Genetic diversity accelerates canine distemper virus adaptation to ferrets

ABSTRACT

RNA viruses adapt rapidly to new host environments by generating highly diverse genome sets, so-called “quasispecies.” Minor genetic variants promote their rapid adaptation, allowing for the emergence of drug-resistance or immune-escape mutants. Understanding these adaptation processes is highly relevant to assessing the risk of cross-species transmission and the safety and efficacy of vaccines and antivirals. We hypothesized that genetic memory within a viral genome population facilitates rapid adaptation. To test this, we investigated the adaptation of the Morbillivirus canine distemper virus to ferrets and compared an attenuated, Vero cell-adapted virus isolate with its recombinant derivative over consecutive ferret passages. Although both viruses adapted to the new host, the reduced initial genetic diversity of the recombinant virus resulted in delayed disease onset. The non-recombinant virus gradually increased the frequencies of beneficial mutations already present at very low frequencies in the input virus. In contrast, the recombinant virus first evolved de novo mutations to compensate for the initial fitness impairments. Importantly, while both viruses evolved different sets of mutations, most mutations found in the adapted non-recombinant virus were identical to those found in a previous ferret adaptation experiment with the same isolate, indicating that mutations present at low frequency in the original virus stock serve as genetic memory. An arginine residue at position 519 in the carboxy terminus of the nucleoprotein shared by all adapted viruses was found to contribute to pathogenesis in ferrets. Our work illustrates the importance of genetic diversity for adaptation to new environments and identifies regions with functional relevance.

IMPORTANCE

When viruses encounter a new host, they can rapidly adapt to this host and cause disease. How these adaptation processes occur remains understudied. Morbilliviruses have high clinical and veterinary relevance and are attractive model systems to study these adaptation processes. The canine distemper virus is of particular interest, as it exhibits a broader host range than other morbilliviruses and frequently crosses species barriers. Here, we compared the adaptation of an attenuated virus and its recombinant derivative to that of ferrets. Pre-existing mutations present at low frequency allowed faster adaptation of the non-recombinant virus compared to the recombinant virus. We identified a common point mutation in the nucleoprotein that affected the pathogenesis of both viruses. Our study shows that genetic memory facilitates environmental adaptation and that erasing this genetic memory by genetic engineering results in delayed and different adaptation to new environments, providing an important safety aspect for the generation of live-attenuated vaccines.

INTRODUCTION

Many RNA viruses adapt rapidly to new environments while maintaining their essential functions. Since morbilliviruses are highly contagious and replicate extensively in their respective hosts (1, 2), they are well suited to investigate the underlying genetic plasticity. While the measles virus (MeV) only infects humans and certain nonhuman primates (1), the canine distemper virus (CDV) affects a broad range of carnivores (3).

Typically, morbillivirus transmission occurs via aerosols. CDV transmission may also occur through direct contact, urine, and/or feces (4–6). The viruses initially infect and amplify in immune cells in the respiratory tract. Infected immune cells spread the virus to the immune organs via the lymphatic system (7–9). In the later stages of infection, the viruses eventually infect respiratory epithelial cells, which is associated with the development of the characteristic rash, respiratory, and often gastrointestinal signs and is essential for transmission to new hosts (5, 7). Despite a higher lethality and increased potential of neuroinvasion, a CDV infection in ferrets (Mustela putorius furo) reproduces many aspects of the clinical disease seen in MeV-infected patients (5, 10, 11). Thus, this model is frequently used to characterize the genetic determinants of morbillivirus virulence (9, 12, 13).

Effective and safe live-attenuated vaccines exist for MeV and several veterinary morbilliviruses, including CDV (14, 15), and vaccination campaigns were the cornerstone of successful rinderpest eradication (16). The continued efficacy of vaccines originally developed in the 1960s illustrates the high overall genetic and antigenic stability of these viruses (17). In contrast, selective pressure results in the rapid emergence of escape mutants, providing considerable genetic plasticity (18–20). For example, wild-type MeV acquired the ability to use CD46 as an additional entry receptor within a few passages in cells that lack the wild-type receptors SLAM and nectin-4 (21); CDV SLAM-binding preference rapidly adapted from the canine to the human protein in vitro when cultivated in human SLAM-expressing cells (22). Moreover, a previously Vero cell-adapted CDV wild-type isolate reverted to virulence after three in vivo passages in ferrets (12).

Morbilliviruses have a negative-sense, single-stranded RNA genome of around 16,000 bases. The viral RNA-dependent RNA polymerase transcribes viral mRNAs and replicates the genome, introducing approximately 10⁻⁴ mutations per nucleotide (23, 24). Consequently, morbillivirus replication produces a dynamic population of closely related genetic variants, often referred to as quasispecies (25–27). The genetic diversity and the ability to produce this heterogeneity are critical for virulence and adaptation of most RNA viruses since they facilitate the emergence of escape variants and genetic variants with improved fitness (28–30). The distribution of genetic variants within the quasispecies depends on the history of the respective virus isolate and can be considered as genetic memory (31). There is increasing evidence that this memory can influence the response to selective constraints (32–34). However, the role of diversity in the pathogenesis of morbilliviruses is less well understood.

Here, we hypothesized that genetic diversity and memory contribute to the rapid adaptation and pathogenesis of CDV in ferrets. To test this hypothesis, we relied on an attenuated, Vero cell-adapted CDV stock (strain 5804) generated more than 20 years ago (12) and a recombinant clone derived from this virus (r5804) (12). We asked three specific questions: first, will a repetition of a passaging experiment in ferrets of the Vero cell-adapted 5804 virus stock result in a similar adaptation process and generate a virulent CDV strain (5804P) as observed previously (12)? Second, will a recombinant virus based on the original attenuated virus (r5804) be able to generate a virulent virus population through passaging in ferrets? Third, if both passaging experiments result in virulent viruses, will these viruses have adapted through shared mutations? To answer these questions, we compared the quasispecies dynamics of the attenuated, Vero cell-adapted 5804 strain and its recombinant derivative r5804 in response to host selective pressure in ferrets using RNA sequencing.

