viruses Article Genotyping Porcine Circovirus 3 (PCV-3) Nowadays: Does It Make Sense? Giovanni Franzo 1, * , Eric Delwart 2,3 , Robert Fux 4 , Ben Hause 5 , Shuo Su 6 , JiYong Zhou 7 and Joaquim Segalés 8,9,10 1 Department of Animal Medicine, Production and Health (MAPS), Padua University, 35020 Legnaro, Italy 2 Vitalant Research Institute, San Francisco, CA 94118, USA; Eric.Delwart@ucsf.edu 3 Department of Laboratory Medicine, University of California, San Francisco, CA 94118, USA 4 Institute for Infectious Diseases and Zoonoses, Department for Veterinary Sciences, LMU Munich, 80539 Munich, Germany; robert.fux@micro.vetmed.uni-muenchen.de 5 Kansas State Veterinary Diagnostic Laboratory, Kansas State University, Manhattan, KS 66506, USA; hausebm@gmail.com 6 MOE Joint International Research Laboratory of Animal Health and Food Safety, Jiangsu Engineering Laboratory of Animal Immunology, Institute of Immunology and College of Veterinary Medicine, Nanjing Agricultural University, Nanjing 210095, China; ssh5658485@163.com 7 MOA Key Laboratory of Animal Virology and Department of Veterinary Medicine, Zhejiang University, Hanghou 310058, China; jyzhou@zju.edu.cn 8 Departament de Sanitat i Anatomia Animals, Facultat de Veterinària, UAB, 08193 Bellaterra, Spain; joaquim.segales@irta.cat 9 UAB, Centre de Recerca en Sanitat Animal (CReSA, IRTA-UAB), Campus de la Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain 10 OIE Collaborating Centre for the Research and Control of Emerging and Re-emerging Swine Diseases in Europe (IRTA-CReSA), 08193 Bellaterra, Barcelona, Spain * Correspondence: giovanni.franzo@unipd.it  Received: 31 January 2020; Accepted: 25 February 2020; Published: 28 February 2020  Abstract: The discovery of a globally distributed porcine circovirus (Porcine circovirus 3; PCV-3) has led to intense research activity and the production of a large amount of molecular data. Different research groups have proposed several, not always concordant, genotypes for this virus. While such categories could aid an easier interpretation of PCV-3 molecular epidemiology, any classification, to be useful in practical settings, must be univocal and of help in the understanding of underlying biological features and epidemiology. Based on these premises, the possibility of defining PCV-3 genotypes was evaluated on the broadest available dataset of PCV-3 complete genome (n = 357) and open reading frame 2 (ORF2, n = 653) sequences. Genetic distance and phylogenetic clustering were selected as the main objective criteria. Additional factors, including the number of within-cluster sequences, host and geographic clustering, concordance between different genomic regions, and analysis method were also taken in account to generate a classification that could be effectively applied in research and diagnostic settings. A maximum within-genotype genetic distance of 3% at the complete genome and 6% at the ORF2 levels, bootstrap support higher than 90%, and concordance between analysis methods allowed us to clearly define two clades which could be potentially defined as genotypes. Further subdivision was not suggested due to the absence of a meaningful association between PCV-3 and its biological/epidemiological features. Nevertheless, since one of the clades included two strains only, thus far we formally propose the definition of only one PCV-3 genotype (PCV-3a). The established criteria will allow us to automatically recognize other genotypes when more strain sequences are characterized. Keywords: PCV-3; genotypes; classification; ORF2; genome Viruses 2020, 12, 265; doi:10.3390/v12030265 www.mdpi.com/journal/viruses Viruses 2020, 12, 265 2 of 13 1. Introduction Porcine circovirus 3 (PCV-3) is a recently identified member of the genus Circovirus. It is featured by a single-stranded circular DNA genome of 1999–2001 nt. Similar to other circoviruses, two main open reading frames (ORFs) have been identified. ORF1 is considered the most conserved region of the circovirus genomes. It is located on the positive strand and codes for a single replicase protein (Rep) of 296–297 aa for PCV-3 [1–3]. ORF2 is located on the negative strand and encodes the Cap protein (about 214 aa), the only constituent of the viral capsid, which is considered the most variable and most immunogenic viral protein [1]. Nevertheless, it must be stressed that current knowledge on PCV-3 immunology is very limited and essentially based on comparison with porcine circovirus 2 (PCV-2) and/or on in silico evolutionary studies [4]. Another putative ORF (ORF3) would code for a putative 231-aa protein, whose function has not been elucidated yet; it has been predicted to be oriented in the opposite direction of ORF1 [2]. Since its first identification in the United States thorough a metagenomics approach [2,3], a remarkable amount of epidemiological and molecular data has been accumulated regarding PCV-3. This virus has been reported in all continents (with Africa and Australia being the only exceptions so far) at moderate to high rates of detection (>10%), depending on the particular study, country, and farm (reviewed in Klaumann et al., 2018 [1]). At the same time, PCV-3 DNA has been found in samples from pigs with several clinical syndromes, including respiratory, reproductive, gastrointestinal, and neurological disorders [2,3,5,6]. Nevertheless, the virus has also been detected in healthy animals [7]. Therefore, at the present state, although there are strong suspicions of associations with reproductive disease and multisystemic inflammation [8], the causative role of PCV-3 in the observed clinical conditions has not been confidently defined [9]. In addition, the circulation of strains with different virulence has not been reported. Besides domestic pigs, PCV-3 