Distinct evolutionary patterns of endemic and emerging parvoviruses and the origin of a new pandemic virus

Significance

Comparing the evolution of a virus in its reservoir host with that seen in a new host will reveal the special circumstances that allow epidemic emergence. A feline parvovirus (FPV) that jumped to dogs in the mid-1970s formed canine parvovirus (CPV), which has circulated worldwide through today. The evolutionary rate of FPV in its original hosts was much lower than that of CPV in dogs, and the mutational patterns seen in the different hosts were distinct. Early CPV isolates differed from the ancestral FPV clade in several key host range mutations. These results highlight the complex biology associated with epidemic emergence, including host-specific rates of lineage evolution and complex origins of host-adaptive mutations.

Abstract

Emergence of epidemic viruses in new hosts threatens both human and other animal populations, and often involves virus evolution to overcome barriers that normally prevent efficient infection and spread in that host. After transfer the separated viruses will evolve in parallel as they spread within the original and new hosts. Here we examine the details of a virus involved in such a host-jumping event, where we define the natural evolution of feline panleukopenia virus (FPV) over 60 y, clarify the origins of the new pandemic canine parvovirus (CPV) that arose in the 1970s, and compare the separate evolution of those viruses over 47 y in cats or dogs. Several live-attenuated FPV vaccine viruses originated from early-1960s isolates or were a recombinant of an early virus, and many sequences in databases proved to be vaccine-derived. The sequences of wild viruses showed that FPV-like strains evolved at less than one-third the rate observed for CPV in dogs, and the higher rate of CPV evolution has been consistent since 1979, when a genetic variant became widespread. The common ancestor of the CPV lineage was related to FPVs from Europe and contained several unique host-adaptive capsid changes associated with canine transferrin receptor type-1 binding. Although the FPV vaccine strains are around 60 y old, little selection for antigenic variation was observed. The distinct evolutionary patterns of these closely related viruses circulating for decades in different hosts emphasize the complex evolution associated with viral epidemic emergence and spread in endemic and new hosts.

Control of new viral diseases requires a clear understanding of their evolutionary dynamics allowing epidemic emergence, or of changes in antigenic structures that allow immune evasion (1). Viruses that emerge to cause epidemics or pandemics in new hosts represent significant threats to human and animal health (2). Our knowledge of how such viruses overcome the barriers that usually prevent emergences is incomplete (3–5), in particular the information about the ancestral virus and its evolutionary patterns in the original reservoir hosts (6). Comparing the evolution and biology of viruses in their original and new hosts can show how such emergence events have occurred, and how we might forestall similar outbreaks in the future (7).

Canine parvovirus (CPV) has been most recently classified as Protoparvovirus carnivoran 1 (8), and its emergence from the feline panleukopenia virus (FPV) as a new virus in dogs occurred around 50 y ago (4, 9). Diseases in cats similar to those caused by FPV were reported around 1887 (10), and in the 1920s (11). In the late 1940s and early 1950s, a new disease seen among farmed mink was associated with an FPV-like parvovirus (12). FPV was first isolated in tissue culture in the early 1960s in the United Kingdom, and the sample was from a captive snow leopard (13). That virus was used to prepare attenuated viruses and since the 1970s attenuated FPV vaccine viruses have been included in the core vaccines recommended for all kittens (14).

In the mid-1970s a variant of FPV emerged in dogs that was named CPV type-2 (CPV-2) to distinguish it from the previously described minute virus of canines (MVC; now Bocaparvovirus carnivoran 1). Although first reported in many regions of the world during 1978 (15–18), CPV-2 strain was likely circulating in Western Europe for a few years before 1978, since anti-CPV antibodies were detected in sera collected from dogs in Europe between 1974 and 1977 (19–21). In other regions of the world, the first CPV-antibody positive sera were collected in early- to mid-1978 (22–24). Between 1979 and 1981 the original strain was replaced globally by a variant which contained several mutations and therefore was named CPV-2a (25, 26). In contrast to the CPV-2 strain, the CPV-2a strain had gained the feline host range (27, 28). Despite the feline host range of CPV-2a, most infections of cats since 1978 are due to FPV (29). Both FPV and CPV strains can naturally infect and cause disease in other hosts among the Order Carnivora (30).

