Bovine-Like Coronaviruses Isolated from Four Species of Captive Wild Ruminants Are Homologous to Bovine Coronaviruses, Based on Complete Genomic Sequences^▿

History of fecal samples collected from CWRs.

The history of the outbreaks of CoV-associated diarrhea and the fecal sampling from sambar deer (Cervus unicolor), a waterbuck (Kobus ellipsiprymnus), a white-tailed deer (Odocoileus virginianus), and a giraffe (Giraffa camelopardalis) were described in our previous reports (27, 59). Briefly, fecal samples were collected from three sambar deer and one waterbuck during a WD outbreak from 1993 to 1994 and from a sable antelope (Hippotragus niger) and a giraffe in 2003 in a wild-animal habitat in Ohio. Fecal samples from a white-tailed deer originated from sporadic diarrhea cases in a wildlife farm in 1994. No cattle were present at either of these premises; cattle herds were within a 2- to 5-mile radius but without direct or any known indirect contact or shared services. All samples were CoV positive by immune EM (IEM) and positive in BCoV antigen capture enzyme-linked immunosorbent assay (ELISA). A single diarrheic fecal sample from a sable antelope was submitted to our laboratory in 2003. The sable antelope was housed in a separate barn about 0.5 miles from the giraffe barn and developed similar signs of diarrhea within 1 or 2 weeks preceding the giraffe diarrhea outbreak. Within the park, the same individuals were responsible for feeding the various animals, and the same dump truck that was used to haul the manure was also used to haul fresh bedding.

CoV strain names and abbreviations.

To be consistent with strain designations for other reported wild-ruminant CoVs (i.e., giraffe CoV [GiCoV]), new or revised strain designations were used for the CWR CoVs. The strains include the original waterbuck CoV (WbCoV) strain OH-WD358 and strains OH-WD358-GnC (GnC indicates that the strain is Gn calf adapted) and OH-WD358-TC (TC indicates that the strain is cell culture adapted); the original sambar deer CoV (SDCoV) strain OH-WD388 and strains OH-WD388-GnC and OH-WD388-TC; white-tailed deer CoV (WtDCoV) strain OH-WD470; and sable antelope CoV (SACoV) strain SACoV-OH1. Different strain names for CWR CoVs were used in our previous paper (59): KI-WB for WbCoV-OH-WD358, KI-D2 for SDCoV-OH-WD388, and WTD for WtDCoV-OH-WD470.

Virus isolation and cell culture passage, plaque induction, and virus purification.

The virus isolation and cell culture passage, plaque induction, and virus purification procedures were performed as described previously (4, 27, 28, 59). Briefly, diluted original fecal samples were clarified by centrifugation and filtered through 0.2-μm filters to remove bacterial contamination. Monolayers of human rectal tumor 18 (HRT-18) cells (59) were inoculated with serial dilutions of the filtered samples, and infection with SDCoV and WbCoV was confirmed by cytopathic effects (observed after the second passage), cell culture immunofluorescence (CCIF) staining, reverse transcription-PCR (RT-PCR), and antigen-capture ELISA. WbCoV-OH-WD358-TC and SDCoV-OH-WD388-TC were plaque purified after seven serial passages on HRT-18 cells and used for RNA extraction and sequencing.

Experimental inoculation of calves.

The preparation of the fecal sample inocula, inoculation methods, and sample collection were described previously (27, 59). Briefly, four 1-week-old Gn calves were inoculated oronasally with 1:10-diluted, filtered original fecal samples from each of the four wildlife species. After necropsy (at day 3 postinoculation for Gn calf inoculated with SDCoV fecal sample and day 6 postinoculation for Gn calf inoculated with WbCoV fecal sample), large intestine contents collected from experimentally infected calves were used as CoV source material for direct RNA extraction and sequencing when appropriate.

RT-PCR.

Total RNA was extracted from clarified and filtered fecal samples (original and from Gn calves) by using an RNeasy minikit (Qiagen, Valencia, CA) according to the manufacturer's instructions. A one-step RT-PCR assay was performed as previously described (9, 27).

IEM.

Briefly, supernatants from 20% fecal samples were mixed with Gn-calf antiserum to the BCoV-Mebus strain, reacted at 4°C for 18 h, and then concentrated by ultracentrifugation and used for IEM as previously described (50).

ELISA and CCIF.

