Genus-Specific Substitution Rate Variability among Picornaviruses ^,

Abstract

Picornaviruses have some of the highest nucleotide substitution rates among viruses, but there have been no comparisons of evolutionary rates within this broad family. We combined our own Bayesian coalescent analyses of VP1 regions from four picornaviruses with 22 published VP1 rates to produce the first within-family meta-analysis of viral evolutionary rates. Similarly, we compared our rate estimates for the RNA polymerase 3D^pol gene from five viruses to four published 3D^pol rates. Both a structural and a nonstructural gene show that enteroviruses are evolving, on average, a half order of magnitude faster than members of other genera within the Picornaviridae family.

TEXT

Members of the Picornaviridae family are the most common cause of human viral infections in developed countries (28, 39). Human picornaviruses produce symptoms ranging from mild respiratory illness to hemorrhagic conjunctivitis, myocarditis, acute flaccid paralysis, and neonatal organ failure (19, 27, 28, 33, 40, 41). Veterinary picornaviruses, such as foot-and-mouth disease virus (FMDV), encephalomyocarditis virus (EMCV), and porcine teschovirus (PTV), can have devastating effects on livestock (5, 8, 21).

Although picornaviruses such as poliovirus (PV) are known to evolve more rapidly than other viruses with single-stranded RNA (ssRNA) genomes (9, 29), little research has been conducted to investigate how they evolve more rapidly than other viruses with similarly error-prone RNA-dependent RNA polymerases (29, 57) or if certain picornaviruses evolve more rapidly than others. Understanding the evolutionary potentials and constraints of these important pathogens is imperative for the development of durable vaccines and effective treatment plans for individual pathogens (42). As even small, 3-fold differences in RNA virus mutation rates can have dramatic consequences, such as driving a population into lethal mutagenesis (7), similar differences in long-term evolutionary rates could indicate significantly dissimilar evolutionary potentials.

While viral evolutionary rates were previously calculated only by linear regression, modern simulation software such as BEAST (11) allows for the estimation of more complex models of viral evolution. These Bayesian coalescent programs can produce both estimated mean rates of evolution and 95% credibility intervals (CIs) that provide a measure of the variability around mean rates. Instead of comparing single-point estimates, now nonoverlapping CIs provide the strongest evidence that genes or organisms are evolving at different rates (11). Many substitution rate estimates have been published for picornaviruses, especially for the antigenically significant VP1 gene, which encodes the most external of the picornavirus structural proteins and interacts with cellular receptors (37, 49). Based on sequence availability in GenBank, four novel analyses were conducted, measuring the rate of evolution of the VP1 gene for two enteroviruses and producing the first rate estimates for the type species of the genera Cardiovirus and Teschovirus. Fewer previous analyses and more limited GenBank data were available for other genes. We conducted five novel analyses of the rate of evolution of the 3D^pol polymerase gene for two enteroviruses and the type species of Aphthovirus, Hepatovirus, and Parechovirus.

Partial VP1 gene sequences of two human enterovirus B serotypes (coxsackievirus B2 [CVB2] and CVB4), encephalomyocarditis virus, and porcine teschovirus, of the genera Enterovirus, Cardiovirus, and Teschovirus, respectively, were obtained from GenBank (Table 1; Fig. 1). Partial 3D^pol gene sequences of a human enterovirus A serotype (enterovirus 71 [EV71]), a human enterovirus C serotype (PV type 1 [PV1]), a foot-and-mouth disease serotype (FMDV-A), hepatitis A virus (HAV), and human parechovirus (HPeV), of the genera Enterovirus, Aphthovirus, Hepatovirus, and Parechovirus, respectively, were obtained from GenBank (Table 1; Fig. 1). Dates of isolation (years) were obtained from GenBank or from the paper that described the virus's isolation (2, 8, 13, 16–18, 22, 26, 31, 32, 35, 38, 45, 49, 52, 59, 61–63, 66, 68, 69). Dates are given in the taxon labels in supplementary figures S1 to S9. We excluded sequences from viruses that had been extensively passaged in the lab prior to sequencing. For each virus, partial gene sequences were manually aligned using Se-Al version 2.0a11 (A. Rambaut, Institute of Evolutionary Biology, University of Edinburgh, United Kingdom; http://tree.bio.ed.ac.uk/). Each alignment included VP1 or 3D^pol sequences from as many dated isolates as possible and was trimmed to preserve the reading frame. No recombination was detected by RDP 3.44 (43).

Table 1.