The parvovirus capsid is a 26 nm diameter T = 1 icosahedron assembled from 60 copies of a combination of the overlapping VP1 and VP2 proteins (31). The small repeated exposed structure recognizes the host transferrin receptor type-1 (TfR), antibodies, and the sialic acid N-glycolylneuraminic acid (Neu5Gc) (32–35). The capsid packages a linear single-stranded DNA genome (ssDNA) of about 5,120 bases which encodes the nonstructural (NS) and the capsid or viral proteins (VP), including those from splice variant mRNAs or protease cleavage—NS1, NS2, VP1, VP2, and VP3 (36). Transmitted through a fecal-oral route, the virus is robust and can survive in the environment for weeks so it may be moved long distances on contaminated fomites (37).

Cell infection involves binding to the transferrin receptor type-1 (TfR) (38, 39). TfR is a homodimeric type-2 membrane glycoprotein which binds to iron-loaded transferrin (Tf), as well as other host proteins (ferritin and homeostatic iron regulator protein [also known as High FE2+ (HFE)], and it is a receptor for other pathogens (40, 41). The capsids of CPV and FPV bind TfR through a small surface of the TfR apical domain, as defined by cryoelectron microscopy (cryoEM) (42) and by mutational analysis of the capsid and receptor (43, 44). The canine host range of CPV-2 (and CPV-2a) is associated with a small number of capsid changes which allow binding to the canine TfR and infection of dog cells (45, 46). Some differences between FPV and CPV also altered antibody binding epitopes (32, 47–49) or the capsid interactions with the sialic acids (35, 50). FPV capsids do not bind the canine TfR or infect canine cells, due largely to the presence of a novel N-glycosylation in the canine TfR apical domain that falls within the capsid binding site (51). The mutations in CPV controlling canine host range allow that capsid to bind to the glycosylated canine TfR (47).

The parvovirus capsid rapidly induces protective antibodies (52) which recognize two dominant antigenic structures on the capsid surface (32–34, 53, 54). Many capsid mutations alter both antibody and TfR binding (33, 51, 55, 56), so that antigenicity and host range are coupled targets of selection in parvoviruses.

Previous studies examining the evolutionary patterns of FPV from cats or related hosts mostly examined the VP2 genes, and did not systematically identify the vaccine strains in the datasets (57–61). Recent studies have shown global distribution of FPV strains, and only a few mutations have spread widely—including the VP2 residue 91 Ala to Ser substitution (62, 63). We have recently defined 47 y of evolution of the full-genomes from natural infections of CPV in detail (64), and here we compare that to 60 y of FPV. We identify the origins of several live-attenuated vaccines in current use, reveal the closest relatives of CPV-2 among the FPV-like viruses, and reveal different rates and patterns of evolution between the two viral lineages during their decades of parallel spread.

Materials and Methods

FPV Sample Collection.

Seventeen FPV-containing samples since 1964 were obtained from original host tissues or cell cultured isolates; these samples had been stored in our laboratory or obtained from other collections. The samples sequenced in this study are described in SI Appendix, Table S1 (and denoted in Fig. 1 with black triangles), along with other FPV full-genome sequences obtained from the GenBank NCBI database.

Fig. 1.

Evolutionary relationships among FPV full-genome sequences. ML tree of full-length FPV genomes (n = 63), in addition to an early CPV-2 emergent clade (n = 5). Newly isolated sequences are denoted with a triangle. Bootstrap support values of >90% are indicated with a white diamond at the nodes, scale represents nucleotide divergence. Branch tips are denoted with shapes colored by geographic origin (continent) and by sequence source (C = cell culture-adapted virus; H = host-derived virus, and V = vaccine-derived virus). A timeline of samples (below; colored by country as in the tree) shows the temporal and geographical origins of natural host-derived sequences analyzed.

We also sequenced six commercially available live-attenuated FPV vaccine viruses obtained in 2022 and 2023. PureVax and Solo-Jec Feline 3 (Boehringer Ingelheim) were obtained in the United States. Virbagen Felis RCP (Virbac), Versifel CVR-T (Zoetis), PureVax RCP (Boehringer Ingelheim), and Nobivac RCP (Intervet International BV) were obtained in Germany.

FPV Genome Amplification, DNA Library Preparation, and Sequencing.

