Nidovirus Sialate-O-Acetylesterases

Many viruses achieve reversible attachment to sialic acid (Sia) by encoding envelope glycoproteins with receptor-binding and receptor-destroying activities. Toroviruses and group 2 coronaviruses bind to O-acetylated Sias, presumably via their spike proteins (S), whereas other glycoproteins, the hemagglutinin-esterases (HE), destroy Sia receptors by de-O-acetylation. Here, we present a comprehensive study of these enzymes. Sialate-9-O-acetylesterases specific for 5-N-acetyl-9-O-acetylneuraminic acid, described for bovine and human coronaviruses, also occur in equine coronaviruses and in porcine toroviruses. Bovine toroviruses, however, express novel sialate-9-O-acetylesterases, which prefer the di-O-acetylated substrate 5-N-acetyl-7(8),9-di-O-acetylneuraminic acid. Whereas most rodent coronaviruses express sialate-4-O-acetylesterases, the HE of murine coronavirus DVIM cleaves 9-O-acetylated Sias. Under the premise that HE specificity reflects receptor usage, we propose that two types of Sias serve as initial attachment factors for coronaviruses in mice. There are striking parallels between orthomyxo- and nidovirus biology. Reminiscent of antigenic shifts in orthomyxoviruses, rodent coronaviruses exchanged S and HE sequences through recombination to extents not appreciated before. As for orthomyxovirus reassortants, the fitness of nidovirus recombinant offspring probably depends both on antigenic properties and on compatibility of receptor-binding and receptor-destroying activities.

The first and arguably most critical step of any viral infection is the attachment of the virion to the host cell. The molecules to which virions bind comprise a diverse collection of cell surface proteins, lipids, and carbohydrates (1). Quite a number of en-teric and respiratory viruses use sialic acid (Sia) 1 either as their sole receptor or as an initial attachment factor (2).
Sialic acids, a generic name for a large family of 9-carbon negatively charged monosaccharides, typically occur as terminal residues of glycoconjugates. Of the variety of Sias that occur in nature, many result from differential O-acylation at C4, C7, C8, and C9 of the parental molecule 5-N-acetylneuraminic acid (Neu5Ac) (2,3). Sias make ideal receptors because of their accessibility and abundance, but their wide distribution also has a downside. High avidity binding of virions to nonpermissive cells, to soluble sialylated compounds, or to sialylated mucins of the mucus barrier all might lead to loss of infectivity. In addition, loss of infectivity might result from virus selfaggregation. To ward against such hazards, some enveloped RNA viruses have developed an elegant strategy, which allows high affinity yet reversible binding to Sia receptors; they specify for virion-associated receptor-destroying enzymes (RDEs). So far, two types of viral RDEs have been identified: sialidases (neuraminidases) and sialate-O-acetylesterases.
RDEs with sialidase activity are found in para-and orthomyxoviruses (4,5). For example, the orthomyxoviruses, influenza virus A and B, express two envelope glycoproteins: the hemagglutinin (HA), which mediates binding to Neu5Ac and subsequent fusion of viral and host cell membranes, and the neuraminidase (NA), which catalyzes the removal of Neu5Ac from sialoglycoconjugates. A functional balance between the antagonistic HA and NA activities is critical for efficient entry (5).
An RDE with sialate-4-O-acetylesterase activity was recently discovered in infectious salmon anemia virus (genus Isavirus, family Orthomyxoviridae), a segmented negative-stranded RNA virus of teleosts (11). There is no obvious sequence identity between influenza C virus HEF and infectious salmon anemia virus hemagglutinin-esterase protein (HE), except for the regions that contain the catalytic residues (12). However, influenza C virus HEF1 does share considerable identity with the nidovirus HEs (i.e. class I envelope glycoproteins expressed by toroviruses and certain coronaviruses) (13)(14)(15)(16).
Coronaviruses and toroviruses (separate genera in the family Coronaviridae, order Nidovirales) are evolutionarily related, enveloped (ϩ)-strand RNA viruses of mammals and birds (17)(18)(19). All toroviruses identified so far carry an HE gene (20), but among coronaviruses, only those belonging to (sub)group 2 do. The latter are exemplified by mouse hepatitis virus (MHV) and bovine coronavirus (BCoV). In SARS-associated human coronavirus (HCoV-SARS), which is an early phylogenetic split-off of group 2 (21), and in group 1 and 3 coronaviruses, represented by porcine transmissible gastroenteritis virus and avian infectious bronchitis virus, respectively, the HE gene is absent.