has been detected in other domestic species, including dogs [10,11], cattle [12], and mice [13], and wild species, including wild boars [14–16], chamois, roe deer, and associated ticks [15]. However, only wild boar seems to be infected at high prevalence and might play a significant role in PCV-3 epidemiology. The identification of a potential association between biological and epidemiological features (e.g., virulence, geographic distribution, and host tropism) with genetic features is of great interest. The willingness to provide a simplified depiction and interpretation of PCV-3 genetic features has led to the definition of different genetic groups, typically named PCV-3a, PCV-3b, and PCV-3c, or Groups A and B [17–19]. Similar to what initially occurred for PCV-2 [20], no consensus is present and different research groups have proposed independent classification criteria and schemes, leading to a certain confusion among veterinarian and researchers. In order to be practical and useful, the PCV-3 classification must not be for its own sake, but should be of help in the interpretation of the underlying viral features and, thus, in the understanding of its epidemiology and potential control measures [21]. Based on these premises, the applicability of a PCV-3 genotyping proposal is evaluated in the present work, offering standardized criteria based on a consensus shared by different research groups operating around the world. 2. Material and Methods 2.1. Database Preparation All available PCV-3 complete genome sequences were downloaded from Genbank (Accessed 19/10/2019). Similarly, all ORF2 sequences were downloaded and only complete ones were maintained for further analysis. When available, collection year, country, and host were extracted from the record and associated to the sequence. Complete genome sequences were aligned using MAFFT [22], while Cap-encoding ones were translated at amino acid level, aligned using MAFFT, and then back-translated to nucleotides using Viruses 2020, 12, 265 3 of 13 TranslatorX [23]. All poorly aligned sequences and those reporting ambiguous nucleotides, frame-shift mutations, and premature stop codons were removed from the alignment. Potential recombinant strains were identified using RDP4 [24] and removed from the alignment. The RDP4 settings for each method were adjusted to account for the dataset features according to the RDP manual recommendations. Only recombination events detected by more than one method with a significance value lower than 0.01 (p-value < 0.01) and Bonferroni correction were accepted. The presence of undetected recombination breakpoints was evaluated using the SBP algorithm implemented in HYPHY [25]. To facilitate the analysis, reduce the computational burden, and increase tree interpretability, only unique sequences were maintained in the alignment. Therefore, only one sequence was selected as representative of all identical ones. 2.2. Phylogenetic and Cluster Analyses The relationship between PCV-3 strains was evaluated by phylogenetic analyses. Three different tree reconstruction methods were used: neighbour joining (NJ) using MEGA X [26], maximum likelihood (ML) using RaxML [27], and Bayesian inference (BI), using MrBayes [28]. The best substitution method was selected for both complete genomes and ORF2 sequences based on the Bayesian Information Criteria (BIC) calculated using Jmodeltest [29]. The robustness of clusters identified by NJ and ML was assessed performing 10,000 bootstrap replicates. For the Bayesian analysis, two independent runs, including one cold and three heated MCMCMC chains, were run for 10,000,000 generations, sampling model parameters and trees every 5000 generations. The runs convergence and mixing were visually inspected and results accepted only if estimated sample size (ESS) was higher than 200 and the potential scale reduction factor (PSRF) approached 1. A consensus tree was obtained after discarding the first 25% of estimated trees as burn-in. The posterior probability of each clade, estimated calculating the proportion of the time that each bipartition was found in the posterior trees, was used as a measure of clade robustness. The presence of reliable clusters that could be considered as genotypes was evaluated according to two criteria: raw genetic distance and bootstrap/posterior probability support. Clusters were identified and marked using ClusterPicker [30] by evaluating different combinations of raw genetic distances (i.e., maximum within-cluster genetic distance) and bootstrap support. 2.3. Association between PCV-3 Phylogeny and Epidemiological Features To assess the association between qualitative traits (i.e., geographical location and host) accounting for phylogenetic uncertainness, different statistics (parsimony score (PS), association index (AI) and monophyletic clade size (MC) statistics) were calculated using BaTS [31]. A Bayesian analysis was performed on the whole ORF2 dataset. Since a limited number of complete ORF2 sequences were available for hosts other than domestic swine, a partial ORF2 dataset was also created and analysed. The considered region was arbitrarily selected in order to achieve the best compromise between the sequence length and the number of included sequences. The posterior trees were used for BaTS analysis after discarding the first 20% of the trees as burn-in. The abovementioned statistics were calculated across all these trees and the observed median (µ obs) value was selected as the final outcome. The distribution under the null hypothesis of no trait–phylogenesis association was obtained by randomizing the tip–trait association 1000 times without replacement for each tree of the posterior distribution. For each randomized dataset the statistics medians (µ null) were