http://orcid.org/0000-0003-2991-217X
Figures
Abstract
Members of the genus Circovirus are host-specific viruses, which are totally dependent on cell machinery for their replication. Consequently, certain mimicry of the host genome features is expected to maximize cellular replicative system exploitation and minimize the recognition by the innate immune system. In the present study, the analysis of several genome composition and codon bias parameters of circoviruses infecting avian and mammalian species demonstrated the presence of quite distinctive patterns between the two groups. Remarkably, a higher deviation from the expected values based only on mutational patterns was observed for mammalian circoviruses both at dinucleotide and codon levels. Accordingly, a stronger selective pressure was estimated to shape the genome of mammalian circoviruses, particularly in the Cap encoding gene, compared to avian circoviruses. These differences could be attributed to different physiological and immunological features of the two host classes and suggest a trade-off between a tendency to optimize the capsid protein translation while minimizing the recognition of the genome and the transcript molecules. Interestingly, the recently identified Porcine circovirus 3 (PCV-3) had an intermediate pattern in terms of genome composition and codon bias. Particularly, its Rep gene appeared closely related to other mammalian circoviruses (especially bat circoviruses) while the Cap gene more closely resembled avian circoviruses. These evidences, coupled with the high selective forces apparently modelling the PCV-3 Cap gene composition, suggest the potential recombinant origin, followed or preceded by a host jump, of this virus.
Citation: Franzo G, Segales J, Tucciarone CM, Cecchinato M, Drigo M (2018) The analysis of genome composition and codon bias reveals distinctive patterns between avian and mammalian circoviruses which suggest a potential recombinant origin for Porcine circovirus 3. PLoS ONE 13(6): e0199950. https://doi.org/10.1371/journal.pone.0199950
Editor: Ulrich Melcher, Oklahoma State University, UNITED STATES
Received: April 27, 2018; Accepted: June 15, 2018; Published: June 29, 2018
Copyright: © 2018 Franzo et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
1. Introduction
Several studies have demonstrated the presence of a relevant genomic signature in dinucleotide frequencies in different organisms [1–3]. For example, TpA is broadly under-represented in eukaryotic chromosomes, potentially because of its low thermodynamic energy, the high degree of degradation of UpA dinucleotides by ribonucleases in mRNA [4], or the presence of TA as part of many regulatory signals and stop codons [4]. Similarly, CpG dinucleotide scarcity in vertebrates is thought to be partially due to cytosine methylation. In fact, methylated cytosines are prone to spontaneous deamination to thymines, leading to the dinucleotide TpG [5] However, DNA conformation, such as secondary structure, and dinucleotide stacking energies can be involved in this bias [6].
Consequently, besides genomic structural constraints and chemical features, other factors affecting the genome stability, like environmental conditions (e.g. pH, temperature, metal concentration, etc.), can be involved in shaping the overall dinucleotide composition [1].
A similar signature pattern has been observed in term of codon usage. Due to the degeneracy of the genetic code, the 20 amino-acids are actually coded by 61 codons. However, the frequencies of synonymous codon usage appear non-random and different species exhibit more or less marked preferences [7]. Two non-conflicting hypotheses have been advocated to explain this scenario: the mutational (or neutral) and the selectionist hypotheses. The first one poses that the codon usage bias is ascribable to the genome composition and mutational patterns non-randomness. While some studies have actually demonstrated that the level of GC content (and more generally the genome composition) can explain part of the codon bias differences between different organisms [8,9], there are some clear evidences that natural selection must also be involved [7]. Supporting the selectionist hypothesis, the codon choice has been linked to translational levels, efficiency and accuracy. A role as an additional level of regulation, tuning the levels of protein abundance and their appropriate folding, has also been proposed [10–12]. Remarkably, a direct effect on organism fitness has been experimentally proven [13]. The concomitant action of these two forces, the so called ‘‘mutation-selection-drift balance model of codon bias” is currently the most commonly accepted theory explaining the codon bias, suggesting the selection favouring the most preferred codons and the mutational drift allowing the maintenance of the minor ones [7,14].
Based on these premises, a similar dinucleotide pattern between obliged intracellular parasites, like viruses, and their respective host can be expected. Selective forces should act favouring those individuals mimicking the host genomic composition to maximize the exploitation of the host cell machinery while minimizing at the same time the recognition by the defence system [15]. Moreover, parasites necessarily share the same environmental conditions of the host, thus it can be hypothesized that comparable forces act on both genomes. Surprisingly, the analysis of 86 viromes and microbiomes revealed that dinucleotide frequencies could allow an effective clustering of the biome based on its origin, suggesting that the environment is actually acting by selecting, directly or indirectly (i.e. favouring a limited number of dominant microorganisms), specific genomic pattern [16]. Similarly, a relation between virus and host genome has been demonstrated by several authors [17,18] and evidences of viral codon bias adaptation to the host one after host jump have been reported [19,20].
Circovirus genus includes species characterized by a monopartite, circular, ssDNA genome of about 1800 to 2000bp. Despite certain among-species variability, two main proteins are encoded in the viral genome: the Rep protein, involved in the host DNA polymerase mediated rolling circle replication, and the Cap one, constituting the viral capsid [21].
