key: cord-0784837-npryvcy3
authors: Reguera, Juan; Mudgal, Gaurav; Santiago, César; Casasnovas, José M.
title: A structural view of coronavirus–receptor interactions
date: 2014-12-19
journal: Virus Res
DOI: 10.1016/j.virusres.2014.10.005
sha: 834f268f16aa77c2b85de55125d528b6a110dd9a
doc_id: 784837
cord_uid: npryvcy3

In the coronavirus (CoV), the envelope spike (S) glycoprotein is responsible for CoV cell entry and host-to-host transmission. The S is a multifunctional glycoprotein that mediates both attachment of CoV particles to cell surface receptor molecules as well as membrane penetration by fusion. Receptor-binding domains (RBD) have been identified in the S of diverse CoV; they usually contain antigenic determinants targeted by antibodies that neutralize CoV infections. To penetrate host cells, the CoV can use various cell surface molecules, although they preferentially bind to ectoenzymes. Several crystal structures have determined the folding of CoV RBD and the mode by which they recognize cell entry receptors. Here we review the CoV–receptor complex structures reported to date, and highlight the distinct receptor recognition modes, common features, and key determinants of the binding specificity. Structural studies have established the basis for understanding receptor recognition diversity in CoV, its evolution and the adaptation of this virus family to different hosts. CoV responsible for recent outbreaks have extraordinary potential for cross-species transmission; their RBD bear large platforms specialized in recognition of receptors from different species, which facilitates host-to-host circulation and adaptation to man.

For productive entry into host cells, viruses attach to specific cell surface receptor molecules (Casasnovas, 2013; Marsh and Helenius, 2006) . Selection of an entry receptor is governed by precise interactions that mediate efficient virus attachment to the cell surface as well as productive cell infection. Viruses can use a large number of cell surface molecules to penetrate host cells (Backovic and Rey, 2012) ; these molecules are the main determinants of virus tropism and pathogenesis. Receptor-binding motifs in viruses are subject to changes promoted by immune surveillance, which can target key receptor-binding residues during neutralization of virus infection. It is thus relatively common that a virus evolves to use distinct cell entry receptors over the course of an infection, or that related viruses use different cell surface molecules for host cell entry (Stehle and Casasnovas, 2009) . This is the case of coronavirus (CoV), whose use of distinct entry receptor molecules is responsible for their broad host range and tissue tropism (Gallagher and Buchmeier, 2001; Masters, 2006) . Some CoV have remarkable capacity for cross-species transmission which is linked to virus Fig. 1 . The CoV S glycoprotein and CoV cell surface receptors. Scheme of a CoV S with the functional domains in the S1 and S2 regions, which are exposed in the virus envelope. The N-terminal signal peptide (SP) and the transmembrane region (TM) are also shown. The N-terminal domain (NTD) that can act as a receptor-binding domain (N-RBD) and the canonical CoV RBD in the C-terminal portion of S1 are indicated. The heptad repeat regions (HR1 and HR2) and the putative fusion peptide (FP) are marked in S2. The arrowhead indicates the putative protease cleavage site in some CoV. Cell entry receptor molecules identified for the indicated CoV (Belouzard et al., 2012) are shown beneath their respective RBD regions. Sialic acids recognized by TGEV and IBV should be considered attachment factors. et al., 2003) . S2 contains the fusion peptide and two conserved heptad repeat regions, HR1 (N-terminal) and HR2 (C-terminal) ( Fig. 1) , which form a coiled coil structure important for S trimerization and the fusion reaction during CoV cell entry (Supekar et al., 2004; Xu et al., 2004) . The fusion peptide is N-terminal from the HR1 in the S2 sequence ( Fig. 1 ), but the HR1-HR2 coiled coil structure places it close to the transmembrane region. As in other enveloped viruses, the initiation of the fusion reaction requires partial disassembly of the trimeric spikes and the exposure of the fusion peptide for binding to the host cell membrane (Belouzard et al., 2012; Beniac et al., 2007; Harrison, 2005) . In some MHV variants and in the SARS-CoV, the S protein is processed into S1 and S2 fragments by cell proteases, which facilitate the fusion process and cell entry (Belouzard et al., 2012; Glowacka et al., 2011; Huang et al., 2006) . The S of alpha-CoV is not processed. Receptor-mediated endocytosis and exposure to low pH is a necessary step for entry of TGEV, hCoV-229E and SARS-CoV (Masters, 2006) . Other CoV, such as MHV and hCoV-NL63, do not require a low pH step for fusion, and the entry processes is mediated by receptor binding on the cell surface (Huang et al., 2006; Sturman et al., 1990) . CoV can thus follow different entry pathways to penetrate host cells (Belouzard et al., 2012) ; receptor, low pH and proteases are three major inducers of membrane fusion, and CoV use them differentially for cell entry. Mutations in the S1 and S2 fragments indicate that differences among CoV entry routes are probably related to variations in S trimer stability (Gallagher and Buchmeier, 2001) . Nonetheless, the conformational changes in the CoV S that lead to membrane fusion and cell entry have not been defined.

The S1 region is largely variable in sequence and length, and is specialized in recognition of cell surface receptors ( Fig. 1) (Li, 2012; Masters, 2006) ; it has several discrete modules or domains that can fold independently Du et al., 2013; Godet et al., 1994; Li et al., 2005a; Reguera et al., 2011; Wu et al., 2009) . Receptor-binding domains (RBD) can be located at the Nand/or C-terminal moieties of the S1 region (Li, 2012; Peng et al., 2011) (Fig. 1) . The S glycoprotein N-terminal domain (NTD) can function as a RBD (N-RBD); it can be the only S1 domain engaged in receptor recognition or, in conjunction with C-terminal RBD, can broaden tissue tropism of certain CoV. As entry receptors, the N-RBD can recognize sialic acids in some cases ( Fig. 1) (Peng et al., 2011) , whereas it binds to carcinoembryonic antigen cell adhesion molecules (CEACAM) in MHV (Williams et al., 1991) . The NTD in TGEV is responsible for its enteric tropism, absent in the related porcine respiratory CoV (PRCV) that lacks this domain (Sanchez et al., 1992) . The NTD region adopts a galectin-like structure in two beta-CoV, and its fold might be conserved in alpha-and gamma-CoV, since glycan-binding activity has been reported for the three genera (Li, 2012; Schultze et al., 1996) .

In most CoV, the major determinants of cell tropism are found in the C-terminal portion of the S1 region (Masters, 2006) . These RBD can usually fold independently of the rest of the S, and can be expressed as a single domain with all receptor-binding determinants (Du et al., 2013; Reguera et al., 2011; Wong et al., 2004; Wu et al., 2009) . Sequence and structure of the RBD vary considerably among CoV, and they recognize distinct entry receptors (Fig. 1) . Several CoV of the genus alpha, including TGEV and hCoV-229E, use aminopeptidase N (APN) for cell entry (Delmas et al., 1992; Yeager et al., 1992) , whereas hCoV-NL63 binds to the human angiotensin-converting enzyme 2 (ACE2) (Wu et al., 2009 ). In the beta-CoV, the SARS-and the MERS-CoV use ACE2 and dipeptidyl peptidase 4 (DPP4, CD26) receptors, respectively (Li et al., 2003; Raj et al., 2013) . APN, ACE2 and DPP4 are membrane-bound ectoenzymes with multiple functions such as angiogenesis, cell adhesion and blood pressure regulation (Boonacker and Van Noorden, 2003; Crackower et al., 2002; Mina-Osorio, 2008) . The three proteins catalyze peptide-bond hydrolysis of short peptides. The reason for CoV use of ectoenzymes as entry receptors is unclear; it might be linked to their abundance on epithelial cells rather than on their peptidase function, which does not appear to be essential for CoV cell entry (Li et al., 2005c) . Virus-binding regions in these ectoenzymes are distant from the catalytic site (Li et al., 2005a; Lu et al., 2013; Reguera et al., 2012; Wang et al., 2013; Wu et al., 2009) .

