2.1. Sample Collection

The investigated saliva and rectal swab samples originated from 206 huntable animals, including red fox (n = 165), European badger (Meles meles, n = 24), golden jackal (Canis aureus, n = 13), stone marten (Martes foina, n = 3) and raccoon dog (Nyctereutes procyonoides, n = 1). The samples were collected by professional hunters in Central Europe, in Northeastern Hungary (Borsod-Abaúj-Zemplén county, Taktaköz) between December 2021 and March 2022. The saliva and rectal swabs were placed in the same tube natively and immediately stored at −20°C until delivery to the National Laboratory of Virology, University of Pécs, Pécs, Hungary, where they were processed.

2.2. Nucleic Acid Extraction and PCR

The sample swabs were washed in 400 μL phosphate-buffered saline (PBS) with extensive vortexing. Total RNA was extracted using Direct-Zol RNA MiniPrep Kit (Zymo Research, USA), and, following reverse transcription, the samples were screened with universal nested PCR primers for coronaviruses [35]. The first PCR of the nested system was carried out with the Luna Universal Probe One-Step RT-qPCR Kit (New England Biolabs, Ipswich, MA, USA). The PCR mixture of 20 μL contained 10 μL 2x Luna Universal Probe One-Step Reaction Mix, 3 μL nuclease-free water, 1 μL Luna WarmStart RT Enzyme Mix, 0.5 μL of each primer mix (200 nM of equimolar mixture: PC2S2−1, PC2S2−2, and 900 nM of equimolar mixture: PC2AS1-1, PC2AS1-2, PC2AS1-3), and 5 μL of the nucleic acid. The cycling protocol consisted of the steps: reverse transcription at 55°C for 10 min; denaturation at 95°C for 1 min; a touchdown protocol with 10 cycles, with the steps denaturation at 94°C for 20 s, annealing starting at 62°C with a decrease of 1°C per cycle for 30 s, and extension at 60°C for 40 s; 30 cycles with the steps denaturation at 94°C for 20 s, annealing at 52°C for 30 s, and extension at 60°C for 40 s; a final extension step at 60°C for 5 min. The second PCR was executed using GoTaq G2 Flexi DNA Polymerase Kit (Promega, WI, USA). In this case the PCR mixture of 25 μL contained 5 μL 5x Green GoTaq Flexi Reaction Buffer, 13.75 μL Nuclease-free Water, 0.25 μL GoTaq G2 Flexi DNA Polymerase (5 U/µL), 1 μM dNTP mix, 2 μL MgCl2 Solution (2.5 mM), 0.5 μL of each primer (80 nM of equimolar mixture: PCS-1, PCS-2, and 400 nM of primer: PCNAS), and 2 μL of the nucleic acid. The cycling protocol consisted of the steps: denaturation at 94°C for 2 min; 30 amplification cycles of denaturation at 94°C for 20 s, annealing at 60°C for 30 s, and extension at 72°C for 30 min; a final extension step at 72°C for 5 min. The PCR products were analyzed by standard agarose gel electrophoresis.

The sequence gaps, remained after assembling the reads generated with next-generation sequencing (NGS) were completed by PCR. Oligonucleotides were designed based on the contigs and used for amplification with DreamTaq DNA polymerase (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions (data not shown). The oligos were used for direct sequencing, executed by service companies.

2.3. Enrichment of Viral Sequences

Before NGS, an enrichment step was applied to increase the proportion of the viral origin reads [36]. The original samples containing coronavirus sequences were centrifuged at 16,000 × g for 10 min, and 160 μL of each supernatant was filtered through a centrifuge membrane filter with pore size of 0.45 μm (Ultrafree-CL Centrifugal Filter, Merck Millipore, Darmstadt, Germany). One hundred and fifty μL of the filtered samples were treated with a cocktail of 1 μL micrococcal nuclease (New England Biolabs, Ipswich, MA, USA), 2 μL of benzonase (Merck Millipore, Darmstadt, Germany), 4.5 µL of Turbo DNase, and 15.5 µL Turbo DNase Buffer (Thermo Fisher Scientific, Waltham, MA, USA) for 2 hours at 37°C, and were extracted with the Direct-zol RNA MiniPrep Kit (Zymo Research, USA).

