|Home | About | Journals | Submit | Contact Us | Français|
The spatial organization of metastable paramyxovirus fusion (F) and attachment glycoprotein hetero-oligomers is largely unknown. To further elucidate the organization of functional fusion complexes of measles virus (MeV), an archetype of the paramyxovirus family, we subjected central predictions of alternative docking models to experimental testing using three distinct approaches. Carbohydrate shielding through engineered N-glycans indicates close proximity of a membrane-distal, but not membrane-proximal, section of the MeV attachment (H) protein stalk domain to F. Directed mutagenesis of this section identified residues 111, 114, and 118 as modulators of avidity of glycoprotein interactions and determinants of F triggering. Stalk-length variation through deletion or insertion of HR elements at positions flanking this section demonstrates that the location of the stalk segment containing these residues cannot be altered in functional fusion complexes. In contrast, increasing the distance between the H head domains harboring the receptor binding sites and this section through insertion of structurally rigid α-helical domains with a pitch of up to approximately 75 Å downstream of stalk position 118 partially maintains functionality in transient expression assays and supports efficient growth of recombinant virions. In aggregate, these findings argue against specific protein-protein contacts between the H head and F head domains but instead support a docking model that is characterized by short-range contacts between the prefusion F head and the attachment protein stalk, possibly involving H residues 111, 114, and 118, and extension of the head domain of the attachment protein above prefusion F.
Paramyxoviruses infect cells through fusion of the viral envelope with target cell membranes. For all members of the Paramyxovirinae subfamily, this involves the concerted action of two envelope glycoproteins, the fusion (F) and attachment (H, HN, or G, depending on the Paramyxovirinae genus) proteins. Both proteins feature short lumenal tails, a single transmembrane domain, and large ectodomains. The F protein, in type I orientation, forms homotrimers, while homodimers or homotetramers have been suggested as functional units for attachment proteins of different Paramyxovirinae subfamily members (7, 14, 28, 41, 49, 50, 66). For entry, upon receptor binding, the attachment protein is considered to initiate a series of conformational rearrangements in the metastable prefusion F protein (15, 77), which ultimately brings together transmembrane domains and fusion peptides and, thus, donor and target membranes (3, 32, 45, 53, 80).
Multiple studies have demonstrated that specific interactions between compatible F and attachment proteins of paramyxovirinae are imperative for the formation of functional fusion complexes (6, 29, 36, 42, 43, 56, 75). However, the molecular nature of these interactions and the spatial organization of functional glycoprotein hetero-oligomers remain largely unknown. Individual ectodomain and partial ectodomain crystal structures have been obtained for different paramyxovirus F (13, 76, 77) and attachment (8, 14, 17, 28, 35, 79) proteins, respectively. For F, a stabilized human parainfluenza virus type 5 (HPIV5) ectodomain that is believed to represent a prefusion conformation folds into a globular head structure that is attached to the transmembrane domains through a helical stalk consisting of the membrane-proximal heptad repeat B (HR-B) domains (77). For the attachment protein, a globular head that harbors the receptor binding sites is considered to be connected to the transmembrane region through extended stalk domains (34, 78). Crystal structures of isolated head domains have been solved for several paramyxovirus attachment proteins, including measles virus (MeV) H, and reveal the six-blade propeller fold typical of sialidase structures (8, 14, 17, 28, 79). However, morbilliviruses recognize proteinaceous receptors (for MeV, the regulator of complement activation [CD46] and/or signaling lymphocytic activation molecule [SLAM], depending on the virus strain) (21, 40, 46, 51, 64, 65). X-ray data do not extend to the stalk domains, but circular dichroism analysis (78) and structure predictions (36, 78) support an α-helical coiled-coil configuration of the stalk.
The nature of individual residues that engage in specific intermolecular interactions between glycoproteins of paramyxovirinae prior to refolding has been studied most extensively for the attachment protein. The stalk domains of several paramyxovirus HN proteins have been implicated in mediating specificity for their homotypic F proteins (18, 20, 43, 63, 70, 72). We have found that this extends to MeV and canine distemper virus H and, thus, to paramyxovirinae recognizing proteinaceous receptors (36), supporting the general hypothesis that F-interacting residues may reside in the stalk region of the attachment protein (30, 78).
Considerably less information concerning the nature of F microdomains that mediate attachment protein specificity is available. Among the few exceptions are peptides derived from Newcastle disease virus (NDV) and Sendai virus F HR-B domains, which interact with soluble variants of the respective HN proteins in vitro (25, 67). Multiple domains have been suggested to mediate specificity of HPIV2 F for its HN (69). However, a conclusive N-glycan shielding study (43) and structural information (77, 78) argue against direct contacts between NDV F HR-B domains and HN in native glycoprotein complexes. Thus, the role of individual HPIV2 F residues in HN binding is unclear (25, 43).
