Novel Porcine Epidemic Diarrhea Virus (PEDV) Variants with Large Deletions in the Spike (S) Gene Coexist with PEDV Strains Possessing an Intact S Gene in Domestic Pigs in Japan: A New Disease Situation

Since late 2013, after an absence of seven years, outbreaks of porcine epidemic diarrhea virus (PEDV) infection have reemerged and swept rapidly across Japan, resulting in significant economic losses. In this study, we report the emergence, mixed infection, and genetic characterization of 15 novel field PEDV variants with large genomic deletions. The sizes of deletion varied between 582 nt (194 aa) and 648 nt (216 aa) at positions 28–714 (10–238) on the S gene (protein). Among 17 PEDV samples isolated from individual pigs, all of them contained at least two distinct genotypes with large genomic deletions, and 94.1% of them were found to consist of strains with an intact S gene. These variants were found in eight primary and nine recurrent outbreaks, and they might be associated with persistent PEDV infection in the farms. Full-length S and ORF3 genes of eight variants derived from 2 samples were characterized. This is the first report of mixed infections caused by various genotypes of PEDV and would be important for the studies of viral isolation, pathogenesis, and molecular epidemiology of the disease.


Introduction
Porcine epidemic diarrhea (PED) is a highly contagious enteric disease characterized by vomiting, acute watery diarrhea, and in particular, a high mortality rate in suckling piglets, resulting in substantial economic losses [1]. The etiologic agent of this disease is the PED virus (PEDV), which is an enveloped, positive-sense, single-stranded RNA virus that belongs to the order Nidovirales, family Coronaviridae, and genus Alphacoronavirus. The disease was initially reported in England in 1971 [2] and Belgium in 1978 [3]. Since then, it has been identified in many swine-producing countries such as those in Europe and Asia, notably Belgium, Czech a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Ethics statement
The samples were obtained from a naturally infected animal in the field by qualified veterinarians as a part of normal veterinary care and diagnostic testing procedures. Samples of intestine used were collected from dead piglets and the fecal samples were non-invasively collected immediately after excretion. Therefore, no aggressive operation had been conducted against pigs for sampling purpose. No piglets or other animals were sacrificed for the purposes of this study. The University of Miyazaki DNA recombinant Committee approved the protocol for our cloning work of PEDV (protocol number 2014-464).
Amplification of the partial S gene, full-length S gene, and ORF3 gene Reverse transcription was performed using random primers and oligo-dT primers from Reverse Transcription System Kits (Promega, Madison, WI, USA). To determine the complete S gene sequence, five primer sets (CS1-CS5, Table 1) were used to amplify five DNA fragments spanning the entire S gene. These primers were designed on the basis of the published sequences of reference PEDV strains [5,9,33]. EmeraldAmp MAX PCR Master Mix kit (Takara Bio, Japan) containing high fidelity DNA polymerases was used for performing the PCR amplifications under the following conditions: denaturation at 94˚C for 2 min; 35 cycles of denaturation at 94˚C for 30 s, annealing at 54˚C for 30 s and extension at 72˚C for 1 min; a final extension step at 72˚C for 10 min. When a partial S gene was amplified using the primer pair CS1-F/CS1-R from generated cDNA, one or two distinct bands of unexpected shorter sizes were observed in each PCR product in addition to the predicted band (Fig 1). The same result was observed when one-step RT-PCR tests (AccessQuick RT-PCR System kit, Promega Corp, WI, USA) were performed from extracted RNA and PCR tests using other kits containing high fidelity DNA polymerases (KOD FX, KOD FX neo, and Blend-Tag Plus Kits, Toyobo Co., Japan) were performed from cDNA templates to amplify the CS1 fragment. The unexpected and expected DNA bands were also observed and distinguished upon performing PCR amplifications with longer fragments containing the first portion of the S gene (primer pairs CS1-F/CS2-R, CS1-F/CS3-R, and CS1-F/CS4-R). The PCR products