Structure of the measles virus hemagglutinin bound to the CD46 ...

articles

Structure of the measles virus hemagglutinin bound to the CD46 receptor

© 2010 Nature America, Inc. All rights reserved.

César Santiago1, María L Celma2, Thilo Stehle3,4 & José M Casasnovas1 The highly contagious measles virus infects millions of individuals worldwide, causing serious disease in children of developing countries. Infection is initiated by attachment of the measles virus hemagglutinin (MV-H), a glycoprotein anchored to the virus envelope, to the host cell receptors CD46 or signaling lymphocyte activation molecule (SLAM). Here we report the crystal structure of MV-H in complex with a CD46 protein spanning the two N-terminal domains. A unique groove at the side of the MV-H -propeller domain, which is absent in homologous paramyxovirus attachment proteins, engages residues in both CD46 domains. Key contacts involve a protruding loop in the N-terminal CD46 domain that carries two sequential proline residues (PP motif ) and penetrates deeply into a hydrophobic socket in MV-H. We identify a similar PP motif in SLAM, defining a common measles virus recognition epitope in the CD46 and SLAM receptor proteins. Despite the availability of a vaccine, measles virus (MV) remains a worldwide cause of morbidity and mortality1,2. MV is an enveloped virus that belongs to the Morbillivirus genus of the paramyxovirus family. The lipid bilayer of the MV particle contains two glyco proteins: the hemagglutinin (MV-H) and the fusion protein (MV-F). MV-H is a disulfide-linked homodimer that mediates the attachment of MV to cell surface receptors, whereas MV-F triggers fusion of the viral and cellular membranes1. The complement-regulatory protein CD46, which is expressed in all nucleated human cells, was first iden tified as a cellular receptor for the Edmonston strain of MV3,4. The extracellular region of CD46 contains four short consensus repeats (SCR1–SCR4), and only the membrane-distal, N-terminal SCR1 and SCR2 participate in the interaction with the MV-H glycoprotein5. In combination with binding studies, the structural analysis of a CD46 protein fragment comprising SCR1 and SCR2 revealed an extended MV binding surface that spans both repeats and includes a protruding and critical MV binding epitope with two consecutive prolines (Pro38 and Pro39) at the base of SCR1 (refs. 6,7). SLAM (CD150), originally identified as a membrane glycoprotein expressed on lymphoid cells, was later found to also serve as a receptor for MV8. The extracellular region of SLAM consists of two immunoglobulin (Ig)-like domains, and MV binds to the N-terminal domain of the protein9. Clinical MV isolates appear to interact specifically with SLAM, although CD46binding MV variants can also be isolated from patients with mea sles10,11. Initially, the ability to use the CD46 receptor was linked to the presence of two distinct amino acid substitutions in the MV-H protein: N481Y12 and S546G13. However, additional mutations such as N390I, N416D, T446S, T484N and E492G also appear to increase the viral binding affinity for CD46 (refs. 14–16). The CD46 and SLAM receptors bind to overlapping sites on the MV-H protein, but with

distinct binding kinetics17,18. Extensive mutagenesis studies of MV-H identified several residues involved in CD46 and SLAM recogni tion14,19,20, and these were used to delineate receptor-binding regions on the structure of MV-H18,21. There is also evidence for a third MV receptor, with unknown identity, on polarized epithelial cells22,23. In order to define the structural basis for MV-H recognition of CD46, we crystallized an Edmonston MV-H protein variant bound to a human CD46 fragment comprising domains SCR1 and SCR2 and determined the structure of this complex at 3.1-Å resolution. Analysis of the contacts explains the critical role of the two residues that enable the use of CD46 by a subset of MV strains12,13. The structure provides relevant insights into the interaction of MV with CD46 as well as SLAM and illuminates the strategy followed by MV to switch receptor-binding specificity. RESULTS Overall structure of the complex We crystallized a version of MV-H that included the C-terminal globu lar region but lacked the cysteine residues that mediate the formation of a disulfide-linked homodimer in complex with a fragment com prising the SCR1 and SCR2 domains of CD46 (see Online Methods). The CD46 binding affinity of the MV-H protein used here (Kd of 0.2 µM)17 was significantly higher than that reported for the Edmonston B strain (Kd of 2.2 µM)18. Our MV-H protein bears a glycine instead of a serine at position 546, a residue substitution that confers high CD46 binding affinity13. Diffraction data extending to 3.1-Å resolution were used for structure determination (Table 1 and Online Methods). The asymmetric unit of the crystals contained an MV-H dimer, consisting of two protomers, that is bound to two CD46 molecules (Fig. 1a). Each MV-H protomer assumes a six-bladed (β1–β6) β-propeller fold, each blade having four β-strands (s1–s4). The relative positions of

1Centro

Nacional de Biotecnología, Consejo Superior de Investigaciones Científicas, Campus Universidad Autónoma, Madrid, Spain. 2Virology Unit, Hospital Ramón y Cajal, Madrid, Spain. 3Interfaculty Institute for Biochemistry, University of Tuebingen, Tuebingen, Germany. 4Department of Pediatrics, Vanderbilt University School of Medicine, Nashville, Tennessee, USA. Correspondence should be addressed to J.M.C. ([email protected]). Received 18 June; accepted 29 September; published online 13 December 2009; doi:10.1038/nsmb.1726

124

VOLUME 17 NUMBER 1 january 2010 nature structural & molecular biology

articles Table 1 Data collection and refinement statistics MV-H–CD46 Data collection Space group Cell dimensions a, b, c (Å) Resolution (Å) Rmerge I / σI Completeness (%) Redundancy


Refinement Resolution (Å) No. reflections Rwork/Rfree No. atoms Protein Ligand/ion (SO42–) Water B-factors Protein MV-H (A/B) CD46 SCR1 (D/C) CD46 SCR2 (D/C) Ligand/ion (SO42−) Water R.m.s. deviations Bond lengths (Å) Bond angles (°)

