Genetic characterization and structural insights of canine astroviruses circulating in South Korea

Abstract

Background

Canine astrovirus (CAstV), classified as Mamastrovirus 5, has been detected worldwide in dogs with gastrointestinal. Still, information on the genetic and structural properties of strains circulating in South Korea has remained limited. In this study, the molecular, structural, and antigenic characteristics of Korean CAstVs were investigated using full-length ORF2 sequences from 13 strains detected in dogs between 2022 and 2023.

Results

CAstV was detected in 16.5% (13/79) of intestinal or fecal samples submitted to the Animal and Plant Quarantine Agency for diagnostic testing (2022–2023). Phylogenetic analysis based on capsid amino acid sequences assigned all Korean strains to Mamastrovirus 5 and classified them into lineages 1, 2, and 4, with the majority of strains clustering within lineages 1 and 4. A putative recombinant strain, JBN/22D16-1, was identified by phylogenetic discordance between ORF1b and ORF2, as well as by recombination analysis, which revealed a group 2–like polymerase backbone and a lineage one capsid gene with a breakpoint near the ORF1b–ORF2 junction. Capsid structures were predicted with AlphaFold and subjected to all-atom quality assessment, which supported a conserved three-domain organization and stereochemically reliable models across all strains. Spike-domain comparisons revealed a conserved β-barrel core and β-strand topology, with most structural differences confined to peripheral β-elements and loop or turn segments. Shannon entropy analysis of the spike domain showed that high-variability sites were clustered in discrete regions within a broadly conserved scaffold. Predicted linear B-cell epitopes were mapped mainly to these variable, surface-exposed segments. They showed lineage-associated patterns, including a recurrent GQKSNSQY-containing central epitope in lineage four and a conserved C-terminal epitope in lineage 1.

Conclusions

Taken together, these findings provide new insight into the genetic, structural, and antigenic characteristics of CAstVs currently circulating in South Korea.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12917-026-05370-6.

Background

Astroviruses (AstVs) are small, non-enveloped, positive-sense single-stranded RNA viruses with a genome size ranging from approximately 6.4 to 7.7 kb [1, 2]. First identified in 1975 from the feces of children with diarrhea, AstVs have since been detected in over 80 species of mammals and birds, underlining their broad host range and zoonotic potential [3]. The Astroviridae family is divided into two genera: Mamastrovirus, which infects mammals, and Avastrovirus, which infects birds [4, 5]. In mammals, AstVs are primarily associated with asymptomatic infectious or mild gastroenteritis. In contrast, in avian species, they are linked to a broader spectrum of pathologies, including enteritis, hepatitis, and nephritis [6]. The genome typically encodes three open reading frames (ORFs): ORF1a and ORF1b, which produce non-structural proteins, including the RNA-dependent RNA polymerase (RdRp), and ORF2, which encodes the capsid protein that mediates host interaction and determines antigenicity [7].

Canine astrovirus (CAstV), classified as Mamastrovirus 5, was first identified by electron microscopy in the feces of diarrheic puppies in the early 1980s. Subsequent molecular studies have detected CAstV in North America, Europe, Asia, and South America, indicating a global distribution [8]. Several studies have reported a higher detection rate of CAstV in dogs presenting with gastroenteritis compared to asymptomatic individuals, suggesting a potential association between CAstV infection and canine diarrhea [1, 9, 10]. However, co-infections with other enteric viruses, including canine parvovirus (CPV), canine coronavirus (CCoV), and canine distemper virus (CDV), are frequently observed [1, 8, 10, 11]. Thus, the specific pathogenic role of CAstV in canine gastrointestinal disease remains to be elucidated.

Although multiple lineages of CAstV have been reported worldwide, data on their epidemiology (including detection in reported cases) and genetic diversity in South Korea remain limited. A previous study conducted in Seoul and Gyeonggi-do identified a low detection rate (~ 2.1%) in canine fecal samples, but phylogenetic analysis revealed close clustering with European CAstVs [12]. These CAstV strains were tentatively classified as belonging to Mamastrovirus 5; however, the sequence obtained was limited to a short fragment (~ 230 bp) of the ORF2 region, hindering comprehensive genetic characterization. Recently, a surveillance study in South Korea identified CAstV-related sequences in raccoon dogs, raising the possibility of interspecies transmission and the presence of wildlife reservoirs [13]. Furthermore, CAstV has been identified in China, a neighboring country, where substantial genetic variation and recombination among strains have been reported [1]. To date, no study has sequenced the full-length ORF2, evaluated structural conservation, or investigated antigenic determinants of CAstV strains circulating in Korean dog populations. Here, we address these gaps by performing an integrated genomic and structural analysis of CAstV strains collected from dogs in South Korea between 2022 and 2023, thereby providing novel insights into CAstV diversity in South Korea.

Materials and methods

Sample collection

Intestinal or fecal samples were obtained from dog carcasses submitted to the Animal and Plant Quarantine Agency (APQA) for post-mortem diagnostic examination from multiple regions between 2022 and 2023. In total, 79 canine samples were tested for CAstV.

Viral RNA extraction and cDNA synthesis

RNA was extracted using a hybrid protocol [14] of TRIzol reagent (Thermo Fisher Scientific, Massachusetts, United States) and the WizPrep™ Viral DNA/RNA Mini Kit (Wizbiosolutions, Korea), and finally eluted in 30 µL of elution buffer. Viral RNA was denatured at 65 °C for 3 min, followed by cDNA synthesis using random hexamers and oligo dT primers with the AccuPower^® RocketScript™ RT Master Mix (RNase H Minus) from Bioneer, South Korea.

