Coronavirus Pathogenesis and the Emerging Pathogen Severe Acute Respiratory Syndrome Coronavirus

Taxonomy

The name “coronavirus,” coined in 1968, is derived from the “corona”-like or crown-like morphology observed for these viruses in the electron microscope (318). In 1975, the Coronaviridae family was established by the International Committee on the Taxonomy of Viruses. Recently, at the 10th International Nidovirus Symposium in Colorado Springs, Colo., in June 2005, it was proposed that the Coronaviridae family be divided into two subfamilies, the coronaviruses and the toroviruses, the latter of which cause enteric diseases in cattle and possibly in humans. The Coronaviridae family, along with the Arteviridae and Roniviridae families, form the Nidovirales order. The Arteviridae family includes swine and equine pathogens, and the Roniviridae family is composed of invertebrate viruses (64, 88).

Coronaviruses are divided into three genera (I to III), usually referred to as groups and based on serological cross-reactivity (218) (Table 1); more recent genome sequence analysis has confirmed this grouping (115). Group I coronaviruses include animal pathogens, such as TGEV of the pig, porcine epidemic diarrhea virus (PEDV), and feline infectious peritonitis virus (FIPV), as well as the human coronaviruses HCoV-229E and HKU1, which cause respiratory infections (see below). Group II also includes pathogens of veterinary relevance, such as BCoV, porcine hemagglutinating encephalomyelitis virus, and equine coronavirus, as well as human coronaviruses viruses OC43 and NL63, which, like HCoV-229E, also cause respiratory infections. Group II also includes viruses that infect both mice and rats. MHV is often studied as a prototype coronavirus; MHV is a group of highly related strains causing a variety of diseases, such as enteric disease, hepatitis, and respiratory disease, as well as encephalitis and chronic demyelination. Rat sialodacryoadenitis coronavirus also belongs to group II. There has been controversy about whether SARS-CoV defines a new group of coronaviruses or whether it is a distant member of group II (as discussed in “CORONAVIRUSES AS EMERGING PATHOGENS: SARS-CoV” below); given the data to date (113, 117), we have listed SARS-CoV in group II in Table 1. Group III thus far includes only avian coronaviruses, such as IBV, turkey coronavirus, and pheasant coronavirus (38). Recently, using reverse transcription-PCR (RT-PCR), coronavirus sequences were detected in the graylag goose (Anser anser), feral pigeon (Columbia livia), and mallard (Anas platyrhynchos) (147); phylogenetic analyses of the replicase and nucleocapsid (N) sequences suggest that these viruses are members of group III, but as yet they have not been isolated or characterized.

Coronavirus Diseases

Coronaviruses cause acute and chronic respiratory, enteric, and central nervous system (CNS) diseases in many species of animals, including humans (218). The pathogenesis of a few of these will be reviewed below.

Spike Protein

Hemagglutinin-Esterase Protein

The coronavirus HE glycoprotein forms a second type of spike, smaller (5 to 7 nm) than the spike protein peplomers, on the envelopes of some group II coronaviruses (151, 356). HE is synthesized as 42-kDa apoprotein, glycosylated to 65 kDa, and disulfide linked to form a homodimer; when expressed, the BCoV HE displays hemagglutinating and esterase activities (28, 151). The MHV HE displays 30% sequence homology to the HA1 subunit of the hemagglutinin-esterase fusion protein of influenza C virus (199). HE is thought to have been obtained via homologous RNA recombination involving a group II coronavirus before the split off of SARS-CoV, which does not encode an HE protein (287). Coronavirus HE proteins have not received much attention in the past, probably because they are nonessential for replication in tissue culture or, in the case of MHV, for virulence in mice.

Group II viruses that encode HE can be divided into two groups based on their sialic acid substrate specificity. BCoV, HCoV-OC43, and MHV-DVIM recognize and bind to Neu5,9Ac₂ (278), and encode HEs with sialate-9-O-acetylesterase activity (286, 325, 326). MHV strains S and JHM bind to Neu4,5Ac₂ and encode HEs with sialate-4-O-acetylesterase activity (259, 344). However, it is not clear whether it is HE or the spike that directly binds to sialic acid-containing receptors, and it has been argued that for BCoV, HCoV-OC43, and MHV, it is the spike protein rather than the HE that has the binding activity (165, 278, 344). However, HE proteins of some coronaviruses have been demonstrated to be sialic acid-specific lectins, as demonstrated by hemagglutination and/or hemadsorption assays, thus supporting the role of HE in receptor binding (244, 279, 359). For MHV-DVIM, hemagglutinating activity is associated with HE and not with spike (296; R. J. de Groot, personal communication). For BCoV, both HE and spike recognize the same receptor determinant of 9-O-acetylneuraminic acid on host cells (253). While HE is nonessential for replication in cell culture, spike is necessary and sufficient to mediate entry. Thus, the role of HE in coronavirus infection is still not clear and merits further investigation.

For MHV, the HE protein is expressed by a minority of strains; among these are the weakly pathogenic MHV-S, some isolates of the highly neurovirulent JHM strain (355), and enteropathogenic strains such as DVIM (344). While the highly tissue culture-adapted MHV-A59 genome encodes an HE protein (282), due to multiple mutations, the HE protein is not expressed and the gene is referred to as a pseudogene. Expression of the viral HE glycoprotein is not necessary for virulence in the animal as evidenced by the fact that MHV-A59 causes encephalitis and hepatitis, as well as demyelination, while it does not express HE. In tissue culture, viruses expressing HE have a relative growth disadvantage with respect to viruses that do not express HE. During serial passage in culture, there is a selection for variants in which mutations in the HE gene preclude insertion into the viral membrane (190).