An indirect antigen capture ELISA and CCIF using a pool of three monoclonal antibodies directed against the S, N, and HE structural proteins of BCoV strain DB2 were used to detect CoV antigens and infectious CoV, respectively, in original and Gn-calf fecal suspensions (28).

The sequencing procedures were described in detail previously (27) and applied to all original and Gn-calf fecal suspensions and plaque-purified WbCoV-OH-WD358-TC and SDCoV-OH-WD388-TC.

Sequence analyses.

The reference CoV genome sequences from GenBank compared for phylogenetic analyses are summarized as follows. (i) The group 1 CoVs used included HCoV-229E (NC_002645), HCoV-NL63 (NC_005831), porcine epidemic diarrhea virus CV777 (NC_003436), feline infectious peritonitis virus 79-1146 (NC_007025), and transmissible gastroenteritis virus M6 (DQ811785). (ii) The group 2a CoVs used included HCoV-0C43 ATCC VR-759 (AY391777), HCoV-HKU1 (NC_006577), MHV-A59 (NC_001846), BCoV- Mebus (U00735), BCoV-Quebec (AF220295), BCoV-LUN (AF391542), BCoV-ENT (NC_003045), BCoV-DB2 (DQ811784), and GiCoV (EF424623). (iii) The group 2b CoV used was SARS-Tor2 (NC_004718). (iv) The group 3 CoV used was infectious bronchitis virus Beaudette (NC_001451).

Sequence alignment and phylogenetic analysis were performed as described previously (27).

Nucleotide sequence accession numbers.

The sequences determined in the present study were submitted to GenBank and assigned accession numbers as follows: sable antelope coronavirus US/OH1/2003, EF424621; waterbuck coronavirus US/OH-WD358-TC/1994, FJ425184; waterbuck coronavirus US/OH-WD358-GnC/1994, FJ425185; waterbuck coronavirus US/OH-WD358/1994, FJ425186; white-tailed deer coronavirus US/OH-WD470/1994, FJ425187; sambar deer coronavirus US/OH-WD388-TC/1994, FJ425188; sambar deer coronavirus US/OH-WD388/1994, FJ425189; and sambar deer coronavirus US/OH-WD388-GnC/1994, FJ425190.

Genetic characterization of bovine-like CoVs from CWRs.

Full-length genome sequences of the bovine-like CoVs from four CWR species, waterbuck (Kobus ellipsiprymnus), sambar deer (Cervus unicolor), white-tailed deer (Odocoileus virginianus), and sable antelope (Hippotragus niger) (originating from the 2003 outbreak and used for comparison with our previously sequenced GiCoV originating from the same 2003 outbreak [27]), were obtained and analyzed. The isolation of CoVs from the sambar deer, waterbuck, white-tailed deer, and giraffe (from the same outbreak as SACoV) fecal samples and their antigenic properties were described previously (27, 59). We also performed full-length genome sequencing and genetic characterization for two cell culture-adapted (WbCoV-WD358-TC and SDCoV-WD388-TC) and two Gn-calf-adapted (WbCoV-WD358-GC and SDCoV-WD388-GC) CWR CoV strains. Unlike the GiCoV and other CWR CoVs, we failed to adapt the sable antelope bovine-like CoV (SACoV-OH1) to HRT-18 cell cultures or to successfully passage it in Gn calves, suggesting that this strain was nonviable.

The results of analysis of the eight full-length CoV genome sequences revealed that their organization and structure resembled those of known group 2a CoVs, particularly BCoV (Fig. 1). The genomes were organized into 10 open reading frames (ORFs) and contained the indicated total numbers of nucleotides: WbCoV OH-WD358 (31,010 bp), WbCoV OH-WD358-GnC (30,962 bp), WbCoV OH-358-TC (30,995 bp), SDCoV OH-WD388 (31,011 bp), SDCoV OH-WD388-GnC (31,009 bp), SDCoV OH-WD388-TC (30,995 bp), WtDCoV OH-WD470 (31,020 bp), and SACoV-OH1 (30,995 bp). The observed variations in sequence length among the CWR bovine-like CoVs were mainly due to different numbers of nucleotides at the ends of the 5′-untranslated region (5′-UTR) and the 3′-UTR. All of our sequences were 9 to 16 nucleotides shorter in the 5′-UTR and from 53 nucleotides shorter (OH-WD358-GnC) to 5 nucleotides longer (OH-WD470) in the 3′-UTR; the reason for the differences is unclear, but it could be related to the sequencing strategy.

