Dromedary fecal and serum samples used in the study were leftover specimens that had been submitted for pathogen screening (feces) or preventive health screening (serum) to Central Veterinary Research Laboratory (Dubai, United Arab Emirates) during January�?"July 2013. The fecal and serum samples were not obtained from the same animals. None of the dromedaries tested were known to have had contact with bats or horses.

We tested a total of 293 fecal samples: 232 from teenage and adult dromedaries (Camelusdromedarius) (>1 year of age) and 61 from dromedary calves (<1 year of age). Among the 293 samples, 6 were collected in January, 209 in February, 5 in March, 39 in April, 16 in May, 7 in June, and 11 in July 2013.

We tested a total of 59 serum samples: 55 from teenage and 4 from adult dromedaries. The serum samples were collected from female dairy farm or racing dromedaries. The dairy dromedaries were purchased from various countries (e.g., Saudi Arabia, Oman, Sudan, and Pakistan); the number of dairy dromedaries from each country and their lengths of stay in Dubai were not known. The racing dromedaries were from Dubai Emirate.

Viral RNA extraction was conducted as described (20,23). Initial CoV screening was performed by amplifying a 440-bp fragment of CoV RNA-dependent RNA polymerase (RdRp) gene by using conserved primers (5�?�-GGTTGGGACTATCCTAAGTGTGA-3�?� and 5�?�-ACCATCATCNGANARDATCATNA-3�?�and degenerative bases N, R, and D, representing A/C/T/G, A/G, and A/G/T, respectively. After the novel CoV was detected in samples, subsequent screening was performed by using specific primers 5�?�-ACTATGACTGGCAGAATGTT-3�?� and 5�?�-TAATAAGGCGACGTAACATA-3�?�. To amplify a 126-bp fragment of RdRp gene, reverse transcription PCR (RT-PCR) and DNA sequencing were performed as described (20,23).

The fecal samples from 3 dromedaries tested positive for CoV. These samples were cultured in HRT-18G, Vero E6, Caco-2, and LLC-MK2 cell lines.

Three complete genomes of DcCoV UAE-HKU23 (265F, 362F, and 368F) were amplified and sequenced as described (5,12). RNA extracted from fecal specimens was used as template, and a database of CoV genes and genomes (CoVDB, http://covdb.microbiology.hku.hk) (25) was used for sequence retrieval. Sequences were assembled and edited to produce final sequences.

We used EMBOSS Needle (http://www.ebi.ac.uk/Tools/psa/emboss_needle/) to compare the nucleotide sequences of the genomes and the deduced amino acid sequences with those for other CoVs. A neighbor-joining phylogenetic tree with 1,000 bootstraps was constructed by using the Jones-Taylor-Thornton substitution model; gamma distribution among sites was conducted in MEGA5 (26).

According to our protocol (19,20), all samples positive for DcCoV UAE-HKU23 by RT-PCR were subjected to quantitative RT-PCR (qRT-PCR).The assay was performed, as described (27), using a real-time 1-step qRT-PCR with DcCoV UAE-HKU23 primers 5�?�-ATAGCGGCTACACGTGGTGTT-3�?� and 5�?�-TCCCAGCCGCCATAAAACT-3�?� and probe 5�?�-(FAM) CTGTTGTTATAGGCACCACT (BHQ1)-3�?�. To generate calibration curves, we prepared a series of 6 log₁₀ dilutions equivalent to 10¹�?"10⁶ copies per reaction mixture and ran them in parallel with the test samples.

