Sample collection

All herds were affected by NPD, despite routine vaccination of sows and gilts against ETEC. Two herds (A and E) were satellites in the same sow pool; two others (I and J) were located in proximity to each other and partially shared the same personnel. A total of 50 diarrhoeic animals and 19 healthy controls were included in the study; there were five diarrhoeic animals and two healthy controls from each herd (except for herd I where only one control piglet was obtained). Piglets originated from 60 different litters (three to seven per herd). Pigs treated with antibiotics were excluded. The 45 females and 24 males were all cross-breeds. Diarrhoeic pigs ranged in age from 0.5 to 6 days (median age, 1 day) and the controls from 1 to 5 (median age, 2 days), see Table 1.

The animals were transported alive to the Section of Pathology, Swedish University of Agricultural Sciences (SLU) where they were anaesthetised with an intramuscular injection of tiletamine and zolazepam, 6 mg/kg (Zoletil ^® 50, Virbac Animal Health, Carros Cedex, France) and euthanized by intracardiac injection of pentobarbital sodium, 100 mg/kg (Pentobarbital for Euthanasia 20% w/v solution for injection, Pharmasol Ltd, Andover, UK). Tissue specimens from the distal jejunum (40 cm proximal to the ileocaecal junction, including both intestinal tissue and contents) were collected immediately after euthanasia, snap frozen in liquid nitrogen and stored at -70°C.

Sample preparation

Samples were prepared according to [24] with modifications. In brief, a 2.5 mm transverse section of the distal jejunum from each pig was placed in 1 ml PBS with 5% bovine serum albumin (BSA, Sigma Aldrich, St. Louis, MO, USA). The samples were subjected to three consecutive cycles of freeze-thawing on dry ice and homogenisation by the Omni TH Motor (Omni International, Kennesaw, GA, USA) using soft tissue Omni Tip^™ Plastic Homogenizer Probes (Omni International). After homogenisation, samples were centrifuged for 10 min at 600 x g. The supernatants were collected and filtered through a 0.45 μm sterile filter (Merck Millipore, Billerica, MA, USA) and split into three fractions—one for extraction of viral RNA, one for extraction of viral DNA and one to be stored untreated at -70°C. Samples were kept on ice/dry ice slurry throughout the process. The RNA and DNA samples were treated with a nuclease cocktail containing TURBO^™ DNase (2 U/μL; Thermo Fisher Scientific, Waltham, MA, USA), RNase Cocktail^™ Enzyme Mix (RNase A; 500 U/mL) and RNase T1 (20,000 U/mL; Thermo Fisher Scientific) in 1x TURBO^™ DNase Buffer. RNA and DNA were extracted as previously described [25]. Extracted RNA and DNA were stored at -70°C. A process control was prepared together with the clinical samples. This control contained PBS/5% BSA and was treated the same way as the samples, yielding one control each for RNA and DNA extractions. These controls were used to evaluate possible contaminants from the laboratory environment and reagents during preparation of the clinical specimens.

Sequence Independent Single Primer Amplification

Sequence Independent Single Primer Amplification (SISPA) was performed according to Rosseel et al. 2013 [26]. Two mock metagenomes containing viruses from various families and genome types were used as positive controls—one for DNA viruses and one for RNA viruses. To check for contamination during the amplification, several non-template controls (NTCs) were included in every amplification set. The NTCs, process control, and positive controls, were processed next to the other samples as described below. In short, samples were tagged with two different SISPA tags (FR20RV-12N, 5’-GCCGGAGCTCTGCAGATATCNNNNNNNNNNNN-3’ and K-12N, 5’-GACCATCTAGCGACCTCCACNNNNNNNNNNNN-3’, respectively) in separate reactions. RNA samples were heated to 95°C for five minutes in the presence of the SISPA primer-tags and then put on ice. Samples were subsequently converted to cDNA using Superscript III (Thermo Fisher Scientific) according to manufacturer’s instructions. DNA samples were heated to 95°C for five minutes in the presence of the SISPA primer-tags and dNTPs, then put on ice. After 2 min, the Klenow Fragment (3’→ 5’ exo⁻) (New England Biolabs, Ipswich, MA, USA) was added and samples were incubated according to manufacturer’s instructions. SISPA was performed in two separate reactions for all samples, one with the FR20RV primer (5’-GCCGGAGCTCTGCAGATATC-3’) and one with the K primer (5’-GACCATCTAGCGACCTCCAC-3’) as previously described [26]. However, the polymerase was the KAPA HIFI HotStart ready mix (KAPA Biosystems, Wilmington, MA, USA).