Despite the fact that this genus was recognized during the 70s [22,23], its clinical relevance was limited to avian species until the beginning of the nineties; Beak and feather disease virus (BFDV), Pigeon circovirus (PiCV) and Goose circovirus (GoCV) were already known as responsible for relevant diseases, but of marginal economic relevance [24]. It was with the emergence of the Porcine circovirus 2 (PCV-2) that this genus rose as one of the major concerns for veterinary medicine. Due to the advances in diagnostic and sequencing techniques, members of Circovirus genus have been described in several animal species and the number of recognized species has currently increased to 29 species [25]; However,their clinical relevance is often unknown or negligible [26]. The aim of the present study was to investigate the features of dinucleotide and codon bias patterns in mammalian and avian viruses of the genus Circovirus to assess a potential association with the host tropism. Even if dedicated studies on codon usage have been published on some Circovirus species [27–29], no comprehensive comparative analysis relating these viruses with the respective host has currently been performed.
In 2016, a new porcine circovirus, tentatively named Porcine circovirus 3 (PCV-3), was discovered [30] and found in tissues or serum of pigs suffering from different clinical conditions [30–33]; moreover, PCV-3 has also been detected in healthy animals [34]. Therefore, it is still too early to assess if PCV-3 is able to cause disease or not [35]. Interestingly, this new species appears distantly related to all known circoviruses and display, particularly in the capsid gene, a comparable amino-acid distance with mammalian (PCV-2) and avian (DuCV) infecting viruses [30]. Consequently, as a secondary study objective, specific analyses based on viral genomic feature evaluation were performed to provide further insights into the origin of PCV-3.
2. Materials and methods
2.1 Dataset
The whole collection of virus sequences classified into the genus Circovirus was downloaded from the NCBI Taxonomy browser (accessed 15/10/2017). In-house developed Python scripts were used for gene and features extraction, benefiting from the Biopython library functions [36].
Rep and Cap coding sequences were selected if their length was greater than 150 codons and non-terminal stop codon, undetermined nucleotides and out of frame mutations were absent. For each viral strain the sequence was extracted and annotated with the following metadata: accession number, viral and host species, country and date of collection. To homogenize the nomenclature, the reported host name was substituted by the scientific name. Additionally, the Class, Order, Family and Genus of the host were obtained through the NCBI Taxonomy and added to the previously described metadata.
2.2 Viral genome composition analysis
For each sequence the following statistics were obtained: content of each nucleotide (in percentage), total GC content (GC) and in codon positions 1 (GC1), 2 (GC2) and 3 (GC3).
The presence of a statistically significant difference among considered groups was evaluated using the Kruskal-Wallis test followed by Mann-Whitney Test with Bonferroni correction. The significance level was set to p<0.01.
The rho statistic was computed for each dinucleotide pair using the R library seqinr [37]. Briefly, the rho is the frequency of dinucleotide (xy) divided by the product of frequencies of nucleotide (x) and nucleotide (y) and it is expected to be equal to 1.00 when dinucleotide (xy) is formed by chance. To evaluate if some dinucleotide pairs were significantly over- or under-represented, a Z-score was calculated. The Z statistic is the normalization of the rho statistic by its expectation and variance according to a given random sequence generation model (i.e. nucleotide bases shuffling with replacement in their respective codon position).
2.3 Relative synonymous codon usage (RSCU) and effective number of codons (Nc)
The RSCU was calculated using the seqinr package in R. This statistic, indicative of codon bias, is calculated based on the number of times a particular codon is observed, relative to the number of times that the codon would be observed assuming a uniform synonymous codon usage. Consequently, the expected value is 1 in absence of any codon bias while synonymous codons with values lower than 0.6 or greater than 1.6 are regarded as under or over-represented, respectively [19,38].
The Nc values were calculated using the ENCPrime program [39,40]. This summary statistic represents the total number of different codons used in a sequence and can thus range between 21 (only one codon used for each amino-acid) and 60 (all synonymous codons are uniformly used) [12]. A second parameter, the Ncp statistic, also ranging between 21 and 60, was calculated to account for the effect of genome composition on codon bias [12,39]. Obtained Nc and Ncp values were plotted against their GC3 content and compared with the expected Nc distribution under the assumption that it is determined only by GC3 content.
2.4 Neutrality plot
The GC content in the first two codon positions (GC12) of each sequence was plotted against the respective GC3 content and the corresponding linear regression was estimated. This analysis aimed to evaluate the influence of mutational pressure and natural selection on codon usage patterns. If a statistical association would be demonstrated between GC12 and GC3, and the regression coefficient is close to 1, the mutational bias is assumed to be the predominant force driving the codon bias patterns. On the contrary, a regression slope close to 0 suggests the presence of selective pressure acting on and shaping the codon bias evolution. In this sense, the regression coefficient can be interpreted as a quantitative measure of the mutation-selection equilibrium [41–43].
2.5 Principal component analysis (PCA) and hierarchical clustering
The principal component analysis [44] was performed on the RSCU values, after centring and scaling, of the Cap and Rep gene datasets independently, using the prcomp function of the stats library in R [45]. The same approach was used selecting the dinucleotide rho statistics as variables.
Similarly, a hierarchical cluster analysis was performed on the same databases (i.e. RSCU and rho values) using a correlation-based dissimilarity distance. Briefly, the correlation among considered variable profiles was calculated for each sequence pair and converted in a dissimilarity measure (1-cor(X)[j,k]), where j and k are the j'th and k'th object (i.e. viral strain). The hierarchical clustering was calculated using as agglomerative method an average linkage using the hclust function of the stats library.
2.6 Host class prediction
Two different methods were developed to predict the taxonomic class of the infected host based on the viral genome composition (i.e. rho and RSCU). Particularly, a Linear Discriminant Analysis (LDA) and a Random Forest (RF) analysis were validated and their discriminatory capabilities were assessed calculating the Accuracy and the Cohen-K coefficient using a 10 fold cross-validation approach. Since the understanding of PCV-3 origin was one of the aims of the study, PCV-3 sequences were excluded from the training datasets and used only during the host class prediction step. All analyses were performed using the library caret and the relative dependencies [46].