The role of the nidoviral HE proteins and their importance for viral replication is still subject to debate. In the case of BCoV, infection in the presence of esterase inhibitors results in reduced virus titers (22), whereas HE-specific antibodies neutralize the virus (23). Other observations, however, indicate that the HEs are nonessential for replication in tissue culture cells (24,25). In fact, during propagation in vitro of murine coronaviruses, HE expression is readily lost (13,26). 2 Moreover, infection experiments with such HE-deficient MHV laboratory strains would suggest that, at least for these coronaviruses, the HE protein is also dispensable in vivo.
Entry of coronaviruses and toroviruses (i.e. attachment and subsequent fusion of viral and host cell membranes) is mediated by the spike envelope glycoprotein S (27,28). MHV, HCoV-SARS, and group 1 coronaviruses, transmissible gastroenteritis virus, and HCoV-229E use specific cell surface glycoproteins as their main receptor (29 -32), as is probably the case for all other coronaviruses and toroviruses. However, there is accumulating evidence that coronaviruses in addition use Sia receptors (33)(34)(35)(36)(37)(38)(39). Contrary to what might be expected from their similarity to influenza C virus HEF or from their name, the coronaviral HEs seem to play only a minor role in viral attachment to Sias. Rather, the spike proteins appear to be the major Sia receptor-binding proteins (35,38,39). Equine torovirus Berne, which lacks a functional HE gene, induces hemagglutination (40), suggesting that toroviruses utilize Sia as an attachment factor as well and again implicating the S protein as the main candidate for Sia receptor binding.
Toroviruses and the group 2 coronaviruses apparently acquired their HE genes from an unknown source via independent heterologous RNA recombination events (13)(14)(15). Conceivably, the acquisition of an HE protein increased viral fitness by allowing optimal usage of specific Sias as attachment factors. Still, our understanding of the evolution, function, and substrate specificities of the nidovirus HEs is incomplete. Both the toro-and the coronaviral HE genes have diverged considerably (20,38). Coronavirus HEs have been described, which in their sequence are clearly distinct from those analyzed so far and for which the substrate specificity remains to be determined. With respect to the specificities of the toroviral HEs, no information is available at all. Here, we present a comprehensive study of nidovirus HE proteins based upon a combination of phylogenetic, biochemical, and molecular virological techniques. In addition to the sialate-9-O-and sialate-4-O-acetylesterases described so far, we report the identification of a novel type of HE, which occurs in bovine toroviruses (BToVs) and displays a preference for Sias with a di-O-acetylated glycerol side chain. We already noted that exchange of HE genes via homologous RNA recombination is common among toroviruses (20). We now show that naturally occurring rodent coronaviruses also exchange their HE as well as their S genes to an extent not appreciated before. At least on one occasion, this must have resulted not only in an antigenic shift but also in a shift in HE substrate specificity. The recombinational exchanges of surface glycoprotein genes in nidoviruses are reminiscent of the reassortment of HA and NA genome segments in influenza A and B viruses. Parallels between orthomyxo-and nidoviral biology with respect to attachment, envelope glycoprotein function, and consequences thereof are discussed.

MATERIALS AND METHODS
Phylogenetic Analyses-HE sequences were extracted from the NCBI data base for BToV strains Breda (BRV), B145 and B150 (accession numbers Y10866, AJ575379, and AJ575380), porcine torovirus (PToV) strains P-MAR, P10, and P4 (AJ575363, AJ575366, and AJ575364), MHV strains A59, JHM, and 2 (NC_001846, AAA46442, and Discrepancies were noted between sequences deposited for MHV-DVIM in the NCBI data base, indicating that virus stocks have been switched. The DVIM HE nucleotide sequence of Sugiyama et al. (AB008939) is 100% identical to the HE gene of MHV-S, as determined in our laboratory, 3 but only 73% identical to the DVIM HE sequence of Compton and Moore (AF091734). We therefore characterized the MHV-DVIM variant of the Compton laboratory (kindly provided by Dr. S. R. Compton). Reverse transcription-PCR products, together spanning a 10.8-kb region comprising the 3Ј-end of ORF1b through the N gene, were cloned in pGEM-T Easy (Promega) and subjected to sequence analysis using the ABI Prism BigDye Terminators version 3.0 cycle sequencing kit (Applied Biosystems) and an ABI Prism 3100 genetic analyzer. The nucleotide sequence, determined in both orientations and for each region on at least two independent clones, was deposited in the EMBL data base (AY771998). Throughout, we use "DVIM" to refer to the Compton variant.