RESULTS

Low genetic diversity of the recombinant virus compared to the tissue culture-adapted isolate

To investigate the influence of genome diversity on host adaptability and pathogenesis, we generated two CDV stocks. One was a Vero cell-adapted, attenuated CDV strain (5804), which was isolated from an infected dog and then passaged in Vero cells (12). The other stock was its recombinant derivative r5804, which is a single cDNA clone derived from 5804 (12). We first determined their consensus sequences and genome diversities by total RNA sequencing, achieving an average coverage of 9,095 reads/nucleotide for 5804_P0 and 5,789 reads/nucleotide for r5804_P0 across the 15,690 nucleotides of the CDV genome (Fig. 1A). The higher read coverage at the 3′ end of the genome (Fig. 1A, top panel), especially the nucleoprotein (N) and phosphoprotein (P) gene regions, reflects the transcription gradient during morbillivirus replication.

Fig 1.

Input virus sequence analysis. (A) Genome coverage and variant sites. RNA sequencing coverage of the Vero-adapted CDV 5804 (blue) and recombinant CDV r5804 (red) P0 genomes (top graph). The coverages were mapped to the CDV 5804 reference genome (AY386315.1) (middle), and variant sites are depicted as lines at their respective genome position (bottom). (B) Total number of variant sites in the Vero cell-adapted and recombinant viruses. Bars represent the total number of variant sites for each virus. Darker gray scales show higher frequencies of variant alleles. (C) Mean Shannon entropy as a measure for genome diversity. Shannon entropy was calculated for the Vero-adapted 5804 (blue) and the recombinant r5804 (red) input viruses.

We then constructed two consensus sequences (5804_P0 and r5804_P0) and identified major alleles with frequencies above 50% and minor alleles with frequencies between 1% and 50%. In both stocks, variable sites were distributed throughout the genome (Fig. 1A, lower panel). 5804_P0 had 63 variant sites, while r5804_P0 had only 32 variant sites (Fig. 1B). Six of these variable sites were shared between the two viruses. 5804_P0 had more variable sites at higher frequencies than r5804_P0, resulting in a twofold difference in calculated mean Shannon entropy, a measure for the total genome diversity between both viruses (Fig. 1C). The lower diversity of r5804_P0 is the result of its generation from a single cDNA clone, whereas the 5804_P0 stock was the result of an extended serial passaging process. Notably, r5804_P0 exhibited four major allele differences compared to 5804_P0, which resulted in amino acid mutations (Table 1). As anticipated, our r5804_P0 virus had the most major alleles identical to its consensus cDNA sequence (GenBank No. AY386315.1) and only one amino acid (position 141 in P) differed here. In contrast, the Vero cell-adapted 5804_P0 stock had major allele differences in three of the four positions. Notably, the genetic diversity of these positions in our virus preparations reflected to a high degree the genetic diversity of CDV, as observed by the comparison of 220 CDV full-length sequences deposited in GenBank (Table 1; Data S1).

TABLE 1.

Non-synonymous major allele differences between the input viruses

Genome position (nt)	5804_P0 alleles (frequency)	r5804_P0 alleles (frequency)	Diversity in 220 GenBank sequences	Protein, amino acid position	Amino acid
2221	T (>99%)	T (9%)	T (97%)	P protein, 141	Fa
	C (<1%)	C (91%)	C (3%)		L
3469	A (75%)	Not detected	A (>99%)	M protein, 13	D
	C (2%)	Not detected	Not detected		A
	G (22%)	G (100%)	G (<1%)		Ga
4365	T (100%)	Not detected	T (97%)	M protein, 312	S
	Not detected	A (100%)	A (3%)		Ta
9436	A (63%)	Not detected	Not detected	L protein, 136	D
	G (37%)	G (100%)	G (100%)		Ga

^a Amino acids in the published reference sequence (GenBank No. AY386315.1).

Higher genetic diversity accelerates adaptation to a new host

To test whether the varying genome diversities in the 5804_P0 and r5804_P0 stocks led to differences in a host adaption process, both virus inocula were serially passaged in ferrets (Fig. 2A) as previously described (12). Toward this, two animals were infected with the respective inoculum and euthanized on day 7 after infection. Spleen, lung, and lymph node tissue were homogenized. The collected tissue homogenates were titrated individually, and the pooled material used for the infection of the next group was analyzed by RNA sequencing. The next group of two animals was infected intranasally with at least 1 × 10⁶ TCID₅₀ of virus derived from the pooled homogenates. Of note, due to insufficient titers in the organ homogenate, an additional expansion in cell culture on VerodogSLAMtag cells was required for the first two passages of the recombinant viruses (r5804_P1 and r5804_P2), which likely had an impact on the genome diversification (see Discussion).

Fig 2.

Ferret passages of Vero-adapted and recombinant viruses. (A) Schematic overview of experimental design. Throughout the passaging experiment, animals were sacrificed 7 days post-infection, and tissues were harvested for further analyses. Groups of two ferrets were infected with 1 × 10⁶ TCID₅₀ of the respective input viruses. The next groups were infected with equal volumes of lymph nodes, spleen, and lung homogenates at a dose of 1 × 10⁶ TCID₅₀. (B) Tissue titers at different passages. Each symbol represents an individual animal, and titers are expressed as log₁₀ TCID₅₀/g tissue 7 days post-infection.

5804 exhibited substantial replication in lymph nodes, spleen, and lung, and recovered virus titers moderately increased through the passaging experiment (Fig. 2B, blue circles). In contrast, r5804 exhibited strongly reduced replication in the first two passages, where it was only recovered from the lymph nodes of one animal (Fig. 2B, red triangles). Only after the third passage, r5804 was also recovered from spleen and lung tissues.