The observation that HE expression is nonessential for the viral life cycle suggests that HE may play a role during infection of the animal (286). It has long been speculated that HE may play a role in acute and/or chronic disease induced by MHV, possibly as a determinant of cellular tropism (353, 354, 358), or may aid spread of the virus by augmenting attachment and/or exit from the cell (151). There are early data both supporting and arguing against this hypothesis. A higher level of mortality as well as increased infection of neurons was associated with a JHM variant that expressed high levels of HE compared with a variant that expressed less HE (353). Taguchi et al. (300) found that HE-expressing variants of MHV-JHM were selected for during propagation in cultured neural rat cells. Moreover, anti-HE monoclonal antibodies, when administered to mice, attenuated the acute encephalitis (354). However, in contrast to these studies, Lai and coworkers reported that in JHM-infected mice, viral variants defective for HE accumulate in the brain and spinal cord (172, 358). These studies were carried out before reverse genetics were available for MHV, and thus they were not able to distinguish between effects of HE and the influence of other genes in the comparison of various MHV isolates. A recent study compared the pathogeneses of isogenic recombinant viruses expressing a wild-type HE protein, expressing an HE protein in which the acetyl esterase activity has been eliminated by mutation, and not expressing the HE polypeptide. Surprisingly both viruses that expressed HE polypeptides (with or without a functional acetyl esterase activity) were more virulent when inoculated intracranially into mice (149). This result would be consistent with a model in which HE may enhance virus attachment and spread by binding to sialic acid-containing receptors and would suggest that the sialic acid binding domain is separate from the esterase domain. Similarly, in the case of influenza C virus hemagglutinin-esterase fusion protein, it has been proposed that there are two separate regions for binding to sialic acid, one for receptor binding and another for the catalytic (esterase) activity (130). We hypothesize that for MHV infection of the CNS, it is the binding activity of HE that augments spread of the virus in certain cell types and that at least in the CNS, the esterase activity is not important for enhanced spread. The esterase activity may be more important in other organs, such as the respiratory tract, where the virus may need to pass through mucus or have the ability to detach cells that may not be productively infected, both of which are believed to be functions of neuraminidases. In the case of influenza virus, it has recently been shown that the specificity of the neuraminidase (esterase) for a particular type of sialic acid determines the cell subtype infected within the respiratory track and hence the pathogenic outcome (210). Thus, by analogy, the HE of MHV may also play a role in tropism.

Membrane Protein

The M protein is the most abundant virion membrane protein. Aside from its role in viral assembly, the coronavirus M protein is believed to have functions in host interactions. It may be O glycosylated (groups I and III) or N glycosylated (group II). While glycosylation is not essential for viral assembly or infectivity (72), the glycosylation state of M protein is likely to play a role in virus-host interaction. For TGEV, the M protein has been shown to have interferogenic activity, and mutations in the M protein ectodomain that impair N glycosylation decrease this activity (175). For MHV, the selection of recombinant viruses with N, O, or no glycosylation demonstrated that while the glycosylation state of M protein does alter the ability to replicate in vitro, it may affect the ability to induce IFN-α in vitro and also to replicate in the liver in vivo (72).

Nucleocapsid Protein

In addition to its role as structural protein, N protein plays a role in transcription and also in pathogenesis. Expression of N protein is necessary for efficient recovery of virus from infectious cDNA clones (363, 364) and recently has been shown to enhance the replication of HCoV-229E genome RNA (277). The N protein of MHV has been implicated in fulminant hepatitis (230). Infection of mice with the highly hepatotropic MHV-3 strain stimulates expression of the fgl2 gene, which encodes a novel immune procoagulant molecule, fibrinogen-like protein 2, expressed in Kuppfer cells and endothelial cells of the liver. The ability to upregulate transcription of this gene maps to the nucleocapsid gene and correlates with the development of fulminant hepatitis (77, 230). While MHV proteins are generally restricted to the cytoplasm, the nucleocapisd proteins of coronaviruses representing groups I, II, and III were shown to localize to the nucleolus as well as to the cytoplasm (343). This report suggests that N protein induces a cell cycle delay or arrest, most likely in the G₂/M phase, possibly by inhibition of cytokines.

Small Envelope Protein

The coronavirus E protein is an integral membrane protein (365). Along with the M protein, E plays an important role in viral assembly (324). E protein, when expressed alone or when expressed together with M, forms virus-like particles. Surprisingly, it was possible to select a recombinant MHV with a deletion of the E gene. Such a recombinant MHV has low infectivity and replicates poorly, indicating that while it is nonessential for MHV, E plays an important role in production of infectious virus (168). The E protein of TGEV, however, is essential; disruption of the E gene within TGEV proteins is lethal (65, 233). Apart from its role in virus assembly, E protein has additional functions during infection. It has recently been demonstrated that the E protein of SARS-CoV has cation-selective ion channel activity (337). While the role of this activity is as yet unknown, the E protein ion channel could function at the site of budding to enhance viral morphogenesis and assembly. E protein appears also to play a role in host-virus interaction, specifically in induction of apoptosis. E induces apoptosis in vitro in MHV-A59-infected 17Cl-1 cells via a caspase-dependent mechanism; such apoptosis is suppressed by a high level of Bcl-2 expression. Inhibition of MHV-induced apoptosis promotes virus production late in infection, suggesting that apoptosis may be a host response that limits the level of virus production (3). Whether this occurs in vivo during infection of the mouse has not yet been reported. Similarly, the E protein of SARS-CoV has been shown to induce apoptosis when expressed in Jurkat T cells, and this activity is inhibited by expression of the antiapoptotic protein Bcl-xL (348). Those authors suggest that T-cell apoptosis may contribute to the SARS-CoV-induced lymphopenia that is observed in most SARS patients.

Internal Protein

The genomes of several group II coronaviruses, including MHV, contain an internal ORF within the nucleocapsid gene (97, 174). This ORF, translated in the +1 reading frame with respect to the N protein, encodes a mostly hydrophobic 23-kDa polypeptide. The I gene product is expressed in MHV-infected cells and found within the virions as well. Selection and characterization of recombinant viruses in which the I gene is disrupted demonstrated that I protein is not essential either for the replication of MHV in tissue culture or for pathogenesis in the mouse (97). However, the I gene does confer a small-plaque morphology and may have an as-yet-unknown subtle role in pathogenesis.

Replicase Proteins

The replicase proteins could affect tropism and pathogenesis by determining the rate of viral replication, perhaps via interactions with noncoding 5′ and/or 3′ UTR sequences in the viral genome, with cell type-specific factors, or with elements of the immune response. The several enzymatic activities that have been predicted to be encoded in ORFs 1a and 1b, as described above, could be involved in subverting many aspects of host cell metabolism (378). With the availability of infectious cDNA clones, the replicase gene is now available for genetic analysis, and information concerning the role of the replicase proteins is likely to be forthcoming. For example, recently a one-amino-acid substitution in nsp14 (ExoN or p59), a protein with exonuclease activity, was shown to drastically reduce the virulence of a recombinant A59 without affecting the in vitro replication; the mechanism of this attenuation is not known (290).

The p28 protein of MHV, encoded at the extreme 5′ terminus of the genome and thus processed from the N terminus of pp1a, when expressed transiently in several different cultured cell types prevents cell cycle progression from G₀/G₁ to S phase (42).

Studies of A59/JHM chimeric MHVs, in which the A59 replicase gene is expressed with the JHM structural genes and vice versa, demonstrated that the replicase is not an important determinant of the difference in tropism and pathogenesis between the two strains (severe encephalitis versus hepatitis). It is rather the 3′ portion of the genome that is responsible for pathogenic properties (Navas-Martin and Weiss, unpublished data). This is consistent with the observation that the ability of MHV to induce fulminant hepatitis maps to the nucleocapsid gene (230).

Group-Specific Proteins

Like other RNA viruses, all coronaviruses encode, in addition to structural proteins and replicase proteins, small nonessential proteins of unknown function. There are many examples of such proteins, encoded by RNA viruses, which interact with and compromise the alpha/beta interferon response; among these are VP35 of Ebola virus (11), the V proteins of several paramyxoviruses (235, 360), and the NS1 and NS2 proteins of human and bovine respiratory syncytial viruses (289, 319). While there are not yet any specific examples of coronavirus proteins being involved in antihost defense by subverting the host innate immune response, it is quite plausible that one or more coronavirus nonessential proteins serve such an “accessory” function during infection in vivo.

The coronavirus genes encoding these proteins are sometimes referred to as “small ORFs” or “group-specific” genes, as they are conserved among each group of coronaviruses. The MHV genome contains ORFs 2a, 4, and 5a; the proteins encoded in these three ORFs appear to be nonessential for replication. In some strains of MHV, such as A59, ORF 4 is interrupted and becomes ORFs 4a and 4b (333), and there are reports of an MHV (JHM strain) isolate with a deletion of ORF 2a (280) as well as an MHV strain S isolate with deletions of most of ORFs 4 and 5a (357). The intriguing question of whether one or more of these ORFs encodes a protein with a role in pathogenesis remains unresolved. The use of reverse genetics makes it possible to address this question by the selection of isogenic viruses differing only in expression of a particular ORF. A recombinant MHV (JHM strain) lacking gene 4 has been shown to be as neurovirulent as the wild type in mice (232). In contrast, a recombinant MHV lacking ORFs 2a, 4, and 5a is avirulent in mice; however, because replication of this virus is inefficient in vitro, it is not possible to determine if the attenuation in mice is due to a specific function of a viral gene product or to a more general inability of the virus to replicate efficiently (73). There are a few examples of recombinant viruses in which elimination of expression of ORFs may confer changes in pathogenic phenotype. A recombinant virus with a deletion of ORF 2a (as well as the HE pseudogene) was attenuated despite its ability to replicate to similar titers as wild-type virus in vitro; this suggests that ORF 2a could play a role in virulence. In support of a role in pathogenesis for the ORF 2a-encoded 30-kDa protein is the report of a recombinant MHV in which a mutation within ORF 2a is associated with attenuation in animals (290). ORF 2a encodes a (putative) cyclic phosphodiesterase; it is intriguing to speculate that such an activity may participate in compromising the host response. It has been speculated that, in analogy to their cellular homologues, the cyclic phosphodiesterase along with the predicted ADP-ribose 1"-phosphatase activity, encoded in ORF 1a (discussed above), may mediate consecutive steps in processing of tRNA splicing products (58, 378).

In the case of the porcine coronavirus TGEV, expression of gene 7 is not essential for replication; however, a recombinant virus in which the expression of gene 7 was abrogated displays reduced virulence with less virus replication in the lung and gut, suggesting that the gene 7 product does influence in vivo replication and virulence (234). ORFs 3a and 3b of TGEV are also not essential for replication (217, 336). Various PRCoV isolates have deletions within ORFs 3a and 3b, suggesting that the loss of expression of these ORFs may be associated with attenuation; however, the attenuation in these strains is associated with deletions in the spike gene rather than in ORFs 3a and 3b. Furthermore, a recombinant with a deletion of these ORFs demonstrates wild-type tropism, replicates efficiently in animals, and displays only a very small reduction virulence (90, 274, 323).