The nucleocapsid proteins of SARS-CoV (betacoronavirus lineage B), Pi-Bat CoV HKU5 (betacoronavirus lineage C), and rousettus bat coronavirus (Ro-BatCov HKU9; betacoronavirus lineage D) were obtained as described (2,16). To produce a plasmid for DcCoV UAE-HKU23 nucleocapsid protein purification, primers 5�?�-CATGCCATGGGCATGTCTTTTACTCCTGGTAAGC-3�?� and 5�?�- CCGCTCGAGTATTTCTGAGGTGTTTTCAG-3�?� were used to amplify the gene encoding the nucleocapsid protein of DcCoV UAE-HKU23. The sequence encoding aa 1�?"672 of the nucleocapsid protein was amplified and cloned into the NcoI and XhoI sites of expression vector pET-28b(+) (Merck KGaA, Darmstadt, Germany). Recombinant nucleocapsid protein was expressed and purified by using the Ni²⁺-loaded HiTrap Chelating System (GE Healthcare Life Sciences, Little Chalfont, UK) according to the manufacturer�?(tm)s instructions. The nucleocapsid protein of MERS-CoV was cloned and purified by using the method described above with primers 5�?�-GGAATTCCATATGATGGCATCCCCTGCTGCACCTC-3�?� and 5�?�- ATAAGAATGCGGCCGCATCAGTGTTAACATCAATCATT-3�?�.

Western blot analysis was performed as described (5) with 1.5 I1/4g of purified (His)₆-tagged recombinant nucleocapsid protein of DcCoV UAE-HKU23 and 1:2,000, 1:4,000, and 1:8,000 dilutions of dromedary serum samples. Antigen�?"antibody interaction was detected by using 1:4,000 diluted horseradish peroxidase�?"conjugated Goat Anti-Llama IgG (Life Technologies, Carlsbad, CA, USA) and the ECL fluorescence system (GE Healthcare Life Sciences).

To determine the possibility of cross-reactivity between antibodies against the nucleocapsid protein of DcCoV UAE-HKU23 and that of other betacoronavirus lineages, we tested 3 serum samples that were positive for antibody against the nucleocapsid protein of DcCoV UAE-HKU23. We used 1.5 I1/4g of purified (His)₆-tagged recombinant nucleocapsid protein of 3 betacoronaviruses (SARS-CoV [lineage B], Pi-BatCoV HKU5 [lineage C], and Ro-BatCoV HKU9 [lineage D]) and 1:2,000 dilutions of serum samples. To determine the presence of antibodies against the nucleocapsid protein of MERS-CoV, we tested the serum samples by using 1.5 I1/4g of purified (His)₆-tagged recombinant nucleocapsid protein of MERS-CoV and 1:2,000 dilutions of serum samples. Antigen�?"antibody interaction was detected as described above.

Anti�?"MERS-CoV antibody detection by indirect immunofluorescence was performed as described (28) with minor modifications. In brief, Vero cells infected with MERS-CoV were prepared as described (28). Camel serum samples were screened at a dilution of 1:160 on infected and noninfected control cells. Antigen�?"antibody interaction was detected by using fluorescein isothiocyanate�?"conjugated Goat Anti-Llama IgG (Life Technologies). Serum samples positive at a screening dilution of 1:160 were further titrated with serial 2-fold dilutions. The indirect immunofluorescence antibody titer was the highest dilution giving a positive result.

The neutralization antibody test was performed as described (28). In brief, starting with a serum dilution of 1:10, we prepared serial 2-fold dilutions of serum in 96-well microtiter plates. For each serum dilution, 0.05 mL of the dilution was mixed with 0.05 mL of 200 MERS-CoV 50% tissue culture infectious doses and incubated at 37A�C for 1.5 h in a CO₂ incubator. Then 0.1 mL of virus�?"serum mixture was inoculated in duplicate wells of 96-well microtiter plates with preformed monolayers of Vero cells and further incubated at 37A�C for 3�?"4 days. Cytopathic effects were observed by using an inverted microscope on days 3 and 4 after inoculation. The neutralizing antibody titer was determined as the highest dilution of serum that completely suppressed the cytopathic effects in at least half of the infected wells.

The number of synonymous substitutions per synonymous site, Ks, and the number of nonsynonymous substitutions per nonsynonymous site, Ka, for each coding region between each pair of strains were calculated by using the Nei-Gojobori method (Jukes-Cantor) in MEGA5 (26). Divergence time was calculated on the basis of RdRp gene sequence data by using a Bayesian Markov Chain Monte Carlo approach as implemented in BEAST version 1.8.0 (http://beast.bio.ed.ac.uk), as described (15,19,29,30). Bayesian skyline under a relaxed-clock model with an uncorrelated exponential distribution was adopted for making inferences because Bayes factor analysis for the RdRp gene indicated that this model fitted the data better than other models tested.