Sequencing methods have been reported previously (64). Briefly, DNA isolated using the E.Z.N.A. Tissue DNA Kit (cat. no D3396-01) was used in PCR using the Q5 High Fidelity DNA Polymerase (New England BioLabs) as specified by the manufacturer. We amplified two overlapping genomic segments that covered the NS1 and VP1 reading frames using primers: 5’-CCGTTACTGACATTCGCTTCTTG-3’ and 5’-GAACTGCTCCATCACTCATTG-3’, and 5’-CATCCATCAACATCAAGACCAAC-3’ and 5’-CTTAACATATTCTAAGGGCAAACCAACCAA-3’. After purification 0.5 ng of each fragment was pooled together as input DNA (1 ng DNA total) to construct barcoded sequencing libraries with the Nextera XT DNA Library Preparation Kit (Illumina). Libraries were multiplexed and sequenced using MiSeq 2x250 Illumina sequencing. Raw read data can be found in NCBI Sequence Read Archive (SRA) under BioProject PRJNA1288508 (SI Appendix, Table S2).

Illumina Read Processing, Analysis, and Bioinformatics.

Raw sequencing reads were trimmed using BBDuk (https://jgi.doe.gov/data-and-tools/software-tools/bbtools/bb-tools-user-guide/bbduk-guide/) to remove adaptors, PCR primers, and low-quality regions from reads. Reads were merged and mapped to an FPV reference sequence (consensus accession no. M38246) using Geneious Prime. Reads were error-corrected and normalized to 5,000-fold coverage per site using BBNorm (https://jgi.doe.gov/data-and-tools/software-tools/bbtools/bb-tools-user-guide/bbnorm-guide/). An average of ~5,000-fold coverage per site was obtained for a majority of the genome. Consensus genome sequence accession numbers are in SI Appendix, Tables S1 and S2.

FPV Full-Genome and VP2 Phylogenetic Analyses.

We determined the evolutionary relationships among the full-genome consensus sequences (n = 63), including 22 sequences generated in this study (six from FPV vaccines), along with 41 FPV sequences from GenBank for which isolation location, date, and host information were available (SI Appendix, Table S1). To determine the most closely related FPV sequences to the early CPV emergence, we included sequences of 1978 and 1979 CPV-2 isolates in the phylogenetic analysis. Maximum likelihood (ML) phylogenetic analysis was performed using PhyML (65) or IQ-TREE (66), employing a general time-reversible (GTR) substitution model, gamma-distributed (Γ) rate variation among sites, and bootstrap resampling (1,000 replications). Trees were rooted to the earliest FPV sample available (FPV-4, 1964, accession no. EU659112).

We also examined trees prepared using midpoint rooting or rooted to an outgroup (porcine parvovirus, 64.1% identical, or Kilham rat parvovirus, 63.2% identical) to reveal the patterns shown.

Sequences of highly cell cultured viruses and identified vaccine virus genomes from the dataset were removed to allow analysis of the natural infection-derived sequences (n = 40) as both full-length genomes, and as NS1 and VP2 sequences separately. We also analyzed a larger dataset of FPV VP2 sequences from the GenBank NCBI database (n = 183). Comparison with the CPV natural genomes was done using a dataset of 212 sequences examined in a previous study of that virus (64).

Mutation and Selection of FPV and Comparison with CPV.

Mutations in FPV genomes found in wildtype natural infections, vaccine virus, or highly cell-cultured samples were identified by comparison with the earliest FPV (FPV-4). Widespread polymorphisms in wildtype natural infections were counted if found in ≥10% of samples (at least 4 of 40). The rates of evolution of natural FPV (n = 40) and CPV (n = 212) genomes were determined using root-to-tip genetic distance by regression of each ML tree by date (year) in the TempEst v.1.5.3 software (67), and by the Bayesian Markov chain Monte Carlo (MCMC) approach run in BEAST v.1.10.4 (68). MCMC sampling was performed with a relaxed clock, a generalized time-reversible (GTR) substitution model with gamma distribution in four categories (Γ4), at a minimum of 100 million runs, with duplicates combined in Log Combiner v1.10.4. We confirmed statistical convergence with an effective sample size [ESS] ≥ 200 (consistent traces, removed burn-in at 10 per cent) in Tracer v1.7.2 (69). Parameter estimation is given as Highest Posterior Density (HPD) interval.