To follow changes in the viral genome across ferret passages, we performed RNA sequencing on all viral passages. During passaging, we observed an initial increase in the number of variant sites in P1 of 5804, but their number and frequencies declined during the further passages (Fig. 3A through D). Interestingly, r5804 exhibited a different evolution. The total number of variant sites remained low in P1 and P2 and only increased rapidly after the third passage (Fig. 3E through H). These different adaptation processes between the two viruses are reflected in the root-mean-square deviation (RMSD) describing the deviation of sequence from its origin (Fig. 3I) and the mean Shannon entropy, a measure for the total genome population diversity (Fig. 3J), calculated for each data set. The initial stagnation in the diversity of r5804 and the loss of variants in the second passage are in line with the inefficient replication of the early passages of this virus in vivo and indicate that a bottleneck effect was the reason for the hampered growth of the early passages.

Fig 3.

Comparison of major allele development during the course of passaging. (A–G) Genome-wide variant development (percent allele frequencies) of the Vero-adapted 5804 (left) and the recombinant r5804 (right) across passages: (A) 5804_P0, (B) 5804_P1, (C) 5804_P2, (D) 5804_P3, (E) r5804_P0, (F) r5804_P1, (G) r5804_P2, and (H) r5804_P3. Each dot represents an individual variant site. Variant site graphs are aligned to the viral genome (schematics on the bottom). Non-coding regions are depicted in light gray and coding regions in dark gray. (I) RMSD from input viruses. Harvested tissue homogenates were analyzed by RNAseq. The RMSD was calculated as a mean deviation to the respective input virus over the whole genome. (J) Viral diversity was calculated for each passage as the mean Shannon entropy for each virus.

Genetic diversity contributes to adaptation and increased virulence

We next asked whether the observed genome diversity of the input and adapted viruses after three passages correlated with their ability to cause more severe clinical disease in ferrets. For this, naive ferrets were infected with equal doses of the respective viruses (5804_P0, 5804_P3, r5804_P0, and r5804_P3) and monitored daily for clinical signs of infection, including weight loss, fever, and development of rash. Twice per week, blood was drawn for detailed viremia analysis. The animals were euthanized when they reached clinical end points. No lethality (Fig. 4A) or severe clinical signs of a CDV infection (Fig. 4B through E) were observed in animals infected with either of the P0 viruses. Aside from a short-lived leukopenia that appeared in all infected animals (Fig. 4E), the only clinical sign of infection was the slightly elevated mean body temperature around 9 days post-infection in animals infected with the Vero cell-adapted input virus (Fig. 4B, light blue).

Fig 4.

In vivo pathogenesis of P0 and P3 CDV 5804 and CDV r5804. Ferrets were infected with 1 × 10⁶ TCID₅₀ of the respective viruses. Animals infected with the Vero-adapted 5804 are depicted in blue [light blue for P0 (N = 5) and dark blue for P3 (N = 4)]. Animals infected with recombinant r5804 are depicted in red [light red for P0 (N = 3) and dark red for P3 (N = 5)]. (A) Survival curve. (B) Body temperatures during the infection period. Lines represent mean values within each group, and open symbols represent the values of individual animals. Dotted lines represent threshold values. Temperatures above 39°C are considered increased body temperature, and temperatures above 40°C are considered fever. (C) Body weight development during the infection period relative to the start weight. Lines represent mean values within each group, and open symbols represent the values of individual animals. Dotted lines represent threshold values. 5% is considered mild, and more than 10% is considered severe weight loss. (D) Heat map indicating development of rash in individual animals. Severity of rash is indicated by darker red scales, while white boxes indicate the absence of rash, black boxes indicate that an animal was dead, and gray boxes indicate that the respective animal was not further monitored at the indicated time point. (E) White blood cell counts. (F) Development of peripheral blood mononuclear cell (PBMC)-associated viral blood titers per 1 × 10⁶ PBMCs during the infection period. Lines are mean values, and error bars are standard deviations for each group.

In contrast, both passaged viruses (P3) caused pronounced disease in the infected animals. Animals infected with the 5804_ P3 virus developed severe fever above 40°C (Fig. 4B, dark blue) at the end of the second week post-infection. The ferrets lost weight (Fig. 4C, dark blue) and developed a generalized rash (Fig. 4D). All animals infected with the 5804_P3 virus were euthanized between days 13 and 15 post-infection once they had reached pre-defined experimental end points (Fig. 4A, dark blue). Importantly, the adapted recombinant virus (r5804_P3, dark red) caused disease in infected animals, but the clinical signs were less severe than those caused by the non-recombinant derivative 5804_P3 (Fig. 4B through E). All animals developed rash in the course of infection, but it remained localized in all but one case (Fig. 4D). One animal in this group lost about 10% body weight (Fig. 4C), developed a generalized rash (Fig. 4D), and eventually had to be euthanized (Fig. 4A). The other four animals in this group all cleared the infection and recovered.

All four viruses reached peak viremia 7 days post-infection (Fig. 4F). Infected peripheral blood mononuclear cells (PBMCs) were found as early as day 3 and remained detectable until 14–17 days post-infection. Generally, P3 viruses replicated to higher blood titers than their P0 derivatives, highlighting the successful adaptation to the new host in both cases. Moreover, the viruses 5804_P0 and 5804_P3 reached higher viremia titers than the respective recombinant viruses r5804_P0 and r5804_P3.

To test whether the recombinant virus would eventually become as virulent as 5804_P3, we generated three more passages of this virus in ferrets and assessed the virulence of r5804_P4 and r5804_P6 (Fig. 5A through F). Notably, additional passaging did not further increase the virulence of r5804 as compared to the third passage of this virus. All animals infected with r5804_P4 or r5804_P6 survived (Fig. 5A) and only developed mild clinical signs (Fig. 5B through E). Peak viral load in PBMCs was comparable for r5804_P3, P4, and P6, and the later passages were cleared even more rapidly than r5804_P3 (Fig. 5F).

Fig 5.