Amino acid alignments were generated with Multalin, employing the Dayhoff symbol comparison table, and refined manually. Gaps were excluded. Phylogenetic trees were created by the neighbor-joining method, implemented in the program MEGA version 2.1 (44); the reliability of each interior branch was tested by bootstrap analysis with 1000 resamplings.
The number of synonymous substitutions per synonymous site (K s ) was estimated by sliding window analysis for overlapping 500-nucleotide gene segments with a 250-nucleotide step size with the program K-Estimator (45,46), and D ss profiles were generated as described (20).
Construction of HE Expression Vectors-Expression plasmid pBS-BRV-HE and reverse transcription-PCR cloning of the HE genes from torovirus strains B145, B150, P-MAR, P10, and P4 were described previously (15,20). The HE gene of MHV-S was reverse transcription-PCRamplified from total intracellular RNA, isolated from infected cells. 2 cDNA clones of the HE genes of MHV-DVIM, ECoV-NC99, and BCoV Quebec (47) were kindly provided by Drs. S. R. Compton, D. Brian, and A. A. Potter, respectively. The HE protein of BCoV Quebec differs from that of strain Mebus by a single substitution, Val 103 3 Leu, within a motif FYEGVN (residues 99 -104), which is strictly conserved among nidovirus and influenza C virus O-acetylesterases; since this substitution might affect enzymatic activity, Leu 103 was changed back to Val via site-directed mutagenesis. All HE genes were inserted into pBluescript vectors (Stratagene) downstream of the bacteriophage T7 RNA promoter, either as PCR amplicons generated with primer sets listed in Table I (ECoV, MHV-DVIM, and MHV-S) or by recloning restriction fragments from pcDNA3HE (BCoV) (47) or pGEM-T Easy vectors (all torovirus HEs). All constructs were sequenced. Inadvertent nucleotide substitutions were corrected in accordance with published sequences.
Heterologous Expression of Nidovirus HE Proteins and Substrate Specificity Assays-OST7-1 cells (ϳ1 ϫ 10 6 ) were infected with recombinant vaccinia virus vTF7-3 and, at 1 h postinfection, transfected with 2.5 g of plasmid DNA as described (50). At 8 h postinfection, the cells were washed with phosphate-buffered saline and lysed in 200 l of phosphate-buffered saline, 1% Triton X-100. Nuclei and cell debris were pelleted by centrifugation for 2 min at 14,000 rpm and 4°C. Supernatants were tested for sialate-O-acetylesterase activity with p-nitrophenyl acetate (pNPA) (Sigma) as a substrate (42). Lysates from mocktransfected vTF7-3-infected cells were included as negative controls. The specific enzymatic activity per ml of lysate, corrected for endogenous acetylesterase activity present in mock-transfected cells, was calculated as described (7).
Substrate specificity assays were initially performed with equal amounts of specific enzymatic activity based upon hydrolysis of pNPA (7). However, as the activity of the acetylesterases toward pNPA proved of limited predictive value with respect to activity toward Sia derivatives, all assays were instead performed with fixed amounts of cell lysates.
Assays with glycosidically bound Sias were performed essentially as described (42). Peak identities of the chromatograms were confirmed by MS analysis. Spectra were obtained on a Micromass Q-Tof Ultima Global mass spectrometer (Waters) with electrospray ionization and direct infusion. Dried samples were dissolved in 50% aqueous acetonitrile containing 1% NH 4 OH and infused at a rate of 10 l/min. Spectra were recorded in the negative ion mode for 5 min, using a scan time of a Position with respect to the EToV Berne or MHV-A59 genome. b Positions as counted from the initiation codon of the respective nidovirus HE genes.