Building on the observation that MeV H is able to engage in productive heterotypic interactions with F proteins derived from some but not all isolates of closely related canine distemper virus, we have recently identified residues in morbillivirus F (MeV F residue 121) and H (H stalk residues 110 to 114) that interdependently contribute to physical MeV glycoprotein interaction and F triggering for fusion (36). While these residues could mediate reciprocal glycoprotein specificity through long-range effects, molecular modeling of the MeV H stalk in an α-helical conformation has posited F residue 121 at the same level above the viral envelope as H residues 110 to 114, making direct contacts structurally conceivable (36). This spatial organization of functional fusion complexes furthermore provides a comprehensive explanation for previous demonstrations of a specific role for attachment protein stalk domains of paramyxovirinae in functional and physical interactions with F (18, 43, 63, 70, 72). However, this “staggered-head” model mandates positioning the globular head of the attachment protein above the prefusion F trimer (36), as opposed to a suggested “parallel-head” alignment of the glycoproteins (31, 47). The latter is mostly based on transmission electron microscopy micrographs of viral particles apparently showing glycoprotein spikes of equal length (33). Unfortunately, these images lack the resolution for an identification of the molecular nature of the spikes (attachment or F protein) or the distinguishing between densely packaged H and F head domains of different heights and laterally aligned head domains. Indeed, a recent single-particle reconstruction based on cryo-electron microscopy images of HPIV5 particles revealed that defined spikes correspond to F in a postfusion conformation, which was interpreted as a product of possible premature F refolding (38). These two-dimensional images of heavy-metal-stained particles did not reveal F spikes in a prefusion conformation. Rather, a dense surface layer was considered to correspond to prefusion glycoprotein hetero-oligomers (38). In addition to further-advanced image reconstructions, biochemical assessment of alternative docking modes is imperative for the elucidation of the organization of functional fusion complexes of paramyxovirinae.
In this study, we subjected central predictions of the hypothetical alignment models to experimental analysis. By employing carbohydrate shielding, directed mutagenesis, and variation of the length of the H stalk domain, we examined the proximity of different regions of the H stalk to F, probed a role of individual residues around the previously identified H stalk section from positions 110 to 114 in the formation of functional fusion complexes, tested the effect of varying the length of the H stalk membrane proximal and membrane distal to this section, and explored the general possibility of whether specific contacts between the prefusion F and H head domains are required for F triggering. Experimental data were interpreted in the light of a working model of MeV glycoprotein hetero-oligomers prior to receptor binding.
All cell lines were maintained at 37°C and 5% CO2 in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Vero (African green monkey kidney epithelial) cells (ATCC CCL-81) stably expressing human SLAM (Vero-SLAM cells) (52), and baby hamster kidney (BHK-21) cells stably expressing T7 polymerase (BSR-T7/5 [BHK-T7] cells) (10) were incubated at every third passage in the presence of G-418 (Geneticin) at a concentration of 100 μg/ml. Lipofectamine 2000 (Invitrogen) was used for all cell transfections. To prepare MeV stocks, Vero-SLAM cells were infected at a multiplicity of infection (MOI) of 0.001 PFU/cell and incubated at 37°C. Cells were scraped in Opti-MEM (Invitrogen), virus was released by two freeze-thaw cycles, and titers were determined by 50% tissue culture infective dose (TCID50) titration according to the Spearman-Karber method (59) as described previously (56). To prepare stocks of modified vaccinia virus Ankara expressing T7 polymerase (60), DF-1 (chicken embryo fibroblast) cells (ATCC CRL-12203) were infected at an MOI of 1.0 PFU/cell, and cell-associated viral particles were harvested 40 h postinfection.
Base vectors for all transient MeV glycoprotein expression experiments and directed-mutagenesis approaches were plasmids pCG-F and pCG-H, encoding glycoproteins derived from the MeV Edmonston strain under the control of the constitutive cytomegalovirus promoter (12). For point mutations, heptamer insertions, or heptamer deletions, directed mutagenesis following the QuikChange protocol (Stratagene) was employed using appropriate primers (individual sequences are available upon request). All changes were confirmed by DNA sequencing, and expression of mutagenized proteins was verified by immunoblotting. To generate different-length tandem repeats of the H stalk downstream of stalk nucleotide 351, a staggered-priming mutagenesis strategy was developed. Using a pCG-H variant harboring an additional copy of stalk residues 84 to 90 at residue 118 (H-1187x) as the template, mutagenesis was carried out with primer 5′-cactgacctagtgaaattcATAGAACACCAAGTGAAAGATatcgagcatcaggtcaagg, which consists of H stalk nucleotides 333 to 351 (boldface), the HR insertion (uppercase letters), and, for the H-1187x template, stalk nucleotides 250 to 268 and 352 to 370 (boldface and underlined), corresponding to amino acids 84 to 90 and the additional copy of this HR inserted at residue 118, respectively. This reaction efficiently (approximately 25 to 50% of clones analyzed) yielded H constructs with extended-length tandem stalk insertions at stalk position 118 (see Fig. Fig.6B6B for individual sequences). To transfer an H variant harboring a 41-residue stalk elongation (H-11841x) into a full-length cDNA copy of the MeV Edmonston genome, a PacI/SpeI fragment containing the H-encoding open reading frame (ORF) was replaced in p(+)MeV-NSe (58) with this variant, resulting in plasmid p(+)MeV-H 11841x. An additional stop codon was added to the H-11841x ORF to render the resulting genome compatible to the rule-of-six that applies to many members of the paramyxovirus family, including MeV (26). An enhanced green fluorescent protein (eGFP)-encoding ORF was furthermore inserted as an additional transcription unit upstream of the nucleocapsid-encoding ORF through replacement of a PacI/NotI fragment with that derived from p(+)MeV-eGFP (22), resulting in plasmid p(+)MeV-eGFP-H 11841x.