Longer passaging of r5804 does not further change virulence and pathogenesis in ferrets. (A) Ferrets were infected with 1 × 10⁶ TCID₅₀ of the respective viruses. Animals infected with r5804_P0 are depicted in light red (N = 3), with r5804_P3 in dark red (N = 5), r5804_P4 in turquoise (N = 2), and r5804_P6 in yellow green (N = 2). Data for r5804_P0 and r5804_P3 are the same as in Fig. 4 and repeated here for easier comparison of the different groups. (A) Survival curve. (B) Body temperatures during the infection period. Lines represent mean values within each group, and open symbols represent the values of individual animals. Dotted lines represent threshold values. Temperatures above 39°C are considered increased body temperature, and temperatures above 40°C are considered fever. (C) Body weight development during the infection period relative to the start weight. Lines represent mean values within each group, and open symbols represent the values of individual animals. Dotted lines represent threshold values. 5% is considered mild, and more than 10% is considered severe weight loss. (D) Heat map indicating the development of rash in individual animals. Severity of rash is indicated by darker red scales, while white boxes indicate the absence of rash, black boxes indicate that an animal was dead, and gray boxes indicate that the respective animal was not further monitored at the indicated time point. (E) White blood cell counts. (F) Development of PBMC-associated viral blood titers per 10⁶ PBMCs during the infection period. Blood from r5804_P4 and P6 animals was not sampled after day 14, and thus no white blood cell or viremia data are available. These animals showed small patches of skin irritation that persisted for several days afterward, and clinical examinations were performed up until the time points indicated in panel D.

These data show that the adaptation process of both the recombinant r5804 and the Vero cell-adapted 5804 to ferrets is accompanied by an increase in virulence. After three consecutive passages, both viruses induced a disease typical for a CDV infection. However, clinical disease induced by the recombinant derivatives remained less severe even after three additional in vivo passages. Moreover, the 5804_P3 generated here reproduced the pathogenesis profile of CDV 5804P generated in the original study from 2003 and of the recombinant CDV r5804P generated based on the consensus sequence of the adapted virus (12).

Adaptation to ferrets results in distinct genetic changes for each virus

We next asked whether shared mutations in the two adapted viruses are associated with more efficient in vivo replication (Fig. 6). For this, we compared all variant sites in each passage for 5804 (Fig. 6A) and the recombinant r5804 (Fig. 6C) viruses in our RNAseq data sets. The total number of variant sites and their frequency distribution for each virus and passage (Fig. 6B and D) followed a similar trend as the mean Shannon entropy calculations (Fig. 3J). Substitutions of major alleles (frequencies > 50%) occurred at 10 sites in 5804 (Fig. 6A; Table 2). Four allele changes (positions 2168, 9137, 9173, and 9436) were identified as minor alleles of at least 1% frequency in the input virus (5804_P0), and their frequencies gradually increased over the passages (Fig. 6A). Moreover, 5 out of 10 mutations in 5804_P3 caused amino acid substitutions in different gene products (Fig. 6E; Tables 2 and 3). Notably, our passaging of 5804 reproduced nearly perfectly the result of the passaging experiment of the same inoculum 20 years ago (12). Seven out of ten consensus changes (positions 1662, 2168, 6856, 9137, 9173, 9436, and 10935) were shared between this and the historic passaging experiment of 5804 (Table 2), including all five non-synonymous changes (positions 1662, 2168, 6856, 9436, and 10935).

Fig 6.

Comparison of genetic variability and major allele development. (A–D) Genome-wide variant development of the Vero-adapted 5804 (A and B) and the recombinant r5804 (C and D) across passages. (A and C) Genome proportions reflect variant frequency in different genome regions during the entire experiment. Each line represents an individual variant, and passages are shown in separate rows. Color intensity indicates variant frequency with white illustrating frequencies below the detection limit of 1% and black above 99%. Variants with a change in the major allele compared to the input sequence are indicated with the respective genome position. (B and D) Variant site counts at different passages. The variants were grouped by frequency analogous to the heat maps. (E) Distribution of major allele changes between input and P3 viruses. The position in the genome of major allele changes is indicated. Non-synonymous changes are shown as blue (5804_P3) and red (r5804_P3) bars; gray bars indicate synonymous changes.

TABLE 2.

Major allele changes in Vero cell-adapted CDV 5804 upon adaptation to ferrets

Genome position	Protein, amino acid change	Reproducibility compared to CDV 5804P (AY386316.1)	Input frequency in 5804_P0
1662	N protein, G519R	Yes	<1%
2168	P/V proteins, I123T	Yes	1%
	C protein, S119R
4483	M-F UTR, non-coding sequence	No	Not detected
6856	F protein, T641I	Yes	<1%
8755	H protein, synonymous	No	Not detected
9137	L protein, synonymous	Yes	31%
9173	L protein, synonymous	Yes	31%
9436	L protein, D136G	Yes	37%
10935	L protein, P636T	Yes	<1%
13406	L protein, synonymous	No	Not detected

TABLE 3.

Major allele changes in recombinant CDV r5804 upon adaptation to ferrets

Genome position	Protein, amino acid change	Reproducibility compared to CDV 5804P (AY386316.1)	Input frequency in 5804_P0
1662	N protein, G519R	Yes	<1%
3436	M protein, T2I	No	Not detected
5801	F protein, synonymous	No	Not detected
7411	H protein, synonymous	No	<1%
10870	L protein, I614T	No	<1%
13896	L protein, R1623S	No	<1%

In contrast, passaging of the recombinant virus r5804 resulted in the accumulation of a different set of variants (Table 3). Major allele changes occurred at six sites (Fig. 6C; Table 3), and none of these were found in the inoculum above the error threshold of 1% (Fig. 6C). All mutations gained through the passaging of r5804 appeared suddenly with frequencies above 99%, indicating their de novo generation and rapid selection through a bottleneck effect. Four of the six major allele changes resulted in amino acid substitutions in r5804_P3 (Table 3). Additional passaging of r5804 did not introduce novel major allele changes (Fig. 6C, P3 vs P6).

