Multiple Strategies Reveal a Bidentate Interaction between the Nipah Virus Attachment and Fusion Glycoproteins

ABSTRACT

The paramyxoviral family contains many medically important viruses, including measles virus, mumps virus, parainfluenza viruses, respiratory syncytial virus, human metapneumovirus, and the deadly zoonotic henipaviruses Hendra and Nipah virus (NiV). To both enter host cells and spread from cell to cell within infected hosts, the vast majority of paramyxoviruses utilize two viral envelope glycoproteins: the attachment glycoprotein (G, H, or hemagglutinin-neuraminidase [HN]) and the fusion glycoprotein (F). Binding of G/H/HN to a host cell receptor triggers structural changes in G/H/HN that in turn trigger F to undergo a series of conformational changes that result in virus-cell (viral entry) or cell-cell (syncytium formation) membrane fusion. The actual regions of G/H/HN and F that interact during the membrane fusion process remain relatively unknown though it is generally thought that the paramyxoviral G/H/HN stalk region interacts with the F head region. Studies to determine such interactive regions have relied heavily on coimmunoprecipitation approaches, whose limitations include the use of detergents and the micelle-mediated association of proteins. Here, we developed a flow-cytometric strategy capable of detecting membrane protein-protein interactions by interchangeably using the full-length form of G and a soluble form of F, or vice versa. Using both coimmunoprecipitation and flow-cytometric strategies, we found a bidentate interaction between NiV G and F, where both the stalk and head regions of NiV G interact with F. This is a new structural-biological finding for the paramyxoviruses. Additionally, our studies disclosed regions of the NiV G and F glycoproteins dispensable for the G and F interactions.

IMPORTANCE Nipah virus (NiV) is a zoonotic paramyxovirus that causes high mortality rates in humans, with no approved treatment or vaccine available for human use. Viral entry into host cells relies on two viral envelope glycoproteins: the attachment (G) and fusion (F) glycoproteins. Binding of G to the ephrinB2 or ephrinB3 cell receptors triggers conformational changes in G that in turn cause F to undergo conformational changes that result in virus-host cell membrane fusion and viral entry. It is currently unknown, however, which specific regions of G and F interact during membrane fusion. Past efforts to determine the interacting regions have relied mainly on coimmunoprecipitation, a technique with some pitfalls. We developed a flow-cytometric assay to study membrane protein-protein interactions, and using this assay we report a bidentate interaction whereby both the head and stalk regions of NiV G interact with NiV F, a new finding for the paramyxovirus family.

INTRODUCTION

The paramyxovirus family includes many important negative-sense, single-stranded enveloped RNA viruses, such as measles (MeV), mumps (MuV), respiratory syncytial (RSV), Newcastle disease (NDV), parainfluenza (PIV) Nipah (NiV), and Hendra (HeV) viruses and human metapneumovirus (hMPV) (1). With very few exceptions, paramyxoviruses require two distinct viral envelope glycoproteins to facilitate both virus-host cell membrane fusion during viral entry and spread within infected individuals through cell-cell fusion (syncytium formation): the attachment (G, H, or hemagglutinin-neuraminidase [HN]) and fusion (F) glycoproteins. The attachment glycoprotein binds the cell receptor, in turn activating F to execute virus-host cell membrane fusion, facilitating entry of the viral genome into the host cell (2). While we are beginning to understand some aspects of the paramyxoviral membrane fusion process, for most paramyxoviruses the specific regions of the attachment and fusion glycoproteins that interact during this process remain elusive (3). Since the membrane fusion process is responsible for both viral infection and spread from cell to cell, characterization of the attachment and fusion glycoprotein interactive regions is crucial to our understanding of viral infections and to creating methods to combat them.

NiV is an emerging zoonotic biosafety level 4 (BSL4) pathogen of the Henipavirus genus that causes respiratory distress and encephalitis within infected individuals and has mortality rates in humans of 40 to 90% (4). There is currently no approved vaccine or treatment available. For NiV entry, the attachment glycoprotein G binds the host cell receptor ephrinB2 or ephrinB3 (5–7), causing several conformational changes in G (8, 9) that in turn trigger a series of conformational changes in F that executes virus-host cell membrane fusion (2, 10, 11). As for most paramyxoviruses, little is currently known concerning the region(s) of G and F that interacts during membrane fusion (1, 3).

NiV G is a type II transmembrane glycoprotein that consists of an N-terminal cytoplasmic tail that is attached to a transmembrane domain, followed an ectodomain composed of a stalk region (G_stalk) and then by a C-terminal receptor-binding globular head region (G_head) (Fig. 1A). G forms a dimer of homodimers, mediated through disulfide bonds formed via three cysteine residues within the NiV G stalk (12). NiV F is a type I transmembrane glycoprotein consisting of an N-terminal ectodomain containing a hydrophobic fusion peptide that inserts into target host cell membranes during membrane fusion, physically followed by two helical regions (HRs), termed HR1 and HR2, and by a transmembrane domain and C-terminal cytoplasmic tail (1, 13). NiV F is produced as an uncleaved homotrimer precursor, F₀, which is endocytosed and cleaved by cathepsin L into two subunits, F₁ and F₂, that remain disulfide bond linked (13, 14) (Fig. 1B).

FIG 1.

Overview of the NiV G and NiV F glycoproteins. (A) Schematic of NiV G. CT, cytoplasmic tail; TM, transmembrane domain; HA, hemagglutinin tag. Numbers represent amino acid residues. (B) Schematic of NiV F. HR, heptad repeat region; FP, fusion peptide. The dashed line represents the cleavage site. A cytoplasmic AU1 or F₂ subunit FLAG tag was used for detection. (C) Schematic of the soluble G ectodomain construct, G_ecto. Igκ, immunoglobulin kappa light chain sequence. (D) Schematic of the soluble G head construct, G_head. (E) Schematic of the soluble G stalk construct, G_stalk. (F) Schematic of the soluble F ectodomain construct F_tri. The GCNt motif represents a trimeric coiled-coil domain.