containing the CS1 fragment were individually gel-extracted and cloned into pCR2.1-TOPO vectors (Thermo Fisher Scientific, US). The cloning reaction was then chemically transformed into One Shot TOP10 To obtain the sequence of the full-length S gene and ORF3 gene of PEDV strains coexisting in a single PEDV sample, a fragment containing these two contiguous genes was amplified using the primer set FS-F/ORF3-R and the EmeraldAmp MAX PCR Master Mix kit (Takara Bio, Japan). The PCR conditions were as follows: denaturation at 94˚C for 2 min; 35 cycles of denaturation at 94˚C for 30 s, annealing at 55˚C for 30 s and extension at 72˚C for 5 min; a final extension step at 72˚C for 10 min. The expected size of this amplified product was 4955 bp. Subsequently, the PCR products were gel-extracted and cloned into pCR-XL-TOPO plasmid vectors (TOPO XL PCR Cloning Kit, Invitrogen, US). After transformation into competent cells, plasmid DNA isolation was performed as described previously. The purified plasmid DNA was sequenced using the primer sets CS1-CS5 and the primer set ORF3-F/ORF3-R [15]. However, when the fragment containing entire S and ORF3 genes was amplified, PCR products derived from only two samples JKa-295 and JMi-277 were visualized distinctly in agarose gel 0.8%. Therefore, these two samples were selected for cloning and sequencing of the fulllength S and ORF3 genes.

Sequencing analysis
All sequences were determined in both directions, and sequencing reactions were performed using BigDye Terminator v3.1 Cycle Sequencing Kits and an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems, US). Nucleotide and deduced amino acid sequences were edited, assembled, and analyzed using Geneious v9.0.4 software (http://www.geneious.com) and molecular evolutionary genetics analysis (MEGA) software version 6.0 [34]. The resultant nucleotide sequences were deposited in GenBank under the following accession numbers: KU363044-KU363130. Phylogenetic trees based on the nucleotide sequences of the full-length S gene and ORF3 gene were constructed by the maximum likelihood method using the Tamura-Nei substitution model with a discrete gamma distribution; then gaps/missing data treatment was done by complete deletion, and bootstrap tests of 1000 replicates in the MEGA program.

Results and Discussion
A total of 248 field samples (70.7%) collected from animals at 145 farms (82.4%) experiencing acute diarrhea in six prefectures were positive for PEDV. The result indicated that PEDV has a high prevalence rate in Japanese pig herds. Notably, PED outbreaks were commonly observed in PED-vaccinated farms, and there was a significant difference in the mortality rate (0-100%) for piglets and durations of diarrhea among PEDV-infected farms. This prompted critical questions regarding the molecular characteristics and virulence of circulating PEDV as well as efficacy of PEDV vaccines that have been used in Japan.
To monitor heterogeneity among PEDV strains, 52 PEDV-positive samples were selected for further sequence analyses. However, when CS1-F/CS1-R primers were used for PCR to amplify the first portion of the S gene, one or two distinct bands of unexpected size (approximately 0.2-0.3 kbp) were observed in each PCR product of 17 PEDV samples (32.7%) in addition to the predicted band (872 bp). The PCR products were individually gel-extracted, cloned, and sequenced.
Sequencing results illustrated that the unexpected bands were fragments with large deletions in the S gene, and a total of 15 novel PEDV genotypes with large deletions were identified ( Table 2 Although two genotypes feature 606-nt deletions, the deletions were located at two distinct positions (94-699 and 46-651) in the S gene. Furthermore, the genotype with a 203-aa deletion at residues 10-217 featured a 5-aa insertion (YAMFF) at the position of the deletion. Interestingly, the identical nucleotide sequence corresponding to the insertion (TACGCCATGTTCTTT) existed in the middle of the S gene (nt 2091-2105). Additionally, all PEDV variants of the S protein identified in this study were predicted to have signal peptide cleavage sites using the SignalP 4.1 server (http://www.cbs.dtu.dk/services/SignalP-4.1). This indicates that the S proteins of the variants are most likely expressed as structural proteins.