2.4. NGS

RNA library was generated using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA). Briefly, 10 ng of the RNA was used as input for fragmentation and for cDNA generation using random primers. Thereafter, the cDNA was end-prepped and adapter-ligated, and then the libraries were amplified according to the manufacturer’s instructions. The quality of the libraries was checked on an Agilent 4200 TapeStation System using D1000 Screen Tape (Agilent Technologies, Palo Alto, CA, USA), while the quantity was measured on Qubit 3.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Illumina sequencing was performed on a NovaSeq 6000 instrument (Illumina, San Diego, CA, USA) in a 2 × 151 cycles run configuration.

2.5. Bioinformatics

Illumina reads were de novo assembled with MEGAHIT v.1.2.9. The contigs were checked with the basic local alignment search tool (BLAST) against the NCBI GenBank database. The closest related sequences were used as references for mapping of the raw Illumina reads with the built-in mapper of Geneious Prime v.2024.0.3. The resulting sequences were checked manually to reveal possible errors. Raw reads were remapped to the resulting genomes for another manual check. As indicated above, the sequence gaps were filled with reads generated by direct sequencing. The genomes were annotated with Geneious Prime v.2024.0.3 and the Open Reading Frame Finder (https://www.ncbi.nlm.nih.gov/orffinder/).

The Kraken2 v1.0 (database PlusPF) was used for the taxonomic classification of raw reads, and the results were visualized with KronaTools and Pavian [37–39].

The generated genomic sequences were edited and aligned with the MUSCLE algorithm of the AliView software, or the MAFFT algorithm of the Geneious Prime software [40]. Pairwise identity values and neighbor-joining phylogenetic trees were generated with the MEGA 11 software [41]. Maximum likelihood phylogeny was investigated with the PhyML 3.0 online tool [42]. Recombination analysis was carried out with the Recombination Detection Program 4, using the RDP, GENECONV, BootScan, MaxChi, Chimera, SiScan, and 3Seq methods [43].

3.1. PCR and Metagenomics

Using the diagnostic nested PCR system, coronavirus sequences were detected in two red fox samples of a juvenile female (shot on 22.01.2022) and a juvenile male (shot on 14.02.2022) [35]. Both samples were further investigated with metagenomics tools.

Altogether 13,800,556 and 18,228,825 microbial reads could be identified from the processed sample fox13 and fox113, respectively. Despite the filtration and DNase/RNase digestion, the number of reads matching viruses was relatively low. According to the Kraken2 program, 3138 reads derived from sample fox13, and 2456 reads obtained from sample fox113 mapped to mammalian virus sequences, mostly to coronaviruses (Coronaviridae), as well as to picornaviruses (Picornaviridae) and circoviruses (Circoviridae) in the case of the sample fox113 (Figure 1). The number of bacteriophage sequences was relatively low (4.7% and 6% for the samples fox 13 and 113, respectively).

3.2. Coronaviruses

One long, de novo assembled sequence contig, generated from both red fox samples, matched with CCoV sequences. Altogether, 6459 (average depth of 34) and 4317 reads (average depth of 23) mapped to the CCoV references (chosen by BLAST analyses) in the case of the sample fox13 and sample fox113, respectively. Completion of the sequence gaps with direct sequencing resulted in a 28,855 (sample fox13) and a 28,955 (sample fox113) nt long, near-complete genome sequence that showed 95.1% genome-wide identity with each other. The sequence of the CCoV strain fox13 represented ≤94.3% overall nt identities with that of the closest related references, CCoV/GD/2020/X9 (GenBank acc. no. MZ320954, from a Chinese bamboo rat or Rhizomys sinensis, China, 2020), 61/22 (GenBank acc. no. OX335623, from a dog, UK, 2022), and CB/05 (GenBank acc. no. KP981644, from a dog, Italy, 2005). In case of CCoV strain fox113, this value was ≤94.1%, if compared to the sequence of the reference strains 11/22 (GenBank acc. no. OX335541, from a dog, UK, 2022), 61/22, and CCoV/GD/2020/X9. The 5’ and 3’ ends of both the CCoV fox13 and CCoV fox113 genomes were incomplete, with missing a 6 nt and an 84 nt long overhang, respectively, as revealed by comparisons to CCoV complete genome sequences (strains CB/05 and 11/22).