The relative numbers of nonsynonymous (dN) and synonymous (dS) nucleotide substitutions per site in NS1 and VP2 ORFs of FPV and CPV natural isolates were calculated using the SLAC (Single-Likelihood Ancestor Counting) method [Datamonkey package, https://datamonkey.org/ (70)]. A 95% CI was calculated for the mean dN/dS ratio and represented as error bars.

Structural Analysis of FPV Capsid Mutations and Receptor or Antibody Binding Sites.

To model the homologous FPV mutations, the canine TfR structure was predicted using AlphaFold3 (71) and placed into the cryoEM density of the black-backed jackal (bbj) TfR in complex with the CPV-2 capsid using rigid-body fitting (EMDB ID: 20002) (PDB ID: 6OAS) (42). The bbjTfR structure was used since that is closely related to the canine TfR, but lacks the additional glycosylation found on that receptor, and was used in the TfR:capsid costructure. The N-glycan moieties [(Hex)2(HexNAc)1(NeuAc)1(Man)3(GlcNAc)2] on the canine TfR residue N384 were modeled into the TfR–CPV complex structure using the CHARMM-GUI Glycan Modeler (72). The effects of viral mutations on virus:ligand interactions were examined in the context of the cryo-EM structure by using Chimera X (73). Those included the bbjTfR in complex with the CPV-2 capsid (PDB ID: 6OAS) (42) and the structures of canine antibody Fab–capsid complexes (PDB IDs: 9E7W and 9E60) (74, 75).

Statistical Analyses.

Data were analyzed using GraphPad Prism (GraphPad Software, Inc., La Jolla, CA).

Results

Sixty Years of FPV Evolution and Identification of Vaccine Viruses.

Here we reconstruct the evolutionary history of FPV and define the relationship between vaccine viruses and other isolates. The FPV sequences generated in this study, including those of live-attenuated vaccines widely used in the United States and Western Europe, allowed us to generate a maximum-likelihood (ML) tree of full-length FPV genomes (total n = 66), along with three CPV-2 genomes (Fig. 1). The tree topology of the major FPV clades (and position of the CPV-2) was consistent whether rooted to the earliest sample, using mid-point rooting, or rooting to distantly related (~64% identical) protoparvovirus outgroups. Most nodes had limited bootstrap support within the phylogeny, reflecting the fact that the sequences share >98.5% nucleotide identity overall. Most viruses were isolated from domestic cats, but some were collected from other susceptible hosts, including mink, raccoons, pine marten, river otter, pandas, and several other felids.

Most vaccine virus sequences clustered with each other and with other early isolates (13). One vaccine virus (Solo-Jec Feline 3) grouped with more recent FPV samples in the full-genome phylogeny (Fig. 1). However, that was a recombinant, with a VP2 gene related to other vaccines and early 1960s isolates (Fig. 2A) and had a distinct NS1 sequence. A number of sequences reported as wildtype FPV strains clustered with the vaccine viral sequences and were, therefore, likely derived from vaccine viruses (Fig. 1).

Fig. 2.

FPV vaccines have similar VP sequences and FPV natural sequences show convergent organization across genome and absence of natural recombination. (A) ML tree of expanded FPV VP2 sequences (n = 183) show the close relationship of all vaccine sequences in the capsid. ML trees of FPV natural sequences (n = 40) were generated for both NS1 (B) and VP2 (C) ORFs. Tree structure and clades match that seen with full-length genomes (Fig. 1).

After removing vaccine-derived and highly cell-culture passaged viruses, a collection of 40 FPV wildtype viruses was analyzed as complete genomes. The sequences fell into two distinct clades (Clades A and B) (Figs. 1 and 2), and while viruses showed a global distribution, North American isolates tended to fall within Clade A and those from Europe-Asia in Clade B. We also separately analyzed the NS1 and VP2 open reading frames, and those showed no obvious recombination lineages among that set of wildtype FPVs (Fig. 2 B and C).

The Origin of CPV and Canine Adaptation.