Recombinant FIPVs with deletions of either ORFs 3a, 3b, and 3c or ORFs 7a and 7b multiply efficiently in cell culture but show an attenuated phenotype in the cat. This suggests that one or more of the proteins encoded in these ORFs may play a role in virulence (124). A recombinant IBV in which expression of gene 5 was eliminated replicates to a similar extent as wild-type virus in tissue culture, demonstrating that neither of the two proteins encoded in gene 5 is essential for replication and that these are candidate accessory proteins (34).

The human SARS-CoV genome encodes multiple small open reading fames (ORFs 3a, 3b, 6, 7a,7b, 8a, 8b, and 9b) that are presumed to encode eight group-specific accessory proteins (205, 287). There are no reports of proteins in the NCBI database with homology to any of these ORF-encoded protein. Like the group II coronavirus I protein gene, the SARS ORF 9b is within the nuclecocapsid gene of SARS-CoV and at a similar position (287); however, the predicted proteins has no homology with the I protein, and whether it is a minor virion protein is not known. It is possible that some of these ORFs, perhaps the smaller ones, may not encode authentic proteins; in fact, ORFs 3b, 6, and 8b appear to be in a poor context for translation (158). ORFs 8a and 8b may have been created by a deletion in ORF 8 of the animal SARS-CoV isolates; thus, ORF 8 may not be necessary for replication in humans (122) (see “CORONAVIRUSES AS EMERGING PATHOGENS: SARS-CoV” below).

Two of these ORFs have been shown to be expressed during infection, and the encoded proteins have been partially characterized. ORF 7a encodes a 122-amino-acid protein predicted to contain a cleaved signal sequence and a C-terminal transmembrane helix, indicating that it is likely to be a type I membrane protein. ORF 7a protein is expressed in Vero E6 cells and in the lungs of patients but is not present in virions (47, 96, 227, 306). ORF 7a is localized to the perinuclear region of both infected and transfected cells; it colocalizes with endoplasmic reticulum and intermediate compartment markers, consistent with the endoplasmic reticulum retrieval motif in the C-terminal tail domain (96). Confocal microscopy suggests that the short cytoplasmic tail and transmembrane domain function in trafficking of the ORF 7 protein within the endoplasmic reticulum and Golgi network (227). Analysis of the crystal structure of the N-terminal ectodomain of ORF 7a protein demonstrates a compact seven-stranded beta sandwich structure similar to that of members of the IgG superfamily (227).

The 274-residue protein encoded by ORF 3a is predicted to contain three transmembrane domains. It is found in virus particles, and in infected Vero E6 cells it is localized to perinuclear regions as well as the plasma membrane and with intracellular virus particles. The ORF 3a protein is present in two processed forms as well as the full-length protein. The ORF 3a protein can interact with the M, E, and S structural proteins, as well as with the ORF 7a protein (143, 306).

There are reports that ORF 7a (305), ORF 3a (177), and ORF 3b (366), in addition to E protein (see above) and N protein (298), each induce apoptosis. Since all of these studies involve overexpression of individual proteins, it is difficult to determine which if any play this role during infection; the analysis of recombinant viruses with specific mutations will be important to definitely demonstrate a role for any particular protein in apoptosis.

Recent data indicate that these small ORFs may vary among SARS isolates. For example, there are reports that in the genomes of SARS-CoV isolates from humans there are deletions within a single ORF 8, resulting in two ORFs, 8a and 8b (122, 287) (see “CORONAVIRUSES AS EMERGING PATHOGENS: SARS-CoV” below). In fact, other than differences in spike gene sequences, this was the most striking difference observed between the genomes of animal and human isolates of SARS-CoV. In another study, an “in-frame” deletion of 45 nucleotides was found to occur in ORF 7b after three passages of SARS-CoV in tissue culture (311). A deletion in ORFs 6 to 8 was observed in the adaptation of SARS-CoV to primate cell culture (251). While it is tempting to speculate that one or more of these ORFs participate in adaptation to the human host and/or subversion of the host innate immune response, there are no conclusive data on these issues.

Severe Acute Respiratory Syndrome

In February 2003 the World Health Organization (WHO) received reports from China of a new respiratory illness outbreak in Guangdong Province (341). However, the first cases of “an atypical pneumonia” emerged in Guangdong Province (China) in late 2002 (238).

A novel virus was isolated from patients' lungs and sputa and cultivated on a monkey kidney cell line (Vero E6) (81, 82, 162, 240). The sequence information demonstrated that this was a previously unrecognized coronavirus (267, 287). Proof that this virus is the etiologic agent for SARS was provided by results of infections carried out in nonhuman primates, in which Koch's postulates were fulfilled (102). The SARS epidemic was officially controlled by July 2003 (6, 100). The epidemic was controlled only by strict isolation of patients. By the end of the epidemic, the CDC and WHO reported more than 8,000 cases, with more than 800 deaths worldwide (162). SARS cases were reported from 29 countries, mostly in Asia, although North America was also affected, most notably Toronto, Canada. A total of 156 suspected SARS cases were reported in the United States, although only 8 cases had confirmed serologic evidence of SARS-CoV infection (250).

SARS infection exhibits a wide clinical course, characterized mainly by fever, dyspnea, lymphopenia, and lower respiratory tract infection (229, 317). Concurrent gastrointestinal symptoms and diarrhea, with active replication of SARS-CoV in both the small and large intestines, are common (19, 179, 182, 240). Infected individuals have slightly decreased platelet counts, prolonged coagulation profiles, and mildly elevated serum hepatic enzymes. Although the route of transmission has not been clearly established, airborne droplets from infected patients may be the main route of transmission (375). Blood and fecal-oral transmission has been suggested to be the route of transmission for one index case (252).