P 2221 81.4, 105.8, 208.7 20–3.1 (3.2–3.1)* 4.8 (38) 22.7 (2.1) 96 (96) 3.7 (3.2)

15–3.1 31,909 22.5/25.9 8,342 8,227 15 8 159 155/155 204/198 140/141 181 94 0.008 1.497

*Values in parentheses are for highest-resolution shell. Diffraction data comes from a single crystal (Na7). The B-factors for the four different molecular chains (A–D) and for the SCR1 and SCR2 repeats of CD46 are shown.

the two MV-H protomers in the dimer are identical to those seen in the structure of a ligand-free disulfide-linked homodimeric MV-H protein18 (Supplementary Fig. 1), even though our protein lacks the cysteine residues engaged in homodimerization. The two dimers can be superimposed with an r.m.s. deviation of 1.33 Å (791 equivalent resi dues). This value is similar to the r.m.s. deviation of 1.31 Å (396 equiva lent residues) for the superposition of the protomers only. Moreover, the two MV-H dimer subunits bury a similar solvent-accessible surface area (about 1,000 Å2) at their interfaces. Therefore, the dimeric organization of MV-H observed in our crystals is likely to be representative of the organization of the protein on the virus envelope.

The MV-H dimer engages two CD46 molecules at equivalent l ocations that are distant from the dimer interface (Fig. 1a). Only a single MV-H protomer interacts with a single CD46 molecule in each case, so that MV-H dimerization is unlikely to influence CD46 recog nition. The overall orientation of the complex shows that each MV-H dimer subunit would easily be able to engage two CD46 molecules at the cell surface, leading to an increase in binding avidity (Fig. 1a). Complex formation buries surface areas of 1,065 Å2 and 1,045 Å2 in each of the two MV-H protomers, respectively, similar to the buried surface areas in the bound CD46 molecules (1,110 Å2 and 1,090 Å2, respectively). Contacts between CD46 and MV-H are very similar in both copies of the complex, and the small differences in buried surface areas are due to slightly differing contacts between the SCR1 of CD46 and the loops at the top of the MV-H β-propeller domain (Fig. 1b). The CD46 region buried by the interaction with MV-H includes the lower portion of SCR1, the SCR1-SCR2 interdomain interface, and one side of SCR2 (Fig. 1b). Contacts mediated by SCR1 are limited to a region at the base of the domain, centered at the D'D loop. Although the base of SCR1 is fixed as a result of its identical interactions with the MV-H protein in both copies of CD46, the remaining regions of the SCR1 domains had somewhat different conformations because of distinct crystal contacts (Supplementary Fig. 2a), which correlates with their relatively poor electron-density maps and enhanced ther mal mobility (B-factors in Table 1). In contrast, the electron density for SCR2 was well defined. The interactions with MV-H cover almost the entire side of SCR2 (Fig. 1b), resulting in substantially lower ther mal mobility of this domain compared with SCR1 (Table 1). In the complex with MV-H, the CD46 two-domain module assumes a lightly bent conformation, with an angle of about 155° between SCR1 and SCR2 (Fig. 1b and Supplementary Fig. 2b). The overall orientation of the two domains is similar to that of the extended CD46 structure seen in the complex with the adenovirus 11 (Ad11) fiber knob24, and it is quite distinct from the severely bent conformation adopted by the ligand-free CD46 protein7 (Supplementary Fig. 2b). Thus, the substantial interdomain flexibility among concatenated SCR domains allows the adaptation of the CD46 molecule to the distinct binding surfaces provided by the Ad and MV attachment proteins. The virus-receptor binding interface The MV-H–CD46 interface can be divided into three contact regions (Fig. 1b, blue, magenta and light blue spheres). Critical contacts between MV-H and CD46 residues are shown in Figure 2, whereas additional contacts are included in Supplementary Table 1.

a

b

Figure 1 Crystal structure of the MV-H MV membrane protein bound to the CD46 receptor. (a) Representation of the dimeric MV-H–CD46 Stalk Stalk β5 complex present in the asymmetric unit of Top Top SCR1 the crystals. Ribbon drawings of the MV-H SCR1 β6 β4 molecules are shown with a rainbow coloring β5 scheme, whereas Cα traces of the CD46 molecules are shown in blue. The six β-sheets building the MV-H β-propeller domain β1 are labeled. The C-terminal residues of the SCR2 SCR2 molecules, shown as spheres, are followed by β2 β4 a polypeptide stalk in the full-length MV-H β3 β4 SCR3 protein or the SCR3 and SCR4 domains in the SCR4 SCR3 CD46 molecule. Putative location of the virus Bottom Cell membrane SCR4 and cell membranes are shown. (b) Detailed view of the MV-H–CD46 interface. Two side views, differing by 120°, are shown. CD46 residues buried by the interaction with MV-H are shown with spheres colored in dark blue (contact region 1, the D'D loop of SCR1), magenta (contact region 2, the SCR1-SCR2 interdomain region) and light blue (contact region 3, residues at SCR2).