There is currently strong evidence that the NiV G stalk plays a central role in F triggering. A headless NiV mutant, in addition to headless mutants for PIV5, NDV, MuV, and MeV, can trigger their respective F homologues to execute membrane fusion (8, 15–17), implying that the G/H/HN stalk interacts with F during F triggering. Mutation of the stalk cysteine residues important for NiV G oligomerization eliminates fusion (12). Similarly, addition of extra cysteine residues to the MeV H stalk prevents fusion, likely by creating extra disulfide bonds that disrupt proper H oligomerization and folding and/or conformational changes (18). Further, mutations to hydrophobic residues in the MeV H stalk disrupt fusion (19), and elongated MeV H constructs can still allow fusion while truncated stalk mutants cannot (20). For several paramyxoviruses, chimeric HN proteins with the stalk of one viral HN and the head of another cause fusion only when they are expressed with an F protein homologous to the HN protein stalk segment, suggesting that the HN stalk, and not the head, is responsible for F interactions (21–26). Additions of N-glycans disrupting the stalks of PIV5 and NDV HN resulted in a reduction or complete loss of fusion and altered interactions with F (27, 28). Interestingly, however, the addition of N-glycans to the NiV G stalk did not disrupt its interaction with NiV F even though fusion was disrupted (29), suggesting that, unlike for other paramyxoviruses, for NiV the head, in addition to the stalk, may interact with NiV F. Additionally, a chimera with the NiV G stalk and NDV HN head was unable to trigger fusion when coexpressed with either NiV F or NDV F, while the reciprocal chimera was able to trigger fusion when expressed with NDV F but not with NiV F, further suggesting that fusion triggering for the henipaviruses may be more complex than for other paramyxoviruses, such as those expressing HNs (21).

There are two proposed models for paramyxoviral (and particularly the Paramyxovirinae subfamily) glycoprotein-glycoprotein interactions. Which model applies to each virus appears to correlate with the type of receptor utilized by the virus. In the association model, proposed for paramyxoviruses such as PIV5 and NDV that bind sialic acid receptors, F and HN do not appear to interact until receptor binding occurs, after which HN-F interactions ensue, triggering fusion (30–32). In contrast, viruses with proteinaceous receptors, such as the henipaviruses NiV and HeV or the morbilliviruses MeV and canine distemper virus (CDV), are suspected to follow a dissociation model. In this model G and F or H and F interact prior to receptor binding, upon which the two glycoproteins relatively dissociate, releasing F to execute membrane fusion (2, 3, 11, 33–35). Though it was previously thought that G/H clamps F in a prefusion form and that dissociation releases F for fusion, recent studies for MeV and CDV suggest that this is not the case (36, 37). Instead, it has been proposed that binding of H to the cell receptor triggers conformational changes within H that alter H-F interactions, triggering fusion (38, 39). This recently proposed mechanism for the morbilliviruses is consistent with the mechanism we proposed for NiV (8).

Traditionally, viral protein-protein interaction experiments have heavily relied on the use of coimmunoprecipitation (co-IP) assays, but this method can be misleading for membrane proteins since the membrane itself in the micelles produced by the detergent used can serve as a linker between the proteins under analysis. Additionally, co-IP relies on detergents that can potentially alter protein conformations and protein-protein interactions. Here, we developed a new flow-cytometric strategy to complement co-IP to study membrane protein-protein interactions. This new strategy could detect interactions between various different protein constructs. The strategy utilizes a soluble form of the glycoprotein (either soluble G or soluble F) in combination with a wild-type (wt; full-length transmembrane protein) version of its partner protein (F or G, respectively). The soluble protein will then be detected on cell surfaces only when bound to its partner transmembrane glycoprotein. Although the flow-cytometric strategy requires the caveat that one of the proteins must be soluble (and thus not full-length), this strategy, in combination with co-IP, can strengthen the validity of the conclusions drawn regarding transmembrane protein-protein interactions.

Additionally, we created truncated NiV G constructs that revealed that both the NiV G head and stalk regions are involved in F interactions. This bidentate interaction differs from other interactions known for paramyxoviruses, where only the stalk region is shown to interact with F, and thus sets the henipaviruses mechanistically apart from other paramyxoviruses. Using a soluble F construct, we also determined that the ectodomain of F is sufficient for the G-F interactions. Additionally, NiV G deletion mutants revealed regions in the NiV G head dispensable for the G-F interactions. Altogether, these results provide important insights into paramyxoviral and henipaviral glycoprotein-glycoprotein interactions and the membrane fusion process and provide a new strategy to study membrane protein-protein interactions.

MATERIALS AND METHODS

Cell culture.

293T and PK13 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum (FBS). 293T cells were obtained from the American Type Culture Collection, and PK13 (porcine fibroblast) cells were a gift from Irvin Chen at the University of California, Los Angeles.

Plasmid constructs.

Codon-optimized wild type (wt) NiV F with an AU1 tag, NiV G, or the G ectodomain (G_ecto; stalk and globular head regions) tagged with hemagglutinin (HA) were expressed in pcDNA3.1 plasmids as described previously (5). G_stalk and G_head mutants were constructed with an Igκ secretion signal added to C-terminal residue 71 and 177, respectively (Fig. 1A and C to E). A soluble NiV F construct, F_tri, with a C-terminal trimeric coiled-coil (GCNt) motif after residue 494 and a FLAG tag at the cleavage site, was cloned into a pCAGGs vector (Fig. 1B and F) using methods similar to those described previously for morbilliviruses (40). NiV F and G deletion mutants were created by site-directed mutagenesis using a QuikChange kit (Stratagene). G deletion mutants were created using G_head as a template so that the stalk would not be a confounding factor.