The branch leading to CPV-2 fell within a group of Clade B FPV sequences from Europe collected at various times after the emergence of CPV (Fig. 1) (25). That FPV—CPV branch included twelve nonsynonymous and ten synonymous substitutions across the genome (Table 1). Nine of the nonsynonymous substitutions fell within the capsid VP2 open reading frame (Fig. 3A). Differences in the VP2 capsid between the ancestral FPV and the earliest CPV-2 sequences included surface residues (80, 93, 232, 323, and 564) adjacent to the threefold spike, some of which influence the canine host range (Fig. 3B). The CPV-specific mutations (and subsequent CPV-2a replacement) within the capsid protein gene have clear effects on cell binding and infection using the nonglycosylated TfRs from cats, raccoons, and black-backed jackals, as well as the glycosylated canine TfR (Fig. 3C) (32, 42, 47, 49, 51).

Table 1.

Alignment uncovered variations between consensus FPV lineage to earliest emergent CPV-2 genomes

Nucleotide	s/ns	ORF	Amino Acid
C743T	ns	NS1	T248I
T1224C	s	NS1	I408
A1479G	s	NS1	V493
var1620T	s	NS1	V540
G1926A	ns	NS1	M152V
var2002G	s	NS1	N/D668
C2110T	ns	VP1	L4F
A2753G	ns	VP2	K80R
A2760G	s	VP2	V82
A2793C	ns	VP2	K93N
C2816T	ns	VP2	T101I
T2822C	ns	VP2	V103A
G3208A	ns	VP2	V232I
T3213C	s	VP2	Y233
G3481A	ns	VP2	D323N
A3552G	s	VP2	E346
A3555G	s	VP2	A347
G3637A	ns	VP2	D375N
T3681C	s	VP2	T389
A4137C	s	VP2	A541
A4205G	ns	VP2	N564S
C4217G	ns	VP2	A568G

Fig. 3.

Molecular and structural evolution of parvovirus host shifts. Mutations on the branch leading to CPV-2 emergence. (A) Nonsynonymous VP2 mutations fixed during host shifts from likely FPV ancestor to earliest CPV-2 isolates (red) to CPV-2a pandemic (green). (B) Direct comparison of FPV-CPV-2 VP2 side chain changes at threefold spike of capsid, in relation to the TfR. (C) Structure of CPV capsid in interaction with canine TfR (including N-glycan structure). Positions in VP2 of mutations arising during shift to CPV-2 (red) and CPV-2a (green) are highlighted on the structure. All mutations present on the branch leading to emergent CPV-2 are listed in Table 1.

Evolutionary Rates of FPV and CPV.

Examining the 47 y of parallel spread in cats and dogs since 1978 showed significant differences in the rates of evolution of the virus in the original versus the new host. Substitution rates of natural isolates of FPV and CPV (n = 40 and n = 212, respectively) were estimated using both root-to-tip divergence (Fig. 4A) and a Bayesian Coalescence Model (Fig. 4B). Those approaches showed FPV variation rates of 5.78 × 10⁻⁵ and 5.11 × 10⁻⁵ [HPD 2.03 × 10⁻⁵ –7.369 × 10⁻⁵] substitutions/site/year, and CPV rates of 2.181 × 10⁻⁴ and 2.184 × 10⁻⁴ [HPD 1.855 × 10⁻⁴ – 2.532 × 10⁻⁴] substitutions/site/year. The CPV divergence rate was therefore 3 to 4 times that seen for FPV (Fig. 4). In addition, the evolutionary rates of FPV and CPV lineages each appeared relatively constant during the entire periods analyzed, with a key exception being seen in the first 2 y of CPV spread, where several mutations were all fixed together in the ancestor of the CPV-2a lineage (Fig. 3).

Fig. 4.

FPV and CPV diverge in evolutionary rates in each host. (A) Divergence rates of FPV (n = 40) and CPV (n = 217) among recent natural sequences were generated from ML trees analyzed using root-to-tip regression in TempEst. (B) Substitution rates were calculated following a Coalescence Model analysis of natural FPV and CPV sequences. (C) Calculated dN/dS ratios of respective gene ORFs among natural FPV and CPV sequences (±SEM).

Patterns of Variation in FPV Genomes and Comparison to CPV.