nature structural & molecular biology VOLUME 17 NUMBER 1 january 2010

125

articles a

C

b

D β4-β5

SCR1

SCR1

β5

Cys60

Arg547

Ser112

c Lys488

D′

β4 Leu500

s4

E′

Tyr61

s3

Asn80

Gly546 Tyr481 DE loop

Glu63 Pro39 s2-s3

D′

Leu464 Tyr543

Tyr541

s3 Thr498

β5 s42

Phe483 Tyr481

β4

s4

Cys65

SCR2

Cys65

Tyr83 Tyr67

β4

SCR2 s4 Val451


Figure 2 Contacts at the MV-H–CD46 interface. Ribbon drawing of the three MV-H–CD46 interacting regions described in Figure 1b. CD46 and MV-H are shown in blue and orange, respectively. Buried residues participating in critical contacts are depicted in stick representation and labeled in the three panels, with oxygens and nitrogens shown in red and blue, respectively. Side chains of MV-H residues engaged in hydrogen bonds (black dashed lines) with CD46 are highlighted in magenta, whereas cysteines are yellow. (a) Contact region 1, showing relevant contacts between MV-H and CD46 residues at the D'D loop of SCR1 (Ile37–Leu40). Loops connecting the β-strands s2 and s3 (s2-s3) of blade β4 and the blades β4 and β5 (β4-β5) are marked. (b) Contact region 2, showing interactions of MV-H with the CD46 SCR1-SCR2 interdomain region. This surface includes contacts between MV-H and CD46 residues linking Cys60 in SCR1 and Cys65 in SCR2, as well as the DE loop of SCR2. (c) Contact region 3, showing interactions between MV-H and the SCR2 of CD46. The N-acetylglucosamine (NAG) residue attached to Asn80 of CD46 is shown in stick representation with carbons colored green. Additional contacts are listed in Supplementary Table 1.

Interactions in contact region 1 are restricted to the base of the CD46 SCR1, including the protruding D′D loop. Four D′D-loop residues (Ile37, Pro38, Pro39 and Leu40) form a plug that penetrates deeply into a hydrophobic socket at the interface between blades β4 and β5 (Fig. 2a). Comparison with the structures of ligand-free MV-H proteins reveals conformational changes in the socket that allow the accom modation of the Pro38 and Pro39 residues (P-P motif) at the edge of the SCR1 D′D loop (Supplementary Fig. 3). The plug is sandwiched between the side chains of Leu464 and Leu500 in blade β4 and Tyr541 and Tyr543 in blade β5. Residue Pro39, which was earlier shown to be critical for the interaction with MV-H6, sits onto the tip of the loop joining β-strands s2 and s3 at the blade β4 (Fig. 2a). Contact region 2 involves the SCR1-SCR2 interface of CD46, and in particular the residues connecting the C-terminal cysteine of SCR1 (Cys60) to the N-terminal cysteine of SCR2 (Cys65) as well as a neigh boring loop (Fig. 2b). The SCR1-SCR2 interface is tightly locked between hydrophobic residues (Tyr481-Phe483) at blade β4 and a β-turn connecting strands s3 and s4 at blade β5. This CD46 region appears to contribute significantly to the virus-receptor binding spe cificity through a substantial number of hydrogen bonds with the MV-H protein (Fig. 2b), engaging the key Tyr481 and Gly546 deter minants for high-affinity binding of MV-H to CD46 (refs. 12,13). The Tyr481 hydroxyl forms a hydrogen bond with the main chain carbonyl of Cys65 in CD46 (Fig. 2b and Supplementary Fig. 4). Its replacement with asparagine would eliminate this and other interactions with nearby residues (Supplementary Table 1). The mutation S546G increased the MV-H binding affinity for CD46 (ref. 13). Although both residues could make similar interactions with the MV-H residue Glu63 via their main chain amino groups (Fig. 2b), it is likely that glycine introduces additional flexibility, facilitating binding to CD46 (Supplementary Fig. 3). Contact region 3 involves almost the entire side of the CD46 SCR2, which rests on the MV-H β4 sheet and buries about 55% of the total interacting surface (Fig. 2c). The side chain of CD46 residue Tyr67 lies at the center of the contact region, packing against MV-H resi due Val451. The substitution V451E abrogates CD46 binding12, most likely because of steric clashes of the glutamate side chain with CD46. Above Tyr67, the side chain of Tyr83 lies close to the MV-H residue Tyr481, and the glycan attached to Asn80 (Fig. 2c, green) interacts with the side chain of the MV-H residue Lys488. The contribution of 126

the N-linked glycan to the interaction explains the reported require ment of this glycan for MV binding to CD46 (ref. 25). A unique groove in MV-H mediates receptor recognition Structures of several paramyxovirus attachment proteins, including the hemagglutinin neuraminidase of Newcastle disease virus (NDV) protein and the Nipah virus G (NiV-G), have been determined in complex with receptors26–28. Both engage receptors via a pocket at the top of the β-propeller domain, independent of whether the receptor is a car bohydrate or a protein. Although the MV-H β-propeller domain also has a pocket at the equivalent location, this pocket is wider, less deep, and shielded from interactions by an N-linked glycan at the top of the domain18,21. As a result, CD46 does not engage the pocket at the center of the β-propeller but instead binds to its side (Fig. 1). The MV-binding fragment of CD46 rests on a concave surface gener ated by the β4 and β5 sheets (Fig. 1b). The protruding D′D-loop region in SCR1, which carries the PP motif, penetrates deeply into a recessed socket, whereas the remaining receptor-interacting footprint extends around Tyr481 and below onto the β4 blade (Fig. 3a, white). However, the SLAM-binding residues are preferentially located on the β5 blade18–21 (Fig. 3b and Supplementary Fig. 5) rather than at the β4 blade. Some MV-H residues buried by CD46 binding were reported to interact with SLAM (pink in Fig. 3b), such as Leu526 in the socket. The mutation A527G affecting the floor of the socket substantially reduced SLAM recognition by MV-H19. Most of the described neutralizing antibody epitopes in MV-H surround rather than directly overlap with the CD46 receptor-binding groove (Fig. 3a). In contrast, several SLAM-interacting MV-H residues in the loops at the rim of the socket (Ser532, Arg533 and Phe552) are recognized by neutralizing antibodies (Fig. 3a,b). The conformation of the receptor-binding region of MV-H, espe cially around the connection between the blades β4 and β5, is quite distinct from the corresponding regions in the attachment proteins of other paramyxoviruses (Supplementary Fig. 5). The recessed regions that accommodate CD46 do not exist in homologous paramyxovirus proteins (Fig. 3c). A kink at Phe483 (see Fig. 2b and red spheres in Fig. 3c) bends the polypeptide chain (s42) following the β4s4 strand away from the blade β5 in the MV-H protein, allowing for the formation of the deep socket that engages the D′D loop of CD46 (Fig. 3). In contrast, the corresponding polypeptide chains (s42) in the NiV-G and NDV-HN proteins are more linear and do not allow


