Population Dynamics and Rates of Molecular Evolution of a Recently Emerged Paramyxovirus, Avian Metapneumovirus Subtype C

Abstract

We report the existence of two distinct sublineages of avian metapneumovirus (MPV) subtype C, a virus which has caused serious economic loss in commercial turkey farms in the United States. This subtype is closely related to human MPV, infects multiple avian species, and is globally distributed. The evolutionary rates of this virus are estimated to be 1.3 × 10⁻³ to 7 × 10⁻³ substitutions per site per year, and coalescent estimates place its emergence between 1991 and 1996. The four genes examined show a concordant demographic pattern which is characterized by a rapid increase in population size followed by stable population grown until the present.

Avian metapneumovirus (aMPV), in the subfamily Pneumovirinae of the family Paramyxoviridae, causes an acute respiratory disease characterized by nasal discharge, inflamed eyes, facial congestion, and swelling of the sinuses in turkeys and is the etiological agent of “swollen-head syndrome” in chickens (35, 54). The disease is more severe in the presence of coinfection with bacteria and Newcastle disease virus (27, 52). aMPV contains a nonsegmented, single-stranded, negative-sense RNA genome with the gene order 3′-leader-N-P-M-F-M2-SH-G-L-trailer-5′ (31). Based on genetic and serological properties, aMPV can be classified into four distinct subgroups (A, B, C, and D) (4, 28, 50). After its first appearance in South Africa in 1978 (8), isolates belonging to different subgroups of aMPV, mainly A or B, were reported in Israel (3), Mexico (14), Morocco (19), Jordan (42), Brazil (13), Japan (34, 49), and many European countries (2, 20, 23, 32, 35, 37, 38, 54). Although this virus has been circulating globally since 1978, there were no reports of aMPV in the United States until it was identified during an outbreak at a commercial turkey farm in Colorado in 1996 (43) and subsequently at commercial turkey farms in Minnesota (22, 45, 46). Based on phylogenetic and serological properties, these U.S. isolates are placed in subgroup C, a new subgroup of aMPV (5, 6, 9, 45, 46) which shows higher sequence similarity to human MPV (hMPV) than to other aMPV subgroups (21, 25, 33). Subtype C aMPV already has caused an estimated annual loss of $15 million to the turkey farms in Minnesota (22). Although aMPV subtype C recently emerged in commercial turkey farms, a survey conducted between 2000 and 2002 demonstrated that aMPV subtype C also infects wild birds (e.g., American coots, American crows, Canada geese, cattle egrets, and pigeons) in Georgia, South Carolina, Arkansas, and Ohio (53). Further, the appearance of aMPV subtype C outside the United States (in 1999 in France [51] and in 2005 in South Korea [30]) in different avian species suggests that this recently emerged subgroup of aMPV is geographically distributed across a wide range of avian hosts. Reports of aMPV during the seasonal outbreaks on several commercial turkey farms in and around Minnesota (22, 45, 46), together with the recent findings of aMPV subtype C in wild avian species (53) and its wide geographic distribution (30, 51), suggest that this subtype could be a potential threat to poultry industries around the world. In this paper, we provide a detailed investigation of the demographic pattern and rates of evolution of this virus and forces that promote its rapid dispersal into different hosts and regions.

We used Bayesian Markov chain Monte Carlo analyses based on the number and temporal distribution of genetic differences among viruses sampled at different times (15, 16) to estimate evolutionary rates and past population dynamics (18). This coalescent-estimate-based approach has been widely used to infer virus population dynamics (7, 29, 36). From a megablast (1) query, we retrieved all of the available aMPV subtype C nucleocapsid (N), phosphoprotein (P), fusion (F), and second matrix (M2; open reading frames M2-1 and M2-2) sequences (see Table SA in the supplemental material). Additionally, sequences obtained from clinical samples from turkeys reported for the M2 (12), N and P (11), and F (10) genes were also used. Sequences were aligned with DAMBE ver. 4.5.6 (55) and Mesquite ver. 1.12 (http://mesquiteproject.org). We reconstructed maximum-likelihood (ML)-based phylogenies with the nucleotide sequence data for each of the four genes. The hMPV sequence (accession no. AF371337) was used as an outgroup because it is more closely related to aMPV subtype C than to the other aMPV subgroups (25, 33). The appropriate nucleotide substitution model for each gene was selected by the hierarchical likelihood ratio tests implemented in Modeltest ver. 3.7 (41). ML analysis was carried out with the heuristic search option, implementing stepwise addition with 100 random addition replicates and tree bisection-reconnection branch swapping with PAUP* ver. 4.0b10 (47). With the same program, nodal supports were estimated with 100 nonparametric bootstrap replicates. The inferred tree was visualized with FigTree ver. 1.12 (http://tree.bio.ed.ac.uk/software/figtree/).

