Canine Distemper Virus and Measles Virus Fusion Glycoprotein Trimers: Partial Membrane-Proximal Ectodomain Cleavage Enhances Function

Abstract

The trimeric fusion (F) glycoproteins of morbilliviruses are activated by furin cleavage of the precursor F₀ into the F₁ and F₂ subunits. Here we show that an additional membrane-proximal cleavage occurs and modulates F protein function. We initially observed that the ectodomain of approximately one in three measles virus (MV) F proteins is cleaved proximal to the membrane. Processing occurs after cleavage activation of the precursor F₀ into the F₁ and F₂ subunits, producing F_1a and F_1b fragments that are incorporated in viral particles. We also detected the F_1b fragment, including the transmembrane domain and cytoplasmic tail, in cells expressing the canine distemper virus (CDV) or mumps virus F protein. Six membrane-proximal amino acids are necessary for efficient CDV F_1a/b cleavage. These six amino acids can be exchanged with the corresponding MV F protein residues of different sequence without compromising function. Thus, structural elements of different sequence are functionally exchangeable. Finally, we showed that the alteration of a block of membrane-proximal amino acids results in diminished fusion activity in the context of a recombinant CDV. We envisage that selective loss of the membrane anchor in the external subunits of circularly arranged F protein trimers may disengage them from pulling the membrane centrifugally, thereby facilitating fusion pore formation.

Fusion of cellular membranes is a fundamental biological process (21). The mechanisms of membrane fusion have been characterized in detail by using viral fusion glycoproteins (vFGps) (19). Type I vFGps include the fusion proteins of retroviruses, filoviruses, coronaviruses, orthomyxoviruses, and paramyxoviruses (7, 10, 46). Type II vFGps are exemplified by the envelope proteins of flaviviruses and alphaviruses (18, 25). Both type I and type II vFGps are arranged as trimers at fusion: rings of five to six type II homotrimers inserted in the target membrane have been visualized and modeled in atomic detail (11, 12), and a similar arrangement of type I homotrimers has been postulated (27). Ring-like structures may facilitate the concerted activation of vFGps and the synchronized release of their conformational energy.

Type I vFGps are synthesized as monomers but trimerize after their cotranslational insertion into the membrane of the endoplasmic reticulum, glycosylation, and folding (19). Following trimerization, type I vFGps are cleaved by host proteases, an essential step in their activation (13, 23, 34). Cleavage and activation of most vFGps rely on the ubiquitous intracellular protease furin, but the activation of vFGps of certain para- and orthomyxoviruses depends on tissue-specific proteases that determine tropism (14, 15, 44). On proteolytic processing, vFGps are in a metastable state, essentially primed for fusion. Activated type I vFGps are composed of a membrane-anchored and a membrane-distal subunit, which are named F₁ and F₂, respectively, in paramyxoviruses.

The trimer-of-heterodimer complexes formed by paramyxoviral F proteins are remarkably similar to those formed by the other type I vFGps of orthomyxo- and retroviruses (2, 40). In particular, the membrane-anchored F₁ subunit contains two hydrophobic domains, the fusion peptide and the transmembrane (TM) segment. These domains are adjacent to conserved heptad repeats, designated HRA and HRB, respectively. Distinctive differences among vFGps also exist. In the paramyxoviral F proteins, only a few linker amino acids separate the fusion peptide from HRA, and the HRB-TM linker peptide is also very short (Fig. 1). The linker regions are much longer in other vFGps (35).

FIG. 1.

Scheme of the CDV F protein and sequences of the membrane-proximal regions of three paramyxoviral F proteins (A) and Western blot analysis of different paramyxoviral F proteins (B and C). (A, top) Linear drawing of the CDV F protein. This protein is synthesized with a long amino-terminal precursor sequence that is cleaved posttranslationally prior to cleavage activation of the F₀ precursor into the disulfide-linked F₁ and F₂ subunits (53). Hydrophobic regions are indicated by hatched boxes. F protein subunits are labeled as follows: signal peptide (SP), F₀ precursor, F₁ and F₂ subunits, and the F_1a and F_1b fragments that result from the newly identified cleavage. The disulfide bond (SS) connecting the F₁ and F₂ subunits is shown below the F protein scheme. The boxes marked HRA and HRB indicate the positions of the corresponding heptad repeats. (Bottom) Alignment of sequences surrounding the putative cleavage regions of CDV, MV, and MuV F proteins. Identical residues are indicated by dots, hydrophobic residues in the first and fourth (a and d) positions of HRB are in bold, and predicted transmembrane domain residues are italicized. (B and C) Characterization of the different F protein forms by Western blot analysis. Proteins were extracted from purified CDV or MV particles (par) or Vero cells transfected (tr) with the CDV, MV, or MuV F protein expression plasmids or infected (inf) with CDV or MV, or control uninfected cells (ctr). Lysates were separated by reducing SDS-PAGE and blotted onto polyvinylidene difluoride membranes. The F proteins were revealed with anti-cytoplasmic tail (B) or anti-F₄₃₁ (C) serum. The positions of F₀, F₁, F_1a, and F_1b are indicated on the left.

The mechanism of activation of paramyxoviral F proteins is similar to that of the human immunodeficiency virus envelope (Env) protein: it occurs at neutral pH and starts with binding of the viral attachment protein to its receptor (8, 49). The viral attachment protein then interacts with F, and this interaction leads to the rearrangement of HRA and HRB and the formation of a stable six-helix bundle structure after insertion of the hydrophobic fusion peptide into the target membrane (2).

