Gluc-inducible expression system was generated by introduction of the Gluc gene into the retro-tight-X-hygro retroviral transfer vector.

Cell lines stably expressing Env were produced by the following procedure. 293FT cells were transfected with MLV packaging construct CMV-MLV-GagPol (Walter Mothes, Yale University), CMV-driven VSV-G, and a packageable MLV genome that contained the Env mutations (Fig. 1B). Medium was collected and used to transduce fresh 293FT cells (for the large-scale cell library, this was done at a low multiplicity of infection [MOI]). After 48 h, cells were treated with puromycin to generate a pure cell population stably expressing the relevant Env mutants. To generate cell lines that stably expressed both Env and HIV GagPol, the stable cell lines expressing Env point mutants, which were generated using the above-described procedure, were transduced with virus containing an NL4-3-derived HIV-CMV-Blast-GFP proviral vector defective for Vif, Vpr, Vpu, and Env. CMV-Blast-GFP was inserted in place of Nef. After 48 h, these cells were treated with blasticidin to create pure cell populations stably expressing both Env and HIV-GagPol.

HEK-293FT (Invitrogen), 293T mCAT-1 (Walter Mothes, Yale University), 293T TVA, and 293T mCAT-1 cells stably expressing a TRE-Gaussia luciferase (Gluc) promoter (20) were maintained in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM glutamine, 1 mM sodium pyruvate, 10 mM nonessential amino acids, and 1% minimal essential medium vitamins.

Infectivity assay.

293FT cells were transfected in six-well plates with 500 ng of HIV-CMV-GFP, or 300 ng of CMV-MLV-GagPol and 200 ng CMV-GFP-MLV genome, and 500 ng of Env expression plasmid using 3 μg of polyethylenimine (PEI) per microgram of DNA (33). The medium was changed 6 to 12 h posttransfection to remove residual transfection reagent. The supernatant was collected 24 h following medium exchange and frozen at −80°C for at least 3 h to lyse cells contained within the supernatant. After supernatants were thawed at 37°C in a water bath, samples were centrifuged at 1,500 × g for 5 min to pellet any cellular debris. Five hundred microliters of the supernatant was added to target cells. Cells were collected at 48 h, fixed with 4% paraformaldehyde, and analyzed using an Accuri C6 flow cytometer.

For infectivity using stable cell lines expressing MLV Env, cells were transfected with 700 ng of HIV-CMV-GFP or with 500 ng of CMV-MLV-GagPol and 200 ng of CMV-GFP-MLV genome. Following transfection, the above-described procedure was used to assess infectivity.

Immunoblotting.

293FT cells stably expressing viral proteins were used for Western blotting. Viral samples were pelleted through a 20% sucrose cushion for 2 h at 20,000 × g at 4°C. Residual medium and sucrose were aspirated from the pellet, and samples were resuspended in 6× SDS-PAGE loading buffer. The equivalent of 1 ml of viral supernatant was analyzed by 10% discontinuous SDS-PAGE. Cell samples were detached using 10 mM EDTA–phosphate-buffered saline (PBS) solution and pelleted at 500 × g for 10 min. Pellets were resuspended in RIPA buffer (10 mM Tricl-Cl [pH 8.0], 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, and 140 mM NaCl), and 5 to 10% of the lysate was combined with 6× SDS-PAGE loading buffer and analyzed by 10% discontinuous SDS-PAGE. Proteins were transferred onto a 0.45-μm polyvinylidene difluoride (PVDF) membrane. The membrane was blocked with 5% nonfat dried milk in PBS containing Tween 20 (PBS-T) and probed with rabbit anti-GFP antibody diluted 1:5,000 (Sigma) and mouse anti-HIV p24 hybridoma medium diluted 1:500 (AIDS Research and Reference Program, Division of AIDS, NIAID, NIH; HIV-1 p24 hybridoma [183-H12-5C]). Primary antibody incubations were performed overnight on a rocker at 4°C. Blots were washed with phosphate-buffered saline and 0.1% Tween 20 (PBS-T) and then probed with horseradish peroxidase (HRP)-conjugated anti-rabbit and anti-mouse antibodies diluted 1:10,000 (Sigma) for probing Env and p24, respectively. Visualization of the membranes was performed using Luminata Classico and Crescendo Western HRP chemiluminescence reagents. Imaging was performed using a LAS3000 image analyzer from Fujifilm.

Cell-to-cell fusion assay.

293FT cells stably expressing Env protein were transfected with 500 ng of tTA expression plasmid in a six-well plate. The medium was changed 6 to 12 h posttransfection to remove residual transfection reagent. Transfected cells were cocultured with an equal number of 293T mCAT-1 TRE Gluc cells for 48 h. Twenty microliters of sample supernatant from the cocultured cells was assayed in duplicate for Gluc content with 50 μl of 10 μM coelenterazine in 0.1 M Tris (pH 7.4) and 0.3 M sodium ascorbate.