Metabolic Labeling and Radioimmunoprecipitation Assay-Metabolic labeling of cells and subsequent radioimmunoprecipitation assay with rabbit-␣-BToV BRV serum or with serum from a piglet naturally infected with PToV strain P-MAR (54) were performed as described (15,55). Most variation between and among orthomyxo-and nidoviral HEs occurs in domain R, and there is no obvious conservation of the HEF receptor-binding site in the nidovirus proteins. In domain E, the catalytic Ser-His-Asp triad (Ser 57 , His 355 , Asp 352 ) is strictly conserved. The HEF residues Gly 85 and Asn 117 , of which the NH group and side chain, respectively, together with the NH group of Ser 57 form the oxyanion hole (9), are maintained in most nidoviral HEs, but in those of the BToVs, Gly 85 is substituted by Ser (see Supplemental Fig. 1).

RESULTS AND DISCUSSION
In phylogenetic trees, the HE proteins of corona-, toro-, and influenza C viruses divide into three separate monophyletic clusters ( Fig. 1; see Table II for abbreviations of virus names). This tree topology is consistent with the notion that the ancestors of torovirus, coronavirus, and influenza C virus acquired their HEs through independent heterologous recombination events (15) and that in each case the introduction of an HE gene into the viral genome occurred only once. Consequently, the coronavirus HE genes must all have evolved from one ancestral gene, which was acquired by a group II coronavirus predecessor after the split-off from HCoV-SARS. Similarly, the torovirus HEs all stem from a single ancestral O-acetylesterase.
Whereas sequence variation among influenza C virus HEFs is limited, the corona-and torovirus HEs have diverged considerably (up to 40%). Four distinct torovirus HE lineages can be distinguished, represented by the HEs of BToV-BRV, BToV-B150, PToV-P4, and PToV P-MAR. A similar branching into four groups is seen for the coronavirus HEs ( Fig. 1) (20). Previously, only two lineages were recognized based upon phylogenetic relationships and substrate specificity, with those of lineage A (represented by the HE of MHV-S and RCoV-SDAV) exhibiting sialate-4-O-acetylesterase activity and those of lineage B (represented by BCoV and HCoV-OC43) displaying sialate-9-O-acetylesterase activity (38). Our current analysis shows that, in fact, two additional lineages, C and D, can be distinguished (represented by the HEs of MHV-DVIM and MHV-2 and that of ECoV strain NC99, respectively), for which the substrate specificity remains to be determined (see below).

Torovirus HEs: PToV and BToV HE Proteins Differ in Their
Preference for O-Acetylated Sialic Acid Subsets-To gain more insight into the evolution of the nidoviral O-acetylesterases, we studied the substrate specificities of the HEs of coronaviruses MHV-DVIM and ECoV and of representatives of each of the four torovirus lineages. Since toroviruses, with Berne virus as the sole exception, cannot be propagated in tissue culture, the various HE proteins were produced in the vaccinia virus-based vTF7-3 expression system. Successful expression of each viral acetylesterase was confirmed by radioimmunoprecipitation as-  say (BToV strain BRV and B145 and PToV strain P-MAR) and/or by demonstrating enzymatic activity in cell lysates with pNPA as a substrate ( Fig. 2 and Table III). Note that antisera against BToV-BRV also detected the HE of BToV-B145 but not that of BToV-B150 ( Fig. 2A), illustrating that the two BToV HE lineages (Fig. 1) are antigenically distinct.