Alignments and results of substitution rate analyses of picornaviruses

Gene	Viral speciesa	No. of taxa	No. of ntb	Date range	Substitution model	Clock model	TMRCA (ybp)c	Substitution rate (×10−3)c	dN/dS
VP1	Human enterovirus B (CVB2)	51	231	1946–2008	GTR+I+Γ	Relaxed	72 (60–100)	5.27 (3.57–7.06)	0.04
	Human enterovirus B (CVB4)	110	300	1959–2007	TrN+I+Γ	Relaxed	75 (60–95)	5.73 (4.18–7.29)	0.03
	Encephalomyocarditis virus	27	210	1986–2008	K80+I	Strict	247 (98–824)	1.61 (0.56–2.78)	0.03
	Porcine teschovirus	46	702	1957–2007	GTR+I+Γ	Strict	507 (205–832)	1.62 (0.63–2.75)	0.10
3Dpol	Human enterovirus A (EV71)	153	999	1986–2010	TN93+I+Γ	Relaxed	120 (97–139)	5.53 (4.29–6.67)	0.05
	Human enterovirus C (PV1)	51	534	1982–2006	GTR+I+Γ	Relaxed	45 (31–69)	11.68 (8.12–14.53)	0.03
	Foot-and-mouth disease virus (FMDV-A)	24	642	1971–2009	GTR+I+Γ	Strict	152 (33–424)	1.45 (0.70–2.24)	0.05
	Hepatitis A virus	25	402	1976–2008	TN93+I+Γ	Strict	746 (429–995)	0.89 (0.46–1.31)	0.03
	Human parechovirus	142	657	1975–2009	GTR+I+Γ	Relaxed	124 (88–212)	2.96 (1.88–3.92)	0.04

^aAbbreviations are given for enterovirus and aphthovirus serotypes.

^bnt, nucleotides.

^cMean substitution rates are shown with lower and upper 95% credibility interval bounds in parentheses.

Fig. 1.

Representations of the nine picornavirus genomes analyzed (∼7 to 8 kb; ViralZone, Swiss Institute of Bioinformatics). Shading indicates portions of the VP1 or 3D^pol gene used in the study from the following viruses: coxsackieviruses B2 and B4 (CVB2 and CVB4), encephalomyocarditis virus (EMCV), porcine teschovirus (PTV), enterovirus 71 (EV71), poliovirus type 1 (PV1), foot-and-mouth disease virus type A (FMDV-A), hepatitis A virus (HAV), and human parechovirus (HPeV).

Modeltest version 3.7 (55) determined the best-fitting nucleotide substitution model for each alignment (by Akaike's information criterion). Estimated nucleotide substitution rates and maximum clade credibility (MCC) trees for each alignment were obtained using BEAST (11). Three demographic models (constant population size, exponential growth, and Bayesian skyline) and both strict and uncorrelated lognormal relaxed molecular clock models were used as priors in simulations. The marginal likelihoods of these six analyses were compared using Bayes factors in Tracer version 1.5 (http://tree.bio.ed.ac.uk/software/tracer/). To ensure accuracy, two independent 200-million-chain runs (of four Markov chain Monte Carlo [MCMC] chains each) were performed for each set of priors. An additional control simulation with an empty alignment and the best-fitting priors was performed to ensure that the priors alone were not determining the results. All of our controls indicated that our BEAST results were informative and reproducible. A comparison between the selected MCC trees and bootstrap-supported (1,000 replicates) maximum likelihood (ML) trees created with the respective best-fit models of nucleotide substitution using PAUP* version 4.0b8 (D. L. Swofford, Sinauer Associates, Sunderland, MA) showed largely consistent relationships among isolates (see Figures S1 to S9 in the supplemental material).

The relaxed molecular clock was the best-fitting prior for CVB2, CVB4, EV71, PV1, and HPeV (logBF > 10 [BF is Bayes factor]), but the strict molecular clock was preferred for EMCV, PTV, FMDV-A, and HAV (logBF > 1.8; Table 1). The constant demographic model was preferred for FMDV-A, HAV, and HPeV (logBF > 2), the exponential model was the best fitting for PV1 and EV71 (logBF > 2), and the Bayesian skyline model was the best fitting for the remaining viruses (logBF > 2). The times to the most recent common ancestor (TMRCA) varied from very short time scales for the enterovirus serotypes (all of their 95% CI ranges coalesce within 140 years before the present [ybp]) to hundreds of years for the species-level analyses of EMCV, PTV, and HAV (Table 1). The faster-coalescing enterovirus genes have higher substitution rates than both VP1 and 3D^pol of the nonenteroviruses, with nonoverlapping credibility intervals (Table 1).

