key: cord-0959283-zgoi2lkv
authors: Stehle, Thilo; Dermody, Terence S.
title: Structural evidence for common functions and ancestry of the reovirus and adenovirus attachment proteins
date: 2003-02-28
journal: Rev Med Virol
DOI: 10.1002/rmv.379
sha: cb855d6eac206bc35994244aefbb430623a75d58
doc_id: 959283
cord_uid: zgoi2lkv

The crystal structure of the reovirus attachment protein, σ1, reveals a fibre‐like structure that is remarkably similar to that of the adenovirus attachment protein, fibre. Both proteins are trimers with head‐and‐tail morphology. They share unique domain structures and functional properties including defined regions of flexibility within the tail and an unusual symmetry mismatch with the pentameric viral capsid protein into which they are inserted. Moreover, the receptors for reoviruses and adenoviruses, junctional adhesion molecule 1 and coxsackievirus and adenovirus receptor, respectively, also share key structural and functional properties. Although reoviruses and adenoviruses belong to different virus families and have few properties in common, the observed similarities between σ1 and fibre point to a conserved mechanism of attachment and an ancient evolutionary relationship. Copyright © 2003 John Wiley & Sons, Ltd.

Specific attachment of viruses to cell-surface receptors is the initial step in viral infection. Although there exists substantial diversity in the types of receptors recognised by different viruses, all viruses must possess a strategy to penetrate cell membranes. Enveloped viruses achieve this through fusion of viral and cellular membranes, and many enveloped viruses employ a common strategy for this fusion process based on the formation of hydrophobic, triple-helical coiled-coil structures that are thought to insert into the cell membrane and initiate fusion. The identification of highly conserved trimeric coiled-coil structures in the glycoproteins of several enveloped viruses (e.g. Ebola virus, HIV and influenzavirus) has led to a model for their conserved mechanism of action (reviewed in [1] ). In contrast to enveloped viruses, much less is known about the attachment and entry strategies used by nonenveloped viruses, a group that includes several important human pathogens. As is the case for enveloped viruses, identification of conserved structural features in nonenveloped viruses will likely facilitate an understanding of common mechanisms used by these viruses for attachment and cell entry.

The recently determined crystal structure of the reovirus attachment protein, 1, [2] reveals unanticipated similarities to the structure of fibre, the adenovirus attachment protein [3] . Although both adenoviruses and reoviruses are icosahedrally shaped viruses without envelopes, they differ dramatically in capsid composition and particle architecture. Most importantly, adenoviruses have a double-stranded (ds) DNA genome, whereas reoviruses contain dsRNA. Both viruses, however, interact with cell-surface receptors using trimeric, fibre-like molecules located at the 12 vertices of the virion icosahedron. The structural similarities between the attachment proteins of these viruses suggest conserved features of action and highlight a distant evolutionary relationship.

Reoviruses form nonenveloped, icosahedral particles that contain a segmented dsRNA genome [4] . The virions measure about 850 Å in diameter and are composed of eight structural proteins. The l1, l2, l 3, m2 and 2 proteins form the 'core' or inner capsid particle, whereas the viral outer capsid is formed by m1, 1 and 3. The m1 and 3 proteins comprise the bulk of the outer capsid. They are tightly associated, with m1 providing the membrane penetration machinery and 3 acting as a protective cap for m1 [5] . The 1 protein protrudes from the 12 vertices of the virion icosahedron [6, 7] and serves as the viral attachment protein [8, 9] . It is a long, fibre-like molecule with head-and-tail morphology and several discrete regions of flexibility [7] . The 1 tail partially inserts into the virion via 'turrets' formed by the pentameric l2 protein, while the 1 head projects away from the virion surface [6, 10] .

Crystal structures of the reovirus core [11] , isolated 3 [12] , the 3/m1 complex [5] , and a C-terminal fragment of 1 [2] have provided a detailed view of the organisation of the reovirus particle and the atomic structures of its key components. However, the structural changes that accompany reovirus attachment and entry, and its transition from the virion to subvirion intermediate and core particle [10] , are poorly understood. The 1 protein is key to precise delineation of these events as it facilitates viral attachment and undergoes major conformational changes during viral disassembly [6, 10] . The 1 protein binds two types of receptors and attaches to cells using an adhesionstrengthening mechanism [13] . A domain in the 1 tail binds to cell-surface carbohydrate [8] , which is known to be -linked sialic acid for serotype 3 reoviruses [14] [15] [16] [17] , and the 1 head binds to junctional adhesion molecule 1 (JAM1) [9] . The 1 protein plays a pivotal role in determining strainspecific disease patterns in animal models [18, 19] , most likely by selective recognition of cell-surface receptors. It also is responsible for strain-specific differences in the efficiency of virus-induced apoptosis [20] [21] [22] .