Surface labeling.

293FT cells stably expressing Env protein were detached using 10 mM EDTA-PBS. Cells were centrifuged at 500 × g for 10 min at 4°C and resuspended in 1% bovine serum albumin (BSA)-PBS blocking solution for 20 min. Subsequently, cells were centrifuged at 500 × g for 10 min at 4°C and resuspended for 1 h in 10 mM EDTA-PBS, 1% goat serum, and primary anti-GFP Alexa-Fluor 647 antibody diluted 1:1,000 (Life Technologies). After incubation, cells were centrifuged at 500 × g for 10 min at 4°C, resuspended in PBS, and analyzed using an Accuri C6 flow cytometer.

RNA harvesting and cDNA synthesis.

Cells grown on 10-cm culture dishes were harvested from each of the four libraries using 10 mM EDTA-PBS. Cells were pelleted at 500 × g for 10 min at 4°C and resuspended in TRIzol reagent (Sigma). Total RNA was harvested using the manufacturer instructions. Pellets were air dried and resuspended in RNase- and DNase-free water.

For cDNA synthesis, 2 μg of the purified RNA was mixed with 1 μl of 100 μM poly(dT) primers and incubated at 70°C for 5 min, chilled on ice, and combined with 4 μl of reverse transcription reaction buffer (Promega), 1 μl of 10 mM deoxynucleoside triphosphate (dNTP) mix, 2 μl of 25 mM MgCl₂, 0.5 μl of 40-U/μl RNase inhibitor (Promega), and 1 μl of 200-U/μl Moloney MLV (MMLV) reverse transcriptase. The reaction volume was brought up to 20 μl with distilled water (dH₂O) and incubated at 42°C for 2 h and then 85°C for 5 min.

Illumina sequencing and analysis.

cDNA libraries were diluted 1:2 with RNase- and DNase-free water, and 5-μl volumes of the diluted samples were used as the template for PCR amplification to introduce Illumina adapters and sequencing indices for multiplexing. Samples were analyzed using Illumina MiSeq (University of Missouri DNA Core Facility) 2 × 150-nucleotide (nt) paired-end reads. Reads were paired using FLASh (http://ccb.jhu.edu/software/FLASH/). Data processing was performed using a locally installed FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/). Fastx clipper was used to trim the 3′ constant region from sequences, and a stand-alone script was used to trim 5′ constant regions. During this process, sequences shorter than 150 nucleotides, or nontrimmed sequences, were discarded. Trimmed sequences were then filtered for high-quality reads using the FASTQ quality filter. Sequences with a Phred quality score less than 30 (99.9% base calling accuracy) at any position were eliminated. Finally, nucleotide sequences were translated to amino acid sequences using the TranSeq program in a locally installed EMBOSS package (http://www.ebi.ac.uk/Tools/st/emboss_transeq).

Preprocessed sequences were then further analyzed using the FASTAptamer toolkit (http://burkelab.missouri.edu/fastaptamer.html) (34). FASTAptamer-Count was used to determine the number of times each sequence was sampled from the population. Each sequence was then ranked and sorted based on overall abundance, normalized to the total number of reads in each population, and directed into FASTAptamer-Enrich. FASTAptamer-Enrich calculates the fold enrichment ratios from a starting population to a selected population by using the normalized reads-per-million (RPM) values for each sequence. After generating the enrichment file, nucleotide changes in each sequence were determine by using FASTAptamer-Distance, which identifies how many mutations each sequence has relative to a wild-type (WT) input sequence. Sequences with fewer than 5 RPM in the starting population or with more than 6 nucleotide mutations were filter removed from the population. Only sequences with a single nonsynonymous amino acid substitution with up to 6 nucleotide substitutions were considered for the final analysis. The fitness index for each substitution was determined by taking the average enrichment or depletion value of all sequences with that specific amino acid mutation and plotting it as a heat map (Fig. 2). Maximal enrichment was set to that of the WT for the specific selection and was colored gold. Maximal depletion was set to the average fitness index of all of the randomly introduced stop codons that occurred in the MPER or MSD for the specific selection and was set to blue.

FIG 2.

NGS selection analysis. NGS results from the large-scale infectivity selection (A), cell-to-cell fusion selection (B), and incorporation selection (C) were analyzed. Blue indicates that a nonsynonymous substitution at the indicated position was depleted following the specified biological selection. Gold indicates that the substitution either was enriched or showed minimal change in sequence representation following the specified biological selection. The portion indicated as “STOP” depicts the observed locations and enrichment of randomly introduced stop codons for each of the libraries. The minimal apparent fitness portion displays the lowest fitness index for each position within the indicated library. Enrichment values for all of the individual mutations are shown in the supplemental material.