CAstV detection and full-length capsid amplification

Pan-AstV screening was performed using a broadly reactive primer set targeting a 422-bp fragment of the RdRp gene, as described in a previous study [15]. PCR products were purified using the PCR Clean-up kit (Cosmogenetech, Korea) and subjected to Sanger sequencing through a commercial sequencing service (Macrogen, Korea).

To amplify the complete capsid gene (ORF2), a gene-specific primer (GSP) within the partial RdRp region was paired with a reverse primer targeting the conserved stem-loop 2-like motif (s2m), enabling amplification of the full-length approximately 2,700 bp ORF2 sequence. PCR cycling conditions consisted of an initial denaturation step at 94 °C for 5 min, followed by 35 cycles of denaturation at 94 °C for 1 min, annealing at 57 °C for 1 min, and extension at 72 °C for 3 min, and a final extension step at 72 °C for 10 min, followed by cooling at 8 °C. Amplicons were purified using the same clean-up kit and sequenced using the BIT Seq Standard service (Bionics, Korea). Details of all primers used in this study are provided in Table 1.

Table 1.

Primer information used in this study

Primer name		Position	Sequence		Band Size (bp)	Reference
Astrovirus screening	AS1-F1	RdRp	F1	GAR TTY GAT TGG RCK CGK TAY GA	422	[15]
	AS2-F1			GAR TTY GAT TGG RCK AGG TAY GA
	AS-R		R	GGY TTK ACC CAC ATN CCR AA
	AS3-F1		F2	CGK TAY GAT GGK ACK ATH CC
	AS4-F2			AGG TAY GAT GGK ACK ATH CC
ORF2	CAstV 23Q45	RdRp	F	TATCAAGAAACTCCGCTGG	3000	Used in this study
	CAstV 22D16-1			TATCAAGAAACTCCGTTGG
	CAstV 23D104			TATTAAAAAACTCCGCTGG
	CAstV 22Q14			TATCAAGAAATTACGCTGG
	CAstV 22D17-1			TATCAAGAAACTACGCTGG
	s2m_rev	ORF2	R	CCC TCG ATC CTA CTC GG		[16]
ORF1ab	CaAstV 5’ consensus	ORF1a	F	CTCTAGTATGTGTTCAGTGCC	4000	Used in this study

Sequences data of CAstVs

Reference sequences representing diverse Mamastrovirus strains were retrieved from the NCBI GenBank database (https://www.ncbi.nlm.nih.gov/genbank/). Multiple sequence alignments (MSA) were performed using MAFFT (v7.526). All CAstV sequences obtained in this study have been submitted to the NCBI GenBank database. The details of CAstV strains used in the current study are provided in Table S1.

Genetic analysis of CAstVs

Phylogenetic analysis was conducted based on amino acid (AA) sequences of the RdRp and the complete ORF2 genes. Maximum-likelihood (ML) trees were generated with IQ-TREE (v 2.3.6), with the best-fit substitution model selected automatically by ModelFinder. Node support was assessed using 1,000 bootstrap replicates.

Pairwise AA identity analysis was performed to assess the genetic relatedness among CAstV strains. Pairwise identity values were calculated from the resulting MSA. Identity was defined as the proportion of identical residues at aligned positions, excluding sites with gaps only. A custom Python3 (v 3.10.12) script, incorporating BioPython (v 1.85) [17] and NumPy (v 2.2.4) [18], was created to automate the generation of the complete identity matrix. To quantify dissimilarity, identity values (%) were converted into distances using the formula: distance = 100 – identity.

To investigate potential recombination events of CAstV strains detected in South Korea, a comprehensive analysis was performed using their nucleotide sequences with the Recombination Detection Program (RDP) 4.0 [19]. Recombination analysis was conducted using a multi-method approach implemented within RDP4, incorporating RDP, GENECONV, BootScan, MaxChi, Chimaera, SiScan, and 3Seq algorithms. A p-value threshold of 0.05 was applied, and only events supported by at least six independent methods were considered statistically significant recombinant. Recombination breakpoints were identified and mapped across the ORF2 region, and putative parental strains were inferred based on the highest sequence similarity. A similarity plot was generated using SimPlot v.3.5.1 to identify potential recombination breakpoints, with a sliding window and step sizes of 200 base pairs (bp) and 20 bp, respectively.

Protein structure prediction

Full-length ORF2 AA sequences from 13 CAstV strains (2022–2023) collected in this study (Table S1) were analyzed. ORF2 lengths ranged from 825 to 842 residues after signal peptide removal. Full-length monomeric models were predicted using the AlphaFold 3 server [20]. For each strain, we retained the single top-ranked structural model and extracted the predicted local-distance difference test (pLDDT), predicted Template Modeling score (pTM), ranking score, the number of recycles used during model refinement, and the expected fraction of intrinsically disordered residues from the corresponding JSON outputs (Table S4). pLDDT is reported on a 0-100 scale, indicating that a higher value represents greater local confidence.

To prepare the AlphaFold models for downstream coordinate curation and all-atom validation, we first converted the mmCIF outputs to PDB format using iotbx. Hydrogen atoms were added and standardized with phenix. Reduce before PDB tidying with phenix.pdbtools [21]. All-atom stereochemical quality was then assessed in Phenix v1.21.2–5419 using the Phenix software. molprobity, phenix.clashscore, phenix. ramalyze, and phenix. rotalyze to obtain the Molprobity score, clashscore (severe steric clashes per 1,000 atoms), Ramachandran favored/allowed/outlier percentages, and rotamer outlier percentages [22, 23] (Table S5).