Structural analysis of the C-terminal half of 1 (residues 246-455), which includes the JAM1binding 1 head, reveals an elongated, trimeric structure with maximum dimensions of 120 Å in length and 50 Å in width [2] . Each monomer contains two domains: a slender tail and a compact head ( Figure 1 ). Residues 246-309 form the tail, a triple -spiral that contains sequence repeats characterised by conserved hydrophobic and glycine or proline residues [2, 3] . Each repeat consists of two short -strands connected by a four-residue -turn that has either a proline or a glycine residue at its third position. A surface-exposed, variable loop links successive repeats, and trimerisation generates the highly regular -spiral structure. The remaining 1 residues (310-455) assemble into a compact -barrel that forms the head. The 1 trimer features a distinct kink between the three-fold axes of the head and tail domains (Figure 1) . Although this kink is most likely introduced by crystal packing forces, it indicates that the 1 trimer possesses a high degree of flexibility between the second and third -spiral repeats in the tail.

Adenoviruses are important human pathogens responsible for a variety of gastrointestinal, ocular and respiratory infections [23] . They contain dsDNA genomes and form particles with a diameter of about 1000 Å [24] . The adenovirus capsid includes 240 copies of the trimeric hexon protein and 60 copies of the pentameric penton protein, which is located at the 12 vertices of the icosahedral particle. The penton protein forms the penton base complex, which serves as the anchor for the trimeric attachment protein, fibre, an elongated structure that protrudes from the virion surface. The fibre can be subdivided into a thinner 'shaft' and a more globular 'knob'. Adenovirus serotypes 2, 5 and 12 bind to the coxsackievirus and adenovirus receptor (CAR) [25, 26] by using a binding site located in the knob [27] [28] [29] . Crystal structures of the knob have been determined for serotypes 2 [30] and 5 [31] , as well as for the non-CAR-binding serotype 3 [32] . Structures are also available for a complex between the serotype 12 knob and the N-terminal domain of CAR [33] and for a trimeric fragment that comprises the knob and a portion of the shaft of the serotype 2 fibre [3] . The latter structure revealed that the shaft forms a triplespiral ( Figure 1 ). Although only a short -spiral was seen in the crystallised fragment, residues 45 to 392 of the serotype 2 fibre are predicted to form a continuous 22-repeat -spiral with residues 399-582 comprising the C-terminal knob [3] .

The overall appearance of the reovirus 1 protein is strikingly similar to that of the adenovirus fibre ( Figure 1 ). Moreover, the two proteins share many architectural features. Both are trimeric, elongated fibrous structures of roughly equivalent length that contain multiple repeats of an N-terminalspiral followed by a globular domain with eight major -strands ( Figure 1 ). The 1 tail is predicted to also contain an -helical coiled-coil N-terminal to the -spiral [2] , whereas the fibre shaft is most likely composed entirely of -spiral repeats [3] . The two proteins are to date the only structures known to contain such triple -spiral motifs, and thus they use a unique design for trimer formation. Most trimeric protein structures are based on -helical coiled-coils, in which three amphipathic -helices are wrapped around each other [2] and adenovirus fibre (right) [3] . The three monomers within each trimer are shown in red, orange and blue. Both proteins have head-and-tail morphology, with a triple -spiral forming the tail and an eight-stranded -sandwich domain forming the head. This and subsequent figures were prepared using the program RIBBONS [66] Reovirus and adenovirus attachment proteins Reovirus and adenovirus attachment proteins 125 125

