Identification of Key Residues in Virulent Canine Distemper Virus Hemagglutinin That Control CD150/SLAM-Binding Activity

Abstract

Morbillivirus cell entry is controlled by hemagglutinin (H), an envelope-anchored viral glycoprotein determining interaction with multiple host cell surface receptors. Subsequent to virus-receptor attachment, H is thought to transduce a signal triggering the viral fusion glycoprotein, which in turn drives virus-cell fusion activity. Cell entry through the universal morbillivirus receptor CD150/SLAM was reported to depend on two nearby microdomains located within the hemagglutinin. Here, we provide evidence that three key residues in the virulent canine distemper virus A75/17 H protein (Y525, D526, and R529), clustering at the rim of a large recessed groove created by β-propeller blades 4 and 5, control SLAM-binding activity without drastically modulating protein surface expression or SLAM-independent F triggering.

Paramyxoviruses are enveloped nonsegmented negative-strand RNA viruses that inject their genetic information into target cells by fusing their lipid envelope with the plasma membrane of the host cell at a neutral pH. Plasma membrane fusion activity is achieved by the concerted action of two viral membrane-bound glycoproteins. The attachment protein (hemagglutinin [H], hemagglutinin-neuraminidase [HN], or attachment [G], depending on the viral genus) is thought to bind a host cell surface receptor, in turn activating the fusion (F) protein, which will then undergo large-scale structural rearrangements, leading to plasma membrane fusion activity (9, 10, 19). In addition, both viral surface glycoproteins may mediate fusion activity between two contacting neighboring cells (22, 27). Virus-induced cell-cell fusion activity eventually leads to multinucleated cell formation (also termed syncytium formation) and, ultimately, to cell lysis.

The crystal structure of the measles virus hemagglutinin (MeV-H) has recently become available (3, 7, 8). Interestingly, the overall β-propeller structure consisting of six β-sheets was well conserved compared to already determined paramyxovirus HN structures (4, 12, 29). The canine distemper virus H (CDV-H) protein has a short N-terminal cytoplasmic tail followed by a transmembrane domain and a large C-terminal ectodomain (1). It is suggested that the ectodomain consists of a stalk region with an α-helical coiled-coil configuration (13, 28) that supports a globular head domain containing the receptor recognition site and antigenic regions of the protein (11).

Recently, site-directed mutagenesis aimed at identifying residues throughout the MeV-H ectodomain that might selectively control membrane fusion activity in a receptor-dependent manner (CD150/SLAM, CD46, or a yet-unidentified putative epithelial cell receptor [EpR]) was conducted. Indeed, four key residues, located in two connected microdomains (site 1 and site 2) on MeV-H globular head β-propeller blade 5, were necessary to uphold SLAM-dependent fusogenicity. Mutations in each one of the four amino acids resulted in a selective inhibition of SLAM-dependent fusion activity (H-SLAM-blind; HSB [25]). Interestingly, the latter quartet of residues were subsequently demonstrated not to be involved in SLAM-binding activity but presumably were involved in controlling SLAM-dependent F triggering (14). An additional residue (isoleucine 194), located within MeV-H β-propeller blade 6 but in contact with site 2, was next shown to govern interaction with the universal morbillivirus SLAM receptor (14). Consequently, the corresponding residues of both microdomains were mutated in the H protein of the virulent CDV strain 5804P and were also demonstrated to control SLAM-dependent fusion activity (24), although for CDV, full ablation of fusion activity required the substitutions in both microdomains and in two additional neighboring amino acids (CDV-H residues in site 1, D526, I527, S528, and R529; in site 2, Y547 and T548). Moreover, using a CDV-H 3D homology model, the two microdomains were demonstrated to be in very close proximity to one another (compared to those of MeV-H) but not in direct contact (24). Subsequently, a recombinant CDV bearing a SLAM-blind H protein was reported to be completely attenuated in ferrets, a phenotype associated with reduced immunosuppression and lack of neurovirulence (26). However, the precise molecular mechanisms sustaining HSB-dependent lack of fusion support activity was not elucidated and remains to be determined.

A75/17 CDV HSB exhibits a defect in SLAM-binding activity.

