Abstract
A phylogeny has been calculated by maximum likelihood comparisons of the concatenated consensus protein sequences of 29 tobamoviruses shown to be non-recombinant. This phylogeny has statistically significant support throughout, including its basal branches. The viruses form eight lineages that are congruent with the taxonomy of the hosts from which each was first isolated and, with the exception of three of the twenty-nine species, all fall into three clusters that have either asterid or rosid or caryophyllid hosts (i.e. the major subdivisions of eudicotyledonous plants). A modified Mantel permutation test showed that the patristic distances of virus and host phylogenies are significantly correlated, especially when the three anomalously placed viruses are removed. When the internal branches of the virus phylogeny were collapsed the congruence decreased. The simplest explanation of this congruence of the virus and host phylogenies is that most tobamovirus lineages have co-diverged with their primary plant hosts for more than 110 million years, and only the brassica-infecting lineage originated from a major host switch from asterids to rosids. Their co-divergence seems to have been 'fuzzy' rather than 'strict', permitting viruses to switch hosts within major host clades. Our conclusions support those of a coalesence analysis of tobamovirus sequences, that used proxy node dating, but not a similar analysis of nucleotide sequences from dated samples, which concluded that the tobamoviruses originated only 100 thousand years ago.