Co-infection Status of Novel Parvovirus's (PPV2 to 4) with Porcine Circovirus 2 in Porcine Respiratory Disease Complex and Porcine Circovirus Associated Disease from 1997-2012.

As global pig health diseases, porcine respiratory disease complex (PRDC) and porcine circovirus associated disease (PCVAD) generate substantial economic losses despite pigs been vaccinated against the primary causative virus, highlighting the importance of understanding virome interactions and specifically co-factor infections. Established primary endemic pathogens for PRDC include porcine circovirus 2 (PCV2), porcine reproductive and respiratory syndrome virus (PRRSv) and swine influenza virus (SIV), and PCV2 aetiology in interaction with other co-infecting viruses can result in PCVAD. Porcine parvovirus (PPV) 1 is a well characterised virus with an available vaccine preventing reproductive failure in sows. However, whilst novel PPV 2 to 7 viruses have been identified since 2001, their viral pathogenic potential in clinical and subclinical disease remains to be determined. Therefore, this study has sought to develop a better understanding of their potential role as associated co-infections in PRDC and PCVAD by examining archival samples for the presence of PCV2 and the novel parvoviruses PPV2-4 from clinically diseased pigs across production age stages. Epidemiologically, the novel PPV2 was found to be the most prevalent within the fattener age group with PPV2-4 statistically associated with pig respiratory disease and enteric ulcers. Additionally, statistical modelling by latent class analysis (LCA) on veterinary pathology scored pigs found a clustering co-factor association between PPV2 and PCV2, suggesting the novel PPV may be involved in PRDC and PCVAD. Phylogenetic analysis of novel PPVs revealed the PPV2 capsid evolution to be diverged from the original strains with a low nucleotide homology of 88-96% between two distinct clades. These findings determine that novel PPV 2-4 viruses are statistically associated as co-infectors in a diseased pig population, and significantly detected PPV2 clustering co-infection frequency with PCV2 in PRDC and PCVAD diseased pigs through LCA analysis.


| INTRODUC TI ON
The novel porcine parvoviruses (PPV2-PPV4) are emerging viral agents and globally endemic in domestic pig and wild boar populations (Afolabi et al., 2018;Cadar, Csagola, et al., 2013;Cságola et al., 2012;Huang et al., 2010;Zhang et al., 2014). Evidence of emerging PPV association with clinical disease has been broadly reported Opriessnig et al., 2014;Saekhow & Ikeda, 2014). However, as these single-stranded (ss) DNA novel viruses are not cell culture adapted, experimental infection has been hindered, resulting in limited knowledge of viral pathogenesis or the relevance of these viruses both clinically and economically. Porcine circovirus type 2 (PCV2), a ssDNA virus, is the known causative agent of the multi-factorial disease, post-weaning multisystemic wasting syndrome (PMWS) in pigs (Allan et al., 1999;Segales, 2012). PCV2 disease is facilitated by a co-factor infection suppressing the pig immune system allowing PCV2 to replicate to a higher rate. Since the initial characterization of PMWS, further disease associations have been documented and referred to as PCV2-associated disease (PCVAD) to encompass a wide variety of clinical symptoms and pathologies including enteric, respiratory, porcine dermatitis nephropathy syndrome (PDNS) and reproductive and congenital disorders (Opriessnig et al., 2007). Additionally, a multi-factorial disease was recognized as porcine respiratory disease complex (PRDC) and is observed in pig herds where respiratory symptoms with coughing, fever, anorexia and gastric ulcers occur in recently weaned piglets and in older pigs. PRDC can be attributed to a wide range of viruses such as PCV2 and potentially PCV type 3 (Kedkovid et al., 2018), PRRSv and swine influenza virus (SIV) which is known to cause significant damage to the epithelial respiratory airways where secondary bacterial pathogens can then cause infection (Opriessnig et al., 2011).
A next-generation sequencing (NGS) study on PRDC porcine lung samples has associated the presence of PPV2, PPV3, PPV6 and torque teno virus to PRDC diseased pigs (Qin et al., 2018).
The novel porcine parvoviruses PPV2-4 may have emerged approximately seventy years ago and possess a high nucleotide substitution rate similar to RNA viruses (Cadar, Lorincz, et al., 2013).
Through frequent recombination, the positive selection of porcine parvoviruses has the capacity to further adapt and circulate amongst pig herds, which merits continuous phylogenetic mapping and the monitoring of disease association. PPV1 is considered an endemic virus in pig herds causing enteritis and dermatitis with gilts routinely vaccinated to prevent the vertically transmitted disorder, SMEDI (stillbirth, mummification, embryonic death and infertility) (Cartwright et al., 1969;Dea et al., 1985;Duhamel et al., 1991;Kresse et al., 1985;Mengeling et al., 2000).
PPV2, the earliest novel PPV, was discovered in 2001 during a hepatitis E virus porcine sera study (Hijikata et al., 2001).
Subsequently in 2006, a South Eastern China study revealed a more divergent PPV2 strain (CN Strain) originating from archival samples from pigs with clinical symptoms of 'high fever disease' . Phylogenetically, PPV3 is more closely related to PPV2  and was initially reported in 2007 from pig samples collected in Hong Kong (Lau et al., 2008). PPV4 which has three ORFs, similar to the genus bocaparvovirus, was initially identified in 2010  in lung lavage samples (collected in 2005) of wasting pigs during a PCVAD outbreak in North Carolina (Cheung et al., 2007). PPV5 was discovered in the USA in 2013 and PPV6 in China in 2014 (Ni et al., 2014), and is both identified as closely related to PPV4 although the genomes lack the additional ORF3 found in PPV4 . More recently, PPV7 was discovered in 2016 by next-generation sequencing (NGS), a divergent Parvoviridae linage, most closely related to turkey parvovirus and Eidolon helvum parvovirus 2 was classified in the proposed genus Chapparvovirus (Palinski et al., 2016).
The current study sought to investigate the potential statistical association of the novel PPVs 2-4 with PCV2, the causative agent of PMWS. This was performed through a retrospective prevalence study of the pathological, molecular and phylogenetic analyses of PPV2-4 across different pig production age groups. Clinically, diseased pigs' samples over the period from 1997 to 2012 (either side of the introduction of the PCV2 vaccine in ~2006) with scored veterinary pathology and tested for the presence of PPV2-4 and PCV2 were analysed by LCA statistical modelling. This analysis revealed a clustering statistical association of PPV2 with PCV2 in diseased cohort of pigs with PCVAD and PRDC.

