Receptor interaction, fusion, and entry.
Coronaviruses attach to specific cellular receptors via the spike protein (Table
1). The first identified coronavirus receptor was CEACAM 1, utilized by MHV (
141). Viral attachment triggers a conformational change in the spike protein that promotes the fusion of viral and cellular membranes (
212,
369). While there are no crystal structures available for any coronavirus spike, it is believed that it may undergo changes similar to those of other type I fusion proteins, such as influenza virus hemagglutinin and human immunodeficiency virus gp120, in order to mediate fusion of viral and cellular membranes.
The coronavirus spike protein plays vital roles in viral entry, cell-to-cell spread, and determining tissue tropism. Coronavirus entry is, in general, not pH dependent, and thus it has been believed to occur directly at the plasma membrane and not via an endosomal route (Fig.
3). However, there are data suggesting that an endosomal route may be utilized by some viruses (
156,
219). Entry of SARS-CoV is inhibited by lysosomotropic agents, suggesting an endosomal route of entry (
285,
349). Furthermore, this inhibition may be overcome by protease treatment of virus that has attached to the cell. This, along with the observation that infection is blocked by inhibitors of the pH-sensitive endosomal protease cathepsin L, suggests that there is a requirement for cleavage of the SARS-CoV spike during entry through the endosomes (
213,
284). Furthermore, entry at the plasma membrane following protease treatment is more efficient than entry by the endosomal route (
213). Those authors suggested that SARS-CoV spike may be cleaved by the proteases produced by inflammatory cells present in the lungs of SARS patients and thus enter cells by the more efficient plasma membrane route (
213). The highly hepatotropic MHV-2 strain may enter the cell by an endosomal route similar to that used by SARS-CoV. MHV-2, like SARS-CoV, encodes an uncleaved spike protein and is sensitive to lysosomotropic agents; however, trypsin treatment of cell-associated MHV-2 spike overcomes inhibition by lysosomotropic agents (Z. Qiu and S. R. Weiss, unpublished data). This suggests that entry at the cell surface may require a cleavage of spike in the viral membrane, while endosomal entry may provide for cleavage during entry. Finally, coronaviruses with cleaved spikes may also enter the cell by the endosomal route. For example, while wild-type MHV-JHM enters cells in culture by a pH-independent pathway, the OBLV60 mutant of JHM is inhibited by lysosomotropic agents and is believed to enter though a lysosomal pathway (
221). Interestingly, OBLV60 is highly attenuated and exhibits restricted spread during infection of the murine central nervous system (
239,
316).
In general, the host range of coronaviruses is extremely narrow. The ability of a coronavirus to replicate in a particular cell type depends solely on the ability to interact with its receptors (
139). For example, murine coronavirus replicates in murine cells and not in human and hamster cells; however, once nonpermissive cells are transfected with the cDNA encoding MHV receptor, they become susceptible to MHV infection (
85). Several coronavirus receptors have been identified. The group I coronaviruses human HCoV-229E, feline FIPV, and porcine TGEV all use aminopeptidase N (APN), a zinc-binding protease, of their respective host species as their receptors (
352) (Table
1). There is some ability to recognize the corresponding APN receptor of another species; for example, HCoV-229E can utilize either human APN or feline APN as a receptor but cannot use porcine APN (
334,
335). The receptor used by the murine coronavirus group is carcinoembryonic antigen-cell adhesion molecule (CEACAM) (CD66a) (
43,
44,
84,
226). CEACAMs are glycoproteins possessing two or four immunoglobulin-like extracellular domains followed by a transmembrane domain and a cytoplasmic tail (
226). They are involved in the intercellular adhesion and development of hepatocellular, colorectal, and epithelial tumors (
13) and are expressed primarily on the epithelial and endothelial cells of the respiratory tract and intestines, as well as on other tissues (
111,
265). The observation that transgenic mice with a knockout of the CEACAM1 gene are resistant to infection demonstrates that this is likely the only receptor for MHV (
131). Interestingly, CEACAM1 is expressed at a low level in the brain, a major site of infection of some MHV strains, suggesting that low levels of receptor may be sufficient for mediating MHV entry. Expression of receptor has been demonstrated on only one central nervous system cell type, microglia; the receptor is downregulated on microglia during infection (
257). MHV spread for the highly neurovirulent JHM strain may be enhanced by receptor-independent spread (
103,
104) and/or by the expression of the hemagglutinin-esterase proteins (see below).
Other group II coronaviruses, such as BCoV, OC43, and porcine hemagglutinating encephalomyelitis virus, bind to 9-O-acetylated sialic acid-containing receptors (
159,
253). It is not clear, however, what the specific receptor molecules are, and little is known about the entry process.