Origin of SARS-CoV

Coronavirus biological vectors are not known. However, serological and genetic evidence from various studies supports a zoonotic origin of SARS-CoV (140). This hypothesis was first based on epidemiological reports demonstrating that early patients with SARS in Guangdong Province were exposed to live wild game animals held in markets serving the restaurant trade (375). In order to identify animals carrying SARS-CoV, a range of domestic and wild mammals in Guangdong Province were examined. Interestingly, SARS-like viruses genetically and antigenically related to the human SARS coronavirus were detected by RT-PCR in the nasal and fecal swabs of civet cats (Paguma larvata) and a raccoon dog (Nyctereutes procyonoides). Serological evidence of infection was found in these species and also in a Chinese ferret-badger (Melogale moschata) (122). Interestingly, animal traders working with live animals in these markets had high seroprevalence for both the human and animal SARS coronaviruses, although they did not have a history of SARS-like disease (122). SARS-like viruses isolated from animals had more than 99% homology with human SARS-CoV. However, compared with the animal viruses, the human SARS coronavirus isolated both from the early phase of epidemics and from May 2003 exhibited deletions in ORF 8 that differed in length (from 29 to 82 nucleotides at the early phase to a 415-nucleotide deletion resulting in the loss of the whole ORF 8 region at the late phase of the outbreak) (375). It is unknown whether deletions in ORF 8 confer an advantage to adaptation to humans or whether ORF 8 is dispensable in humans but not in animals. Poon et al. (250) speculated that the animal precursor of SARS coronavirus is thus likely to be inefficient at infecting humans, and repeated exposure to the precursor animal virus might lead to the abortive infection or antigenic stimulation that results in the observed serological response in the animal handlers. Thus, the live-animal markets were probably the site where the interspecies transfer of the animal virus to humans occurred. In this sense, molecular epidemiological studies have suggested several introductions of the animal coronavirus into humans. For example, phylogenetic analysis of the few human cases that occurred in December 2003 in Guandong Province revealed that this SARS-CoV is much closer to the palm civet CoV than to any SARS-CoV isolated from humans in the early epidemic (51, 288). Whether SARS-CoV has a reservoir in a wild animal species (142) remains to be further evaluated. In addition, ferrets and domestic cats are susceptible to SARS-CoV infection (207). Those authors reported that when noninfected ferrets and cats were housed with an infected animal, transmission occurred and viral titers increased gradually. Interestingly, efforts from several groups trying to identify a putative wild animal reservoir have led to the discovery of new animal coronaviruses in bats (249) and birds (147). The possibilities of interspecies transmission and wildlife reservoirs are not a new ideas for coronaviruses. Group I porcine, canine, and feline coronaviruses are antigenically related and may even be host range mutants derived from a common ancestor. In support of this idea, coronavirus infection may occur across species in pigs, dogs, and cats. Furthermore wild and domestic carnivores such as dogs and foxes may serve as reservoirs for TGEV (272).

The discovery of ACE2 as a receptor for SARS-CoV (187) is a major advance in our understanding of how SARS-CoV enters the cells and has allowed the elucidation, at the molecular level, of SARS-CoV cross-species transmission. Two groups have compared the spikes proteins derived from the 2002 to 2003 outbreak (TOR2), from the less severe 2003 to 2004 outbreak (GD), and from palm civets (SZ3) (122, 188, 256). A representative spike protein from the mild 2003 to 2004 outbreak and one from palm civets mediate more efficient infection of cells expressing palm civet ACE2 than of cells expressing human receptor. In contrast, spike protein from the severe 2002 to 2003 outbreak efficiently binds and utilizes both receptors. These data are consistent with the absence of human-to-human transmission during the 2003 to 2004 outbreak and with recent transmission of SARS-CoV from palm civets to humans (122, 288). This difference in the ability of civet and human SARS isolates to utilize human receptor is associated with amino acid substitutions in the RBD (see “Spike Protein” in “ROLE OF CORONAVIRUS PROTEINS IN PATHOGENESIS” above).

There has been some controversy about SARS-CoV classification. Early after its discovery, SARS-CoV was suggested to define a new group (IV) within the coronavirus groups (162, 205, 267). However, based on sequence comparisons and the observation that regions of ORF 1a of SARS-CoV contain domains that are unique to the group II coronaviruses, it has been suggested that it is more directly related to group II viruses, along with the bovine coronavirus, the human OC43 virus, and the murine coronavirus (MHV) (287). Thus, Gorbalenya et al. (117) have placed SARS within group II coronaviruses, in a subgroup IIb. Using bioinformatics methods, several groups have suggested possible recombination events for the origin of SARS (203, 261, 292, 371). Rest and Mindell (261) have suggested that the SARS-CoV polymerase may be a result of recombination. Stavrinides and Guttman (292) have reported a possible past recombination event between mammalian-like and avian-like parent viruses. Those authors suggested a mammalian-like origin for the replicase protein, an avian-like origin for the matrix and nucleocapsid proteins, and a mammalian-avian mosaic origin for the spike protein. However, the origin of SARS-CoV by recombination of mammalian and avian viruses seems unlikely. While recombination is one of the hallmarks of coronaviruses and there is evidence that may be it is a major force in coronavirus evolution, there is no evidence for recombination between members of different groups of coronaviruses. Consistent with this, Masters and colleagues have demonstrated, using chimeric recombinant viruses, that the 3′ UTR of SARS-CoV may substitute functionally for that of MHV, while 3′ UTRs from prototype group I or group II coronavirus cannot (113). Thus, the evidence to date suggests that SARS-CoV belongs within group II coronaviruses.