Core, spike, and acidic tail domains were identified using uniform rule-based thresholds. The core-spike boundary was defined as the first residue after a local pLDDT minimum for which at least 10 consecutive residues had a pLDDT value ≥ 55 (conservatively ≥ 60). The spike-acidic tail boundary was determined as the earliest 10–15 residue window in which pLDDT values were < 60 and coincided with acidic enrichment, defined as a combined aspartate/glutamate (D + E) fraction > 0.20. Residue indices were treated as 1-based, and numerical summaries of all boundary assignments are presented in Table S4.

Visualization of spike domain

Thirteen spike domain PDB sets of CAstV strains were subjected to downstream structural analysis. PDB files were parsed programmatically using the Biopython PDB module [17, 24]. Secondary structure annotation was obtained by running mkdssp, which is the modern implementation of the Define Secondary Structure of Proteins (DSSP) algorithm, on the final spike domain PDB files [25]. Hydrogen bond-based DSSP assignments were used in their standard eight-state form, and continuous stretches of residues annotated with the ‘E’ state were operationally defined as β-strands. In contrast, positions annotated with the ‘T’ state were classified as turns.

The spike domain structure of JBN/15 − 6 was designated as the reference template for all comparative analyses. All remaining target structures were rigid-body superimposed onto this reference using the cealign algorithm implemented in PyMOL v3.2.0a0. A standard camera matrix was obtained from the reference structure using the get_view command and applied to each aligned structure via set_view, thereby fixing the viewing orientation, zoom level, and slab clipping parameters across all models.

Variability analysis of spike domain and B-cell antigenic epitope prediction

AA sequence variability of spike domain among thirteen CAstV strains was quantified using the Shannon entropy (H) method on the Protein Variability Server (PVS) [26]. Briefly, the AA sequences of the spike domain were aligned by MAFFT, and the alignments were input into PVS by “first sequence in alignment” mode.

Linear B-cell epitopes were predicted for the thirteen CAstV spike domain AA sequences using the SVMTriP, a machine learning-based tool for epitope prediction based on a support-vector-machine (SVM) model [27]. The epitope length was fixed at 20 AA; all remaining parameters were kept at their default values. Only peptides flagged as “recommended” by SVMTriP were retained as candidate linear B-cell epitopes.

Results

Detection of CAstV

Between 2022 and 2023, CAstV was detected in 13 of 79 intestinal or fecal samples (16.5%) obtained during post-mortem examination of dog carcasses submitted to APQA from multiple regions in South Korea (Table 2). Positive samples were identified from Gyeonggi-do (10/13; Goyang-si and Yangpyeong-gun), Chungcheongbuk-do (Eumseong-gun, 1/13), Daegu (1/13), and Daejeon (1/13). Among the CAstV-positive samples, eight were classified as companion dogs and five as free-roaming dogs. Clinical signs were not consistently available through cases; however, enteritis/diarrhea-related findings were recorded in three companion dog cases (JBN/22Q14, JBN/22D15-6 and JBN/22D127-1).

Table 2.

Summary of CAstV-positive samples and CAstV positivity rate

Sample No. (Sample type)	Region	Medical history	Year	Dog category (Clinical sign)*	CAstV-positive rate
JBN/22Q14 (intestine)	Gyeonggi-do (Goyang-si)	CPV, CDV	2022	Companion dog (Enteritis, Diarrhea)	16.5% (13 / 79)
JBN/22D15-6 (intestine)	Gyeonggi-do (Goyang-si)	CPV, CDV	2022	Companion dog (Enteritis, Diarrhea)
JBN/22D16-1 (intestine)	Gyeonggi-do (Goyang-si)	CPV, CCoV	2022	Companion dog
JBN/22D17-1 (intestine)	Gyeonggi-do (Goyang-si)	CPV, CAV, CCoV	2022	Companion dog
JBN/22Q69-2 (intestine)	Chungcheongbuk-do (Eumseong-gun)	CPV	2022	Free-roaming dog
JBN/22D67 (fecal)	Gyeonggi-do (Goyang-si)	CPV, CCoV, CDV	2022	Companion dog
JBN/22D127-1 (intestine)	Gyeonggi-do (Goyang-si)	CPV	2022	Companion dog (Enteritis, Diarrhea)
JBN/23Q45 (fecal)	Gyeonggi-do (Yangpyeong-gun)	CPV, CCoV	2023	Free-roaming dog
JBN/23Q47 (fecal)	Gyeonggi-do (Yangpyeong-gun)	CPV, CCoV	2023	Free-roaming dog
JBN/23Q48 (fecal)	Gyeonggi-do (Yangpyeong-gun)	CPV, CCoV	2023	Free-roaming dog
JBN/23Q50 (fecal)	Gyeonggi-do (Yangpyeong-gun)	CPV, CCoV	2023	Free-roaming dog
JBN/23D80 (fecal)	Daegu (Metropolitan City)	CPV, CCoV	2023	Companion dog
JBN/23D104 (fecal)	Daejeon (Metropolitan City)	CCoV	2023	Companion dog

^*Only gastrointestinal symptom–related findings were recorded under “Clinical sign.”

