Altered Gut Microbiota Profiles in Sows and Neonatal Piglets Associated with Porcine Epidemic Diarrhea Virus Infection

Porcine epidemic diarrhea virus (PEDV) is a devastating cause of diarrhea in pigs worldwide. Most of studies have focused on molecular and pathogenic characterization of PEDV, whereas there were limited studies in understanding the role of gut microbiota (GM) in viral-associated diarrhea. Here, using the Illumina MiSeq platform, we examined and compared the impact of PEDV infection on the GM of sows and their piglets less than 10 days old. Our results showed that PEDV caused alternations in the structure and abundance of GM from levels of phylum to genus, and even species. For sows, a significant decrease of observed species was found in diarrheal sows than that in healthy sows (p < 0.05). The unweighted and weighted UniFrac distances also revealed considerable segregations of GM structure among healthy, asymptomatic, and diarrheal sows. For piglets, Bacteroidetes, the dominant bacteria in healthy piglets, were replaced by Firmicutes in asymptomatic and diarrheal piglets. The abundances of Fusobacteria and Proteobacteria were also remarkably increased in asymptomatic piglets and diarrheal piglets when compared to those of the healthy piglets. Our findings demonstrated that PEDV infection caused severe perturbations of GM, reduced probiotic bacteria, and enriched pathogenic bacteria.

have also demonstrated GM perturbations from phylum to genus levels 28,29 . Both aforementioned studies showed that Fusobacteria was higher in the PEDV infected pigs than in the healthy groups, and either Verrucomicrobia 28 or most commensal bacteria [29][30][31] were found significantly lower in infected piglets due to the dysbiosis caused by PEDV. However, both studies just focused on the nursing or >10-day-old suckling piglets, but did not observe the GM variations in their mother sows and newborn piglets under 10 days old. Therefore, further investigations on GM perturbations in PEDV-infected sows and newborn piglets are indeed needed.
In the present study, we attempted to elucidate the GM profiles in naturally PEDV infected sows and healthy sows, and the offspring delivered by these sows. We firstly presented a systematically comparative evaluation on the microbiota profiles of healthy sows and their relevant offspring by using the universal primers covering 16 S rRNA (small subunit of the ribosome in bacteria) V1~V3 hypervariable regions and the Illumina MiSeq platform. We then analyzed the GM alterations in sows and <10-day-old newborn piglets with symptomless and diarrhea during variant PEDV infection.

Results
Sequencing consistency of mock community (MC). To evaluate the primers 27 F and 534 R for 16 S rRNA V1~V3 hypervariable regions and analytic workflows for MiSeq next generation sequencing (NGS) platform, a MC was constructed by using 15 known bacterial species belonging to 13 genera, in three phyla ( Table S1a). The equimolar pooled genomic DNAs of MC along with that of clinic samples were amplified and then sequenced by an Illumina MiSeq paired-end (PE) 300 run. After quality control, the reads were processed using Quantitative Insights Into Microbial Ecology (QIIME) v 1.9.0. At the phylum level, 99.27% reads were assigned to three phyla (Firmicutes, Proteobacteria, and Actinobacteria); 92.35% assigned reads hit the 13 genera of MC (Table S1b). The abundance of pooled bacteria determined by NGS showed an average of 85% agreement rate with that of qPCR (Detailed information was provided in Appendix 1. Experimental Procedures). Hence, it was proved that the primers 27 F and 534R-based PCR could accurately amplify the 16 S rDNA of the microflora and thus the methodology established was appropriate for unveiling the microbial communities in clinic samples.