3. Results
3.1Datasets
A total of 2555 Rep and 4424 Cap sequences were included in the final dataset. A complete list of the accession numbers and related information (e.g. virus species, host taxonomy, etc.) are provided as S1 Data.
Unfortunately, the limited number of strains collected from host classes different from Aves and Mammalia precluded the execution of meaningful comparisons. Consequently, unless otherwise stated, the analyses were focused on the circoviruses infecting mammals and birds.
3.2 Genome composition
In the Rep gene, a statistically significant difference was demonstrated between circoviruses infecting different hosts in the mean value of all considered parameters (p-value<0.001). Globally, PCV-3 demonstrated a quite distinct pattern, being more closely related to avian or mammalian infecting viruses depending on the specific parameter studied (Fig 1). The only exception was represented by the A content (p-value = 0.82), where no significant differences were present with respect to circoviruses infecting the Aves class.
Density plot of the different genome composition parameters colour coded accordingly with the specific class category (i.e. Aves: 705 Cap and 933 Rep; Mammalia: 3705 Cap and 1601 Rep). PCV-3 (111 Cap and 40 Rep) has been reported in blue. Both Rep (top) and Cap (bottom) genes have been analysed.
Although a significant diversity was demonstrated between avian and mammalian infecting viruses in the Cap gene (p-value<0.001), the PCV-3 genome composition appeared globally overlapping with the one of avian circoviruses (Fig 1). No statistically significant differences were found between these and PCV-3 in A (p-value = 0.019), C (p-value = 0.043) and GC3 content (p-value = 0.78).
Comparable results were obtained evaluating the dinucleotide composition. A less evident distinction was evident in the Rep gene a between circoviruses infecting Aves and Mammalia and among PCV-3 and the other two viral groups (S1 Fig). On the other hand, a clearer resemblance between PCV-3 and the avian circoviruses was evident in the capsid gene for practically all dinucleotide pairs (S1 Fig). The Z-score calculation in the Rep gene evidenced no under- or over-represented dinucleotide pair with the following exception: CpC, GpA and TpG were over-represented in mammalian infecting circoviruses while the CpG and TpC were under-represented in mammalian and avian circoviruses, respectively. CpG and GpG were slightly under- and over-represented in the PCV-3 Rep gene (S2 Fig). In the Cap gene, only mammalian circoviruses showed a significant deviation of dinucleotide frequency from what was expected; particularly, ApA, CpC, CpT and TpG were significantly over-represented while CpG and TpC were under-represented. Remarkably, a stronger bias was observed in the Cap gene compared to the Rep one. The TpT pair was the only dinucleotide over-represented in PCV-3 (S2 Fig).
3.3 Codon usage
Codon usage analysis showed a globally lower bias in the Rep gene compared with the Cap gene. The codons with a RSCU <0.6 or >1.6 are reported in Table 1 and S3 Fig. Briefly, 8, 9 and 14 codons were under-represented in the Rep of viruses infecting Aves, Mammalia and in PCV-3 while 4, 8 and 9 were over-represented, respectively. In the Cap gene, 9, 14 and 19 codons were under-represented in viruses infecting Aves, Mammalia and in PCV-3 while 7, 9 and 12 were over-represented, respectively. PCV-3 had a globally higher bias in several codons; however, the remarkably lower number of available sequences increased the likelihood of more extreme values.
3.4 Nc plot
The RSCU results were substantially confirmed by the Nc analysis, being Nc values significantly (p-value<0.001) higher (i.e. lower codon bias) on average in the Rep than in Cap gene (p-value<0.001) and globally lower in mammalian infecting viruses compared to avian ones (p-value<0.001) (Fig 2). Circoviruses infecting mammals demonstrated a higher deviation of the Nc values from the expectation based on GC3 (S4 Fig). On the contrary, avian circoviruses Nc globally mimicked the expected pattern, at least for the Rep gene. Interestingly, when the background nucleotide composition was accounted, an overall reduction in the gap between expected and observed values was observed. This reduction was particularly evident for the Ncp value of PCV-3 in the Rep gene, which substantially moved to the expected range. Similarly, the Ncp values of the avian circoviruses overlapped the expected ones in the Rep gene, while a significant deviation was still present for some avian strains in the Cap gene and in both Rep and Cap genes of the Mammalia infecting circoviruses.
Scatterplot reporting the relationship between Nc and Ncp and GC3 content for the Rep and Cap genes. Avian and Mammals circoviruses and PCV-3 have been color-coded. The line representing the expected Nc values, which would result from GC composition being the only factor influencing the codon usage bias, has been superimposed.
3.5 Neutrality plot
A statistically significant relationship between GC12 and GC3 was found in the Rep gene of avian circoviruses (b = 0.27; p-value<0.001), mammalian circoviruses (b = 0.47; p-value<0.001) but not in that of PCV-3 (p-value = 0.26), although a certain correlation seemed present (b = 0.21). Similar results were obtained for the Cap of avian circoviruses (b = 0.29; p-value <0.001) but not of mammalian circoviruses, which slope was remarkably lower (b = 0.13, p-value < 0.001), and of PCV-3, demonstrating no relationship between GC12 and GC3 (b = 0.02;p-value = 0.75). Consequently, the mutational drift accounted for about 25% of the codon bias in avian circoviruses, 40% and 15% in the Rep and Cap of mammalian circoviruses, and was negligible for PCV-3, although some evidences of its action were present in the Rep gene.
