Episodic evolution of RNA viruses.

Abstract

The molecular clock and the neutral allele theory have been intimately intertwined since both were conceived in the 1960s (1-3). The molecular clock embodies the observation that rates of molecular are nearly constant; the neutral allele theory provides a mechanistic explanation for the constancy. Like most relationships, that between the neutral theory and the molecular clock has had its share of problems. Ironically, in the same year that Kimura and Ohta stated that Probably the strongest evidence for the [neutral allele] theory is the remarkable uniformity for each protein molecule in the rate of mutant substitutions in the course of evolution (4), they also showed that the rate of substitution was not, in fact, constant (5). Their and subsequent studies have shown that the ratio of the variance to the mean number of amino acid substitutions is 2.5 to 7.5. The neutral theory predicts that the ratio should be one. For some, this is sufficient to reject the neutral theory for protein evolution. For others, it is not. Kimura, for example, thinks that 2.5 is close enough to one that those who reject the theory are being picayunish (6). As this debate is in the realm of population genetics, it is naturally both acrimonious and esoteric. But some of the itssues are fundamental to the way we view molecular evolution. In particular, they are relevant to the conclusion that the phylogeny of the polymerase-associated phosphoprotein (P) of vesicular stomatitus virus (VSV) is an example of'positive Darwinian evolution as claimed by Nichol et al. (7) in this issue. In this remarkable study, the of the VSV is shown to exhibit variations in rates of substitutions that exceed anything seen to date. For example, one branch leading from the root of the VSV phylogenetic tree to a virus collected in Costa Rica in 1987 has 23 substitutions. Another branch from the (same) root to a virus collected in the United States, also in 1987, has 217 substitutions. Barring any complications (there are some; see below), the numbers of substitutions in these two branches should be approximately the same if substitutions are neutral. Equally remarkable is the complete lack of correlation between the year of isolation of a VSV virus and its relatedness to other isolates. In studies of other species, viruses tend to branch off of their phylogenetic tree in the temporal order in which they were isolated, just as they would were there a molecular clock. The order of branching for VSV reflects geography rather than time. Successive branches come off as the isolates come from more northern localities. Within a locality, there is sometimes little variation between individuals collected at widely spaced time intervals. For example, two isolates separated in time by 15 years, one from Guatemala and one from El Salvador, differed by only two nucleotides. Is such a pattern suggestive of positive Darwinian selection? This seems entirely reasonable. VSV is found in domestic mammals and in insect vectors. As the virus moved from its origins in Central America, it would have to adapt to life in different insect vectors. As the polymerase-associated phosphoprotein must function with the polymerase of the insect host, it would be expected to evolve rapidly and, quite possibly, in a way that reflects geography more than time. Could the neutral theory account for the data? Probably. The rate of substitution per generation under the neutral theory is equal to the mutation rate per generation (k = u). Thus, variability in rates of substitution could be due to variability in generation times and/or mutation rates. As far as I am aware, there are no studies directly examining such variation in VSV. But if the generation time of the VSV varies wildly-say, by 100fold in different localities-then the variation in the number of substitutions is entirely expected under the neutral model. If the virus should have a dormant phase, then even greater variation might be seen. The fact that some isolates separated by 15 years differ by only two nucleotides forces us to consider this possibility. Much has been done to remove the generation-time effect in the analysis of the variability of rates of substitution for eukaryotic genes (8, 9). This effect is typically lumped into a category called lineage effects. When all lineage effects are removed, significant stochastic variation in replacement substitution rates remains when comparisons are made between orders of mammals. There is a suggestion that silent rates may vary as well, although less so than replacement ates. The technique used to remove lineage effects requires data from several loci. A similar approach in VSV, as more loci become available, may allow us to rule out variation in generation times or mutation rates as a likely cause for the rate of variation. In a recent essay (10), Kimura has summarized the evidence for the neutral theory. One bit is relevant to VSV. If the rate of neutral is equal to the mutation rate (k = u), then RNA viruses should evolve about one-million times faster than their mammalian hosts, as the virus mutation rate is higher by that amount. Many RNA viruses do evolve that fast, supporting the neutral theory (as well as all other models that are mutation-rate limited). Moreover, many viruses, including VSV, have silent substitution rates that are higher than replacement (amino acid changing) rates. Thus, RNA virus looks like the more familiar eukaryotic DNA evolution, only it is much, much faster. Now, Nichol et al. (7) claim that the substitutions in the polymerase-associated phosphoprotein are selected. Although they do not discuss replacement and silent rates separately (other than to report that the former is 67% of the latter), they do express the view that most of the is due to natural selection. If so, then perhaps the rapid of silent sites is due to the need to match the codon usage of the virus to that of the everchanging insect vectors. The paper of Nichol et al. (7) is yet another example that conclusive proofs for or against the neutral theory are elusive, particularly when the inferences are indirect. Hopefully, the extreme variation in VSV substitution rates will motivate someone to look directly for functional effects of the substitutions. This, in the last analysis, is the only fully satisfying approach to learning the causes of molecular evolution.