Vero cells were transfected with 2 μg of plasmid DNA encoding MeV H constructs as indicated. After being washed in cold phosphate-buffered saline (PBS), cells were incubated in PBS with 0.5 mg/ml sulfosuccinimidyl-2-(biotinamido)ethyl-1,3-dithiopropionate (Pierce) for 20 min at 4°C, followed by washing and quenching for 10 min at 4°C in Dulbecco's modified Eagle's medium. Samples were lysed in immunoprecipitation buffer, and lysates were cleared by centrifugation for 45 min at 20,000 × g and 4°C. Biotinylated proteins were adsorbed to immobilized streptavidin for 120 min at 4°C, washed two times in buffer A (100 mM Tris, pH 7.6, 500 mM lithium chloride, 0.1% Triton X-100) and then buffer B (20 mM HEPES, pH 7.2, 2 mM EGTA, 10 mM magnesium chloride, 0.1% Triton X-100), and incubated in urea buffer (200 mM Tris, pH 6.8, 8 M urea, 5% sodium dodecyl sulfate [SDS], 0.1 mM EDTA, 0.03% bromphenol blue, 1.5% dithiothreitol) for 30 min at 50°C. Samples were then subjected to SDS-polyacrylamide gel electrophoresis (PAGE) and immunoblotting using a polyclonal antiserum directed against the H cytosolic tail (54). For densitometric quantification, immunoblots were decorated with anti-rabbit IRDye 800CW conjugate (LI-COR) and analyzed using an Odyssey infrared imager and Odyssey software (LI-COR). Amounts of surface steady-state levels of cellular transferrin receptor that were analyzed in parallel using a monoclonal anti-human transferrin antiserum (Zymed) served as an internal standard.
For PNGase F-catalyzed deglycosylation, H variants were expressed and subjected to surface biotinylation as described above. Following streptavidin pull-down, samples were split into two equal aliquots, proteins were denatured (0.5% SDS, 400 mM dithiothreitol; 95°C for 10 min), and aliquots were subjected to PNGase F deglycosylation (50 mM sodium phosphate, pH 7.5, 1% NP-40; 1 h at 37°C) or mock treatment. All samples were then analyzed by SDS-PAGE and immunoblotting using antiserum directed against the cytosolic H tail.
For expression of a soluble SLAM-murine immunoglobulin G (sSLAM-mIgG) fusion protein, a cDNA sequence encoding the first 233 amino acids of the SLAM ectodomain was isolated through reverse transcription-PCR from Vero-SLAM cells, and the sequence was confirmed. This fragment was subcloned in frame to the 5′ end of an ORF encoding murine IgG heavy chains that is present in the RCASBP(A) vector, a replication-competent retrovirus vector derived from avian leukosis virus (23). After transfection of the RCASBP(A) construct into permissive DF-1 cells, sSLAM-mIgG was purified from culture supernatants (2.5 liters, total volume) through affinity chromatography using immobilized protein G, ultrafiltration, and dialysis against PBS. The final purity of sSLAM-mIgG determined by SDS-PAGE and Coomassie blue staining was ≥90%, and concentrations of purified material were adjusted to 1 mg/ml.
To determine the ability of MeV H variants to bind the SLAM receptor, SLAM- and CD46-negative CHO cells were transfected in a six-well plate format with MeV H constructs (5 μg) and an eGFP expression plasmid (1 μg) as a marker for transfection. Thirty-six hours posttransfection, cells were lifted off and washed twice in PBS supplemented with 3% fetal calf serum. Cell populations were split into two equal aliquots and incubated with either sSLAM-mIgG or a monoclonal antibody mix directed against epitopes in the MeV H ectodomain (Millipore) for 1 h on ice. Following decoration of cells with anti-mouse Fc-specific allophycocyanin conjugate (Jackson Immunoresearch) for 30 min on ice, samples were analyzed using a FACSCanto II cytometer and FlowJo software. Fluorescence intensity was determined for the GFP-positive cell population.
To quantify fusion activity, an effector Vero-SLAM cell population (3 × 105 cells/well) was cotransfected with 2 μg each of H and F expression plasmids, and target Vero-SLAM cells (2 × 105 cells/well) were transfected with 2 μg of the reporter plasmid encoding firefly luciferase under the control of the T7 promoter. Single transfections of plasmids encoding MeV F served as controls. Two hours posttransfection, the effector cells were infected with modified vaccinia virus Ankara expressing T7 polymerase at an MOI of 1.0 PFU/cell. Following incubation for 16 h at 37°C, in the case of effector cells in the presence of 100 μM fusion inhibitory peptide (Bachem), target cells were detached and overlaid on washed effector cells at a 1:1 ratio and incubated at 37°C. Five hours postoverlay, cells were lysed using Bright Glo lysis buffer (Promega), and the luciferase activity was determined using a luminescence counter (PerkinElmer) and the Britelite reporter gene assay system (PerkinElmer). The instrument's arbitrary values were analyzed by subtracting the relative background provided by values of the controls, and these values were normalized against the reference constructs indicated in the figure legends. On average, background values were <1.5% of the values obtained for reference constructs.