Among the CAstV-positive samples, co-detection with other viral pathogens was frequently recorded, most commonly involving CPV and/or CCoV (Table 2). The most frequent co-detection pattern observed was CAstV in combination with both CPV and CCoV, detected in eight samples (JBN/22D16-1, JBN/22D17- 1, JBN/22D67, JBN/23Q45, JBN/23Q47, JBN/23Q48, JBN/23Q50, and JBN/23D80). Notably, JBN/22D17-1 tested positive for CPV, Canine Adenovirus (CAV), and CCoV, representing a triple co-detection. Similarly, JBN/22D67 harbored CPV, CCoV, and CDV. In the remaining samples, CPV alone was detected in two samples (JBN/22Q69-2 and JBN/22D127-1), CPV with CDV was observed in two samples (JBN/22Q14 and JBN/22D15-6), and CCoV alone was detected in one sample (JBN/23D104).

Phylogenetic analysis of CAstVs

Phylogenetic and pairwise AA identity analyses were performed based on the RdRp partial AA sequences and capsid protein AA sequences. All the CAstV strains belonged to MAstV 5 and were classified into four distinct lineages (Lineage 1–4) (Fig. 1a), based on phylogenetic relationships and previously established CAstV classifications (1). Phylogenetic analysis of the RdRp partial AA sequences exhibited two different groups, provisionally designated group 1 (G1) and group 2 (G2) (Fig. 1b). Within-group similarity was high, with G1 strains showing more than 97.51% AA identity. In G2 strains, all AA sequences of RdRp exhibited at least 98.35% identity, except for one CAstV strain displaying a minimum identity of 94.21% (Table S2). The amino acid sequence similarity between these two groups was approximately 82.99–86.36% in the RdRp partial sequence. Taken together, phylogenetic analyses of RdRp and capsid AA sequences indicate that lineage 1 belongs to G1, while lineages 2–4 correspond to G2 (Fig. 1a and b). Lineage 2 and lineage 4 exhibited high intra-lineage conservation, with mean capsid AA sequence identities of 98.04% and 97.92%, respectively (Table S3). In contrast, lineages 1 and 3 exhibited more heterogeneity, with mean pairwise identities of 93.93% and 90.67%, respectively (Table S3).

Fig. 1.

Phylogenetic relationships of Canine Astrovirus (CAstV) strains based on amino acid sequences of (A) the capsid protein and (B) RNA-dependent RNA polymerase (RdRp). Strains identified in this study are indicated by yellow stars. Colored circles denote the country of origin. Lineages are distinguished by color: Lineage 1 (purple), Lineage 2 (orange), Lineage 3 (sky blue), and Lineage 4 (green). Phylogenetic trees were constructed using the maximum likelihood method; branch support was assessed with 1,000 bootstrap replicates. Scale bars represent amino acid substitutions per site

The complete capsid AA sequences of JBN strains were classified into distinct lineages 1, 2, and 4, with the majority clustering within lineages 1 and 4 (Fig. 1a). JBN/23Q45, JBN/23Q47, JBN/23Q48, JBN/23Q50, JBN/23D104, and JBN/22D15-6 strains were grouped into lineage 4, along with previously reported CAstV strains from Hungary, China, and Vietnam. In the case of JBN/22D127-1 and JBN/22D67 strains, they belonged to lineage two clustering with CAstV strains from China, and exhibited high AA sequence conservation. Strains JBN/22Q14, JBN/22Q69-2, JBN/22D16-1, and JBN/22D17-1 clustered within lineage 1, alongside previously reported CAstV strains from China and Hungary. Despite high conservation in RdRp AA sequences between lineage 1 CAstV and raccoon dog AstV (RAstV) strains (> 97.51% identity; Table S2), the capsid protein AA sequences exhibited higher divergence, with pairwise identities falling to as low as 80.47% (Table S3).

Recombination analysis of CAstVs

Based on the phylogenetic analysis, the JBN/22D16-1 strain clustered within G2 in the RdRp-based phylogenetic tree, but it grouped with G1 in the ORF2-based tree, suggesting a potential recombinant. The JBN/22D16-1 strain was selected as a potential recombinant based on RDP4 and a similarity plot analysis, in agreement with phylogenetic inference. The JBN/22Q69-2 strain was the major parent of JBN/22D16-1, and the B194 strain was the minor parent. The recombination breakpoint was located at approximately 730 bp, corresponding to the considered start site of ORF2 (Fig. 2).

Fig. 2.

Recombination analyses of CAstV strains JBN/22D16-1. The x-axis indicates the genomic position, and the y-axis exhibits the similarity between the major parent (blue) and minor parent (red). Recombinant signals were identified by RDP4 and visualized using SimPlot. This plot shows the relative similarity of the query sequence (JBN/22D16-1) to the inferred major and minor parental strains along the partial ORF1b (RdRp)–ORF2 alignment, using a sliding window of 200 bp with a 20 bp step (Kimura 2-parameter model)

Predicted capsid protein structure and confidence assessment

All capsid protein models exhibited high stereochmical quality as evaluated by MolProbity all-atom validation. Across the JBN-CAstV strains, MolProbity scores averaged 1.97 ± 0.11 (range 1.81–2.13), with clash scores of 6.23 ± 0.80, Ramachandran favored residues of 91.41 ± 0.80% (outliers 3.20 ± 0.78%), and rotamer outliers of 1.46 ± 0.52% (Table S4). These values fall within the range typically observed for well-refined experimental structures, indicating that the predicted capsid protein of the JBN-CAstV strain models is stereochemically reliable and suitable for quantitative domain- and residue-level analyses.