3.6 Principal component analysis and hierarchical clustering
After PCA eigenvalues evaluation, the first two principal components (PC1 and PC2) were maintained for both Rep and Cap genes, since they explained a good percentage (i.e. always greater than 45%) of the observed variability. The avian and mammalian circoviruses formed, with few exceptions, easily separable groups when recoded using the PC1 and PC2 obtained from either the RSCU or the rho datasets. PCV-3 represented an interesting exception since it was related to mammalian viruses for the Rep gene, particularly in terms of codon bias, while it appeared more similar to avian infecting species for the Cap gene (Fig 3). Fully comparable results were obtained using the hierarchical clustering approach (S5 Fig). To investigate the possible relationship between PCV-3 and other circoviruses infecting hosts not belonging to the Aves and Mammalia classes, a hierarchical clustering was performed on the whole dataset, which included all the available host taxa. Nevertheless, PCV-3 always clustered with avian or mammalian circoviruses, in accordance with the patterns previously described (data not shown).
Scatter plot based on the first two components of the PCA performed on RSCU and rho values calculated on mammal and avian circoviruses. For interpretation easiness, PCV-3 and Chiroptera circoviruses have been highlighted with different colours. The PCA loading are represented as arrows. The 95% confidence ellipses around clusters are also reported. Both Rep (top) and Cap (bottom) genes have been analysed.
3.7 Predictive methods
The two validated predictive models showed remarkable discriminative capabilities for both the Rep and Cap genes (Fig 4). However, when the host class was predicted for PCV-3 sequences, conflicting results were obtained between the two considered genes. In fact, the Rep gene was classified as “Mammalia” even with a relatively high degree of uncertainness, while the Cap gene was classified in the “Aves” infecting virus category (Table 2).
Distribution of diagnostic performance metrics of RF and LDA evaluated by cross-validation on Cap and Rep datasets.
4. Discussion
Virus existence and maintenance lie on an intimate relationship with their host, since they depend on the same cell machinery, share the same physical and biochemical environment, and struggle with each other for survival. Consequently, viruses are expected to mirror (or at least be influenced by) the host genomic features, which have a huge impact on genome structure, RNA transcription and stability, protein translation and folding [11]. A marked adaptation to the “host environment” appears particularly realistic for ssDNA viruses since they totally depend on the host cell machinery for replication and are devoid of the panel of proteins used by other, more complex viruses, to interfere with the host immune response [47]. The results of the present study largely confirm this host-adaptation, since circoviruses infecting avian and mammalian species show a quite distinct pattern in terms of genome composition, dinucleotide frequency and codon bias.
Avian circoviruses show a globally higher C, G and particularly CpG content compared with mammalian ones, whose genome was proven to be deprived of these nucleotides. Remarkably, the avian genome does not differ significantly from the mammalian one with regard to CpG content and overall GC percentage [48]. Thus, other explanations to the different circovirus composition patterns must be claimed besides simple genomic mimicking. The fitness and spreading of ssDNA viruses depend on a rapid replication, anticipating the development of an effective host immune response [47]. Therefore, a high CpG dinucleotide content was proposed to be deleterious for these viruses since it can slow down the duplication and transcription processes because of the high stacking energy of CpG dinucleotide pair [47]. However, avian species typically exhibit a higher body temperature than mammals [49,50] and the different thermodynamic environment could provide enough free energy for an efficient replication, while the greater GC content would guarantee the stability of relevant secondary structure [51,52]. The deamination of cytosine to thymines has been proposed to explain the CpG under-representation in vertebrate genomes. However, the methylation of actively replicating virus genomes, although proven [53], is still a poorly documented phenomenon which frequency and relevance remain unknown [54].
Interestingly, a similar scenario was described for influenza A virus after its host jump from birds to human, leading to the 1918 pandemic. Since then, an overall decrease in CpG content of influenza viruses was observed, which has been attributed to an attempt to reduce the Toll-like receptor (TLR) (potentially TLR3, TLR7, TLR8 and RIG-I) viral recognition mediated by CpG motifs of the RNA molecules [15]. Remarkably, the ancestral 1918 influenza virus strain and modern avian derived strains appear to induce a more marked innate immune response [55]. A similar mechanism could be involved also in the different dinucleotide pattern observed in avian and mammalian circoviruses. In vertebrates, unmethylated CpG motifs are involved in the recognition of viral DNA genome mediated by the TLR-9. Interestingly, this TLR has been deleted by the avian genome [56] and no orthologue gene has been found [57]. In avian species, a comparable function is carried out by the TLR-21, which appears sensitive to the same motifs [57,58]. However, a differential activation has been reported between TLR-9 and TLR-21 when stimulated by pathogens [59,60] and, therefore, a different virus-host interaction may take place. If these physiological differences are actually responsible for differential evolutionary pressures acting on virus evolution, needs further investigation.
The Cap gene demonstrated a marked codon bias in avian and especially in mammalian circoviruses. In the latter, a relevant deviation from the expected Nc based on GC3 was observed. Accordingly, the neutrality plot analysis comes out on the side of a prominent action of natural selection on mammalian circovirus Cap codon bias, whereas mutational drift is more involved in the Rep gene evolution and, more generally, in the evolution of avian circoviruses as well. While an exhaustive explanation of the different patterns observed in the two animal classes is challenging, the evidence that evolution appears to be directed towards the selection of CpG depleted synonymous codons, particularly in highly expressed capsid protein, suggests a trade-off between a tendency to optimize the capsid protein translation while minimizing the recognition of the genome and the transcript molecules.