Rabbit antiserum production.

Rabbit antiserum (monoclonal antibody 92 [MAb92]) was produced by immunizing New Zealand White rabbits through electroporation with codon-optimized plasmids expressing NiV M (matrix protein), F, and G as previously described (5, 34).

Flow cytometry.

EphrinB2/ephrinB3 (ephrinB2/B3) receptor-negative PK13 cells in six-well plate wells were transfected with 2.0 μg of total DNA/well. The fluorescent protein DsRed was used as a transfection marker. As negative controls, cells were transfected with pcDNA3 alone, DsRed alone, or wt NiV F or G alone. All other cell samples were transfected with 0.2 μg of DsRed. As additional negative controls for the proteins under analysis, cells were transfected with soluble F or G constructs without the complementary wt full-length F or G, with 0.2 μg of G_ecto, G_head, G_stalk, or F_tri, and with 1.6 μg of pcDNA3. Test cells were transfected with 1.5 μg of wt NiV F and 0.3 μg of G_ecto, G_head, or G_stalk or with 1.5 μg of wt NiV G and 0.3 μg of F_tri. PIV5 glycoproteins were also used as controls; PIV5 F was used with G_ecto, and PIV5 HN was used with F_tri transfected in the ratios listed above. Transfected cells were collected at 20 to 24 h posttransfection and incubated for 1 h at 4°C with 1 primary antibodies (MAb92 or anti-FLAG for NiV F, anti-HA for NiV G, MAbF1a for PIV5 F, and MAbs HN1b and HN4b for PIV5 HN) diluted 1:200 to 1:1,000, followed by three washes in fluorescence-activated cell sorting (FACS) buffer (1% FBS with phosphate-buffered saline [PBS]). Cells were next incubated with fluorescent anti-mouse or anti-rabbit Alexa Fluor 647 or 488 antibody (Life Technologies, NY) diluted 1:200 for 30 min at 4°C, followed by an additional two washes. Cells were fixed in 0.5% paraformaldehyde (PFA) and read on a flow cytometer (Guava easyCyte8 HT; EMD Millipore, MA). Analysis was done by first gating for live cells based on forward scatter (FSC) and side scatter (SSC) to measure cell size and granularity, respectively. Live cells were then selected for DsRed (yellow channel) expression to select a population of live transfected cells. Fluorescence intensities of these live transfected cells were then viewed for both the 647/red channel (representing G) and 488/green channel (representing F). Populations positive and negative for F (for measuring G_ecto cell membrane retention) or G (for measuring F_tri cell membrane retention) were gated separately. The G or F geometric mean fluorescent intensity (MFI) was measured for each respective population, and then MFIs of the positive population were normalized to the MFIs of the negative population. Compensation using cells expressing DsRed, F, or G alone was performed as needed.

Western blot analysis.

Transfected 293T cells expressing wt or mutant NiV G in the presence or absence of wt NiV F or expressing wt NiV G in the presence or absence of wt or mutant NiV F were lysed in radioimmunoprecipitation assay (RIPA) buffer (Millipore) supplemented with complete protease inhibitor (cOmplete Mini; Roche). Lysates were separated on a 10% gel for SDS-PAGE analysis. Immunoblotting was performed using mouse anti-AU1 (BioLegend), rabbit anti-HA (Bethyl Laboratories), or mouse anti-FLAG (Sigma) antibody at dilutions of 1:250 to 1:2,000. Fluorescent secondary antibodies were diluted 1:1,000 to 1:2,000, and blots were imaged on a Bio-Rad Chemidoc 249 imager (8, 41).

Coimmunoprecipitation.

Cells transfected with NiV wt or mutant F and G expression plasmids were lysed in RIPA buffer as described above. Cell lysates were incubated with 40 μl of anti-HA microbeads (Miltenyi Biotech) at 4°C for 30 min with rotation. Lysates were purified and eluted over microcolumns (Miltenyi Biotech), with RIPA buffer used for all wash steps. Proteins from cell lysates or column elutions were separated by 10% polyacrylamide gel electrophoresis (PAGE). F and G proteins were detected by Western blotting, as detailed above (8).

Cell-cell fusion.

293T or Vero cells in six-well plates were transfected with plasmids expressing DNA of wt or mutant NiV F and wt or mutant NiV G (3:1 ratio; 2 μg of total DNA) using Lipofectamine transfection reagent (Thermo Fisher Scientific). At 18 to 24 h posttransfection cells were fixed in 0.5% paraformaldehyde. Syncytia were counted microscopically, with a syncytium defined as four or more nuclei per cell (8, 41).

RESULTS

A new flow-cytometric strategy that uses soluble NiV G and full-length F is capable of detecting membrane protein-protein interactions.

To complement co-IP methodologies, we aimed to detect G-F interactions via flow cytometry, using an HA-tagged soluble NiV G construct, G_ecto, which contains the G ectodomain (stalk and globular head regions) but no transmembrane or cytoplasmic region (Fig. 1C) (5). Receptor-negative PK13 cells were transfected with the red fluorescent protein DsRed as a transfection marker, NiV F, and wild type (wt) G or G_ecto. As a control, PK13 cells were transfected with pcDNA3 in the place of NiV F to account for any binding of soluble G to the cell surface of PK13 cells although these are considered receptor-negative cells (8, 42, 43). We observed that G expression was generally lower in the presence of F, likely due to the cell's protein production machinery having to divert energy to produce two different proteins (F and G) instead of one (G only). To account for this we transfected PK13 cells with plasmids expressing either 0.2 μg of DsRed/0.2 μg of G (or G_ecto)/1.6 μg of pcDNA3 or 0.2 μg of DsRed/0.3 μg of G (G_ecto)/1.5 μg of F DNA, as we determined that these ratios resulted in similar levels of G_ecto expression in the presence or absence of F. At 18 to 24 h posttransfection cells were collected and stained with a rabbit monoclonal antibody, MAb92, to detect F and with a mouse anti-HA polyclonal antibody to detect G/G_ecto, followed by staining with Alexa Fluor 488 and 647 secondary antibodies, respectively (Fig. 2A). Cells transfected with pcDNA3 alone, DsRed alone, only F, or only G were used as compensation controls during flow-cytometric analysis.