Correlative SEM imaging.

The method for imaging the distribution of MLV Env on the cell surface has been described previously (10). Briefly, 293T TVA cells were plated onto coverslips coated with patterned gold and cotransfected with 1 μg of a late-domain-defective HIV-1 Gag expression vector along with 1 μg of GFP-tagged Env expression vectors using PEI. Transfected cells were fixed with 4% paraformaldehyde roughly 36 h posttransfection, and the locations of individual transfected cells on the finder grids were recorded. Cells were labeled with primary mouse anti-GFP (Sigma) and 12-nm gold-conjugated anti-mouse secondary antibody (Jackson ImmunoResearch). Cells were then fixed with 2.5% glutaraldehyde, dehydrated in ethanol, critical point dried, coated with carbon, and imaged with a Hitachi S-4700 field emission scanning electron microscope (FE-SEM). Quantified enrichment was determined by dividing the density of gold particles at viral assembly sites by the density of gold particles at nonassembly sites. Virus-associated density was calculated by counting the gold nanoparticles associated with virus and then dividing this number by the total viral surface area, including the surface underneath the assembly site (n5πr², where n is the number of viruses in the field of view and r is 70 nm). Nonviral density was calculated by counting the gold nanoparticles not associated with virus and then dividing this number by the total nonviral surface area (total surface area of the field of view − nπr², where n is the number of viruses and r is 70 nm). For each Env construct analyzed, 10 images were quantitated.

RESULTS

Creating a stable cell library expressing mutant MLV envelopes.

To determine whether the MPER or MSD of MLV Env contributes to the selective incorporation of MLV Env to viral assembly sites, we generated a densely mutagenized library targeting 50 contiguous amino acid positions (Fig. 1A). The goals were to identify amino acids necessary for active Env incorporation and to differentiate selection pressures associated with active incorporation from those associated with fusogenicity during subsequent infection. Random mutagenesis was achieved by synthesizing and assembling oligonucleotides that contained a positional mutation rate of 3%, correlating to approximately 4.5 nucleotide changes per assembled amplicon. Full-length, double-stranded DNA was digested with unique restriction sites that flank the randomization region and ligated into a retroviral vector (Fig. 1B) and then transformed into E. coli. Plasmid DNA was pooled from 10⁵ to 10⁶ colonies to generate the initial plasmid library. Pooled plasmids were transfected into 293FT cells along with plasmids expressing viral packaging components (MLV Gag/GagPol) and the promiscuous viral glycoprotein VSV-G. Medium from the transfected cells was used to transduce fresh 293FT cells. VSV-G was included in this step to enable generation of stable cell lines through retroviral transduction without relying upon the mutant MLV Env carried in the library. Because a low multiplicity of infection (MOI) was used (<0.1), the vast majority of cells were transduced by one or fewer viruses and therefore expressed a single variant of MLV Env, along with packageable RNA encoding that glycoprotein. Transduced cells were selected for puromycin resistance to create the cell library used in subsequent experiments described below.

We chose a truncated version of MLV Env (Δ25) (9) for library construction because removal of these residues results in a constitutively active glycoprotein population due to the loss of the negative regulation provided by the R peptide (35–37), allowing for study of fusogenic activity. The MLV Env used also contained a GFP tag within the proline-rich region of the protein for use in antibody capture assays. Insertions of GFP or other sequences in this location are well tolerated and have no discernible effect on Env functions (38–40).

Mutant library selection and next-generation Illumina sequencing.

We coupled next-generation sequencing to large-scale biological selections to identify mutant glycoproteins that are impaired for one or more of three specific functions: fusogenic activity, viral infectivity, and incorporation into viral particles. The library was subjected to three independent selection procedures (described below) in which the mutated env gene was maintained only if the encoded Env met the functional criteria of the particular selection. Following these biological selections, the mutated gene segments were recovered by PCR for Illumina sequencing. The sampling frequencies (in reads per million [RPM]) of sequences present following selection were compared with those of sequences prior to selection to determine the apparent fitness of each mutant glycoprotein.