articles a

Top

b

6 55

DE

16

l41

Ep21

32

c

Arg533

r5 Se

Phe552 β5

β5

β5

Ala527

Ep22

Leu526

Arg547 Tyr524

-s4

s3

Ep1 Asp521

Tyr481

s42

4 s3-s

s42 s4

β4

β4

β4

Figure 3 A unique elongated groove in MV-H mediates receptor recognition. (a) CD46 and neutralizing antibody binding footprints on the MV-H protein. Surface representation of the MV-H protein, colored in gold, with the residues buried by the interaction with CD46 shown in white (Supplementary Fig. 5). The bound D′D loop of the CD46 SCR1 is shown in stick representation, with carbons colored in blue. The MV-H Tyr481 residue near the center of the CD46 binding surface is labeled. Neutralizing antibody epitopes (Gly491 for 16CD11, Ser532 and Arg533 for 16DE6, Arg533 for 55 and Phe552 for I41), and epitopes related to both MV virulence and neutralization (“noose epitope” residues 380–400 as Ep1, residues 190–200 as Ep2 1 and residues 571–579 as Ep22) are shown in black41–45. (b) CD46 and SLAM binding footprints on the MV-H protein. The MV-H protein with the CD46 binding footprint in white is shown as in a. MV-H residues engaged in SLAM binding are colored in red. Some of these residues are also buried by CD46, and they are colored in pink and labeled19,20. SLAM-binding residues recognized by neutralizing antibodies in the loops at the rim of the socket are labeled in black, whereas the MV-H residue Ala527 at the bottom of the socket is marked in red. (c) Absence of a socket in homologous paramyxovirus attachment proteins. Superposition of MV-H, shown both as ribbon and transparent surface representation in orange, with the structures of NiV-G (PDB 2VSM (ref. 27)) and NDV-HN (PDB 1E8T (ref. 26)), shown as blue and green ribbon drawings, respectively. The superposition reveals structural differences at the groove where CD46 binds (see Supplementary Fig. 5). Regions showing significant conformational differences, such as the β-strand s42 in blade β4 and the s3-s4 β-turn in blade β5, are labeled. The bent polypeptide chain connecting the s4 and s42 strands in MV-H is shown with red spheres.

for the formation of a similar socket (Fig. 3c, blue and green ribbons). Moreover, the s3-s4 β-turn at the blade β5 approaches the end of the β4s4 strand in the NiV-G and NDV-HN proteins because of the highly twisted conformation of the β-strands, which differs from the less twisted structures adopted by the β5s3 and the β5s4 strands in MV-H (Fig. 3c). The unique conformations of those two regions on the side of the β-propeller domain create a glycan-free groove on the MV-H protein that is used for receptor recognition. A common MV recognition motif in CD46 and SLAM Binding studies and data presented in Figure 3 show that CD46 and SLAM bind to overlapping sites in the MV-H protein17–19,21,29, although residues required for MV binding to SLAM cluster at the

β5 blade of the β-propeller domain18,21,29 (Supplementary Fig. 5). Therefore, it is likely that the N-terminal MV-binding domain of SLAM will also bind into or near the recessive surface built by the blades β4 and β5 on the side of the MV-H β-propeller domain. The N-terminal domain of SLAM contains a sequence (PPRY) that resembles the MV binding residues (YIPP) in the D′D loop of CD46 (Supplementary Fig. 6). Although the sequences differ in their direc tion, they both contain two sequential proline residues and a nearby tyrosine that is located N-terminally to the two prolines in CD46 and C-terminally to them in SLAM. The N-terminal domain of SLAM is likely to fold into a variable-type Ig (IgV) domain30 (Supplementary Fig. 6). Modeling the SLAM IgV domain shows that the PPRY sequence is likely located at the C″ edge of the IgV domain (Fig. 4a),

a

b

Binding ratio (%)

Figure 4 A common MV recognition motif in the CD46 and SLAM receptors. (a) Ribbon drawing 100 MV-H 80 of a model of the N-terminal IgV domain of 60 40 SLAM with the β-strands shown in red and 20 the loop with the PPRY sequence represented 0 SLAM +IPO3 +A12 as a stick model (carbons in orange, oxygens 140 n Fc IPO3 A12 MV-H in red and nitrogens in blue). The model was c 120 prepared as described in Online Methods. A P G F C′ C stick representation of the D′D loop bearing the 100 P YIPP sequence of CD46, with carbons in blue, E D B 80 Y is shown on the right. The N- and C-terminal 60 R ends of the loop in both MV receptors are n 40 c marked with lowercase letters, showing their 20 inverse orientation. (b) MV-H binding activity of SLAM protein mutants with substitutions 0 SLAM PP/GG R/A Y/A in the PPRY sequence. Normalized binding of soluble SLAM-Fc wild-type (SLAM) and mutant proteins (PP/GG, R/A and Y/A) to plastic-bound MV-H protein, SLAM mAbs (IPO3 and A12) or an Fc antibody, used as internal control to monitor protein concentration (see Online Methods). Mean and s.d. of data from three to six different experiments are shown. The inset shows inhibition of SLAM binding to MV-H by the IPO3 and A12 SLAM mAbs. Normalized binding of soluble SLAM-Fc to MV-H in the absence (SLAM) or presence of the mAbs is plotted. Mean and s.d. of three different experiments are shown. Binding ratio (%)