Having established model quality, we next examined domain-resolved confidence patterns in the predicted capsid. AlphaFold-derived pLDDT values supported a conserved three-domain organization across the thirteen JBN-CAstV strains. The domain-mean pLDDT values followed the hierarchy spike, core, tail, with mean values of approximately 81, 78, and 64, respectively (Table S4). Despite modest variation in absolute score among JBN-CAstV strains, the position of domain boundaries was highly reproducible. The core starts between residues 111 and 123, spike starts between residues 483 and 486, whereas acidic tail onsets are mapped predominantly to residues 715–721, with only rare extension to 730–738 (Table S5). Residue-wise pLDDT profiles showed a consistent local minimum at the core-spike junction followed by recovery to ≥ 55–60 over the next 10–20 residues, and a subsequent decline below < 60 at acidic tail initiation, coincident with enrichment of acidic residues (Fig. 3).

Fig. 3.

Per-residue pLDDT profiles of CAstV capsid proteins. Line plots show the per-residue pLDDT confidence scores of full-length capsid protein from 13 JBN-CAstV strains. Each curve represents a single strain, with higher values indicating greater structural confidence. Vertical dotted lines mark the approximate boundaries of the spike and tail domains inferred from AlphaFold predictions and sequence annotations

Topology metrics further indicated a conserved capsid fold. The pTM values were narrowly distributed (0.49 ± 0.01), suggesting that overall capsid topology is preserved despite sequence divergence. In contrast, the fraction of residues belonging to low-confidence/disordered regions varied modestly (range 0.22–0.24) and showed no clear relationship with pTM (Table S5).

Structure analysis of CAstV spike protein

Rigid-body alignment of the JBN-CAstV strains spike domain models to a single reference (JBN/22D15-6) under an identical view matrix revealed a compact β-barrel core that overlaid closely across all JBN-CAstV strains. The β-barrel core overlapped with minimal deviation after rigid-body alignment, indicating strong structural conservation of the spike domain scaffold across JBN-CAstV strains. In contrast, peripheral β-elements (top-β-hairpins and side-sheets) varied markedly in length, radial displacement, and parallel packing between strains. A deliberately strict, rule-based classification scheme was used, and strands that did not meet these unambiguous criteria were assigned to an “other” category, thereby reducing the risk of over-interpretation of marginal or very short segments (Fig. 4).

Fig. 4.

Structural visualization of the spike protein of Korean CAstV strains. (A), (B) and (C) correspond to lineage 1, 2 and 4 strains, respectively. β-sheet elements are shown with the following colors: β-barrels, red; top β-hairpin, blue; top-β-hairpin, purple; side sheets, green; others, yellow

DSSP-derived turn assignments were collapsed into single “TT” markers per continuous run to simplify visualization. TT markers were concentrated in loop segments that cap or flank the β-barrel core and decorate the solvent-exposed spike surface. In all JBN-CAstV strains, the spike domain was consistently maintained at approximately 230 residues in length, and the number of β-strands generally ranged from 14 to 18. Within this conserved structural framework, the peripheral strand categories (top-β-hairpins and side-sheets) accounted for most of the structural differences among JBN-CAstV strains.

When JBN-CAstV strain spike domain structures were aligned according to lineage, a high degree of inter-lineage structural similarity was observed. The β-barrel core overlapped with minimal deviation within each lineage, and both the number and ordering of DSSP-defined β-strands were almost invariant. Peripheral β-elements and TT markers also exhibited lineage-consistent patterns in terms of length and position.

Variability analysis and B-cell antigenic epitope prediction

The multiple sequence alignment of the JBN-CAstV strain spike domains comprised approximately 253 AA sequences. The site-wise Shannon entropy analysis showed that sequence variability was irregularly distributed across the spike domain, with a mean entropy of 0.62 and values ranging from 0 (fully conserved) to 1.91 (highly variable; Fig. 5). Of the 253 alignment sequences, 115 sites (45.5%) were classified as highly conserved (H ≤ 0.4) among the JBN-CAstV strains. In contrast, 43 sites (17.0%) exhibited high entropy (H ≥ 1.2), indicating distinct mutational hotspots embedded within a conserved scaffold. The remaining 95 sites (37.5%) were classified as intermediately variable (0.4 < H < 1.2). Overall, conserved sites were scattered throughout the spike domain, whereas high-entropy residues were more frequently observed in the N-terminal (approximately residues 1–70) and C-terminal (approximately residues 190–253) regions.

Fig. 5.

Shannon entropy analysis of the Korean CAstV strains spike domain. Shannon entropy values were calculated from the aligned spike-domain amino acid sequences of 13 JBN-CAstV strains. The plot displays residue-wise variability across the spike domain, with higher entropy values indicating more variable sites

The predicted epitopes consistently focused on three distinct regions of the spike domain: (i) an N-terminal region, (ii) a central region comprising two partially overlapping clusters (75–95 and 111–130), and (iii) a C-terminal region. These epitope regions were repeatedly identified across the JBN-CAstV strains, with SVMTriP scores ranging from 0.88 to 1.00, indicating high prediction confidence (Table 3).

Table 3.