The identification of the CoV entry receptors and the RBD in the S glycoprotein led to structural characterization of the CoV-receptor interaction. RBD-receptor complexes have been determined for prototype alpha-(TGEV and hCoV-NL63) and beta-CoV (MHV, SARS-and MERS-CoV). RBD regions are targets of antibodies (Ab) that neutralize CoV infection, and their epitopes overlap receptorbinding motifs (Godet et al., 1994; He et al., 2005; Hwang et al., 2006; Pak et al., 2009; Prabakaran et al., 2006; Reguera et al., 2012) . Some structural studies have determined how neutralizing Ab prevent CoV cell entry and infection. In this review, we will summarize the currently determined CoV-receptor complex structures, highlighting the distinct receptor recognition modes in this virus family.

The alphacoronavirus (alpha-CoV) genus is a group of important animal and human viruses subdivided into several lineages (de Groot et al., 2011) . The alpha1 lineage comprises two types of canine (cCoV and cCoV-NTU336) and feline (fCoV and FIPV) CoV, (Reguera et al., 2012) . ␤-Strands (numbered) are shown in light or dark blue, coils in orange, and the helix in red; a ␤-bulge at ␤-strand 5 is in magenta. N-and C-terminal ends on the terminal side of the structure are indicated in lowercase letters. The Asn residues at glycosylation sites and the attached glycans defined in the structure are shown as a ball-and-stick model, with carbons in yellow. Cysteine residues and disulphide bonds are shown as green cylinders. The two ␤-turns at the ␤-barrel domain tip are labeled. Ribbon diagrams of the PRCV and hCoV-NL63 RBD structures are shown in B and C, respectively. The structures of these domains were determined in complex with the APN (PRCV, PDB ID 4F5C) and ACE2 (NL63, PDB ID 3KBH) (Reguera et al., 2012; Wu et al., 2009) . Receptor-binding surfaces in the RBD are shown in pink or red (tyrosine or tryptophan residues) and were generated by the RBD residues that contact the respective receptor molecules in the structures.

PRCV and TGEV; another lineage includes hCoV-229E and hCoV-NL63, and other members of the genus alpha are porcine epidemic diarrhea virus (PEDV) and some bat CoV.

TGEV, one of the most studied alpha-CoV, has enteric and respiratory tropism. The enteric tropism is linked to its NTD, since a deletion mutant of TGEV (the homologous PRCV) shows only respiratory tropism (Sanchez et al., 1992) . NTD binding to an attachment factor (sialic acid) is thought to be responsible for its enteric tropism (Schultze et al., 1996) . TGEV, PRCV and the related animal alpha1-CoV use APN for host cell entry (Fig. 1) . APN is also the receptor for hCoV-229E (Delmas et al., 1992; Yeager et al., 1992) , one of the first human CoV discovered, which is responsible for common colds (Kahn and McIntosh, 2005) . The related hCoV-NL63 does not bind to APN and recognizes the cell surface ACE2 ectoenzyme ( Fig. 1 ) (Smith et al., 2006) , like the SARS-CoV (Li et al., 2003) . The cell surface receptor of PEDV and other alpha-CoV are currently unknown.

The alpha-CoV RBD are modules of ∼150 residues that locate near the C-terminal portion of the S1 region ( Fig. 1) (Breslin et al., 2003; Godet et al., 1994; Wu et al., 2009) . The RBD can be expressed independently of the S; binding studies with receptors and Ab show that the RBD preserves its native conformation and binding specificity (Reguera et al., 2011; Wu et al., 2009) . Preparation of single RBD proteins facilitates their crystallization in complex with receptors and Ab.

The crystal structures of hCoV-NL63, PRCV and TGEV RBD have been determined (Reguera et al., 2012; Wu et al., 2009) . They show a single domain unit that has a ␤-barrel fold with two highly twisted ␤-sheets (Fig. 2) . In one ␤-sheet, three ␤-strands (␤1, ␤3 and ␤7) run parallel ( Fig. 2A) . The three RBD have three disulphide bonds. In the crystal structure of the TGEV RBD, solved at high resolution, the bent ␤-strand 5 (␤5) crosses both ␤-sheets ( Fig. 2A) . N-linked glycans cluster at one side of the ␤-barrel; the opposite side is not glycosylated and might be closer to other S protein domains. N-and C-terminal ends of the RBD are located on the same side of the domain (terminal side); at the opposite side, two ␤-turns form the tip of the barrel in the TGEV RBD ( Fig. 2A ). This region of the ␤-barrel domain contacts the receptor (see below) and its conformation in the APN-binding RBD of TGEV and PRCV differs from the ACE2-binding region in the hCoV-NL63 domain ( Fig. 2B and C). These differences probably determine the distinct receptorbinding specificities of alpha-CoV. The TGEV or PRCV RBD tips are formed by two protruding ␤-turns (␤1-␤2 and ␤3-␤4), each bearing a solvent-exposed aromatic residue (tyrosine or tryptophan) ( Fig. 2A and B ). In contrast, the hCoV-NL63 RBD tip has a slightly recessed conformation, with the aromatic residues at the center of the receptor-binding surface (Fig. 2C ).

Crystal structures have been reported for complexes of alpha-CoV RBD with the APN and ACE2 ectodomains (Reguera et al., 2012; Wu et al., 2009) . The RBD of these viruses contact receptor regions distal to the cell membrane (Fig. 3) .

The APN ectodomain is composed of four domains (DI-DIV), is heavily glycosylated and forms dimers through extensive DIV-DIV interactions ( Fig. 3A ) (Reguera et al., 2012) . Each APN monomer has an RBD bound in the crystal structure of the PRCV RBD-APN complex (Fig. 3A) . The bidentate, protruding tip contacts the APN, and the exposed side chains of the tyrosine and tryptophan residues penetrate small cavities of the APN ectodomain. The tyrosine side chain fits between an ␣-helix and a carbohydrate N-linked to the APN, whereas the bulky tryptophan is in a narrow cavity formed at the DII-DIV junction (Fig. 3A ). In addition to the tyrosine, other RBD residues contact the first N-acetyl glucosamine (NAG) linked to the porcine APN Asn736, and fix the glycan conformation. The CoV tyrosine and tryptophan residues are critical for TGEV RBD binding to APN (Reguera et al., 2012) , and preliminary results indicated that they are essential for virus entry and infection (unpublished data). CoV recognition of APN is species-specific, and specificity is associated to the APN N-linked glycan that interact with the RBD ␤1-␤2 turn in the structure (Reguera et al., 2012; Tusell et al., 2007) . Porcine, feline and canine alpha-CoV with a tyrosine at the ␤1-␤2 turn recognize APN proteins bearing the glycan. The large degree of sequence conservation in the RBD tip of alpha1-CoV also suggests a highly conserved APN recognition mode (Reguera et al., 2012) . hCoV-229E does not have a tyrosine in its RBD ␤1-␤2 turn, however, and it recognizes the human APN that lacks this glycosylation (Reguera et al., 2012) . Pig APN molecules are shown with domains in orange (N-terminal DI), yellow (DII), red (DIII) and green (C-terminal DIV), as well as the N-terminal ends near the putative location of the cell membrane. The RBD is shown as ribbon and surface drawings in blue and cyan, with the APN-binding tyrosine and tryptophan residues at the RBD tip in red. (B) Ribbon drawing of the hCoV-NL63 RBD-ACE2 complex (PDB ID 3KBH) (Wu et al., 2009 ). The ACE2 molecule is shown with the two lobes in green (N-terminal) and orange (C-terminal). The RBD is shown as ribbon and surface drawings in blue, with the ACE2-binding residue in pink and the aromatic residues that contact the receptor in red. The N-and C-terminal ends of the receptor molecules are marked in lowercase letters, N-linked glycans are shown as sticks with carbons in yellow, and the zinc ion at the catalytic sites of APN and ACE2 as cyan spheres. For A and B, details of key virus-receptor binding motifs are shown beneath the complex structures. Interaction of the PRCV RBD ␤1-␤2 and ␤3-␤4 turns (shown as sticks) at the domain tip with cavities in the APN (ribbon and surface drawings). The tyrosine at the ␤1-␤2 turn contacts APN residues and the NAG carbohydrate (yellow surface), which is N-linked to pig APN Asn736. The tryptophan side chain at the ␤3-␤4 turn penetrates between DII and DIV. Interaction of the concave center of the hCoV-NL63 RBD tip with the ACE2 ␤4-␤5 turn. Lys535 at the tip of the ACE2 turn is labeled. The ACE2 ␣-helices ␣1 and ␣10 contact the most exposed regions of the RBD loops. Sides chains of buried residues in the virus-receptor interfaces are shown with oxygens in red and nitrogens in blue in this and the following figures; hydrogen bonds are dark dashed lines. (Reguera et al., 2012; Tusell et al., 2007) . The conformation of this alpha-CoV RBD tip differs from that of alpha1-CoV, suggesting that hCoV-229E recognition of APN must be unique. It is nonetheless likely that this human alpha-CoV preserves a protruding tip for binding to small APN cavities. hCoV-NL63 RBD interacts with the ACE2 ectodomain opposite to the way that the alpha-CoV bind to APN. The hCoV-NL63 RBD has a blunt tip that contacts protruding regions of the receptor (Fig. 3B ). In the middle of the interacting surface, the depressed center of the RBD tip contacts a unique receptor ␤-turn (␤4-␤5), which interacts with a tyrosine and a tryptophan in the virus protein (Fig. 3B ). The rims of the RBD tip bind to two ␣-helices of the ACE2 receptor. Specificity is determined by several hydrogen bonds that engage amino and carbonyl groups in the main chains of the interacting molecules (Fig. 3B ).