Sequencing data of clinical samples. Thirty-four newborn piglets under different healthy conditions
(healthy piglet (HP, N = 4), asymptomatic piglet (AP, N = 15), and diarrheal piglet (DP, N = 15)) along with their respective mothers (health sow (HS, N = 3), asymptomatic sow (AS, N = 4), and diarrheal sow (DS, N = 4) were sampled. The 45 samples produced a total of 2,152,604 raw pair-end 300-bp reads from the MiSeq runs. After quality filtering and demultiplexing, 1,674,142 clean reads were obtained with an average of 31 (Table S2). The low abundance OTUs under 0.005% cutoff were removed according to the criteria set by a previous report 32 . To reduce the bias generated by sequencing depth, the reads of each sample were normalized to 15,400 by QIIME script single_rarefaction.py (the minimum read number of the samples assigned to OTUs was 15,400 in this study) ( Table 1). Alpha diversity indices, including Shannon, Chao1, Good's coverage, Observed species, and Simpson, were introduced to describe the sequencing data in clinical samples ( Fig. 1 and Table S3). Rarefaction curves of these indices plateaued a saturation phase after 5,000 reads per sample, suggesting adequate reads of all samples for the microbiota investigation (Fig. S1). The values of Good's coverage of all these subjects were >95%, indicating sufficient sequencing depth for the investigation of gut microbiota in the six groups in this investigation.
At the phylum level, the percentiles of dominant phyla of gut microbes in HS were 47.39% for Bacteroidetes, 34.57% for Firmicutes, 5.74% for Proteobacteria, 4.64% for Spirochaetes, 4.05% for Tenericutes, and 1.36% for Verrucomicrobia, respectively, while in HP, the abundances of these bacteria were 76.29%, 21.18%, 2.12%, 0.02%, 0.03%, and 0%, respectively (Table S4a). Fibrobacteres, Planctomycetes, Cyanobacteria, Actinobacteria, Lentisphaerae, Thermi, Fusobacteria, and Synergistetes were in low abundances in both HS and HP samples. The HS had significant higher abundances of Proteobacteria, Spirochaetes, Tenericutes, Verrucomicrobia, and Fibrobacteres than that in HP, while the HP showed a very high abundance of Bacteroidetes. At the genus level, an even sharper segregation of GM structure was observed between HS and HP. The HP had 15.29% unclassified OTU at genus level. In contrast, unclassified OTU in the HS was as high as 60.11%. Bacteroides, the most abundant genus in HP, accounted for about half of the abundance on genus level in HP, while there was only 1.39% in HS (Table S4b). A greater percentage of Prevotella, Parabacteroides, and Lactobacillus were observed in HP, while Oscillospira, YRC22, CF231, Streptococcus, Ruminococcus, and Treponema were higher in HS. These discrepancies on GM structure and abundance might explain the segregations on PCoA.
Shifts in gut microbial diversity and abundance in healthy, asymptomatic and diarrheal sows. A total of 3,403 OTUs were clustered based on the 327,099 non-chimeric reads from sow samples (HS: among the three groups (Table S3). The observed species was significantly lower in DS than that in HS (p < 0.05). The unweighted and weighted UniFrac distances revealed sharp segregations among these groups, indicating a core division in GM community and abundance among HS, AS and DS samples ( Fig. 2B and E). Meanwhile, the UPGMA tree also showed the similar result (Fig. S2A).
Bacteroidetes, Firmicutes, Proteobacteria, Spirochaetes, and Tenericutes in HS, AS and DS were the predominant phyla, accounting for more than 96% of the overall abundance of bacteria (Table S5a), and each of the remaining phyla accounted for <3% in relative abundances. A significant increase in the ratio of Firmicutes/Bacteroidetes was observed in AS when compared to that in HS (p < 0.05). As compared, the ratio of Firmicutes/Bacteroidetes in DS was irregular and showed a slight higher than that in HS on average. A significant increase of Actinobacteria was found in AS, represented by Coriobacteriaceae. The abundance of Proteobacteria was decreased from HS to AS and DS, while the percentage of Spirochaetes was increased. Despite a small quantity of Verrucomicrobia, a trend of reduction was observed from HS to AS and DS.