In 2016, a new porcine circovirus (PCV-3) was discovered in pigs. The recent identification and the low genetic diversity of the currently sequenced strains, which would suggest a recent PCV-3 origin, conflict with the absence of closely related circoviruses. Remarkably, the analysis of the genome composition, dinucleotide frequency and codon bias led to cluster the capsid gene of this virus together with avian circoviruses. On the contrary, a resemblance was observed between PCV-3 Rep gene and other mammalian circoviruses. These results were further confirmed by two independent classification methods that performed excellently on other known circoviruses. Although the development of host-prediction tools was beyond the scope of the present study, the accurate results provided by the two methods demonstrated that codon bias and genome composition were informative enough to predict the viral tropism and, indirectly, support the effect of host environment in shaping viral genome evolution. Surprisingly, the CpG content in the PCV-3 Cap gene substantially overlaps the one of avian circoviruses, which is in sharp contrast with the hypothesized role of CpG depletion in reducing mammal innate immunity activation. Moreover, while the PCV-3 Rep gene effective number of codons can be explained mainly by genome composition background (as shown in Fig 2), other forces appear to remarkably act on the Cap gene. Accordingly, the PCV-3 Cap gene was the only one where absolutely no correlation was demonstrated between GC12 and GC3 content. Therefore, the presence of a strong selective pressure shaping the PCV-3 Cap gene patterns can be confidently stated; this scenario is fully compatible with the recent introduction in a new environment (i.e. from avian to mammals species), as demonstrated for other viruses experiencing a recent host jump [15,20]. The role of recombination in the emergence of this new virus can therefore be suggested. In fact, although the clustering with mammalian circoviruses appeared globally weak, particularly at dinucleotide level, PCV-3 exhibited a rather surprising similarity with some bat circoviruses in the Rep region, either in codon usage and dinucleotide frequency. Members of the order Chiroptera are reservoirs of several viruses and are considered the source of many emerging diseases [61]. Many biological features enable them to carry a diversity of viruses. They represent about 20% of all mammalian species [62], hence providing a remarkable genetic heterogeneity. Since their ancient origin (about 52.5 million years ago), many viruses could have progressively co-evolved with bats [63]. Moreover, the absence of a bone marrow producing B cell as well as other peculiarities in the immune system (reviewed in Baker et al., 2013) [64] provide a favourable immune environment for viruses to survive and being maintained in these species [65]. Finally, their worldwide distribution, social behaviour and ability to fly guarantee advantageous conditions for the genesis of huge viral populations and their spreading [63]. Despite no clear evidences are available about the bat role as mixing vessels for avian and mammalian viruses, some data suggest their potential susceptibility to both types of viruses. Serological data have reported a seroprevalence of about 30% against avian influenza subtype H9 in Ghanaian bats [66] and little brown bats (Myotis lucifugus) were proven to co-express both avian and human type influenza receptors in their respiratory and gastrointestinal systems [67]. Remarkably, different species of bat circoviruses have different genome composition, ranging from mammalian- to avian-like (as shown in Fig 3). Therefore, the possibility to harbour genetically distant viruses could have favoured the emergence of recombination events. While these results can-not be automatically used to infer a bat role in PCV-3 emergence and the intrinsically poorly informative genetic data may hinder definitive conclusions, they at least support the plausibility of the offered hypotheses. Unfortunately, the knowledge of the virosphere is still at its infancy and the lack of information hampers both more precise identification of the bat role in avian-like virus evolution as well as the understanding of the PCV-3 origin.
In conclusion, the present study demonstrates the presence of quite distinctive patterns in genomic composition, dinucleotide frequency and codon bias between circoviruses infecting mammalian and avian species. Although several forces appears to be in place, including the mutational bias, a significant trade-off between the reduction of host innate immune response recognition and the maximization of translation efficacy, particularly of the capsid protein, seems to be the driving forces shaping circovirus genomic evolution. Moreover, the analysis of these parameters allowed to speculate a potential recombinant origin, followed (or preceded) by a host jump, of PCV-3. The genome of this virus appears to result from the combination of a mammalian-virus (likely a bat-circovirus) Rep gene with an avian circovirus-like Cap gene.
Supporting information
S1 Data. List of selected sequences and metadata.
List of Rep and Cap sequences used in the study. Additional metadata including viral and host species taxonomy are reported.
https://doi.org/10.1371/journal.pone.0199950.s001
(XLT)
S1 Fig. Density plot of dinucleotide pairs.
Density plot of the different dinucleotide pairs colour coded accordingly with the specific class category. PCV-3 has been highlighted in blue.
https://doi.org/10.1371/journal.pone.0199950.s002
(PDF)
S2 Fig. Z-score of dinucleotide pairs.
The mean value (points) and 95CI (error-bars) of the Z-score for different dinucleotide pairs are reported and colour-coded according to the animal class. Both Rep (top) and Cap (bottom) genes have been analysed. Z-score higher and lower than 1.96 (i.e. statistically different from 0) have been highlighted by dotted lines.
https://doi.org/10.1371/journal.pone.0199950.s003
(PDF)
S3 Fig. Relative synonymous codon usage.
The mean value (points) and 95CI (error-bars) of RSCU of the analyzed viral strains are reported and colour-coded according to the animal class. Both Rep (top) and Cap (bottom) genes have been analysed. The values corresponding to overrepresented and underrepresented codon thresholds have been reported as dotted lines.
https://doi.org/10.1371/journal.pone.0199950.s004
(PDF)
S4 Fig. Deviation of Nc values from expectation.
Boxplot reporting the deviation of the Rep and Cap genes Nc values from the expectations based on GC3. Avian and Mammals circoviruses and PCV-3 have been colour-coded.
https://doi.org/10.1371/journal.pone.0199950.s005
(PDF)
S5 Fig. Hierarchical clustering.