The three-dimensional structure of a recombinantly expressed ectodomain of the Newcastle disease virus F protein has been determined by X-ray crystallography at 3.3 Å resolution (6), and the three-dimensional structure of the proteolytically processed F₁ and F₂ Sendai virus trimeric complex has been determined by electron cryomicroscopy at 16 Å resolution (26). These studies, together with low-resolution structural analyses that identified “lollipop”- and “cone”-shaped F protein trimers in human respiratory syncytial virus (3), and mutational studies of the HRA and HRB domains of simian virus 5 (SV5) (41) on the one hand indicated that the F proteins of paramyxoviruses have similar structures and on the other identified structurally and functionally different conformers and intermediates.

Despite a wealth of knowledge, no consensus about the mechanism of fusion by type I vFGps has been reached. Nevertheless, one fusion model agrees with most of the experimental data and is widely accepted. It predicts that fusion protein trimers are organized in rings before fusion. Upon concerted activation, the fusion peptides are propelled over a distance of many nanometers into the target membrane. In the second step, the HRBs fold back on the HRAs, forming a six-helix bundle while bringing the TM segment into proximity with the fusion peptide, pulling the two membranes together and forming the fusion pore. This model, however, does not explain how bending towards the ring center is achieved by the refolding proteins in spite of rotational symmetry of the trimers.

Therefore, alternative models of the fusion mechanisms have been considered. One proposes that the fusion peptide inserts into the viral rather than the target membrane and that when the fusion proteins are grouped in a ring around a central patch of lipid, a dimple is created that destabilizes the membrane and initiates the fusion reaction (24). Another model considers that fusion peptides from the same trimer insert simultaneously into both the target and the viral membrane and then zipper up the trimeric coiled coil to pull the two membranes together, ultimately destabilizing the bilayer and causing lipid exchange (47).

We show here that the ectodomain of the type I vFGps of three paramyxoviruses undergo partial membrane-proximal cleavage and that proteolytic processing enhances fusion function. We suggest that membrane-proximal cleavage of the external trimer subunit may disengage it from pulling the membrane centrifugally, thereby enhancing fusion efficiency.

MATERIALS AND METHODS

Cells and virus purification.

Vero cells (ATCC CCL-81) were used for all experiments. The cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) with 5% fetal calf serum (Invitrogen). CDV Onderstepoort strain (CDV_OS) and MV Edmonston strain (MV_Edm) were propagated in Vero cells.

To obtain purified viral particles, two 175-cm² dishes were infected with each virus and incubated at 32°C until at least 80% of the cells were in syncytia. The supernatant was cleared by centrifugation at 3,000 rpm in a tabletop centrifuge (Sorvall) for 20 min at 4°C and overlaid onto a 20 to 60% sucrose gradient in TNE buffer (10 mM Tris [pH 7.8], 100 mM NaCl, 1 mM EDTA). After centrifugation at 28,000 rpm in a TH641 rotor (Sorvall) for 1.5 h at 4°C, the interphase was transferred into a fresh tube, adjusted to 20% sucrose in TNE buffer and the centrifugation was repeated at 28,000 rpm for 1.5 h at 4°C. The pellet was resuspended in 200 μl of TNE buffer, aliquoted, and stored at −70°C. The titer of each preparation was determined, and sample volumes for Western blot analysis were adjusted accordingly.

Construction of expression plasmids.

Plasmid pCG-F (55) constituted the basis for all CDV F protein mutants. Soluble F proteins with increasing deletions in the F₁ C terminus were generated by PCR (Expand High Fidelity PCR system; Roche Biochemicals) with the common forward primer TTTGGATCCGGTCAACCAGGTCCACCAGCCAGG, which introduces a BamHI site upstream of the coding region (indicated in italics), and the reverse primers TTTGCATGCTCATTACTTGTCATCGTCATCCTTGTAGTCATTAAAGGAAGAGCGCCTAACCGTCTC (pCG-sF₆₀₅), TTTGCATGCTCATTACTTGTCATCGTCATCCTTGTAGTCGATCTGGTTAGAGGAGTCTATCAG (pCG-sF₅₉₅), and TTTGCATGCTCATTACTTGTCATCGTCATCCTTGTAGTCAGCATCCAGTTTCTTAAGGGCG (pCG-sF₅₈₅), adding the Flag peptide (underlined) between the end of the truncated F₁ subunit and two stop codons, which are followed by a SphI site (italics). Plasmid pCG-FTryps (54) was used as the template for the generation of the corresponding soluble F proteins pCG-sF_tryps/605, pCG-sF_tryps/595and pCG-sF_tryps/585, in which the furin cleavage motif RRQRR/F was changed into the trypsin cleavage motif RNHNR/F. The mumps virus (MuV) F protein expression plasmid pCG-FMuV was obtained from P. Duprex (unpublished data).