SARS Pathogenesis

The mechanism of injury caused by SARS-CoV infection remains unknown. A SARS disease model was proposed, consisting of three phases: viral replication, immune hyperactivity, and pulmonary destruction (317). SARS pathology of the lung has been associated with diffuse alveolar damage, epithelial cell proliferation, and an increase of macrophages. Multinucleate giant-cell infiltrates of macrophage or epithelial origin have been associated with putative syncytium-like formation that is characteristic of many coronavirus infections (228). The lymphopenia, hemophagocytosis in the lung, and white-pulp atrophy of the spleen observed in SARS patients are reminiscent of those reported for fatal influenza virus subtype H5N1 disease in 1997 (312). Strikingly, the presence of hemophagocytosis supports a cytokine deregulation (99). It is widely considered that SARS is a viral pneumonia. However, SARS patients may also exhibit gastrointestinal symptoms (313) and splenic atrophy and lymphadenopathy (79). Diarrhea is a very frequent finding in SARS patients (30 to 40% of patients). SARS-CoV replicates in enterocytes, with minimal disruption of the intestinal architecture. The absence of intestinal inflammation has been speculated to be a result of upregulation of transforming growth factor β (48) and an antiapoptotic host cellular response in the intestinal epithelial cells (241). Recent findings based on autopsies of SARS patients proposed that SARS is a systemic disease with widespread extrapulmonary dissemination, resulting in viral shedding in respiratory secretions, stools, urine, and even sweat (78, 95).

Proinflammatory cytokines released by stimulated macrophages in the alveoli may have a role in the pathogenesis of SARS. SARS-CoV infection of macrophages in vitro leads to the initiation of viral replication and viral protein synthesis, but replication is abortive and no virus particles are produced. In contrast to the case for influenza A virus and HCoV-229E, no IFN-α/β response is detected in macrophages, despite the induction of the expression of chemokines such as CXCL10/IFN-γ-inducible protein 10 and CCL2/monocyte chemotactic protein 1 (50).

Strikingly, it has been recently reported that SARS-CoV replicates in peripheral blood mononuclear cells (PBMCs) from SARS patients (184). Reghunathan et al. (258) have analyzed the gene expression profiles of PBMCs from 10 SARS patients compared with healthy controls. Analysis of gene expression of PBMCs of SARS patients by using a microarray platform that includes more than 8,000 gene sequences suggests that the response of SARS patients seems to be mainly an innate inflammatory response, rather than a specific immune response against a viral infection. Those authors did not find significant upregulation of major histocompatibility complex class I genes or major cytokines, including IFNs (IFN-α, IFN-β, and IFN-γ), or genes involved in complement-mediated cytolysis. They concluded that the immune response against the SARS-CoV may be different from that in other viral infections or that the virus may be using an unusual strategy to evade the host immune system and cause the pathogenesis and mortality.

Lymphopenia and increasing viral load in the first 10 days of SARS suggest immune evasion by SARS-CoV. The lack of an IFN-β response in SARS-CoV-infected cells has been reported in vitro, using human primary myeloid-derived dendritic cells (176) and the epithelial 293 cell line (291). Law et al. (176) proposed a mechanism of immune evasion by SARS-CoV in DCs, based on their findings of low expression of antiviral cytokines (IFN-α, IFN-β, IFN-γ, and IL-12p40), moderate upregulation of proinflammatory cytokines (tumor necrosis factor alpha [TNF-α] and IL-6), and significant upregulation of proinflammatory cytokines (MIP-1α, RANTES, IP-10, and MCP-1). Spiegel et al. (291) demonstrated that SARS-CoV escapes interferon-mediated growth inhibition by preventing the induction of IFN-β through interfering with the activation of IFN regulatory factor 3. The mechanism of lymphopenia remains unclear. The rapid decrease in both CD4 and CD8 T cells may be associated with an adverse outcome (185, 338). Due to the absence of ACE2 expression in T- and B-cell lymphocytes (126), it seems unlikely that SARS-CoV-induced lymphopenia may be caused by direct viral infection. Rather it has been suggested that apoptosis of uninfected lymphocytes may lead to the acute lymphopenia observed in SARS patients. In this sense, various SARS-CoV proteins have been suggested to induce apoptosis in vitro. These include ORF 7a (305), ORF 3a (177), and ORF 3b (366), in addition to E protein (see above) and N protein (298) (see “Group-Specific Proteins” in “ROLES OF CORONAVIRUS PROTEINS IN PATHOGENESIS” above).

Nonhuman Primate Models

The cynomolgus macaque (Macaca fascicularis) was the first animal subjected to experimental infection with SARS-CoV. Osterhaus and colleagues (102, 164), demonstrated that SARS-CoV fulfils Koch's postulates to a related host (cynomolgus macaque), leading to the development of a disease comparable to that observed in humans. Macaques were infected with 10⁶ 50% tissue culture infective doses (TCID₅₀) of SARS-CoV intratracheally, intranasally, and on each conjunctiva. Some macaques became lethargic from 3 days postinfection (p.i.) onwards and developed a temporary skin rash and respiratory distress. Viral RNA was detected using RT-PCR in sputum and in nasal and pharyngeal swabs on days 2 to 6 p.i. Necropsies were performed at day 6 p.i. Macaques developed interstitial pneumonia of differing severity, some of them with necrosis of alveolar and bronchiolar epithelia. Occasional multinucleated cells of approximately 30 nuclei (syncytia), similar to those observed in SARS-infected human lungs, were present in the lumens of alveoli and bronchioles. SARS-CoV RNA was detected in lungs by RT-PCR, and virus was also isolated. SARS-CoV antigen was detected in type 2 pneumocytes in lungs. SARS-CoV RNA was occasionally detected in cerebrum, duodenum, kidney, nasal septum, skin, spleen, stomach, trachea, tracheobronchial lymph node, and urinary bladder.