Vero-SLAM cells were transfected with 4 μg each of plasmid DNA encoding MeV F and H variants as specified for the individual experiments and incubated in the presence of 100 μM fusion inhibitory peptide. At 36 h posttransfection, cells were washed three times with cold PBS and treated with DTSSP (1 mM final concentration in PBS) for two hours at 4°C, followed by the addition of Tris, pH 7.5, to a final concentration of 20 mM for quenching (15 min, 4°C). Cells were then lysed in radioimmunoprecipitation assay buffer (1% sodium deoxycholate, 1% NP-40, 150 mM NaCl, 50 mM Tris-Cl, pH 7.2, 10 mM EDTA, 50 mM NaF, 0.05% SDS, protease inhibitors [Roche], 1 mM phenylmethylsulfonyl fluoride), and cleared lysates (20,000 × g; 30 min at 4°C) were incubated with specific antibodies directed against epitopes in the MeV H ectodomain (Millipore) at 4°C. After precipitation with immobilized protein G at 4°C, samples were washed two times each in buffers A and B as described above, eluted in urea buffer, and subjected to SDS-PAGE and immunoblotting. Coprecipitated F was visualized in a ChemiDoc XRS digital imaging system (Bio-Rad) using a polyclonal antiserum against the F cytosolic tail (37) and rabbit IgG light chain-specific horseradish peroxidase conjugate. For densitometry, signals were quantified using the QuantityOne software package (Bio-Rad).
Recombinant MeV was generated as described previously (58) with the following modifications: BHK-T7 cells were cotransfected by calcium phosphate precipitation with a plasmid carrying a cDNA copy of the recombinant MeV genome and plasmids encoding the MeV L, N, or P protein. All constructs were under the control of the T7 promoter. Helper cells were overlaid on Vero-SLAM cells 76 h posttransfection, and emerging virus particles were transferred to fresh Vero-SLAM cells. The integrity of recombinant MeV particles was then confirmed by reverse transcription-PCR and DNA sequencing of the modified area using Superscript II reverse transcriptase and random hexamer primers for first-strand synthesis.
For multistep growth curves, cells were infected in a six-well-plate format with recombinant MeV Edmonston (recMeV-Edm) variants as specified at an MOI of 0.001 PFU/cell. To ensure equal MOIs for each variant analyzed, all input stocks were prediluted to approximately 104 TCID50/ml in Opti-MEM (Invitrogen) prior to infection, and titers were reconfirmed by TCID50 titration. Sixty minutes postinfection, inocula were replaced with fresh growth medium. Cell-associated viral particles were harvested every 12 h as described above and subjected to TCID50 titration. For documentation of the viral cytopathic effect (CPE), plates were photographed at a magnification of ×200.
Previously described structural models of MeV F and H proteins were used in this analysis (36, 57). Briefly, a homology model of the MeV F protein was built on the basis of the coordinates reported for the prefusion HPIV5 F X-ray crystal structure (77). The MeV H structural model was built by connecting the X-ray coordinates available for residues 154 to 607 forming the head domain (14, 28) to a coiled-coil model of the MeV H stalk (residues 58 to 153), which was generated by mapping these residues on the α-helical scaffold of cortexillin I (11) as described previously (36). In the present study, a slightly different mapping of the stalk residues was used to better accommodate the stutter in the HR sequence at residues I99 and L105. A single-residue gap was introduced in the alignment after P108, placing this residue in “a” position of the helical wheel. Proline residues 94 and 127 are also expected to introduce kinks in the helix structure in agreement with electron microphotographs of the related HPIV5 HN (78). Due to the lack of native X-ray information, however, these are not represented in the schematic models. For the modeling of N-glycan additions, a sugar with the following structure was prepared in three dimensions by using Maestro software (Schrodinger, LLC, New York, NY): asparagine-GlcNAc-β(1,4)-GlcNAc-β(1,4)-Man[-α(1,3)-Man-β(1,2)-GlcNAc-β(1,4)-Gal-α(2,6)-NeuNAc] [-α(1,6)-Man-β(1,2)-GlcNAc-β(1,4)-Gal-α(2,6)-NeuNAc]. For a reasonable low-energy conformation of the carbohydrate, a Monte Carlo conformational search of 1,000 steps was performed with MacroModel 9.5 software (Schrodinger, LLC, New York, NY) using the OPLS2005 force field with generalized Born/surface area water solvation. The low-energy conformation was attached to the proteins by superimposing the backbone atoms of asparagine to the target glycosylation position on the MeV H stalk. Structural models of MeV H stalk deletion mutants were generated by removing HR regions from the stalk model, save for one flanking residue on either side of the deletion (residues 85 to 89 for HR 84 to 90 and residues 119 to 123 for HR 118 to 124). The flanking residues were then used to superimpose the backbone atoms of the downstream structure to the upstream structure, generating one construct (H-Δ84-90) with residues 58 to 83 attached to residues 91 to 607 and another construct (H-Δ118-124) with residues 58 to 117 attached to residues 125 to 607. MeV H stalk insertions were generated by duplicating the desired section of the stalk plus one flanking residue on either side for superimposing (residues 83 to 91, 117 to 125, and 83 to 118). These duplicate sections were inserted into the downstream structure and aligned by superimposing the backbone atoms of the overlapping residues. The upstream structure was then superimposed in the same manner to the C-terminal end of the newly extended stalk, generating the constructs H-847x, H-1187x, and H-11841x.
With the overarching goal of further elucidating the spatial organization of paramyxovirus envelope glycoproteins in a prefusion state, we subjected key predictions of hypothetical docking models to experimental testing.