16CD11


127


articles where it would be easily accessible for interactions with MV-H. We note that a similar region is used by the homologous IgV domain of CD4 to engage the HIV gp120 glycoprotein31,32. In CD46, the side chain of Tyr36 serves to stabilize the semicircular conformation of the D′D loop via a hydrogen bond and hydrophobic contacts7 (Fig. 4a, blue). The tyrosine in the SLAM sequence could stabilize a circular loop in a similar manner, which would also allow the projection of the two prolines into solution (Fig. 4a, orange). Thus, it is conceivable that the PP motif at the edge of the IgV domain of SLAM could penetrate the hydrophobic MV-H socket in a manner similar to that observed in the complex with CD46. To test the validity of the proposed role for the PP motif in MV-H binding, we carried out mutagenesis of the SLAM PPRY sequence and analyzed the effect of the amino acid substitution (PP/GG, R/A and Y/A) on receptor binding to the MV-H protein (Fig. 4b). The SLAM PP/GG mutant bound very poorly to the MV-H protein, and binding of the R/A and Y/A mutants was significantly lower (by about 70%) than that of the wild-type protein. Thus, our binding data provide clear evidence for a critical contribution of the SLAM PPRY sequence in MV recognition and support our hypothesis that CD46 and SLAM share a key virus-binding motif with two sequential prolines. SLAM monoclonal antibodies (mAbs) blocking receptor binding to MV-H (Fig. 4b, inset) appear to recognize the PP sequence, as they show decreased binding to the PP/GG mutant (Fig. 4b). Our data therefore identify a structurally related MV-H binding epitope in CD46 and SLAM, perhaps explaining why MV can use both receptors, which are otherwise structurally unrelated, to gain entry into host cells. DISCUSSION Subtle modifications in the receptor-binding properties of virus coat proteins can drive the emergence of new pathogens with altered infectivity, tissue tropism or host range33–35. MV represents a wellestablished model for receptor-specificity switching. Whereas all or most MV variants appear to bind to the lymphocyte-specific SLAM receptor, single-residue substitutions (N481Y or S546G) in the MV-H protein enable MV to use the ubiquitous CD46 receptor for entry into cells12,13. This adaptation in receptor specificity may allow for a broader tissue tropism to MV, which could be essential for spreading of a measles infection throughout the host36. The data shown here illuminate the strategy used by MV to engage CD46, and they also provide a platform for understanding the forces that underlie the receptor-specificity switch in MV. MV-H binds to the CD46 receptor via a groove at the side of the β-propeller domain. Earlier studies had shown that a proline in the D′D loop of SCR1 is a key player in this interaction6, and the crystal structure of the MV-H in complex with CD46 demonstrates that this residue is part of a hydrophobic plug that inserts into a socket at one end of the MV-H groove. In addition, the other end of the groove appears to help deter mine the virus-receptor binding specificity through numerous addi tional interactions. Notably, the SLAM sequence contains a PP motif that likely also protrudes from the edge of the N-terminal MV-binding domain and that may bind to the MV-H socket in a similar fashion. Our mutational studies clearly support a critical role for this motif in SLAM binding to MV-H, and the location of SLAM-binding residues at or near the MV-H socket makes it tempting to speculate that the socket is also involved in binding the PP motif of SLAM. The CD46 binding groove would also be engaged in MV-H binding to the MV receptor identified in epithelial cells22,23, as MV-H residues Leu482, Phe483, Tyr541 and Tyr543 involved in recognition of this receptor contact CD46 (Supplementary Table 1). Thus, no fewer than three 128

receptors are likely to engage the same MV-H surface, perhaps using similar strategies but with subtle differences in the actual contacts. Our structural analysis also illustrates the critical role of the two resi dues that were found to be primarily responsible for the switch that allows MV-H to engage CD46 (refs. 12,13). The tyrosine side chain introduced by the N481Y mutation makes precise contacts with the SCR1-SCR2 interdomain interface. We note that, in the CD46 bound to adeno virus, the interdomain region is also a key point of contact for the viral protein24. The S546G mutation likely results in added flexibility to the β5s3-β5s4 region, required for CD46 binding (Supplementary Fig. 3). The β-propeller architecture is common to all paramyxovirus cell attachment proteins characterized to date, independent of the type of receptor they recognize18,21,26–28,37,38. It is also found in numer ous ligand-binding proteins, such as integrins and G proteins39,40, where it often forms a platform that can engage different proteins. The propeller architecture features a recessed pocket at the hub, and this pocket is used by the paramyxovirus for recognition of both sialic acid and protein receptors26–28,37,38. MV-H is the first documented paramyxovirus attachment protein deviating from this binding mode, using a different region at the side of the β-propeller for receptor recognition. Notably, interactions with CD46 involve MV-H residues that are located in a recessed binding groove as well as a socket. Thus, MV-H conserves the recessed nature shared by many receptor-binding sites in virus proteins, which protects receptor-binding residues from immune neutralization. Indeed, the binding sites of MV-H neutralizing antibodies surround rather than overlap with the CD46 binding footprint (Fig. 3a). The socket that accommodates the D′D loop of CD46 is especially recessed and protected by long, flexible loops. In many ways, the architecture of the CD46-binding socket therefore resembles the recessed centers of the β-propeller domains, which are also surrounded by long and flexible loops. In the case of the MV-H–CD46 complex, however, the receptor-binding surface extends beyond the socket into the shallow groove. Viral residues within this extended groove can have certain variability whereas less accessible residues are conserved, which leads to an extended MV tropism by increasing CD46 binding affinity and preserving SLAM binding. The relatively variable and concave surface onto which the CD46 interdomain interface and the SCR2 dock represents an area well suited to accommodate diverse receptor molecules sharing a common motif binding to an inaccessible socket in the MV-H protein. Therefore, the use of an extended surface on the side of the β-propeller domain for receptor binding by MV forms the basis for a strategy to extend the virus tissue tropism by receptorspecificity switching. Methods Methods and any associated references are available in the online version of the paper at http://www.nature.com/nsmb/. Accession codes. Protein Data Bank: Coordinates and structure fac tors for the MV-H–CD46 complex have been deposited with accession code 3INB. Note: Supplementary information is available on the Nature Structural & Molecular Biology website. Acknowledgments We are grateful to F. Pazos for assistance with secondary structure prediction and to R. Fernandez-Muñoz for helpful discussions. We acknowledge the European Molecular Biology Laboratory, the Deutsches Elektronen Synchrotron and the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities. This work has been supported by grants from the Ministerio de Ciencia e Innovación (BFU2005-05972 and BFU2008-00971) to J.M.C. T.S. acknowledges support from SFB-685.


articles AUTHOR CONTRIBUTIONS C.S. and J.M.C. designed the constructs. C.S. prepared the proteins and crystallized the MV-H–CD46 complex. C.S. and J.M.C. contributed to data collection and structure determination. C.S., T.S. and J.M.C. performed structure refinement and model building. C.S., M.L.C., T.S. and J.M.C. contributed to analysis of the data and preparation of the manuscript.