Based on the proposed morbillivirus F/H-induced fusion model, plasma membrane fusion activity may be regulated by the hemagglutinin at multiple levels. Indeed, surface density, receptor-binding efficiency, and the strength of lateral interaction with F and F triggering are suggested to be major factors that may modulate fusion activity. To discriminate between these possibilities, we mutated the six previously identified residues shown to ablate SLAM-dependent fusion activity (24) in the hemagglutinin protein of the highly virulent and demyelinating A75/17 CDV strain. Subsequently, a FLAG-tag sequence was fused to the C-terminal part of HSB (HSBeF) as well as the parental H protein (HwteF) and the hemagglutinin derived from the vaccine CDV strain (HOPeF) to allow for easier cell surface determination. Importantly, the tag sequence did not impede Hwt and HOP fusion support functions. Furthermore, we confirmed that the six previously identified residues were also fully ablating SLAM-dependent fusion promotion mediated by the H protein of the virulent A75/17 CDV (data not shown).

The three tagged hemagglutinins were next transfected in Vero cells, and surface expression was determined by flow cytometry and surface biotinylation. From this set of experiments, it was apparent that all three H proteins were properly targeted to the plasma membrane, although HSB was characterized by a slight reduction in transport competence and HOP elicited enhanced surface expression compared to that of Hwt (Fig. 1A and C). We noticed a difference in the electrophoretic mobility between Hwt and HOP, very likely resulting from alternative processed N-glycosylation sites (Fig. 1A) (21). Moreover, it appeared that total expression of HOPeF was substantially reduced compared to that of HwteF and HSBeF, thus suggesting some defects in protein stability and/or accelerated degradation. However, we did not further investigate this phenomenon, since we focused our study on the amount of the three H proteins specifically expressed at the cell surface. Subsequently, we investigated the ability of all three hemagglutinins to interact with the canine SLAM receptor by using a newly established semiquantitative SLAM-binding assay. To this purpose, a soluble form of the N-terminally hemagglutinin-tagged SLAM molecule (HA-sSLAM) (23) was engineered and expressed in 293T cells. This molecule specifically bound to Hwt but not to empty vector-transfected Vero cells, as demonstrated by immunofluorescence staining using an anti-hemagglutinin monoclonal antibody (MAb) (Fig. 1B). Thus, following the addition of HA-sSLAM to H-expressing cells, we decorated the cells with an anti-hemagglutinin MAb, and mean fluorescence intensities (MFI) were recorded by flow cytometry to monitor the HA-sSLAM-binding efficiencies to the various hemagglutinins. The latter values were further normalized to MFI values obtained from anti-FLAG MAb-stained cells (to control for any H surface expression differences). Finally, the ratio obtained from the HA-sSLAM/Hwt interaction was arbitrarily set at 100%.

FIG. 1.

Assessment of SLAM-binding activity with the different H mutants. (A) The A75/17 CDV-H SLAM-blind protein reaches the cell surface. Cell surface expression (CSE) of various H glycoproteins was determined by surface biotinylation. Biotinylated H proteins from transfected Vero cells were precipitated overnight with streptavidin coupled to agarose beads, separated by reducing SDS-PAGE and blotted onto nitrocellulose membranes. A polyclonal anti-H antibody (2) was employed to reveal the surface-exposed H antigenic materials. Total lysates (TL) taken prior to precipitation were subjected to immunoblotting, and H antigenic materials were detected using the identical polyclonal anti-H antibody. (B) Vero cells were transfected with an empty (pCI) or Hwt-expressing plasmid. The supernatant (filtered and concentrated) of soluble canine hemagglutinin-tagged SLAM molecules produced in 293T cells was then added to the cells at 4°C for 1 h. This was followed by immunofluorescence staining using an anti-hemagglutinin MAb and the addition of an Alexa Fluor 488-conjugated secondary antibody, validating the specificity of our assay. Images were taken using a confocal laser microscope (Olympus). (C) Vero cells were transfected with an empty vector or Hwt and derivative H mutants. Soluble SLAM molecules were added as indicated above, and SLAM-binding activity was calculated as the ratio of mean fluorescence intensities obtained with the anti-hemagglutinin MAb values (staining for sSLAM) normalized to the levels obtained with the anti-FLAG MAb (staining for H). Values recorded for Hwt/SLAM-binding efficiency were set at 100%. Cell surface expression (CSE) of each H protein was then determined by measuring the different H protein-binding activities to the anti-FLAG MAb, as recorded by flow cytometry. Values obtained for HwteF were set at 100%. Next, for the quantitative fusion assay, a Vero-SLAM cells population was infected with MVA-T7 (multiplicity of infection [MOI] of 1). In parallel, a second population (Vero cells) was transfected with the different H proteins, a plasmid encoding wt F (F^wt), and a plasmid containing the luciferase reporter gene under the control of the T7 promoter. Twelve hours after transfection, the two cell populations were mixed and seeded in fresh plates. After 2.5 h at 37°C, fusion was quantified by measuring the luciferase activity produced. For each experiment, the value for the F^wt/H^wt combination was set to 100%. Means and standard deviations from three independent experiments performed in duplicate are shown. (D) Syncytium formation after cotransfection of Vero-SLAM cells with plasmid DNA encoding various CDV-H proteins and F^wt. Mock-transfected cells (pCI) received CDV F^wt-encoding plasmid and empty vector; representative fields of view were photographed 24 h posttransfection.