Apparent fitness was estimated as follows. Sequences with low read counts (<5 RPM) in the preselection library or with >6 nucleotide changes per sequence were discarded prior to analysis. To avoid complications from compensatory mutations, sequences that contained more than one nonsynonymous amino acid substitution were also removed. Roughly 2,800 unique sequences in the starting library fulfilled these three criteria. All of the observed amino acid changes were represented by multiple sequences carrying additional synonymous (silent) changes. For instance, the starting cell library contained 15 unique sequences that encoded a single E122D substitution, with an average of ∼1 additional synonymous nucleotide change per sequence. For each unique sequence, the ratio of read frequencies (in RPM) between the starting library and the selected library was calculated, and then the ratios of all of the unique sequences that led to a given single amino acid change were averaged to give the apparent fitness index of that amino acid substitution. For the E122D mutant discussed above, we observed an average apparent fitness index of ∼1.3 in the infectivity selection, compared with ∼3 for the wild type. For the output heat maps in Fig. 2 to 4, the maximum allowable index value (gold) was set to that of the wild-type amino sequence, and the minimum (dark blue) was set to the value observed for sequences in which a stop codon preceded the MSD. Shown above each data set are the minimal observed apparent fitness for each position (“Max. Sens.”) and the apparent fitness values for stop codons observed at those positions (“STOP”) (Fig. 2).

FIG 4.

Transient expression of mutants that displayed striking selection patterns in NGS analysis. Shown are infectivity results from 293FT cells transfected with 500 ng of the indicated Env mutant, 300 ng of MLV GagPol, and 200 ng of CMV-GFP-MLV genome expression plasmids. Open bars indicate Envs with a truncated (Δ25) cytoplasmic tail. Infectivity is shown relative to the appropriate wild-type (WT) control (∼28% ± 9% mean GFP-positive cells for Δ25CT Env). The NGS analysis is shown below the respective amino acid substitution and depicts apparent infectivity fitness. Data are averages from three independent experiments.

Infectivity selection.

To select for MLV Env mutants that were capable of generating infectious particles, the cell library was transfected with a plasmid expressing the viral packaging components; viral glycoprotein was supplied by the library clone. Viral particles produced by transfected cells were collected and used to transduce 293T mCAT-1 cells, which express the receptor for MLV Env. Following puromycin selection, the mutated region was recovered by PCR and deep sequenced. In this selection scheme the wild-type sequences had an average enrichment of ∼3.0, whereas sequences that introduced a stop codon prior to the MSD had enrichment of ∼0.1 (i.e., strongly depleted). Introduction of a stop codon preceding the MSD would result in a nonfunctional protein product that would be turned over by the cell. Therefore, we would expect all stop codons randomly introduced into this region to show strong depletion values, as was observed for all selections. The fact that even the introduction of stop codons generally did not reduce enrichment to 0.0 most likely reflects low-level complementation from cells that were transduced with more that one genome during library production. This selection was the most sensitive of the three selection conditions examined because it placed the greatest functional demands on the encoded Env proteins. Specifically, this selection required MLV Env clones both to be incorporated into viral particles and then subsequently to be fusogenically active. At least one substitution at every position was depleted at least 2-fold in the selected population compared to the wild type (Fig. 2A, Max. Sens.). As expected, introducing a charged residue (R, K, E, or D) anywhere in the membrane-spanning domain severely depleted sequence representation in the functionally selected population, while mutations to amino acids with similarly sized aliphatic side chains were relatively well tolerated. The MPER was less sensitive to change than the MSD overall, but several positions were clearly very sensitive to mutation. In particular, changes to the two membrane-proximal tryptophan residues were highly detrimental to viral infectivity. The last 5 amino acids of the protein were the least sensitive to change.

Fusion selection.

To identify mutants that retained fusogenic activity, the starting library of transduced cells was cocultured with mCAT-1 cells that expressed resistance to hygromycin. Because the library of Env mutants was constructed with a truncated CT, the R peptide was no longer able to negatively regulate fusogenicity, which resulted in a constitutively fusogenic glycoprotein population. Therefore, any cells expressing mutant MLV Env that promoted cell-to-cell fusion would have formed syncytia with the mCAT-1 cells, resulting in resistance to both puromycin and hygromycin. After cocultured cells were treated with both drugs, the mutagenized region was recovered using PCR and deep sequenced. This selection was the least stringent among the three tested in this study. Wild-type sequences had an average enrichment of ∼1.4, while sequences with a stop codon preceding the MSD had an average enrichment value of ∼0.8. Unlike the other two assays, this assay was a negative selection, and incomplete killing of nonfused cells may have contributed to background. Although the dynamic range of this selection was much narrower than the other two selection criteria, the output was remarkably robust, as only ∼10% of sequences had an apparent fitness below 0.9, but addition of a stop codon anywhere in the MPER or MSD reduced the apparent fitness index below this value.

In general, many of the substitutions in the selected fusion population had a minimal impact on fusogenic activity. Nonetheless, some positions were strikingly sensitive to mutation, such as S124, R134, and S143, whereas others displayed more modest patterns of sensitivity, such as T139, T140, P148, and several residues in the MSD. These results are in agreement with our recent examination of residues in the membrane-proximal ectodomain that are critical for MLV Env fusogenic activity (41).

Incorporation selection.