Hierarchical clustering obtained from rho and RSCU values of the Rep and Cap genes. The different animal groups have been colour coded. For graphical reasons, only a subset (i.e. a maximum of 5 randomly selected sequences for each viral species) is represented.
https://doi.org/10.1371/journal.pone.0199950.s006
(PDF)
References
- 1. Karlin S, Campbell AM, Mrázek J. COMPARATIVE DNA ANALYSIS ACROSS DIVERSE GENOMES. Annu Rev Genet. 1998;32: 185–225. pmid:9928479
- 2. Campbell A, Mrázek J, Karlin S. Genome signature comparisons among prokaryote, plasmid, and mitochondrial DNA. Proc Natl Acad Sci U S A. 1999;96: 9184–9. Available: http://www.ncbi.nlm.nih.gov/pubmed/10430917 pmid:10430917
- 3. Gentles AJ, Karlin S. Genome-Scale Compositional Comparisons in Eukaryotes. Genome Res. 2001;11: 540–546. pmid:11282969
- 4. Beutler E, Gelbart T, Han JH, Koziol JA, Beutler B. Evolution of the genome and the genetic code: selection at the dinucleotide level by methylation and polyribonucleotide cleavage. Proc Natl Acad Sci U S A. 1989;86: 192–6. Available: http://www.ncbi.nlm.nih.gov/pubmed/2463621 pmid:2463621
- 5. Bird AP. DNA methylation and the frequency of CpG in animal DNA. Nucleic Acids Res. 1980;8: 1499–504. Available: http://www.ncbi.nlm.nih.gov/pubmed/6253938 pmid:6253938
- 6. Yakovchuk P, Protozanova E, Frank-Kamenetskii MD. Base-stacking and base-pairing contributions into thermal stability of the DNA double helix. Nucleic Acids Res. 2006;34: 564–574. pmid:16449200
- 7. Hershberg R, Petrov DA. Selection on codon bias. TL—42. Annu Rev Genet. Annual Reviews; 2008;42 VN-r: 287–299. pmid:18983258
- 8. Chen SL, Lee W, Hottes AK, Shapiro L, McAdams HH. Codon usage between genomes is constrained by genome-wide mutational processes. Proc Natl Acad Sci U S A. National Academy of Sciences; 2004;101: 3480–5. pmid:14990797
- 9. Knight RD, Freeland SJ, Landweber LF. A simple model based on mutation and selection explains trends in codon and amino-acid usage and GC composition within and across genomes. Genome Biol. 2001;2: 1–13.
- 10. Parmley JL, Huynen MA. Clustering of Codons with Rare Cognate tRNAs in Human Genes Suggests an Extra Level of Expression Regulation. Petrov DA, editor. PLoS Genet. Public Library of Science; 2009;5: e1000548. pmid:19578405
- 11. Chaney JL, Clark PL. Roles for Synonymous Codon Usage in Protein Biogenesis. Annu Rev Biophys. 2015;44: 143–166. pmid:25747594
- 12.
Roth A, Anisimova M, Cannarozzi GM. Measuring codon usage bias. Codon Evolution: Mechanisms and Models. Oxford University Press; 2012. pp. 189–217. https://doi.org/10.1093/acprof:osobl/9780199601165.003.0013
- 13. Carlini DB. Experimental reduction of codon bias in the Drosophila alcohol dehydrogenase gene results in decreased ethanol tolerance of adult flies. J Evol Biol. Blackwell Science Ltd; 2004;17: 779–785. pmid:15271077
- 14. Shah P, Gilchrist M a. Explaining complex codon usage patterns with selection for translational efficiency, mutation bias, and genetic drift. Proc Natl Acad Sci U S A. National Academy of Sciences; 2011;108: 10231–10236. pmid:21646514
- 15. Greenbaum BD, Levine AJ, Bhanot G, Rabadan R. Patterns of evolution and host gene mimicry in influenza and other RNA viruses. PLoS Pathog. Public Library of Science; 2008;4: e1000079. pmid:18535658
- 16. Willner D, Thurber RV, Rohwer F. Metagenomic signatures of 86 microbial and viral metagenomes. Environ Microbiol. Blackwell Publishing Ltd; 2009;11: 1752–1766. pmid:19302541
- 17. Bahir I, Fromer M, Prat Y, Linial M. Viral adaptation to host: a proteome-based analysis of codon usage and amino acid preferences. Mol Syst Biol. 2009;5: 311. pmid:19888206
- 18. van Hemert FJ, Berkhout B, Lukashov V V. Host-related nucleotide composition and codon usage as driving forces in the recent evolution of the Astroviridae. Virology. Academic Press; 2007;361: 447–454. pmid:17188318
- 19. Wong EHM, Smith DK, Rabadan R, Peiris M, Poon LLM. Codon usage bias and the evolution of influenza A viruses. Codon Usage Biases of Influenza Virus. BMC Evol Biol. BioMed Central Ltd; 2010;10: 253. pmid:20723216
- 20. Franzo G, Tucciarone CM, Cecchinato M, Drigo M. Canine parvovirus type 2 (CPV-2) and Feline panleukopenia virus (FPV) codon bias analysis reveals a progressive adaptation to the new niche after the host jump. Mol Phylogenet Evol. 2017;114: 82–92. pmid:28603036
- 21. Hulo C, de Castro E, Masson P, Bougueleret L, Bairoch A, Xenarios I, et al. ViralZone: a knowledge resource to understand virus diversity. Nucleic Acids Res. 2011;39: D576–D582. pmid:20947564
- 22. Tischer I, Gelderblom H, Vettermann W, Koch MA. A very small porcine virus with circular single-stranded DNA. Nature. Nature Publishing Group; 1982;295: 64–66.