Soon after the identification of SARS-CoV, the receptor for this virus was identified as angiotensin-converting enzyme 2 (ACE2). ACE2, like APN, the group I coronavirus receptor, is a zinc metalloprotease (
187). Human CD209L, a C-type lection (also called L-SIGN, DC-SIGNR, and DC-SIGN2), when expressed by transfected Chinese hamster ovary cells, renders the cells highly susceptible to SARS-CoV infection; however, it is significantly less efficient than ACE2 in mediating entry (
145). SARS-CoV S protein is able to interact with the lectin DC-SIGN; while DC-SIGN binding enhances infection of ACE2-bearing cells, it cannot alone mediate entry in the absence of ACE2. Thus, the interaction of SARS-CoV with this lectin on dendritic cells (DCs), which are not permissive for infection, may augment transmission of SARS to its target cells (
135). Surprisingly, it was recently shown that the newly identified group I human coronavirus NL63 also uses ACE2 as its receptor (
136).
The spikes of some coronaviruses mediate cell-to-cell fusion of infected cells as well as virus/cell fusion during entry, presumably by a similar mechanism (
369) (
212). However, viral entry and cell-to-cell fusion do display some differences in requirements. For example some MHV-JHM spikes can mediate cell-to-cell fusion in the absence of CEACAM, while entry requires the CEACAM receptor. Furthermore spike proteins that have mutations that eliminate cleavage into S1 and S2 subunits carry out cell-to-cell fusion very inefficiently; however, they mediate entry into susceptible cells with similar efficiency as wild-type virus (
75,
114,
181). Similarly, the MHV-2 strain encodes an uncleaved spike protein and does not carry out cell-to-cell fusion; this virus infects cells efficiently in vitro and causes severe hepatitis in vivo (
70,
132,
150). The spike of MHV-A59, which is usually cleaved during replication in cell culture, is not cleaved when recovered from brains or livers of infected mice, suggesting that cleavage is not a prerequisite for infection for strains that express cleaved spike (
133) and that entry of MHV into some types of cells in vivo may require an endosomal route of infection.
The heptad repeat domains and the putative fusion peptide are believed to play important roles in the fusion process (
103). This has been explored most for the MHV spike. Substitution of charged amino acids for hydrophobic ones in HR1 (and within a candidate fusion peptide) eliminates the ability to induce cell-to-cell fusion (
198). Mutations in the leucine zipper domain within HR2 inhibit the ability of spike to oligomerize and to inhibit cell-to-cell fusion (
197). Amino acid substitutions at L1114 within the HR1 domain of the JHM spike (L1114R or L1114F) are particularly intriguing in that they have been reported in multiple studies, in association with several mutant phenotypes. An L1114R substitution is one of three mutations believed to contribute to the low pH dependence for viral entry of the OBLV60 variant of JHM as well as its neuroattenuation and restriction to olfactory bulbs during infection of mice (
105). Furthermore, L1114R alone was sufficient to cause restriction of recombinant MHV to the olfactory bulbs during infection of mice (
316). Substitutions at L1114 have been identified in the spike of an attenuated monoclonal-antibody-resistant mutant (
327) and a soluble-receptor-resistant mutant (
269,
270). Interestingly, L1114R and L1114F substitutions were identified as secondary mutations in several recombinant viruses expressing A59/JHM chimeric spike proteins (
248,
316). The soluble-receptor-resistant mutant of JHM, srr7, (expressing a spike containing L1114F) demonstrated increased stability of the S1/S2 interaction, the loss of the ability to induce CEACAM-independent fusion (
301), and altered interactions with the receptor CEACAM1
b (an allele of CEACAM 1
a expressed by resistant SJ/L mice) as well as resistance to neutralization by soluble CEACAM1
a receptor (
211,
212). Similarly, the L1114R mutation results in loss of receptor-independent fusion along with neuroattenuation. In support of the idea that the RBD interacts with S2, a mutation in the RBD could functionally suppress the effects of an L1114R mutation in HR1 of srr7 that affected the ability to use CEACAM1
b (
211). Thus, small changes within the HR domains (for example, a single amino acid substitution at L1114) may result in major alterations in spike/receptor interaction and hence in virus entry and finally pathogenesis in vivo.