Figure 1C documents that Hwt and HOP readily bound HA-sSLAM, whereas in striking contrast, HSB totally lost any binding activity. To map the residue(s) in HSB affecting HA-sSLAM interaction, we generated a series of single-mutated H proteins, encompassing residues in site 1 and 2 microdomains as well as neighboring amino acids. Each single H mutant was next assessed for its capacity to bind soluble SLAM. Strikingly, it appeared that H proteins with residues Y525, D526, and R529 mutated exhibited a phenotype identical to that of HSB. Interestingly, while amino acids D526 and R529 were part of the six residues previously reported to inhibit SLAM-dependent fusion activity in CDV, residue Y525 was not. Conversely, no mutation within the site 2 microdomain of CDV-H substantially affected SLAM-binding activity (Fig. 1C).

We next determined whether the loss of HA-sSLAM-binding activity correlated with F-triggering impairments. For this purpose, Vero-SLAM cells were transfected with the various H mutants together with Fwt-expressing plasmids, and fusion activity was assessed 1 day posttransfection by using the luciferase reporter content mix assay as well as the microscopic cell-cell fusion assay. As expected, all H mutants exhibiting defective sSLAM-binding activity (HSB, H Y525A, H D526A, and H R529A) also elicited strongly impaired plasma membrane fusion activity (Fig. 1D). Interestingly, while all SLAM-binding-deficient H proteins exhibited a lack of (or very weak) fusion promotion activity by the content mix assay, only HSB totally ablated fusion in the microscopic cell-cell fusion assay (compare Fig. 1C and D). We believe that such differences might be explained by the long time interval typically used in the cell-cell fusion assay compared to that used in the content mix fusion assay (24 h versus 2 h, respectively), which may increase the potential of poor fusion support H mutants to achieve sufficient mass action to finally trigger F. Taken together, our studies identified three residues in Hwt that were required to allow efficient binding to HA-sSLAM, which correlated with their inability to promote proper F triggering.

Combining the H Y525A substitution with D526A and/or R529A results in complete SLAM-dependent fusion inhibition.

In order to investigate any synergistic deficiency in fusion activity by all three identified sSLAM-binding residues, three double mutants and one triple H mutant were generated. Figure 2 illustrates that any combination involving a mutation at position Y525 in Hwt drastically impaired fusion activity. An assessment of soluble SLAM interaction with the latter H mutants revealed that all three double mutants and the triple mutant completely lost SLAM-binding activity, a result which correlated with their fusion promotion deficiencies (Fig. 2A). Conversely, the double H mutant bearing the combined substitutions at positions 526 and 529 did exhibit residual binding to sSLAM and moderate fusion activity (Fig. 2A and B). While the precise mechanism sustaining the unexpected phenotype elicited by the H double mutant (D526A/R529A) merits further investigation, we speculate that the two alanine substitutions, when combined, help H to fold into a specific conformational state that recovers a minor capacity to bind SLAM.

FIG. 2.

Combinations involving the H Y525A mutations drastically reduced cell-cell fusion-promoting activity. (A) Cell-cell fusion activity induced by the different double and triple H mutants (identical to Fig. 1D). White arrowheads highlight rare syncytia mediated by the different H/F combinations. (B) Investigation of H/SLAM-binding activity. Assays of cell surface expression, HA-sSLAM binding, and quantitative cell-cell fusion were performed as mentioned above (Fig. 1C). Means and standard deviations from three independent experiments performed in duplicate are shown.

SLAM-binding-deficient H mutants can support F triggering.

Finally, it was investigated whether the intrinsic capacity of the H mutants that were unable to bind SLAM efficiently correlated with additional impairments in F triggering (i.e., in signal transduction to activate F). However, as stipulated above, Hwt does not induce cell-cell fusion in Vero cells if SLAM is not artificially expressed, suggesting that binding efficiency to a yet-unidentified receptor in Vero cells is too low to enable wild-type (wt) F triggering. To overcome this problem, we employed a recently described F mutant (L372A), which is characterized by a reduced energy threshold required to activate the fusion process (17). Hwt expressed in combination with this destabilized F mutant in the absence of the receptor SLAM achieved sufficient fusogenicity to allow the monitoring of cell-cell fusion induction (16, 17). The various H proteins were thus coexpressed in Vero cells with F L372A, and their fusion-promoting capacities were assessed 1 day posttransfection. Importantly, Fig. 3A demonstrates that in Vero cells, all SLAM-binding-deficient H mutants were able to support syncytium formation, although fusion remained very limited, as was also the case for Hwt.