A set of CDV F protein mutants were generated by site-directed mutagenesis (Stratagene), in which the residues between E597 and L610 were changed to alanine in groups of two or three, resulting in pCG-F597/9 (A1), pCG-F600/1 (A2), pCG-F602/4 (A3), pCG-F605/7 (A4), and pCG-F608/10 (A5). The combination of mutants A3 and A4, A4 and A5, and A3 through A5 yielded pCG-F602/7 (A3/4), pCG-F605/10 (A4/5) and pCG-F602/10 (A3-5). To interrupt extended alanine stretches, the two additional constructs pCG-F602S/10 (S3-5) and pCG-F602/10S (3-5S) were generated, in which apolar residues in either A3/4 (S3-5) or A5 (3-5S) were changed to serine instead of alanine. Finally, the following mutants were obtained: pCG-FΔ602/5 (Δ1) and pCG-F606-9 (Δ2), which lack the corresponding residues, and pCG-F605/10MV (M4/5) and pCG-F602/10MV (M3-5), in which the respective CDV F protein residues are replaced by the corresponding MV F protein residues. The sequences of all constructs were verified (ABI Prism 377 DNA Sequencer; Perkin-Elmer Applied Biosystems).

Construction and recovery of recombinant viruses.

The mutation A3-5 was introduced into the context of the full-length CDV genome of the vaccine strain Onderstepoort with the internal restriction sites AflII and PacI. The fragment to be inserted was assembled by overlap extension PCR (20), combining the F gene downstream of the AflII site with the part of the untranslated region between the F and H gense that is located upstream of the PacI site. After sequence verification of the inserted fragments, recombinant viruses were recovered as described previously with an MVA-T7-based system (55). First, small syncytia were detected around 12 days after transfection, compared to 6 days in the control transfected with standard virus cDNA. For each virus, three syncytia were picked, transferred to fresh Vero cells in six-well plates, and expanded into 75-cm² flasks with 10 ml of DMEM supplemented with 2% fetal calf serum. When the cytopathic effect was pronounced, the cells were scraped into the medium and subjected once to freezing and thawing. The cleared supernatants were used for all further analysis.

Fusion assay.

The quantitative fusion assay based on luciferase as the reporter gene (53) was used. Briefly, Vero cells were transfected with the different F expression plasmids together with pCG-H/OL and pTM1-luc at a molar ratio of 1:1:0.7 with Lipofectamine 2000 (Gibco-BRL). For each transfected well, a second well of Vero cells was infected with modified vaccinia virus Ankara expressing the T7 polymerase (MVA-T7) (31) with a multiplicity of infection of 1. Twelve hours after transfection or infection, the cells were detached with 50 μl of 0.25% trypsin-EDTA (Gibco-BRL), resuspended in 1 ml of fresh DMEM with 5% fetal calf serum, and transferred into two wells of a 24-well plate. Following visual grading of the fusion activity, luciferase activity was determined with the Steady-Glo luciferase assay system (Promega) and a 96-well plate reading luminometer (Topcount-NXT; Packard). A fraction of each lysate was mixed with an equal amount of 2× Laemmli sample buffer (Bio-Rad) containing 100 mM dithiothreitol and subjected to Western blot analysis.

Western blot analysis.

Vero cells were seeded into 12-well plates, transfected with the different constructs or infected with the respective viruses, and incubated at 37°C for 36 h or until profound cytopathic effect was observed. For the analysis of cellular proteins, cells were washed twice with ice-cold phosphate-buffered saline (PBS; Invitrogen) before adding 100 μl of lysis buffer (150 mM NaCl, 1.0% Triton X-100, 50 mM Tris-HCl, pH 8.0) with complete protease inhibitor (Roche Biochemicals). After incubation for 20 min at 4°C, the lysates were cleared by centrifugation at 5,000 × g for 15 min at 4°C, and the supernatant was mixed with an equal amount of 2× Laemmli sample buffer (Bio-Rad) containing 100 mM dithiothreitol. For the analysis of soluble proteins, cells were washed twice with PBS, and 500 μl of DMEM without fetal calf serum was added 12 h after transfection. The supernatant was exchanged every 24 h for the next 3 days, and the removed supernatant was cleared by centrifugation at 5,000 × g for 15 min at 4°C before mixing with an equal amount of 2× Laemmli sample buffer (Bio-Rad) containing 100 mM dithiothreitol. Samples were incubated for 10 min at 95°C, followed by fractionation on a sodium dodecyl sulfate (SDS)-15% polyacrylamide gel (Bio-Rad) and blotting on polyvinylidene difluoride membranes (Millipore). After blocking with 1% blocking reagent (Roche Biochemicals) overnight, the membranes were incubated with the antibody recognizing the epitope of interest. Following incubation with a peroxidase-conjugated secondary antiserum, the membranes were subjected to enhanced chemiluminescence (ECL) detection (Amersham Pharmacia Biotech). Band intensities were determined with NIH Image software version 1.6. Films from at least three independent experiments were used to determine the F₁/F_1b ratios.

Surface biotinylation.

Vero cells were seeded into 12-well plates, transfected with the different constructs, and incubated at 37°C for 36 h. Cells were shifted to 4°C and washed once with cold PBS before 125 mg of EZ-Link Sulfo-NHS-LC-biotin (Pierce) dissolved in 0.3 ml of cold PBS was added to each well. After incubation for 20 min on a rocker platform at 4°C, cells were washed with 0.5 M glycine in PBS, followed by the addition of 1 ml of 0.5 M glycine-PBS and incubation for 20 min at 4°C to quench the excess biotin. Cells were lysed in 300 μl of radioimmunoprecipitation assay (RIPA) buffer (150 mM NaCl, 1.0% NP-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 8.0) with complete protease inhibitor (Roche Biochemicals), and further processed as described above. Instead of adding the 2× sample buffer directly, the supernatant was mixed with 50 μl of protein A-agarose beads (Bio-Rad), and the antibody was added at the appropriate concentration. Following incubation at 4°C overnight, the beads were washed three times in RIPA buffer before 30 μl of 2× Laemmli sample buffer (Bio-Rad) containing 100 mM dithiothreitol was added. The samples then underwent Western blot analysis as described above. The membranes were incubated with peroxidase-conjugated streptavidin (Amersham Pharmacia Biotech) and subjected to ECL detection.