To select for mutant glycoproteins that retained the ability to incorporate into viral particles, the starting library was transfected with expression constructs for MLV packaging components and VSV-G. Our approach was to utilize a viral capture assay (42) to identify MLV Env mutants that were incorporated into viral particles. Briefly, an enzyme-linked immunosorbent assay (ELISA) plate adsorbed with anti-GFP antibody was used to immobilize virus from the library that had incorporated the GFP-tagged MLV Env mutants. Importantly, viral particles were also pseudotyped with VSV-G so that overlaid 293FT cells could be transduced by the immobilized virus and acquire puromycin resistance even if the MLV Env used for capture was not fusogenically active. After puromycin selection of the overlaid cells, the mutated region was amplified by PCR and deep sequenced. The dynamic range observed for this selection was greater than that of the fusion selection but was narrower than the infectivity selection, with enrichment values of ∼2.0 for the wild-type sequence and ∼0.2 for sequences with a stop codon preceding the MSD. Mutations resulting in the greatest reduction of Env incorporation clustered within and immediately adjacent to the MSD and at two positions in the MPER (W127 and W137).

Candidate amino acids that contribute to active incorporation into viral assembly sites.

The selections described above do not directly examine whether a given envelope mutant is distributed randomly across the cell surface or localized to viral assembly sites. However, we reasoned that the most informative mutants with respect to this issue would be those that simultaneously retained fusogenic activity (suggesting that the mutation was not detrimental to overall protein folding, surface expression, or processing), abrogated infectivity (suggesting that the mutant was not incorporated into viral particles), and exhibited decreased incorporation into viral particles. Therefore, we correlated apparent fitness across multiple biological selections with respect to these three criteria.

Amino acid substitutions that produced the lowest fitness indices for infectivity (0.51 or below) were used to establish a mask for the substitution matrix. The fitness indices from the incorporation and fusion selections at those positions were then plotted side by side after removal of data points not included within the mask (Fig. 3). Filtering and overlaying the data sets in this manner identified two residues in the MPER (W127 and W137) and 25 of the amino acids from T140 to L163 in which substitutions resulted in severe incorporation defects (blue on the left) without impairing surface display (inferred from fusogenicity; gold on the right). Importantly, upon visual inspection, these results were found to correlate well with the results for the minimal apparent fitness value from the incorporation selection (Fig. 3, top).

FIG 3.

Two tryptophan residues in the MPER and the MSD are predicted to be important for Env incorporation. Depicted are the amino acid substitutions that were highly defective for infectivity. At each of these positions the incorporation fitness is shown on the left and the fusion fitness is shown on the right. The upper bar depicts the maximum sensitivity at each position in the incorporation assay (Inc. Max.). Red dashed lines outline regions that affect incorporation.

Characterization of mutants that displayed striking selection patterns in NGS analysis.

Our next aims were to identify substitutions from both regions (W127 and W137 in the MPER and T140 to L163 in the MSD) that were incorporation defective and to establish whether the apparent fitness calculated from sequence enrichment reflects true biological fitness. To this end, we chose 14 substitutions (shown in Fig. 4) that covered a range of performances based on the NGS analysis in Fig. 3. These point mutants were transiently expressed in 293FT cells and assayed for the ability to promote viral infectivity (Fig. 4). In general, these single-round infectivity data largely recapitulated the results from the infectivity selection. For example, E122K, M146T, and V167F were all predicted from apparent fitness indices to have a minimal impact on infectivity, and indeed, these substitutions displayed full infectivity compared to Δ25CT Env. Additionally, W137L, T140M, and S143A were all predicted to have detrimental effects on function, and all three displayed greatly reduced infectivity (∼80 to 95% reduction). In contrast, two of the point mutants, W127R and I142R both functioned much better than one would have predicted from the NGS analysis (Fig. 2 and 3). A possible explanation for this discrepancy is that these point mutants were fusogenically active and incorporation defective but were incorporated into particles nonspecifically due to high expression levels in the transfection-based assay. To test this possibility, we generated 293FT cell lines that stably expressed the indicated point mutation, thus reducing their copy number and reproducing the conditions of the biological screen, and then we assayed their abilities to promote viral infectivity (Fig. 5A). Under these conditions, W127R and I142R mutants displayed abrogated infectivity, as predicted from the NGS analysis, suggesting that these mutants were likely defective for the ability to incorporate into viral particles when expressed at physiological levels. In the stable cell lines expressing the individual Env proteins, these single-round infectivity data largely recapitulated the results from the infectivity selection.

FIG 5.