FIG. 3.

The intrinsic F-triggering function of all SLAM-binding-deficient H mutants remains active in two SLAM⁻ cell types. Vero cells and primary canine epithelial keratinocytes were transfected with an empty vector (pCI) or Hwt and derivative mutants together with the destabilized F L372A mutants (17). Cell-cell fusion induction was monitored 1 day posttransfection using a confocal laser microscope (Olympus). Representative field of view of pictures taken by phase-contrast microscopy are shown. White arrowheads highlight the syncytia mediated by the different H/F combinations.

To validate the above data with a canine cell system in which the virulent A75/17 CDV efficiently grows using a receptor other than SLAM (5, 18), cell-cell fusion was investigated by coexpressing H wild-type and derivative mutants with F L372A in canine primary SLAM⁻ epithelial keratinocytes. Strikingly, syncytium formation was readily detected with wt H as well as with all SLAM-binding-deficient H mutants (Fig. 3B). Importantly, although we did not directly investigate H/F interactions or H conformational alterations, the above data suggested that subsequent to binding activity, the remaining H-protein fusion-support functions were not drastically impaired. Indeed, fusion promotion efficiency induced by all mutated H proteins was very similar to that of Hwt in both SLAM⁻ Vero cells and SLAM⁻ canine keratinocytes. Consequently, these results provided strong evidence that the major defect characterizing HSB (and derivative mutants) lies in the receptor-binding activity and not in triggering F, although we do not exclude the possibility that the latter function might be slightly modulated, in turn regulating the extent of the fusion process.

In attempts to accurately localize the respective positions of the three identified residues within the globular head, a new three-dimensional (3D) homology model of the A75/17 CDV hemagglutinin was generated. Strikingly, all three residues clustered at the top of the molecule, and residue D526 makes direct contacts with amino acids Y525 and R529 (Fig. 4A and B). While residues Y525 was mostly buried inside the molecule, the side chains of D526 and R529 were exposed at the top of the rim of a large groove (Fig. 4A). Since all three residues were demonstrated to independently modulate SLAM-binding activity, substitution of residue Y525 will have major conformational effects on the structure of the upper part of the groove, thereby influencing the receptor-binding site, where D526 and R529 may reside.

FIG. 4.

Residues Y525, D526, and R529 controlling SLAM-binding activity clustered at the top of a recessed groove within CDV-H (side view). (A) Location of the site 1 microdomain in CDV-H. Residues Y525 (buried inside the molecule), D526, and R529 are highlighted in green, red, and purple, respectively. (B) Detailed view of the cluster of three residues (spherical representation) identified to regulate SLAM-binding activity within the CDV-H 3D model. Figures were generated using PyMOL v0.99 and color coded for clarification. Homology modeling was performed using the sequence alignment of MeV-H and CDV-H^A75/17 and the homologous template of the native H structure of MeV (7) using the automated protein-modeling software on the SWISS-MODEL protein-modeling server (6, 15). (C) Sequence alignment of the relevant segment within the H protein of various morbilliviruses (Morb.). GenBank numbers for each virus sequence are AB016162.1 (measles virus, ICB strain), AF266288.1 (measles virus, Edmonston strain), AB254456.1 (measles virus, SSPE strain kobe-1), BAA01203.1 (CDV, 5804P strain), AY386315.1 (CDV, 5804 strain), AY386316.1 (CDV, A75/17 strain), D85755.1 (CDV, Yanaka strain), AF305419.1 (CDV, Onderstepoort strain), X98291.3 (rinderpest virus, Kabete O strain), Y18816.1 (rinderpest virus, K strain), EU267273.1 (peste des petits ruminants virus, ICV89 strain), FJ648456.1 (phocine distemper virus, DK02 strain), FJ648457.1 (porpoise morbillivirus, IRL88 strain), and NC_005283.1 (dolphin morbillivirus). Bottom, asterisks represent identical residues and colons represent semiconserved residues. CDV-H residues Y525, D526, and R529 and the corresponding amino acids among the different Morbillivirus strains are highlighted in gray.