These experiments tentatively identified the toroviral HEs as sialate-9-O-acetylesterases. We noted, however, a consistent difference between BToV and PToV enzymes; although the HEs of BToV strains BRV and B145 and those of PToV strains P-MAR and P10 hydrolyzed the synthetic substrate pNPA to similar extents, the BToV enzymes were reproducibly less active toward free Sia substrates (Table III). This raised the question of whether a difference in substrate specificity between the PToV and BToV O-acetylesterases might yet exist. We therefore compared the reactivity of these enzymes toward bovine submaxillary mucin (BSM), a model substrate, which contains a mixture of glycosidically bound Sias (Table IV). The HE of PToV P-MAR hydrolyzed 9-O-acetyl-5-N-glycolylneuraminic acid, Neu5,9Ac 2 , Neu5,8Ac 2 , and possibly also Neu5,7Ac 2 but showed little reactivity toward 5-N-acetyl-7(8),9-di-Oacetylneuraminic acid (Neu5,7(8),9Ac 3 ) (Fig. 5A and data not shown). In contrast, the HE of BToV-BRV exhibited only modest activity toward mono-O-acetylated Sias but converted Neu5,7(8),9Ac 3 very efficiently (Fig. 5A), apparently to Neu5,7Ac 2 and Neu5Ac (Fig. 5B). Similar observations were made for the HEs of BToV strains B145 (not shown) and B150 (Fig. 5A). None of the enzymes showed any reactivity when tested with horse serum glycoproteins, a model substrate containing glycosidically bound Neu4,5Ac 2 (not shown). Our findings suggest that the PToV and BToV HEs indeed differ in substrate specificity, displaying distinct preferences for Sia subsets with either a mono-O-or di-O-acetylated glycerol side chain, respectively.
Coronavirus HEs: The HE Protein of ECoV-NC99 Is a Sialate-9-O-acetylesterase-The HE proteins of MHV strains DVIM/-2 (lineage C) and ECoV-NC99 (lineage D) are distinct from each other and from those of the coronavirus HE lineages A and B (Fig. 1). The HE of ECoV is most closely related to the sialate-9-O-acetylesterases of the BCoV/HCoV-OC43 cluster, but since 4-O-acetylated Sias are prominently expressed in the horse (2), we entertained the possibility of ECoV HE being a sialate-4-O-acetylesterase. However, in cleavage assays with the free Sia substrates Neu4,5Ac 2 , Neu5,9Ac 2 , and ␣Neu4,5,9Ac 3 2Me, ECoV HE was indistinguishable from BCoV HE and preferentially cleaved at C9 (Figs. 3 and 4; Table III). Also, when tested with glycosidically bound Sias of BSM, ECoV HE preferentially de-Oacetylated Neu5,9Ac 2 (not shown). These observations identify ECoV-NC99 HE as a classical sialate-9-O-acetylesterase.  2. Heterologous expression of nidovirus HE proteins. A, detection of vTF7-3-expressed nidovirus HE proteins by radioimmunoprecipitation. vTF7-3-infected cells, (mock-)transfected with HE expression plasmids were metabolically labeled. Lysates were subjected to a radioimmunoprecipitation assay with a rabbit antiserum raised against BToV BRV (␣BRV) or with serum from a piglet, naturally infected with PToV P-MAR (␣P-MAR). Samples were analyzed in SDSpolyacrylamide gels. Molecular size markers are given in kDa. B, detection of vTF7-3-expressed nidovirus HE proteins by acetylesterase assay. Lysates from vTF7-3-infected, (mock-)transfected cells were harvested at 8 h postinfection. Samples (10 l) were tested for acetylesterase activity by incubation with 1 mM pNPA in phosphate-buffered saline in 1-ml reaction volumes. Cumulative specific hydrolysis of pNPA (in nmol), measured at 405 nm and corrected for background acetylesterase activity in mock-transfected cells (y axis), is plotted against reaction time (x axis). sialate-4-O-acetylesterase. However, in contrast to MHV-S HE, DVIM HE did not cleave the 4-O-acetyl groups from the free Sia substrates Neu4,5Ac 2 and ␣Neu4,5,9Ac 3 2Me (Figs. 3 and 4) nor from glycosidically bound Neu4,5Ac 2 present in horse serum glycoproteins (not shown). Instead, DVIM HE hydrolyzed the 9-O-acetyl moieties from free Neu5,9Ac 2 and from ␣Neu4,5,9Ac 3 