Published online at http://www.nature.com/nsmb/. Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/. 1. Griffin, D.E. Measles virus. in Fields Virology Vol. 1 (eds. Fields, B.N. et al.) 1551–1585 (Lippincott, Williams & Wilkins, Philadelphia, 2007). 2. Rota, P.A., Featherstone, D.A. & Bellini, W.J. Molecular epidemiology of measles virus. Curr. Top. Microbiol. Immunol. 330, 129–150 (2009). 3. Naniche, D. et al. Human membrane cofactor protein (CD46) acts as a cellular receptor for measles virus. J. Virol. 67, 6025–6032 (1993). 4. Dorig, R.E., Marcil, A., Chopra, A. & Richardson, C.D. The human CD46 molecule is a receptor for measles virus (Edmonston strain). Cell 75, 295–305 (1993). 5. Manchester, M. et al. Measles virus recognizes its receptor, CD46, via two distinct binding domains within SCR1–2. Virology 232, 1–12 (1997). 6. Buchholz, C.J. et al. Mapping of the primary binding site of measles virus to its receptor CD46. J. Biol. Chem. 272, 22072–22079 (1997). 7. Casasnovas, J.M., Larvie, M. & Stehle, T. Crystal structure of two CD46 domains reveals an extended measles virus-binding surface. EMBO J. 18, 2911–2922 (1999). 8. Tatsuo, H., Ono, N., Tanaka, K. & Yanagi, Y. SLAM (CDw 150) is a cellular receptor for measles virus. Nature 406, 893–897 (2000). 9. Ono, N., Tatsuo, H., Tanaka, K., Minagawa, H. & Yanagi, Y. V domain of human SLAM (CDw150) is essential for its function as a measles virus receptor. J. Virol. 75, 1594–1600 (2001). 10. Manchester, M. et al. Clinical isolates of measles virus use CD46 as a cellular receptor. J. Virol. 74, 3967–3974 (2000). 11. Erlenhofer, C., Duprex, W.P., Rima, B.K., ter Meulen, V. & Schneider-Schaulies, J. Analysis of receptor (CD46, CD150) usage by measles virus. J. Gen. Virol. 83, 1431–1436 (2002). 12. Lecouturier, V. et al. Identification of two amino acids in the hemagglutinin glycoprotein of measles virus (MV) that govern hemadsorption, HeLa cell fusion, and CD46 downregulation: phenotypic markers that differentiate vaccine and wildtype MV strains. J. Virol. 70, 4200–4204 (1996). 13. Shibahara, K., Hotta, H., Katayama, Y. & Homma, M. Increased binding activity of measles virus to monkey red blood cells after long-term passage in Vero cell cultures. J. Gen. Virol. 75, 3511–3516 (1994). 14. Tahara, M., Takeda, M., Seki, F., Hashiguchi, T. & Yanagi, Y. Multiple amino acid substitutions in hemagglutinin are necessary for wild-type measles virus to acquire the ability to use receptor CD46 efficiently. J. Virol. 81, 2564–2572 (2007). 15. Rota, J.S., Wang, Z.D., Rota, P.A. & Bellini, W.J. Comparison of sequences of the H, F, and N coding genes of measles virus vaccine strains. Virus Res. 31, 317–330 (1994). 16. Schneider, U., von Messling, V., Devaux, P. & Cattaneo, R. Efficiency of measles virus entry and dissemination through different receptors. J. Virol. 76, 7460–7467 (2002). 17. Santiago, C., Björling, E., Stehle, T. & Casasnovas, J.M. Distinct kinetics for binding of the CD46 and SLAM receptors to overlapping sites in the measles virus hemagglutinin protein. J. Biol. Chem. 277, 32294–32301 (2002). 18. Hashiguchi, T. et al. Crystal structure of measles virus hemagglutinin provides insight into effective vaccines. Proc. Natl. Acad. Sci. USA 104, 19535–19540 (2007). 19. Masse, N. et al. Measles virus (MV) hemagglutinin: evidence that attachment sites for MV receptors SLAM and CD46 overlap on the globular head. J. Virol. 78, 9051–9063 (2004). 20. Vongpunsawad, S., Oezgun, N., Braun, W. & Cattaneo, R. Selectively receptor-blind measles viruses: identification of residues necessary for SLAM- or CD46-induced