Production of antibodies.

For the detection of CDV and MV F proteins, a rabbit antipeptide antiserum recognizing the 14 carboxy-terminal amino acids was used (5). A similar rabbit anti-F ectodomain peptide serum was generated. Towards this end, the peptide QVGSRRYPDAVYLHR(C), corresponding to MV F amino acids 431 to 445 with a carboxyl-terminal cysteine, was synthesized and coupled to keyhole limpet hemocyanin. This conjugate was used to produce a rabbit antiserum. Moreover, a peptide antiserum was raised against the MuV F protein. Towards this end, the peptide (C)NTISSSVDDLIRY, corresponding to the 13 carboxy-terminal residues, was coupled to keyhole limpet hemocyanin. A commercially available monoclonal antibody (M2; Sigma) was used to detect the Flag-tagged proteins.

Membrane purification and trypsin digestion.

Two wells of a six-well plate seeded with Vero cells were infected with CDV_OS or transfected with plasmids expressing the unaltered CDV F protein, a CDV F protein in which the furin consensus sequence was mutated to a trypsin consensus sequence (pCG-Ftryps; T), or a CDV F protein that contains an endoplasmic reticulum retention sequence in its cytoplasmic tail (pCG-F_ER) (54). Thirty-six hours after transfection or infection, the cells were washed twice with Tris-buffered saline (TBS; 50 mM Tris-HCl [pH 7.5], 150 mM NaCl) and 0.45 ml of 0.1× TBS was added to each well. After incubation for 10 min at 4°C, cells were scraped into the supernatant and transferred into an Eppendorf tube, and 0.05 ml of 10× TBS was added. The lysate was precleared by centrifugation at 2,500 rpm for 20 min in a minicentrifuge at 4°C. The supernatant was layered on 2.7 ml of 250 mM sucrose-TBS and centrifuged for 30 min at 37,000 rpm in a T61 rotor (Sorvall). The pellet was used for all further experiments.

For proteolysis with trypsin, the pellet was resuspended in 80 μl of TBS with 2 mM tetracaine and 2 mM CaCl₂ and divided into two equal aliquots. Trypsin was either omitted or added to a final concentration of 25 μg/ml. After incubation for 1 h on ice, 30 μg of aprotinin per ml was added to terminate the proteolysis. Following an additional incubation for 20 min on ice, all samples were adjusted to 0.5% NP-40 and solubilized for 20 min on ice. The samples were centrifuged for 15 min at 14,000 rpm in a minicentrifuge at 4°C, and the supernatant was mixed with an equal amount of 2× Laemmli buffer (Bio-Rad) with 100 mM dithiothreitol and subjected to Western blot analysis as described above.

RESULTS

Small F protein carboxy-terminal fragment is detected in transfected and infected cells and in purified particles of three paramyxoviruses.

With an F protein cytoplasmic tail-specific antiserum, we occasionally observed a small band in Western blots of lysates of cells expressing the F proteins of two paramyxoviruses, the morbilliviruses MV and CDV. To better characterize this fragment, we analyzed lysates of cells transfected with F protein expression plasmids or cells infected with MV or CDV or purified viral particles. With the same antiserum, we detected an approximately 10-kDa band in lysates of transfected and infected cells as well as purified viral particles (Fig. 1B). Some variability in band size or a double band was detected in some samples. We suspect this may be due to further proteolytic degradation of the small fragment. Detection of this F protein carboxy-terminal fragment in all samples indicates that it is produced in both transfected and infected cells and that it is incorporated into viral particles. We suspect that this small fragment has been previously overlooked due to its size and its lack of reactivity against antibodies raised against the F protein ectodomain.

We then asked whether the complementary fragment could also be detected. For this we raised a serum against a peptide corresponding to a hydrophilic sequence of the MV F₁ ectodomain. With this antiserum, we detected an approximately 30-kDa band in purified viral particles (Fig. 1C, band F_1a; this antiserum does not react with the CDV F protein). The fact that a strong F₁ band was maintained in all lysates (Fig. 1B and C) indicates that only a fraction of all F proteins undergo the additional cleavage. We refer to the newly identified fragments as F_1a and F_1b. Since the CDV F protein is more stable than the MV F protein, further analyses focused on it.

We then asked whether an F_1b-like fragment is produced by a representative of another Paramyxoviridae subfamily (30), the rubulavirus MuV. Indeed, with an antiserum raised against the 13 carboxy-terminal residues of the MuV F protein, a small fragment was detected in a cell lysate (Fig. 1B, right panel). We determined the ratio of cleaved (F_1b) to uncleaved (F₁) proteins by densitometry. In purified virions we measured an average 1:2 ratio for CDV (Fig. 1B, left panel, and data not shown) and an apparently lower ratio for MV, in the range of 1:3 to 1:4 (Fig. 1B, center panel). However, the MV F_1a to F₁ ratio also ranged around 1:2 (Fig. 1C), suggesting that approximately one-third of morbillivirus F proteins incorporated in viral particles have undergone the additional cleavage.