Comparing the ratios of nonsynonymous to synonymous (dN/dS) mutations among wildtype FPV and CPV sequences over their circulation showed minimal selection and no significant differences between each virus in their respective hosts. The ratio for NS1 genes was somewhat greater for FPV (0.26 ± 0.06) than seen for CPV (0.17 ± 0.03), while those for the VP2 genes were similar (0.13 ± 0.03 versus 0.16 ± 0.03, respectively) (Fig. 4C). There were 45 widespread mutations or polymorphisms among the FPV genomes, including 32 that were synonymous and 13 that were nonsynonymous (Fig. 5 and SI Appendix, Table S3). Of the synonymous mutations, 12 fell within the nonstructural genes (NS) and 16 within the VP coding region. For the 13 nonsynonymous mutations, 6 were found in the NS1 gene, 2 were found to impact both the NS1 and NS2 overlapping reading frames, 2 were in the NS2 gene region alone, 1 was found in the VP2 gene, and 2 were in the small alternatively translated (SAT) alternate open reading frame within the VP2 gene. Compared to CPV the FPV sequences acquired more synonymous and fewer nonsynonymous changes (Fig. 5) (64). Many (12 of 23) of the nonsynonymous substitutions in CPV were in the VP1/2 gene, compared to one of 13 in the FPV VP1/2 (Fig. 5). The functions of most mutations within the FPV NS1, NS2, and SAT genes are not known, but host-specific effects may be possible, as these proteins interact with host-derived cellular proteins (76–81).

Fig. 5.

Mutational variation in wildtype FPV genomes. FPV genomes were aligned and categorized by source (wildtype, cell culture, vaccines). Widespread polymorphic variants among wildtype natural infections (≥10%) were determined among FPV (n = 40) and CPV (n = 217) samples. Mutation counts by type and ORF position are tabled; a full listing of widespread variants is found in SI Appendix, Table S3. FPV mutations are visualized on the genome using pins (nonsynonymous = colored squares, synonymous = white squares). Nucleotide numbering begins at the first coding site of the NS1/2 gene, and amino acid numbering is based on the translational start site of each respective gene.

FPV samples collected from hosts other than cats showed few specific sequence patterns (57, 59, 82, 83), apart from a VP2 Gly229Glu (G229E) mutation in giant panda isolates (84, 85).

Capsid Mutations, FPV Vaccines, and Antigenic Variation.

The only consistent variant that became widespread among wildtype FPVs since the 1960s was VP2 Ala91Ser (A91S). That is exposed in the threefold spike (Fig. 6A) and falls within the TfR footprint (Fig. 6B) and the binding site of canine antibody Fab2C5 (Fig. 6C) and is adjacent to antibody Fab7C8 footprint (Fig. 6D).

Fig. 6.

Widespread FPV VP2 variants and their relative position in the virus capsid surface in context with TfR and antibody footprints. (A) Reconstruction of the FPV capsid, showing axes of symmetry (oval = twofold axis, triangle = threefold axis, and pentagon = fivefold axes), and a roadmap of a single icosahedral asymmetric unit (PDB ID: 1C8G). The table below the roadmap identifies the widespread variants shared among wildtype infection samples (red) or majority of vaccine strains (gray). (B) Reconstruction of the high-resolution structure of the parvovirus capsid in complex with TfR (green, PDB ID: 6OAS) with corresponding roadmap where variants are outlined in red relative to the TfR footprint (green). (C) Reconstruction of the FPV capsid in complex with a canine-derived monoclonal antibody as Fab: 3G6 (yellow, PDB ID: 9E7W) with corresponding roadmap where variants are outlined in red relative to the Fab3G6 footprint (yellow). (D) Reconstruction of the FPV capsid in complex with a canine-derived monoclonal antibody as Fab: 7C8 (purple, PDB ID: 9E60) with corresponding roadmap where variants are outlined in red relative to the Fab7C8 footprint (purple).

Live-attenuated FPV vaccines analyzed were all closely related to early FPV isolates, including the snow leopard strain from around 1964 (FPV-2) (Fig. 1). The vaccine strains often have unique or shared nonsynonymous changes in NS1 and NS2. Two VP2 nonsynonymous changes were shared among the majority of the vaccine strains, including the Val232Ile (V232I) change on the threefold spike and Val562Leu (V562L) near the twofold axis—and both were exposed on the capsid surface (Fig. 6A). Residue VP2 232 fell within the TfR binding footprint, while VP2 residue 562 was immediately adjacent to it (Fig. 6B). Similarly, the FPV vaccine variant at residue 232 is within or close to both canine antibody binding sites (54), while VP2 residue 562 is adjacent to one of these sites (Fig. 6 C and D).