Surprisingly, mixed infection with multiple variants was found in all PEDV samples (17/17) collected from individual pigs in primary or recurred PED outbreaks. Particularly, in sample JKa-295, we found nine coexisting genotypes including a genotype with an intact S gene and eight genotypes with deletions of 194, 197, 201, 204, 205, 211, 215, and 216 aa (Table 3). Identical genotypes were also isolated from different farms of various prefectures that are geographically distant from each other. In addition to the variants, PEDV strains with an intact S gene were detected in 94.1% of the samples (16/17), with no intact gene detected in sample JAo-56. These results revealed that complex mixed-genotype infection of PEDV had occurred in individual pigs and farms.
Full-length S and ORF3 genes of PEDV strains derived from samples JKa-295 and JMi-277 were cloned and sequenced. Identical nucleotide sequences of both S and ORF3 genes were distinguished and excluded, resulting in the identification of 11 individual strains (Table 3). Three strains that were obtained had an intact S gene (4161 nt, 1386 aa), whereas eight other variants possessed deletions of 582 (194 aa), 591 (197 aa), and 645 (215 aa) nt. Notably, the variants with a 197-aa deletion were identical to that of US strain PC177-P2, which was proposed to be obtained during adaptation of the virus in cell culture [17]. In addition, the 194-aa deletion from the variants in this study was similar to that of the Tottori2 strain that has been recently isolated from domestic pigs in Japan [19]. However, apart from Tottori2, all variants with a 194-aa deletion in our study were found to coexist with other variants as well as strains with an intact S gene.
The PEDV S protein is a type I glycoprotein [35]. Its identified epitope regions include the COE domain (499-638 aa) and epitopes SS2 (748-755 aa), SS6 (764-771 aa), and 2C10 (1368-1374 aa) [36][37][38]. It was demonstrated that the large deletions found in the S gene of the variants did not occur in these four regions, and there was no significant difference in the epitope regions between the PEDV variants and PEDV strains with an intact S gene (S1 Fig).
Sequencing analysis indicated that strains found in a single PEDV sample displayed genetic diversity in their S genes. In sample JMi-277, three strains with an intact S gene had 99.64-99.78% nucleotide sequence identity with each other and 99.64-99.89% nucleotide sequence identity with three variants possessing the 197-aa deletion when pairwise distances were measured using MEGA v6.05 software. Within JKa-295, three variants with the 197-aa deletion shared 99.72-99.92% nucleotide sequence identity with each other but had 99.38-99.55% nucleotide sequence identity with JKa-295fSde204Co25, which possessed the 204-aa deletion (S1 Table). On the contrary, all strains found in JKa-295 and JMi-277 had the greatest   [39]. Furthermore, phylogenetic analysis based on the full-length S gene classified these Japanese strains into the North American (highly virulent) type as shown in Fig 2A. It is suggested that the variants originated from North American type of PEDV. The sequencing results illustrated that the ORF3 gene of the 11 strains in JKa-295 and JMi-277 had the same length of 675 nt, and it did not possess any deletion found in known PEDV vaccine strains [15,27,40]. All of the variants with large deletions have the same ORF3 nucleotide sequence excluding JKa-295fSde194Co25. Nucleotide identity of the ORF3 gene between the variants and intact S strains was 98.81-99.41% which was lower than the 28 field isolates (99.6-100%) collected from 5 prefectures in Japan during 2103-2104 [15]. Eight distinct SNPs between the variants and the strains with an intact S gene were also identified at positions 136, 154, 178, 228, 302, 561, 645, and 657 in the ORF3 gene (S2 Fig). Phylogenetic analysis revealed that the strains with an intact S gene and the variants fell into two distinct clades (Fig 2B). The result indicated that the variants were not vaccine-related, and might have evolved independently as compared to the intact S strains which coexisted in the individual samples.