Following PCR, the products were cleaned using the GeneJET PCR purification kit (Thermo Fisher Scientific) and subsequently subjected to cleavage using EcoRV (New England Biolabs). Cleaved products were repurified using the GeneJET PCR purification kit (Thermo Fisher Scientific) and quantified using the Qubit^™ 2.0 Fluorometer (Thermo Fisher Scientific) and Qubit^® dsDNA HS Assay Kit (Thermo Fisher Scientific). Fraction size was checked by gel electrophoresis [27]. Finally, SISPA products from the DNA and RNA fractions with K and FR20RV primers were pooled per piglet and control sample (the positive control with mock metagenomes, the process control, and the NTCs from the SISPA reaction). Thereafter, samples from diarrhoeic piglets within the same herd were pooled, resulting in three samples per herd.

Library preparation and sequencing

Sequencing was performed using the Ion Proton system (Thermo Fisher Scientific) at the National Genomics Infrastructure in Uppsala. 0.4 μg DNA from each sample was fragmented using the S2 system from Covaris (Covaris Inc., Woburn, USA). End repair and adaptor ligation were performed by the AB Library builder (Thermo Fisher Scientific). Samples were amplified according to the Ion Xpress^™ Plus gDNA Fragment Library Preparation protocol and selected with a target size of 250–310 bp (Blue PippinTM, Sage Science, Beverly, MA, USA). Library size and concentration were assessed by a Bioanalyzer High Sensitivity Chip (Agilent Technologies, Santa Clara, CA, USA) and by the Fragment Analyzer system (Advanced Analytical, Ankeny, IA, USA). Samples were pooled together in sets of six, followed by template preparation using the Ion Proton^™ Template OT2 200 v3 Kit on either the Ion OneTouch^™ 2 system or the Ion Chef system (Thermo Fisher Scientific). Samples were then sequenced on the Ion Proton^™ system using the Ion Proton^™ Sequencing 200 v3 Kit on Ion PI v2 chips (200 bp read length, Thermo Fisher Scientific). The Ion Reporter system was running Torrent Suite 4.2.1 at the time of analysis.

Bioinformatics

Taxonomic classification.

Taxonomic classification was performed on read level using Kraken version 0.10.5-beta [32] and the BLASTx equivalent of Diamond (version 0.7.10) using the non-redundant (nr) database from NCBI (version of 2016-01-20) [33]. The same reads were used for predicting protein sequences with FragGeneScan (FGS) 1.19, and the resulting sequences were scanned, using HMMER3, for the viral profiles Hidden Markov Models (HMM) of the vFam database (February 2014 release) [34,35]. The Kraken database used for the classification is an expanded in-house database covering all sequences from the viral (VRL) and phage (PHG) divisions of GenBank (NCBI), as well as the bacterial and archaeal sequences from RefSeq [36]. See Fig 2 for an overview of the data analysis pipeline. The results are only presented descriptively since the individual samples from the healthy animals were not directly comparable with the pooled diarrhoeic samples.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Schematic illustration of the bioinformatic pipeline.

Raw data from the sequencing platform, in FASTQ format, was processed by PrinSeq to select high quality reads. The fractions of high quality reads were then mapped towards the porcine genome, using Bowtie 2, to remove host sequences. The unmapped reads were then classified by three different methods, Kraken, Diamond and HMM/vFam. The resulting three datasets were grouped in two datasets, one representing reads classified on nucleotide and protein level and one representing reads classified only on protein level.

https://doi.org/10.1371/journal.pone.0151481.g002

Detection of mammalian viruses in porcine distal jejunum

Table 3 reports classification of viral sequences on nucleotide level (Kraken) and protein level (Diamond and HMM/vFam). Viral families detected on both nucleotide and protein level included sequences belonging to eight different virus families infecting mammals: Adenoviridae, Anelloviridae, Astroviridae, Caliciviridae, Circoviridae, Parvoviridae, Picornaviridae, and Reoviridae. Viral sequences in healthy samples included hits towards Adenoviridae (n = 3), Anelloviridae (n = 1), Astroviridae (n = 1), Caliciviridae (n = 1), Circoviridae (n = 8), Parvoviridae (n = 2), Picornaviridae (n = 10), and Reoviridae (n = 2). Viral sequences detected in pooled diarrhoeic samples were Anelloviridae (n = 1), Circoviridae (n = 1), Picornaviridae (n = 4), and Reoviridae (n = 2). More than one mammalian virus family was present in ten out of 19 samples from healthy piglets and in two out of 10 pooled samples from diarrhoeic piglets. In four samples from individual healthy piglets (2A, 1B, 2B, 1C, and 1G), sequences belonging to three different mammalian virus families were detected. See Table 4 for a summary of the virus families found in healthy and diarrhoeic samples respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Overview of samples and viruses detected at family level in the distal jejunum of neonatal piglets with and without diarrhoea.

