Genotypes 3 and 4 are enzootic and zoonotic and can infect a number of species 
. Adaptation of each strain to a range of hosts 
should lead to a greater demand for genetic changes in the genome. Indeed, genotypes 3 and 4 were found here to have the higher fraction of variable sites in ORF1 and ORF2.N, even though most polymorphic sites were negatively selected. A larger fraction of invariant sites in genotype 1 leads to reduced intra-genotype heterogeneity and most probably reflects adaptation to a single host. While geographic constraints and isolation should lead to genetic diversity the lower geographic diversity with greater number of hosts and higher genetic diversity seen in genotype 4 versus genotype 1 may reflect a higher genetic diversity for adaption to a greater number of hosts in genotype 4 
Until recently, HEV was only known to infect humans, swine, deer, boar and avians 
. These observations suggested a simple model of evolution, in which mammalian HEV infected humans and artiodactyls (swine, deer and boars). The discovery of HEV variants in rabbits 
and rats 
indicates, however, a more complex model of evolution for HEV. Using HEV sequences from avian and rat to root genotypes 1–4, we showed that HEV lineages can be split into two clades, genotype 1/2 and genotype 3/4, or anthropotropic and enzootic HEV genotypes, although this result should be viewed with some caution because of the single genotype 2 sequence used in the calculation. Divergence time analysis shows that the ancestor of genotypes 1–4 split into anthropotropic and enzootic genotypes about 536 to 1344 years ago (range; 536 to 865 years ago, ORF2.N and 816 to 1344 years ago, ORF1). It is not possible to order this split to determine whether the ancestor was anthropotropic or enzootic, although the rooting of these sequences with a sequence from HEVinfecting a rat suggests the ancestor was enzootic. The split which led to human and swine variants occurred at about the same time. The TMRCA for the avian/mammalian HEV ancestor and for the rat/[human/swine] HEV ancestor suggests that the most ancient split resulted in avian and mammalian variants. The sequence from HEV infecting a rat suggests that mammalian HEV has adapted to different mammalian species over time, and there may be mammalian intermediates leading from the avian/mammalian ancestor to humans and swine, which have not been discovered. It should be noted, however, that addition of novel HEV sequences may affect estimates for TMRCA, if these sequences would affect genetic heterogeneity of HEV genotypes.
There is a disparity between the estimated TMRCA values for modern HEV and the suggested evolutionary history. Intuitively, the estimated TMRCA values should be greater than these estimates. The disparity in ORF1 may, in part, be caused by the removal of the polyproline region, although its evolution cannot be accurately estimated by the methods used. A recent analysis of mitochondrial genomes shows that calculated divergence times were two to six-fold shorter than the true dates 
. This effect may also hold for viral RNAs. Such apparent underestimate in viral TMRCA has been recently discussed 
. It was suggested that underestimates of viral divergence times may be caused by limitations of the models to adequately reflect evolutionary events. It is possible that virus-host cospeciation may have resulted in lower substitution rates over the long run. It is also possible that viruses have histories dating back tens of thousands to millions of years but early members have gone extinct and been replaced by the modern variants; thus, divergence times only estimate times to the appearance of these more modern variants. The time for origination of more extensive contacts between humans as well as between humans and swine also suggests a more ancient TMRCA for the appearance of HEV genotypes 1–4 as swine were domesticated about 11,000 years ago 
and urbanization started about 7,000 years ago 
. The most direct way to resolve this issue is to obtain viral RNA from ancient humans or animals infected with HEV as has been done recently for other pathogens 
and to create a more diverse sampling of HEV sequences from around the world.