Independently of the lateral arrangement of the head domains, schematic juxtapositions of F and H structures predict that membrane-proximal regions of the H stalk are not in direct contact with the F trimer. For membrane-distal stalk sections, however, the staggered-head (Fig. (Fig.1A,1A, left panel), but not the parallel-head model (Fig. (Fig.1A,1A, right panel), implies close proximity of the prefusion F head domain. Carbohydrate shielding through insertion of additional N-glycosylation sites is reportedly capable of examining the contribution of protein domains to activity (24, 68) and has been used in a conclusive study to test candidate contact zones between NDV HN and F (43). We employed this strategy to probe the proximity of F from the H stalk domain at a membrane-proximal and membrane-distal position. Since the exact rotational orientation of the H stalk is unknown, N-glycosylation sites were introduced separately for each position at three consecutive residues: residues 70 to 72 were chosen as targets for membrane-proximal N-glycans guided by the prediction of the staggered-head model that carbohydrates at these positions will not impair the interaction of H with F, and residues 110 to 112 were chosen for membrane-distal N-glycans based on our previous identification of the role of MeV H section 110 to 114 in mediating F specificity (36).
Of the membrane-proximal set, only the engineered site at position 72 was recognized efficiently without affecting intracellular transport competence, as demonstrated by comparing the electrophoretic mobilities of these H variants with that of standard H before and after PNGase F deglycosylation of cell surface-exposed material (Fig. (Fig.1B).1B). Subsequent full characterization was therefore limited to the corresponding variant, H-72-Nglyc. Equivalent analysis of the electrophoretic mobility of the membrane-distal set confirmed glycosylation at each of the targeted positions, 110, 111, and 112 (Fig. (Fig.1C).1C). Recognition of the engineered glycosylation sites was highly efficient for all four constructs, since immunoblotting revealed homotypic mobility patterns that lacked subspecies with Edmonston H (H-Edm)-like mobility prior to PNGase F treatment.
Assessments of plasma membrane steady-state levels through surface biotinylation and of SLAM binding through flow cytometry demonstrated essentially unchanged intracellular transport competence and ability to interact with SLAM receptor for all four H variants with engineered N-glycans (Fig. (Fig.2A).2A). To establish a receptor binding assay, we generated an sSLAM-mIgG fusion protein that specifically blocks MeV entry, confirming its ability to bind MeV H (Fig. (Fig.2B).2B). Figure Figure2C2C exemplifies that, following binding of this sSLAM-IgG construct to H-expressing cells by flow cytometry, the ratios of SLAM-specific mean fluorescence intensity to H-specific mean fluorescence intensity provide an assessment of receptor (SLAM) binding competence. For a control, an I194S mutant (H-I194S) that reportedly is unable to bind to SLAM was assessed (48). This construct returned a SLAM-to-H signal ratio of 0.03, compared to 1.07 obtained for standard H (Fig. (Fig.2C),2C), confirming specificity of the assay.
Quantification of cell-to-cell fusion using a luciferase reporter-based content mixing assay demonstrated that fusion activity and, thus, F triggering were virtually unaffected by the additional N-glycan at H stalk position 72 (Fig. (Fig.2A).2A). In contrast, N-glycans at position 110, 111, or 112 completely blocked F triggering (Fig. (Fig.2A),2A), as implied by Fig. Fig.1A1A.
To assess whether this lack of functionality extends to altered physical interaction of the glycoproteins, we further advanced a coimmunoprecipitation (co-IP) assay that we have previously developed for MeV H and F (36, 55, 56). The general strategy of coprecipitating one envelope glycoprotein with antibodies directed against the other after lysis of the producer cells has been applied to multiple members of the paramyxovirus family (2, 5, 19, 44, 75). However, in this setup, the possibility that removal of the stabilizing lipid bilayer induces conformational changes in the metastable prefusion F trimer (76) prior to co-IP, potentially distorting results, cannot be excluded. To assess envelope glycoprotein hetero-oligomers in situ prior to membrane lysis, we examined membrane-impermeable chemical cross-linkers. Covalently linking surface-expressed MeV glycoprotein hetero-oligomers with membrane-impermeable, reducible DTSSP prior to membrane extraction, immunoprecipitation of H under stringent conditions, and detection of coprecipitated F allowed specific analysis of plasma membrane-embedded envelope glycoprotein complexes (Fig. (Fig.2D).2D). Having established this assay, we found that N-glycans at stalk position 111 (representative of N-glycan-mediated shielding of membrane-distal stalk positions) essentially block the formation of functional fusion complexes (<10% of those observed for unchanged H and F). This observation underscores the concept that DTSSP-mediated MeV glycoprotein cross-linking does not generate artificial H and F complexes. In contrast, N-glycans at position 72 still allow glycoprotein interaction, although at a reduced level compared to standard complexes (Fig. (Fig.2E2E).
These data demonstrate that insertion of additional N-glycans at a membrane-proximal or -distal position of the H stalk does not result in a loss of receptor binding capacity or intracellular protein retention through gross H misfolding. The block of F triggering and glycoprotein hetero-oligomerization by membrane-distal, but not membrane-proximal, N-glycans indicates close proximity of this H stalk section to F.
To assess whether a specific subset of residues in the previously identified linear H stalk segment from positions 110 to 114 determines effective F triggering, we next subjected this domain to alanine-scanning mutagenesis. In a first round, all five residues or only the first three or last two residues were mutagenized. The surface expression and SLAM binding ability of these H variants were unchanged, but none was capable of triggering F in cell-to-cell fusion assays (Fig. (Fig.3).3). This extends the essential role of the segment in F triggering to homotypic MeV glycoprotein pairs and indicates that residues in both subsections examined individually (positions 110 to 112 and positions 113 and 114) are required for the formation of functional fusion complexes.