We then compared the amino acid sequences in the regions preceding the predicted transmembrane domain of CDV, MV, and MuV (30). The hydrophobic residues in positions 1 and 4 (Fig. 1A, bottom) of the HRB domain allow alignment of the three sequences, but otherwise little conservation exists. Downstream of the HRB amphipathic helix, five to seven amino acids precede the hydrophobic region predicted to cross the membrane. A single residue (serine) was conserved among the three viruses.

F_1a/b cleavage requires endoplasmic reticulum transit and F_1/2 cleavage.

To determine the sequence of events occurring during F protein processing and maturation, we relied on two mutants, F_tryps, in which the furin cleavage sequence is changed to a trypsin cleavage sequence, resulting in a protein that is efficiently processed and transported to the cell surface in its uncleaved form; and F_ER, in which an endoplasmic reticulum retention signal has been introduced (54). Both F_ER and F_tryps can be cleaved by trypsin after cell lysis. Membrane preparations were used for these experiments because they were less prone to nonspecific proteolytic cleavage (data not shown).

In membrane protein extracts of CDV-infected cells or cells transfected with the plasmid expressing the unaltered F protein, trypsin digestion led to the disappearance of F₀, no marked increase in F₁, and slightly reduced F_1b (Fig. 2, lanes CDV± and F±), suggesting that trypsin may not cleave at the F_1a/b junction. In the case of F_tryps, F₀ was the dominant form detected in the undigested sample (Fig. 2, lane F_tryps−). Trypsin digestion of this protein resulted in increases of not only F₁ but also F_1b, suggesting the possibility that F_1/2 cleavage is a prerequisite for F_1a/b processing (Fig. 2, lane F_tryps+). While F_ER was also predominantly detected in its uncleaved F₀ form in the undigested sample (Fig. 2, lane F_ER−), mainly F₁ was detected in the trypsin-digested aliquot (Fig. 2, lane F_ER+). In summary, these findings indicate that F_1a/b cleavage is a late event in F protein maturation that requires endoplasmic reticulum transit and F_1/2 cleavage.

FIG. 2.

Trypsin digestion of CDV F protein mutants. Western blot analysis of crude membrane preparations of Vero cells infected with CDV or transfected with different F protein expression plasmids was performed. Cells underwent hypotonic lysis prior to pelleting through 250 mM sucrose in TBS. The pellet was resuspended and divided into two equal aliquots, one of which was subjected to trypsin digestion (+), while the other served as a control (−). Lysates were separated by reducing SDS-PAGE (15% acrylamide) and blotted onto polyvinylidene difluoride membranes. The F proteins were revealed with an anti-cytoplasmic tail serum. The positions of F₀, F₁, and F_1b are indicated on the left. Note that trypsin digestion reduced the amount of all F protein forms detected regardless of the mutant analyzed. C, control nontransfected cells; F, F_tryps, and F_ER, cells transfected with pCG-F, pCG-F_tryps, and pCG-F_ER, respectively.

F_1a/b cleavage is conformation dependent and occurs between residues 595 and 608.

To localize the F_1a/b cleavage site, we reasoned that we could monitor the loss of a terminal tag sequence added at different positions towards the end of the HRB sequence or further downstream (see scheme in Fig. 3A). We generated F proteins with a Flag tag inserted after amino acid 585 (deleting the last two heptad repeats; Fig. 3B), after residue 595 (deleting the last heptad repeat), or after residue 608 (leaving the whole ectodomain intact). Since the Flag tag was followed by a stop codon, secretion of these proteins was expected. We reasoned that F_1a/b cleavage may result in loss of protein detection, thus allowing us to map the cleavage site position.

FIG. 3.

Scheme of the F protein (A), sequence of its membrane-proximal region (B), and initial mapping of the membrane-proximal cleavage site (C). (A) Scheme of the membrane-bound F protein (left) and of the truncated F proteins sF₆₀₈, sF₅₉₅, and sF₅₈₅. The disulfide bond connecting the F₁ and F₂ subunits is symbolized by a line. The cell membrane is indicated by a gray rectangle, the Flag tag by a white box, and the cleavage site by an arrow. (B) The position of the last residue of different proteins is indicated with an ordinal number and a bent arrow. The hydrophobic a and d residues of HRB are in bold, and predicted transmembrane domain residues are italicized. (C) Western blot analysis of soluble F proteins. Supernatants of Vero cells transfected with the indicated plasmids were subjected to reducing SDS-PAGE (10% acrylamide) and blotted onto polyvinylidene difluoride membranes. Soluble F proteins were detected with a monoclonal antibody directed against the Flag peptide that was added at the truncation site. The positions of F₀ and F₁ are indicated on the left.

With the anti-Flag antibody, we detected the F₁ ectodomain in the supernatant of cells transfected with sF₅₈₅ and sF₅₉₅ but not in the supernatant of sF₆₀₈-transfected cells (Fig. 3C), indicating that the cleavage site is located upstream of residue 608 but downstream of amino acid 595. In control furin cleavage-resistant F proteins, F₀ ectodomains were secreted (Fig. 3C, lanes sF_tryps/585, sF_tryps/595, and sF_tryps/608), confirming that F_1a/b processing depends on F_1/2 cleavage. However, the sF_tryps/608 signal was reduced compared to the intensity of the corresponding bands of sF_tryps/585 and sF_tryps/595, which may reflect inefficient cleavage independent of F_1/2 cleavage or lower stability of this protein form. Thus, the F_1a/b cleavage site is located between residues I595 and S608 in close proximity to the transmembrane region. In contrast to the membrane-bound F protein, the soluble form is completely cleaved membrane proximally, suggesting that membrane anchor loss makes the cleavage site more accessible.