FIG 2.

A novel flow-cytometric method utilizing soluble NiV G is capable of detecting membrane protein interactions. (A) Schematic of the flow-cytometric assay. Soluble G will not be retained at the cell surface when no F is present, resulting in low antibody binding and low fluorescence (left). In the presence of F, however, soluble G can interact with F, is retained at the cell surface, and is detected by antibodies, resulting in fluorescence (right). 1° Ab, primary antibody; 2° Ab, secondary antibody. (B) Fluorescence plots for the flow-cytometric assay. Live cells were chosen based on forward and side scatter, followed by a gate for transfected cells (DsRed-positive cells). Fluorescence intensities of live transfected cells for the 647 (G) and 488 (F) channels were visualized, and a quad gate was used to show the distribution of cells. The lower left quadrant represents F- and G-negative cells, the lower right quadrant represents F-positive cells, the upper left quadrant represents G-positive cells, and the upper right quadrant represents the F- and G-positive cells. pcDNA3 (PC) alone, F alone, G alone, F plus G, and pcDNA3 plus G_ecto are shown as controls. The increase in F- and G-positive cells (upper right quadrant) between the G_ecto versus the F+G_ecto plot demonstrates that the soluble G_ecto is being retained by F at the membrane and detected via flow cytometry. (C) Normalized G MFIs of G_ecto from the F-positive or F-negative gate. Averages and standard deviations are shown (n = 17). *, P < 0.001.

Since G_ecto is soluble and since PK13 cells express very little (if any) ephrinB2/B3 receptors, we observed that G_ecto was not retained at the cell surface autonomously at significant levels, but the levels significantly increased in the presence of F (P < 0.001), suggesting that G_ecto interacts with F and is thus retained at the cell surface and detectable via flow cytometry. Additionally, this increase was not observed in cells transfected with G_ecto and PIV5 F, demonstrating that the interaction is specific for NiV F-G (Fig. 2B). For quantitative analysis, F-positive and F-negative populations were gated separately, the geometric G mean fluorescent intensities (MFIs) were measured for each, and then either the F-positive G MFI was normalized to the F-negative G MFI (Fig. 2C) or the F-positive G MFI was subtracted from the F-negative G MFI (Table 1). The G MFIs were consistently stronger for the F-positive cell population, confirming that G_ecto was retained on the cell surface by the presence of F. Additionally, the percentage of cells in the F- and G-positive quadrant increased when F was present (Table 1).

TABLE 1.

F and G construct combinations and their flow cytometric values

F and G combination	MFIa						% of cells in the F+G-positive quadrant in:				% increase in F+G-positive cells in:
	Cell population				Difference between populations	Normalized value	Fneg population	Fpos population	Gneg population	Gpos population	Fpos populationb	Gpos populationc
	Fpos	Fneg	Gpos	Gneg
F+Gecto	586.3	72.9			513.4	100	0.9	4.2			3.3
F+Ghead	543.2	69.8			473.4	92.2	0.5	5.3			4.8
F+Gstalk	592.4	75.1			517.3	100.8	0.7	4.7			4.0
Ftri+G			882.8	106.5	776.3	100			1.3	6.8		5.5
Ftri+G167			847.3	110.2	737.1	95.0			1.3	7.7		6.4

^aGeometric mean fluorescence intensity (MFI) values were calculated by gating both F-positive (F_pos) and F-negative (F_neg) populations and G-positive (G_pos) and G-negative (G_neg) populations and then measuring the G and F MFIs, respectively, for each. Negative MFI values were then subtracted from positive MFI values. The resulting values for the G and F MFIs were then normalized to those of G_ecto and F_tri, respectively. Values represent one representative experiment.

^bThe percent increase in F+G-positive cells were calculated by subtracting the percentage of F+G-positive cells in the negative control (F-negative population; pcDNA3 + soluble G construct) from the percentage of F+G-positive cells in the F-positive population (F + soluble G construct).

^cThe percent increase in F+G-positive cells was determined by calculating the difference in percentages of cells in the F+G quadrant in the presence (G-positive population) and absence (G-negative population) of G or G₁₆₇.

Reciprocal soluble NiV F and full-length G experiments validate the flow-cytometric strategy to detect membrane protein-protein interactions.

To further test our flow-cytometric strategy, we tested the binding of a soluble NiV F with full-length (wt) NiV G. We created a soluble NiV F construct, termed F_tri, having a FLAG tag on the F₂ subunit directly after the cleavage site (40) and containing a C-terminal GCNt trimeric coiled-coil domain (44) (Fig. 1F). Cells were transfected using the same method as before, with the exception that F_tri and wt G were now used in the place of G_ecto and wt F. Additionally, the wt F plasmid used also contained a FLAG tag in the same position as that in F_tri (Fig. 1B). The wt F-FLAG construct is functional in a cell-cell fusion assay (data not shown; also our unpublished data). At 18 to 24 h posttransfection, cells were collected and stained with mouse anti-FLAG and rabbit anti-HA primary antibodies, followed by Alexa Fluor 488 and 647 secondary antibodies, respectively. Cells were gated into G-positive and G-negative populations, and the geometric mean F MFIs were measured for each population. Once again we observed a clear shift in F- and G-positive cells when F_tri was transfected with G compared to results with F_tri alone (Fig. 3A and B; Table 1). This shift did not occur in cells transfected with PIV5 HN and F_tri (Fig. 3A).