| Samples
A biobank of archival pig samples (n = 695), originating from Northern Ireland, Republic of Ireland, Great Britain and other neighbouring European countries, sourced from 9 different studies and post-mortem submissions (Table S1), dating from 1997 to 2012, was collated. This cohort represented a single sample per animal, comprising of various sample types including lymph tissue, sera, abortion fluid, abortion tissue, organ tissue, nasal swab and faeces (Table   S1). Novel PPVs 2-4 are known to be endemic in 'healthy pigs', and as it is difficult to assess subclinical disease in pig herds, ultimately this study focused on clinically diseased pigs to demonstrate the prevalence and statistical association within this grouping (Streck et al., 2013). The PCV2 vaccine introduced in ~2006 reduces PCV2 viraemia and viral shedding, and 65% of the samples from this study were categorized in the pre-vaccine category. These samples were sourced from veterinary diagnosed clinically diseased pigs representing production groups originating from France (n = 337, 48.5%), Northern Ireland (n = 182, 26.19%), Great Britain (n = 106, 15.25%), Republic of Ireland (n = 47, 6.7%) and Belgium (n = 23, 3.3%). All recorded sample data were reviewed, and where possible, included the history of animal, age, post-mortem veterinary gross pathology, diagnostic tests including PRRS and PCV2 indirect fluorescence assay, and veterinary histopathology. The collated sample data were categorized by production age groups as foetus (0 weeks), pre-weaners (<4 weeks), weaners (>4 weeks to < 10 weeks) and fatteners (>10 weeks).
Samples were processed before nucleic acid extraction. Sera were removed from whole clotted blood and spun at 200 ×g for 10 mins to pellet remaining blood cells. 10% w/v faecal samples were prepared in transport media with three freeze-thawed cycles and spun at 200 ×g for 30 mins 4°C to clarify the supernatant. 0.2 g of tissue samples was trimmed of fatty tissue and homogenized be-

| Detection PCR and sequencing
The detection of PCV2 and the PPVs (PPV1, PPV2, PPV3 and PPV4) (Eurofins, MWG Operon, Germany) (  Table S4), extension at 72°C 30 s with a melt curve before cooling to 40°C. The sample PCR melt-curve analysis was completed with crossing threshold (Ct) >40 considered negative and Ct levels classified into low, medium and high levels ( Table S1). The semiquantitative qPCR data from this study were restrictive as viral loads were not standardized by quantifying in copies per gram and PCR assays were of different specificity and sensitivity. PCR Ct values were accordingly scored as positive or negative for statistical purposes.
Sequencing primer sets were applied (Table S4) to selected samples with a low Ct to amplify the capsid gene. Specifically, the partial variable region of PPV2 VP1 capsid gene and the full VP capsid gene for PPV3 and PPV4 was amplified for phylogenetic analysis. Excised positive amplicons were purified through the Wizard® SV gel and PCR clean-up system (Promega, UK) and sequenced at Eurofins

| Phylogenetic analysis
Multiple sequence alignments (MSA) of novel PPVs were generated using sequences available from Nucleotide-NCBI (National Centre for Biotechnology Information) and sequencing data analysed on Geneious 10v (Biomatters Ltd, Auckland, New Zealand). Sequenced amplicons of the novel PPVs whole and partial capsid were initially aligned on Geneious and then transferred and completed on Mega X (Mega software version 10.0.4) to avail of the alignment tool, multiple sequence comparison by log-expectation (MUSCLE) and phylogenetic tree builder through maximum-likelihood (ML) phylogenetic trees obtaining 1,000 bootstrap replicates. Phylogenetic trees were graphically viewed and edited in Interactive Tree of Life (Letunic & Bork, 2016).