The structure of the genomic region 1ab of the CCoV fox13 and fox113 was comparable with that of the CCoV-II strains, including the closest related sequences of the strain 61/22 and VuCCoV_191134SY (GenBank acc. no. OQ540908, from a red fox, China, 2019). Considering the 1ab, the CCoV fox13 and fox113 represented 96.1% nt and 98% aa identity with each other, and ≤94.5% nt and ≤97.7% aa identity with the references (table in Figure 2).

The NTD of the two novel CCoV strains (669 nt and 223 aa for the strain fox13, 672 nt and 224 aa for the strain fox113) showed moderate, 78.1% nt and 73.9% aa identity with each other. This part of the fox13 CCoV genome represented ≤87.9% nt and 85.2% aa identity with the closest relatives, sequences gained from a Chinese bamboo rat (CCoV/GD/2020/X9), a spotted hyena (Crocuta Crocuta; strain SH36_2004, GenBank acc. no. MF095848, Tanzania, 2004) and a dog (strain B447_ZJ_2019, GenBank acc. no. MT114540, China, 2019). This group of sequences clustered with the CCoV-IIa type strains CB/05 and 450/70 (GenBank acc. no. GU146061, from a dog, Italy, 2007) (Figure 3). The nt and aa pairwise identity values were ≤87.8% and 84.8%, respectively, for the NTD of the CCoV strain fox113 if compared to the reference CCoV sequences identified in the samples of a dog (61/22) and a black-backed jackal (Canis mesomelas; GenBank acc. no. MF095855, Tanzania, 2011). Comparison of the aa alignment revealed the absence of a sialic acid binding domain in the NTD of CCoV fox13, fox 113, and the reference sequences clustering with these [23, 27].

The remaining, 1229 aa long part of the S (downstream of the NTD, referred here as CTD) of the CCoV fox13 and fox113 genomes showed 94.3% nt and 97.5% aa pairwise identity with each other, and even higher pairwise identities with the closest references (table in Figure 3). The high identities and close clustering were maintained through the 3a, 3b, 3c, and E regions, and the CCoV-IIa and CCoV-IIb strains were located on the same branch of the phylogenetic tree (represented by the tree E, Figure 2). ORF3, typical for CCoV-I strains, could not be identified in the CCoV fox13 and fox113 genomes. The relationships of the CCoV fox13 and fox113 strains varied for the M and N. Regarding M, the fox13 CCoV strain clustered with the dog origin strain B447_ZJ_2019 and red fox origin strain VuCCoV_191134SY, while CCoV fox113 was located in a common branch with TGEV, IIa, IIb, and jackal origin strains. In contrast, in the N-based phylogenetic tree, the CCoV fox13 clustered with IIa, IIb, and jackal origin strains, while the fox113 branched with the TGEV, the dog origin 61/22, UCD-1, and human origin CCoV-HuPn-2018 strains (Figure 3). Overall, despite the variable grouping, the genetic distances were low between the novel strains in every region, with the exception of the NTD.

The putative, nonstructural 7a and 7b protein encoding genomic region represented low variability. However, some strains did not have any, or had truncated 7b that may be nonfunctional, or encode proteins with altered functions. A 180 nt and a 183 nt long deletion (interrupted with a short sequence that matches with references) were recorded in the middle part of the fox113 CCoV 7b genomic region, resulting in a 92 aa long protein, if it is functional at all. In contrast, a 459 nt long deletion shortened the 7b of the CCoV fox113, thus the derived protein might be only 60 aa in length. Shortened 7b regions with internal deletion have also been noted for other CCoV-II strains (for example, the sequence with GenBank acc. no. EU856361, from a dog, Italy, 2005).