https://doi.org/10.1371/journal.pone.0151481.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Number of samples from healthy piglets (n = 19) and pooled diarrhoeic samples (n = 10) positive for virus families known to infect mammals.

https://doi.org/10.1371/journal.pone.0151481.t004

In several samples classification of reads to viral families occurred only on protein level. In five samples both Diamond and HMM/vFam detected Circoviridae. In addition, Diamond detected Circoviridae (n = 1), Picobirnaviridae (n = 1) and HMM/vFam detected Adenoviridae (n = 2), Circoviridae (n = 4), Caliciviridae (n = 7), Parvoviridae (n = 2), and Reoviridae (n = 1).

The most common virus family detected was Picornaviridae, which was present in 14 out of 29 samples followed by Circoviridae reported in 9 out of 29 samples. The majority of the Picornaviridae sequences were classified on nucleotide level as species Aichivirus C. Sequences classified as Circoviridae in samples (n = 9) were interestingly absent in all diarrhoeic samples but one (E) and were identified on a nucleotide level as po-circo-like virus 21, 22, 51, porcine stool-associated circular virus 5, and fur seal faeces associated circular DNA virus. In addition, some unassigned small circular ssDNA viruses were found in seven healthy piglets. Reoviridae sequences included hits on nucleotide level against the species Rotavirus A and Rotavirus C. Both Rotavirus A and C were detected in two samples (1B and pool B) and only Rotavirus A in one sample (Pool A). In three samples (1B, 2B, and 2E) 18, 20 and 5 reads, respectively, were classified on nucleotide level as the dsDNA virus immunodeficiency associated stool virus (IAS virus).

The composition of mammalian viruses in the intestine of healthy and diarrhoeic neonatal piglets

A number of different virus families known to infect mammals were detected in the present study (Adenoviridae, Anelloviridae, Astroviridae, Caliciviridae, Circoviridae, Parvoviridae, Picornaviridae, and Reoviridae). These findings agree with recent studies on the porcine faecal virome indicating that the general composition of mammalian viruses in the porcine intestine is similar, despite differences in geographical location and age of the sampled pigs [21–23,40–42]. A notable difference, however, is the absence of the coronaviruses TGEV and PEDV in our study. This dissimilarity was expected since TGEV and PEDV have never been detected in Sweden.

Comparisons of the faecal virome between healthy and diarrhoeic pigs in a larger number of animals have so far been limited to two studies investigating piglets at an age of 19–30 days [21,22]. Similar to the present study, putative co-infections were common. Zhang et al. (2014) [21] reported a mean of 2 viral families in the faeces of healthy pigs (n = 29) compared to 3.8 viral families in diarrhoeic pigs (n = 27). Similarly, Shan et al. (2011) [22] found an average of 2.75 viral families in faecal samples from healthy pigs (n = 24) and 3.7 in diarrhoeic pigs (n = 12). When comparing our results with these previous studies, it should be noted that we used intestinal tissue instead of faeces. In addition, the diarrhoeic samples were pooled in the present study (each pool containing samples from five pigs) and thus the number of viral families in individual diarrhoeic piglets is unknown. Interestingly, compared to previous studies, we found an overall smaller number of different virus families both in samples from healthy individual piglets (0–3, mean = 1.5) and in pooled samples from diarrhoeic piglets (0–2, mean = 0.8). This difference is likely due to the lower age of the piglets in the present study (median age 1 day for diarrhoeic piglets and median age 2 days for controls). An additional factor that could have contributed to the low number of viruses in neonatal piglets is the passive immunity received through colostrum.

Another contradiction with previous studies is that we found a greater number of different virus families in healthy piglets compared to the pooled diarrhoeic samples. Two families or more were detected in 2/10 pooled diarrhoeic samples compared to 10/19 from healthy piglets. This could be related to age, as healthy piglets were slightly older. Alternatively, the diarrhoea could have led to a decreased complexity of the intestinal virome. These results call for further studies including analysis of individual diarrhoeic pigs.