Alpha-CoV use protruding RBD regions to bind APN or recessed surfaces to recognize exposed ACE2 motifs (Figs. 2 and 3). Crystal structures demonstrate that the conformation of the receptor-binding region in the alpha-CoV S must be the principal determinant of its receptor recognition specificity. We recently demonstrated that the RBD tip is a principal antigenic determinant (site A) in the S of TGEV and related alpha-CoV (Reguera et al., 2012) . Potent neutralizing Ab of porcine CoV cluster at site A (Delmas et al., 1990; Sune et al., 1990) . These Ab recognize the RBD tip and bind to the tyrosine or the tryptophan essential for APN binding (Reguera et al., 2012) . These data suggest that the conformation of the alpha-CoV receptor-binding region evolved under pressure from the immune system, particularly in humans, leading to small variations in the way hCoV-229E recognizes the APN protein (Tusell et al., 2007) or to radical changes that modified receptor specificity in hCoV-NL63.

The betacoronavirus (beta-CoV) genus comprises four lineages, A (MHV, hCoV-HKU1 and the beta1-CoV), B (SARS-CoV), C (batCoV and MERS-CoV), and D (batCoV-HKU9) (de Groot et al., 2011) . The most representative CoV prototypes of this genus are hCoV-OC43 (beta1-CoV), MHV, SARS-CoV and the recently identified MERS-CoV. Members of lineage A incorporate an extra, short spike-like glycoprotein in their envelope, the hemagglutinin esterase (HE) (Masters, 2006; Qinghong et al., 2008) .

hCoV-OC43 causes common cold and pneumonia in elderly populations, as well as severe lower respiratory tract infection in immunocompromised patients (Kahn and McIntosh, 2005) . Like bovine CoV (bCoV), another beta1-CoV, it uses sialic acids (N-acetyl-9-O-acetylneuraminic acid, Neu5, 9Ac2) as entry receptors ( Fig. 1 ) (Krempl et al., 1995) . Before SARS, MHV was the most studied beta-CoV in vitro and in vivo, especially in laboratory mouse. MHV strains cause specific inflammations in several mouse organs, such as the neurotropic strains JHM and A59 responsible for acute encephalitis and chronic demyelination in survivors, which serve as a model for the study of multiple sclerosis (Weiss and Leibowitz, 2011) . The MHV cell entry receptor is a member of the CEACAM family ( Fig. 1 ) (Williams et al., 1991) .

The SARS-CoV brought coronavirology to the center of the research community's attention due to a worldwide epidemic with very high mortality rates (Gallagher and Perlman, 2013) . It uses ACE2 as the entry receptor ( Fig. 1 ) (Li et al., 2003) . Epidemiologists believe that SARS virus originated in bats (natural reservoir) was then transmitted to palm civets, ferret badgers, and raccoon dogs (amplification and transmission hosts) and then introduced into man (Li et al., 2005b) . SARS-CoV adaptation to different species and its transmission to humans is linked to subtle changes in the S glycoprotein, which increased its binding affinity for human ACE2 (Li et al., 2005c) .

MERS-CoV emerged in Saudi Arabia a decade after the SARS epidemic. It shares 90% sequence identity with batCoV-HKU4 and -HKU5, and it docks in beta-CoV lineage C (de Groot et al., 2013) . Given this relationship, it is likely that MERS-CoV originated from bats . This CoV uses DPP4 as a cell entry receptor ( Fig. 1) (Raj et al., 2013) . BatCoV-HKU4 recognizes the human DPP4 protein, indicating possible direct transmission from bats to humans Yang et al., 2014) . Recent evidence nonetheless shows involvement of dromedary camels as intermediates in virus transmission from bats to man (Doremalen et al., 2014; Haagmans et al., 2014) . Human-to-human transmission is not frequent, probably because of low DPP4 expression in the human lower respiratory tract .

Several crystal structures show the folding of the SARS-CoV RBD, the mode by which this beta-CoV recognizes ACE2, and how Ab prevent virus binding to the receptor. These studies led to improved understanding of host-host transmission and adaptation of this CoV to humans, and also indicated strategies used by the SARS-CoV to evade neutralization by the immune system.

The SARS-CoV RBD is defined as a ∼200-residue fragment in the C-terminal portion of the S1 region ( Fig. 1) (Wong et al., 2004) . It is composed of two subdomains; the core has a central five-stranded ␤-sheet surrounded by polypeptides that connect the ␤-strands (Fig. 4A, yellow) . It has three small ␣-helices (A-C) and three disulphide bridges. A second subdomain of ∼65 residues inserts between two central ␤-strands of the core (␤4 and ␤7), and is distal to the terminal side of the domain (Fig. 4A , dark-red). This inserted subdomain lies on one side of the core and comprises a central two-stranded ␤-sheet connected by a long loop region; one side of this loop and the ␤-sheet clamp the core. The ␤-sheet, the extensive interactions with the core, and a disulphide bond in the most solvent-exposed region of the subdomain stabilize its structure (Fig. 4A) . One crystal structure of the isolated SARS-CoV RBD shows (Li et al., 2005a) , with the core subdomain in yellow and the inserted subdomain in dark red. The ␤-strands and ␣-helices are labeled with numbers and uppercase letters, respectively. Terminal ends are labeled in yellow and disulphide bonds in green; Asn residues at glycosylation sites and the attached glycans are shown as sticks, with carbons in yellow. SARS-CoV residues that bind to the ACE2 receptor and define the receptor-binding surface are pink. (B) Ribbon drawing of the SARS-CoV RBD-ACE2 complex (PDB ID 2AJF) (Li et al., 2005a) . ACE2 is shown as in Fig. 3B and the RBD as in panel A. The three main ACE2 regions recognized by SARS-CoV are labeled in green. (C) Key virus-receptor binding motifs. ACE2 residues are shown, with carbons in green. In the RBD, receptor-binding tyrosines and an arginine are shown, with carbons in pink, whereas the two critical residues for SARS-CoV adaptation to human ACE2 (Asn479 and Thr487) are shown, with carbons in magenta.

that it can form dimers through the terminal side (Hwang et al., 2006) . The dimerization surface in these crystals is relatively large (∼1000Å 2 buried surface area, BSA/monomer) and the authors proposed that RBD dimers could crosslink S glycoprotein trimers. It is nonetheless unclear whether such oligomers are found on the virus envelope and could recognize ACE2.