Only a single mutation in N (position 1662) was shared by both adapted viruses, resulting in an amino acid change in N from glycine to arginine at position 519 (Fig. 6E; Tables 2 and 3). Interestingly, this mutation occurred de novo during the passaging, as this allele variant was undetectable in both input viruses (Fig. 1A, 3A, E, 6A, and C).

In summary, the detailed analysis of major allelic evolution of the Vero cell-adapted 5804 indicates that pre-existing alleles at low frequencies favor rapid adaptation. This genetic memory can be efficiently erased by generating a recombinant derivative based on the consensus sequence of a genome population. This process reduces the speed of host adaptation, as it requires this virus to explore the genetic landscape through de novo mutations.

Position 519 of the N protein is an important determinant for pathogenesis

To test whether the 519R allele identified in both adapted viruses is important for pathogenesis, we generated novel recombinant CDVs carrying either allele. In the non-pathogenic r5804 (12), the glycine was substituted with arginine, generating r5804 N_G519R. Likewise, arginine was exchanged with glycine in the highly pathogenic r5804P (12), generating r5804P N_R519G. The pathogenesis of the new viruses was tested in vivo in ferrets and compared to their parental counterparts (Fig. 7). The introduction of the arginine residue from the pathogenic virus into the apathogenic virus did not result in a phenotypic change. As with the parental r5804, r5804 N_G519R did not cause clinical disease (Fig. 7A through E). We observed transient skin irritations around the mouth in only one animal infected with r5804 N_G519R (Fig. 7D). Despite the lack of causing clinical disease, both viruses caused the typical leukopenia (Fig. 7E) and replicated to similar titers in PBMCs (Fig. 7F).

Fig 7.

In vivo pathogenesis of CDV r5804, CDV r5804_N_G519R, CDV r5804P, and CDVr5804P_N_R519G. Three ferrets per group were infected with 1 × 10⁶ TCID₅₀ of the respective viruses. Animal data for infection with CDV r5804_N_G519R in mint; CDV r5804P in azure; and CDV r5804P_N_R519G in violet. Data of animals infected with r5084_P0 (black) are the same as presented in Fig. 4 and are shown here again for easier comparison with the other groups. (A) Survival curve. (B) Body temperatures during the infection period. Lines represent mean values within each group, and open symbols represent the values of individual animals. Dotted lines represent threshold values. Temperatures above 39°C are considered increased body temperature, and temperatures above 40°C are considered fever. (C) Body weight development during the infection period relative to the start weight. Lines represent mean values within each group, and open symbols represent the values of individual animals. Dotted lines represent threshold values. 5% is considered mild, and more than 10% is considered severe weight loss. (D) Development of rash in individual animals. Severity of rash is indicated by darker red scales, while white boxes indicate the absence of rash, black boxes indicate that an animal was dead, and gray boxes indicate that the respective animal was not further monitored at the indicated time point. (E) White blood cell count. (F) Development of PBMC-associated viral blood titer per 1 × 10⁶ PBMCs during the infection period. Lines are mean values, and error bars are standard deviations for each group.

In contrast, exchange of the positive arginine residue to glycine in the highly pathogenic r5804P caused an observable phenotypic change. While r5804P N_R519G still induced clinical signs in the form of localized rash (Fig. 7D), transient fever (Fig. 7B), leukopenia (Fig. 7E), and weight loss (Fig. 7C) in the second week after infection, the severity of these signs was reduced compared to those induced by the parental r5804P (compare violet and green lines). Both r5804P and r5804P N_R519G caused transient leukopenia (Fig. 7E) and resulted in similar PBMC titers at peak viremia at day 7 post-infection (Fig. 7F). However, due to the generally milder manifestation of clinical disease observed in r5804P N_R519G-infected animals, these animals did not reach the clinical end points of the study and thus were not euthanized, while two out of three animals infected with the parental r5804P had to be euthanized (Fig. 7A).

Taken together, these results indicate that arginine at position 519 in N protein is one determinant for CDV pathogenesis and disease severity. Exchange of this residue to glycine results in reduced severity of clinical disease. However, arginine at position 519 in N protein alone is not sufficient to enhance the virulence of an attenuated CDV strain.

DISCUSSION

Diverse genome populations allow RNA viruses to adapt rapidly to new environments and hosts (26). Morbilliviruses subsequently infect immune and epithelial cells during a host invasion (35) and exhibit rapid selection of optimized genome subpopulations in different cell types (30). In addition, some morbilliviruses such as CDV infect a wide range of host species and can adapt more readily to new hosts (6). Here, we asked to which extent genome diversity contributes to the rapid adaptation of an in vitro-attenuated CDV strain to ferrets. For this, we passaged a Vero cell-adapted, attenuated CDV isolate (5804), which possessed a highly diverse genome population, and its low-diversity recombinant derivative (r5804) three times in ferrets and performed RNA sequencing of both viruses after each passage. The higher diversity of the Vero cell-adapted virus was associated with a faster gain of virulence compared to the recombinant virus. Several non-synonymous mutations in both viruses were either acquired or selected during in vivo passaging, but they shared only a single mutation in the carboxy-terminal tail region of nucleoprotein (N_tail). By applying reverse genetics, we show that this common mutation modulates CDV pathogenesis in ferrets.

Genetic diversity facilitates adaptation to a new environment

The Vero cell-adapted input virus immediately replicated efficiently in lymphatic tissues and then infected epithelia in the first host. The genome population of this virus was highly diverse, as measured by its mean Shannon entropy, and its diversity increased gradually with each passage, as determined by the RMSD of each passage from the input inoculum. The gain in genetic diversity correlated with increased virus titers and pathogenic properties in vivo. In contrast, the recombinant virus initially replicated inefficiently and was detectable in PBMCs and lymph nodes at very low titers, completely failing to spread to airway epithelia. The genome diversity determined as mean Shannon entropy of this virus was low and only reached a level comparable to that of the 5804 virus after the third passage. At this point, the recombinant virus exhibited a strong increase in viral load in the organs. These observations are consistent with several studies showing that higher genetic diversity of a virus inoculum is a strong determinant of its relative fitness (36–38) and can modulate viral pathogenicity (28, 39). This suggests that CDV requires a threshold genome diversity to acquire the ability to spread efficiently in its host. Consistent with this, we have recently shown that the quasispecies genome population of the closely related MeV alternates between lymphatic and epithelial tissue-adapted viruses (30).