2Me. Puzzlingly, the enzyme showed little reactivity toward any of the glycosidically bound Sias of BSM (Fig.  5). Substrate specificity of DVIM HE might be restricted to a subset of 9-O-acetylated Sias not present in BSM (e.g. efficient cleavage might require additional Sia modifications or might be linkage-dependent). Recently, we have found that treatment of chicken erythrocytes with DVIM HE destroyed BCoV receptors (i.e. Neu5,9Ac 2 ), thus preventing BCoV-mediated hemagglutination; conversely, treatment of rat erythrocytes with BCoV HE abrogated agglutination by MHV-DVIM. 4 The combined observations strongly suggest that MHV-DVIM binds to (a subset of) 9-O-acetylated Sia receptors and that its HE is a sialate-9-O-acetylesterase. We conclude that murine coronaviruses occur in the field in two types, which differ with respect to Sia receptor usage and HE substrate specificity. Evidence for Horizontal Transfer of HE Genes among Coronaviruses-It is important to stress that the phylogenetic relationships in Fig. 1 reflect the evolutionary history of the HE modules and not necessarily that of the viruses from which these genes were derived, since nidoviruses frequently engage in homologous RNA recombination (57). Indeed, there is evidence for recurrent horizontal transfer of HE genes among toroviruses; ancestors of bovine and porcine field strains apparently replaced their original HE sequences via intertypic RNA recombination with hitherto unidentified toroviruses, FIG. 3. Reactivity of heterologously expressed nidovirus HE proteins toward purified free mono-O-acetylated Sias. Free Sia preparations, enriched for Neu5,9Ac 2 (peak 5, upper panels) or Neu4,5Ac 2 (peak 6, lower panels), but also containing 5-N-glycolylneuraminic acid (peak 1), Neu5Ac (peak 2), Neu5,7Ac 2 (peak 3), and Neu5,8Ac 2 (peak 4) and a non-Sia compound (peak 7) were incubated with vTF7-3-expressed HE proteins of coronaviruses (A), porcine toroviruses (B), and bovine toroviruses (C). Sias were derivatized with 1,2-diamino-4,5-methylenedioxybenzene and analyzed by fluorimetric reversed-phase C 18 HPLC. In the chromatograms, fluorescence at 448 nm (y axis) is plotted against retention time (x axis).
As the sequence data available for MHV-DVIM are incomplete and conflicting (also see "Materials and Methods"), we determined the nucleotide sequence for the 3Ј one-third of the genome for the DVIM stock obtained from the Compton laboratory. Phylogenetic trees generated per gene revealed incongruencies, indicative for recombination (Fig. 6A). In the HE gene, MHV strains DVIM and MHV-2 grouped separately from the other rodent coronavirus strains (A59, JHM, and RCoV-SDAV). In the S gene, MHV-DVIM and MHV-2 were also distinct from strains A59 and JHM but now grouped together with RCoV-SDAV. Finally, in the surrounding genes (ORF1b, ns2, E, M, and N), all rodent coronaviruses were closely related and in essence formed one single cluster.
Further support for recombination was obtained by analysis of the distribution of synonymous substitutions (D ss ; Fig. 6B). D ss profiles based upon pairwise comparison of MHV-DVIM and -2 with JHM and A59 showed that within the HE and S genes, K s values (i.e. the estimated number of synonymous substitutions per synonymous site) (46) are Ն6-fold higher than in the flanking genes. Comparison of RCoV-SDAV versus MHV-A59 yielded roughly constant K s values across the ns2, HE, E, M, and N genes, but in the S gene, the number of synonymous substitutions per synonymous site was 5-fold higher. Most interestingly, pairwise comparison of MHV-DVIM and MHV-2 with RCoV-SDAV revealed that these three viruses are very closely related. RCoV-SDAV only differs from these MHV strains in the HE gene as indicated by a 5-fold increase in the K s values. Surprisingly, however, MHV-DVIM and MHV-2 also seem to be recombinant with respect to each other as indicated by an abrupt 3-fold rise in K s values in the HE gene.