Unsurprisingly, recombination analysis of coronavirus near-complete genomic sequences resulted in a number of recombination events. A high probability event listed the CCoV fox13 and fox113, and some dog origin strains (including the pantropic CCoV-IIa reference strain CB05, strain NA/09 with GenBank acc. no. JF682842, and strains 11/22 and 61/22) as recombinants (Figure 4). The affected genomic region spanned the NTD, with possible recombination points in the 3’ end of the 1ab and in the receptor binding 5’ region of CTD. The event implicated the human origin strain CCoV-HuPn-2018 (acc. no. MW591993) as major parent and FeCoV-II as minor. The clustering with the minor parent could be followed through the phylogeny as well; within the NTD and CTD tree, the FeCoV-II (GenBank acc. no. DQ010921) clustered close with the above-mentioned recombinants and IIa strains, but the difference became more apparent in the downstream and upstream genomic regions. The strain CCoV-HuPn-2018 was located on a separate branch of the NTD tree with the CCoV-IIb sequences. It should be noted that parents are difficult to unravel due to the multiple recombination, and the predicted relationships change for the variable genomic regions (Figure 4). Accordingly, CCoV-HuPn-2018, which has also been characterized as a recombinant strain, was phylogenetically localized closer or further to the fox13 and fox113, depending on the investigated genome fragment (Figure 4) [44]. Phylogeny and pairwise comparison suggested the red fox origin strain VuCCoV_191134SY and the dog origin strain 61/22 as the closest relatives of both the CCoV fox13 and fox113 for the 1ab, which changed variably in the downstream regions. The phylogenetic analysis (Figures 2 and 3) was applied to complete ORFs, thus, it did not faithfully reflect the (multiple) recombinations. Recombination was also predicted within the 1ab of the fox13 and fox113 that was more pronounced in the second part of this ORF, affecting an approximately 5000 nt long fragment potentially obtained from the Chinese bamboo rat origin strain CCoV/GD/2020/X9 (Figure 4) [45]. Regarding the region downstream of the S to the 3’ end of the M, IIa and IIb strains might serve as minors for the novel genomes. Then, as the phylogenetic trees also showed, the two novel strains branched somewhat separately, and the probable genetic donors of this region varied, also depending on the sequences included in the analysis; clear relationships could not be established.

3.3. Picodicistrovirus

One contig, obtained by de novo assembly from the reads of the sample fox113, represented ≤87% nucleotide (nt) identity with CPDV strains according to BLAST search. The deeper bioinformatics analysis resulted in compilation of an 8793 nt long, near-complete genome sequence (mapping 1759 reads to it, average depth of 30) that showed the highest overall pairwise identity of 87.5% with a CPDV sequence identified in the feces of a common raccoon dog sampled in China, 2015 (strain RDX, GenBank acc. no. MT498598). The dicistronic organization of the novel sequence, together with the phylogenetic analyses, suggested that the generated sequence indeed derived from CPDV [28] (Figure 5).

Two main ORFs were predicted in the novel genome with the layout typical for the CPDVs [28]. The viral capsid polyprotein (P1) sequence showed 86.6% nt and 96.8% aa identity with that of the strain RDX, while the nt and aa identities were 66.8%–78.5% and 72.2%–93.3%, respectively, with the P1 sequence of other reference CPDV strains (Table 1, Figure 6). The inferred P1 could be cut for four proteins, but a notable difference is in the length of the VP4 of the novel sequence and the available reference sequences (44 aa) compared to the VP4 of RDX (81 aa) [29]. The nt, but not the aa pairwise identities showed significant variances both for P1, and for the other polyprotein, the P2-3 (Table 1): the nt identities remained below 90% for all of the predicted proteins, while the aa identity is ≥95%, except the 3A of the P2-3 (89.7%) (Table 1).

The intergenic region (IR) of the novel sequence, intervening the P1 and P2-3, is 2 nt longer (599 nt) than that of the strain RDX. The two sequences showed high, 95.3% nt identity with each other. The P2-3, which encodes the nonstructural proteins, represented 86.5% nt and 97.3% aa identity with that of the strain RDX, while the identity was 80.9%–83.9% at nt and 91.3%–95.9% at aa level with the P2-3 region of other CPDV strains (Table 1, Figure 6). The RNA-dependent RNA polymerase sequence, encoded in the 3D of the P2-3 region, represented the highest identity with that of the strain RDX (89.8% for nt and 98.3% for aa), as well as with two red fox origin partial CPDV sequences (85.6%–86.6% nt and 98.8%–98.9% aa identity) [10]. The complete genome sequence of the CPDVs identified in red fox in Australia (GenBank acc. no MT833876 and MT833880) has not been determined, thus, further comparisons cannot be executed. When considering the aa identities, 2C (containing NTP binding and helicase motifs) seemed to be a conserved region with uniformly high identities (≥97.4%) in all comparisons (Figure 6). Although the nt identities represented uniformity as well, these were below the value of the aa identities. The coloring in Figure 6 highlighted the overall, relatively large differences between the corresponding nt and aa identities.