Validation of numerous stably expressed point mutations. (A) Open bars indicate infectivity from stable cell lines expressing indicated mutants when transfected with MLV GagPol and a reporter genome. Infectivity is shown relative to that of Δ25CT Env (∼20% ± 5% mean GFP-positive cells for Δ25CT Env). Hatched bars indicate cell-to-cell fusion activity. Cell lines were transfected with a tet-off expression plasmid and cocultured with a permissive cell line expressing a Gaussia luciferase reporter. Luminescence output is depicted as percent relative to that of Δ25CT Env (average absolute RLU for Δ25CT Env was ∼6 × 10⁴ ± 2 × 10⁴). The NGS analysis is shown below the respective amino acid substitution; apparent infectivity fitness is shown on the left, while apparent fusion fitness is shown on the right. (B) Surface expression of Env point mutants. Data from all three assays are the averages of at least three independent experimental replicates.

Next, we characterized the fusogenic activities of the individual Env mutants. Since these mutants have a truncated cytoplasmic tail and therefore lack the regulatory R peptide, they remain constitutively active and could be tested directly for fusogenic activity. Briefly, each cell line stably expressing an Env mutant was transfected with a plasmid expressing the tTA transcription factor and was cocultured with an equal number of 293T mCAT-1 cells expressing TRE-driven Gaussia luciferase (Gluc) (20). If the Env mutants retain fusogenic activity, the transfected cells fuse with the receptor-expressing cells and Gluc induction occurs in a tTA-dependent manner. Gluc induction can then be correlated between mutant Env and Δ25CT controls to assess fusogenic activity (Fig. 5A). As with the infectivity results, the fusogenic activities of these 14 Env mutants correlated with those predicted from the NGS analysis (Fig. 5A). Interestingly, we identified three point mutations, W127R, W137L, and I142R, that displayed robust fusogenic activity but had abrogated infectivity, further supporting our prediction that these mutants are defective for incorporation into viral particles. To ensure that all mutants were similarly expressed on the cell surface compared to Δ25CT, stable cell lines expressing the mutant Envs were surface labeled and analyzed using flow cytometry. As depicted in Fig. 5B, there were no significant differences among the surface expression levels of the mutants compared to those of the controls.

Identification of mutants defective for incorporation into viral particles.

Next, we wanted to test whether any of our candidate mutants were defective for incorporation into viral particles. We sought to stably infect cell lines that express Env point mutants with a selectable Env-defective HIV-1 provirus, rather than to transfect those cells with proviral plasmid DNA. To accomplish this, we took advantage of an Env-defective HIV-1 provirus that contained a fluorescent reporter and a selectable marker (HIV CMV-Blast-GFP). Following transduction and selection for blasticidin resistance, this approach results in all viral proteins being expressed at approximately physiological level, since on average there would only be one copy of env and gag per cell. Because all of the experiments to this point were performed with MLV cores, we first tested whether the MLV Env mutants behaved similarly when using HIV-1 cores and MLV cores. To accomplish this, we transfected the stable Env cell lines with 500 ng of either HIV-1 or MLV GagPol expression constructs and tested infectivity. The results show no discernible difference in infectivity between HIV-1 and MLV cores (Fig. 6A). Next, we transduced the 14 stable cell lines with our HIV CMV-Blast-GFP and selected transduced cells. To test infectious particle production, medium was collected and used to transduce 293T mCAT-1 cells. The infectivity results from stable transduction (HIV CMV-Blast-GFP [Fig. 6B]) closely matched infectivity data described above, wherein Env was stably expressed and MLV GagPol was expression transiently (Fig. 5A).

FIG 6.

Mutant Env infectivities are similar between MLV and HIV cores. (A) Stable cell lines expressing the indicated Env mutant were transfected with either MLV or HIV cores, and infectivity was assessed. Infectivity is shown relative to that of WT Δ25CT controls of the appropriate core (∼18% ± 4% mean GFP-positive cells for Δ25CT Env with MLV cores and ∼25% ± 9% mean GFP-positive cells for Δ25CT Env with HIV cores). (B) Infectivity output from cells stably expressing the indicated Env mutant and HIV GagPol. Infectivity is shown relative to that of WT Δ25CT controls (∼25% ± 8% mean GFP-positive cells for Δ25CT Env). Data are the averages of three independent replicates.

To assess mutant Env incorporation into viral particles, viral supernatants and cell lysates were collected from the stably transduced cell lines for immunoblotting. MLV Env is produced as a gp85 precursor protein and is processed by a cellular protease into 70- and 15-kDa products (43). In the cases of W127R, W137L, and I142R, identified above, there was a significant reduction in the normalized ratio of gp70/Gag present in viral particles compared to the wild-type control (Fig. 7A, Virus). Analysis of cell lysates revealed that the reduction of gp70 in viral pellets for W127R, W137L, and I142R was not due to a reduction of processed gp70 (Fig. 7A, Cell). Because several of the mutants displayed some reduction in viral incorporation, several independent immunoblots were quantified (Fig. 7B). While several point mutants displayed various levels of decreased Env incorporation, these data indicate that W127R, W137L, and I142R reproducibly displayed the greatest reduction of Env incorporation into viral particles (∼75 to 95% reduction relative to the Δ25CT control).