to form a supercoil structure. Sequences that form these amphipathic -helices are characterised by a heptad repeat, in which non-polar residues are found at the first and fourth positions of the repeating unit [34] . Trimerisation motifs based on -sheet structures are far less common and have to date been observed in very few cases. For example, the bacteriophage T4 gp5 and gp12 proteins each contain triple -helices [35, 36] , trimeric structures formed by -strands that are roughly parallel to the trimer axis. The bacteriophage T4 gp12 protein also contains an additional trimeric structure that somewhat resembles the -spirals seen in reovirus 1 and adenovirus fibre, although it is not a true -spiral [35] . The surprising structural similarities between reovirus 1 and adenovirus fibre also extend to their C-terminal domains, both of which are globular structures that contain eight antiparallel -strands ( Figure 2) . A large number of domain folds are based on a core structure of eight antiparallel -strands, ranging from the ubiquitous -sandwich domains of the immunoglobulin (Ig) superfamily [37] to the circular -barrel of the E. coli outer membrane protein (omp) A [38] and the jelly-roll motif found in many viral coat proteins [39] . The principal difference between these folds lies in the manner in which the various -strands are connected (i.e. the 'topology' of each domain). Thus, analysis of the connectivity of a given fold serves to establish a structural relationship. The connectivity of -strands in the 1 head and the fibre knob is unique to these two proteins, demonstrating that they are more similar to each other than to any other protein for which structural information is available. This relationship is underscored by the observation that homology searches [40] with the coordinates of either domain yield the other as the only close structural homologue [2] . One difference between the two folds is that the -strands of the 1 head circularise to form a round -barrel, whereas the fibre knob is a -sandwich domain with two separate -sheets facing each other. The only other known circular -barrel featuring eight antiparallel strands is that of the ompA protein [38] , a membraneembedded domain with very different connectivity ( Figure 2) .

The remarkable homology between the 1 head and the fibre knob extends to other structural features such as a conserved short helix in a loop between -strands D and E ( Figure 2 ) and the spatial relationship of the -sandwich domains to the C-terminal repeat of the -spiral. However, there exist differences between the two domains in the surface loops. The 1 head features extremely short connections between most -strands (except the D-E loop), which results in a highly compact structure. The fibre knob contains about 30 additional residues in comparison to the 1 head, and these residues account for a more elaborate loop structure. This is not surprising as the two proteins bind different receptors.

The 1 trimer is a remarkably dynamic structure. A comparison of the three independent molecules present in the crystals reveals that a linear insertion into the -spiral (residues 291-294 of T3D 1) allows for substantial movement of the head with respect to the tail (Figure 3 ). This mobility is likely increased in the absence of constraining crystal packing forces, and it would allow the 1 head to 'sway' above the reovirus capsid in a physiologic context. Electron microscopic images of full-length 1 show flexibility in a region of the molecule that corresponds to this insertion and also in a second region close to the midpoint of the protein [7] . This latter region of flexibility, which is not included in the crystallised 1 fragment, correlates with the transition from the predicted N-terminal -helical coiled coil to the more C-terminal triple -spiral [2] . The adenovirus fibre does not have an insertion similar to that of 1 beneath the knob domain, and the four turns of -spiral that precede the knob are uninterrupted. However, residues 394-396 of fibre, which follow the last spiral turn in sequence, have poor electron density and are presumed to be flexible [3] . These poorly ordered residues appear to act as a hinge that allows the knob to move with respect to the shaft. The three-fold axes of the fibre knob and shaft differ by 2 [3] , and comparison of the three independent copies of the fibre molecule in the crystal structure demonstrates that this difference translates into significant mobility between the shaft and knob (Figure 3) . Thus, the trimeric knob can move as a whole with respect to the spiral, although the movement is somewhat more modest than that observed for 1. also has several other kinks within its -spiral [41, 42] , and these likely represent additional regions of flexibility.

Flexible regions in both proteins may serve a similar purpose by acting to facilitate interactions with receptors or structural rearrangements during viral disassembly. In the case of adenoviruses, shaft flexibility is a critical determinant of receptor binding as the rigid serotype 37 fibre fails to engage CAR despite the presence of the CAR-binding epitope in the knob [43] . Although similar analyses of 1 have not been performed, it seems probable that mobility between the globular and fibrous regions of both 1 and fibre is required for optimal positioning of the receptor-binding surfaces of these attachment proteins with their cell-surface receptors. It is striking that 1 and fibre use functionally equivalent hinge regions just beneath their globular head domains for this purpose.