In addition to the complete coding region, 924 nt of the 5’ end and 401 nt of the 3’ end of the fox113 genome were determined. Some dog origin reference sequences (GenBank acc. no. JN819202-JN819204) have a shorter 5’ noncoding region of 875–878 nt in length, and a 428–429 nt long 3’ region, if excluding the poly (A) tail. In contrast, the RDX genome has an even longer 5’ noncoding region (1063 nt), while its 3’ end has not been completely determined.

Recombination analysis of the available near-complete CPDV sequences indicated a probable recombination event among dog origin strains affecting almost the entire P1 (recombinant GenBank acc. no. OQ198171, major GenBank acc. no. OQ198169, minor GenBank acc. no. JN819203). Another event listed the novel CPDV fox113 as a recombinant, with dog origin sequences as the major and minor parents (Figure 7, predicted by 5 methods with p-values 10⁻⁷–10⁻¹⁹). The breakpoints were located at the very end of the 5’ noncoding region and in the conserved, central part of the IR, thus, the recombinant region spanned the whole P1. The recombination can be followed through the clustering in the phylogenetic trees and the identity values in the matrix as well (Figures 5 and 6).

3.4. CanineCV

Some short contigs, assembled from the sequence reads obtained from sample fox113, matched with fox circovirus (referred to thereafter as CanineCV) sequences. Remapping to the closest related sequence (strain 55590, GenBank acc. no. KP941114, from fecal sample of a red fox, Croatia, 2014) resulted in the determination of three genomic regions built from 124 reads [31]. Overlapping PCR amplicons were produced and directly sequenced to complement and confirm the metagenomics results. The length and the structure of the novel complete genome (GenBank acc. no. PQ177899) were comparable with those of the other CanineCVs. The 2056 nt long genome contained two main ORFs, one for the rep (912 nt) and one for the cp (813 nt), as well as a 3’ IR (203 nt) and a 5’ IR (128 nt). The typical nonanucleotide motif TAGTATTAC is located upstream of the rep in the 5’ IR.

The genome sequence of CanineCV fox113 shared the highest nt identity of 96.2% with that of the strain 55590. These two sequences are located on a common branch in the phylogenetic tree generated using the near-complete genome sequences of the available CanineCV sequences of wild carnivores (e.g., red fox, arctic fox, jackals, and wolves) and some representative strains identified in dogs (Figure 8). Further closely related sequences have been detected in red fox in Serbia and Canada, in golden jackal in Serbia, as well as in gray wolf in Canada (Figure 8).

The CanineCV strain 55590 was among the closest relatives of the CanineCV fox113 in any comparisons, but the rep and cp (nt and aa) based phylogenetic trees represented differences otherwise. Regarding the rep and the deduced protein sequences, the novel strain clustered with the same references as found in the whole genome-based tree. In contrast, the Serbian, red fox, and golden jackal origin strains are located in separate branches of the cp-based nt and aa trees (Figure 9).

Recombination analysis was carried out to reveal the evolutionary processes accompanied by the exchange of genome fragments among wild carnivore origin CanineCVs. The novel red fox origin strain was involved in recombination events primarily as a major or minor parent. Potential recombination (event no. 1, Figure 10) occurred among CanineCV sequences identified in golden jackal from Serbia and the CanineCV fox113, affecting a 1021 nt long region overlapping most of the rep, the 3’IR, to the 3’ part of the cp. Another recombination event (event no. 2, Figure 10) predicted the participation of the novel, red fox, and a black-backed jackal origin (from Namibia) sequences, resulting in a recombinant detected from golden jackal (from Serbia). The recombinant region was 1039 nt long, covering the complete 3’ IR and cp. The CanineCV fox113 was found a minor parent in a third event affecting a 1167 nt long region containing a long part of the rep (event no. 9, Figure 10), in which case golden jackal origin sequences were identified as recombinants, while the major was a strain collected from a red fox, Norway. These three events included exchange of sequences overlapping most of the rep or cp, thus could be followed through the clustering in the ORF-based phylogeny (Figure 9, strains labeled with blue, orange, and purple circles according to Figure 10).

Additional recombination events were also statistically supported, implying sequences identified in red fox (also the sample fox113), golden jackal, black-backed jackal, Italian wolf, European badger, and dog. The exception was the arctic fox origin CanineCVs, which were grouped separately in each case, as described earlier, with two strains found in gray wolf specimens collected in Canada [46]. Recombination was previously suggested among red fox and dog origin strains, which was confirmed here [34, 47].