The combined observations indicate that the divergence of the rodent coronaviruses was indeed accompanied by multiple recombination events during which both HE and S sequences were replaced. Apparently, related strains as well as hitherto unidentified coronaviruses served as donors. Our current data do not allow conclusions as to which of the MHV strains represent parental types and which represent recombinant progeny. The exchange among rodent coronaviruses of S and HE genes is highly reminiscent of the reassortment of HA and NA genome segments, which gives rise to antigenic shifts in influenza A viruses. We propose that, in analogy, the observed murine coronavirus diversity, brought about by RNA recombination, is a direct consequence of immune pressure for antigenic variation. In the case of the MHV HE proteins, the genetic exchanges would, however, not only have resulted in an antigenic shift, but, at least on one occasion, also in a shift in substrate preference from 4-O-to 9-O-acylated Sias or vice versa.
The Evolution of Nidoviral Acetylesterases: A Complex Tale of Divergence and Recombination-In comparison with other nidovirus structural proteins, the HEs have received little attention, probably because they are dispensable for replication in vitro and, in the case of the coronaviruses, do not belong to the standard protein repertoire. Yet, the mere fact that in the course of evolution, two groups of nidoviruses independently acquired an HE gene indicates that these virion-associated enzymes must provide a huge selective advantage during the natural infection (15). Our current observations strongly support this view: under field conditions, HE genes may be exchanged via homologous RNA recombination both among coronaviruses and among toroviruses, but HE expression per se is strictly maintained. This is in stark contrast with the loss of HE expression, which occurs during virus propagation in tissue culture cells (13,15,26). 2 The evolution of the coronavirus and torovirus HE proteins has yielded a variety of sialate-O-acetylesterases. Current diversity seems to have arisen not only from antigenic drift but also as a result from adaptations to new hosts and/or new target tissues. Under the assumption that HE substrate preference reflects Sia receptor usage, nidoviruses apparently employ Neu5,9Ac 2 , Neu4,5Ac 2 , and Neu5,7(8),9Ac 3 as initial attachment factors. As first reported here, HEs with a preference for the latter Sia subtype occur in bovine toroviruses, both in the BRV/B145 and B150 lineages. The latter observation is of particular interest, since the B150-type BToVs are recombinant viruses, which apparently originated from an exchange of HE sequences between FIG. 5. Reactivity of heterologously expressed nidovirus HE proteins toward glycosidically bound 9-O-acetylated sialic acids of BSM. A, BSM was incubated with vTF7-3-expressed nidoviral HE proteins for 0, 1, 3, and 18 h at 37°C. After acid hydrolysis and derivatization with 1,2-diamino-4,5-methylenedioxybenzene, Sias were analyzed by fluorimetric reversed-phase HPLC. Graphs depict the enzymatic hydrolysis of Neu5,7(8),9Ac 3 (upper panel) and Neu5,xAc 2 (i.e. Neu5,7Ac 2 , Neu5,8Ac 2 , and Neu5,9Ac 2 combined) (lower panel) over time. B, sialic acid composition of BSM after (mock) treatment for 0, 1, and 3 h with BToV-BRV HE. In the bar graphs, relative amounts of Neu5,7(8),9Ac 3 (5,7 (8), 9), Neu5Ac (5), Neu5,7Ac 2 (5,7), Neu5,8Ac 2 (5,8), and Neu5,9Ac 2 (5,9) are depicted as percentages of total BSM Sia content. a B145-type ancestor and an as yet unidentified torovirus donor (20). This raises the question of whether the donor virus already encoded a sialate-7 (8)9-di-O-acetylesterase or if the substrate preference of the newly acquired HE was adjusted after the horizontal gene transfer. Porcine toroviruses at least display a preference for Neu5,9Ac 2 receptors.
The divergence of the coronavirus HEs produced four distinct lineages. Whereas the HEs of lineage A preferentially hydrolyze 4-O-acetylated Sias, those from the three other lineages de-O-acetylate Sias preferentially at C9. Based upon phylogenetic tree topologies ( Fig. 1 and data not shown), we propose that the ancestral coronavirus HE was a sialate-9-O-acetylesterase, which was converted into a sialate-4-O-acetylesterase presumably through the accumulation of mutations in the E domain. The substrate binding site of the coronavirus sialate-4-O-acetylesterases must differ essentially from that of their 9-O-specific homologues; bringing the 4-O-instead of the 9-Oacetyl group in close proximity of the principal nucleophile would require a rotation of the Sia with respect to the catalytic site. In which coronavirus and host the transition in substrate preference took place is not known; nor is the direction of the various recombinational exchanges that occurred during subsequent divergence of the rodent coronaviruses. Possibly, the sialate-4-O-acetylesterase arose in an as yet unknown coronavirus to be introduced into the genome of a DVIM-type MHV via horizontal gene transfer. Irrespective of the precise course of events, it has led to the natural occurrence of two types of closely related murine coronaviruses. These variants differ with respect to substrate preference of their HE proteins, implying that they use different Sia receptors as attachment factors to initiate their infections in the mouse. It is of note that both 4-O-and 9-O-acetylated Sias as well as less common Sias like 5-N-acetyl-9-O-lactylneuraminic acid are present in mouse colon and trachea and in virtually all other target organs of MHV (58). 5 The consequences of Sia receptor specificity for host cell tropism and disease outcome may be considerable (59) and remain to be explored.