The ACE2 ectoenzyme is the cell entry receptor of SARS-CoV (Li et al., 2003) . It is a type I membrane glycoprotein with an N-terminal extracellular domain built of two ␣-helical lobes; the catalytic site with a coordinated zinc ion is located between the two lobes ( Figs. 3B and 4B ). The ACE2 ectodomain shows some conformational movement, and substrate binding to the active site leads to a closed conformation (Towler et al., 2004) . Drug binding to this active site does not affect SARS-CoV binding, in accordance with virus recognition of a single lobe (Li et al., 2005c) (Fig. 4B) .

The SARS-CoV RBD inserted subdomain is the main S glycoprotein receptor-binding motif (Li et al., 2005a) (Fig. 4) ; the ACE2-binding subdomain region forms a curved, elongated surface with the two-stranded ␤-sheet at the bottom (Fig. 4A) . The interaction buries 25 residues and about 860Å 2 of the virus protein, and a similar surface (820Å 2 ) of the ACE2 receptor. The ACE2interactive surface of the SARS-CoV RBD is ∼100Å 2 larger that of hCoV-NL63, consistent with marked differences in kinetic dissociation rate constants, which is an order of magnitude lower in SARS than in hCoV-NL63 (Li et al., 2005c; Wu et al., 2009 ). Both viruses recognize overlapping ACE2 regions, including the Nterminal ␣-helix (␣1) and the ␤-turn formed by ␤4 and ␤5 strands (Figs. 3B and 4B). The central concave SARS-CoV RBD surface cradles the ACE2 N-terminal ␣-helix, whereas the terminal side of the subdomain interacts with the ACE2 ␤4-␤5 turn and ␣10 ( Fig. 4B and C). The interaction includes at least 10 virus-receptor hydrophilic bonds, some of which engage the hydroxyl groups of RBD tyrosines that also mediate non-polar interactions with the receptor (Fig. 4C ). There is an important virus-receptor hydrogen bond interaction between the ACE2 Lys353 carbonyl and the main chain amino group of RBD Gly488 (Fig. 4C ) (Li et al., 2005a) . The lysine side chain amino interacts with RBD main chain carbonyl. This ACE2 lysine is absent in mouse and rat ACE2 proteins, which are not recognized by the SARS-CoV. ACE2 glycosylation is also a determinant of SARS-CoV species specificity (Li et al., 2005c) . A glycan linked to rat ACE2 Asn82 prevents its use as an efficient virus receptor. Deletion of the glycan and the His353/Lys substitution convert rat ACE2 into a SARS-CoV receptor, showing that efficient ACE2 recognition is central to virus infection and host-to-host transmission (Holmes, 2005; Li et al., 2005a,c) .

SARS-CoV emerged from bat CoV and was transmitted through palm civet CoV; cross-species transmission is linked to RBD changes that increased its affinity for human ACE2 (Holmes, 2005; Li, 2013; Li et al., 2005a,c) . Of the residues involved in SARS-CoV RBD binding to ACE2, only a few have a key role in SARS-CoV adaptation to man (Fig. 4C ). Lys479/Asn and Ser487/Thr mutations are two key changes in the SARS-CoV S glycoprotein for infection of human cells. Substitutions in one of these residues increases SARS-CoV RBD binding affinity to human ACE2 by 20-to 30-fold, whereas the double mutation has a synergistic effect, with a 1000-fold increase in affinity (Li et al., 2005c) . The Asn at position 479 is found in some civet CoV; it does not affect binding to civet ACE2, but increases SARS-CoV RBD affinity for the human protein (Li et al., 2005c) . Asn479 contacts the human ACE2 His34 and is relatively close to Lys31 in the N-terminal ␣-helix (Fig. 4C) , which are Tyr34 and Thr31 in civet ACE2. The presence of a positively charged lysine rather than Asn in RBD position 479 does not complement the human ACE2 Lys31 and His34 residues. The crystal structure of SARS-CoV (Hwang et al., 2006; Pak et al., 2009; Prabakaran et al., 2006) . The three RBD-Ab crystal structures were superimposed based on the RBD. The RBD is shown as in Fig. 4A and the variable domains of the Ab in green (80R, PDB ID 2GHW), blue (F26G19, PDB ID 3BGF) and cyan (m396, PDB ID 2DD8). RBD Ile489, which is recognized by the m396 and F26G19 Ab (Pak et al., 2009; Prabakaran et al., 2006) , is black. Side chains of residues that change in scape mutants to the neutralization are shown in red (Rockx et al., 2010) . RBD in complex with human ACE2 demonstrates that the methyl group of the threonine at position 487 establishes specific contacts with the ACE2 Tyr41 and Lys535 side chains, increasing affinity for the human protein (Fig. 4C ) (Li et al., 2005a) . The SARS-CoV that caused sporadic outbreaks in 2003-2004 has serine at position 487 and shows very poor human-to-human transmission. This phenotype was also associated with the Leu472/Pro substitution in the ACE2 contact region of the SARS-CoV RBD (Li et al., 2005a) . Other RBD residues have some influence on cross-species transmission of SARS-CoV (Li, 2013) .

The RBD is a major antigenic determinant in the S glycoprotein of the SARS-CoV (Du et al., 2009) . Potent human and mouse SARS-CoV neutralizing Ab target the RBD and prevent virus infection by blocking its binding to the ACE2 receptor (He et al., 2005; Zhu et al., 2007) . The RBD can elicit broadly neutralizing Ab against diverse isolates, and human monoclonal Ab (mAb) can protect from infection by various zoonotic and human SARS-CoV (He et al., 2006; Zhu et al., 2007) . Several conformational epitopes (I-VI) have been defined in the RBD, some of which are conserved in different species (He et al., 2006) . Epitopes of several neutralizing Ab have been identified by crystal structures of RBD-Ab complexes (Hwang et al., 2006; Pak et al., 2009; Prabakaran et al., 2006) , which show that they overlap with the receptor-binding region (Fig. 5) .

Neutralizing Ab bind to the RBD external subdomain that contacts ACE2 (Fig. 5) . The human mAb m396 is a potent neutralizing Ab of several zoonotic and human SARS-CoV (Zhu et al., 2007) ; it targets a region in the C-terminal side of the RBD inserted subdomain (residues 482-491) that is involved in ACE2 recognition, as well as residues in the RBD core (Fig. 5) (Prabakaran et al., 2006) . The mAb epitope includes RBD residues Ile489 and Tyr491, which contact the receptor directly. A very similar epitope was described for the mouse mAb F26G19 (Fig. 5) (Pak et al., 2009) , which binds to residues 486-492 of the RBD inserted subdomain and some regions of the core. Ile489 is a central residue in the F26G19 epitope (Fig. 5 , black). Epitopes of mAb m396 and F26G19 are thus very similar, and include an exposed ridge in the RBD ACE2-binding region (Fig. 5) ; this S region must be a hot spot for SARS-CoV neutralization.

The crystal structure of the human 80R mAb shows a distinct mode of SARS-CoV neutralization that also prevents virus binding to ACE2 (Fig. 5) (Hwang et al., 2006) . The 80R variable domains make extensive contact with the concave region of the RBD-inserted subdomain (Fig. 5) , mimicking the way that RBD and ACE2 interact. The 80R epitope in the RBD overlaps with the region buried by the N-terminal ␣-helix of the receptor. The total surface buried by the 80R-RBD interaction is larger than the ACE2-RBD surface and is responsible for its high affinity (in the nanomolar range). This mAb makes contact with 29 residues of the receptor-binding subdomain, 17 of which are involved in ACE2 recognition (Hwang et al., 2006) .