Attachment proteins of reovirus and adenovirus are trimers that are anchored into the virion capsid via a pentameric structure, and therefore, they feature a highly unusual symmetry mismatch. The reovirus 1 trimer is inserted into a slot formed by the l2 pentamer, the structure of which is known [11] . The adenovirus fibre is inserted into the penton base, also a pentameric structure. Thus, in both cases a unique A 3 B 5 complex is formed. While symmetry mismatches have been seen in other protein complexes such as the AB 5 toxins [44] , the F1 ATP synthase [45] , or the VP1/ VP2 polyomavirus complex [46] , they are not common. Such mismatches usually indicate interactions of limited strength or specificity, and proteins involved in these mismatches often undergo rearrangement. They also are often hydrophobic in nature. AB 5 toxins such as cholera toxin undergo dramatic conformational changes upon membrane binding [44] , and this change involves exposure of the A subunit. The symmetry mismatch of three , three , and one subunit in F1 ATPase is crucial for its rotational mechanism of action, with the subunit forcing asymmetry onto the 3 3 pseudohexamer and providing a hydrophobic contact area that allows for efficient rotation [45, 47] . The structure of a polyomavirus VP1/VP2 complex, in which a single VP2 molecule engages a hydrophobic channel at the centre of the VP1 pentamer, suggests that some rearrangement must occur during viral entry in order to expose the myristylated VP2 N-terminus [46] . Thus, the conserved A 3 B 5 mismatch in both reovirus and adenovirus indicates a strong potential for structural rearrangement, which may occur upon receptor binding or disassembly. Moreover, the A 3 B 5 stoichiometry suggests that a conserved mechanism underlies these processes. To our knowledge, these structures are the only examples of A 3 B 5 complexes.

Both reovirus and adenovirus bind to receptors that belong to the Ig superfamily. In both cases, the N-terminal domain of the receptor is recognised by sequences in the globular C-terminal domain of the viral protein ( [29] and Forrest and Dermody, unpublished). Crystal structures of the adenovirus receptor CAR [48] and the murine homologue of JAM1 [49] indicate that both proteins form physiologically relevant homodimers. The CAR homodimer features extensive interactions between the GFCC' -sheets of the N-terminal Ig-like (D1) domains [48] (Figure 4) . The corresponding region of murine JAM1 also mediates homodimerisation, with an equally extensive contacting surface [49] (Figure 4 ). In addition, the relative orientations of the monomers in the CAR and JAM1 homodimers are strikingly similar (Figure 4) , demonstrating a close structural relationship between the two receptors.

Although dimeric structures of Ig superfamily members are not uncommon [50] [51] [52] [53] , CAR and JAM1 are the only molecules for which structural information is available that form physiologically relevant homodimers via the GFCC' -sheet of D1. The recently published structure of a coronavirus receptor suggests that it also may use a similar mode of dimerisation [54] . We note, however, that for several receptors of the Ig superfamily the GFCC' -sheet is engaged by viral ligands for complex formation. For example, the HIV glycoprotein gp120 binds to residues on the GFCC' face of the HIV-receptor CD4 [55] . Complexes of coxsackievirus B3 and rhinoviruses with their receptor intercellular adhesion molecule 1 (ICAM1) also demonstrate that residues at the top of D1 of 128 128

T. Stehle and T. S. Dermody T. Stehle and T. S. Dermody ICAM1, a region that includes part of the GFCC' face, mediate virus binding [56, 57] . The structure of a complex between adenovirus fibre knob and CAR shows that the heterodimeric knob-CAR interactions also involve the GFCC' -sheet of the CAR D1 domain [33] . In this last case, it appears that knob binding is incompatible with the homodimeric CAR structure shown in Figure 4 . However, it is not known whether the knob is able to disrupt the CAR dimer or simply recognises monomeric versions of CAR at the cell surface.

The origins of viruses are poorly understood, and establishing evolutionary relationships between them is difficult because of their rapid divergence in sequence. Three-dimensional structures change much more slowly than amino acid sequences, and structural analyses can sometimes reveal a previously unsuspected relationship between viruses. However, the biological relevance of such a link may not be apparent. The jelly-roll motif, for example, has thus far been found in a large number of viruses, but it is not associated with a conserved function. It therefore appears to serve as a scaffold used by many viruses to construct their capsids, similar to the Ig-like domain that is the structural motif of choice for many cell-surface receptors. For cases in which structural homologies parallel conserved function, an evolutionary relationship is more likely, and indeed several examples have been identified. The conservation of key structural and functional features between the alphavirus glycoprotein E1 [58, 59] and the flavivirus glycoprotein E [60] suggests a common progenitor. Structural similarities between the adenovirus hexon protein and the P3 capsid protein of bacteriophage PRD1 also have led to the suggestion of an evolutionary relationship between the two proteins [61] . In both cases, the structural similarities extend beyond the conservation of a single domain and translate into common functional properties. The same is true for the reovirus and adenovirus attachment proteins, and thus we think that the similarities between them point to a conserved mechanism of action and a distant evolutionary relationship, although we cannot exclude the possibility that the similarities have arisen through convergent evolution.