calculated and formed the null distribution. This distribution was used to achieve a p-value by simply evaluating the proportion of simulated values more extreme than the observed one. The significance level was set to p < 0.01. Viruses 2020, 12, 265 4 of 13 3. Results After database refinement, 357 complete genomes and 653 complete ORF2 sequences were included in the study. The geographic and host distribution of the included sequences is reported in Table 1. Table 1. Count of sequences included in the present study classified based on the collection continent/region, country, and host. ORF2: open reading frame 2. Host Continent/Region Country Count Bovine Dog Mouse Unknown Swine Wild Boar China 4 5 4 2 243 258 Japan 3 3 Asia South Korea 21 21 Taiwan 2 2 Thailand 2 2 Central America Mexico 1 1 Denmark 2 2 Complete Germany 14 14 genome Hungary 1 1 Europe Italy 5 5 Spain 13 13 Sweden 1 1 Unknown Unknown 11 1 12 North America USA 1 9 10 Russia Russia 2 2 Brazil 7 1 8 South America Colombia 2 2 Total 4 5 4 14 329 1 357 China 16 18 4 6 482 526 Japan 3 3 Asia South Korea 26 26 Taiwan 2 2 Thailand 3 3 Central America Mexico 2 2 Denmark 2 2 France 2 2 Germany 24 24 ORF2 Europe Hungary 1 1 Italy 5 5 Spain 14 14 Sweden 1 1 United Kingdom 2 2 Unknown Unknown 12 6 18 North America USA 1 9 10 Russia Russia 2 2 Brazil 7 1 8 South America Colombia 2 2 Total 16 18 4 19 595 1 653 The results of ML, NJ, and BI of complete and ORF2 sequences provided overall concordant topologies, featured by a homogeneous genetic group and two distantly related sequences, corresponding to strains sampled in China in 2006 (Figure 1). Several potential genetic distance thresholds and bootstrap values were evaluated in order to define a reasonable number of robust genotypes. Using a combination of 90% bootstrap support (or posterior probability) and a 3% and 6% within-clade maximum genetic distance for the complete genome and ORF2, respectively, two clusters were unequivocally identified, which included the same sequences independently of the selected method and dataset (Figure 1). The first one (Clade 1) comprises the vast majority of sequences, and it is featured by highly homogeneous sequences (complete genome average p-distance = 0.009, range = 0.000–0,026; ORF2 average p-distance = 0.014, range = 0.000–0,024). Viruses 2020, 12, 265 5 of 13 The second identified cluster (Clade 2) includes only the two above mentioned Chinese sequences (complete genome p-distance = 0.023, ORF2 p-distance = 0.057). The two clades are separated by a Viruses 2020, minimum 12, 265 p-distance of 0.078 and 0.121 at the complete genome and ORF2 level, respectively. 5 of 15 Figure 1. Phylogenetic trees obtained on complete genome (upper figure) and ORF 2 (lower figure) Figure using the 1. Phylogenetic maximum trees obtained likelihood on complete (ML), neighbour genome joining (NJ), (upper figure) inference and Bayesian and ORF (BI) 2 (lower figure) approach. using the maximum likelihood (ML), Different clades have been colour coded. neighbour joining (NJ), and Bayesian inference (BI) approach. Different clades have been colour coded. In the complete genome dataset only, another intermediate cluster (Clade 3) was identified Several using the abovepotential mentioned genetic distance criteria. thresholds However, and bootstrap these Clade 3 sequences values were evaluated (GenBank accession in order to numbers define a reasonable KY924473, KY924474 and number of robust KY924475) genotypes. were Using1 aincombination part of Clade the ORF2 based of 90% bootstrap analysis. support Finally, in the(or posterior ORF2 dataset,probability) sequences and a 3% and MF318450 and 6% within-clade MG253680 maximum were not classifiedgenetic within distance forusing any cluster the complete the NJ genome and and ORF2, BI methods, respectively, two clusters were unequivocally identified, which included the respectively. same sequences While independently additional sub-clustersof the selected could method be identified andClade within dataset1, (Figure this was1). The first featured byone (Clade 1) a typically comprises low bootstrapthe vast (i.e., support majority lowerofthan sequences, and it is featured 70%) (Supplementary FigurebyS1)highly and werehomogeneous not consistentsequences among (complete different genome dataset and average analysis p-distance methods. = 0.009, range = 0.000–0,026; ORF2 average p-distance = 0.014, range = 0.000–0,024). A significant The second association was identified cluster (Clade detected between 2) includes tree topology andonly the twoorabove continent mentioned host (Table 2). Chinesewhen However, sequences (complete a reduced datasetgenome p-distance was prepared = 0.023, ORF2 p-distance by down-sampling the number=of0.057). The two sequences clades collected inare Asiaseparated to obtainby a minimum a more balanced p-distance dataset, noof 0.078 and 0.121 association at thecontinent between completeandgenome and ORF2 tree topology level, could berespectively. demonstrated for most geographical regions (Table 2). Even when such association could be In the detected, completeregions it involved genome(i.e., dataset only, Russia another and Southintermediate America) for cluster which(Clade a very3)limited was identified numberusing of the above mentioned criteria. However, these Clade 3 sequences (GenBank sequences was available (Table 1). Similarly, when the number of sequences collected from Sus scrofa accession numbers KY924473, was KY924474 down-sampled, and KY924475) a lower were part ofalthough statistical association, Clade 1 in the ORF2 present, couldbased analysis.between be detected Finally, host in the ORF2 and tree dataset, topologysequences (Table 2). MF318450 and MG253680 were not classified within any cluster using the NJ and BI methods, respectively. While additional sub-clusters could be identified within Clade 1, this was featured by a typically low bootstrap support (i.e., lower than 70%) (Supplementary Figure 1) and were not consistent among different dataset and analysis methods. A significant association was detected between tree topology and continent or host (Table 2). However, when a reduced dataset was prepared by down-sampling the number of sequences collected in Asia to obtain a more balanced dataset, no association between continent and tree topology could be demonstrated for most geographical regions (Table 2). Even when such association could be detected, it involved regions (i.e., Russia and South America) for which a very limited number of sequences was available (Table 1). Similarly, when the number of sequences collected from Viruses 2020, 12, 265 6 of 13 Table 2. Results of the phylogeny–trait (i.e., region and host) association performed on both original and balanced datasets. The global statistics (AI and PS) and the feature-specific ones (MC) are reported. AI: association index; PS: parsimony score; MC: monophyletic clade size. \ Statistic Feature Observed Mean Lower Observed 95% CI Upper Observed 95% CI Null Mean Lower Null 95% CI Upper Null 95% CI p-Value AI 10.624 9.048 12.338 20.408 19.203 21.617 0.000 Global PS 90.378 86.000 94.000 116.707 115.063 118.105 0.000 Asia 3.001 3.000 3.000 1.375 1.090 2.002 0.001 Central America 2.000 2.000 2.000 1.025 1.000 1.102 0.001 All sequences Europe 34.865 25.000 50.000 14.750 11.833 20.940 0.001 MC North America 3.631 3.000 5.000 1.660 1.249 2.111 0.001 Russia 2.769 2.000 3.000 1.150 1.003 1.958 0.001 Geographic location South America 1.996 2.000 2.000 1.000 1.000 1.000 0.001 Unknown 1.973 2.000 2.000 1.000 1.000 1.000 0.001 AI 7.251 5.826 8.717 13.574 12.407 14.658 0.000 Global PS 62.552 59.000 66.000 89.229 85.724 92.304 0.000 Asia 1.348 1.000 2.000 1.313 1.017 2.006 1.000 Central America 2.000 2.000 2.000 1.118 1.000 1.812 0.026 Balanced dataset Europe 6.630 5.000 11.000 4.367 3.541 5.536 0.018 MC North America 3.602 3.000 5.000 2.424 2.011 3.149 0.107 Russia 3.022 2.000 4.000 1.388 1.039 2.017 0.001 South America 2.000 2.000 2.000 1.004 1.000 1.000 0.002 Unknown 1.873 1.000 2.000 1.003 1.000 1.000 0.001 AI 5.290 4.037 6.617 9.511 8.757 10.269 0.000 Global PS 41.388 39.000 44.000 53.611 52.749 53.914 0.000 Bovine 108.773 98.000 146.000 30.972 25.001 40.500 0.002 All sequences Canine 3.612 2.000 4.000 1.107 1.013 1.339 0.001 MC Mouse 6.969 6.000 7.000 1.084 1.007 1.271 0.001 Domestic pig 1.827 1.000 2.000 1.080 1.005 1.284 0.006 Host Wild boar 2.000 2.000 2.000 1.004 1.000 1.014 0.001 AI 3.082 2.345 3.900 7.543 6.736 8.316 0.000 Global PS 31.276 28.000 34.000 50.186 47.890 52.021 0.000 Bovine 4.007 3.000 5.000 1.589 1.166 2.145 0.002 Balanced dataset Canine 17.428 17.000 19.000 4.366 3.386 6.276 0.001 MC Mouse 6.662 5.000 9.000 1.472 1.110 2.062 0.001 Domestic pig 2.310 2.000 4.000 1.481 1.110 2.083 0.160 Wild boar 2.000 2.000 2.000 1.029 1.000 1.110 0.007 Viruses 2020, 12, 265 7 of 13 Viruses 2020, 12, 265 9 of 15 The qualitative evaluation of major clades demonstrated the absence of any geographical pattern. The qualitative Although Asia was byevaluation of represented far the most major clades demonstrated location, the absence strains collected of regions in other any geographical were well pattern. Although Asia was by far the most represented location, strains collected interspersed in the phylogenetic tree (Figure 2a). A similar scenario was demonstrated when in other regions the were well interspersed in the phylogenetic tree (Figure 2a). A similar scenario was demonstrated collection host was considered, with sequences obtained from different hosts mixed in different clades when the (Figure 2b).collection host was considered, with sequences obtained from different hosts mixed in different clades (Figure 2b). Figure 2. Phylogenetic tree reconstructed using the BI on the complete (left figure) and partial (right figure) ORF2 dataset. The collection continent and host have been colour-coded. Figure 2. Phylogenetic tree reconstructed using the BI on the complete (left figure) and partial (right The presence figure) of potential ORF2 dataset. amino acid The collection markers continent andwas hostevaluated have beenon the ORF2 (Cap) dataset. Although colour-coded. a certain correspondence was observed between genotype and amino acid sequence, several exceptions wereThe also presence present and of no reliableamino potential amino acid markers markersdifferentiating was evaluated PCV-3 in separate on the groups ORF2 (Cap) could dataset. be identified (Figure 3). Although a certain correspondence was observed between genotype and amino acid sequence, several exceptions were also present and no reliable amino acid markers differentiating PCV-3 in separate groups could be identified (Figure 3). Viruses 2020, 12, x; doi: FOR PEER REVIEW www.mdpi.com/journal/viruses Viruses 2020, 12, 265 Viruses 2020, 12, x FOR PEER REVIEW 8 of 13 10 of 15 Figure 3. Plot reporting the alignment of the amino acids of the cap gene with respect to their position Figure 3. Plot in the phylogenetic reporting tree. Differentthe alignment amino of thebeen acids have amino acids of theFor colour-coded. capease geneofwith respect to their interpretation, only position in the phylogenetic tree. Different the results of ML-based analysis are reported. amino acids have been colour-coded. For ease of interpretation, only the results of ML-based analysis are reported. 