All three SARS-CoV-neutralizing mAb epitopes overlap with the receptor-binding region in the S protein (Fig. 5) ; efficient virus neutralization is thus achieved by targeting receptor-binding residues and blocking virus binding to ACE2 and thus, cell entry. Virus mutants have been identified that escape mAb neutralization, although these mutants usually cause attenuated infection (Rockx et al., 2010) ; some of the escape mutations map to the RBD inserted subdomain (Fig. 5) and probably affect SARS-CoV binding to ACE2. (Coleman and Frieman, 2014) ; it is thought to have emerged from bats and is transmitted to humans via dromedary camels. Crossspecies transmission is determined mainly by the adaptability of this beta-CoV for different hosts, mediated by subtle modifications in its envelope S protein. MERS and SARS-CoV RBD are structurally similar (Fig. 6) , but use different cell entry receptors; MERS-CoV attach to a distinct ectoenzyme, DPP4 (Raj et al., 2013) . Several crystal structures have defined MERS-CoV RBD and how it binds to its DPP4 receptor Lu et al., 2013; Wang et al., 2013) .

The MERS-CoV RBD is a fragment in the S1 region C-terminal portion (Fig. 1) ; its structure is remarkably similar to the SARS-CoV RBD ( Fig. 6 ) (rmsd of 2.4Å for 132 residues), although they show little sequence identity. The MERS-CoV RBD also has two subdomains (Fig. 6A) , the core with a central five-stranded ␤-sheet and three disulphide bridges, as well as an inserted or external subdomain between two core ␤-strands Lu et al., 2013; Wang et al., 2013) . The central ␤-sheet of the core is surrounded by polypeptides that connect the ␤-strands and contain helical structures (Fig. 6A) . The core has an overall globular shape. The inserted subdomain is distal from the RBD terminal side and has a four-stranded ␤-sheet (Fig. 6A) . The ␤-sheet and a long loop that connects the ␤-strands at one edge of the sheet clamp the core subdomain, as in the SARS-CoV RBD (Fig. 4A ). The cores are more similar in MERS-and SARS-CoV than the external subdomain (Fig. 6B) , which is longer in the MERS (80 residues) than the SARS RBD (65 residues). Because of the extended ␤-sheet, the solventexposed region of the inserted subdomain is broader than that of SARS-CoV. The first (␤6) and last (␤9) ␤-strands of the MERS-CoV inserted subdomain align with the two ␤-strands of the SARS-CoV inserted subdomain, but the other two ␤-strands (␤7 and ␤8) are absent in the SARS RBD (Fig. 6B) . The MERS-CoV inserted subdomain contains a concave surface or small "canyon" formed by the ␤-strands and the loop that connect ␤6 and ␤7 (Fig. 6A) . This "canyon" is very distant from the terminal side and exposed for receptor recognition. It is absent in the SARS-CoV RBD, which contains a long loop in this location (Fig. 6B) . Likely, these differences in the external subdomains are the major determinants of the distinct receptor-binding specificity between the MERS-and SARS-CoV.

DPP4 or CD26 is a multifunctional membrane-bound serine protease (Boonacker and Van Noorden, 2003) . DPP4 is a type II membrane protein that forms homodimers on the surface of different cells (Fig. 7A) . The DPP4 ectodomain has ∼730 amino acids and is composed of two domains, an ␣/␤-hydrolase domain and an eight-bladed ␤-propeller (Fig. 7A) . The substrates bind to a pocket in a central cavity formed between the two domains (Boonacker and Van Noorden, 2003) . The MERS-CoV contacts only the ␤propeller domain (Fig. 7A, green) .

Crystal structures of the MERS-CoV RBD bound to DPP4 demonstrate that the virus attaches to the most membrane-distal region of the ␤-propeller Wang et al., 2013) . One RBD binds to each of the DPP4 monomers in the dimer, away from the receptor dimerization interface (Fig. 7A ). This dimeric virus-receptor complex is similar to the alpha-CoV RBD-APN structure described above . The DPP4 monomers are shown with the N-terminal ␤-propeller domain in green and the C-terminal ␣/␤-hydrolase domain in orange. The RBD molecules are as in Fig. 6A . Labels and glycosylation are as in previous figures. (B) Key virus-receptor binding motifs. The virus-binding DPP4 ␤-propeller blades IV and V are shown in light and dark green, respectively. DPP4 residues are shown, with carbons in green. In the RBD, residues in the small "canyon" that interact with the exposed ␣-helix in the blade linker are shown, with carbons in magenta, whereas those that bind to the DPP4 N-linked glycan (Asn229) are shown, with carbons pink. Some residues in the two receptor-binding motifs and the external subdomain ␤-strands (␤6 − ␤9) are labeled. (Fig. 3A) . The bound RBD does not appear to interfere with DPP4 catalytic activity, binds only to the ␤-propeller domain and away from the regions at which the substrate accesses the active site.

The MERS-CoV RBD engages the DPP4 molecule through the solvent-exposed side of its external subdomain (Figs. 6A and 7A) . It contacts the edges of DPP4 ␤-propeller blades IV and V, including N-linked carbohydrates at blade IV and a helix at the linker between the two blades (Fig. 7B) . It is the largest CoV-receptor interface, and buries 32 residues of the RBD (∼1110Å 2 surface) and of DPP4 (∼1240Å 2 surface). In the two structures reported (PDB ID 4KRo and 4L72) Wang et al., 2013) , the interaction includes between 9-14 hydrogen bonds and 2-3 salt bridges.

A key spot in this virus-receptor interaction includes the RBD contact with the helix that bulges out at the N-terminus of blade V (Fig. 7B ). The small "canyon" in the inserted subdomain cradles the DPP4 helix. This helix contains mostly hydrophobic residues (Ala291, Leu294, Ile295) that lie on a hydrophobic patch in the RBD "canyon", composed of the side chains of Lys502, Leu506, Tyr540, Arg542, Trp553 and Val555, residues located in the three main ␤strands of the subdomain (Fig. 7B) . The side chain amino groups of Lys502 and Arg542 are hydrogen-bonded to the main chain of DPP4. The loop at one rim of the small "canyon" forms polar interactions with the DPP4 ␤-strands in blade V (Fig. 7B ).

An interesting feature of MERS-CoV binding to DPP4, also shown in the PRCV-APN complex (Fig. 3A, bottom) , is the RBD interaction with N-linked receptor carbohydrates (Fig. 7B ). The first three carbohydrates attached to DPP4 Asn229 are well defined in the crystal structures of the MERS-CoV RBD-DPP4 complex Wang et al., 2013) . They interact with several solventexposed residues in the virus protein (Fig. 7B ). The first NAG residue is hydrogen-bonded to RBD Glu536, whereas the second NAG of the glycan stacks onto the aromatic ring of viral Trp535, which strengthens the glycan-virus interaction and probably stabilizes the motif conformation. The third mannose residue in the DPP4 N-linked glycan also interacts with the RBD tryptophan. Another glycan at DPP4 Asn281 in blade IV is very close to the RBD (not shown), but does not interact with the virus protein. The conformation of this last glycan appears to be determined by its interaction with a tryptophan residue (Trp187) in the DPP4 protein Wang et al., 2013) , and could be critical for MERS-CoV RBD binding to DPP4. A highly flexible glycan in this position could prevent this virus-receptor interaction, such as shown for glycosylations in APN or ACE2.

The S glycoprotein NTD can mediate attachment of CoV particles to cell surface molecules (Peng et al., 2011 (Peng et al., , 2012 Schultze et al., 1996; Tsai et al., 2003) . It thus function as an RBD (N-RBD) in certain CoV. The crystal structure of the MHV NTD shows a galectinlike fold (Fig. 8) (Peng et al., 2011) ; the homologous bCoV NTD also has a galectin fold (Peng et al., 2012) . The galectins are a family of lectins with a common ␤-sandwich carbohydrate recognition domain (CRD) (Fig. 8A ). They preferentially recognize N-acetyl lactosamine in cell surface proteins, which binds to conserved residues on one of the CRD ␤-sheets (Fig. 8A) . The CoV NTD is also composed of a central ␤-sandwich formed by two long ␤-sheets with six and seven ␤-strands that is structurally similar to the galectin CRD (Fig. 8B) .