FIG 7.

Specific point mutations in, or near, the membrane-proximal domain display severely abrogated incorporation into HIV particles. (A) (Top) Immunoblot analysis of cell lysates stably expressing both HIV GagPol and the indicated MLV Env point mutant. (Bottom) Immunoblot analysis of HIV particles pseudotyped with the indicated MLV Env point mutant. Particles were collected from supernatants of 293FT cells stably expressing HIV GagPol and the indicated Env. Shown are three independent replicates to highlight reproducibly decreased Env incorporation for specific point mutants. (B) Immunoblot quantification of four independent replicates of Env incorporation into HIV particles. Particles were isolated from supernatants of cells stably expressing HIV GagPol and the indicated Env mutant. The gp70 signal was normalized to the corresponding capsid p24 band; shown are the averages from all four experiments relative to the WT Δ25CT normalized signal.

W127R, W137L, and I142R Env point mutants are not actively incorporated into HIV-1 assembly sites.

Because a low level of gp70 was detectable in HIV-1 particles by immunoblot analysis on W127R, W137L, and I142R mutant Envs, we wanted to investigate whether this residual incorporation was due to active or passive incorporation. We imaged glycoprotein distribution in the presence of assembling HIV-1 particles using correlative SEM with gold immunolabeling and electron backscatter detection. Individual cells were imaged in tandem, first by fluorescence microscopy and then by SEM (44). Retroviral assembly sites were identified in SEM secondary electron images by their distinct spherical shape when using late-domain-defective viral constructs (PTAP changed to AAAA in the HIV-1 p6 domain) (44). Env was visualized by surface labeling using specific antibodies conjugated with gold nanoparticles (12-nm gold particles), which were imaged using electron backscatter detection. As shown in Fig. 8A, immunolabeled WT Δ25CT Env gold particles (indicated by red dots) predominantly overlapped with the HIV-1 assembly sites. In stark contrast, all three Env mutants (W127R, W137L, and I142R) displayed mostly random distribution across the plasma membrane relative to the HIV-1 assembly sites (Fig. 8B to D). Quantification of the distribution of the mutant Envs in relation to HIV-1 assembly sites relative to Δ25CT Env displayed significantly lower enrichment at viral assembly sites (Fig. 8E).

FIG 8.

W127R, W137L, and I142R are not actively incorporated into HIV-1 particles. Correlative SEM was used to visualize the localization of the Env point mutants in relation to HIV-1 assembly sites. Cells were transfected with indicated Env expression constructs along with a budding defective HIV-1 Gag (PTAP mutated to AAAA) expression construct. (A to D) Cells immunogold labeled to visualize Env distribution relative to assembly sites using electron backscatter detection. Red punctate spots indicate the localization of gold nanoparticle-immunolabeled Env on the respective image. (E) Quantification of the fold enrichment of gold particles with HIV-1 assembly sites visualized using SEM coupled with electron backscatter detection. The quantification is of 10 independent images per indicated Env construct. **, P < 0.01.

DISCUSSION

We previously showed that MLV Env is actively incorporated into HIV-1 assembly sites (10) and that the CT is dispensable for active incorporation (9). Additionally, further studies to probe the hydrophobicity of the MSD as a protein component necessary for active incorporation showed no correlation between hydrophobic index and incorporation (20). Attempts to characterize the domains in Gag that are necessary to facilitate active incorporation have also yielded ambiguous results. Interestingly, the MA domain of MLV Gag is not required for active Env incorporation into viral particles (45). In this study, we evaluated the roles of MPER and MSD of MLV Env in facilitating pseudotyping into HIV-1 particles.

Identification of residues in the MSD crucial for infectivity and fusogenicity.