Discussion

Here we directly compare two parvoviruses before and after the emergence of one of those to cause a pandemic that has continued for around 50 y. We show that the original FPVs in cats and related carnivore hosts evolve slowly with few nonsynonymous mutations, that they are distributed among two clades that are partially geographically structured, and that the single branch that contained several host-adaptive coding changes gave rise to the ancestor of the CPV lineage. This highlights the complex processes required for the virus to successfully overcome a host-range barrier and to spread widely in its new host.

An aspect of our analysis is that the FPV genomes were derived from natural infections over 60 y, allowing us to exclude sequences from the widely used attenuated vaccine viruses. This detailed understanding of the FPV and CPV sequence evolution is paired with the structures and functions of their respective capsids, their interactions with TfRs from different hosts, as well as with natural postinfection host antibodies.

FPV Strains Show Distinct Patterns of Variation Compared to CPV.

Over the 47 y of parallel spread in different hosts the FPV sequences evolved at around 0.5 × 10⁻⁴ subs/site/year, compared to 2.2 × 10⁻⁴ subs/site/year for CPV (Fig. 4 A and B). The relatively higher evolutionary rate of CPV sequences could only be resolved most clearly throughout the 47 y since it spread worldwide. It is unclear how, and by what mechanism, the evolution of CPV progressed comparatively during the earlier posttransfer period (1979 to 1989) and later (1989 to 2024). Both the CPV emergence from the FPV-clade and the further emergence of the CPV-2a lineage involved many observed mutations, including several host-adaptive changes, in a relatively short order of time (25, 26, 48). Given the generally low rates of evolution in ssDNA parvoviruses, it is unclear how those groups of mutations arose and were fixed by gradual evolution.

Explanations for the overall faster evolution of the CPV lineage may include multiple nonexclusionary hypotheses. Given the immune naivete of dogs and the worldwide spread of the virus as a pandemic of dogs, an early expansive population size is likely, as well as continuing host or antigenic selection. Greater genetic drift resulting from different epidemiological properties in dogs can be considered. Observations of host-specific rate variation are common (86), and have also been measured for influenza A and SARS-CoV-2 viruses (87, 88). The involvement of additional hosts beyond cats or dogs is another hypothesis—for example, apparent partial intermediate viruses have been detected in raccoons, which may suggest pathways for the acquisition of mutational combinations that would be unfit in either cats or dogs but may have facilitated new potential genotypes (89).

Little parallel (convergent) evolution was observed between the FPV and CPV lineages since they diverged in 1978, with only VP2 V232I being seen among both FPV and CPV sequences (64), indicating there were different selective pressures or genetic drift in each host. The baseline evolutionary rate seen for carnivore parvoviruses (as exhibited by the FPV rate) in this study is similar to that determined for human B19 parvoviruses using combinations of contemporary- and ancient DNA-derived sequences, which were calculated to be 1 to 3 × 10⁻⁵ subs/site/year (90, 91), both examining well-adapted parvoviruses spreading in their normal host.

Changes Selected in Alternative Hosts.

Cat and dog populations vastly outnumber those of other carnivore hosts for these viruses, and so we expect that most FPV and CPV evolution occurs in those animals. No host-specific variation was identified among FPVs collected from snow leopards, wild cats, minks, raccoons, or bobcats, while the variation of VP2 residue 299 (Ala to Glu) may be host-adaptive for giant pandas (84, 85). In contrast, CPVs from noncanine hosts often show variations in capsid structures associated with TfR binding (89, 92), likely needed for efficient binding and infection using the nonglycosylated TfRs in those hosts (32, 47, 51).

Sixty-Year-Old FPV Vaccine Strains Select Little Antigenic Variation.

The FPV live-attenuated vaccines examined here were all derived from one or a few 60-year-old isolates, and those are included in core feline vaccines around the world (14). The natural evolution of FPV included only one or two variations within the two main capsid epitopes, suggesting little or no antigenic selection of the virus infecting cats. In contrast, several coding changes and antigenic variants of CPV have been observed, although there is no clear evidence of immune escape (93, 94). Escape from individual monoclonal antibodies occurs readily in parvoviruses in in vitro studies (95). However, although the host polyclonal antibody repertoire against the CPV (and likely FPV capsid) is limited, that appears sufficient to block infection and circulation of all strains of the virus in its animal hosts (96, 97). The FPV (and CPV) therefore is similar to other viruses where vaccines have retained their efficacy for decades with little immune evasion arising, such as smallpox, measles, and rinderpest viruses (98–100).