To identify specific biomarkers of GM in samples from HS, AS, and DS, the high-dimensional class comparison method of linear discriminant analysis (LDA) effect size (LEfSe) was employed to evaluate the statistical difference of GM abundance from the taxonomic phylum to species. A cladogram representing the structure of the host-microbiota axis showed significant shifts of GM among HS, AS and DS. A total of 41 phylotypes were discovered as high-dimensional biomarkers for separating GM among HS, AS and DS samples. Sixteen of these phylotypes relatively enriched in abundances in HS group. Of which, seven belonged to phylum Bacteroidetes, six were in Proteobacteria, and three in Firmicutes, respectively; eighteen phylotypes in phyla Firmicutes, Bacteroidetes, Fibrobacteres, and Actinobacteria were dominant in AS; and seven phylotypes in phyla Tenericutes and Firmicutes were leading phylotypes in DS ( Fig. 3 and Table S6a). Interestingly, a butyrate-producing bacteria Clostridium butyricum was significantly higher in abundance in HS samples than that in AS and DS samples (p = 0.028). The family Desulfovibrionaceae and genus Desulfovibrio, which is positively associated with inflammation, were observed in higher percentage in AS than that in HS and DS groups.
The diarrheal sows had a significantly different GM structure compared to that in the healthy sows ( Fig. S3A and Table S6b). Fifty-four phylotypes were found significantly different between HS and DS, of which 45 phylotypes were significantly higher in HS samples, and 9 were dominant in DS samples. The differential species in DS samples were mostly in the order Clostridiales in Firmicutes and family Peptostreptococcaceae in Tenericutes. When an overall difference of GM between HS and AS groups was compared, 31 phylotypes were found higher and 7 were lower in the HS group than the AS group ( Fig. S3B and Table S6c). Most of the preponderant bacteria in HS fell into phyla Bacteroidetes and Proteobacteria, while the dominant GM in AS samples fell into phyla Firmicutes and Actinobacteria. The HS gut had a significantly different microbial structure with that in the DS samples ( Fig. S3C and Table S6d). A total of 36 phylotypes were found to be different between AS and DS groups, with 28 higher in AS samples and 7 higher in DS samples. The AS had more Verrucomicrobia in GM, while the DS had more Tenericutes. OTUs in all of the piglet samples (HP: 621, AP: 697, and DP: 519) were generated from 1,259,623 clean reads. Unexpectedly, no significant differences in alpha diversity indices among the groups were observed ( Fig. 1 and Table S3). The unweighted UniFrac distance was not able to separate the HP, AP and DP samples (Fig. 2C); while the weighted UniFrac could slightly separate the HP samples from AP samples as well as the DP samples, indicating a difference in GM abundance between HP and AP/DP samples (Fig. 2F). The UPGMA tree could discriminate HP from other two groups, while couldn't separate AP and DP samples (Fig. S2B). These beta diversity discrepancies between HP and AP/DP revealed the HP had different GM structure and composition than that of AP/DP samples.
The five most abundant phyla, Bacteroidetes, Firmicutes, Proteobacteria, Fusobacteria, and Actinobacteria, accounted for over 99% of all piglet subjects. The remaining phyla accounted for < 1% abundance of bacteria (Table S5b). In particular, the GM in HP samples were largely dominated by Bacteroidetes (78.80% ± 16.38%, s. d.), significantly higher than those in AP and DP samples (p = 0.000), while the AP and DP samples contained a large number of taxa belonging to phylum Fusobacteria (AP: 15 LEfSe algorithm on microbial abundances of GM in piglets showed significant differences between the normal, asymptomatic and diarrheal piglets. Among the three groups, 35 phylotypes from phylum to species were discovered as high-dimensional biomarkers ( Fig. 4 and Table S6e). The HPs had substantial higher abundances of GM in phyla of Bacteroidetes and Proteobacteria, due to an increase of Prevotella, Odoribacter, and Parabacteroides in phylum Bacteroidetes, and Stenotrophomonas, in phylum Proteobacteria, respectively; the DPs had distinct GM compositions mostly belonging to phyla Firmicutes, represented by Enterococcus, Veillonella, and A total of 24 phylotypes were significantly different between HP and AP samples. Of which two were overrepresented in HPs and the remaining 22 phylotypes were reduced in APs. Interestingly, the butyrate-producing bacterium, Clostridium butyricum, was overrepresented in the GM in HP samples. In contrast, the pathogenic Clostridium perfringens was overrepresented in the GM in AP group ( Fig. S4A and Table S6f). Fifty one phylotypes showed significant differences between the GM of HP and DP samples ( Fig. S2B and Table S6g). The biomarkers for separating GM in HP and DP groups at the phylum level were Bacteroidetes and Firmicutes, as the HP samples had higher abundance in Bacteroidetes in GM while the DP samples had more Firmicutes (Fig. S4B and Table S6g). The bacteria in genera Bacteroides and Prevotella made a significant contribution to the predominant phylum of Bacteroidetes in HP samples, while Clostridiales and Enterococceae highly contributed to the increase of Firmicutes in DP samples. Interestingly, Clostridium butyricum, a species in genus Clostridium, was the predominant phylotype in GM of HP group, while potential pathogenic Clostridium perfringens, another species in the same genus, showed higher abundance in the GM of DP group. Compared to the DP samples, the AP samples had 18 distinct phylotypes with either higher or lower abundances ( Fig. S4C and Table S6h).