We identified a number of MLV Env mutations throughout the MPER, MSD, and CTD that impaired fusogenicity. Not surprisingly, all mutations that were severely impaired for fusogenicity were also impaired for infectious particle production. In general, nonfusogenic mutations fell into two categories: those that were also impaired for incorporation and those that were not. The vast majority of the noninfectious mutants that were impaired in both incorporation and fusogenicity contained charged or polar amino acid substitutions within the membrane-spanning domain. Most likely, such mutations cause the transmembrane (TM) protein to be unstable and lead to the loss of both functions. There were five positions (T139, T140, S143, P148, and L163) for which mutations preferentially impaired fusion activity, with minimal impact on accumulation, processing, or incorporation into particles. In the case of P148, it is possible that this residue, in combination with glycine at position 147, forms a molecular hinge that allows for a conformational change within the transmembrane domain (p15E/TM) stalk during activation of the fusion mechanism. Recent cryoelectron microscopy imaging data indicate that the CTD of MLV Env holds the ectodomain in a tight conformation; however, cleavage of the R peptide results in a conformational rearrangement in TM wherein the trimer helices are splayed apart, allowing for fusogenic activation of Env (46). The G147-P148 pair at the N terminus of the MSD may facilitate the molecular reordering of TM during this process. In a recent study where we examined the coordination between the MPER and CTD interfaces that contribute to MLV Env functionality, we independently identified the hydroxyl residues T139, T140, S143, and T144 as being critical modulators of Env fusogenicity without affecting incorporation into viral particles (41). We proposed that these residues contribute a hydrogen-bonding network that stabilizes the TM trimer (47–52). These residues could form interhelical hydrogen bonds within the trimer, resulting in stabilization of the TM domain. This stabilization may allow for the conformational change necessary for transmitting the signal of R-peptide cleavage up the TM stalk. In support of this hypothesis, we found that amino acids with short hydrogen bonding side chains were tolerated at these positions, but bulky nonbonding side chains completely abolished fusogenicity. These hydrogen bonds could also aide in orientation of the monomers within the trimer, establishing a helical interface (53–56). Importantly, mean surface fluorescence and immunoblotting analysis indicated that processivity and trafficking of these mutants are not impaired, supporting the idea that the loss of infectivity is due to the loss of the fusogenicity and not due to an overall gross defect in protein folding (Fig. 4 and 5).

Identification of residues in the MPER and MSD that affect Env pseudotyping.

We identified three MLV Env mutants (W127R, W137L, and I142R) that were defective for infectivity but remained fusogenically active (Fig. 5). This combination of phenotypes suggested that these mutations would display reduced incorporation into viral particles, consistent with the analysis results in Fig. 3. Immunoblotting analysis showed that Env mutants carrying the W127R, W137L, and I142R mutations had significantly decreased gp70 incorporation into viral particles but accumulated to wild-type levels in cell lysates (Fig. 6). Additionally, our studies using correlative SEM with electron backscatter detection confirmed that W127R, W137L, and I142R are not actively incorporated into HIV-1 particles, supporting our hypothesis that these residues are within domains required for pseudotyping MLV Env into HIV particles (Fig. 8).

A majority of the MPER in our target region is predicted to be unstructured, except for a short helical stretch from positions F128 to L131 (Fig. 1A and 3). This helix possibly orients toward the membrane, allowing for W127 interaction with specific lipids that cluster at the viral assembly site once Gag begins to multimerize (Fig. 9). W127, along with W137 and the region from T140 to L163, could act in concert to direct Env trimers to the assembly site through interactions with the lipid environment (Fig. 9). However, the specific lipid properties at the viral assembly site are not well understood.

FIG 9.

Model of the transmembrane portion of an MLV Env trimer interacting with lipids at the plasma membrane. The red asterisks highlight the locations of W127R, W137L, and I142R in the TM domain of MLV Env. We hypothesize that these residues reside in two domains that dictate MLV Env recruitment to HIV-1 assembly sites. MLV Env monomer structural predictions were generated using Phyre² protein fold recognition software. The monomers were assembled into a trimer using the molecular modeling program Chimera based on our prediction. Only the region of TM analyzed in this study is depicted in the trimer. Ex. Cell., extracellular environment.

Both of these regions are primarily hydrophobic. All three residues for which we demonstrated incorporation-defective mutations were large and hydrophobic, and two of these were tryptophan. Structural, biochemical, and biophysical studies performed on the MPER of HIV-1 Env show that the tryptophan residues in this region play a crucial role in Env functionality through direct interactions with cholesterol molecules that cluster at the assembly site. Further, biophysical and biochemical studies on the MPER of HIV-1 Env revealed a pair of molecular helices that immerse themselves in the lipid environment to form a molecular hinge region (57). Therefore, by extension, it is reasonable to suggest that the MLV tryptophan residues that we identified could engage in similar molecular interactions. Taken together, our data suggest that the MSD has a role in MLV Env fusogenicity and that W127, W137, and T140 to L164 are important for Env incorporation into viral particles during assembly.

Finally, the selection strategy outlined here can be readily applied to other targets under diverse selection pressures. It will be particularly valuable in dissecting sequence requirements associated with individual functions for proteins that perform multiple roles. The strategy is expected to minimize difficulties associated with targets that are highly sensitive to traditional mutagenesis, such as the viral capsid protein. Coupling next-generation sequencing technologies with in vivo selection allows for generation of large data sets capable of sampling entire selected populations, providing insight into previously unidentified features of viral protein biology.