The CoV is thought to incorporate this N-terminal galectin-like domain from the host (Peng et al., 2012) . In several CoV such as TGEV, beta1-CoV and IBV, the NTD preserves glycan binding activity, whereas in MHV it binds to a protein, CEACAM1 (Peng et al., 2012; Tsai et al., 2003) . The CoV NTD has diverged from galectins and recognizes proteins or sialic acids rather than N-acetyl lactosamine; the mode of ligand recognition also differs (Peng et al., 2012) . Although the side of the NTD that recognizes cell surface molecules is the same side as the galectin CRD that binds carbohydrates, the top of the carbohydrate-binding ␤-sheet is covered by polypeptides that shape the receptor-binding region in CoV (Figs. 8 and 9A ). In addition, a glycan N-linked to one edge of the ␤-sheet further prevents ligand binding to the carbohydratebinding sheet in galectins. This region is similar in MHV, which binds CEACAM1, and in bCoV, which binds sialic acid, showing that the NTD has evolved in CoV to specifically select cell entry receptors. (Seetharaman et al., 1998) . The ␤-strands in the ␤-barrel are in light or dark blue, and a galactose ligand on the top of the ␤-sheet is shown as sticks, with carbons in yellow. N-and C-terminal ends are indicated in lowercase letters. (B) Ribbon drawing of the MHV NTD structure (PDB ID 3R4D) (Peng et al., 2011) . The ␤-strands in the central galectin-like ␤-barrel are in light or dark blue, and those on the top of the sheet are in pink. The Asn residues at glycosylation sites and the attached glycans defined in the structure are shown as sticks, with carbons in yellow. Cysteine residues and disulphide bonds are shown as green sticks.

MHV is a prototype beta-CoV of the A lineage. It uses CEA-CAM proteins to enter host cells (Williams et al., 1991) . CEACAM are type I membrane proteins of the immunoglobulin superfamily (IgSF), markers of colorectal tumors that contribute to tumorigenesis (Beauchemin et al., 1999) ; in contrast to other CoV receptor proteins, they are not peptidases. The CEACAM mediate homo-and heterophilic cell adhesion. There are two murine CEACAM genes, CEACAM1 and CEACAM2. CEACAM1 has four splice forms, which have two (D1, D4) or four (D1-D4) Ig-like domains in the extracellular region, as well as a transmembrane region and two distinct cytoplasmic tails (Beauchemin et al., 1999) . All four CEACAM1 variants can be used as entry receptors by MHV (Dveksler et al., 1993) . CEACAM1 is also a receptor for virulent Neisseria strains .

CEACAM1 is a member of the IgSF, and the MHV S protein recognizes the N-terminal Ig-like domain 1 (D1), which adopts a variable (V) fold (Tan et al., 2002) . The virus interacts with the CFG ␤-sheet of D1 (Fig. 9B) , the surface commonly engaged in intermolecular interactions by cell surface molecules of the IgSF. The CFG ␤-sheet is formed by the ␤-strands C, C , C on one side and the ␤-strands F and G on the other (Fig. 9B) . About 25 receptor residues, 770Å 2 of its surface, are buried by the MHV protein. Most of the virus-binding residues locate at the D1 C edge and around the FG loop. CEACAM1 has a unique CC loop that protrudes from the CFG ␤-sheet of the Ig-like domain (Tan et al., 2002) . This is a key structural determinant for CEACAM1 recognition by the MHV S protein (Peng et al., 2011) .

The CEACAM1-binding surface is on top of the galectin-like ␤-sandwich in the MHV N-RBD (Fig. 9A) ; its N-terminal portion occupies the top of the receptor-binding surface and contributes 50% of the 24 MHV residues buried by interaction with the receptor. The N-terminal residues form a "socket" that contains a hydrophobic amino acid, Leu160, at the bottom (Fig. 9A) . Ile41 of CEACAM1 is exposed in the D1 CC loop and penetrates the socket (Fig. 9B ). MHV Tyr15, Leu89 and Leu160 contact the Ile41 side chain (Fig. 9C) , and comprise a critical virus-receptor motif (Peng et al., 2011; Tan et al., 2002) . Surrounding residues in the CEACAM1 CC loop, Thr39 and Asp42, form hydrogen bonds with the MHV N-RBD (Fig. 9C) , which confirms the importance of this receptor region in virus recognition.

The N-terminal portion of the MHV N-RBD also contacts other motifs in the C edge of D1. In the C ␤-strand, CEACAM1 Arg47 contributes to binding and establishes hydrogen bonds with the main chain carbonyl oxygens of MHV N-terminal residues. Up to 10 polar interactions contribute to virus-receptor specificity. MHV N-terminal residues interact extensively with the receptor C ␤strand, which runs parallel to the ␤1-strand of the virus domain. CEACAM1 Phe56 in the C ␤-strand appears to be an important residue for the interaction and establishes van der Waals contacts with the virus protein. Another important receptor-binding motif surrounds MHV Leu174 and contacts the loops at the top of the CFG ␤-sheet (Fig. 9B ). This N-RBD region protrudes slightly and is distant from the socket.

The crystal structure of the MHV NTD in complex with CEACAM1 shows how the N-terminal module of a CoV S recognizes a protein receptor. This region has been implicated in the recognition of sialic acids in alpha-(TGEV), beta-(bCoV) and gamma-(IBV) CoV (Fig. 1) . The NTD of these CoV were proposed have a similar fold, which was confirmed by the crystal structure of the bCoV NTD (Peng et al., 2012) . As in the MHV structure (Fig. 8) , the bCoV NTD has polypeptides on the top of the galectin-like ␤-sandwich. The bCoV NTD structure nonetheless lacks the MHV NTD socket, a critical motif for CEACAM1 binding. Differences in the conformation of exposed NTD regions could be responsible for the distinct receptor-binding specificity observed among CoV that use the N-terminal module to bind to cell surface molecules.

The structural studies reviewed here established the basis for understanding receptor recognition diversity in CoV, its evolution and its adaptation to different hosts. CoV RBD folding, conformation of receptor-binding motifs and subtle changes in those motifs determine receptor binding specificity and CoV host range. Two domains of the multifunctional CoV S glycoprotein anchor the virus particles to cell surface molecules for virus penetration of cells (Fig. 1) . The two domains might be exposed in the S1 region for CoV binding to host cell entry receptors (Fig. 10) .

The S glycoprotein NTD can function as an RBD in certain CoV (Fig. 1) , and might have a conserved fold in alpha-, beta-and gamma-CoV (Peng et al., 2012) . This domain has a galectin-like core, which indicates it was incorporated into the CoV S from a host (Li, 2012; Peng et al., 2011) . It has evolved in some CoV to recognize cell surface molecules such as sialic acids, or in MHV to bind the (Peng et al., 2011) . The CEACAM1 N-terminal D1 is shown in green, with the ␤-strands in the receptor-binding CFG ␤-sheet labeled. The side chain of CEACAM1 Ile41 that penetrates the NTD socket is shown as spheres. The MHV Leu160 in the socket and Leu174 that contacts the top of D1 are in red. (C) Key virus-receptor binding motifs. Side chains of some receptor-binding MHV residues are shown, with carbons in pink; the hydrophobic residues in the bottom of the socket and Leu174 are in magenta; the CEACAM1 residues are in green. Ile41 in the CC loop, the most important virus-binding motif in CEACAM-1 (Peng et al., 2011) , is shown as spheres.

CEACAM1 protein (Fig. 1) . CoV NTD has integrated polypeptides and an N-linked glycan on the top of the flat galectin-like ␤-sandwich, which covers the galactose-binding ␤-sheet in galectins (Fig. 8) . The virus-specific conformation of the polypeptides at the top of the NTD probably determine its receptor-binding specificity (Peng et al., 2012) . The MHV NTD contains a socket for specific recognition of a unique structural feature in the CEACAM1 D1 (Fig. 9) . Acquisition of the galectin-like NTD from the host probably expanded CoV host cell tropism, as shown for the TGEV NTD that confers enteric tropism (Schultze et al., 1996) , although MHV and related beta-CoV only use the NTD for recognition of cell surface molecules (Fig. 1) . The receptor-binding function of the S1 C-terminal portion appears to have been lost in these CoV. It would be interesting to explore the conformation of this region, which could provide clues to its presumed lack of function. Fig. 10 . Structural view of the multifunctional CoV S binding to host cell surface receptors. The two domains, NTD and RBD, of the S1 region that CoV use for attachment to cell surface molecules (Fig. 1) docked into the cryo-electron microscopy map (gray) of the trimeric SARS-CoV S (EMD-1423) (Beniac et al., 2006) . Ribbon representations of the SARS-CoV RBD (yellow) and the MHV NTD (blue) alone or bound to ACE2 (Fig. 4B ) and to CEACAM1 D1 (Fig. 9B) , respectively.