Detection of Aichivirus C, eukaryotic small circular DNA viruses, and IAS virus

The most common findings were sequences belonging to Picornaviridae and eukaryotic circular single-stranded DNA viruses classified as Circoviridae. The majority of Picornaviridae-sequences—detected in 10/19 healthy piglets and in 4/10 pooled diarrhoeic samples—were classified at species level as Aichivirus C. A number of studies have reported a high prevalence of Aichivirus C in faeces from both healthy and diarrhoeic piglets [43–45] and an association with diarrhoea is hence controversial. The transmission route of Aichivirus C is thought to be faecal-oral, but the virus has also been detected in sera only three days after birth, demonstrating its capacity to escape the gastrointestinal tract [46]. This, together with the present detection of Aichivirus C in the intestine within 24–48 h from birth, calls for further investigations of possible transplacental or lactogenic transmission routes.

Several sequences in the present study were identified as unclassified Circoviridae (po-circo-like virus 21, 22, 51 and fur seal faeces associated circular DNA virus) or as unassigned small circular ssDNA viruses. These newly discovered unassigned small circular ssDNA viruses, are highly divergent viruses encoding a well-conserved replication initiator protein (Rep) involved in rolling circle replication [47]. Recently, circular Rep -encoding ssDNA (CRESS-DNA) viruses have been identified in a variety of samples including insects, plants, sewage and mammalian faeces, but the origin and host of these viruses are presently unclear [47–50]. Interestingly, we found sequences classified as CRESS-DNA viruses in only clinically healthy piglets. These results agree with the hypothesis that these viruses might be of dietary origin or possibly a commensal part of the eukaryotic virome [49,51].

Interestingly, a small number of sequences from three healthy piglets shared similarities with immunodeficiency -associated stool virus (IASV). This large dsDNA virus has previously been reported in only one study on diarrhoea in humans with advanced-stage HIV infection [52]. The IAS virus is highly divergent from previously characterised viruses and further characterisation of our sequences is required to establish if this in fact is an eukaryotic virus.

Considerations for the use of viral metagenomics in the analysis of clinical samples

As this study relied on a viral metagenomics approach for investigating the virome, the quality of the sequence reads is highly important. The present study generated an average of 220 Mbases of good-quality data per clinical sample, with a mean read length of 185 bases. The NTC and process control followed the same pattern as the clinical samples, but at an analysis level, reported no significant hits towards known viral families. However, the positive control generated 6.3 Gbases of data.

Endogenous viral elements (EVE) were detected within all samples. However, no porcine EVE was detected within the process control and the NTC. In the positive control, porcine EVE sequences were detected which is probably due to the use of porcine cell lines for culturing viruses.

All samples were subjected to three analytical methods suited for detecting sequences in metagenomics datasets, Kraken, Diamond and HMM. As Kraken is only detecting similar sequences at a nucleotide level, two additional methods were used for detecting more distant homologues; the BLASTx equivalent of Diamond and HMMER3 for scanning predicted protein sequences (using FGS) for viral profiles (vFam). Together all three methods enable not only detection of viruses on species level but also a broader detection on family level. As a consequence, some additional viral families were detected only by protein homology. Diamond and HMM/vFam detected putative Circoviridae sequences in five different samples. By classification with Diamond, Picobirnaviridae homologues were classified on species level to porcine picobirnavirus, however this concerns only four reads in one diarrhoeic sample. Porcine picobirnavirus has frequently been detected in pigs but its association to diarrhoea is uncertain [53,54]. Moreover, classifications with only HMM/vFam included additional hits towards Caliciviridae, Adenoviridae, Reoviridae, Parvoviridae and Circoviridae, indicating the presence of proteins homologous to proteins from viruses belonging to these families. However, the results did not show any consistent pattern in relation to the piglets’ health status.

The mammalian intestinal virome in neonates

The present study showed that viral infections were present as early as 24–48 h after birth. The presence of several different eukaryotic viruses in the intestine of healthy piglets suggests that some of the viruses may be regular constituents of the porcine intestinal virome, whereas others may be capable of causing disease under certain circumstances. As an increasing number of eukaryotic viruses are being detected in mammalian tissues without any overt signs of disease in the host, the concept of commensal virus infections has been proposed [19,20]. It has recently also been suggested that eukaryotic viruses may be important for gut homeostasis as the presence of a common enteric RNA virus was able to replace the beneficial functions of commensal bacteria in mice [55].

Studies on the neonatal intestinal virome in mammals younger than a week of age are limited to one recently published study in humans [20]. The number of eukaryotic viruses was low in the first week of life (similar to our study) and thereafter the complexity increases over time. This is in agreement with the development of the porcine intestinal virome, based on the present and previous studies [21,22].