To test the hypothesis of topological congruence among the four genes and with the concatenated tree, we performed the phylogeny-based Shimodaira-Hasegawa test (44) implemented in PAUP. The concatenated tree was constrained and was tested with each individual gene tree. We performed the test on three different data set, (i) 16 unique sequences which include aMPV/MN2B, (ii) 15 unique sequences excluding aMPV/MN2B (which changed clade affiliations), and (iii) 14 unique sequences excluding aMPV/MN2B and aMPV/MN7 (which changed within-clade affiliations). The mean nucleotide distances among the lineages were estimated with MEGA ver. 4 (48). We used a Bayesian Markov chain Monte Carlo approach implemented in BEAST ver. 1.4.7 (17) to estimate evolutionary rates. The overall substitution rates (nucleotide substitutions per site per year) for the N, P, F, and M2 genes were estimated with the Bayesian skyline model, with a relaxed molecular clock (uncorrelated log-normal model) implemented in BEAST (15, 16). Previous studies have shown that the relaxed molecular clock is a better fit to the data than the strict clock (15, 24). Bayesian skyline plots with 10 grouped intervals were reconstructed to infer demographic history. Phylogenies were evaluated with a chain length of 20 million states under the GTR + Γ₄ substitution model, with uncertainty in the data reflected in the 95% high-probability density (HPD) intervals. Convergence of trees was checked with Tracer 1.4 (http://beast.bio.ed.ac.uk/Tracer). For selection analyses, the nucleotide alignment of each gene and the inferred unrooted ML tree were used as input. We used the ML approach implemented in CODEML (PAML package version 3.15) (56) to determine whether any of the codons in each genomic region of aMPV subtype C have evolved under positive selection. The likelihood ratio tests were used to compare models (M1a, M7, M8a) that assume no positive selection (ω < 1) with those models (M2a, M8) that assume positive selection (ω > 1) (57).

The N, P, F, and M2 gene trees presented here (Fig. 1A to D) clearly demonstrated the existence of two distinct clusters within U.S. aMPV group C. Concordantly, the number of fixed nucleotide differences is also consistent with the existence of two distinct subgroups (see Fig. SB in the supplemental material). Based on the nucleotide sequence data of the respective genes, the net percent nucleotide distances between the two sublineages at the N, P, and F genes are 2.56 ± 0.45, 3.26 ± 0.65, and 2.06 ± 0.38, respectively.

FIG. 1.

ML trees, based on the N (A), P (B), F (C), and M2 (D) genes, depicting the existence of two distinct sublineages within aMPV subtype C. The placement of isolate MN2B in the F gene tree differs from its position in the N, P, and M2 trees. ML-based nodal supports are indicated at the bases of the nodes. The trees are midpoint rooted for clarity only. The topological placement of isolates is consistent with the group affiliation when the tree was rooted with hMPV (accession no. AF371337).

The phylogenetic analyses have revealed that while isolate aMPV/MN2B belongs to sublineage C1 based on genes N, P, and M2, in the F gene tree this isolate belongs to sublineage C2. The phylogenetic discordance with regard to the placement of isolate aMPV/MN2B in the F gene tree indicates either that the sequence identification is incorrect, the animal was infected with multiple aMPV strains that were differentially sampled by gene-based PCR, or that this isolate is a possible recombinant. Our phylogeny-based Shimodaira-Hasegawa test confirms that aMPV/MN2B sequences significantly (P < 0.001) change the phylogenetic topology; when isolate aMPV/MN2B was excluded, the concatenated tree and the N and F gene trees were congruent (P > 0.05). When aMPV/MN7 was excluded from analyses, the N, F, and P gene trees were congruent with the concatenated tree (P > 0.05). Within-clade recombination event could be an explanation for the incongruence of the M2 gene tree with the remaining gene trees. Since most of the isolates analyzed in this study were derived by PCR of individual genes and not the full genome (45), each amplified gene might not represent the same viral strain. Therefore, we cannot conclude unambiguously whether recombination or coinfection is the likely factor causing the incongruence of the P, F, and N gene trees with the M2 gene tree. Nevertheless, phylogenetic analyses of each revealed the existence of two distinct sublineages, C1 and C2. Identification of recombinants among the negative-sense RNA viruses would be an important finding, and future efforts should focus on genome sequencing to enable the detection of recombination events.