The S1 C-terminal RBD have unique structures unrelated to host proteins Li et al., 2005a; Reguera et al., 2012; Wu et al., 2009 ) and can thus be considered genuine CoV RBD. Alpha-and beta-CoV RBD adopt two distinct folds, but bind only to ectoenzymes; the NL63-and SARS-CoV bind to the same protein, ACE2 (Fig. 1) . Some of these enzyme features must be essential for CoV entry into host cells. Perhaps they cluster with other proteases that facilitate fusion (Glowacka et al., 2011) . ACE2, APN and DPP4 have distinct structures and functions, but their ectodomains share an inherent conformational flexibility (Boonacker and Van Noorden, 2003; Towler et al., 2004; Xu et al., 1997 ) that could assist in dissociation of the S1-S2 heterotrimer. Trimeric spikes that bind simultaneously to several receptor molecules could disassemble by pulling forces generated during ectodomain movement. The conformation and dynamics of the APN ectodomain vary with the pH (unpublished data), so that endosomal acidification can alter APN conformation during receptor-mediated endocytosis.

Alpha-CoV RBD adopt a conserved ␤-barrel fold (Fig. 2) (Reguera et al., 2012; Wu et al., 2009 ). S1 C-terminal fragments of the IBV gamma-CoV and the bulbul delta-CoV share certain sequence similarity with the alpha-CoV RBD, and could have a similar fold (Reguera et al., 2012) . Crystal structures of alpha-CoV in complex with receptors identified the receptor-binding regions in the RBD (Reguera et al., 2012; Wu et al., 2009 ), which have remarkable structural variability (Fig. 2) . The conformation of the ␤-barrel tip dictates the receptor molecule used by alpha-CoV for host cell entry. RBD with protruding tips determine alpha-CoV attachment to APN, whereas those with blunt RBD tips recognize ACE2 and perhaps other yet uncharacterized receptor molecules. Structures of alpha-CoV RBD in complex with APN or ACE2 show two opposite modes of CoV-receptor recognition (Reguera et al., 2012; Wu et al., 2009) (Fig. 3) . In viruses, recessed surfaces hide conserved receptorbinding residues from antibodies (Casasnovas, 2013; Rossmann, 1989) ; hCoV-NL63 uses a recessed surface to recognize exposed ACE2 motifs, following a receptor-binding strategy similar to the other beta-CoV reviewed here. CoV binding to APN is unique among CoV, and contrasts with the mode of ACE2 and DPP4 recognition. The bidentate, protruding RBD tip of alpha1-CoV, which has two exposed aromatic residues, penetrates small cavities of the APN ectodomain (Fig. 3A) . Similarly to MERS-CoV, the alpha1-CoV also recognizes a dimeric cell surface protein.

The folding of the MERS-and SARS-CoV RBD are similar (Fig. 6 ) Li et al., 2005a; Lu et al., 2013; Wang et al., 2013) . Both have a core with a single ␤-sheet and an additional subdomain that recognizes cell entry molecules. MERS-and SARS-CoV show extraordinary potential for cross-species transmission, related to S binding to distinct orthologous receptor molecules. This is probably linked to the specific structure of their RBD, especially to the extended receptor-binding surfaces of the inserted subdomains (Figs. 4A and 6A) . A few changes in those large surfaces increase affinity for receptor molecules in new hosts, while preserving virus growth (Holmes, 2005; Li et al., 2005c) . Measles virus (MV) follows a similar strategy for recognition of several receptors that facilitate virus growth and transmission (Casasnovas, 2013) . The MV hemagglutinin uses a broad concave surface to bind to three distinct receptor molecules, a unique feature of MV among the paramyxoviruses. The use of a large receptor recognition surface probably enables virus dissemination in tissues and host-to-host virus transmission.

The DPP4-binding surface in the MERS-CoV is larger (∼300Å 2 ) than the ACE2-binding surface in SARS-CoV, which correlates with a larger RBD inserted subdomain. The two CoV use concave surfaces to bind different receptors. MERS-CoV uses a small "canyon" to bind to an ␣-helix in the linker between blades IV and V of the DPP4 ␤-propeller (Fig. 7) , whereas the curved inserted subdomain in SARS-CoV RBD cradles the N-terminal ␣-helix of ACE2 (Fig. 4) . The mode by which these CoV bind to receptors shows similarities to other CoV-receptor interactions, particularly to hCoV-NL63, which also binds to ACE2 (Fig. 3B) . NL63-and SARS-CoV recognize overlapping ACE2 regions, including two helices and a ␤-turn in the virus-binding lobe of the receptor. The ACE2-binding surfaces in both CoV are concave and are distant from the terminal end of the RBD. The receptor-binding surface in SARS-CoV is more extended and curved than in hCoV-NL63, and interacts more extensively with the ACE2 N-terminal ␣-helix; the two residues involved in SARS-CoV adaptation to humans (Asn479 and Thr487) interact directly with the ␣-helix.

MERS-and alpha1-CoV share recognition of carbohydrates Nlinked to their receptors (Figs. 3A and 7B) Reguera et al., 2012; Wang et al., 2013) . In APN, the N-linked glycan is essential for binding and infection of TGEV and related alpha-CoV (Reguera et al., 2012; Tusell et al., 2007) . Receptor glycosylations are important determinants of CoV-receptor recognition, as they can promote or hinder CoV binding to cell entry receptors in certain species (Holmes, 2005; Tusell et al., 2007) , which delimits CoV host range.

The CoV RBD is a major target of neutralizing Ab that prevent virus infections by blocking virus binding to receptors (Hwang et al., 2006; Pak et al., 2009; Prabakaran et al., 2006; Reguera et al., 2012; Zhu et al., 2007) . RBD protein can elicit potent neutralizing Ab and protective immune responses (Du et al., 2009 ). These neutralizing Ab recognize the exposed receptor-binding tyrosine or tryptophan in TGEV or PRCV (Reguera et al., 2012) . In the SARS-CoV, structural studies showed that several neutralizing Ab bind to the receptor-binding subdomain (Fig. 5) (Hwang et al., 2006; Pak et al., 2009; Prabakaran et al., 2006) . These results indicate that the receptor-binding regions are under selective pressure from the immune system. In alpha-CoV, this pressure could mediate the notable conformational changes in the RBD tip (Fig. 2) , which alter receptor-binding specificity. The APN-binding tip in alpha-CoV RBD has exposed receptor-binding residues that are easily targeted by Ab, whereas the recessed ACE2-binding tip in hCoV-NL63 more efficiently hides conserved receptor-binding residues from immune surveillance.

Apha-and beta-CoV RBD folds are distinct but are both unique, with no known homology to host domains Li et al., 2005a; Reguera et al., 2012; Wu et al., 2009) . They are thought to have evolved from a common CoV RBD ancestor (Li, 2012) . They share some common features, such as recognition of glycans Nlinked to receptors, and the presence of parallel ␤-strands (␤1, ␤3, ␤7 in TGEV and ␤2, ␤11 in MERS, Figs. 2A and 6A ). It is tempting to speculate that this precursor RBD had a ␤-barrel fold similar to the alpha-CoV, with a variable tip that accommodated different receptor molecules. In SARS and MERS beta-CoV, the RBD lost the ␤-barrel fold, but maintained two ␤-sheets, one of which forms a large receptor-binding platform with recessed surfaces that bind to specific motifs in cell surface molecules. The receptor-binding subdomains in SARS and MERS beta-CoV appear to specialize in recognition of orthologous receptor molecules. The beta-CoV RBD probably evolved to enhance host-to-host transmission, responsible for the recurrent CoV outbreaks in man.