In case of the unpassaged, recombinant r5804, the low genome diversity and limited intra-host replication may form a bottleneck for adaptation (40). The rapid fixation of a few de novo mutations in this genome suggests that these mutations may be key drivers to overcome this bottleneck. Due to the limited virus yield in the organ homogenates of the first two passages, we had to amplify these viruses in VerodogSLAMtag cells one additional time before infecting the next group of ferrets. While we are aware that the evolutionary path of r5804 is skewed by this additional amplification in tissue culture, and it may have allowed the virus to accumulate additional de novo mutations, reversion to virulence is still unlikely. Typically, the passaging of viruses on Vero cell derivatives or other cell lines leads to attenuation, but it requires many passages (41). However, this fact also strengthens our hypothesis that low genetic diversity of the recombinant virus causes over-attenuation in vivo.

The consensus sequences of the two starting viruses differed in several positions. Large-scale sequence analysis of CDV sequences deposited in GenBank revealed that the alternative alleles found in 5804_P0 and r5804_P0 reflect the diversity of different CDV strains. Remarkably, most of the variable sites in either virus did not mutate further during the passaging experiment, indicating that they are likely inert with respect to the gain in pathogenesis. One exception may exist at position 9436 in the non-recombinant virus. At this position, G was present at 37% frequency and A at 63% in 5804_P0, whereas G was present at 100% in r5804_P0. In both 5804_P3 and r5804_P3, the G allele was fixed. Thus, the G allele in position 9436 corresponding to amino acid residue 136 in the L protein may enhance CDV pathogenesis but alone is not sufficient to explain the increase in pathogenesis through the adaptation process.

Genetic memory accelerates adaptation to new host species

In live-attenuated rabies vaccines, low-frequency variants maintaining the wild-type sequence have been found throughout the adaptation process (42), and there is evidence for genetic memory of specific cell types and anatomic sites in human immunodeficiency virus (43). Thus, genetic memory that encompasses imprints of selective pressures from previously encountered environments (31) may exist within the CDV quasispecies. For MeV, epithelial cell-adapted genomes retained at low frequencies after in vitro adaptation to lymphatic cells have provided evidence that genetic memory contributes to morbillivirus adaptation (30). In addition, the ability of paramyxoviruses to package more than one genome per particle (44) specifically enables the preservation of low-fitness variants within a population. Our study reproduced the emergence of the same set of adaptive mutations as a passaging experiment performed more than 20 years ago (12). We were now able to provide evidence that these mutations were present at low frequencies in the input virus, which we interpret as the genetic memory of this virus isolate. In contrast, the recombinant virus acquired a different set of mutations and needed additional passaging in VerodogSLAMtag cells to achieve sufficient titers for the infection of more animals, even though both viruses were adapted to identical environments. This demonstrates the importance of genetic diversity and memory for adaptation to a new host.

The dose given to animals also plays a deciding role in adaptation. Virus populations transferred during natural transmission (e.g., aerosol or direct contact) impose stringent bottlenecks that greatly reduce the genetic diversity of these inocula (5). Under these conditions, the potential for adaptation is likely very limited. In our study, we used homogenized lungs, lymph nodes, and spleens from infected ferrets and infected successive animals with a large dose (10⁶ TCID₅₀). This has several effects that differ substantially from natural routes of transmission. Most importantly, by combining lymph nodes, lung, and spleen homogenates and administering a larger dose, a much larger pool of genetic diversity can be transferred, which could in turn lead to the selection of rarer genomic variants that would not otherwise appear in natural inocula. However, CDV transmission can also occur through urine or direct contact, which also allows easy high-dose host-to-host transmission (4–6). While highly unlikely to occur with MeV, direct transmission from infected immune organs by predation may also be a real-life possibility for CDV (4).

We also conclude that generating a recombinant virus by reverse genetics based on a consensus cDNA sequence erases the genetic memory of a genetically diverse virus isolate. This is an important safety aspect for the development of new live-attenuated vaccines. Classically, live-attenuated vaccines, such as those against measles and canine distemper, are generated by in vitro passaging of wild-type viruses (41, 45), leading to the accumulation of attenuating mutations. However, low-frequency wild-type alleles may allow these viruses to revert to virulence quickly (45). It should be noted that live-attenuated measles and CDV vaccines have been used successfully in humans and animals, respectively, for more than 60 years, and there has been no evidence that they have reverted to virulent phenotypes. However, in the ever-evolving process of developing safer vaccines against morbilliviruses and other pathogens, our study supports the use of recombinant viruses based on the consensus sequence of attenuated viruses to enhance their safety profiles and further reduce risks associated with reversion to virulence. While probably not necessary for morbillivirus vaccines, this approach may be useful for the generation of vaccines against viruses with a higher propensity for reversion.

A positively charged residue in the N_tail domain is important for virulence

Different passaging experiments based on either high or low variability CDV genome populations in ferrets led to the emergence of the G519R mutation in the N protein. Isolated analysis of this variant in recombinant CDV backbones confirmed its important role in pathogenicity. Nearly all CDV strains analyzed here, including vaccine strains, encode a positively charged lysine or arginine at this position (Data S1). The glycine observed in the input 5804 viruses is rather atypical and indicates its involvement in the attenuation of this virus.