The results of phylogenetic analysis with representatives from all known groups of CoVs demonstrated that the CWR CoVs belong to CoV group 2a, forming a tight cluster together with reference BCoV isolates (data not shown). To define the relatedness between CoVs isolated in our laboratory and known BCoV isolates, we performed phylogenetic analysis with five genomes of reference BCoV strains from different years of isolation and reflecting the evolution of the BCoV genome during the last 35 years. The BCoV strains Mebus and Quebec, isolated in 1972, represent the earliest group; BCoV-DB2, isolated in our laboratory in 1984, represents an intermediate group; and two isolates from 1998, BCoV-LUN and BCoV-ENT, represent the most-recent BCoV reference strains available. The GiCoV recently isolated in our laboratory (27) was the most-recent isolate of a bovine-like CoV and originated from the same outbreak as the SACoV. The resulting phylogenetic tree for these isolates is shown in Fig. 2. Sequences represented on the tree were grouped according to year of isolation, demonstrating the changes accumulated in bovine and bovine-like CoV genomes over time.

The full-length genome identity among the CWR CoVs and BCoV strains was high. The earliest strains (Mebus and Quebec) had the lowest levels of nucleotide identity, 98.4 to 98.8%, to the most-recent BCoV and BCoV-like isolates, including CWR CoVs (Table 1). The nucleotide identity between strain DB2 isolated in the mid-1980s and the early Mebus and Quebec strains was higher (98.8 to 99%), whereas the levels of identity between strain DB2 and the CWR bovine-like CoVs varied from 99.2% for the most-recent GiCoV and SACoV strains to 99.4 to 99.5% for the earlier WbCoV, SDCoV, and WtDCoV strains. The most-recent reference BCoV strains, ENT and LUN (1998), had 99.5 to 99.6% nucleotide identity with the four CWR bovine-like CoVs, whereas they showed only 98.5 to 98.7% identity with the Mebus and Quebec strains. Almost the same percent identity (99.4 to 99.6%) was observed between the CWR CoV strains that originated from different outbreaks in different years. Thus, similar high genomic nucleotide identities were observed among the four CWR bovine-like CoVs and the most-recent reference BCoV strains and similar degrees of divergence were evident between these genomes and the earliest BCoV strains.

No substantial nucleotide differences were noted between the CWR CoV and reference BCoV genomes in the noncoding regions and regions of a putative RNA pseudoknot and a ribosomal slippery site. No variable positions were present in the 5′-UTR compared to this region in more-recent isolates, whereas the earliest strains (Quebec and Mebus) possessed three unique changes (at positions 59, 96, and 104). Only one substitution common for sambar deer, waterbuck, and bovine DB2 CoV sequences was located in the region adjacent to the −1 ribosomal shift site between ORF1a and ORF1b: a silent mutation 83 nucleotides downstream of the heptanucleotide slippery site. Comparison of the 3′-UTR for our isolates and for reference BCoV strains revealed only two mutations unique for WtDCoV (nucleotide positions given are for the Mebus strain): position 30778 of WtDCoV contained thymine; for other strains, cytosine; and for Mebus and Quebec BCoVs, guanine; in position 30855, adenine was replaced by guanine for the WtDCoV strain. There were no differences in length within the coding region of genomes for CoV sequences except for a single-nucleotide deletion in codon 39 of the 4.8-kDa nonstructural protein (nsp) of the SACoV. All CWR isolates contained the same deletion of AATT, 30 to 33 nucleotides upstream of the 12.7-kDa nsp start codon (28077 to 28080 in the Mebus genome) and had a single-nucleotide insertion after guanine 30956, a common feature for recent BCoV and CWR CoV strains that is in contrast to the earlier Mebus and Quebec BCoV strains.

Amino acid differences within the putative CoV proteins.

Next we compared the amino acid sequences of all proteins between the CWR CoV and reference BCoV strains: BCoV-Mebus, BCoV-Quebec, BCoV-DB2, BCoV-LUN, BCoV-ENT, and the GiCoV-OH3 strain (Fig. 3). The most-variable positions were detected when comparing recent bovine and bovine-like CoV isolates with the early BCoV-Mebus and Quebec strains. Among the 81 variable positions in the ORF1a amino acid sequence, only 14 positions were unique for one or more CWR bovine-like CoVs, including the GiCoV strain (27).