FIG 3.

A soluble F construct binds wt G via both flow-cytometric and co-IP analyses. (A) Flow-cytometric plots of the F_tri construct and controls. PC, pcDNA3. G detection is shown on the y axis, and F detection is shown on the x axis. (B) Normalized F MFI values for F_tri from the G-positive or G-negative gate. Averages and standard deviations are shown (n = 9). *, P < 0.001. (C) Cell lysates of cells transfected with either wt or soluble F and wt G. NiV G was blotted with rabbit anti-HA, and NiV F was blotted with mouse anti-FLAG. (D) Pulldown of NiV G and co-IP of F for transfected cells. Antibody blotting was performed as described for panel C.

We then confirmed our results via more standard co-IP assays. 293T cells were cotransfected with wt G and either wt F or F_tri. At 20 to 24 h posttransfection the cells were collected and lysed. Cell lysates were incubated with anti-HA beads and run through a co-IP column (Miltenyi Biotec). Both cell lysates and co-IP eluates were processed by SDS-PAGE and blotted with mouse anti-FLAG (to detect F) and rabbit anti-HA (to detect G) primary antibodies, followed by two distinct fluorescent secondary antibodies of two different fluorescence wavelengths. Co-IP assays confirmed that F_tri interacts with NiV G (Fig. 3C and D). Of note, only F₀ (and not F₂) was detectable by co-IP. We previously demonstrated that cleaved (mature) F can be detected via co-IP (9, 41). Noticeably, no F₂ band was observed for F_tri in the cell lysates (Fig. 3C), which is expected since F_tri is readily secreted and, unlike wt F, is likely not endocytosed for cleavage by cathepsin L (14). These co-IP results confirmed that our flow-cytometric strategies reliably detect protein-protein interactions using complementary approaches and, importantly, that uncleaved F is capable of interacting with G, as previously thought based on co-IP results alone (33–35, 41).

Truncated G constructs demonstrate a bidentate interaction between G and F.

While the NiV G stalk is known to trigger NiV F and likely interacts with F (8), previous evidence based on addition of N-glycans to the G stalk region has suggested that the NiV G head may also interact with F (29), yet no direct evidence of the presumed bidentate interaction currently exists. To provide evidence of a bidentate interaction, we created additional G constructs containing only the G head region (G_head; amino acids [aa] 178 to 602) (Fig. 1D) or containing just the stalk region (G_stalk; aa 71 to 177) (Fig. 1E). These G constructs were tested for their interactions with wt NiV F first using our flow-cytometric assay. Both G_head and G_stalk interacted with wt F via flow cytometry, as shown with G MFIs for each construct normalized to the G_ecto values and also by the increase in the percentages of cells doubly positive for F and G (Fig. 4A and B; Table 1). These results were then confirmed by co-IP (Fig. 4C and D), demonstrating that both the G head and stalk can independently interact with NiV F.

FIG 4.

Both the NiV G head and stalk regions alone are capable of interacting with NiV F. (A) Flow-cytometric plots of the NiV G constructs. PC, pcDNA3 vector-only control. G detection is shown on the y axis, and F detection is shown on the x axis. (B) Normalized G MFI values of the NiV G constructs determined by flow cytometry. Values are normalized to G_ecto. Averages and standard deviations are shown (n = 5). (C) Cell lysates of cells transfected with wt F and either wt or mutant G constructs. NiV G was blotted with rabbit anti-HA, and NiV F was blotted with mouse anti-AU1. Asterisks denote the G protein band. (D) Pulldown of NiV G and co-IP of F for transfected cells. Antibody blotting was performed as described for panel C. (E) Syncytium quantification of wt F transfected with wt or soluble G construct DNA. Values are normalized to those of wt G (n = 3).

Since G-F interactions are thought to precede F triggering and cell-cell fusion events, we then tested if our soluble G constructs were capable of causing cell-cell fusion when coexpressed with wt NiV F. Both 293T and Vero cells were transfected with plasmid DNA of wt F and of either wt or soluble G constructs. After 18 to 24 h cells were fixed, and syncytium formation was assessed microscopically. While the wt G and F combination had considerable syncytium formation in both cell types, none of the cells transfected with soluble G constructs had syncytium formation (Fig. 4E), suggesting that the ectodomain region of G is not sufficient to induce cell-cell fusion.

Soluble NiV F interacts with headless NiV G.

Our lab has previously created a headless NiV G construct consisting of amino acids 1 to 167 (G₁₆₇), which thus contains the G cytoplasmic tail and transmembrane domain that allow it to be expressed at the cell membrane, unlike the G_stalk mutant. This headless mutant is functional as it is able to trigger F to execute membrane fusion and has been useful for F-triggering mechanistic studies (8, 9). To determine if the ectodomain of F alone is capable of interacting with the stalk region of G alone, we tested this mutant, termed G₁₆₇, with our soluble F_tri construct. Flow cytometry revealed an interaction occurring between F_tri and G₁₆₇ (Fig. 5A and B; Table 1), demonstrating that the cytoplasmic tail and transmembrane domain of F are not necessary for the G-F interactions. These results were also confirmed by co-IP (Fig. 5C and D), further supporting our findings that the G head region is not necessary for G-F interactions to occur (5E). This combination of truncated glycoproteins was also unable to induce cell-cell fusion (data not shown).

FIG 5.