A further examination of the individual genotypes using ORF2.N sequences shows that each exhibits its own distinctive population dynamic. The effective number of infections associated with the HEV genotype 1 increased within the last 35 years (~1970) and has been stable for the last twenty years (). Genotypes 3 and 4 showed increases around 1940 to 1945 followed by decreases around 1990 ( and ). This increase in population size coincided with World War II and may have resulted from the movement of some human populations from urban settings to more rural settings 
or more lax sanitation procedures at that time. A potential confounding factor to this conjecture is lowered consumption of pork in Japan during and immediately after the war. However, the effective population size for HEV is related to the rate of exposure rather than to meat consumption. Indeed Tanaka et al
. showed that the effective population of HEV increases throughout the war. They believe the increase in HEV was related to the importation of infected pigs from England in 1900 followed by the growth of the population of imported pigs. Transmission to susceptible native swine should further increase the effective population. The reason for the decrease in the effective populations of genotypes 3 and 4 starting about 1990 is unknown. This decline suggests that the emergence of HEV seen in recent years may be due to greater awareness of the HEV health problem in the world and improved diagnostics rather than an actual expansion of the HEV viral population.
Genotype 3 can be split into three clades; namely, 3.1, 3.2 and rabbit strains. The removal of the rabbit sequences does not modify the skyline plot when all other genotype 3 sequences were analyzed. The 3.1 and 3.2 clades show similar skyline plots with similar trends, although the 3.2 clade did not show the same levels of population increase and decrease as 3.1. This suggests that both genotype 3 clades have experienced similar evolutionary history even though they represent different geographic distributions, with 3.1 being predominantly represented with HEV variants detected in Japan and 3.2 in Europe (Table S1
Almost all the genotype 4 sequences used in the present study were recovered from HEV strains circulating in China or Japan. Thus, the evolutionary history of genotype 4 described here is the history of this genotype in these 2 countries. When the genotype 4 sequences were split into Chinese and Japanese sequences, different dynamics were observed. The Chinese sequences went through an increase in effective population over a period of about 55 years starting around 1890 to plateau over the last 65 years (). The Japanese sequences had a relatively constant effective population from 1900 to about 1990 and then decreased in size. These findings suggest significant differences in evolutionary processes acting in these two countries.
It is interesting to note that all 3 HEV genotypes, 1, 3 and 4, studied here experienced a population expansion. However, only genotypes 1 and 3 showed a rapid increase in the population size over 5–15 years ( and ), with genotype 1 rapidly expanding in late 1970, and genotype 3 in the middle of the 20th century. Genotype 4, however, exhibited a slow increase in the population size over the first half of the 20th century (). A similar slow increase from 1880 to the middle of the 20th century was observed for genotype 3 (). A dramatic decline in the population size for genotype 3 worldwide and genotype 4 in Japan over the last 15 years suggests a significant interruption in transmissions of these viral lineages, which could be associated with reduction in exposure. However, genotype 4 in China and genotype 1 do not show signs of decline, suggesting no dramatic changes in epidemiological processes acting on these lineages over the last 20–30 years. The country-specific HEV evolutionary history observed in this study most probably reflects temporal variations in rates of transmission and/or exposure for HEV strains of the same genotype circulating in different geographic regions. Analysis of population dynamics allows for distinguishing between the rate of detection and variation in the infected population size and presents an important tool for the identification of emerging diseases.
The dataset used in this study has several limitations that may potentially contribute to bias in the calculations presented here. First, the distribution of genotypic sequences is skewed primarily toward genotype 3 and 4 sequences, with only a single genotype 2 sequence being available. Second, the genotype 3 and 4 sequences are obtained mostly from HEV strains circulating in China and Japan, with only a limited number of sequences obtained from other parts of the world. Such geographic sequence representation could bias our analysis toward a representation of the history of HEV mainly in China and Japan. However, as was shown above, the skyline plots share a significant similarity between clades 3.1 and 3.2 of genotype 3, while sequences from these 2 clades have a noticeable difference in geographic origin. Nevertheless, the use of additional genotype 3 and 4 sequences, when available, outside China and Japan should help more accurate define the evolutionary history of HEV and most likely yield longer times to the most recent common ancestor for genotypes 3 and 4 and the genotype 3/4 ancestor. Third, there is no a known history or fossil records that can be used to confirm our analysis. The history of domestication of swine and the process of human urbanization; however, suggest that the Bayesian analysis presented here may be an underestimation of the evolutionary time for HEV.