Subsequent single-residue mutagenesis revealed a complete loss of F triggering by H-F111A and a significant reduction by H-L114A, while changes at each of the other three positions (110, 112, and 113) did not affect fusion activity (≥85% of F triggering by standard H; Fig. Fig.4A).4A). We note that H-F111A and H-L114A showed essentially unaltered surface expression, while approximately 20% (H-T112A and H-D113A) to 30% (H-R110A) reductions were observed for the F triggering-competent mutants (Fig. (Fig.4A).4A). SLAM binding levels of all variants were equivalent to that of standard H (Fig. (Fig.4A4A).
A three-residue distance between individual positions determining F triggering supports the predicted helical character of the attachment protein stalk domains of paramyxovirinae (36, 43, 78), since an α-helical configuration positions residues 111 and 114 in immediate proximity on consecutive turns of the helix (Fig. (Fig.4B).4B). To explore whether neighboring residues on preceding or subsequent turns likewise contribute to F triggering, we extended the alanine mutagenesis to stalk positions 108, 117, 118, and 122, which are predicted to be located at the same face of the helix as residues 111 and 114 (Fig. (Fig.4B).4B). Of these, an I118A mutation likewise abolished fusion activity but did not reduce surface expression or SLAM binding capacity (Fig. (Fig.4A).4A). None of the other changes resulted in major reductions of F triggering or SLAM binding, although the proline-to-alanine change at position 108 caused an approximately 50% reduction in surface expression (Fig. (Fig.4A).4A). In addition to affecting F triggering, the individual mutation of residue 114 or, in particular, 111 or 118 to alanine impaired physical glycoprotein interaction at the plasma membrane (Fig. (Fig.4C4C).
Taken together, these data identify H stalk positions 111, 114, and 118, but not intermittent or flanking residues 108, 112, 113, 117, or 122, as determinants for the formation of functional fusion complexes between MeV H and F. This is consistent with an α-helical configuration of the stalk domains in situ, supporting previous results derived from secondary structure predictions and in vitro analysis of soluble attachment protein ectodomain fragments (36, 43, 78).
The N-glycan shielding and cross-linking results are compatible with close proximity between F and the H stalk section including residues 111, 114, and 118. The corresponding alignment of F and H (staggered-head model) shown in Fig. Fig.5A5A implies that shortening the stalk domain will likely disturb glycoprotein hetero-oligomerization due to steric interference of a lowered H head with F. In contrast to the parallel-head model, the staggered-head alignment furthermore predicts that stalk elongation distal, but not proximal, to residues 111/114/118 should be tolerated, since the relative positions of contacting domains proximal to residue 118 would likely be preserved (Fig. (Fig.5A).5A). Deleting or inserting complete HR domains is anticipated to minimally compromise the predicted overall helical conformation of the stalk while altering its length by approximately 11 Å per HR domain.
To test these predictions, we identified two HRs (residues 84 to 90 and residues 118 to 124) (Fig. (Fig.5A;5A; for amino acid sequences, see Fig. Fig.6B)6B) that flank the 111/114/118 section and are distinct from H residues 91 to 105. Introducing mutations in the latter section was avoided, since this reportedly affects F triggering either through long-range effects or a shared role of this segment and other segments in the formation of functional fusion complexes (16). Our approach furthermore preserved the hydrophobic nature of residue 84, which was likewise identified by Corey and Iorio in the same study to be required for F triggering (16).
Four constructs were generated: constructs with a deletion of the HR from positions 84 to 90 [H-Δ(84-90)] or 118 to 124 [H-Δ(118-124)] or an insertion of an additional copy of the HR from positions 118 to 124 at residue 84 (H-847x) or the HR from positions 84 to 90 at residue 118 (H-1187x). The isoleucine residue at position 118 was conserved in all of these, with the exception of H-Δ(118-124), in which, however, an I118L change maintained the hydrophobic nature at position 118. All H variants reached the surface, although plasma membrane steady-state levels of H-1187x and H-Δ(84-90) were reduced by approximately 30 and 50%, respectively (Fig. (Fig.5B).5B). All four constructs were fully capable of SLAM binding (Fig. (Fig.5B).5B). However, microscopic analysis (data not shown) and quantitative fusion assays revealed that both HR deletions and the membrane-proximal HR insertion completely abrogated F triggering. In contrast, stalk elongation membrane-distal of the 111/114/118 section still allowed the formation of functional fusion complexes, since coexpression of H-1187x with F resulted in approximately 50% fusion activity of standard H and F (Fig. (Fig.5B).5B). Likewise, only stalk elongation downstream of 111/114 (H-1187x) was compatible with glycoprotein hetero-oligomerization, whereas membrane-proximal insertion (H-847x) and HR deletions at either position essentially abolished glycoprotein interactions (Fig. (Fig.5C5C).
These results demonstrate that the overall stalk lengths can be increased, but the distance between the transmembrane domain and stalk section from positions 91 to 118 must be preserved for effective hetero-oligomerization and F triggering.
F triggering by H-1187x suggests that H domains downstream of residue 118 do not engage in functional protein-protein contacts in prefusion hetero-oligomers, thus, further arguing against a parallel-head alignment and indicating a lack of functional contacts between the base of the H and the top of the F head in staggered alignment. Alternatively, structural flexibility of the carboxy-terminal third of the stalk (residues 118 to 154) could compensate for the single HR insertion of the H-1187x variant.