4. Discussion The4.identification Discussion of a new porcine circovirus, likely because of its similarities with the more well-known The and identification economically important PCV-2, circovirus, of a new porcine has stimulated likelya remarkable because of its number of studies similarities withon the more PCV-3, including the genetic characterization of several strains. To date, well-known and economically important PCV-2, has stimulated a remarkable number of studies on different research groups have introduced differentthe PCV-3, including nomenclatures in order to provide genetic characterization of several anstrains. easier-to-understand To date, different depiction research of groups PCV-3 epidemiology [17–19,32]. Such classifications could represent a double-edged have introduced different nomenclatures in order to provide an easier-to-understand depiction of sword when no shared criteria are proposed and PCV-3 epidemiology acceptedSuch [17–19,32]. at the internationalcould classifications level.represent While thea definition double-edged of categories sword when no can surely simplify and enhance the interpretation of PCV-3 epidemiology, the presence shared criteria are proposed and accepted at the international level. While the definition of categories of different nomenclatures can surely cansimplify lead to confusing and enhance andthemisleading resultofinterpretations, interpretation PCV-3 epidemiology, as initially occurredof the presence fordifferent PCV-2 [33,34] and several nomenclatures canother lead viral infections to confusing and[35]. Additionally, misleading resulttheinterpretations, inclusion of strains in separate as initially occurred for groups could erroneously lead to the conclusion that differences do exist even in biological PCV-2 [33,34] and several other viral infections [35]. Additionally, the inclusion of strains in separate features when those groups have not been could demonstrated erroneously lead to thus far. It mustthat the conclusion be differences taken into account do existthat evenit in is biological difficult tofeatures assess different when those biological features have not beensuch as virulence, demonstrated thushost far.adaptation, It must beor tissue taken tropism into accountforthat a virus it isthat difficult to has not assess yet been isolated in cell culture. In consequence, reliable pathogenesis studies different biological features such as virulence, host adaptation, or tissue tropism for a virus are lacking, althoughthat onehasexperimental not yet been infection isolatedusing a PCV-3 in cell culture. infectious clone hasreliable In consequence, been published [36]. studies pathogenesis In addition, are lacking, field data do not indicate that different PCV-3 strains cause different disease although one experimental infection using a PCV-3 infectious clone has been published [36]. In outcomes [1]. With this in mind, addition, the broadest field data available do not indicate dataset that of complete different PCV-3 cause PCV-3 strains ORF2 different and genome sequences disease outcomes [1]. was analyzedWith to evaluate this inthemind, feasibility and practical the broadest utility of available establishing dataset a sharedPCV-3 of complete PCV-3 classification. ORF2 and genome Although sequences the complete was analyzed genome providedthe to evaluate more informative feasibility sites, the utility and practical whole of genome sequencing establishing is PCV- a shared typically3 challenging, classification. especially in routine diagnostic settings. Additionally, distance comparison over the entire virus genomethe Although of complete viruses that are subject genome providedto recombination, more informative a phenomenon widely sites, the whole occurring genome sequencing in PCV-2 [37] and other circoviruses [38], is likely misleading [21]. ORF2 is typically challenging, especially in routine diagnostic settings. Additionally, distance is the most variable PCV-3 comparison genomeover regiontheand likely entire virusthegenome one most affected that of viruses by host-derived are subject toimmune pressureaselection recombination, phenomenon [1,4]. widely occurring in PCV-2 [37] and other circoviruses [38], is likely misleading [21]. ORF2 is the most Viruses 2020, 12, 265 9 of 13 Its sequencing and analysis could therefore represent a reasonable cost/benefit compromise and was thus included in the analysis for more accurate evaluation. Currently, no strict guidelines are present to define taxonomic levels under the species level by the International Committee for the Taxonomy of Viruses (www.ictvonline.com). Nevertheless, among the potential criteria, the present work attempted to adhere to the ones typically followed by official viral taxonomy, i.e., genetic distance and phylogenetic analyses [39,40]. Based on the selected criteria (bootstrap support > 0.9 and maximum genetic distance of 3% and 6% at the complete genome and ORF2 levels), two clades (Clade 1 and Clade 2) could be consistently identified independently from the considered region and analysis method. Particularly, Cluster 1 included strains which were previously divided in two or more genotypes (named PCV-3a, PCV-3b, and PCV-3c, or Groups A and B in previous papers) [17–19,32]. Even if the tree topology could support the definition of additional clades, the presence of several intermediate minor clades and the poor bootstrap support suggest caution in the recognition of these potentially different genotypes. Even though a certain geographic and host clustering could be statistically identified, this seems largely explained by the severely biased sequence availability, featured by a clear predominance of sequences collected from Asia (especially China) and Sus scrofa. Since most of the sequences collected from other hosts or continents (e.g., South and Central America, Russia, etc.) were obtained from single/few studies, an effect of the specific experiment rather than an actual difference in