- 23. Tischer I, Rasch R, Tochtermann G. Characterization of papovavirus-and picornavirus-like particles in permanent pig kidney cell lines. Zentralbl Bakteriol Orig A. 1974;226: 153–167. Available: http://www.ncbi.nlm.nih.gov/pubmed/4151202 pmid:4151202
- 24. Todd D. Avian circovirus diseases: Lessons for the study of PMWS. Veterinary Microbiology. 2004. pp. 169–174.
- 25. King AMQ, Lefkowitz EJ, Mushegian AR, Adams MJ, Dutilh BE, Gorbalenya AE, et al. Changes to taxonomy and the International Code of Virus Classification and Nomenclature ratified by the International Committee on Taxonomy of Viruses (2018). Arch Virol. 2018; pmid:29754305
- 26. Rosario K, Breitbart M, Harrach B, Segalés J, Delwart E, Biagini P, et al. Revisiting the taxonomy of the family Circoviridae: establishment of the genus Cyclovirus and removal of the genus Gyrovirus. Arch Virol. 2017;162: 1447–1463. pmid:28155197
- 27. Xu Y, Jia R, Zhang Z, Lu Y, Wang M, Zhu D, et al. Analysis of synonymous codon usage pattern in duck circovirus. Gene. 2015;557: 138–45. pmid:25497833
- 28. Liu X, Zhang Y, Fang Y, Wang Y. Patterns and influencing factor of synonymous codon usage in porcine circovirus. Virol J. 2012;9: 68. pmid:22416942
- 29. Chen Y, Sun J, Tong X, Xu J, Deng H, Jiang Z, et al. First analysis of synonymous codon usage in porcine circovirus. Arch Virol. Guangdong Provincial Key Lab of Agro-animal Genomics and Molecular Breeding, College of Animal Science, South China Agricultural University, Guangzhou, Guangdong, 510642, People’s Republic of China.; 2014;159: 2145–2151. pmid:24557524
- 30. Palinski R, Piñeyro P, Shang P, Yuan F, Guo R, Fang Y, et al. A Novel Porcine Circovirus Distantly Related to Known Circoviruses Is Associated with Porcine Dermatitis and Nephropathy Syndrome and Reproductive Failure. J Virol. American Society for Microbiology; 2017;91: e01879–16. pmid:27795441
- 31. Phan TG, Giannitti F, Rossow S, Marthaler D, Knutson T, Li L, et al. Detection of a novel circovirus PCV3 in pigs with cardiac and multi-systemic inflammation. Virol J. 2016;13: 1–8.
- 32. Ku X, Chen F, Li P, Wang Y, Yu X, Fan S, et al. Identification and genetic characterization of porcine circovirus type 3 in China. Transbound Emerg Dis. 2017;64: 703–708. pmid:28317326
- 33. Kwon T, Yoo SJ, Park CK, Lyoo YS. Prevalence of novel porcine circovirus 3 in Korean pig populations. Vet Microbiol. 2017;207: 178–180. pmid:28757021
- 34. Zheng S, Wu X, Zhang L, Xin C, Liu Y, Shi J, et al. The occurrence of porcine circovirus 3 without clinical infection signs in Shandong Province. Transbound Emerg Dis. 2017;64: 1337–1341. pmid:28653486
- 35. Franzo G, Legnardi M, Tucciarone CM, Drigo M, Klaumann F, Sohrmann M, et al. Porcine circovirus type 3: a threat to the pig industry? Vet Rec. British Medical Journal Publishing Group; 2018;182: 83. pmid:29351975
- 36. Cock PJA, Antao T, Chang JT, Chapman BA, Cox CJ, Dalke A, et al. Biopython: freely available Python tools for computational molecular biology and bioinformatics. Bioinformatics. Oxford University Press; 2009;25: 1422–1423. pmid:19304878
- 37.
Charif D, Humblot L, Lobry JR, Necsulea A, Palmeira L, Penel S. SeqinR 2.0–1: A contributed package to the R project for statistical computing devoted to biological sequences retrievel and analysis. Structural approaches to sequence evolution. Springer; 2008. p. 268.
- 38. Ma M, Ha X, Ling H, Wang M, Zhang F, Zhang S, et al. The characteristics of the synonymous codon usage in hepatitis B virus and the effects of host on the virus in codon usage pattern. Virol J. BioMed Central; 2011;8: 544. pmid:22171933
- 39. Novembre J a. Letter to the Editor Accounting for Background Nucleotide Composition When Measuring Codon Usage Bias. Amino Acids. 2000;2: 1390–1394. pmid:12140252
- 40.
Novembre J. User Documentation for ENCprime. 2006; 1–6.
- 41. Sueoka N. Directional mutation pressure and neutral molecular evolution. Proc Natl Acad Sci U S A. National Academy of Sciences; 1988;85: 2653–2657.
- 42. Nasrullah I, Butt AM, Tahir S, Idrees M, Tong Y. Genomic analysis of codon usage shows influence of mutation pressure, natural selection, and host features on Marburg virus evolution. BMC Evol Biol. 2015;15: 174. pmid:26306510
- 43. Kumar N, Bera BC, Greenbaum BD, Bhatia S, Sood R, Selvaraj P, et al. Revelation of Influencing Factors in Overall Codon Usage Bias of Equine Influenza Viruses. PLoS One. Public Library of Science; 2016;11: e0154376. pmid:27119730
- 44. Su M, Lin H, Yuan HS, Chu W. Categorizing host-dependent RNA viruses by principal component analysis of their codon usage preferences. J Comput Biol. 2009;16: 1539–47. pmid:19958082
- 45.