Predicted linear B-cell epitopes in the spike domain of Korean CAstV strains

CAstV strains	Lineage	Location	Epitope	Score
JBN/22D16-1	Lineage 1	198–217	AAQNIFVCVGSNTDSIMSVI	1.00
JBN/22D17-1	Lineage 1	198–217	AAQNIFVCVGSNTDSIMSVI	1.00
JBN/22Q69-2	Lineage 1	198–217	AAQNIFVCVGSNTDSIMSVI	1.00
JBN/22Q14	Lineage 1	199–218	AAQNIFVCVGSNTDSIMSVI	1.00
JBN/23D80	Lineage 1	199–218	AAQNIFVCVGSNTDSVMSVI	1.00
		48–67	DEVNFFFKAAHPHAWLTGSA	0.88
JBN/22D67	Lineage 2	76–95	PTAVTFMEIGQKSNSQYLRR	1.00
JBN/22D127-1	Lineage 2	194–213	AAAQNLLVCTSSTVSSTMNI	1.00
		35–54	KVVAKVGQHYDEINFIYRAT	0.90
JBN/22D15-6	Lineage 4	75–94	PSALDFMEVGQKSNSQYLTR	1.00
		111–130	QSVLTFYVGQLSRQIRYSNY	0.88
JBN/23Q45	Lineage 4	75–94	PSALDFMEVGQKSNSQYLTR	1.00
		30–49	YLYHSYNKVVVKIGQHYDEI	0.96
JBN/23Q47	Lineage 4	75–94	PSALDFMEVGQKSNSQYLTR	1.00
		111–130	QSVLTFYVGQLSRQIRYSNY	0.88
JBN/23Q48	Lineage 4	75–94	PSALDFMEVGQKSNSQYLTR	1.00
		111–130	QSVLTFYVGQLSRQIRYSNY	0.88
JBN/23Q50	Lineage 4	75–94	PSALDFMEVGQKSNSQYLTR	1.00
		111–130	QSVLTFYVGQLSRQIRYSNY	0.88
JBN/23D104	Lineage 4	75–94	PSALNFMEVGQKSNSQYLTR	1.00
		30–49	YLYHSYNKVVVKIGEHYDEI	0.92

Putative N-terminal linear B-cell epitope candidates were predicted in a few JBN-CAstV strains. In lineage 1 strain, JBN/23D80, a predicted epitope spanned residues 48–67 (DEVNFFFKAAHPHAWLTGSA). In lineage 2, JBN/22D127-1 harbored an N-terminal candidate at residues 35–54 (KVVAKVGQHYDEINFIYRAT). In lineage 4, two strains (JBN/23Q45 and JBN/23D104) showed predicted epitope candidates at residues 30–49 (YLYHSYNKVVVKIGQHYDEI) and 30–49 (YLYHSYNKVVVKIGEHYDEI), respectively (Table 3).

A central region around residues 75–95 was repeatedly highlighted (Table 3). In lineage 4, five JBN-CAstV strains (JBN/22D15-6, JBN/23Q45, JBN/23Q47, JBN/23Q48, and JBN/23Q50) shared an identical predicted epitope, PSALDFMEVGQKSNSQYLTR (75–94), whereas JBN/23D104 harbored a closely related variant, PSALNFMEVGQKSNSQYLTR (75–94). In addition, the lineage 2 strain JBN/22D67 showed a related epitope at positions 76–95 (PTAVTFMEIGQKSNSQYLRR), differing at several residues but retaining the central GQKSNSQY motif. Collectively, these findings suggest that a GQKSNSQY-centered segment is a recurrently predicted linear epitope candidate among the Korean CAstVs, particularly within lineage 4.

A second central region at residues 111–130 was predicted as a putative linear B-cell epitope candidate in several lineage 4 JBN-CAstV strains. Specifically, JBN/22D15-6 exhibited a predicted epitope at residues 111–130 (QSVLTFYVGQLSRQIRYSNY), and an identical sequence was also predicted in three additional lineage 4 strains (JBN/23Q47, JBN/23Q48, and JBN/23Q50) (Table 3). Collectively, these results indicate that this region represents a recurrently predicted linear epitope candidate within lineage 4.

C-terminal putative linear B-cell epitope candidates were consistently predicted between residues 198 and 218 in all five lineage 1 JBN-CAstV strains (Table 3). Four lineage 1 strains shared an identical sequence (AAQNIFVCVGSNTDSIMSVI), whereas JBN/23D80 showed a closely related variant (AAQNIFVCVGSNTDSVMSVI). In lineage 2, a related candidate was identified in JBN/22D127-1 (residues 194–213; AAAQNLLVCTSSTVSSTMNI), which retained a partially conserved AAQ-(…)-VCT motif but differed substantially in the surrounding residues. Collectively, these results suggest that the C-terminal region harbors a lineage 1-conserved putative linear epitope candidate, with a motif-related variant observed in one lineage 2 strain.

Discussion

In this study, the molecular, structural, and antigenic characteristics of CAstVs circulating in South Korea were comprehensively investigated. Full-length ORF2 sequences from 13 JBN-CAstv strains were analysed together with recombination analysis, predicted spike protein structures, and variability patterns. Through this integrated approach, an overview of CAstV diversity and its structural characteristics has been provided. However, the samples were obtained through post-mortem diagnostic submissions rather than population-based surveillance, and the observed detection rate should not be interpreted as prevalence in the general dog population. Nevertheless, an integrated genomic and structural analysis was performed, and a baseline description of CAstV diversity detected in South Korea was provided, which may help to inform future surveillance efforts and functional studies.