| Statistical analysis
Initial statistical analysis from the whole cohort of samples evaluated possible association between the detection of virus and disease was completed using permutation tests to calculate the significance probability for the classical chi-square test of the independence of rows and columns in a two-dimensional contingency table using GenStat 16.2 (2013) (VSN International Ltd) (Roff & Bentzen, 1989).
All collated results from clinical observations such as wasting, gross pathology, histopathology, diagnostic immunofluorescent assay (IFA) and PCR findings were presented in a scored matrix with positives "1", negatives "0" and no information as a "blank". Chi-squared uni-variate analysis was set with the probability value "P" at the statistical significance level of .05.

| Latent class analysis
LCA was completed using the poLCA package in R (Linzer & Lewis, 2011), which allows for both traditional LCA and latent class regression, where the probability of being in one or another class is influenced by other covariates besides those used to define the classes (Dayton & Macready, 1988). Only UK and Ireland samples for which symptoms were recorded (n = 272) were included in these analyses (Supplementary Data S2). Models were run from 200 or 400 random starting points to ensure, as far as possible, that the final fit is a global maximum. A key question for LCA is how many latent classes are present. This was decided by a combination of advice from veterinary experts, who reviewed the sets of symptoms included in analysis, and for the resulting classes, the Bayesian information criterion (BIC) (Dziak et al., 2012) As with any other analysis, there is a risk of false-negative and falsepositive classifications with LCA, but this is mitigated to the extent possible, by the principled examination of a range of models, and use of expert domain knowledge to guide model construction. All of the models considered are presented in Supplementary Data S2.
The profiling of PCV2 detection from 1997 to 2012 found PCV2 consistently present in all years from 1998; however, a decline in viral presence was recorded when the vaccine introduction (~2006) occurred (Table 1, Figure S5). Interestingly, the PCV2 levels also profiled an increased frequency, per year, post-PCV2 vaccine introduction. The earliest detection of co-infections of the novel PPV's was PPV2 in 1998, whilst PPV3 and PPV4 were first detected in 2000 (Table 1). The novel PPVs were detected in samples prior to their initial discovery dates although a study did also detect PPV2 and PPV3 in the USA in 1998 (Opriessnig et al., 2014). A PPV2 sample, not captured in this data analysis, was sequenced from 1996 and included in the phylogenetic analysis. The PPV4 from this study detected in 2000 is the earliest reported PPV4.
Foetuses and abortion juice tested positive for novel PPVs at a lower prevalence, with only three samples testing positive for PPV2 and one for PPV4. The known vertically transmitted PCV2 virus was present in 24% (21/88) of abortion juice samples and in 53% (16/30) of foetus samples, and although only one placenta sample was tested, it was positive only for PPV2. No PPV2-PPV4 were present in colon, small intestine, thymus and brain samples although these were representative of low sample numbers.

TA B L E 1 Virus prevalence per year
Year n = total samples PCV2 PPV2 PPV3 PPV4

| Viral presence and statistical associations with clinical and pathological observations
A total of 25 different parameters including gross post-mortem observations, pathological/histopathological findings and the results of diagnostic assays (such as the IFA for PRRSv) were used as a scoring matrix of clinically diseased pigs (Table 4). The data set (n = 695 samples) was initially subjected to chi-squared analysis in order to investigate the presence of virus and statistical association with clinical and pathological presentations. A significant association of virus and lung pathologies was observed for PCV2 (p = .012), PPV2 (p = .001) and PPV4 (p = .01). PCV2 was significantly linked to respiratory pathogenesis with respiratory symptoms (p = .012), pneumonia (p = .017) and broncho interstitial pneumonia (BIP) (p = .036).
PPV2 also showed significance to the same respiratory pathogen- associations but PPV2 was statistically associated with lymphoedema (p = .002) with PPV4 also significantly associated with lymphoedema and intracellular inclusions (p = .001). In addition, PPV2 and PPV3 were also significantly associated with peritonitis and all novel PPVs were statistically associated with ulcers detected in the stomach, colon and caecum (p = .001 to p = .004). Kidney disease had the most significant association with PPV4 (p = .001), and PPV2 was less (p = .038), whereas PCV2 and PPV3 had no association.
Porcine dermatitis and nephropathy syndrome (PDNS) were only associated with PPV2 (p = .001) and PPV3 (p = .017) and interestingly none with PCV2. The novel systemic PPVs and specifically PPV2, statistical analysis reveals, significant association to a range of clinical disease and pathologies; however, these associations are unable to infer the PPV's causal role in the associated pathogenesis.