Team RC. R: a language and environment for statistical computing. Vienna (Austria): R Foundation for statistical computing. ISBN: 3-900051-07-0. R Foundation for Statistical Computing; 2011.
- 46. Kuhn M. caret Package. J Stat Softw. 2008;28: 1–26. Available: http://www.jstatsoft.org/v28/i05/paper
- 47. Shackelton LA, Parrish CR, Holmes EC. Evolutionary basis of codon usage and nucleotide composition bias in vertebrate DNA viruses. J Mol Evol. 2006;62: 551–563. pmid:16557338
- 48. Di Giallonardo F, Schlub TE, Shi M, Holmes EC. Dinucleotide Composition in Animal RNA Viruses Is Shaped More by Virus Family than by Host Species. Dermody TS, editor. J Virol. 2017;91: e02381–16. pmid:28148785
- 49.
Sturkie PD. Avian Physiology [Internet]. Springer New York; 1986. https://books.google.it/books?hl=it&lr=&id=TwrVBwAAQBAJ&oi=fnd&pg=PA48&dq=avian+physiology&ots=DZoENb0GH4&sig=IoYk7Bce3T1aeMZsuoMhqIBgotk#v=onepage&q=avian%20physiology&f=false
- 50.
Myers P, Espinosa R, Parr CS, Jones T, Hammond GS, Dewey TA. The animal diversity web. Accessed Oct. 2006;12: 2.
- 51. Wang J, Dong P, Wu W, Pan X, Liang X. High-throughput thermal stability assessment of DNA hairpins based on high resolution melting. J Biomol Struct Dyn. Taylor & Francis; 2016; 1–13. pmid:28024437
- 52. Wang AH-J, Hakoshima T, van der Marel G, van Boom JH, Rich A. AT base pairs are less stable than GC base pairs in Z-DNA: The crystal structure of d(m5CGTAm5CG). Cell. Cell Press; 1984;37: 321–331.
- 53. Bonvicini F, Manaresi E, Di Furio F, De Falco L, Gallinella G. Parvovirus B19 DNA CpG Dinucleotide Methylation and Epigenetic Regulation of Viral Expression. Sinclair AJ, editor. PLoS One. Public Library of Science; 2012;7: e33316. pmid:22413013
- 54. Hoelzer K, Shackelton LA, Parrish CR. Presence and role of cytosine methylation in DNA viruses of animals. Nucleic Acids Res. Oxford University Press; 2008;36: 2825–37. pmid:18367473
- 55. Kobasa D, Jones SM, Shinya K, Kash JC, Copps J, Ebihara H, et al. Aberrant innate immune response in lethal infection of macaques with the 1918 influenza virus. Nature. Nature Publishing Group; 2007;445: 319–323. pmid:17230189
- 56. Temperley ND, Berlin S, Paton IR, Griffin DK, Burt DW. Evolution of the chicken Toll-like receptor gene family: A story of gene gain and gene loss. BMC Genomics. 2008;9: 62. pmid:18241342
- 57. Brownlie R, Zhu J, Allan B, Mutwiri GK, Babiuk LA, Potter A, et al. Chicken TLR21 acts as a functional homologue to mammalian TLR9 in the recognition of CpG oligodeoxynucleotides. Mol Immunol. 2009;46: 3163–70. pmid:19573927
- 58. Brownlie R, Allan B. Avian toll-like receptors. Cell Tissue Res. 2011;343: 121–130. pmid:20809414
- 59. Dalpke A, Frank J, Peter M, Heeg K. Activation of Toll-Like Receptor 9 by DNA from Different Bacterial Species. Infect Immun. 2006;74: 940–946. pmid:16428738
- 60. de Zoete MR, Keestra AM, Roszczenko P, van Putten JPM. Activation of Human and Chicken Toll-Like Receptors by Campylobacter spp. Infect Immun. 2010;78: 1229–1238. pmid:20038539
- 61. Hughes JM, Wilson ME, Halpin K, Hyatt AD, Plowright RK, Epstein JH, et al. Emerging Viruses: Coming in on a Wrinkled Wing and a Prayer. Clin Infect Dis. 2007;44: 711–717. pmid:17278066
- 62. Turmelle AS, Olival KJ. Correlates of viral richness in bats (order Chiroptera). Ecohealth. 2009;6: 522–39. pmid:20049506
- 63. Han H-J, Wen H, Zhou C-M, Chen F-F, Luo L-M, Liu J, et al. Bats as reservoirs of severe emerging infectious diseases. Virus Res. 2015;205: 1–6. pmid:25997928
- 64. Baker ML, Schountz T, Wang L-F. Antiviral Immune Responses of Bats: A Review. Zoonoses Public Health. Blackwell Publishing Ltd; 2013;60: 104–116. pmid:23302292
- 65. Dobson AP. VIROLOGY: What Links Bats to Emerging Infectious Diseases? Science (80-). 2005;310: 628–629. pmid:16254175
- 66. Freidl GS, Binger T, Müller MA, de Bruin E, van Beek J, Corman VM, et al. Serological Evidence of Influenza A Viruses in Frugivorous Bats from Africa. Baker ML, editor. PLoS One. Public Library of Science; 2015;10: e0127035. pmid:25965069
- 67. Chothe SK, Bhushan G, Nissly RH, Yeh Y-T, Brown J, Turner G, et al. Avian and human influenza virus compatible sialic acid receptors in little brown bats. Sci Rep. Nature Publishing Group; 2017;7: 660. pmid:28386114