Soluble F can interact with a headless G mutant. (A) Flow cytometry plots of F_tri with the headless G stalk mutant G₁₆₇. PC, pcDNA3. G detection is shown on the x axis, and F detection is shown on the y axis. (B) Normalized F MFI values determined by flow cytometry. The value for F_tri plus G was set at 100%. Averages and standard deviations are shown (n = 3). (C) Cell lysates of cells transfected with wt G or G₁₆₇ and either wt or soluble F. NiV G was blotted with rabbit anti-HA, and NiV F was blotted with mouse anti-FLAG antibodies. Asterisks denote the G protein band. (D) Pulldown of NiV G and co-IP of F. Antibody blotting was performed as described for panel C. (E) Model of G-F interactions. The headless NiV G₁₆₇ mutant (yellow and black) is capable of interacting (represented by red lines) with the ectodomain of NiV F (green). For simplicity only monomers are shown.

Identification of dispensable regions for G-F interactions.

We next aimed to determine further smaller regions of G and F important or dispensable for the G-F interactions by creating deletion mutants for each glycoprotein, with each mutant lacking anywhere from 5 to 80 amino acids in a specific region. Targeted deletion regions were chosen based on X-ray crystallographic images for NiV and other paramyxoviruses (45–48), and protein regions were chosen that appeared to be exposed and therefore had the potential to interact with their partner glycoprotein or that looped out, suggesting that they could potentially be removed without destroying the overall integrity of the glycoprotein.

We created seven NiV G and 11 NiV F deletion mutants by site-directed mutagenesis (Table 2). NiV G head mutations were made in the G_head mutant background (no stalk) so that the stalk region would not act as a confounding G-F interaction factor. Of these, only three G mutants were expressed and secreted properly (Fig. 6A), as measured by immunoprecipitation of transfected cell lysates. Two of these mutants, G with residues 196 to 211 deleted (G_196–211) and G with residues 372–393 deleted (G_372–393), were stalkless versions of full-length G mutants characterized in previous studies (8, 49) and thus had regions deleted that we suspected could be successfully deleted. Of the 11 F mutants created, none were expressed at the cell surface, as measured by flow cytometry using a variety of different F monoclonal and polyclonal antibodies to account for potential different F conformations (Table 2 and data not shown). Interaction studies of the three expressed G deletion mutants via both flow cytometry and co-IP strategies showed interactions with F (Fig. 6B to D), demonstrating that these three G regions are not required for G-F interactions. Additionally, the number of nonfunctional mutants we observed demonstrates how integral these protein regions are to preserving protein structure and how sensitive these proteins are to alterations (Table 2). None of the G mutants were able to cause fusion, which is expected since the G_head mutant itself is fusion dead (Fig. 4D).

TABLE 2.

Cell expression and G-F interactions of NiV G and NiV F deletion mutants

NiV mutant	Cell expression	G-F interaction determined by:
		Flow cytometry	Co-IP
G constructs (aa)
196–211	+	+	+
340–350	−	−	−
372–393	+	+	+
400–416	−	−	−
419–430	−	−	−
515–520	+	+	+
550–556	−	−	−
F constructs (aa)
61–73	−	−	−
63–70	−	−	−
116–151	−	−	−
248–263	−	−	−
251–260	−	−	−
299–310	−	−	−
357–371	−	−	−
361–444	−	−	−
365–445	−	−	−
374–400	−	−	−
400–428	−	−	−

FIG 6.

Deletion mutants for NiV G reveal regions not necessary for G-F interactions. (A) Western blot analysis of G_head or G deletion mutants. Supernatants from cells transfected with wt NiV F and G mutants were immunoprecipitated with anti-HA beads and processed by SDS-PAGE and Western blot analysis. Wild-type G was used as a negative control as it is not secreted into the supernatant. (B) Flow cytometry interaction assay for the three expressed G deletion mutants, with G MFI values normalized to those of G_head. Averages with standard deviations are shown (n = 3). (C and D) Antibody blotting of cell lysates and co-IP of the G deletion mutants. Blotting was performed as described in the legend of Fig. 4.

Altogether, our results identified regions in NiV G and F dispensable for the G-F interactions: the cytoplasmic tail and transmembrane domains of both G and F and three regions within the head of G. The new flow-cytometric strategy complemented standard co-IP assays and provided evidence that the stalk and head regions of G and the ectodomain of F play important roles in a novel bidentate NiV G-F interaction. In future studies, it will be valuable to elucidate how these G-F interactions change during the course of the membrane fusion process.

DISCUSSION

Coimmunoprecipitation, bifluorescence complementation, and other strategies to study membrane protein-protein interactions have various downsides. Here, we have developed a new flow-cytometric method to detect membrane protein-protein interactions that circumvents these pitfalls. Our strategy utilizes a soluble version of a membrane protein, detectable via antibody staining by flow cytometry only if it interacts with a second full-length protein anchored to the cell surface. Furthermore, we have shown that this assay works using either soluble G or soluble F, with over 4-fold average increases in MFI values, demonstrating the versatility of the assay (Fig. 2 and 3; Table 1), and results were confirmed via co-IP. The new flow-cytometric strategy is useful for observing interactions between two membrane proteins. Importantly, this strategy revealed a bidentate interaction between the NiV head and stalk regions of G and F (Fig. 4).