To probe these alternatives, we further extended the stalk domain membrane distal to position 118. Insertion of a second additional HR domain generated variant H-11814x, with a predicted added stalk length of approximately 22 Å. Compared to H-1187x, the second HR domain addition returned very similar surface expression (approximately 70%) of H-Edm, was fully capable of SLAM binding, and improved F triggering to approximately 80% of the standard H value in cell-to-cell fusion assays (Fig. (Fig.6A).6A). In parallel to our study of stalk extensions of defined length, we developed a staggered-priming mutagenesis strategy to create different-length tandem repeats of H stalk section 84 to 117, always inserted at position 118. This convenient approach yielded two additional H variants (sequences shown in Fig. Fig.6B)6B) that harbor 41-residue (H-11841x) and 55-residue (H-11855x) stalk elongations. The two constructs were fully capable of SLAM binding and showed plasma membrane steady-state levels of approximately 65% and 45% of standard H, respectively (Fig. (Fig.6A).6A). Remarkably, coexpression of F with H-11841x, featuring stalk extensions of nearly 50% of the original lengths (depicted in Fig. Fig.6C),6C), returned approximately 50% fusion activity of standard H. A drastic reduction in F triggering was found for H-11855x. However, microphotographs of cells coexpressing this variant and F (Fig. (Fig.6D)6D) and cell-to-cell fusion assays confirmed some residual activity, equivalent to approximately 10% of the standard H value (Fig. (Fig.6A).6A). As suggested by these findings, all three H variants are capable of physically engaging F. While an approximately 70% reduction of hetero-oligomer formation was observed for H-11855x, co-IP efficiencies of H-11814x and H-11841x with F were similar (nearly 80%) to that observed for H-Edm (Fig. (Fig.6E6E).
In aggregate, these findings demonstrate that membrane-distal insertion into the H stalk of a structurally rigid 41-amino-acid domain composed of consecutive HRs, which is equivalent to a pitch of approximately 75 Å in an α-helical configuration, is largely compatible with the formation of functional fusion complexes.
The remarkable nature of these findings raised the question of whether tolerance of the 41-residue insertion is limited to inducing cell-to-cell fusion in a transient glycoprotein expression setting or extends to sustaining viral entry and efficient production of infectious particles. To test this, we generated a cDNA copy of the MeV Edmonston genome that harbored the H-11841x gene in place of the standard H gene and, to facilitate detection of recombinant virions, an eGFP gene as an additional transcription unit.
recMeV-eGFP-(H-11841x) virions were recovered successfully and subjected to analysis of CPE in cell monolayers and growth in comparison to standard recMeV-eGFP. Figure Figure7A7A shows that cell-to-cell fusion was also the predominant CPE associated with the recMeV-eGFP-(H-11841x) variant. However, final syncytial sizes were reduced, and the incubation time until syncytial disintegration was increased compared to that for the recMeV-eGFP variant. These observations were corroborated by multiple-step viral growth curves. The recMeV-eGFP-(H-11841x) variant grew efficiently to final titers equivalent to those reached by the recMeV-eGFP variant (Fig. (Fig.7B).7B). However, presumably due to reduced lateral spread and, hence, less-rapid breakdown of syncytia in cell culture, recMeV-eGFP-(H-11841x) growth curves showed an extended plateau phase of high virus titers, as opposed to the sharp drop from the peak titer observed for the recMeV-eGFP variant (Fig. (Fig.7B7B).
These data confirm that our observations based on transient expression of H-11841x and F are fully transferable to the context of virus infection. Thus, they underscore that an H stalk elongation of approximately 50%, inserted membrane distal to residue 118, is compatible with productive interaction of H with the prefusion F trimer and the efficient formation of infectious particles.
With an increasing number of high-resolution X-ray structures of envelope glycoproteins of paramyxovirinae at hand, elucidating the spatial organization of functional fusion complexes in situ has emerged as a major task in better understanding the molecular mechanism of infection by paramyxovirinae. In this study, we subjected key predictions derived from hypothetical glycoprotein interaction models that were generated on the basis of crystallographic and biochemical data (14, 16, 28, 36, 43, 77) to experimental evaluation. In aggregate, our results suggest a spatial ectodomain organization in which the prefusion F head domain contacts a membrane-distal section of the H stalk. This requires positioning of the H head domain above prefusion F in a staggered-head alignment. Several lines of evidence support these conclusions.
First, N-glycan shielding at H stalk position 111 blocks both functional and physical glycoprotein interactions, suggesting close proximity of F to this section of the H stalk. While these results could alternatively derive from a long-range, indirect effect of the added glycans, efficient H-111-Nglyc surface expression, unchanged receptor binding capacity, and a rigid α-helical configuration of the stalk supported by secondary structure predictions (36, 43) and circular dichroism analysis of the related HPIV5 HN stalk (78) render this unlikely. Nearly unperturbed physical interaction and F triggering capacity of H-72-Nglyc underscore this conclusion. It is noteworthy in this context that additional N-glycosylation sites are efficiently recognized at three consecutive positions starting with residue 110 but only at position 72 of the more-membrane-proximal 70-to-72 triplet. This suggests that rotational flexibility within the postulated α-helical coiled coils is limited in proximity to the transmembrane domains but increases toward membrane-distal positions. Bulky N-glycans at any of the three membrane-distal positions and/or glycan-induced slight rotation of the stalk helix in the z axis may interrupt precise protein-protein contacts at the H and F interface.