| Latent class analysis
Viral co-infections were examined using logistic regression analysis and graphical 95% confidence interval level (Table 5 (Table 6). Once both viruses were added to the model, there was little remaining evidence for an effect of age (Table 7). The confidence intervals for these effects are wide, reflecting the relatively small number of animals, but the direction and approximate magnitude of the main effects appear to be robust to a range of modelling assumptions.

| Phylogenetic analysis
The partial PPV2 variable region of the VP1 capsid gene sequence

| D ISCUSS I ON
There is a scientific premise concerning pig herd health issues attributed to viral aetiology in cases of multi-factorial PCVAD and PRDC which remains unexplained (Opriessnig et al., 2011;Qin et al., 2018).
The observed concurrent infections of novel PPV and PCV2 resulted in PCV2 and PPV2 being the most predominant (12.9%), whilst other co-infections occurred at a much lower rate (<3.6%) similar to findings from a previous reported USA study (Opriessnig et al., 2014), with the viral distribution across age groups also similar to previous reports (Cságola et al., 2012;Xiao, Gerber, et al., 2013;Xiao et al., 2012). The low-level prevalence of novel PPVs in the pre-weaner and weaner age groups may indicate that the viruses were not vertically transmitted or that piglets acquired protection from maternally derived antibody and is comparable to the epidemiological serological profile study of PPV2 infection with severe respiratory disease, where onset occurred from 28 days post-decline of PPV2 protective maternal antibodies (Cságola et al., 2016). genotypes were not detected (Collins & Zaykalova, 2020). The global prevalence rates of novel PPVs are considerably variable, specifically the prevalence of PPV3 and PPV4 virus, and this variation is compounded by factors not fully understood and may affect the viral epidemiological dynamics, for example the country of origin, sample type, study design, environmental factors and year of sampling. A recent retrospective study (Sun et al., 2015) on PCVAD suspected pigs (mixed samples) in China revealed a similarly higher occurrence of PPV3 (45.11%) and PPV4 (21.56%), with a South African prevalence study of healthy and sick pigs (2015-2016) reporting a higher rate of PPV4 prevalence (43.6%) (Afolabi et al., 2018). Whilst these recent reports are indicative that prevalence rates for PPV3 and PPV4 may have increased, validated standardized assays, phylogenetic analysis, sample age and tissue type would be beneficial in enabling data comparisons to be made from any future studies.
The retrospective yearly virus prevalence, first detected PCV2 in 1998, and whilst profiling a high prevalence, a decline in frequency occurred with the introduction of the PCV2 vaccine (~2006) (Table 1, Figure S5). The impact of vaccine reduced the PCV2 frequency, which continued over several years, followed by a yearly F I G U R E 5 PPV2 phylogenetic analysis. The tree with the highest log-likelihood (−2324.94) is shown and analysis involved 114 nucleotide sequences. There were a total of 330 positions in the final data set and formed two main clades. Figures 5 to 7 phylogenetic analysis. The molecular phylogenetic analysis by maximum-likelihood method for PPV2-PPV4 virus's evolutionary history was inferred by using the maximum-likelihood method based on the Tamura-Nei model (Tamura & Nei, 1993). Initial tree(s) for the heuristic search were obtained automatically by applying neighbour-join and BioNJ algorithms to a matrix of pairwise distances estimated using the maximum composite likelihood (MCL) approach, obtaining 1,000 bootstrap replicates and then selecting the topology with superior log-likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. Evolutionary analyses were conducted in MEGA X (Kumar et al., 2018) in this study detected virus from 1998 although an additional PPV2 sample was sequenced from 1996 and included in the phylogenetic sequencing analysis which may be the earliest reported detection.     Figure 6). The PPV4 capsid remains highly conserved in comparison with PPV2 (Cadar, Csagola, et al., 2013;Cadar, Lorincz, et al., 2013)

ACK N OWLED G EM ENTS
Cheryl Ball 1 and Lorna McCabe 1 .

CO N FLI C T O F I NTE R E S T
There is no conflict of interest.

E TH I C S S TATEM ENT
The authors confirm that the ethical policies of the journal, as noted in the journal guidelines, have been adhered to. The study included sampling of animals for diagnostic purposes and therefore did not require approval from the ethic committee.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in the supplementary material of this article, and the sequenced data are deposited in NCBI GenBank.