Since reovirus and adenovirus face similar challenges in the initiation of an infectious cycle, it is perhaps not too surprising that these viruses have evolved highly similar fibre-like proteins for attachment to host cells. Particles of both viruses are relatively large, and the long fibrelike tail may be required to allow the head to reach cell-surface receptors, which in both cases contain only two Ig-like domains and thus are located close to the cell membrane. Viruses typically engage membrane-distal regions of their receptors, and in many cases these virus-binding epitopes are found on more protruding receptors with more than two domains [55, 56, 62, 63] . Flexibility at a region just below the head domain may facilitate the engagement of receptors by allowing the head to position itself properly for a productive interaction. Thus, a globular head domain with [49] and adenovirus receptor CAR (right) [48] . Both proteins form dimers using the same interface (formed by -strands G, F, C, and C') of the membrane-distal Ig-like domain. The adenovirus fibre knob binds to a monomeric version of CAR [33] Reovirus and adenovirus attachment proteins Reovirus and adenovirus attachment proteins 129 129 numerous protruding loops offers the opportunity to modulate this interaction. Finally, the triplespiral creates a highly stable, regular trimeric structure that nonetheless allows for sequence variations and insertions in its surface-exposed loops. This design feature includes added functionality that is missing in the more common -helical coiled-coil trimerisation motif. In both reoviruses and adenoviruses, different serotypes are distinguished in part by insertions and deletions in the triple -spiral region, and these differences lead to altered surface properties. For example, thespirals of serotype 1 and 3 reoviruses contain binding sites for carbohydrate receptors in two distinct regions [8] , and differences in -spiral loops between adenovirus serotypes are at least partially responsible for serotype-specific antigenic determinants [3] . In both cases, the loops emanating from the -spiral structure are also likely to mediate contacts with other viral capsid proteins.

It is tempting to speculate that the numerous structural and functional parallels between the reovirus and adenovirus attachment proteins, which are reinforced by similarities between their receptors JAM1 and CAR, lead to conserved routes of entry and mechanisms of infection. We note that both receptor proteins are located at tight junctions. JAM1 was originally identified as an intercellular adhesion protein located at tight junctions and named accordingly [64] . CAR was shown recently to be a component of tight junctions between epithelial cells [65] . The fact that both viruses use tight junction proteins as receptors raises the possibility of conserved themes of receptor recognition, internalisation, tropism and disease. Thus, reovirus and adenovirus may have more in common than the striking structural similarities exhibited by their attachment proteins. Future studies will hopefully reveal the full extent of the fascinating relationship between these two pathogens.

Structural basis for membrane fusion by enveloped viruses

Crystal structure of reovirus attachment protein sigma 1 reveals evolutionary relationship to adenovirus fiber

A triple -spiral in the adenovirus fibre shaft reveals a new structural motif for a fibrous protein

Reoviruses and their replication

Harrison SC. Structure of the reovirus membrane-penetration protein, Mu1, in a complex with is protector protein, Sigma 3

Sigma 1 protein of mammalian reoviruses extends from the surfaces of viral particles

Molecular structure of the cell-attachment protein of reovirus: correlation of computerprocessed electron micrographs with sequence-based predictions

Identification of carbohydrate-binding domains in the attachment proteins of type 1 and type 3 reoviruses

Junction adhesion molecule is a receptor for reovirus

Early steps in reovirus infection are associated with dramatic changes in supramolecular structure and protein conformation: analysis of virions and subviral particles by cryoelectron microscopy and image reconstruction

Structure of the reovirus core at 3.6 Å resolution

Structure of the reovirus outer capsid and dsRNA-binding protein sigma 3 at 1.8 Å resolution

Utilization of sialic acid as a coreceptor enhances reovirus attachment by multistep adhesion strengthening

Effect of neuraminidase treatment of cells and effect of soluble glycoproteins on type 3 reovirus attachment to murine L cells

The alpha-anomeric form of sialic acid is the minimal receptor determinant recognized by reovirus

A sigma 1 region important for hemagglutination by serotype 3 reovirus strains

Dermody TS. Mutations in type 3 reovirus that determine binding to sialic acid are contained in the fibrous tail domain of viral attachment protein sigma 1