Of the 293 fecal samples tested, 14 (4.8%) were RT-PCR positive for the CoV RdRp gene; 1 (0.4%) of the samples was from an adult dromedary and 13 (21.3%) were from calves (Table 1). Of the 14 positive samples, 11 were collected in April and 3 were collected in May. Ten of the 14 samples were submitted to the Central Veterinary Research Laboratory for routine checking, and the other 4 were collected because the dromedaries had diarrhea. Sequencing results indicated a 126-nt sequence identical to that of equine CoV, a betacoronavirus 1 species in lineage A (betacoronavirus A1).

Attempts to stably passage DcCoV UAE-HKU23 in cell cultures were unsuccessful; no cytopathic effect or viral replication was detected. Real-time qRT-PCR showed that the amounts of DcCoV UAE-HKU23 RNA ranged from 5.7 A- 10⁴ copies/mL to 9.8 A- 10⁷ copies/mL (median 8.4 A- 10⁵) in the 14 fecal samples positive for DcCoV UAE-HKU23 (Table 1).

The 3 complete genomes of DcCoV UAE-HKU23 (GenBank accession nos. KF906249�?"KF906251) were 31,036 bases and had a G+C content of 37% (Table 2). The genome organization is similar to that of other betacoronavirus lineage A CoVs (Figure 1). Additional open-reading frames (ORFs) coding for nonstructural proteins (NSPs) NS2 and NS5 are found. DcCoV UAE-HKU23 and other CoVs in betacoronavirus lineage A possess the same putative transcription regulatory sequence motif, 5�?�-UCUAAAC-3�?�, at the 3�?� end of the leader sequence and preceding most ORFs (Table 3; Technical Appendix Table 1) (19,30).

The characteristics of putative NSPs in ORF1ab of DcCoV UAE-HKU23 are shown in Technical Appendix Table 2. The ORF1ab polyprotein shared 70.7%�?"99.3% aa identity with polyproteins of other betacoronavirus lineage A CoVs. The predicted putative cleavage sites were conserved between DcCoV UAE-HKU23 and other members of betacoronavirus A1 of the Betacoronavirus genus. The lengths of NSPs 1�?"3, 13, and 15 in DcCoV UAE-HKU23 differed from those in equine CoV, porcine hemagglutinating encephalomyelitis virus, and/or HCoV-OC43 as a result of deletions/insertions.

The amino acid sequence of the predicted spike protein of DcCoV UAE-HKU23 is most similar to that of bovine coronavirus (BCoV) and sable antelope CoV, with which DcCoV UAE-HKU23 has 94.1% similarity (Table 2). A comparison of the amino acid sequences of DcCoV UAE-HKU23 spike protein and BCoV spike protein showed 81 aa polymorphisms, of which 24 were seen within the region previously identified as hypervariable among the spike protein of other betacoronavirus lineage A CoVs (31) (Figure 2); this finding suggests that this region in DcCoV UAE-HKU23 is also subject to strong immune selection. BCoV has been found to use N-acetyl-9-Oacetyl neuramic acid as a receptor for initiation of infection (32). Among the 5 aa that may affect S1-mediated receptor binding in BCoV (31), 2 aa (threonine at position 11 and glutamine at position 179) were conserved in DcCoV UAE-HKU23 (Figure 2). However, at positions 115, 118, and 173, aspartic acid, methionine, and asparagine observed in BCoV and were replaced by serine, threonine, and histidine, respectively, in DcCoV UAE-HKU23. A recent report identified 4 aa acids that were critical sugar-binding residues in the spike protein of BCoV (tyrosine, glutamic acid, tryptophan, and histidine at positions 162, 182, 184, and 185, respectively) (32); all 4 aa were also present in spike protein of DcCoV UAE-HKU23 (Figure 2). Another study identified 7 aa substitutions in the spike protein of BCoV that differed between virulent and avirulent, cell culture-adapted strains (31); 5 of the 7 aa from virulent strains were also conserved in DcCoV UAE-HKU23, and amino acid substitutions were observed in the other 2 aa (threonine�+'valine at position 40 and aspartic acid�+'asparagine at position 470). It has also been reported that an amino acid change at position 531 of the spike protein of BCoV discriminated between enteric (aspartic acid/asparagine) and respiratory (glycine) strains (33). At this position, a threonine was conserved in all 3 genomes of DcCoV UAE-HKU23.