The HE-expressing coronaviruses seem to have adopted a lifestyle that in many ways resembles that of influenza A viruses. There are parallels with respect to virion composition and host cell attachment in that Sia receptor binding, and RDE activities are performed by two different surface glycoproteins (HA and S versus NA and HE, respectively). These envelope proteins occur in multiple subtypes, which differ in antigenic properties and in Sia receptor specificity casu quo substrate preference. Antigenic shifts may occur by exchanging the genes (or parts thereof) for the envelope glycoproteins, in concert or individually, through reassortment or RNA recombination. However, ultimate success of the recombinant offspring would depend not only on its potential to escape host and population immunity but also on proper matching of receptor-binding and RDE activities. Depending on their host species, influenza A virus strains differ in their preference for particular Sia subsets (e.g. avian strains preferentially bind to sialoglycoconjugates containing ␣2,3-linked Sias, whereas human strains use ␣2,6-linked Sia-receptors). A mismatch between receptor pref-FIG. 6. Evidence for multiple exchanges of S and HE sequences during the radiation of rodent coronaviruses. A, rooted neighborjoining trees depicting the phylogenetic relationships among rodent coronavirus strains. Trees were constructed for the 3Ј-most 1476 residues of ORF1b and for the ns2, HE, S, E, M, and N genes with BCoV serving as an outgroup. Confidence values are indicated at the major branching points. Branch lengths are drawn to scale; the scale bar represents 0.05 nucleotide substitution per site. B, distribution of synonymous substitutions (D ss ) in the 1b, ns2, HE, S, E, M, and N genes of rodent coronavirus variants. Sequences were compared pairwise by sliding window analysis; the number of synonymous substitutions per synonymous site (K s ) was estimated for overlapping 500-nucleotide gene segments with a 250-nucleotide step size. D ss profiles were gene-rated by plotting calculated K s values against nucleotide position. As a reference, the genes for 1b, ns2, HE, S, E, M, and N, drawn to scale, are depicted as boxes (top). C, simplified schematic representation of the recombinant genome composition of rodent coronaviruses. Genes are represented as boxes. Related S and HE genes are color-coded. The genome structure indicated by a question mark has not yet been identified in the field. erence and RDE substrate specificity brought about by reassortment causes a decrease in viral fitness, which can be restored by compensatory mutations in HA, NA, or both proteins (5). A similar functional link might exist between the nidovirus S and HE proteins. It is of interest that in rodent coronaviruses, two major types of S proteins can be distinguished (JHM/-S versus DVIM/RCoV; Fig. 6, A and B), that (parts of) S and HE genes are exchanged individually, and that three of four possible S-HE combinations have been found so far (Fig.  6C). Intriguingly, the rodent coronavirus strains MHV-DVIM and RCoV-SDAV express S proteins, which are closely related (90% identity) but have HE proteins with different substrate specificities. Future studies should (i) determine the Sia receptor preference of these viruses, (ii) determine whether also in these viruses the S proteins are the major Sia-binding proteins, and (iii) if so, establish whether the receptor-binding and RDE activities of their respective S and HE proteins match.
Our findings provide a framework for further structure-function analysis of nidovirus sialate-O-acetylesterases, which should aim at the elucidation of the molecular basis for substrate specificity and explore the potential of HE-directed antiviral drugs. Moreover, the HEs may serve as novel tools to study sialic acid biology.