Recent studies of the HR domains provide further evidence confirming that the coronavirus spike is indeed a class I fusion protein (
23). Peptides representing HR1 and HR2 of MHV, when mixed together, assemble into an extremely stable oligomeric complex with both peptides in alpha-helical conformations and antiparallel to each other. In the native protein, such a conformation would be predicted to bring the N-terminal domain of HR1 and the transmembrane anchor into close proximity to facilitate the fusion process. Furthermore, the HR2 peptide was shown to be a potent inhibitor of virus entry, as well as of cell-to-cell fusion. Similar results were obtained for SARS HR domains. SARS-CoV HR1 and HR2 peptides, when mixed, assemble into a similar six-helix bundle; however, this complex was less stable than that of the corresponding MHV complex. The lack of stability may explain why HR2 peptides are less inhibitory for SARS than for MHV (
22).
Role in pathogenesis.
The use of recombinant coronaviruses, including MHV (
223,
246), TGEV (
274), and IBV (
35,
134), has definitively demonstrated that the spike is a major determinant of tropism and pathogenicity. In the case of TGEV, the replacement of the spike gene of an attenuated respiratory strain of TGEV with the spike gene from a virulent enteric strain renders the virus enterotropic (
274). Figure
5 summarizes the mapping of tropism and virulence with A59/JHM chimeric recombinant MHVs. The JHM strain is highly neurotropic, causing severe, usually fatal encephalitis and little if any hepatitis, while the A59 strain causes moderate hepatitis and is only weakly neurovirulent. The replacement of the spike gene in the genome of the A59 strain with the spike gene of the most highly neurotropic isolate of the JHM strain renders the resulting virus highly neurovirulent (
223,
246). The high neurovirulence conferred by the JHM spike is associated with rapid spread through the CNS, which may occur, in part, independently of the CEACAM receptor and the large numbers of infected neurons (
247). However, the resulting chimeric virus (JHM spike in the A59 background) is not as virulent as parental JHM, at least partially because it induces a much stronger CD8
+ T-cell response. JHM fails to induce a strong enough CD8
+ T-cell response to mediate clearance (
201,
260; Iacono et al., unpublished data). The mechanisms that underlie the differences in the immune response in the brain to the closely related A59 and JHM strains are intriguing and not at all understood.
The replacement of the spike protein of the moderately hepatotropic MHV-A59 with the spike of the nonhepatotropic JHM results in recombinant viruses with the ability to induce only minimal hepatitis (
222). Similarly, a chimeric virus with the spike of MHV-2, a highly hepatotropic strain in the A59 background, is highly hepatotropic (
223). Thus, for recombinant viruses with A59 background genes, the ability to induce hepatitis is dependent largely on the ability of the spike to mediate entry into cells of the liver. However, the outcome is somewhat different in chimeras in which background genes are derived from JHM. The replacement of the JHM spike with the A59 spike results in a chimeric virus that causes minimal infection of the liver and induces hepatitis very poorly; thus, the in the presence of JHM background genes, the spike of the A59 strain is unable to mediate efficient infection of the liver. The mechanism by which JHM background genes suppress infection of the liver is intriguing and merits further investigation.
In a similar spike exchange experiment performed with IBV, the ectodomain of the spike protein from the virulent M41-CK strain was used to replace the corresponding region within the apathogenic IBV Beaudette genome. The resulting chimeric virus displays the in vitro cellular tropism phenotype of M41-CK (
35); however, the virus remains apathogenic. Thus, the M41-CK spike is not sufficient to render the chimeric virus virulent (
134). The spike protein is therefore a major determinant of tropism and thus influences pathogenesis; however, the spike alone is not always the main determinant of pathogenesis, and as the data indicate, other genes also contribute to pathogenic phenotypes.
There are many strains of MHV and many isolates of the JHM strain, displaying different levels of neurovirulence. Among the JHM isolates, virulence is correlated with the presence of a long hypervariable domain (see above) within S1. The isolate referred to as MHV-4 (
67) or MHV
SD (
231) has the longest MHV HVR among JHM spikes and is able to induce cell-to-cell fusion and viral spread in the absence of the CEACAM receptor (
103,
104). It is likely that this ability is related to the less stable association of S1 and S2, such that the conformational changes in spike that lead to fusion are more easily triggered, and this in turn is at least partially responsible for its very high neurovirulence (
103,
161). Similarly, deletions, as well as single-site mutations, within the HVR region have been shown to influence neurovirulence (
67,
106,
201,
245).