NS5 of DcCoV UAE-HKU23 shares 83.5%�?"98.2% aa identity with the corresponding NSPs of betacoronavirus A1 members. In murine hepatitis virus, translation of the envelope protein is cap-independent, through an internal ribosomal entry site (19,30). However, a preceding transcription regulatory sequence, 5�?�-UCCAAAC-3�?�, can be identified upstream of the envelope protein in DcCoV UAE-HKU23, as in other betacoronavirus A1 members (Table 3, Technical Appendix Table 1) (19,30). Downstream to nucleocapsid protein, the 3�?�-untranslated region contains a predicted bulged stem-loop structure of 68 nt (nt position 30,747�?"30,814) that is conserved in betacoronaviruses (34). Overlapping with the bulged stem-loop structure by 5 nt, is a conserved pseudoknot structure (nt position 30810�?"30863) that is important for CoV replication.

Phylogenetic trees constructed by using the amino acid sequences of ORF1b polyprotein, spike protein, and nucleocapsid protein of DcCoV UAE-HKU23 and other CoVs are shown in Figures 3�?"5. The pairwise amino acid identities of chymotrypsin-like protease (3CL^pro), RdRp, helicase, spike protein, and nucleocapsid protein are shown in Table 2. In all 3 phylogenetic trees, DcCoV UAE-HKU23 clustered with other members of betacoronavirus A1, including BCoV, sable antelope CoV, equine CoV, HCoV-OC43, giraffe CoV, porcine hemagglutinating encephalomyelitis virus, canine respiratory CoV, and rabbit CoV (RbCoV) HKU14 (Figure 3�?"5).

Nucleocapsid protein of DcCoV UAE-HKU23 was purified. Prominent immunoreactive bands were visible for 31 (52%) of 59 dromedary serum samples; 25 of the 31 samples had titers of 2,000, three had titers of 4,000, and 3 had titers of 8,000 (Figure 6). Serum samples for all 4 adult dromedaries were positive for DcCoV UAE-HKU23 antibodies, and 27 (49%) of the 55 samples for teenage dromedaries were positive. Band sizes were �%^50 kDa, consistent with the expected size of 50.4 kDa for the full-length (His)₆-tagged recombinant nucleocapsid protein. Only very faint bands were observed when the 3 serum samples positive for DcCoV UAE-HKU23 antibodies were incubated with nucleocapsid proteins of SARS-CoV, Pi-BatCoV HKU5, or Ro-BatCoV HKU9, indicating minimal cross-reactivity. This finding concurs with our previous observation that minimal cross-reactivity occurs between CoVs in different lineages in betacoronavirus (16). For MERS-CoV antibody testing, results were positive for 57 (97%) of the 59 samples by Western blot analysis, for all 59 samples by indirect immunofluorescence, and for 58 (98%) of the 59 samples by neutralization antibody test (Table 4).

The Ka, Ks, and Ka/Ks of the various coding regions in DcCoV UAE-HKU23 are shown in Table 5. The Ka/Ks of all the coding regions in DcCoV UAE-HKU23 was <0.5.

By using the uncorrelated relaxed clock model on RdRp gene sequences, we estimated the date of divergence between DcCoV UAE-HKU23 and BCoV to be �%^46 years ago. We estimated that the 3 strains of DcCoV UAE-HKU23 diverged from their most recent common ancestor in March 2010 (the 95% highest posterior density interval, August 2006�?" September 2012) (Technical Appendix Figure 1).