Molecular basis of reovirus virulence: role of the S1 gene

Absolute linkage of virulence and central nervous system cell tropism of reoviruses to viral hemagglutinin

Differences in the capacity of reovirus strains to induce apoptosis are determined by the viral attachment protein sigma 1

Reovirusinduced apoptosis of MDCK cells is not linked to viral yield and is blocked by Bcl-2

Reovirusinduced apoptosis requires activation of transcription factor NF-kappaB

Lippincott-Raven: Philadelphia

Lippincott-Raven: Philadelphia

Isolation of a common receptor for Coxsackie B viruses and adenoviruses 2 and 5

Receptor for the group B coxsackieviruses and adenoviruses: CAR

Cell-binding domain of adenovirus serotype 2 fiber

Characterization of the knob domain of the adenovirus type 5 fibre protein expressed in Escherichia coli

Coxsackievirus and adenovirus receptor amino-terminal immunoglobulin V-related domain binds adenovirus type 2 and fibre knob from adenovirus type 12

Structure of the human adenovirus serotype 2 fiber head domain at 1.5 Å resolution

Crystal structure of the receptor-binding domain of adenovirus type 5 fibre protein at 1.7 A resolution

Structure of the fiber head of Ad3, a non-CAR-binding serotype of adenovirus

Structural analysis of the mechanism of adenovirus binding to its human cellular receptor, CAR

Coiled coils: a highly versatile protein folding motif

Crystal structure of a heat and protease-stable part of the bacteriophage T4 short tail fibre

Structure of the cell-puncturing device of bacteriophage T4

The immunoglobulin fold. Structural classification, sequence patterns and common core

Structure of the outer membrane protein A transmembrane domain

The familiar and the unexpected in structures of icosahedral viruses

Protein structure comparison by alignment of distance matrices

Structure of adenovirus fibre. II. Morphology of single fibres

The fibre of bovine adenovirus type 3 is very long but bent

Structural analysis of a fiber-pseudotyped adenovirus with ocular tropism suggests differential modes of cell receptor interactions

AB(5) toxins: structures and inhibitor design

Reovirus and adenovirus attachment proteins Reovirus and adenovirus attachment proteins

Structure at 2.8 A resolution of F1-ATPase from bovine heart mitochondria

Interaction of polyomavirus internal protein VP2 with the major capsid protein VP1 and implications for participation of VP2 in viral entry

Mechanism of the F1F0-type ATP synthase, a biological rotary motor

Dimeric structure of the coxsackievirus and adenovirus receptor D1 domain at 1.7 A resolution

X-ray structure of junctional adhesion molecule: structural basis for homophilic adhesion via a novel dimerization motif

Structure of murine CTLA-4 and its role in modulating T cell responsiveness

Structural basis of cell-cell adhesion by NCAM

A dimeric crystal structure for the N-terminal two domains of intercellular adhesion molecule-1

Dimeric association and segmental variability in the structure of human CD4

Crystal structure of murine sCEACAM1a[1,4]: a coronavirus receptor in the CEA family

Structure of an HIV gp120 envelope glycoprotein in complex with the CD4 receptor and a neutralizing antibody

The structure of the two amino-terminal domains of human ICAM-1 suggests how it functions as a rhinovirus receptor and as an LFA-1 integrin ligand

Interaction of coxsackievirus A21 with its cellular receptor, ICAM-1

The fusion glycoprotein shell of Semliki Forest virus: an icosahedral assembly primed for fusogenic activation at endosomal pH

Locations of carbohydrate sites on alphavirus glycoproteins show that E1 forms an icosahedral scaffold

The envelope glycoprotein from tick-borne encephalitis virus at 2 A resolution

Viral evolution revealed by bacteriophage PRD1 and human adenovirus coat protein structures

Crystal structure of two CD46 domains reveals an extended measles virus-binding surface

The crystal structure of human CD21: implications for Epstein-Barr virus and C3d binding

Junctional adhesion molecule, a novel member of the immunoglobulin superfamily that distributes at intercellular junctions and modulates monocyte transmigration

The coxsackievirus and adenovirus receptor is a transmembrane component of the tight junction

Ribbon models of macromolecules

We thank Jim Chappell and Pierre Schelling for review of the manuscript and members of our laboratories for helpful discussions. We acknowledge support from Public Health Service awards AI45716 (T.S.) and AI38296 (T.S.D.) and the Elizabeth B. Lamb Center for Pediatric Research (T.S.D).