Discussion
In this study, we determined and compared the fecal microbiome in healthy, asymptomatic, diarrheal sows and their piglets associated with PEDV infection. We systematically investigated the microflora profiles in sows associated with PEDV infection, but also evaluated the bacterial alternations from their healthy, asymptomatic and diarrheal offspring associated with PEDV infection. To our knowledge, this is the first report of using next-generation sequencing technology to compare the gut microbiota in asymptomatic and diarrheal sows naturally infected with PEDV, and their newborn piglets as well. During the outbreaks of severe diarrhea in pigs in 2010, we observed different clinical presentations between sows and suckling piglets, especially the neonates within 10 days old, in the field. Sows manifested a mild diarrhea which last for about one to two days, accompanied with mild anorexia and listlessness. However, neonates aged within 10-day had a severe diarrhea along with dehydration, vomiting and high morbidity and mortality as previously reported 1,9 . Diarrhea in neonates infected by PEDV could last for over 7 days and the treatment of antibiotics is usually inactive, and most of them (50-80%) died from dehydration within a week after infection. This urges us to explore underlying reasons responsible for the huge discrepancies in clinical symptoms and outcomes between PEDV-infected sows and piglets.
The gastrointestinal tract ecosystem is complex, and plays critical role in both health and disease 11,33 . The bacteria colonized in this niche is dynamic and subject to changes based on time, age, breed and many other factors 34 . It has been reported that Firmicutes and Bacteroidetes are the dominant phyla in mammal's gut regardless of fluctuation and influence from multiple factors such as diseases, age, breed, diet, and gender 35 . In this study, we found that phyla Bacteroidetes and Firmicutes were the predominant phyla in the GM from HS, AS, DS, and HP, which was consistent with the aforementioned study. Changes in microbial memberships and proportions were observed due to the infection of PEDV, although the infected sows had just showed one-or two-day diarrheal disease. Furthermore, interesting membership shifts were detected in sows infected with PEDV. Clostridium butyricum was decreased in proportion of AS and DS (p = 0.028). This bacterium could produce short chain fatty acids (SCFAs), mainly butyrate, and contributes to complex carbohydrate breakdown in the gut, which is greatly related to colonic health 36,37 . While the family Desulfovibrionaceae and genus Desulfovibrio had been previously found to be positively associated with inflammation, and also showed inhibition of butyrate oxidation, resulting in disruption of intestinal barrier function 38 . The increased abundance of family Desulfovibrionaceae and genus Desulfovibrio in AS and DS might lead to inflammation in musoca and facilitated the infection of PEDV. Liu et al. (2015) reported a diminishment of Verrucomicrobia in PEDV infected piglets compared to the healthy controls 28 . Despite small proportion of Verrucomicrobia in GM of sows, we also saw a similar trend of decrease of this phylum from HS to AS and DS. Similar to Liu's study, we found the reduction of genus akkermansia accounts mostly for the phylum Verrucomicrobia. It was found that Akkermansia was associated with immune regulation by inducing Foxp3 regulatory T cells, which could attenuate tissue inflammation 39 .