Inhibition of F_1a/b cleavage leads to cell-cell fusion reduction.

To identify residues important for cleavage within the previously defined 14-amino-acid segment, we then systematically mutated or exchanged blocks of two to three neighboring residues (Fig. 4A). The resulting mutants were analyzed in a quantitative cell-cell fusion assay, with luciferase as the reporter gene. While mutants A1 (residues 597 to 599) and A2 (residues 600 and 601) were efficiently cleaved membrane proximally (Fig. 4C, lanes A1 and A2, F_1b signals; Fig. 4A, % processing column) and exhibited wild-type fusion activity (Fig. 4A, % fusion column), a 40 to 50% reduction of fusion efficiency was observed for mutants A3 and A4 (residues 602 to 604 and residues 605 to 607, respectively), which coincided with detection of less F_1b (Fig. 4B, lanes A3 and A4). Mutant A5 (residues 608 to 610) cleavage analysis results were more variable than for the other mutants, as reflected by the standard deviation in Fig. 4A.

FIG. 4.

Sequence, processing, fusion activity, and surface expression of CDV F protein cleavage mutants. (A) Sequence of the region of interest of the original and mutant F proteins, efficacy of their membrane-proximal processing, and fusion activity. The names of the mutant F proteins are indicated on the left. Mutated residues are bold. The end of HRB and the beginning of the transmembrane region (TM) are marked by bent arrows. For quantitative processing assays (means and standard deviation are indicated in the % processing column), Western blots from three independent experiments similar to those shown in panels B and C were evaluated. For quantitative fusion assays, Vero cell monolayers were either infected with MVA-T7 (multiplicity of infection of 1) or transfected with the different F constructs, a plasmid coding for the H protein, and a plasmid containing the luciferase reporter gene under control of the T7 promoter. Twelve hours after transfection, both cell populations were mixed and seeded into fresh plates. After 36 h at 37°C, fusion was quantified by measuring luciferase activity. For each experiment, the value measured for the parental F protein was set to 100%. The means and standard deviations of four independent experiments done in duplicate are indicated in the % fusion column. (B and C) Western blot analysis of CDV F protein mutants. Lysates were separated by reducing SDS-PAGE and blotted onto polyvinylidene difluoride membranes. The F proteins were revealed with an anti-cytoplasmic tail serum. The positions of F₀, F₁, and F_1b are indicated on the left. (D) Surface biotinylation. A duplicate well of the cells used for Western blot analysis was shifted to 4°C, biotin labeled, lysed, and immunoprecipitated overnight with the anti-Fcyt rabbit antipeptide antibody. Samples were separated by reducing SDS-PAGE, blotted onto polyvinylidene difluoride membranes, and probed with peroxidase-coupled streptavidin. (E) Correlation of fusion activity (y axis) with processing efficiency (x axis) for the 15 proteins characterized.

Throughout these experiments, we observed some variability in the amount of F_1b detected for the different mutants (Fig. 4A, % processing column). Nevertheless, the standard F protein and the five mutants that retained 95 to 110% fusion efficiency (A1, A2, M4/5, Δ1, and Δ2; see below) all had 37 to 42% membrane-proximal processing, whereas those combination mutants that had low fusion efficiency (A3/4, A3-5, and 3-5S) had 5 to 7% membrane-proximal processing (Fig. 4A, middle column). Figure 4E shows that the fusion efficiencies of most mutants correlated well with the levels of their F_1a/b processing.

To assess whether the effects of A3, A4, and A5 are cumulative, we generated plasmids with different combinations of these mutations. The combination of A3 and A4 reduced fusion activity to near the background (Fig. 4A, A3/4, % fusion column), and little F_1b was detected (Fig. 4B, lane A3/4). In contrast, fusogenicity and F_1b production of the A4/5 combination mutant remained in the range of A4 (Fig. 4A, % fusion and % processing columns). For the 3-5 triple combination mutant, in addition to A3-5 with eight consecutive alanines, two constructs with serines interrupting the alanine stretches were generated (3-5S and S3-5, Fig. 4A). Fusion activity and membrane-proximal processing of mutants A3-5 and 3-5S were similar to those of A3/4, while those of S3-5 were in the range of A4 and A4/5 (Fig. 4A, % fusion and % processing columns). Since surface expression of all mutants reached at least wild-type levels (Fig. 4D) and was thus not a limiting factor, these results indicate that the residues most relevant for cleavage are situated between S602 and G607.

Different primary amino acid sequences can be efficiently cleaved.

Having noted that the primary sequences of CDV and MV differ in the cleavage region, we asked whether MV sequences could nevertheless functionally substitute the corresponding CDV segment. Towards this end, we replaced residues 602 to 607 or residues 602 to 610 with the corresponding MV F protein amino acids (Fig. 4A, M4/5 and M3-5, respectively). In contrast to the alanine mutant A3/4, M4/5 displayed fusion activity similar to that of standard F (Fig. 4A, % fusion) and was efficiently cleaved (Fig. 4C, lane M4/5); the fusogenicity of M3-5 was only slightly reduced (Fig. 4A and C, lanes M3-5). The fact that two divergent primary sequences can be cleaved with similar efficiency suggests that an exposed structure may be the primary determinant of proteolytic cleavage.