A literature review yielded 22 picornavirus partial and full VP1 substitution rates, summarized in Fig. 2. Published VP1 substitution rates for coxsackievirus B5 (CVB5), echovirus 9 (E9), echovirus 11 (E11), echovirus 30 (E30), HAV, and HPeV were obtained via BEAST analyses similar to those used in this study (15, 23, 34, 46–48); those for EV71, FMDV-A, and FMDV-O were obtained via analyses performed in TipDate (58), a precursor to BEAST (29); and the remaining rates were estimated via linear regression (3, 4, 44, 50, 66, 70, 71). These mean rates of enterovirus VP1 evolution range from 3.40 × 10⁻³ to 1.19 × 10⁻² nucleotide substitutions per site per year (ns/s/y), and mean VP1 rates for nonenteroviruses range from 9.76 × 10⁻⁴ to 2.79 × 10⁻³ ns/s/y. The average of the 18 enterovirus mean rates was 6.50 × 10⁻³ ns/s/y (standard deviation [SD] = 2.61 × 10⁻³), while the average of the eight nonenterovirus mean rates was four times lower at 1.60 × 10⁻³ ns/s/y (SD = 5.33 × 10⁻⁴). The only overlap of substitution rates between enteroviruses and all other picornaviruses occurs as the upper boundary of the HPeV 95% CI overlaps with the estimates for several enteroviruses, coxsackievirus A16 (CVA16), CVB2, E30, EV71, and swine vesicular disease virus (SVDV).

Fig. 2.

Comparison of the VP1 nucleotide substitution rates of CVB2, CVB4, EMCV, and PTV (shown in bold) to published VP1 rates of other picornaviruses. Enteroviruses for which substitution rates are shown include serotypes coxsackievirus A16 (CVA16) (71), enterovirus 71 (EV71) (4, 29), coxsackievirus B5 (CVB5) (23), echovirus 9 (E9) (46), echovirus 11 (E11) (46), echovirus 30 (E30) (46, 47), coxsackievirus A24 (CVA24) (3), poliovirus type 1 (PV1) (50), enterovirus 70 (EV70) (64), and swine vesicular disease virus (SVDV) (70). Additionally, substitution rates for foot-and-mouth disease virus type A, type C, and type O (FMDV-A, FMDV-C, and FMDV-O) from the genus Aphthovirus (29, 44), hepatitis A virus (HAV) from the genus Hepatovirus (36, 48), and human parechovirus (HPeV) from the genus Parechovirus (15) are shown. Ninety-five percent confidence (EV71 and FMDV; calculated with TipDate) or credibility (all others; calculated with BEAST) intervals are shown where available. dN/dS ratios for the alignments that produced these rate estimates are shown in italics above, where available. Dashed line delineates the enteroviruses from other genera.

Only four 3D^pol substitution rates have been published, all from BEAST analyses, and from three different human enterovirus B (HEV-B) serotypes (47). The mean rates of enterovirus 3D^pol evolution range from 5.53 × 10⁻³ ns/s/y to 1.17 × 10⁻² ns/s/y, and mean rates for that of nonenteroviruses range from 8.89 × 10⁻⁴ ns/s/y to 2.96 × 10⁻³ ns/s/y. The average of the six enterovirus mean rates was 7.99 × 10⁻³ ns/s/y (SD = 2. 71 × 10⁻³), while the average of our three nonenterovirus mean rates was again more than four times lower at 1.77 × 10⁻³ ns/s/y (SD = 1.07 × 10⁻³). HPeV is again the only source of overlap, as its 95% CI includes the lower CI boundary of enteroviruses E9 and E30.

Overall, the mean rates of genomic evolution of both human and veterinary enteroviruses are consistently higher than those of members of the two other human-infecting genera (Hepatovirus and Parechovirus) and three veterinary genera in the Picornaviridae (Fig. 2 and 3). These higher rates are evinced despite a wide range of mean rates; there was more variability among mean rates of enterovirus evolution than among those of isolates from five other genera (the standard deviation for enterovirus rates was four times [VP1] and two times [3D^pol] greater than for the other genera). Our novel and collected VP1 rates are similar to published rates of the entire P1 structural region (15, 29, 30), and our 3D^pol rates echo that of an adjacent nonstructural region of a veterinary enterovirus (29). The striking similarity between evolutionary rates of structural and nonstructural picornavirus genes has not previously been discussed in the literature, but for each virus the rates were remarkably consistent with the estimated rates for both regions. Despite undoubtedly different selection pressures and potentially divergent evolutionary histories due to recombination (60), the VP1 and 3D^pol genes of individual picornaviruses appear to share an evolutionary rate. As more picornavirus whole-genome sequences become available, it will be interesting to see if additional genes support these evolutionary patterns.