fusion and their localization on a new hemagglutinin structural model. J. Virol. 78, 302–313 (2004). 21. Colf, L.A., Juo, Z.S. & Garcia, K.C. Structure of the measles virus hemagglutinin. Nat. Struct. Mol. Biol. 14, 1227–1228 (2007). 22. Leonard, V.H.J. et al. Measles virus blind to its epithelial cell receptor remains virulent in rhesus monkeys but cannot cross the airway epithelium and is not shed. J. Clin. Invest. 118, 2448–2458 (2008). 23. Tahara, M. et al. Measles virus infects both polarized epithelial and immune cells by using distinctive receptor-binding sites on its hemagglutinin. J. Virol. 82, 4630–4637 (2008). 24. Persson, B.D. et al. Adenovirus type 11 binding alters the conformation of its receptor CD46. Nat. Struct. Mol. Biol. 14, 164–166 (2007). 25. Maisner, A. et al. The N-glycan of the SCR 2 region is essential for membrane cofactor protein (CD46) to function as a measles virus receptor. J. Virol. 70, 4973–4977 (1996). 26. Crennell, S., Takimoto, T., Portner, A. & Taylor, G. Crystal structure of the multifunctional paramyxovirus hemagglutinin-neuraminidase. Nat. Struct. Biol. 7, 1068–1074 (2000). 27. Bowden, T.A. et al. Structural basis of Nipah and Hendra virus attachment to their cell-surface receptor ephrin-B2. Nat. Struct. Mol. Biol. 15, 567–572 (2008). 28. Xu, K. et al. Host cell recognition by the henipaviruses: crystal structures of the Nipah G attachment glycoprotein and its complex with ephrin-B3. Proc. Natl. Acad. Sci. USA 105, 9953–9958 (2008). 29. Stehle, T. & Casasnovas, J.M. Specificity switching in virus-receptor complexes. Curr. Opin. Struct. Biol. 19, 181–188 (2009). 30. Ono, N., Tatsuo, H., Tanaka, K., Minagawa, H. & Yanagi, Y. V domain of human SLAM (CDw150) is essential for its function as a measles virus receptor. J. Virol. 75, 1594–1600 (2001). 31. Kwong, P.D. et al. Structure of an HIV gp120 envelope glycoprotein in complex with the CD4 receptor and a neutralizing human antibody. Nature 393, 648–659 (1998). 32. Wang, J. et al. Atomic structure of a fragment of human CD4 containing two immunoglobulin-like domains. Nature 348, 411–418 (1990). 33. Bergelson, J.M. et al. Coxsackievirus B3 adapted to growth in RD cells binds to decay-accelerating factor (CD55). J. Virol. 69, 1903–1906 (1995). 34. Hueffer, K. & Parrish, C.R. Parvovirus host range, cell tropism and evolution. Curr. Opin. Microbiol. 6, 392–398 (2003). 35. Martinez, M.A., Verdaguer, N., Mateu, M.G. & Domingo, E. Evolution subverting essentiality: dispensability of the cell attachment Arg-Gly-Asp motif in multiply passaged foot-and-mouth disease virus. Proc. Natl. Acad. Sci. USA 94, 6798–6802 (1997). 36. Manchester, M., Naniche, D. & Stehle, T. CD46 as measles receptor: form follows function. Virology 274, 5–10 (2000). 37. Lawrence, M.C. et al. Structure of the haemagglutinin-neuraminidase from human parainfluenza virus type III. J. Mol. Biol. 335, 1343–1357 (2004). 38. Yuan, P. et al. Structural studies of the parainfluenza virus 5 hemagglutininneuraminidase tetramer in complex with its receptor, sialyllactose. Structure 13, 803–815 (2005). 39. Xiong, J.P. et al. Crystal structure of the extracellular segment of integrin alpha Vbeta3. Science 294, 339–345 (2001). 40. Sondek, J., Bohm, A., Lambright, D.G., Hamm, H.E. & Sigler, P.B. Crystal structure of a G-protein β γ dimer at 2.1A resolution. Nature 379, 369–374 (1996). 41. Ertl, O.T., Wenz, D.C., Bouche, F.B., Berbers, G.A. & Muller, C.P. Immunodominant domains of the measles virus hemagglutinin protein eliciting a neutralizing human B cell response. Arch. Virol. 148, 2195–2206 (2003). 42. Hu, A., Sheshberadaran, H., Norrby, E. & Kövamees, J. Molecular characterization of epitopes on the measles virus hemagglutinin protein. Virology 192, 351–354 (1993). 43. Hummel, K.B. & Bellini, W.J. Localization of monoclonal antibody epitopes and functional domains in the hemagglutinin protein of measles virus. J. Virol. 69, 1913–1916 (1995). 44. Liebert, U.G. et al. Antigenic determinants of measles virus hemagglutinin associated with neurovirulence. J. Virol. 68, 1486–1493 (1994). 45. Ziegler, D. et al. Protection against measles virus encephalitis by monoclonal antibodies binding to a cystine loop domain of the H protein mimicked by peptides which are not recognized by maternal antibodies. J. Gen. Virol. 77, 2479–2489 (1996).