The bacteria colonization in piglet intestine is believed to begin immediately after birth, and is mainly influenced by their mothers as previously reported 40 . The mother's feces, skin, and the surrounding environment contain a large amount of bacteria, which are likely to be the source of intestinal microbiota of piglets 41 . The factors that have been recognized to be associated with enteritis and diarrhea in suckling piglets included viruses (e. g. rotavirus, coronavirus), bacteria (Escherichia coli, Clostridium perfringens, and Clostridium difficile) and parasites (Cryptosporidium spp., Giardia spp., Cystoisospora suis and Strongyloides ransomi). In this study, we found that the microbiota in the gut of neonates was greatly affected with PEDV infection. For instance, Bacteroidetes abundance was significant dropped from HP group (78.80% ± 17.33% s. d.) to both AP (23.05% ± 20.15% s. d.) and DP (21.09% ± 23.11% s. d.) groups. PEDV infection has also caused the increase of abundances in Firmicutes, Proteobacteria, and Fusobacteria, respectively. The significant increase in abundance of Enterobacteriaceae was the major reason for the increase of Proteobacteria in AP and DP groups. In addition, it was reported that Fusobacteria was greatly increased in the GM of PEDV-infected pigs 28,29 . A similar finding, a sharp rise of Fusobacteria abundance in GM of AP and DP, was also observed in GM of piglets in this study. However, the increased abundance of genus Fusobacterium was the key contribution for the rise of phylum Fusobacteria, which was in accordance with Koh's report (2015) 28 , but inconsistent with Liu's findings (2015) 27 . Fusobacteria, as typically dominant obligate anaerobic bacteria clustered in mouth, have been identified in a wide range of clinical anaerobic infections including oral, dental, gut and brain, and correlated positively with catarrhal appendicitis of mucosal surface [42][43][44] . Recently, it was reported that Fusobacterium plays a vital role in facilitating colorectal cancer 45 , and Fusobacterium nucleatum could definitely cause intestinal inflammation 46 . Perhaps the proportion increase of Fusobacterium in AP and DP could cause bowel inflammation. We observed a higher abundance of Bacteroidetes in GM of piglets than that in sows, which confirmed a previous study that the Bacteroidetes decreased as the swine gain weight and aged 47 . Our observations of decreased Bacteroidetes and increased Firmicutes amount and the ratio of Firmicutes/Bacteroidetes in asymptomatic and diarrheal samples from both sows and piglets are in agreement with a recently published study on pigs 47 . The bacteria in phylum Bacteroidetes are dominant bacteria of GM in healthy piglets, which includes those with high proportions in the predominant genera Bacteroides, Prevotella, and Parabacteroides. It is presumed that the shifts in GM composition and abundance correspond to the changes of diet and essential metabolic functions 48 . Another healthy associated bacterium, Clostridium butyricum, was significantly higher in abundance in HP than that in AP and DP. Studies revealed that Prevotella, Bacteroides, and Coprococcus are short-chain fatty acid producing bacteria and Clostridium butyricum are butyrate-producing bacteria. Therefore, the decreased proportion of these bacteria may lead to the de-stabilized of gut microbiome and may facilitate the infection of PEDV. In addition to the decrease of probiotic bacteria in asymptomatic and diarrheal pigs, we observed the increase of pathogenic bacteria in these infected pigs. Recently, Clostridium perfringens and Clostridium difficile were found to be ones of the most common bacterial species associated with diarrhea in pigs worldwide 49 . In this study, the relative abundances of Clostridium perfringens in both diarrheal sows and piglets were 10 to 15 times higher than those of healthy piglets. The Clostridium perfringens type C is rarely found in intestine of healthy piglets, but as a primary pathogen, it can also colonize on gut following other pathogens, such as transmissible gastroenteritis virus (TGEV), rotavirus, and PEDV 49 . Clostridium perfringens might be also as a primary pathogen to advance the infection of PEDV. Following the infection of PEDV, it may aggravate the PED as a secondly colony. Unlike sows, a decrease trend in proportion of Verrucomicrobia was found in the HP samples when compared to that in AP and DP samples. To date, there are no reports of a relationship between hosts and Verrucomicrobia. Dubourg et al. 50 reported an increase level in the colonization of human gut by Verrucomicrobia due to dysbiosis, i.e., microbial imbalance of gut microbiota after broad-spectrum antibiotic treatment. Therefore, the Verrucomicrobia detected in the swine gut might be due to the use of antibiotics, since antibiotics are usually used in pig farms for preventive or therapeutic purposes.