distribution is likely. This hypothesis is supported by the relevant reduction in statistical significance of the observed trait–genotype association when a more balanced dataset was artificially created. Future sequencing of strains collected from other hosts and non-Asian countries would be of help to further elucidate these issues. Nevertheless, when the most relevant genetic groups were evaluated, no clear association could be observed between specific clades and host or geographic distribution. These results confirm previous studies, which highlighted a broad PCV-3 distribution and a massive circulation involving countries all around the world [1,4]. Similarly, no evidence of a differential virulence has been reported to date. Setting a rational number of man-made categories onto a substantial continuum of viral evolution (i.e., genetic variability) requires pragmatism in order to ensure its utility and applicability by the virology community [21]. Therefore, the definition of genotypes should account not only for genetic features, but also for virus features that are designed to assist studies of its biology and/or epidemiology. Based on these considerations, no evolutionary, epidemiological, or biologic factors seem to justify the further division of Clade 1 in additional genotypes. This conclusion was also supported by the analysis of the amino acid profile. While initial studies suggested the presence of reliable amino acid markers [18], the updated dataset revealed that although a certain association between genotype and phenotype can be identified, several exceptions are present, likely due to random mutations or convergent evolution. Clade 2 included only two distantly related sequences collected in 2006 from two different Chinese farms. The farms experienced high mortality; however, they were also affected by a highly pathogenic porcine reproductive and respiratory syndrome. No further sequences of PCV-3 were found in the same farms (Dr. Jue Liu, Beijing Academy of Agriculture and Forestry Sciences, China, personal communication). In consequence, understanding the actual epidemiological role of these strains is challenging. Similar to some “minor” PCV-2 genotypes, these strains could represent either recently emerged variants or the last descendant of previously circulating genotypes, probably like the case of PCV-2c [41–43]. The limited knowledge on PCV-3 molecular epidemiology could also justify the low Clade 2 frequency. However, the ancient origin of PCV-3 [4] and the extensive sequencing activity, especially in China, lessen the likelihood of this latter hypothesis. While Clade 1 and Clade 2 were recognized on both complete genome and ORF2 sequence analyses, an intermediate cluster could be detected at complete genome level only, as recently described by Liu et al., 2019 [44]. The reasons behind these conflicting results could be several, including differential forces acting on the Rep and Cap gene or the presence of undetected recombination events. Despite efforts were made to identify and exclude potential recombinant strains, accurate recombinant identification is challenging, especially dealing Viruses 2020, 12, 265 10 of 13 with highly homogeneous viral species like PCV-3. The actual weight of recombination in PCV-3 evolution and if its occurrence could also explain the typically low support within Clade 1 remains to be established. In conclusion, obtained results suggest that the current and sometimes conflicting PCV-3 classification schemes are poorly supported by sequence and biological data. With the aim of providing robust criteria that could fit an ever-increasing number of PCV-3 sequences in the future, allowing easy and automatic updates, the authors of the present work suggest the following threshold for the definition of PCV-3 genotypes: bootstrap support (or posterior probability) > 0.9, maximum genetic distance of 3% and 6% at the complete genome and ORF2 levels, and concordant results between ORF2 and the complete genome. The last criterion is dictated by practical reasons, since it would allow the genotyping of the PCV-3 strains using the ORF2 sequence only, allowing an easy classification even during routine diagnostic activity. Similarly, the congruence between NJ and more complex phylogenetic approaches could allow for fast but reliable results in everyday routine. According to this pragmatic criteria, a reference dataset has been provided (Supplementary data 1). Finally, it is suggested to formally accept a genotype only if at least five sequences are available, corresponding to approximately 1% of the analyzed PCV-3 sequences. This approach should allow a focus on more widespread genotypes and avoid the risk of defining poor quality sequences or extremely low-fitness strains as separate genotypes. Based on these settings, we formally define only Clade 1 as PCV-3a, while Clade 2 (or other clusters that could be detected in the following years), could be recognized as a formal genotype if a larger number of strains are characterized in the future. While the relatively high number of sequences collected from countries worldwide suggests that the present data are representative of PCV-3 molecular epidemiology, other fit variants could emerge and spread, requiring a prompt classification. The nomenclature has been selected to mirror that of the well-established PCV-2 genotype classification [43]. We hope that the unified classification scheme proposed in the present study will establish a “common language” among different research groups and diagnostic laboratories. At the same time, a larger effort must be established to provide more representative and structured sampling activity and to increase the sharing of properly annotated sequences in freely available databases. Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4915/12/3/265/s1. Author Contributions: Conceptualization, G.F. and J.S.; methodology, G.F.; formal analysis, G.F.; investigation, G.F., E.D., B.H., S.S., and J.Z.; resources, G.F., E.D., B.H., S.S., and J.Z. and J.S.; data curation, G.F.; writing—original draft preparation, G.F. and J.S. ; writing—review and editing, E.D., R.F., B.H., S.S., J.Z.; supervision, J.S.; project administration, G.F. and J.S.; funding acquisition, G.F. and J.S. All authors have read and agreed to the published version of the manuscript. Funding: This research was founded by grant BIRD187958/18 from the Department of Animal Medicine, Production and Health, University of Padua. J. Segalés acknowledges the E-RTA2017-00007-00-00 INIA Project from the Instituto Nacional de Investigación y Tecnologia Agraria y Alimentaria (Spanish Government) and the CERCA Programme/Generalitat de Catalunya (Spain). Conflicts of Interest: The authors declare that they have no competing interests. Data Availability Statement: All used sequences are available in Genbank and the accession number, including a suggested reference dataset, are reported in supplementary data 1. References 1. Klaumann, F.; Correa-Fiz, F.; Franzo, G.; Sibila, M.; Núñez, J.I.; Segalés, J. Current Knowledge on Porcine circovirus 3 (PCV-3): A Novel Virus with a Yet Unknown Impact on the Swine Industry. Front. Vet. Sci. 2018, 5, 315. [CrossRef] [PubMed] 2. Palinski, R.; Piñeyro, P.; Shang, P.; Yuan, F.; Guo, R.; Fang, Y.; Byers, E.; Hause, B. A Novel Porcine Circovirus Distantly Related to Known Circoviruses Is Associated with Porcine Dermatitis and Nephropathy Syndrome and Reproductive Failure. J. Virol. 2016, 91. [CrossRef] [PubMed] Viruses 2020, 12, 265 11 of 13 3. Phan, T.G.; Giannitti, F.; Rossow, S.; Marthaler, D.; Knutson, T.P.; Li, L.; Deng, X.; Resende, T.; Vannucci, F.; Delwart, E. Detection of a novel circovirus PCV3 in pigs with cardiac and multi-systemic inflammation. Virol. J. 2016, 13, 184. [CrossRef] [PubMed] 4. Franzo, G.; He, W.; Correa-Fiz, F.; Li, G.; Legnardi, M.; Su, S.; Segalés, J. A Shift in Porcine Circovirus 3 (PCV-3) History Paradigm: Phylodynamic Analyses Reveal an Ancient Origin and Prolonged Undetected Circulation in the Worldwide Swine Population. Adv. Sci. 2019, 6, 1901004. [CrossRef] 5. Ku, X.; Chen, F.; Li, P.; Wang, Y.; Yu, X.; Fan, S.; Qian, P.; Wu, M.; He, Q. Identification and genetic characterization of porcine circovirus type 3 in China. Transbound. Emerg. Dis. 2017, 64, 703–708. [CrossRef] 6. Shen, H.; Liu, X.; Zhang, P.; Wang, L.; Liu, Y.; Zhang, L.; Liang, P.; Song, C. Genome characterization of a porcine circovirus type 3 in South China. Transbound. Emerg. Dis. 2017, 65, 264–266. [CrossRef] 7. Zheng, S.; Wu, X.; Zhang, L.; Xin, C.; Liu, Y.; Shi, J.; Peng, Z.; Fu, F.; Yu, J.; Sun, W.; et al. The occurrence of porcine circovirus 3 without clinical infection signs in Shandong Province. Transbound. Emerg. Dis. 2017, 64, 1337–1341. [CrossRef] 8. Arruda, B.; Piñeyro, P.; Derscheid, R.; Hause, B.; Byers, E.; Dion, K.; Long, D.; Sievers, C.; Tangen, J.; Williams, T.; et al. PCV3-associated disease in the United States swine herd. Emerg. Microbes Infect. 2019, 8, 684–698. [CrossRef] 9. Franzo, G.; Legnardi, M.; Tucciarone, C.M.; Drigo, M.; Klaumann, F.; Sohrmann, M.; Segalés, J. Porcine circovirus type 3: A threat to the pig industry? Vet. Rec. 2018, 182, 83. [CrossRef] 10. Sun, W.; Wang, W.; Xin, J.; Cao, L.; Zhuang, X.; Zhang, C.; Zhu, Y.; Zhang, H.; Qin, Y.; Du, Q.; et al. An epidemiological investigation of porcine circovirus 3 infection in dogs in the Guangxi Province from 2015 to 2017, China. Virus Res. 2019, 270, 197663. [CrossRef] 11. Zhang, J.; Liu, Z.; Zou, Y.; Zhang, N.; Wang, D.; Tu, D.; Yang, L.; Deng, Z.; Yang, Y.; Jiang, P.; et al. First molecular detection of porcine circovirus type 3 in dogs in China. Virus Genes 2017, 54, 140–144. [CrossRef] 12. Wang, W.; Sun, W.; Cao, L.; Zheng, M.; Zhu, Y.; Li, W.; Liu, C.; Zhuang, X.; Xing, J.; Lu, H.; et al. An epidemiological investigation of porcine circovirus 3 infection in cattle in Shandong province, China. BMC Vet. Res. 2019, 15, 60. [CrossRef] 13. Jiang, S.; Zhou, N.; Li, Y.; An, J.; Chang, T. Detection and sequencing of porcine circovirus 3 in commercially sourced laboratory mice. Vet. Med. Sci. 2019, 5, 176–181. [CrossRef] 14. Franzo, G.; Tucciarone, C.M.; Drigo, M.; Cecchinato, M.; Martini, M.; Mondin, A.; Menandro, M.L. First report of wild boar susceptibility to Porcine circovirus type 3: High prevalence in the Colli Euganei Regional Park (Italy) in the absence of clinical signs. Transbound. Emerg. Dis. 2018, 65, 957–962. [CrossRef] 15. Franzo, G.; Grassi, L.; Tucciarone, C.M.; Drigo, M.; Martini, M.; Pasotto, D.; Mondin, A.; Menandro, M.L. A wild Circ-ulation: High presence of Porcine circovirus 3 in different mammalian wild hosts and ticks. Transbound. Emerg. Dis. 2019, 66, 1548–1557. 16. Klaumann, F.; Dias-Alves, A.; Cabezon, O.; Mentaberre, G.; Castillo-Contreras, R.; Lopez-Bejar, M.; Casas-Díaz, E.; Sibila, M.; Correa-Fiz, F.; Segalés, J. Porcine circovirus 3 is highly prevalent in serum and tissues and may persistently infect wild boar (Sus scrofa scrofa). Transbound. Emerg. Dis. 2018, 66, 91–101. [CrossRef] 17. Franzo, G.; Legnardi, M.; Hjulsager, C.K.; Klaumann, F.; Larsen, L.E.; Segalés, J.; Drigo, M. Full-genome sequencing of porcine circovirus 3 field strains from Denmark, Italy and Spain demonstrates a high within-Europe genetic heterogeneity. Transbound. Emerg. Dis. 2018, 65, 602–606. [CrossRef] 18. Fux, R.; Söckler, C.; Link, E.K.; Renken, C.; Krejčí, R.; Sutter, G.; Ritzmann, M.; Eddicks, M. Full genome characterization of porcine circovirus type 3 isolates reveals the existence of two distinct groups of virus strains. Virol. J. 2018, 15, 25. [CrossRef] 19. Li, G.; Wang, H.; Wang, S.; Xing, G.; Zhang, C.; Zhang, W.; Liu, J.; Zhang, J.; Su, S.; Zhou, J. Insights into the genetic and host adaptability of emerging porcine circovirus 3. Virulence 2018, 9, 1301–1313. [CrossRef] 20. Segalés, J.; Olvera, A.; Grau-Roma, L.; Charreyre, C.; Nauwynck, H.; Larsen, L.E.; Dupont, K.; McCullough, K.; Ellis, J.; Krakowka, S.; et al. PCV-2 genotype definition and nomenclature. Vet. Rec. 2008, 162, 867–868. [CrossRef] 21. Simmonds, P. Methods for virus classification and the challenge of incorporating metagenomic sequence data. J. Gen. Virol. 