129


ONLINE METHODS Protein preparation and crystallization of the MV-H–CD46 complex. We obtained the recombinant soluble MV-H protein from an Edmonston virus vari ant grown in CV-1 cells46 and comprising the C-terminal globular region of the virus protein, residues 179–617, and an influenza hemagglutinin A (YPYDVPDYA) epitope followed by a linker region (GAQPARSPGIRG) and a thrombin recog nition site (LVPRGS) at the N terminus. The N-terminal end of the MV-H protein is 23 residues shorter than that of the reported ligand-free monomeric MV-H structure (PDB 2RKC21), although they both lack the Cys154 engaged in MV-H homodimerization17 present in the structure of a disulfide-linked MV-H homodimer (PDB 2ZB618). We produced the soluble MV-H protein in stably transfected Chinese hamster ovary lectin 3.2.8.1 (CHO Lec) cells and purified the protein from cell supernatants as described earlier17. The CD46 receptor fragment contains residues 1–126 of the mature protein, which comprise the two N-terminal SCR1 and SCR2 domains, and it was produced in CHO Lec cells as previously reported7. Both CD46 and MV-H bear high mannose–type carbohydrates. The two proteins were concentrated to ~20 mg ml−1 in each case, and the complex was prepared by incubation of the two proteins at a MV-H/CD46 molar ratio of 1:1.2 for 2 h at room temperature (22–24 °C). Crystallization of the complex was per formed using the hanging-drop vapor diffusion method at 20 °C and with a total protein concentration of ~12 mg ml−1. The best crystals grew from a precipitant solution containing 12% (w/v) PEG-8000, 0.2 M ammonium sulfate, 2% (v/v) PEG-400, 1% (w/v) 1,2,3-heptanetriol, 1 mM sodium tungstate (Na2WO4) and 0.1 M MES buffer (pH 6.5). Presence of the two proteins in the crystals was con firmed by gel electrophoresis and MS (data not shown). The crystals were frozen with a solution similar to that used for crystallization but containing 20% (w/v) PEG-8000 and 20% (v/v) ethylene glycol as cryoprotectant. Data were collected in the WB7B beamline at Deutsches Elektronen Synchrotron (wavelength of 0.845 Å) at 100 K and processed with the HKL package47. Data collection and refine ment statistics are shown in Table 1. The crystals contain two macromolecular complexes in the asymmetric unit and about 64% of solvent content. Structure determination. The structure was determined with the molecular replacement method using the program PHASER48,49 and search models derived from the known MV-H and CD46 protein structures7,18,21,24. After initial rigidbody refinement of the six modules present in the asymmetric unit (two copies each of the MV-H, SCR1 and SCR2 domains), we performed several cycles of manual rebuilding, each followed by refinement with the program PHENIX-1.3 (http://www.phenix-online.org). The refinement strategy included bulk sol vent correction, positional, adp, simulated annealing and TLS refinement. NCS restraints were included for MV-H and SCR2 of CD46 but not for the SCR1 domain. This domain was poorly defined in the electron density maps, and it had marked conformational differences in the two molecules of the asymmetric unit, except for the bottom region that contacts MV-H (Supplementary Fig. 2a). The N- and C-terminal ends of the MV-H protein were disordered. The N-terminal residues of the MV-H molecules A and B are Thr193 and Ser184, respectively, whereas the C-terminal residues are Met602 and Cys606. Additional absent resi dues in the refined structure are colored gray in the Supplementary Fig. 5. All the residues are in allowed regions of the Ramachandran plot. MV-H molecules A and B are in complex with CD46 molecules D and C, respectively.

nature structural & molecular biology

Binding studies with soluble MV-H and SLAM proteins. Experiments were carried out with plastic-bound MV-H protein or SLAM mAbs and with soluble SLAM-Fc fusion proteins described in detail previously17. SLAM mutants were generated by PCR and confirmed by sequencing. The SLAM PP/GG, R/A and Y/A mutant proteins contained the residue substitutions Pro-Pro→ Gly-Gly, Arg→Ala and Tyr→Ala, respectively. Wild-type and mutant SLAM-Fc fusion proteins were prepared by transient transfection in 293T cells, and the protein in the cell supernatants was quantified by ELISA with anti-human IgG1–Fc polyclonal antibodies (Fc Abs, DAKO) using a purified Fc fusion protein of known concentration as control. The MV-H, SLAM mAbs IPO3 (SCB) or A12 (Pharmingen) at 5 µg ml−1 in PBS were bound to plastic in 96-well plates (Nunc) for 1 h at 37 °C. The wells were subsequently blocked with 2% (w/v) BSA in PBS for 1 h at 37 °C. Supernatants containing SLAM-Fc protein at four different protein concentrations were added to duplicate wells with bound MV-H or SLAM mAbs and incubated for 1 h at 37 °C. Wells were then washed, and binding of the fusion protein was monitored using a horseradish peroxidase–labeled Fc Ab by optical density at 490 nm. Soluble Fc proteins were used at ranges of 15–0.5 µg ml−1 and 2.5–0.06 µg ml−1 for MV-H and antibody binding experiments, respectively. Background binding of an unrelated Fc fusion protein to wells with bound MVH or SLAM mAbs was subtracted. Average of the binding obtained for all four SLAM-Fc protein concentrations was determined for each experiment. Antibody inhibition of SLAM binding to MV-H was carried out with SLAM-Fc protein at 10 and 5 µg ml−1 in the absence or presence of SLAM mAbs at 10 µg ml−1. Modeling of the N-terminal IgV domain of SLAM. The SLAM protein was first aligned with homologous domains of known structure (PDB 2IF7 (ref. 50), PDB 2PKD (ref. 51) and PDB 1EAJ (ref. 52)) based on both sequence and on secondary structure (Supplementary Figure 6). The model was generated with the program modeller (http://salilab.org/modeller/modeller.html), using the PDB 1EAJ (ref. 52) structure as a template and using the alignment shown in Supplementary Figure 6. Illustrations and structure analysis. Figures of the structure were prepared with PyMOL (http://pymol.sourceforge.net/). Buried residues and surface areas were deter mined with the PISA server (http://www.ebi.ac.uk/msd-srv/prot_int/pistart.html). 46. Stern, L.B., Greenberg, M., Gershoni, J.M. & Rozenblatt, S. The hemagglutinin envelope protein of canine distemper virus (CDV) confers cell tropism as illustrated by CDV and measles virus complementation analysis. J. Virol. 69, 1661–1668 (1995). 47. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326 (1997). 48. Read, R.J. Pushing the boundaries of molecular replacement with maximum likelihood. Acta Crystallogr. D Biol. Crystallogr. 57, 1373–1382 (2001). 49. Collaborative Computational Project, N. The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D 50, 760–763 (1994). 50. Cao, E. et al. NTB-A receptor crystal structure: insights into homophilic interactions in the signaling lymphocytic activation molecule receptor family. Immunity 25, 559–570 (2006). 51. Yan, Q. et al. Structure of CD84 provides insight into SLAM family function. Proc. Natl. Acad. Sci. USA 104, 10583–10588 (2007). 52. van Raaij, M.J.,Chouin, E.,van der Zandt, H.,Bergelson, J.M. & Cusack, S. Dimeric structure of the coxsackievirus and adenovirus receptor D1 domain at 1.7 Å resolution. Structure 8, 1147–1155 (2000).

doi:10.1038/nsmb.1726