It is important to note that residues H Y525, D526, and R529 are invariably conserved among all morbillivirus strains so far tested (Fig. 4C). This is consistent with the notion that all morbilliviruses use SLAM as the main in vivo receptor as well as with the findings demonstrating that MeV-H residues Y529, D530, and R533 (corresponding to residues Y525, D526, and R529 in CDV-H) were initially discovered to control SLAM-dependent fusion activity (25). However, in striking contrast to our findings, MeV-H residues Y529, D530, and R533 (in site 1) and Y549 (in site 2) were demonstrated not to be involved in SLAM-binding activity, whereas isoleucine 194, a residue located in the MeV-H site 2 microdomain of β-propeller blade 6, governed interaction with a soluble form of SLAM (14). The authors concluded that the MeV-H quartet residues (Y529, D530, R533, and Y549) might be involved in triggering a signal required to activate F rather than in receptor-binding activity (14). Interestingly, the most effective substitution at position 194 in MeV-H reported to reduce SLAM-binding activity was achieved by I194S (14). In CDV-H, however, the corresponding residue at that location is a serine (S194), suggesting that this residue did not sustain an identical key regulating role. We nevertheless mutated serine residue 194 in CDV-H into alanine, but neither fusion activity nor SLAM-binding activity was found to be substantially altered (not shown). Thus, while in MeV-H, residue S194 located in the site 2 microdomain within β-propeller blade 6 was documented to control interaction with SLAM (probably through a long-range effect), in CDV-H, this appears to be regulated by a cluster of three residues within site 1 of β-propeller blade 5. The reason for the latter discrepancy between the two closely related morbilliviruses is not clear but may be explained by the fact that we used a full-length membrane-bound H protein, while Navaratnarajah and colleagues employed a soluble version of H to investigate SLAM/H interactions (14). Indeed, the absence of the cytosolic tail and the transmembrane domains may alter the conformational state(s) of the H globular head. Alternatively, we cannot exclude the possibility that the origin of the hemagglutinin used (wild type in case of CDV and vaccine in case of MeV) may result in subtle differences in the specific residues required to interact with SLAM.

Interestingly, Santiago and coworkers (20) recently successfully crystallized the MeV-H protein (Edmonston strain) bound to the CD46 receptor. In the crystal structure, CD46 binds to a unique glycan-free groove created by β-propeller blades 4 and 5 on one side of the MeV-H globular head. Importantly, not only one, but three main contacting regions between the two molecules were identified, of which one was lying in a recessed socket in the center of the groove. Based on previously identified receptor-binding regulating residues and an additional site-directed mutational analysis of soluble SLAM molecules, the authors suggested that the large recessed groove might be required to engage no fewer than three receptors (20). The latter MeV-H/CD46 atomic structure additionally suggests that MeV-H residue I194 constitutes a long-range determinant for productive interaction with SLAM. Interestingly, the large recessed groove is well conserved in our CDV-H 3D model and, as mentioned above, residues Y525, D526, and R529 clustered at the top of the molecule on the rim of the groove. Strikingly, when the corresponding CD46 footprints on MeV-H (20) were illuminated in our model, it appeared clearly that the cluster of three identified residues was not involved in any of these domains, even though they are adjacent to two residues shown to be involved in direct MeV-H/CD46 contacts (in CDV-H, 499 and T524; in MeV-H, E503 and T528) (Fig. 5). We speculate that the cluster of three residues may define a primary (or secondary) microdomain that exclusively drives interaction with SLAM and may therefore explain why selective receptor-recognizing hemagglutinins have been successfully generated.

FIG. 5.

Residues Y525, D526, and R529 did not overlap with residues within CDV-H that correspond to CD46 footprints on MeV-H (side view) (20). Residues Y525 (buried inside the molecule), D526, and R529 are shown in red. Residues in CDV-H corresponding to residues in MeV-H β-propeller blade 4 that contact CD46 are represented in orange. Residues in CDV-H corresponding to residues in MeV-H β-propeller blade 5 that contact CD46 are shown in blue. Residues in CDV-H corresponding to residues in MeV-H that were shown to be engaged in hydrogen bonding with CD46 are highlighted in pink. The figure was generated using PyMOL v0.99.

In conclusion, we identified three key residues located at the top of the globular head domain of the virulent A75/17 CDV-H protein which govern SLAM-binding activity without substantially influencing protein expression and SLAM-independent F triggering. Considering that the three identified H residues are invariably conserved in all morbillivirus strains, we believe that combining the MeV-H crystal structure and derived 3D homology models will provide relevant templates for the rational design of small molecules that could inhibit the spread not only of virulent CDV but also of morbilliviruses in general.