In this study, variants with large deletions in the S gene were found in eight primary and nine recurrent outbreaks from 16 pig farms, and they mostly (94.1%) coexisted with PEDV strains with an intact S gene. In four farms (Te, Sa, Is, and At), the variants were not found in the primary outbreaks, but they appeared in subsequent pandemics. This indicated that PEDV variants could be transmitted from farm to farm and/or that the proportion of the variants in samples collected at the primary outbreaks was insufficient to permit detection of the variants by PCR. Additionally, PED outbreaks in Japan since late 2013 were proposed to be caused by the invasion of recent overseas strains [15,31]. Taken together, it is possible that in the recent pandemic, both the PEDV variants and the strains with an intact S gene were introduced together from other countries, after which they spread across Japan. However, the earliest PED outbreak of the variants possessing the large S deletions occurred on Kagoshima on December 2013 (sample JKa-292) while the first PED case in Japan was reported earlier in Okinawa on October 2013. Presently, no PEDVs possessing the large deletions in the N-terminus of the S gene analogous to the Japanese variants was reported in other countries except PC177-P2 which was concluded to derive from cell adaptation in vitro. Therefore, the variants might have arisen spontaneously during the PEDV spreading in Japan. Further studies are required for the confirmation.
In this study, due to the coexistence of the variants and PEDV strains with an intact S gene in all of the samples excluding JAo-56, it might not be accurate to evaluate the virulence of individual strains based solely on the information from clinical signs and mortality rates recorded from the pig farms. In JAo-56, we could not identify any PEDV strain with an intact S gene. In the farm from which this sample was collected, the sows had never been vaccinated with modified live PEDV. Piglets in the farm manifested various severities of diarrhea from mild to severe, and the mortality rate reached 72.4% (670/926). The mortality rate was much higher than the mortality rate recorded in the epidemic during which Tottori2 was isolated (0%) [19]. Therefore, the large deletion in the variants observed in this study might not necessarily lower the virulence of this virus. Further investigations are needed to elucidate the virulence of variants in PED outbreaks.
After the primary outbreaks occurred in 2013 and the first half of 2014, there was a sharp decrease of PED cases that were reported during the second half of 2014 and 2015 in Japan. Among the few recurrent PED outbreaks reported in 2014 and 2015, all the samples collected from pig farms in Miyazaki (6 farms) and Aichi (2 farms) were found to contain the PEDV variants. In those farms, diarrhea in piglets was observed for a long time (up to three months), and outbreaks of PED were reported to frequently recur in intervals from three weeks to several months (S2 Table). Although many strict measures for establishing a high level of security were implemented such as quarantine of reposition, strict visitor policies, banning of unwashed vehicles, controlling of carcasses, movement of the caretakers on the farms and using virucidal disinfectants; PED still recurred in the farms. Therefore, it is suggested that mixed infection of PEDV variants with large genomic deletions may play an important role in the persistence of PEDV in pig farms.