Co-infections with other viruses in the presence of CAstV have previously been reported to be associated with an increased risk of more severe gastrointestinal disease [28, 29]. Co-detection with CPV was observed more frequently in Table 2, and it has previously been reported that concurrent infection with CAstV and CPV can be associated with hyperhemorrhagic diarrhea, particularly in cases involving CPV-2 variants 2b and 2c [30]. CPV infection has been shown to preferentially target rapidly dividing epithelial cells in the intestinal crypts, and disruption of the intestinal mucosal barrier with villous atrophy and subsequent malabsorption has been described [31, 32]. In parallel, involvement of lymphoid tissues and bone marrow precursor cells has been associated with marked leukopenia and broader immune perturbation in affected dogs [31]. Accordingly, in CPV-positive samples, compromised barrier integrity and/or altered immune status could potentially influence the detection patterns of concurrent enteric viruses, including CAstV. However, further studies are needed to determine this hypothesis using quantitative approaches (e.g., viral load measurements) and to directly evaluate virus–virus and virus–host interactions. Moreover, co-detection with CCoV and CDV was also detected in CAstV-positive samples (Table 2). Co-detection of CAstV with CCoV (often together with other major enteric viruses) has been frequently reported in dogs with gastroenteritis, including a study in which most CaAstV-positive samples were co-infected with CCoV [11]. However, recent molecular surveys have also described numerous co-detection/co-infection events in which no consistent additional clinical signs were observed [29, 33, 34]. Likewise, CCoV was not detected in enteritis/diarrhea-related samples that were positive for CAstV in the present study. Co-detection of CDV also was detected in CAstV-positive samples in the present study (Table 2). Similar multi-agent enteric detections that include CDV have been reported in diarrheic dogs, including studies in which CAstV-positive samples frequently carried concurrent CCoV and CPV-2 together with CDV [11, 35]. CDV infection is characterized by lymphoid tropism and can be associated with leukocyte depletion/lymphopenia and broader immune suppression, the presence of CDV could potentially increase the likelihood that additional pathogens are detected concurrently [36, 37].

In the current study, evolutionary relationships among JBN-CAstV strains were primarily interpreted based on ORF2-encoded capsid AA sequences. This approach is supported by the current astrovirus classification system, which is based primarily on the capsid protein for species and variant classification [38]. As emphasized by the ICTV Astroviridae Study Group and subsequent reviews, early classification schemes based on virion morphology and host species proved inadequate once sequence data accumulated: genetically similar astroviruses were frequently detected in different host species, whereas substantial genetic diversity was revealed among viruses infecting a single host [38, 39]. To resolve this, a shift toward capsid-based criteria was recommended: strains belonging to the same astrovirus species are expected to share > 75% AA identity across ORF2, whereas more divergent capsid sequences are assigned to different species, and additional “variant” levels have been proposed when capsid-based distances exceed defined thresholds [38].

In the ORF2-based phylogeny, the JBN-CAstV strains were distributed in three lineages of MAstV5—lineages 1, 2, and 4—with most strains clustering within lineages 1 and 4. This lineage distribution is broadly consistent with previous studies, which have identified four major ORF2 AA lineages; however, the relative contribution of each lineage has been shown to vary by region [1, 29, 40]. In China, most contemporary strains have been reported in lineages 2–4, with lineage one detected only rarely or not at all. In contrast, in Ecuador, viruses belonging to lineages 1, 2, and 4, as well as a putative Thai-derived lineage, have been detected in dogs with gastroenteritis [1, 29]. The finding that JBN-strains are mainly distributed in lineages 1 and 4, therefore, suggests a regional pattern in which multiple globally identified lineages co-circulate but with distinct local dominance. Furthermore, for CAstV lineage 1, although the Korean RAstV strains clustered, this cluster was separated from CAstV lineage 1 strains by long branch lengths and by approximately 83% amino acid similarity in the ORF2 AA sequence (Table S3). A previous study of RAstV in Korea reported a similar pattern, with raccoon dog MAstV5 strains showing about 75–84% ORF2 AA identity to CAstV capsids, high conservation in the core domain, and higher variation in the spike domain [13]. Taken together, these findings were considered to indicate that the raccoon dog viruses are a raccoon dog–associated MAstV5 cluster most closely related to CAstV lineage 1, rather than belonging to the established canine lineages. Although no JBN-CAstV strain belonged to CAstV lineage 3 in this study, the reference sequences classified as lineage 3 were more heterogeneous than those in the other lineages. In the ORF2-based phylogenetic tree, CAstV lineage 3 strains were split into several smaller clusters, and the ORF2 AA similarity between these clusters was lower than that observed within lineages 1, 2, and 4 (Fig. 1 and Table S3). Two Korean CAstV strains reported in 2010 [12], for which only partial ORF2 AA sequences are available, were located in CAstV lineage 3. Because only partial capsid sequences were reported, the exact position of these early Korean strains and their relationship to other lineage 3 strains cannot be defined clearly. Taken together, these findings suggest that CAstV lineage 3 may contain several sub-lineages rather than a single uniform group, and that additional full-length ORF2 and whole-genome sequences will be required before any revision of the current CAstV lineage 3 classification is considered.

Interestingly, the JBN22D16-1 strain shared high AA similarity in the RdRp region with lineage 2,3, and 4 CAstV strains and clustered together (G2) with them in the ORF1b-based phylogeny. However, this strain did not belong to the same lineage (2, 3, or 4) when ORF2 AA sequences were analysed, but belongs to lineage 1 (G1 based on ORF1b-AA-phylogeny; Fig. 1). This discordance between RdRp- and capsid-based phylogenies was likewise evident in the recombination analysis (Fig. 2). This discordance is consistent with the evolutionary patterns of mamastroviruses, where repeated recombination near the ORF1b–ORF2 junction produces viruses that retain a similar polymerase backbone while acquiring capsid genes from different viruses [41–43]. Although complete genome sequences would be required to confirm such events in JBN/22D16-1 definitively, the pattern observed in this study is consistent with a modular mode of mamastrovirus evolution. It supports the use of ORF2 capsid sequences as the primary criterion for routine lineage assignment. These patterns are compatible with potential cross-species transmission, although these hypotheses have not yet been experimentally evaluated [2, 44]. Taken together, these observations support the idea that ORF1b mainly tracks deeper evolutionary relationships, whereas ORF2 more readily reflects the faster-changing antigenic and ecological features of circulating astroviruses. In addition, the presence of such a G2–lineage one mosaic genome suggests that capsid exchange can occur on a relatively conserved polymerase backbone in canine astroviruses circulating in South Korea. Such events may contribute to the evolutionary diversity of CAstV in this region.