To complete the characterization of the requirements for efficient cleavage of CDV F at this site, we generated two mutants in which four amino acids, either residues 602 to 605 (Δ1) or 606 to 609 (Δ2), were deleted (Fig. 4A, bottom two lines). Both mutants were efficiently cleaved and displayed wild-type fusion activity (Fig. 4A, % fusion columns). These results concur with the MV F segment substitution analysis in suggesting that the main cleavage requirements are close proximity to the transmembrane region and a tertiary structure compatible with protease accessibility and cleavage.

Mutation of the membrane-proximal region in the viral context causes reduced syncytium formation.

To assess the importance of the F_1a/b cleavage in the viral context, we introduced mutation A3-5 into the infectious cDNA clone of the vaccine strain Onderstepoort. Recovery of the resulting virus, CDV-F_A3-5, was delayed: syncytia appeared 12 days after transfection, compared to 6 days for the positive control. Moreover, the growth phenotype was dominated by small syncytia. A comparative growth analysis revealed that the cytopathic effect was delayed by 48 to 72 h in CDV-F_A3-5. As shown in Fig. 5, large syncytia formed 2 days after inoculation in the control infection with standard CDV (Fig. 5B), whereas syncytia of similar size were detected 4 days after inoculation with CDV-F_A3-5 (Fig. 5H). Moreover, in CDV-F_A3-5 the size and numbers of syncytia never reached the levels of the parental virus (Fig. 5, compare panels D and H). We also found that the kinetics of virus production was slightly delayed and the CDV-F_A3-5 titers were slightly reduced (data not shown), but virus production was still efficient even when the cytopathic effect was minimal. In summary, these findings demonstrate that a recombinant CDV with nine membrane-proximal amino acids mutated to alanine has reduced fusion function and suggest that reduction of F_1a/b cleavage also negatively affects fusion efficiency in the viral context. However, we cannot exclude that additional mechanisms contribute to the fusion reduction of this mutant.

FIG. 5.

Cytopathic effects and cell fusion of a standard CDV (A to D) and a recombinant virus with mutated membrane-proximal cleavage site (E to H). Vero cells were infected at a multiplicity of infection of 0.01 with the parental virus CDV or the cleavage-impaired virus CDV-F_A3-5 and photographed at 24 h (A and E), 48 h (B and F), 72 h (C and G), and 96 h (D and H).

DISCUSSION

We have shown that the F protein ectodomain of three paramyxoviruses is cleaved in a short linker region situated between HRB and the hydrophobic membrane region after cleavage of the inactive precursor F₀ into the F₁ and F₂ subunits. In the morbillivirus CDV, this cut occurs in about one third of the molecules. Mutation of certain membrane-proximal residues results in cleavage inhibition and correlates with fusion activity reduction.

Multilevel control of fusion efficiency.

Trimeric type I vFGps share structural characteristics and mechanisms controlling fusion efficiency. For example, truncation of the cytoplasmic tail results in rapid and extensive cell-to-cell fusion in type D retroviruses (4), human immunodeficiency virus (39, 59), and paramyxoviruses (5, 29). Moreover, in paramyxoviruses the lateral interactions of the vFGps and the attachment protein modulate the efficiency with which the signal elicited by receptor binding is transmitted (8, 16, 38, 49). In addition, N-linked glycans modulate fusion efficiency in paramyxo- and orthomyxoviral vFGps (33, 54).

Membrane-proximal F₁ protein cleavage is another level at which viral fusion can be controlled. Cleavage and activation of the vFGps of certain para- and orthomyxoviruses depends on tissue-specific proteases and determines tropism by restricting fusion to selected tissues (15, 23, 44). Analogously, F_1a/b cleavage may modulate fusion efficiency in different tissues. However, we do not know which protease cleaves in the membrane-proximal region. The primary sequence does not give hints, because it is highly variable even among pairs of closely related paramyxoviruses such as the morbilliviruses CDV and MV (Fig. 1) or the rubulaviruses MuV and simian virus 5 (data not shown). Thus, it seems likely that structural elements, and possibly different proteases, may be involved in membrane-proximal cleavage. Analogously, it has been proposed that structural elements as well as recognition sequences contribute to the cleavage susceptibility of cell surface proteins whose ectodomains are cleaved membrane proximally (1, 17).

Two common characteristics emerge from the mutational analysis of the vFGps membrane-proximal region of the three paramyxoviruses simian virus 5 (58), human parainfluenza virus 2 (52), and now CDV: (i) insertion of additional amino acids is not tolerated, while up to four residues can be deleted without negatively affecting fusion activity; (ii) a majority of residues are polar, and an increase in apolar amino acids results in reduction of fusion activity. In contrast, an unusually high concentration of aromatic amino acids exists in the membrane-proximal regions of the vFGps of retroviruses, filoviruses, orthomyxoviruses, rhabdoviruses, alphaviruses, and flaviviruses (48). It is known that in the lentiviral vFGps, certain membrane-proximal tryptophan residues are essential for fusion (42) and may have a membrane-partitioning function. Since there are no membrane proximal tryptophan residues in MV, CDV, and MuV, paramyxoviral F proteins cannot rely on those to distort membranes at fusion.