Defective interfering (DI) particles are a common observation in many virus families [41]. DI particles are defective for replication in the absence of the product of the deleted gene, and its replication requires the presence of the complete functional virus particle to co-infect a cell with it, in order to provide the lost factors [42]. The PEDV variants wherein large genomic deletions were observed may be DI particles. However, we speculated otherwise because the reported PEDV strains PC177-P2 and Tottori2 possessing the same deletions found in the variants of this study could replicate well in cell culture and cause cytopathic effect [17,19]. Besides, the mutant strain PC177-P2 can propagate well and induce diarrhea in piglets [43]. A deletion of 215 aa (position  in the N-terminal domain of the S protein analogous to the deletion observed in the variants revealed a not impaired propagation of PEDV in vivo [44]. Moreover, only variants with large S deletions were found in the field sample JAo-56, which indicated that the variants were able to independently propagate well in pigs. Additionally, all the deletions found in the variants are in-frame and the S proteins were predicted to express as a functional structure protein. Taken together, it is suggested that the variants with the large S deletions in this study were not DI particles. Functions of the spike protein N-terminal domain of PEDV and other alphacoronaviruses were previously elucidated [44]. The loss of this domain occurred in a number of alphacoronaviruses correlates with a loss or reduction of enteric tropism. The most obvious example of this phenomenon is transmissible gastroenteritis virus (TGEV) that has a dual tropism targeting the respiratory and gastrointestinal tract of the pig. Meanwhile, porcine respiratory coronavirus (PRCV) is a natural variant of TGEV which lacks the S protein N-domain. It has reduced considerably its enteric tropism and primarily replicated in the respiratory tract then, produced an attenuated disease in comparison with parental TGEV strains. Another typical example is feline coronaviruses that have 2 pathotypes: feline enteric coronavirus (FECV) causing a clinical mild or asymptomatic disease and feline infectious peritonitis (FIPV) inducing a highly pathologic disease. FIPV is thought to be derived from FECV by natural mutation. However, compared to FECV, PIPV has a large deletion in the S protein N-domain which results in the loss of tropism for enterocytes [44,45]. These examples revealed an important role of alphacoronavirus spike N-domain for viral replication in the enteric tract. The N-terminus (residues 1-249) of the S protein was demonstrated to be responsible for the sialic acid binding activity of PEDV [44]. The loss of enteric tropism observed in TGEV was specifically associated with the loss of the sialic binding activity that located in the N-domain. However, how the sialic acid binding activity affects the tropism of these alphacoronaviruses remains unclear. The S protein N-domain of some PEDV strains were thought to be dispensable for replication of PEDV in vitro [44][45][46]. The large deletions in the S protein of the PEDV variants in this study indicated that this domain might also be dispensable in vivo. In the present study, the characteristics of large deletions found in the Japanese variants were highly similar to those observed in the PRCV that have a 224 or 227-aa deletion in the N-terminal domain of the S protein [47]. PRCV can persist in pig herds throughout the entire year, or it can disappear temporarily from a herd for several months and reappear subsequently. The emergence and widespread prevalence of PRCV lessened the clinical impact of TGE in the United States and Europe due to cross-protective immunity with TGEV [48]. Thus, the PEDV variants identified in this study might have biological characteristics analogous to PRCV and may persist in the pig farms of Japan. Unfortunately, we did not collect lung tissues from PEDV-infected pigs in these outbreaks, and virus isolation for the individual variants has not yet been completed. Therefore, it remains unclear whether the PEDV variants found in this study exhibited altered virulence, pathogenicity, and tissue tropism from the strains with an intact S gene. In future studies, we will continuously seek to answer this question.

Conclusions
This is the first report describing PED outbreaks with mixed infections of various genotypes. Our study identified 15 field PEDV genotypes with large deletions in the S gene that have been circulating in Japan from 2013 to 2015. The full-length S gene and ORF3 gene of eight variants and three strains with an intact S gene were characterized. The variants coexisted with intact S strains in high frequency and they might be associated with persistent PEDV infection in pig farms in Japan. Such a reemerging and mixed infection of the variants in this study have put PED in a new situation indicating this disease has become more complex in terms of viral isolation, pathogenesis, and epidemiology. The mechanism by which the PEDV variants were generated and evolved remains unclear. Hence, further studies are required to elucidate biological properties such as changes in tissue tropism and the virulence patterns of the new variants.  Table. Nucleotide sequence identity based on the full-length spike genes of 11 Japanese PEDV field strains identified in this study and PEDV reference strains Ã . (DOCX) S2 Table. Status of pig farms where the PEDV samples using in this study were collected. (DOCX)