Three other labs have attempted to develop nonhuman primate models. It is worth noting that the utility of the cynomolgus macaque model remains controversial. Rowe et al. (268) have demonstrated that cynomolgus as well as rhesus macaques develop a mild self-limited respiratory infection very different from that observed in humans. Clinical symptoms were mild and localized to the upper respiratory tract, consisting of mild cough, sneezing, and slightly decreased activity on days 2 to 3 after virus challenge. Animals did not demonstrate signs of respiratory distress and become asymptomatic by day 8 to 10 p.i. In contrast to the studies reported by Osterhaus and colleagues, necropsies were done at days 12 to 14 p.i. (instead of day 6 p.i.). Microscopic examination revealed patchy areas of mild interstitial edema and alveolar inflammation, with occasional areas of intra-alveolar edema, interspersed with normal lung histology. There was no evidence of syncytia or alveolar damage, as has been reported for humans. No viral antigen was detected in lungs by immunohistochemical analysis.

In addition to the studies described above, McAuliffe et al. (216) have evaluated the level of replication and serologic response to experimental infection with SARS-CoV in three species of Old World monkeys (African green, rhesus, and cynomolgus macaques). Although clinical illness was not present in any of these species, African green, rhesus, and cynomolgus monkeys were susceptible to infection with SARS-CoV. The level of replication in the respiratory tracts of African green monkeys was greater than that in cynomolgus monkeys, which was higher than that in rhesus macaques. The neutralizing antibody response correlated with the level of virus replication. Histopathologic examination of African green monkey lungs demonstrated interstitial pneumonitis in association with SARS-CoV replication on day 2 p.i., which was resolving by day 4 p.i. SARS-CoV antigen was present in bronchiolar epithelial cells and type I pneumocytes. The ability of primary infection to prevent reinfection was evaluated in African green monkeys by SARS-CoV challenge 2 months after the primary infection. These animals showed no evidence of viral replication or enhanced disease.

Qin et al. (255) investigated the susceptibility of rhesus macaques to SARS-CoV infection through nasal cavity inoculation. In contrast with previous studies, macaques were followed for a 60-day period postinfection. Viral RNA could be detected in all the infected monkeys up to 16 days after infection. Histopathological changes of interstitial pneumonia were found in the lungs during the 60 days after inoculation. Interstitial pneumonia was less prominent at later times, indicating resolution in these animals. The neutralizing antibody response persisted in the animals at day 60 p.i.

The differences in disease outcome observed in cynomolgus macaques after experimental infection with SARS-CoV are intriguing. Rowe at al (268). have proposed a model of disease to reconcile apparent differences in the macaque experiments reported to date. Taking all the data together, those authors believe that the clinical manifestations observed in macaques are not sufficiently robust to be useful in evaluating pathogenesis or assessing therapeutic efficacy. SARS-CoV replicates in the upper respiratory tract and lungs of macaques, although lung pathology and recovery of virus are more evident when animals are necropsied at early time points (i.e., days 2 to 6 p.i.) than at later time points (days 12 to 14 p.i.). In contrast to the previous studies (216, 268), Qin et al. (255) followed the macaques for a longer period (60 days p.i.) and reported interstitial pneumonia (with milder histopathology after 30 days p.i.) through the study. Interestingly, the lesions in the lung, although milder than the histopathological changes observed in SARS patients, were consistently observed in all animals, with hyaline membranes, lung edema, and desquamation of alveolar lining cells in some infected macaques. This is the first study that follows nonhuman primates for such a long period, and it suggests that SARS-CoV infection in macaques deserves further investigation.

The marmoset (Callithrix jacchus) is a New World nonhuman primate that supports SARS-CoV replication in the respiratory tract (119). All animals developed multifocal mononuclear cell interstitial pneumonitis, accompanied by multinucleated syncytial cells, edema, and bronchiolitis in most animals. SARS-CoV antigen localized primarily to infected alveolar macrophages and type 1 pneumocytes by immunohistochemistry. Viral RNA was detected in all animals from pulmonary tissue extracts obtained at necropsy. Interestingly, viral RNA was also detected in tracheobronchial lymph node and myocardium, together with inflammatory changes, in some animals.

Cat Model

Domestic cats (Felis domesticus) support SARS-CoV replication. Natural asymptomatic infection in cats was first documented during the community outbreak at Emory Gardens, Hong Kong (http://www.who.int/sars/en/whoconsensus.) Martina et al. (207) studied the susceptibility of cats to experimental SARS-CoV infection. Cats inoculated intratracheally with 10⁶ TCID₅₀ do not develop clinical signs, although cats shed virus from the pharynx starting at 2 days p.i. and continuing until day 10. The virus was isolated from nasal swabs on days 4 and 6 p.i., whereas rectal swabs were negative. SARS-CoV was isolated from the trachea and lungs, although titers were low (1 × 10³ TCID₅₀/ml), peaking at days 6 to 8 p.i. Some infected cats developed mild pulmonary lesions. Cats seroconvert by day 28 p.i.

Ferret Model

The ferret (Mustela furo) is been exploited as an animal model for other respiratory infections, such as influenza, because ferrets develop a clinical illness that resembles human influenza, with rhinorrhea, fever, and sneezing following intranasal inoculation. To date, there is evidence that ferrets support SARS-CoV replication and that lungs of infected ferrets show some pulmonary lesions milder than those observed in cynomolgus macaques (although no pathological analysis was reported in this study) (207). However, ferrets did not develop fever or respiratory signs, but they were lethargic. At the time this review was written, another group has reported results in ferrets that disagree with those of the above-mentioned study (331). In contrast to the first study (207), in the latter study ferrets were totally asymptomatic although they supported SARS-CoV replication. Ferret models are currently been studied by other groups, and future research will clarify their use as a model for SARS pathogenesis.