Among the predicted amino acid sequences for all viral proteins, three substitutions in the spike glycoprotein were unique to all CWR bovine-like CoVs compared to this protein in the reference BCoV strains N499S, P501S, and S510T (Fig. 3). Amino acid differences among the CWR CoVs were represented by point mutations, excluding the nucleotide deletion in the 4.8-kDa nsp leading to protein synthesis termination at amino acid (aa) 39 for both SACoV and GiCoV.

The isolates that originated from the same location and outbreak were very similar. However, for the SACoV-OH1 strain from the sable antelope, the 5-aa deletion in the spike protein (as detected for the contemporary GiCoV) was not observed. Throughout all the protein sequences, these isolates differed in only five positions: aa 3711 in ORF1a; aa 507, 542, and 942 in spike; and aa 107 in the 12.7 kDa nsp. In addition, in the small membrane protein of SACoV, a single-nucleotide substitution of G for C was present in the start codon, probably leading to start of translation from the methionine codon 3. Five of these seven mutations were typical only for the SACoV (Fig. 3). In addition, a serine at position 507 of the spike protein was unique for the SACoV as compared to the other BCoV and bovine-like CoV sequences available from GenBank for this region of spike (91 total). Remarkably, only two amino acid differences were observed between the WbCoV and SDCoV 1994 isolates and these were located in the most-conserved proteins, ORF1b (aa 2313) and N (aa 55).

Compared to other CoVs, the WtDCoV possessed three unique amino acid changes (I113V, L147F, and N149S) localized in the spike protein (aa 100 to 150), which is part of the putative receptor binding domain (RBD). Interestingly, two of these changes (L147F and N149S) were present only in WtDCoV and in all recent (2002 to 2003 season) NCD (isolated from newborn calves) and WD (isolated from cattle 1 to 8 years old) BCoV isolates from South Korea (total of 29 sequences) (data not shown). Another substitution, I113V, was only present in half of the Korean BCoVs (data not shown) and additionally in strains R-AH65 (respiratory) and E-AH65 (enteric), BCoVs from young adult cattle recently characterized in our laboratory (71).

Among all CoV strains compared, ORF1b (except for BCoV-Quebec), membrane (M), the proteins of the replicative complex (ORF1a), nucleocapsid (N), and hemagglutinin-esterase (HE) were the most-conserved proteins, with 1.0 to 1.8 substitutions per 100 aa, whereas the nsp's (4.8 kDa, 4.9 kDa, 12.7 kDa, and 32 kDa) with unknown function and subunit S1 of the spike protein were the most variable: the substitution rate for subunit S1 of the spike protein was 6.1 per 100 aa according to the consensus sequence. The amino acid differences among the small nsp's were even higher than for S1, but these differences were caused by a single-nucleotide deletion (4.8 kDa nsp) or substitution (4.9 kDa nsp). For example, the amino acid substitution rate for the 4.9-kDa nsp of the WtDCoV was 32.6 per 100 aa, whereas the nucleotide substitution rate was only 3 substitutions per 100 nucleotides. Although small nsp's are highly variable among BCoV and bovine-like CoV strains, these differences are the result of minor changes at the nucleotide level.

Adaptational changes in putative CoV proteins.

To define the effect of cell culture adaptation in the cloned HRT-18 cell line and Gn-calf passage at the amino acid level, we compared SDCoV and WbCoV sequences from the original fecal sample to the Gn-calf-passaged and cell culture-adapted CoV genomes. Five amino acid differences were observed between the original and the cell culture-adapted WbCoV (aa positions 200 and 2026 in ORF1a, aa 2337 in ORF1b, and aa 412 and 436 in the spike protein), whereas no differences were observed between the original CoV and the calf-passaged strain. For the SDCoV, there were six amino acid differences (aa 200 and 2026 in ORF1a; aa 237 in HE; and aa 244, 482 and 531 in spike) between the original and cell culture-adapted strains and only one difference was observed in the Gn-calf-passaged genome, an ambiguity in aa 33 in the 4.9-kDa nsp (Fig. 3). Of interest, both changes in the ORF1a sequence of the cell culture-adapted strains, amino acid substitutions A200S and S2026L, were common for both WbCoV and SDCoV.

Also noteworthy, the phylogenetic tree based on the full-length spike protein sequences of all BCoV strains available in GenBank (total of 48 strains) (data not shown) resulted in the WtDCoV 1994 isolate grouping with the nine recent Korean WD strains (from 1- to 8-year-old cattle) that are distantly related to most U.S. and Canadian BCoV reference strains, including WD isolates. However, the rest of the CWR CoV isolates analyzed in the present study clustered together with North American reference strains.