Many studies have attempted to elucidate paramyxovirus glycoprotein interactions due to their high relevance to viral entry, yet the specific regions of HN/H/G that interact with F remain predominantly elusive. For the fusion protein, a hydrophobic loop of PIV5 F, specifically residues 391 to 395, was shown to interact with HN (50). Additionally, 21 amino acids in the PIV5 F head showed specificity for the simian virus 41 (SV41) HN protein (51). Alterations to MeV F residues Y349, Q383, and L394 all resulted in reduced co-IP with MeV H, and residues in the base of the MeV F head were found to interact with H (52). For the attachment protein, mutations of residues 174 to 183 in the HeV stalk region disrupted interactions with HeV F (53), residues 104 to 125 of the MeV H stalk were shown to be important for F interactions (19), and mutations in residues 94 to 97 of MeV H increased avidity with F (54). Alteration of the domain of residues 84 to 105 in NDV HN reduces both F interactions and fusion (28), with mutations to the specific HN residues 89, 90, and 94 producing decreased F interactions (32). A region in the PIV5/SV41/human PIV2 (hPIV2) HN head also appears to play a role in keeping the HN stalk in a conformation favorable for F triggering though the head itself does not appear to be directly interacting with F (23, 55). Additionally, complementation studies using MeV and canine distemper virus (CDV) showed that residues 110 to 114 in the MeV H stalk allow specificity for CDV F (56). The general consensus appears to be that the stalk of HN/H/G is predominantly responsible for interactions with the respective F proteins. In spite of all of these studies, it remains unclear for most paramyxoviruses (i) what overall regions of the two glycoproteins interact, (ii) what roles each of the interactive regions plays, and (iii) how these interactions change during the membrane fusion process.

While for most paramyxoviruses it appears that the stalk of G/H/HN alone interacts with F, here we showed for NiV G that both the head and stalk regions interact with F (Fig. 4). This finding was suggested by Zhu et al. based on the observation that disrupting the NiV G stalk domain through the addition of N-glycans disrupted fusion but not G-F interactions (29). Additionally, Mirza et al. showed that a chimera expressing an HN stalk with an NiV G head still interacted with NiV F (21). Here, our data, combined with data of Zhu et al. and Mirza et al., provide clear evidence of a bidentate interaction for NiV and suggest that the henipavirus fusion-triggering process is relatively more complex than that of other paramyxoviral genera.

We previously showed that the NiV G head undergoes a series of receptor-induced conformational changes necessary for F triggering. First, a conformational change occurs near the receptor binding site, followed by a conformational change in the base of the head, resulting in exposure of a C-terminal domain of the stalk that triggers fusion. Mutants unable to make these conformational changes are fusion deficient, highlighting how essential these conformational changes are for the fusion-triggering process (8, 49). It is possible that the G head interacts with F while it undergoes its conformational changes, followed by the stalk exposure and G stalk-F interaction (Fig. 7A to C). Alternatively, it is possible that both the G head and stalk regions interact simultaneously with F prior to G receptor binding (Fig. 7B to D). We proposed that the G head helps stabilize the stalk, preventing it from premature triggering (8), and the reverse has been proposed as well, i.e., that the stalk helps keep the head in a prereceptor-bound conformation (29, 49). Additionally, the C-terminal portion of the stalk has been proposed as a linker region, transmitting a signal between the receptor binding head and the stalk (9, 15, 57, 58). Similar models have been proposed for other paramyxoviruses such as MeV (59, 60). The specifics of the interactions that occur between the G head, G stalk, and F, therefore, are quite complex and warrant further investigation.

FIG 7.

Models for G-F interactions leading to F triggering. (A) Prior to ephrinB2 (blue) receptor binding, the head region of NiV G (purple) interacts (represented by red lines) with NiV F (green) while the C-terminal domain of the stalk is relatively covered. (B) Upon receptor binding, the G head undergoes conformational changes that result in exposure of the C-terminal portion of the stalk (yellow), resulting in the G-F interactions switching from the G head to the G stalk. (C) Interaction of the G stalk C-terminal domain with F triggers F to undergo its own conformational changes, which execute virus-host cell membrane fusion. (D) Alternatively, both the head and N-terminal stalk regions interact with F prior to receptor binding, after which conformational changes in G result in exposure of the G stalk C-terminal domain, which triggers F and executes membrane fusion as shown in panels B and C. For simplicity, ephrinB2, NiV G, and NiV F are represented as monomers.

Using a soluble F construct, we showed that the ectodomain of NiV F is sufficient for interacting with NiV G, a finding similar to previous results with MeV (39). Furthermore, the soluble NiV F mutant is capable of interacting with G₁₆₇, a headless G mutant, further supporting the observation that the head region of G is not required for G-F interactions to occur. These results were confirmed by co-IP (Fig. 5). Importantly, although the headless G₁₆₇ and soluble F_tri mutants are capable of interacting, they were incapable of executing cell-cell fusion (data not shown), indicating that the minimal interactions of these two glycoprotein mutants are not sufficient for rendering membrane fusion. These data also indicate that the transmembrane and cytoplasmic regions of F play a role(s) in membrane fusion. Since most elements necessary for F triggering are likely in the ectodomain and thus included in the F_tri mutant protein, our data also suggest that the transmembrane and cytoplasmic regions of F may play roles late in the fusion process, as in fusion pore formation or expansion, as previously suggested by analysis of cytoplasmic tail NiV F mutants (33, 43).

It is interesting that while all of our soluble G constructs demonstrated an interaction with NiV F, none was able to cause cell-cell fusion (Fig. 4). This suggests (i) that anchoring of NiV G to the membrane is necessary for fusion triggering and/or (ii) that the transmembrane domain and/or cytoplasmic tail of G is involved in the fusion process, possibly through signal transduction. These results also corroborate that G-F interactions are not sufficient for fusion activity.

The majority of the deletion mutants we created were not expressed at the cell surface (Table 2), suggesting that the NiV glycoproteins are extremely sensitive to even small deletions. These proteins are extremely reliant on proper folding and conformational change capabilities (2, 8), and even single amino acid changes can result in lack of expression or altered fusion capabilities (8). While the majority of our deletion mutants were not expressed, we were able to identify three regions of NiV G not necessary for G-F interactions to occur (Fig. 6). NiV G residues 196 to 211 and residues 372 to 393 have been previously characterized as regions in the NiV G head important for fusion triggering (8), but their roles in G-F interactions were previously not determined. It is interesting that though these regions appear to be important for G conformational changes and thus proper F triggering, they do not appear to be necessary for the interactions between these glycoproteins.