Mutations within both the RBD of S1 and the heptad repeat domains within S2 have been show to influence pathogenesis. Mutations in the RBD are likely to affect the interaction between spike and the host cell and could thus affect viral entry and tropism, while mutations in the heptad repeats are likely to affect tropism by altering the fusion mechanism. Variation in the amino-terminal portions of the spike has also been noted in TGEV and IBV; the attenuated porcine coronavirus PRCoV has a deletion in the amino-terminal portion of S1 compared with the virulent TGEV (
91). A single amino acid substitution within the RBD, S310G, is responsible for enhanced neurovirulence of a JHM isolate (
231) Furthermore, a single Q159L amino acid substitution in this region eliminates the ability of MHV-A59 to infect the liver while having no effect on neurovirulence (
180,
181). The observation that a one-amino-acid substitution in the RBD can confer a complete loss of tropism to the liver while not affecting infection of the brain, while using the same CEACAM receptor, suggests that other cell surface molecules may serve as cofactors or coreceptors in an organ-specific way. An E1035D substitution within HR1 may overcome the Q159L mutation, since a spike with both of these substitutions confers hepatotropism upon a recombinant MHV-A59 (
224). In support of the idea that the RBD may interact with the HR domains, escape mutants selected by resistance to a monoclonal antibody mapping to the receptor binding domain of S1 had point mutations in the region of HR2, suggesting an interaction between these two physically distant portions of the spike (
121). Furthermore, mutations within S1 may also affect host range; 21 amino acid substitutions and a 7-amino-acid insertion within the N-terminal region of spike, but downstream of the RBD, allow MHV infection of the usually resistant hamster, feline, and monkey cells (
309).
SARS-CoV is believed to have jumped to humans from civets (see “CORONAVIRUSES AS EMERGING PATHOGENS: SARS-CoV” below). The adaptation of SARS-CoV to humans likely involved changes within the RBD. In comparison of the spike protein from civets and from humans, there are six amino acid differences within the RBD of the spike. The spike protein of civet SARS-CoV has low affinity for the human ACE2 SARS-CoV receptor. Substitution of two amino acids within the RBD of the human spike protein with those of the civet spike (N479K/T487S) almost abolishes the ability to infect (using the single-round infection assay) human cells expressing the SARS-CoV receptor. Conversely, substitution of the two residues within the civet spike with the human amino acids confers the ability to infect cells expressing the human receptor. Thus, it is likely that amino acids 479 and 487 are important for receptor interaction and hence species specificity and that selection of viruses with substitutions of these residues allowed the adaptation of SARS-CoV to humans (
188,
256).
For MHV, most of the H-2
b-restricted T-cell epitopes thus far identified are encoded in the spike gene. The MHV spike encodes an immundominant CD8
+ T-cell epitope (S510 to S518), located within the HVR (and therefore absent from the many strains and variants with deletions in the HVR, such as A59) and an additional subdominant CD8
+ epitope (S598 to S605) that is expressed by all MHVs. Mutation within the immunodominant epitope of the MHV spike has been reported as a mechanism to escape the immune response and achieve viral persistence; such epitope escape mutants selected in suckling mice were more virulent than wild-type virus (
243). However when the same inactivating mutation was introduced into a recombinant virus, the resulting virus ranged from slightly to significantly attenuated in weanling mice, depending on the genetic background of the virus and the strain of the mouse infected (
200). Under similar conditions, inactivating mutations within a foreign CD8 T-cell epitope (gp33 from lymphocytic choriomeningitis virus), introduced into recombinant MHVs in a nonessential gene, were readily selected in weanling mice previously immunized against this epitope (
54). Thus, the likelihood of epitope escape occurring depends on multiple factors, such as the location of the epitope within an essential versus a nonessential protein and its effect on function of the protein, the background genes of the virus, and the age and strain of mouse (
55,
152,
201). CD4 T-cell epitopes have been identified in the spike (
127,
322) as well as the M (
346) and N (
322) proteins of MHV (
242) and in the N proteins of porcine TGEV (
4) and avian coronaviruses (
20).
Studies with chimeric A59/JHM recombinant viruses demonstrated that genes other than the spike play a major role in determining tropism. In fact, JHM genes eliminate the ability of a virus expressing the A59 spike to cause hepatitis, and this is not due to the replicase but rather to genes in the 3′ end of the genome (
223; S. Navas-Martin and S. R. Weiss, unpublished data) (see Fig.
5). Furthermore, the extent of T-cell response to recombinant MHVs, and thus the likelihood for viral clearance to occur, is not determined by the spike gene, but rather by background genes, again encoded in the 3′ end of genome (Iacono et al., unpublished data). Thus, other viral structural genes clearly influence pathogenic outcome dramatically, and these are discussed below.
SARS-CoV spike protein may play a role in pathogenesis by inducing interleukin-8 (IL-8) in the lungs via activations of MAPK and AP-1 (
40). Such an activity was mapped to amino acids 324 to 688 of the SARS-CoV spike. This activity was detected in epithelial cells and fibroblasts by using baculovirus-expressed SARS-CoV spike; the location of the sequencing responsible for this activity overlaps with the RBD, suggesting that attachment to the ACE2 receptor may trigger this activation (
40).