The mutation is located in the unstructured flexible carboxy-terminal N_tail domain (46), which, for MeV, is described as the site of interaction with the viral P and the M proteins (47, 48). It contains three short motifs referred to as box 1–3 that are conserved among the Paramyxoviridae (47). The biological role of box 1 remains to be elucidated. Box 2 contains the P protein binding site (47, 49, 50), and the M protein interaction site is mapped to box 3 (48). This region comprises the nine carboxy-terminal amino acids of N [515-VYND(G/R/K)ELLN-523]. M protein binding to the N protein seems to depend on the negatively charged amino acids and two leucine residues (48, 51). Notably, a G in position 519 emerged as a compensation mutation after truncation of the spacer between box 1 and box 2 (52). The R519G substitution caused a loss of polymerase activity, indicating that residue 519 may be important for proper attachment of the viral polymerase to the nucleocapsid. Finally, reducing the length of the spacer between box 1 and box 2 in CDV r5804P has the same attenuating effect as introducing the R519G mutation in box 3 (53), indicating a functional connection between these motifs. Taken together, our findings establish the importance of a positively charged amino acid at position 519 of N protein for CDV pathogenesis.

Our study provides evidence for the importance of genetic diversity for the pathogenesis of an RNA virus by combining high-throughput sequencing technology with in vivo pathogenesis studies in an authentic animal host. As exemplified by the morbillivirus CDV, genetic memory exists in the form of low-frequency variants in the genome population and accelerates adaptation processes to a new host. Genetic memory can be efficiently erased by reverse genetics approaches to generate recombinant viruses from cDNA clones. This aspect is important for the safety evaluation of novel live-attenuated vaccines. Finally, our study identifies a crucial function of the viral nucleoprotein required for viral pathogenesis, which can be used to design novel vaccine candidates for CDV and related paramyxoviruses.

MATERIALS AND METHODS

Cells and viruses

VerodogSLAMtag (12) cells were maintained in Dulbecco’s modified Eagle medium (Sigma-Aldrich) supplemented with 5% fetal bovine serum (Thermo Scientific), 200 mM L-glutamine (Biochrom AG), 50 U/mL penicillin, and 50 µM Streptomycin (Gibco). The cells constitutively expressed the canine SLAM (cSLAM) receptor from the vector pCGdogSLAMtagZeo and were kept under selective pressure with the supplementation of 500 µg/mL Zeocin.

The Vero cell-adapted wild-type CDV isolate 5804 and a low passage of its recombinant derivative r5804 were used as input viruses for the passaging experiment (12). For this study, aliquots of the same stock from the 2003 study were used. Titers of virus stocks, homogenized tissues, and isolated peripheral blood mononuclear cells were quantified in VerodogSLAMtag cells by limited dilution method and expressed as 50% tissue culture infectious doses (TCID₅₀) as previously described (12).

Plasmid construction and recombinant viruses

The genomic cDNA plasmids p5804-N_G519R and p5804P-N_R519G were generated by the transfer of the SbfI/HpaI from p5804 and p5804P (12) into the shuttle vector pI.18 (13). The cloned shuttle vectors were subjected to mutagenesis PCR with primer pairs 5804 G1662A-sdm fwd (CCCGT CTATA ATGAT AGAGA GCTAC TCAAT T), 5804 G1662A-sdm rev (AATTG AGTAG CTCTC TATCA TTATA GACGG G) and 5804P A1662G-sdm fwd (CCCGT CTATA ATGAT GGAGA GCTAC TCAAT T), 5804P A1662G-sdm rev (AATTG AGTAG CTCTC CATCA TTATA GACGG G), respectively. The mutated sequences were reintroduced into their respective former genomic cDNA plasmid.

Recovery of recombinant virus and stock production

Recombinant viruses were generated by transfection of 1 × 10⁶ HEK-293T cells with 10 µg of the respective genomic cDNA plasmid and helper plasmids expressing measles virus N (500 ng), P (100 ng), and L (40 ng) proteins and T7 RNA polymerase (500 ng), as previously described (7). Transfected 293T cells were overlaid onto VerodogSLAMtag cells in 10-cm dishes, and syncytia were allowed to develop at 37°C. Single syncytia were then picked and inoculated into single wells of VerodogSLAMtag cells in 6-well dishes. Cytopathic effect was allowed to develop at 37°C until approximately 80% of the well was affected. The culture medium was then removed, and the cellular material was scraped together in a minimal volume of OptiMEM and transferred to fresh VerodogSLAMtag cells in 15-cm dishes for virus stock amplification. We defined the virus stocks produced from these dishes as “Passage 1” or P1. All titers were determined using the TCID₅₀ method.

Animal experiments and assessment of virulence

Male and female ferrets (Mustela putorius furo) 16 weeks or older from the Paul-Ehrlich-Institute in-house breeding colony without prior immunity to CDV were used for all experiments. Consecutive passages were performed in two ferrets per group.

The initial group was inoculated with 1 × 10⁶ TCID₅₀ of 5804 or r5804 intranasally (200 µL volume) under medetomidine and ketamine anesthesia. Animals were sacrificed on day 7 after infection, which corresponds to the time of peak cell-associated viremia. Spleen, lymph nodes, and lungs were homogenized and titrated separately, and an even mixture of tissue homogenate-derived virus with an infectious dose of at least 1 × 10⁶ TCID₅₀ was used to inoculate the next group of animals.

In case of viral titers below 1 × 10⁶ TCID₅₀/mL, the tissue homogenate-derived virus was expanded once on VerodogSLAMtag cells, and animals were infected with 1 × 10⁶ TCID₅₀ of the resulting virus stock.

All ferrets were examined daily for clinical signs of CDV infection including fever, rash, activity, and depression. Twice a week, animals were weighed to monitor weight change, and blood was drawn from the superior vena cava under general anesthesia (medetomidine with atipamezole reversal). PBMC-associated viral titers were determined by limited dilution and expressed as TCID₅₀/10⁶ PBMCs. Total leucocyte counts were determined by 1:10 dilution of whole blood in 3% acetic acid and counted in a hemocytometer. Pathogenesis over the entire course of disease was examined in three to five ferrets per virus. The animals were monitored as described above until they recovered or reached pre-defined clinical end points and were then humanely euthanized.