Studies of membrane protein-protein interactions often rely on co-IP assays. Though widely used, a co-IP assay can be inconclusive for membrane proteins as it relies on the formation of micelles. Since within each micelle there can be numerous copies of both proteins present, the micelle itself can act as a linker to yield misleading results. Additionally, co-IP uses detergents, which can denature the proteins, potentially further altering protein-protein interactions. Co-IP studies for MeV have eliminated the micelle issue by using chemical cross-linking of the surface glycoproteins prior to protein lysis, followed by co-IP with harsher detergents (20, 38, 61). This method can be difficult to develop, however, due to the complexity and specificity of cross-linking agents necessary for successful cross-linking to occur. Efforts to develop a cross-linking co-IP assay for NiV in our laboratory were unsuccessful though it is possible that attempts using a more exhaustive list of cross-linking agents would be more fruitful. Chemical cross-linking, therefore, is preferable to traditional co-IP but is not possible for every protein-protein complex. Additionally, chemical cross-linking still relies on the use of detergents and thus does not eliminate the potential protein-denaturing caveat.

An additional assay commonly used to study protein-protein interactions is bimolecular fluorescence complementation (BiFC). This method uses a fluorescent protein that is split into two fragments that individually do not fluoresce. These fragments are then attached to two different proteins of interest. If the two proteins interact, the two fragments are physically brought together, resulting in a fluorescent signal that can be measured and quantified. Though this method has worked effectively for many different applications, it also has downsides. Background fluorescence signals can be high as self-assembly of the fluorescent protein fragments can occur independently of protein-protein interactions, resulting in false fluorescence signals. Steric hindrance can prevent the fluorescent fragments from interacting, and adding the fragments may disrupt the structural or functional integrity of the proteins of interest. Additionally, the type of fluorescent protein, along with the proper attachment site, must be determined empirically, thus making BiFC a potentially time-consuming method (62, 63).

Other common methods for studying protein-protein interactions are fluorescence colocalization and cryo-electron microscopy. Both of these methods can visualize that proteins are in the same vicinity, but they currently lack the resolution to definitively show that an interaction actually occurs. An ideal method for determining G and F interactions would be crystallization. To date, however, efforts to cocrystallize HN/H/G and F have been unsuccessful for the paramyxoviruses. Thus, there are many methods currently being used to investigate protein-protein interactions, but each method has its pitfalls. Our flow-cytometric strategy complements existing approaches and adds to the small arsenal available while avoiding many of the disadvantages described for other methods.

Though a qualitative interaction assay is very useful for demonstrating membrane protein interactions, ideally the new flow-cytometric assay would be quantitative. Attempts with previously characterized point mutants in the F cytoplasmic tail (33), as well as N-glycan mutants in both G (35) and F (34), however, failed to demonstrate a correlation between flow-cytometric values and previously determined F/G avidities for these mutants, even for mutants shown to have significantly decreased interactions via co-IP (data not shown). It is possible that with future troubleshooting this flow-cytometric strategy could be more quantitative, but there are several challenges to be overcome. More stringent buffer and wash conditions may help, but in order to keep cells alive during antibody staining, these options are limited. Additionally, the relatively receptor-negative PK13 cells have a transfection efficiency of only 20 to 40%, so using another cell type may be more effective. However, ephrinB2/B3 receptor-negative cell type options are limited. Attempts with both HEK 293T cells and Chinese hamster ovary (CHO) cells were unsuccessful as background binding was too high (data not shown), while PK13 cells yielded the lowest background levels. Techniques such as the use of small interfering RNA (siRNA) or CRISPR/Cas9 editing to block the ephrinB2/B3 receptor in these cells or addition of exogenous ephrinB2/B3 ligand to block the receptor may be alternative approaches. It should also be noted that although we observed an increase in the percentage of F- and G-positive cells for both soluble G and soluble F constructs when their full-length partners were present, this increase was still relatively small compared to the overall number of transfected cells. It is possible that further optimization could provide a greater increase. It is also possible, however, that the transient nature of the G-F interaction and the need of trapping it prior to G-F dissociation may explain the low sensitivity of the assay and the low yield of cells positive for both F and G. In addition, a caveat of the flow-cytometric assay, particularly when the assay is used singly, is that only one of the two membrane glycoproteins is full-length, making it somewhat uncertain that the interactions seen are entirely true. Thus, we must still rely on a combination of approaches to yield definitive conclusions.

In summary, we have developed a new strategy for studying membrane protein interactions without the use of detergents and used this strategy to demonstrate a novel bidentate interaction whereby both the NiV G head and stalk regions interact with NiV F, a finding so far uniquely demonstrated for NiV among the paramyxoviruses though it is possible that this bidentate interaction occurs in other paramyxoviruses as well, particularly in other henipaviruses. Additionally, we showed that the NiV F ectodomain is sufficient for G-F interactions, determined the regions of NiV G not necessary for G-F interactions to occur, and demonstrated the vulnerability of both G and F to alterations. Summarizing, G-F interactions are required, but not sufficient, for membrane fusion to occur. A combination of G-F interactions, F-triggering domains, and cell-anchoring elements must coexist for membrane fusion to ensue. How the nature of the bidentate interaction changes during the membrane fusion process will be a fascinating direction for future study. These findings add to our understanding of the paramyxovirus membrane fusion process, provide a new G-F interaction model within this virus family, and provide a new tool for future studies of viral and nonviral transmembrane protein-protein interactions.