The structural comparisons among the JBN-CAstV strains provide insight into how sequence divergence relates to potential functional and antigenic differences. Overall, despite substantial heterogeneity in ORF2 AA sequences in each lineage, all predicted spike-domain structures from the JBN-CAstV strains shared a conserved β-barrel scaffold (Fig. 4). This preservation of the core fold is consistent with structural studies of human astrovirus capsids, which have shown that the spike domain adopts a conserved β-sandwich core. In contrast, most sequence and structural diversity is concentrated in surface-exposed loops at the periphery [45–47]. The rigid-body superposition of the predicted spike domains showed that most between-strain differences were confined to peripheral β-elements, particularly the top β-hairpins and side-sheets. In contrast, the central β-barrel displayed only minimal deviation among lineages. These results are similar to a previous study on astrovirus spike structures, in which localized changes occur mainly in surface-exposed regions around a conserved spike core [48].

Conserved residues (low-entropy) were scattered through the spike, with a tendency to cluster within β-strand elements of the core. In contrast, higher-entropy positions occurred in several discrete segments spanning the N-terminal, central, and C-terminal regions (Fig. 5). Many of these variable segments overlapped with residues predicted by DSSP to adopt loop or turn (“TT”) conformations and to be surface-exposed. In a previous study, surface-exposed loop regions were identified as the most variable portions of the spike [45–48]. They were proposed to contribute to antigenic drift and lineage-specific divergence, whereas the conserved β-barrel forms a stable core required to maintain the integrity of the spike scaffold [46, 49]. Consistently, many of the variable sites in our alignment were located within loop- or turn-like regions and were predicted to be surface accessible. In contrast, the most conserved residues tended to cluster along the β-strands.

The three major putative linear epitope clusters—an N-terminal region around residues 30–70, two central clusters spanning residues 75–95 and 111–130, and a C-terminal region around residues 198–218—were all located in stretches where entropy values were generally above the median (Fig. 5 and Table 3). Lineage-associated patterns were most apparent in the central and C-terminal regions. In particular, the 75–95 cluster was repeatedly predicted in lineage 4 strains and was centered on the conserved GQKSNSQY motif, with most lineage 4 sequences sharing an identical peptide and one strain showing a closely related single-residue variant. A related peptide was also predicted in the lineage 2 strain JBN/22D67, suggesting partial motif conservation across lineages.

For the C-terminal region, a putative linear epitope candidate around residues 198–218 was conserved across lineage 1 strains, whereas a motif-related candidate was observed in one lineage 2 strain (JBN/22D127-1) at a nearby but not identical position, with substantial sequence differences in the surrounding residues. In the previous goose astrovirus studies, linear B-cell epitopes located in conserved regions of the capsid protein have been utilized to generate diagnostic monoclonal antibodies and to develop epitope-based ELISA or IFA assays for serological detection [50, 51]. Likewise, the recurrently predicted motifs identified here—particularly the GQKSNSQY-centered segment in lineage 4 and the conserved C-terminal candidate in lineage 1—may provide useful starting points for developing CAstV serological reagents. However, because these epitopes were predicted in silico, further experimental validation will be required to confirm antigenicity and diagnostic performance.

Conclusion

Collectively, these data indicate that canine astroviruses circulating in South Korea are genetically diverse Mamastrovirus 5 strains, most of which fall into lineages 1 and 4, and that they include at least one putative recombinant (JBN/22D16-1) carrying a G2 polymerase backbone and a lineage one capsid. The combined use of structural modeling, entropy analysis, and linear B-cell epitope prediction identifies specific strains and epitope regions—such as the GQKSNSQY-containing cluster in lineage four and the conserved C-terminal epitope in lineage 1—as promising candidates for future functional studies on diagnostics, antigenicity, and receptor usage. Together, these results outline the current genetic and structural profile of canine astroviruses in South Korea and require further in-depth study to reveal their antigenic properties, receptor interactions, and roles in lineage-specific viral ecology.

Abbreviations

AstVs

Astroviruses

CAstV

Canine astrovirus

ORFs

Open reading frames

RdRp

RNA-dependent RNA polymerase

CPV

Canine parvovirus

CCoV

Canine coronavirus

CDV

Canine distemper virus

CAV

Canine Adenovirus

GSP

Gene-specific primer

s2m

Stem-loop 2-like motif

MSA

Multiple sequence alignments

ML

Maximum-likelihood

AA

Amino acid

RDP

Recombination Detection Program

pLDDT

Predicted local-distance difference test

pTM

Predicted Template Modeling score

G1

Group 1

G2

Group 2

RAstV

Raccoon dog AstV

Authors’ contributions

CGJ wrote the manuscript. JKO conceptualized the manuscript. HYJ and KKL provided the canine samples. CGJ and SBC completed the methodology, where all authors analyzed the resultant dataset. JKO edited and reviewed the manuscript. All authors have read and approved the final manuscript.

Funding

The research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2019R1A6A1A03033084).

Data availability

The data is provided within this article and in additional files.

Declarations

Ethics approval and consent to participate

Canine samples analysed in the present study were provided by the Animal and Plant Quarantine Agency through a collaborative research arrangement. The authors also confirmed that ethical approval is not essential to use animal materials received for diagnostic purposes in accordance with the guidelines of the Jeonbuk National University-Institutional Animal Care and Use Committee (JBNU-IACUC).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.