Rodent Models

The use of small animal models, in particular mice, to study the pathogenesis of SARS-CoV has obvious advantages over use of the nonhuman primate models. First, a mouse model, if successful, would be less expensive than a macaque model. Second, mice can be genetically engineered, allowing the study of specific components of the host immune response (or others) against viral infection. In addition, mouse models for other human infectious diseases have been successfully developed (31). In this respect, the identification of the human receptor for SARS-CoV made possible the generation of receptor-expressing transgenic mice.

BALB/c, C57BL/6, and 129SvEv mice have been evaluated by several labs as possible animal models for SARS. It seems that the route of inoculation was a key factor to demonstrate SARS-CoV replication in mice. Whereas Martina et al. (207) reported that SARS-CoV did not infect mice by intracranial inoculation, Subbarao et al. (295) demonstrated that SARS-CoV does infect mice (reported first for BALB/c mice [295] and subsequently for C57BL/6 mice [110]) by intranasal inoculation. The intranasal route was selected by those authors because SARS is a respiratory illness in humans. Intranasal inoculation of BALB/c mice resulted in virus recovery from the upper and lower respiratory tracts. However, mice did not develop clinical signs and even continued to gain weight. The virus replicated to higher titers in the lung than in the nasal turbinates, although it was cleared from the respiratory tract by day 7 p.i. The peak of viral replication was seen in the absence of disease at day 2 p.i., and viral antigen and SARS-CoV RNA were detected by immunohistochemistry and in situ hybridization in bronchiolar epithelial cells during the peak of viral replication. Mice exhibited mild and focal peribronchiolar mononuclear inflammatory infiltrates, with no significant pathology in other organs. Interestingly, primary infection conferred protection in the upper and lower respiratory tracts from subsequent SARS-CoV challenge, and antibody alone protected against viral replication.

C57BL/6 mice also support SARS-CoV replication, with a peak of viral replication on day 3 p.i. and clearance by day 9 p.i (110). Although C57BL/6 mice did not develop respiratory illness, similarly to BALB/c mice, viral RNA localized to bronchial and bronchiolar epithelia. SARS-CoV RNA was detected by RT-PCR at days 1, 3, 5, 7, and 9 p.i. in the lung, brain, heart, liver, and spleen (but not in the kidney). Brain was the only organ for which there was an increase in the number of positive samples with increasing time after infection. Remarkably, live virus was isolated from brain at days 9, 12, and 15 p.i., and SARS-CoV RNA was detected in the hippocampus by in situ hybridization.

Consistent with the above reports, Hogan et al. (137) demonstrated replication of SARS-CoV following intranasal inoculation in both 129SvEv and Stat1^−/− mice (which have impaired IFN responses). Both strains of mice developed bronchiolitis and patchy interstitial pneumonia. However, whereas 129SvEv immunocompetent mice resolved the infection, Stat1-deficient mice progressed to develop diffuse interstitial pneumonia and systemic spread of the virus to the liver and spleen. These results suggested that IFNs may have a role in controlling SARS-CoV infection. In agreement with these data, IFNs inhibit the replication of SARS-CoV in vitro (reviewed in reference 56) and in animal studies (123). In contrast, Glass et al. (110) have demonstrated that SARS-CoV is rapidly cleared in beige, CD1^−/−, and RAG1^−/− mice, suggesting that, at least in B6 mice, the virus is cleared by a mechanism independent of NK, NK-T, and B and T lymphocytes.

In contrast to young BALB/c and B6 mice, aged SARS-infected BALB/c mice showed elevated levels of IFN-α and -γ and TNF-α, suggesting that the proinflammatory cytokine response may be responsible for SARS-associated events (263).

Golden Syrian hamsters are susceptible to SARS-CoV infection after intranasal inoculation (264). Similarly to the mouse model, hamsters did not exhibit clinical disease; the peak of replication virus occurred on day 2 p.i., and virus was cleared by day 7 p.i. In contrast to the BALB/c mouse model, in which the virus is detected only in the respiratory tract and is cleared by day 5, hamsters demonstrate a longer duration of viral shedding from the upper respiratory tract (day 14 p.i), a transient viremia, spread to extrapulmonary tissues (liver and spleen), and most significantly, inflammation in the respiratory tract, which is associated with viral replication. The humoral immune response, as measured by mean serum neutralizing antibody titers 28 days after virus administration, is more robust in hamsters than in mice or African green monkeys.

Taking all the data together, it can be argued that none of the above-mentioned animals (mice, hamsters, nonhuman primates, or ferrets) reproduce the SARS-induced disease observed in humans. Respiratory tracts of these animals can be experimentally infected with SARS-CoV; however, infections remain subclinical, and pathological findings are not consistent with the human disease. For instance, intraalveolar edema, pneumocyte necrosis, or hyaline membranes were not observed in mice (not even in the Stat1^−/− mice). Therefore, animal models for SARS pathogenesis need to be further developed. One of the avenues that require more investigation is to further adapt SARS-CoV to various animal species. This adaptation will likely lead to more pathogenic SARS variants in a particular host. Variants of influenza virus and Ebola virus, which are pathogenic for mice, have been selected by serial passage in the mouse (24, 29, 330). Experience with other coronaviruses such as MHV suggests that adaptation will occur by serial passage (8, 46, 224, 309). It should be emphasized that these animals (in particular mice and macaques) are currently being exploited for vaccine studies. However, our understanding of SARS pathogenesis will be hampered until animal models in which clinical disease and pathological findings mimic SARS human disease are developed.