2015, 96, 1193–1206. [CrossRef] [PubMed] Viruses 2020, 12, 265 12 of 13 22. Standley, K. MAFFT Multiple Sequence Alignment Software Version 7: Improvements in Performance and Usability. Mol. Boil. Evol. 2013, 30, 772–780. 23. Abascal, F.; Zardoya, R.; Telford, M.J. TranslatorX: Multiple alignment of nucleotide sequences guided by amino acid translations. Nucleic Acids Res. 2010, 38, W7–W13. [CrossRef] [PubMed] 24. Martin, D.P.; Murrell, B.; Golden, M.; Khoosal, A.; Muhire, B. RDP4: Detection and analysis of recombination patterns in virus genomes. Virus Evol. 2015, 1, vev003. [CrossRef] 25. Kosakovsky Pond, S.L.; Frost, S.D.W.; Muse, S.V. HyPhy: Hypothesis Testing Using Phylogenies. Bioinformatics 2005, 21, 676–679. [CrossRef] [PubMed] 26. Kumar, S.; Stecher, G.; Li, M.; Knyaz, C.; Tamura, K. MEGA X: Molecular Evolutionary Genetics Analysis across Computing Platforms. Mol. Boil. Evol. 2018, 35, 1547–1549. [CrossRef] 27. Stamatakis, A. RAxML version 8: A tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 2014, 30, 1312–1313. [CrossRef] 28. Ronquist, F.; Teslenko, M.; Van Der Mark, P.; Ayres, D.L.; Darling, A.E.; Hoehna, S.; Larget, B.; Liu, L.; Suchard, M.A.; Huelsenbeck, J.P. MrBayes 3.2: Efficient Bayesian phylogenetic inference and model choice across a large model space. Syst. Boil. 2012, 61, 539–542. [CrossRef] 29. Darriba, D.; Taboada, G.L.; Doallo, R.; Posada, D. JModelTest 2: More models, new heuristics and parallel computing. Nat. Methods 2012, 9, 772. [CrossRef] 30. Ragonnet-Cronin, M.; Hodcroft, E.; Hue, S.; Fearnhill, E.; Delpech, V.C.; Brown, A.L.; Lycett, S.J. Automated analysis of phylogenetic clusters. BMC Bioinform. 2013, 14, 317. [CrossRef] 31. Parker, J.; Rambaut, A.; Pybus, O.G. Correlating viral phenotypes with phylogeny: Accounting for phylogenetic uncertainty. Infect. Genet. Evol. 2008, 8, 239–246. [CrossRef] 32. Fu, X.; Fang, B.; Ma, J.; Liu, Y.; Bu, D.; Zhou, P.; Wang, H.; Jia, K.; Zhang, G. Insights into the epidemic characteristics and evolutionary history of the novel porcine circovirus type 3 in southern China. Transbound. Emerg. Dis. 2017, 65, e296–e303. [CrossRef] 33. Cortey, M.; Olvera, A.; Grau-Roma, L.; Segalés, J. Further comments on porcine circovirus type 2 (PCV2) genotype definition and nomenclature. Vet. Microbiol. 2011, 149, 522–523. [CrossRef] 34. Franzo, G.; Cortey, M.; Olvera, A.; Novosel, D.; De Castro, A.M.M.G.; Biagini, P.; Segalés, J.; Drigo, M. Revisiting the taxonomical classification of Porcine Circovirus type 2 (PCV2): Still a real challenge. Virol. J. 2015, 12, 131. [CrossRef] 35. Valastro, V.; Holmes, E.C.; Britton, P.; Fusaro, A.; Jackwood, M.W.; Cattoli, G.; Monne, I. S1 gene-based phylogeny of infectious bronchitis virus: An attempt to harmonize virus classification. Infect. Genet. Evol. 2016, 39, 349–364. [CrossRef] 36. Jiang, H.; Wang, D.; Wang, J.; Zhu, S.; She, R.; Ren, X.; Tian, J.; Quan, R.; Hou, L.; Li, Z.; et al. Induction of Porcine Dermatitis and Nephropathy Syndrome in Piglets by Infection with Porcine Circovirus Type 3. J. Virol. 2019, 93, e02045-18. [CrossRef] 37. Franzo, G.; Cortey, M.; Segalés, J.; Hughes, J.; Drigo, M. Phylodynamic analysis of porcine circovirus type 2 reveals global waves of emerging genotypes and the circulation of recombinant forms. Mol. Phylogenet. Evol. 2016, 100, 269–280. [CrossRef] 38. Lefeuvre, P.; Lett, J.-M.; Varsani, A.; Martin, D.P. Widely Conserved Recombination Patterns among Single-Stranded DNA Viruses. J. Virol. 2008, 83, 2697–2707. [CrossRef] 39. Simmonds, P.; Adams, M.J.; Benkő, M.; Breitbart, M.; Brister, J.R.; Carstens, E.B.; Davison, A.J.; Delwart, E.; Gorbalenya, A.E.; Harrach, B.; et al. Virus taxonomy in the age of metagenomics. Nat. Rev. Microbiol. 2017, 15, 161–168. [CrossRef] 40. Walker, P.J.; Siddell, S.G.; Lefkowitz, E.J.; Mushegian, A.R.; Dempsey, D.M.; Dutilh, B.E.; Harrach, B.; Harrison, R.L.; Hendrickson, R.; Junglen, S.; et al. Changes to virus taxonomy and the International Code of Virus Classification and Nomenclature ratified by the International Committee on Taxonomy of Viruses (2019). Arch. Virol. 2019, 164, 2417–2429. [CrossRef] 41. Dupont, K.; Nielsen, E.; Bækbo, P.; Larsen, L.E. Genomic analysis of PCV2 isolates from Danish archives and a current PMWS case-control study supports a shift in genotypes with time. Vet. Microbiol. 2008, 128, 56–64. [CrossRef] Viruses 2020, 12, 265 13 of 13 42. Franzo, G.; Cortey, M.; Castro, A.M.; Piovezan, U.; Szabó, M.P.J.; Drigo, M.; Segalés, J.; Richtzenhain, L.J. Genetic characterisation of Porcine circovirus type 2 (PCV2) strains from feral pigs in the Brazilian Pantanal: An opportunity to reconstruct the history of PCV2 evolution. Vet. Microbiol. 2015, 178, 158–162. [CrossRef] 43. Franzo, G.; Segalés, J. Porcine circovirus 2 (PCV-2) genotype update and proposal of a new genotyping methodology. PLoS ONE 2018, 13, e0208585. [CrossRef] 44. Liu, X.; Zhang, X.; Li, X.; Ouyang, T.; Niu, G.; Ouyang, H.; Ren, L. Genotyping based on complete coding sequences of porcine circovirus type 3 is stable and reliable. Infect. Genet. Evol. 2020, 78, 104116. [CrossRef] © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).