Structural studies reviewed here have established the basis for understanding receptor recognition diversity in CoV, its evolution and adaptation to different hosts. These studies have identified sites of vulnerability in the CoV S that should guide the development of anti-virals and vaccines to prevent CoV infections.

Buried surfaces and residues at the molecular complex interfaces were determined with the PISA server (http://www.ebi.ac.uk/ msd-srv/prot int/pistart.html). Fig. 2A was prepared with Ribbons (Carson, 1987) , Fig. 10 with Chimera (Pettersen et al., 2004) and the other structure representations with PyMOL software (pymol.org).

Virus entry: old viruses, new receptors

Redefined nomenclature for members of the carcinoembryonic antigen family

Mechanisms of coronavirus cell entry mediated by the viral spike protein

Architecture of the SARS coronavirus prefusion spike

Conformational reorganization of the SARS coronavirus spike following receptor binding: implications for membrane fusion

Identification of a receptor-binding domain of the spike glycoprotein of human coronavirus HCoV-229E

The multifunctional or moonlighting protein CD26/DPPIV

The coronavirus spike protein is a class I virus fusion protein: structural and functional characterization of the fusion core complex

Human coronavirus 229E: receptor binding domain and neutralization by soluble receptor at 37 degrees C

Ribbon models of macromolecules

Virus-receptor interactions and receptor-mediated virus entry into host cells

Crystal structure of the receptor-binding domain from newly emerged Middle East respiratory syndrome coronavirus

Coronaviruses: important emerging human pathogens

Angiotensin-converting enzyme 2 is an essential regulator of heart function

Virus Taxonomy: Ninth Report of the International Committee on Taxonomy of Viruses

Middle East respiratory syndrome coronavirus (MERS-CoV): announcement of the Coronavirus Study Group

Aminopeptidase N is a major receptor for the entero-pathogenic coronavirus TGEV

Four major antigenic sites of the coronavirus transmissible gastroenteritis virus are located on the amino-terminal half of spike glycoprotein S

Host species restriction of Middle East respiratory syndrome coronavirus through its receptor, dipeptidyl peptidase 4

The spike protein of SARS-CoV-a target for vaccine and therapeutic development

Identification of a receptor-binding domain in the S protein of the novel human coronavirus Middle East respiratory syndrome coronavirus as an essential target for vaccine development

Several members of the mouse carcinoembryonic antigen-related glycoprotein family are functional receptors for the coronavirus mouse hepatitis virus-A59

Public health: broad reception for coronavirus

Coronavirus spike proteins in viral entry and pathogenesis

Evidence that TMPRSS2 activates the severe acute respiratory syndrome coronavirus spike protein for membrane fusion and reduces viral control by the humoral immune response

Major receptor-binding and neutralization determinants are located within the same domain of the transmissible gastroenteritis virus (coronavirus) spike protein

Recombination, reservoirs, and the modular spike: mechanisms of coronavirus cross-species transmission

Middle East respiratory syndrome coronavirus in dromedary camels: an outbreak investigation

Mechanism of membrane fusion by viral envelope proteins

Cross-neutralization of human and palm civet severe acute respiratory syndrome coronaviruses by antibodies targeting the receptor-binding domain of spike protein

Receptor-binding domain of severe acute respiratory syndrome coronavirus spike protein contains multiple conformation-dependent epitopes that induce highly potent neutralizing antibodies

Structural biology: adaptation of SARS coronavirus to humans

SARS coronavirus, but not human coronavirus NL63, utilizes cathepsin L to infect ACE2-expressing cells

Structural basis of neutralization by a human anti-severe acute respiratory syndrome spike protein antibody, 80R

History and recent advances in coronavirus discovery

Analysis of cellular receptors for human coronavirus OC43

Evidence for a common evolutionary origin of coronavirus spike protein receptor-binding subunits

Receptor recognition and cross-species infections of SARS coronavirus

Structure of SARS coronavirus spike receptor-binding domain complexed with receptor

Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus

Bats are natural reservoirs of SARS-like coronaviruses

Receptor and viral determinants of SARS-coronavirus adaptation to human ACE2

Molecular basis of binding between novel human coronavirus MERS-CoV and its receptor CD26

Virus entry: open sesame

The molecular biology of coronaviruses

The moonlighting enzyme CD13: old and new functions to target

Structural insights into immune recognition of the severe acute respiratory syndrome coronavirus S protein receptor binding domain

Crystal structure of mouse coronavirus receptor-binding domain complexed with its murine receptor

Crystal structure of bovine coronavirus spike protein lectin domain

Coronaviruses post-SARS: update on replication and pathogenesis

UCSF Chimera-a visualization system for exploratory research and analysis

Structure of severe acute respiratory syndrome coronavirus receptorbinding domain complexed with neutralizing antibody

Structure of coronavirus hemagglutinin-esterase offers insight into corona and influenza virus evolution

Dipeptidyl peptidase 4 is a functional receptor for the emerging human coronavirus-EMC

MERS: emergence of a novel human coronavirus

Antigenic modules in the N-terminal S1 region of the transmissible gastroenteritis virus spike protein

Structural bases of coronavirus attachment to host aminopeptidase N and its inhibition by neutralizing antibodies

Escape from human monoclonal antibody neutralization affects in vitro and in vivo fitness of severe acute respiratory syndrome coronavirus

The canyon hypothesis: hiding the host cell receptor attachment site on a viral surface from immune surveillance

Genetic evolution and tropism of transmissible gastroenteritis coronaviruses

Transmissible gastroenteritis coronavirus, but not the related porcine respiratory coronavirus, has a sialic acid (N-glycolylneuraminic acid) binding activity

X-ray crystal structure of the human galectin-3 carbohydrate recognition domain at 2.1-A resolution

Human angiotensin-converting enzyme 2 (ACE2) is a receptor for human respiratory coronavirus NL63

Specificity switching in virus-receptor complexes

Conformational change of the coronavirus peplomer glycoprotein at pH 8.0 and 37 • C correlates with virus aggregation and virus-induced cell fusion

Mechanisms of transmissible gastroenteritis coronavirus neutralization

Structure of a proteolytically resistant core from the severe acute respiratory syndrome coronavirus S2 fusion protein

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

ACE2 X-ray structures reveal a large hinge-bending motion important for inhibitor binding and catalysis

The N-terminal domain of the murine coronavirus spike glycoprotein determines the CEACAM1 receptor specificity of the virus strain

Mutational analysis of aminopeptidase N, a receptor for several group 1 coronaviruses, identifies key determinants of viral host range

Critical determinants of host receptor targeting by Neisseria meningitidis and Neisseria gonorrhoeae: identification of Opa adhesiotopes on the N-domain of CD66 molecules

Structure of MERS-CoV spike receptor-binding domain complexed with human receptor DPP4

Bat origins of MERS-CoV supported by bat coronavirus HKU4 usage of human receptor CD26

Chapter 4-coronavirus pathogenesis

Receptor for mouse hepatitis virus is a member of the carcinoembryonic antigen family of glycoproteins

A 193-amino acid fragment of the SARS coronavirus S protein efficiently binds angiotensin-converting enzyme 2

Crystal structure of NL63 respiratory coronavirus receptor-binding domain complexed with its human receptor

Structural basis for coronavirus-mediated membrane fusion: crystal structure of mouse hepatitis virus spike protein fusion core

Cryptic and regulatory epitopes in CD13/aminopeptidase N

Receptor usage and cell entry of bat coronavirus HKU4 provide insight into batto-human transmission of MERS coronavirus

Human aminopeptidase N is a receptor for human coronavirus 229E

Potent cross-reactive neutralization of SARS coronavirus isolates by human monoclonal antibodies

We thank C. Mark for editorial assistance. GM is a PhD fellow of the La Caixa Foundation International Fellowship Program (La Caixa/CNB). The work was supported by a grant from the Spanish Ministry of Science (BFU2011-23940 to JMC). We acknowledge support by the European Synchrotron Research Facility through the BAG-Madrid project.