Second, mutagenesis of individual residues around the previously identified H stalk section from positions 110 to 114 (36) has highlighted three residues (111, 114, and 118), predicted to be located on consecutive turns of the MeV H stalk α-helix as determinants of both effective F triggering and, in particular, in the case of residues 111 and 118, glycoprotein interaction. Previously, Corey and Iorio demonstrated that mutation of MeV H stalk residue I84, L92, I98, or I99 likewise affects F triggering (16). Based on very efficient coprecipitation of these mutants with F, they concluded that these residues likely contribute to F triggering through indirect effects or form one of several interaction domains. Since it is likely a requirement that the F trimer separates at some stage in fusion from the attachment protein for F refolding to proceed (73, 74), increased strength of hetero-oligomer interaction might be sufficient to reduce or block fusion. While indirect effects could likewise result when an alanine point mutation is inserted at position 111, 114, or 118, the block of F triggering by these H variants combined with a strong reduction of glycoprotein interactions in situ makes it alternatively conceivable that these residues engage in short-range functional contacts required for the initiation of F refolding.
Lastly, shortening the MeV H stalk by either membrane-proximal or membrane-distal residues 84 to 117 blocks both hetero-oligomerization and triggering. This is compatible with both the parallel (required protein-protein contacts are disrupted)- and staggered (steric constraints between F head and a lowered H head)-alignment hypotheses of glycoprotein head domains. Importantly, though, increasing the length of the H stalk membrane proximal to residue 84 by only ~11 Å disrupts hetero-oligomerization and F triggering, while an insertion of the same length membrane distal to residue 118 still allows the formation of functional fusion complexes. This demonstrates that the position of the section from positions 84 to 118 relative to the donor membrane, and thus relative to the F trimer, is critical for productive hetero-oligomerization. Tolerance to stalk elongation at position 118 renders specific protein-protein contacts between the H head and F unlikely, in particular, since structurally rigid insertions with a predicted pitch of approximately 75 Å in an α-helical configuration are compatible with the formation of functional fusion complexes in the context of both transient glycoprotein expression and recombinant virus particles. This finding furthermore supports the idea that H and F oligomers separate after successful initiation of F rearrangements, given that in the case of continued hetero-oligomer interaction, the 75-Å H extension would likely keep the target membrane outside the range of the F fusion peptide.
The fact that large stalk insertions with very strong α-helical propensity are tolerated membrane distal to residue 118 furthermore lends support to secondary structure predictions (36) extending a rigid, helical configuration found for the PIV5 HN stalk (78) to MeV H. Insertions equivalent to 1.9 (7 residues), 3.9 (14 residues), or 11.4 (41 residues) helix turns are compatible with the formation of functional fusion complexes. This again suggests some rotational flexibility within the H stalk as considered above on the basis of successful insertion of N-glycans at positions 110, 111, and 112. A drop in F triggering and surface expression as a consequence of a 55-residue insertion supports, however, the idea that stalk insertions of progressive length eventually have a destabilizing effect on H.
If H stalks enjoy some rotational flexibility and residues in stalk section 84 to 118 (presumably residues 111, 114, and 118) engage in short-range contacts for F triggering, how is receptor binding by the H head domain (48, 71) signaled to the contact zone despite the presence of a long (41-residue) stalk insertion? Interestingly, no major conformational changes between crystal structures of PIV5 HN, HPIV3 HN, and henipavirus G, solved alone or in complex with their receptors, were observed (8, 9, 74). Although receptor binding was originally proposed to initiate major changes in the NDV HN protein (17, 61, 62), subsequent work has demonstrated that major HN rearrangements are unlikely to be required for NDV F triggering (39). Since the dimer-of-dimer interface of PIV5 HN is substantially smaller than that within the dimer interaction, it has been hypothesized as a possible mechanism for F triggering that receptor binding may result in partial disassembly of the tetrameric head of the HN attachment protein, in turn changing the interaction with F and inducing triggering (78). Some conformational changes upon receptor binding have also been implicated for Nipah virus and Hendra virus G (1, 4). For the latter, it was proposed that these changes may affect the dimer-of-dimer interface (6).
By analogy, partial or complete disintegration of hypothetical MeV H homotetramers upon receptor binding would not require complex signaling through the stalk domain and, thus, be reconcilable with the partially maintained functionality of the H-11841x variant. However, isolated MeV H head domain fragments have crystallized as monomers (14) and homodimers (28). In this context, it should be noted that similarly soluble head domains of henipavirus G attachment proteins are also monomeric in solution and crystals (8, 74). Nevertheless, native G is predicted to exist in a tetrameric dimer-of-dimers configuration (7). Elucidating the order of physiological, membrane-embedded MeV H oligomers before and after receptor binding therefore emerges as a central task in understanding the mechanism of MeV F triggering. While we represent H structures as homodimers in this study for simplicity and in light of the available X-ray data, we remain open to the hypothesis that the tetramer constitutes the physiological oligomer of MeV H. Certainly, reorganization of a hypothetical dimer-of-dimer oligomer rather than an intricate signaling mechanism could provide a straightforward explanation as to how engagement of distinct binding sites in the H-Edm head domain by CD46 or SLAM (48, 71), and even of single-chain antibody moieties carboxy-terminally attached to MeV H through flexible linkers (27), results in effective F triggering.
We thank J. J. Yoon and A. Shapiro for help with support experiments of the study and A. L. Hammond for critical reading of the manuscript.
M.A.B. was funded by the Molecular Mechanisms of Microbial Pathogenesis training grant T32 AI007470 from the NIH. This work was supported by U.S. Public Health Service grant AI071002 (to R.K.P.) from the NIH/NIAID.
Published ahead of print on 5 August 2009.