All of the isolates from Colorado, where the first outbreak occurred in early 1996, belong to subgroup C1. Despite their different geographic origin and host (pheasant), the South Korean aMPV isolates were 99% identical to subgroup C1 isolates from turkeys in the United States. In contrast, most of the isolates from Minnesota are restricted to subgroup C2. Further, our analyses unequivocally suggest that the split within aMPV subtype C leading to C1 and C2 is neither host specific nor associated with geography because both clades are represented by isolates from a common host (turkey), as well as from the same region (Minnesota). The presence of Minnesota isolates in both subgroups C1 and C2 suggests two possible scenarios. Either (i) the Colorado isolates were introduced into the commercial turkey farm in Minnesota and the virus adapted rapidly to the new environment and hosts, or (ii) wild birds are the reservoir of aMPV subtype C and frequently reintroduce it into domestic avian species. It is likely that transport of poultry is partially responsible for the dissemination of this subtype because sequences from South Korean pheasants are closely related to those obtained from Colorado turkeys.

Given the ranges of the evolutionary rate estimates for the N, P, F, and M2 genes (Table 1), it is not possible to unambiguously conclude whether the rates of these four genes are different. Our results do place aMPV subtype C among the fastest-evolving RNA viruses based on reported evolutionary rates (26, 39, 40, 58). The recent emergence of aMPV subtype C with successful adaptation to new hosts and regions and large populations for virus expansion are possible explanations for such high evolutionary rates. There was no evidence that positive selection was acting on any of the genes analyzed (Table 1), as might be anticipated for a multihost, rapidly evolving virus.

TABLE 1.

Inferred evolutionary rates and TMRCAs for different genes of aMPV subtype C

Gene	na	Evolutionary rate (10−3/site/yr)b	TMRCA (yr)b	ωc
N	31	4.47 (1.91-7.04)	9.26 (9.00-14.56)	0.0924
P	32	7.01 (3.34-10.71)	9.48 (9.00-14.48)	0.3719
F	28	1.39 (0.56-2.37)	13.93 (9.00-29.95)	0.4171
M2	34	6.14 (3.59-9.00)	9.21 (9.00-10.90)	0.5009,
(M2-1, M2-2)				0.8273

^aSample size.

^bValues in parentheses are 95% HPD ranges.

^cω, ratio of nonsynonymous to synonymous substitutions.

Based on these evolutionary rates, the time to the most recent common ancestor (TMRCA), where both C1 and C2 coalesce, is estimated to be between 9 and 14 years (Table 1). Our analyses also indicate that the TMRCAs for the isolates within C1 and C2 are 9.22 (95% HPD, 9.00 to 10.40 years) and 7.04 (95% HPD, 6.37 to 8.24 years) years, respectively. Both measures are consistent with the epidemiological data (43). Despite the differences in the demographic pattern prior to 1999 (6 years before 2005) revealed by each gene (Fig. 2), the population history of each of the four genes shows a sudden increase in population size in 1999 and the population remained stable through 2005.

FIG. 2.

Bayesian skyline plots depicting population size in relation to time in years before 2005 for aMPV subtype C. Population size estimates (solid lines, Ne × g; the product of effective population size and generation time in years) are expressed on a logarithmic scale (y axis). The dotted lines give the 95% HPD intervals of the estimates. The green, yellow, blue, and red lines are the estimates based on the N, P, F, and M2 genes, respectively. All four of the genes revealed a consistent demographic history of the virus consistent with the timing of a sudden population expansion.

The large degree of demographic pattern concordance and the lack of positive selection inferred from different genes suggest that strong selective sweeps or recombination events are not responsible for high evolutionary rates. It is also apparent that none of the genes show any sign of a recent bottleneck or decline in population size (Fig. 2). These lines of evidence suggest that the preventive measures undertaken to date are not effective enough to constrain viral population dynamics. Nevertheless, the relatively stable population size of aMPV and the emergence of genetic subtypes in a recent Minnesota outbreak indicate that there is a reservoir for future epidemics in domestic poultry and potential spread to a wide range of avian species. Further, the close phylogenetic relatedness of aMPV subtype C (25, 33) with hMPV indicates that this clade may also have zoonotic potential.