Unexpectedly, we observed a high relative abundance of Bacteroides fragilis in all three piglet groups. A similar finding was previously reported in young children with diarrhea 16 . Enterococci were presented in both asymptomatic and diarrheal samples, including AS and AP. However, none of the sequences from HS and HP samples were assigned to this genus. Enterococcus is a large genus of Gram-positive lactic acid bacteria of the phylum Firmicutes, commonly resident in the intestines of humans and animals, including the pig 51 . Important clinical infections caused by Enterococcus include urinary tract infections, bacteremia, bacterial endocarditis, diverticulitis, and meningitis 52 . It has been reported that there is an association of Enterococci to neonatal porcine diarrhea, and also association with resistance to several antibiotics 53,54 , which could explain why the inefficiency of antibiotic treatment of these diarrheal piglets.

Conclusion
In conclusion, our study demonstrates that dysbiosis in gut microbiota occurs in asymptomatic and diarrheal sows and their piglets. This dysbiosis caused by PEDV infection resulted in unbalance and translocation of gut microbiota from phylum to genus, or even species levels. The abundances of inflammatory or diarrhea associated pathogenic bacteria were significantly increased while the beneficial bacteria, e.g., butyrate-producing bacterium, were reduced in asymptomatic and diarrheal sows. PEDV infection also switched the dominant bacteria in gut of piglets. Fusobacteria and Proteobacteria were remarkably increased in AP and DP, while the abundances of short-chain fatty acid producing bacteria and the butyrate-producing bacterium were remarkably decreased in DP and AP. Our findings from this study demonstrated that the gut microbiota were altered due to PEDV infection, which caused the severe perturbations of GM in both sows and piglets through diminishing probiotic bacteria and enriching pathogenic bacteria. Clinical samples. Clinical samples were collected from December 2013 to March 2014, from an independent commercial farm with about 600 sows in Jiangxi, China, which suffered with severe diarrhea (for details, please refer to in the supporting information, SI). Thirty-four piglets under different healthy conditions (healthy piglets (HP, N = 4); asymptomatic piglets (AP, N = 15), and diarrheal piglets (DP, N = 15), were sampled, along with their respective mothers (health sow (HS, N = 3), asymptomatic sow (AS, N = 4), and diarrheal sow (DS, N = 4). Fresh feces were immediately collected, transported to our laboratory on ice and then stored at −80°C until use. The details on study design are described in Supporting Information, SI. Data processing, bioinformatics analysis and statistical analysis. The raw reads were put for quality control after base calling, and then were subjected to pipeline QIIME v1.9.0 (http://qiime.org/tutorials/illu-mina_overview_tutorial.html) for OTU picking and taxonomy 56 . Alpha and beta diversities were also analyzed by UPGMA and PCoA (for details, see Appendix 1. Experimental Procedures). An algorithm for discriminating high-dimensional biomarker of genomic features, LEfSe (http://huttenhower.sph.harvard.edu/galaxy) was used to discover and explain the differences of microbiome from phylum to species levels between PEDV infected and non-infected groups as described previously 57 . The R packages Vegan, multcomp and MASS were used to perform statistical analysis. ANOVA was used to test the difference of alpha diversity indices and OTU/bacteria abundances between samples by using R package multcomp 58 .