Mechanism of fusion pore formation.

Fusion induced by most paramyxoviruses relies on two viral proteins, the attachment (H or HN for hemagglutinin and hemagglutinin-neuraminidase, respectively) protein that contacts the cellular receptor(s), and the F protein that coalesces the viral with the cellular membrane (40). In this respect, paramyxoviral fusion is different from fusion induced by the influenza virus hemagglutinin or the retroviral Env; both hemagglutinin and Env can act alone to fuse membranes. Figure 6A shows the cellular receptor and viral proteins involved in morbillivirus fusion: one of the F trimer subunits is cleaved membrane proximally (Fig. 6A, black arrow). Figures 6B and C show how membrane-proximal cleavage of the F protein ectodomain may facilitate the fusion process.

FIG. 6.

Model of the morbillivirus fusion mechanism. (A) Viral and cellular proteins involved in fusion; (B) target membrane invasion; and (C) pore formation. (A, bottom) Viral membrane (gray) with the attachment (H) protein tetrameric complex (green) and the F protein trimer. (Top) Cellular membrane (gray) with the receptor (dark blue). The H protein has a six-blade propeller structure (9, 56) and consists of two noncovalently linked dimers of covalently linked dimers (36, 37). The main receptor protein for both MV and CDV is the signaling lymphocytic activation molecule (SLAM, also called CD150) of the immunoglobulin superfamily (50, 51). The contact surface areas both on SLAM (32) and the attachment protein (45, 56) have been defined. The F protein consist of a fusion peptides (blue horizontal cylindrical structure), a short linker (light gray cylinder), HRA (yellow cylinders numbered 1 to 3), the body of the F₂ and F₁ subunits (large elongated red object), HRB (orange cylinders numbered 4 to 6), another short linker (a gray continuous cylinder in two monomers; an interrupted gray and blue cylinder in the cleaved monomer), and the TM segment with the cytoplasmic tail (blue thin cylindrical structure). (B) Six circularly arranged F protein trimers in the target membrane invasion conformation. (C) Trimers in their most stable six-helix bundle conformation arranged circularly around a fusion pore. Three of the six trimers have been removed for visual clarity. The inset is a coronal section of the six-helix bundle. For details, see the text.

The model of morbillivirus fusion is based on the assumption (21, 57) that type I vFGp trimers are arranged into a ring-like structure at the site where the fusion pore will form. It postulates that the fusion process begins with the contact of a cellular receptor (SLAM; Fig. 6A, top) with the H protein tetrameric complex. This contact may elicit an H protein conformational change, transmitting the signal laterally to the F protein trimer (38, 49). The F protein trimer, in a metastable conformation after cleavage and activation, may then propel the fusion peptide over a distance of several nanometers towards the tip of the molecule and into the target membrane (Fig. 6B, top). In the resulting intermediate conformation, the extended F protein is connected with both membranes and the HRAs form a trimeric coiled-coil (Fig. 6B, top). In a second step, the HRBs fold back and dock on the trimeric coiled coil, repositioning the TM segments near the fusion peptides while forming the fusion pore (Fig. 6C).

The general model of type I vFGps fusion does not explain how bending of the trimers toward the pore center is achieved in spite of the rotational symmetry of their structures. The model shown in Fig. 6B and C illustrates how cleavage of the external subunits in a ring of trimers facilitates fusion pore formation by breaking symmetry. The trimers are poised to reach the most stable six-helix bundle conformation: in this process HRB 5 will dock between HRA 1 and HRA 2, HRB 6 between or HRA 2 and HRA 3, and HRB 4 between HRA 1 and HRA 3 (scheme in the inset of Fig. 6C). In the six-helix-bundle formation process HRB 4 and HRB 6 will drag the lower membrane towards the upper membrane that is concomitantly pulled down by the trimeric coiled coil. If still connected with its TM segment, the subunit with HRB 5 would counteract the action of the two others. However, membrane-proximal cleavage disengages the HRB 5 subunit from pulling the membrane centrifugally; the protein body flips over and connects to the trimeric coiled coil from the top (HRB 5 in Fig. 6C, center and inset), whereas its TM segment and cytoplasmic tail are left behind in the membrane (Fig. 6C, bottom).

Why does membrane-proximal cleavage remain partial? Our data suggest that a major determinant for cleavage is an exposed structure. We propose that cleavage of a single subunit may induce a conformational change of the trimer, interfering with efficient cleavage of the two other subunits. Self-limiting cleavage proximal to the membrane may also favor the reconfiguration of the F protein trimers in a ring.

Can membrane-proximal cleavage modulate the function of other type I vFGps? We have shown that the MuV F protein is cleaved proximal to the membrane, and it is therefore conceivable that the F protein of other rubulaviruses, including simian virus 5, undergoes similar processing. Viruses whose glycoproteins undergo elaborate proteolytic trimming, such as the filoviruses Ebola virus and Marburg virus (22, 43), are other candidates for partial membrane-proximal vFGps cleavage.

How do type I vFGps that are not processed membrane proximally efficiently execute fusion? Structural data have recently suggested that in type II vFGps, a long interdomain linker permits independent domain rotation and allows spontaneous symmetry breaking (28). It is possible that certain type I vFGps have adopted a similar mechanism.