RNA purification and RNA sequencing

For RNA sequencing, RNA was isolated from cell-associated infectious virus released from infected cells by freeze-thawing. A volume of 100 µL virus stock aliquots was processed with Trizol reagent using the Zymo Direct-zol RNA kit (Zymo Research Europe GmbH, Freiburg, Germany). To extract RNA from tissues, 100 mg pieces were homogenized in 350 µL RLT buffer and processed according to the instructions of the RNeasy Mini kit manual (Qiagen, Hilden, Germany).

For library preparation, 500 µg of purified RNA was processed with the Illumina TrueSeq StrandedTotal RNA kit (Illumina, San Diego, USA), including depletion of ribosomal RNA via Ribo-Zero Gold rRNA Removal kit (Illumina, San Diego, USA) following the manufacturer’s protocols. Libraries generated with single-indexed adapters were then validated and sequenced on the NovaSeq/HiSeq with 2 × 150 base pairs (Genewiz, Leipzig, Germany).

RNA sequencing analysis

Adapter sequences were removed from raw fastq files using SeqPurge version 0.1-1000-gb4d1b1c (54). Quality trimming was disabled (-qcut 0 -ncut 0), and adapter recognition sequences (read 1: GATCG GAAGA GCACA CGTCT GAACT CCAGT CACNN NNNNA TCTCG TATGC CGTCT TCTGC TTG; read2: AGATC GGAAG AGCGT CGTGT AGGGA AAGAG TGTAG ATCTC GGTGG TCGCC GTATC ATT) were provided. Quality trimming and filtering were performed using Atropos version 1.1.17 (55) with minimum base quality cutoffs of 30 at both sides of the read (-q 30,30) and discarding reads with Ns and a length below 30 bp after trimming (-max-n 0 -m 30).

Mapping was performed with BWA mem version 0.7.12-r1039 (56) using default parameters unless stated otherwise. Host sequences in ferret- and cell culture-derived samples were removed by mapping quality-controlled reads against either the ferret (Mustela putorius furo, RefSeq assembly accession: GCA_000215625.1) or the green monkey genome (Chlorocebus sabeus, RefSeq assembly GCA_000409795.2), respectively, specifying the minimum seed length (-k 31). Unmapped reads were extracted using samtools version 1.7 (57) and bamToFastq version 2.17.0 (58) and subsequently mapped to the CDV reference genome.

The CDV reference was obtained by mapping 5804 used for inoculation reads against the CDV NCBI reference sequence (RefSeq assembly GCA_000854065.1) and extracting the majority consensus sequence.

Host-free alignments were deduplicated using Picard Tools MarkDuplicates (http://broadinstitute.github.io/picard) and left aligned using GATK LeftAlignIndels version 4.0 (59). We obtained an average read coverage of at least 5,500 reads per site for the two input viruses. Animal-derived samples were sequenced to a mean depth of at least 1,800 reads per site and a median of 1,400 reads per site. Due to the low virus load in the tissues of the first r5804 passage, only a mean read depth of 600 reads per site was achieved. Variant calling was performed with LoFreq version 2.1.3 (60) using default parameters and discarding variants with minor allele frequencies below 1%.

Calculation of diversity and deviation

The diversity of all samples was calculated using the formula for Shannon entropy provided by Gregori et al. (61).

$${\displaystyle {\displaystyle H_S\left(p\right)=-\sum _{i=1}^H{p}_i\log \left(p_i\right)}}$$

Toward this, variant frequencies for each allele at every site (pi) were entered in the formula.

To calculate the mean Shannon entropy, the sum was divided by the genome length of 15,690 base pairs. For the analysis of divergence to the reference sequence, variant sites throughout the genome were compared to the same site in the respective Vero cell-adapted or recombinant inoculum. To this end, the formula for root-mean-square deviation provided by Li et al. (62) was used. P[X_i,p] and P[Y_i,p] are the frequencies for each nucleotide (p) for the sample and the reference (X and Y) at each genome position (i). Since we did not call indels in our sequencing data, the formula was modified to include only nucleotide substitutions.

$${\displaystyle {\displaystyle \mathrm{R}\mathrm{M}\mathrm{S}\mathrm{D}{\left(X,Y\right)}_i=\sqrt{\frac{\sum _{p=\left(A,T,C,G\right)}{\left(P\left[X_{i,p}\right]-P\left[Y_{i,p}\right]\right)}^2}{4}}}}$$

To calculate the number of variant sites, all variants in one genome position were counted as one site. Afterward, variant sites were subdivided into major allele substitutions (frequency ≥ 50%), minor variants (frequency < 5%), and the remaining allele changes (frequency ≥ 5%–50%).

Statistical analyses

Generation of figures and statistical analyses were performed using GraphPad Prism 6.0 software. Statistical significance for viral kinetics was calculated using the two-way ANOVA, and P-values < 0.05 were considered significant.

DATA AVAILABILITY

The RNAseq data generated in this study have been deposited on the servers of the National Library of Medicine of the National Institutes of Health and are publicly available under BioProject accession number PRJNA1124512 and Gene Expression Omnibus Series GSE270448. The deposited data include raw fastq files deposited in the Sequence Read Archive (accession numbers SRR29428510–SRR29428518) and processed data (base count tables for CDV-aligned reads) deposited in the Gene Expression Omnibus database (accession numbers GSM8343105–GSM8343113).

ETHICS APPROVAL

Animal experiments were authorized by the Regierungspräsidium Darmstadt and performed in compliance with German and European regulations.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/jvi.00657-24.

Supplemental Data 1. jvi.00657-24-s0001.txt.

FASTA alignment of 220 full-length CDV sequences deposited in GenBank.

Supplemental legend. jvi.00657-24-s0002.pdf.

Legend for Supplemental Data 1.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.