All These Viruses May Share a Recent Common Ancestor, and CPV-2 Is Most Closely Related to European FPV Strains.

Phylogenetic and rate analysis of all the FPV-derived (including CPV) sequences suggests that those viruses may share a common ancestor that existed in the late 19th century. The first reports of feline ataxia due to cerebellar hypoplasia in two kittens were from 1887 (10), while other FPV-associated diseases in cats and raccoons were reported in the 1920s and 1930s (11, 101). However, backextrapolation of the sequences collected since the 1960s may not be accurate, and these viruses may have been circulating for much longer. For comparison, while recent human B19 parvovirus sequences also appeared to converge upon a common ancestor in the 19th century (86, 102), ancient DNA in human remains up to ~7,000 y old showed that similar strains existed at that time (90, 91). Sequences of FPVs derived from ancient cat samples would reveal a more accurate rate if those could be found.

The CPV-2 strain was recognized, and the first viral sequences were collected, when it spread worldwide in 1978 (9, 17, 18). However, serological testing showed that the virus was circulating in dogs in Western Europe between 1974 and 1978 (19–21), and the ancestor of the CPV lineage fell within a group of FPVs from Europe (Figs. 1 and 2). As those European FPV isolates were collected after 1978 (between 2006 and 2016), they would be descendants of viruses similar to the CPV ancestor. The true ancestor may still be sought among archival tissues collected from dogs, cats, or related hosts in Europe or Eurasia during the early to mid-1970s. This analysis also rules out the possibility that CPV emerged from an FPV vaccine (Fig. 1) (103, 104).

Evolution of the Canine Host Range.

While dogs would have been consistently exposed to FPV, CPV is only known to have emerged once to cause an epidemic. The host barrier of dogs was high, with several host-adaptive changes being present in the long branch between FPV and CPV-2—including VP2 residues 93, 103, 323, and 375 (Fig. 3A) (32). While some variation of those sites has been reported [e.g., VP2 93 Lys in a CPV background (60), VP2 93 Asn in an FPV background (61)] those are rare exceptions. Canine host range resulted from the virus adapting the capsid contacts with the canine TfR, including both protein–protein and protein–glycan interactions (38, 45, 51, 105), and key mutations were associated with small changes in capsid structure, which together allowed them to engage the canine TfR to allow efficient entry and infection (Fig. 3 B and C) (42, 55).

The slow evolution of FPV in its normal hosts indicates that acquiring several canine host range mutations involved a different evolutionary process. Parvoviruses differ from the RNA viruses that are often suggested to be most likely to emerge due to their high mutation rates. The ssDNA parvovirus genomes are replicated by host cell DNA polymerase during cellular S-phase (most commonly pol δ), likely without the error-correcting mechanisms of host genome replication (106), and their error rates of replication are likely about 10-fold lower than most RNA viruses. Although polymorphisms are rarely detected within deep-sequenced parvovirus DNA samples (64), mutant genomes are being generated at a level where they can be quickly selected—including by host cells or neutralizing antibodies (64, 95). It is likely that the several canine-adaptive mutations found in CPV-2 (and then in CPV-2a) all arose through normal processes of mutation, but with strong selection for the use of the canine TfR to bind and infect dogs.

Summary.

Comparing the evolution of a virus in its reservoir, the process of host jumping, and the posttransfer evolution shows that viral emergence in a new host required a special event that allowed the rapid acquisition of several mutations to overcome the high host range barrier posed by dogs. Similar to other examples of evolution involving punctuated equilibrium, those mutations would not have arisen and been selected in the normal course of evolution of the FPV but were likely due to strong selection within dogs (possibly with involvement of other intermediate hosts). After pandemic emergence, the higher evolutionary rate and increase of nonsynonymous changes in CPV suggests that there is continuing selection in dogs, even decades after the host jump. These general features are similar to those seen for some other emerging viruses such as HIV-1 and influenza virus, where many different mutations and factors were required for the virus to successfully become a pandemic pathogen in a new host (6).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

DNA sequences data have been deposited in Genbank, NCBI Sequence Read Archive (SRA) (PRJNA1288508) (107).