Fig. 3.

Comparison of the 3D^pol nucleotide substitution rates of EV71, PV1, FMDV-A, HAV, and HPeV (shown in bold) to published 3D^pol rates of human enterovirus B serotypes echovirus 9 (E9), 11 (E11), and 30 (E30) (47). All are shown with 95% credibility intervals. dN/dS ratios are shown in italics above.

While previous investigations of RNA virus evolutionary-rate variability have been focused on the differences between viral families (29), our results demonstrate that significant long-term substitution rate variation can exist between related genera. The assumption that family members have identical rates underlies the recent estimations of long-term evolutionary rates of whole viral families, including those of the Luteoviridae (51), Potyviridae (20), and sobemoviruses (14). Our results discourage the continued presumption that evolutionary-rate differences among related genera or species must be negligible.

The immense serotype diversity of human pathogens within the genus Enterovirus (1) has created an imbalanced availability of sequence data within Picornaviridae. The genus Enterovirus is also by far the largest and most diverse genus in the family (2), which has allowed for more rate analyses of enteroviruses than of other genera. While all enterovirus substitution rates are of serotypes and most other available substitution rates are of species, the substitution rates of an enterovirus species should closely resemble the average of its serotype rates (30). However, our analyses showed that slower-evolving nonenterovirus species have much longer TMRCAs, which means they have had more time to become saturated at synonymous positions. Saturation reduces long-term substitution rate estimates and could explain some of the difference between fast-evolving enterovirus serotypes and the lower rates for species. Among the veterinary picornaviruses, three aphthovirus serotypes (FMDV-A, -C, and -O) still evolve significantly more slowly than the enterovirus serotype (SVDV). This serotype-to-serotype comparison confirms that the higher evolutionary rates of enteroviruses are not solely due to a difference in taxonomic scale (20).

Three common explanations for high per-year substitution rates are high per-generation mutation rates, high replication rates (increasing the number of generations per year), and positive selection (12). We tested whether our enterovirus alignments experienced positive selection, using the single likelihood, ancestor-counting, codon-based ML method on the Datamonkey web server (54). The estimated ratio of nonsynonymous to synonymous evolutionary changes (dN/dS ratio) for each of our alignments was very low (≤0.1; Table 1), similar to published picornavirus ratios (Fig. 2 and 3), indicating strong purifying selection on both the VP1 and 3D^pol genes. No codons were found to be under positive selection in our nine analyses. Purifying selection is common in long-term RNA virus evolution: 73% of nucleotide substitutions from 46 RNA viruses were synonymous (29). While it is evident that negative selection on picornavirus genes does not preclude high nucleotide substitution rates, we have no evidence that positive selection on enteroviruses could explain their higher rates of evolution than those of other genera.

Due to their error-prone RNA-dependent RNA polymerases (RdRp), ssRNA viruses are known for high mutation rates, which are on the order of 10⁻⁵ to 10⁻³ mutations per nucleotide per replication event (10, 12, 25, 30, 57). In the absence of selection, the per-year substitution rate is solely a function of the per-replication mutation rate, so higher mutation rates can directly translate into higher substitution rates. It is possible that enteroviruses have evolved higher mutation rates than other picornaviruses. It has been suggested that the unusually low substitution rate of hepatitis A virus is due to its having evolved a significantly lower mutation rate than other picornaviruses (6). Indeed, the enterovirus poliovirus has a very high mutation rate that can be lowered through mutations in the RdRp (53, 67), indicating that its mutation rate is evolvable. On the other hand, atypical substitution rates within the family have also been attributed to tissue tropism (48). Tissue tropism can be easily linked to a virus's replication rate, as viruses that infect slowly dividing tissue will have a lower replication rate than viruses that infect rapidly dividing tissue (48). Because affinity for the gastrointestinal tract is a defining characteristic of enteroviruses, it is possible that higher enterovirus substitution rates are due to their primary tropism for intestinal tissue, which has the highest turnover rate of all adult mammalian tissues (24, 65). This speculation is also supported by the fact that the fastest-evolving nonenterovirus, HPeV, frequently infects enteric tissue (15, 56) and was once classified as an enterovirus (15, 56). The mechanistic basis of the high enterovirus substitution rates remains an area of future research.