The 3 Retroviridae classes I, II and III exhibit phenotypic differences that delineate a network never before reported between Ty3/Gypsy and Retroviridae LTR retroelements. This new scenario reveals how the diversity of vertebrate retroviruses is polyphyletically recurrent into the Ty3/Gypsy evolution, i.e. older than previously thought. The simplest hypothesis to explain this finding is that classes I, II and III trace back to at least 3 Ty3/Gypsy ancestors that emerged at different evolutionary times prior to protostomes-deuterostomes divergence. We have called this "the three kings hypothesis" concerning the origin of vertebrate retroviruses.

As already shown, gag polyprotein and the PR depict a new scenario as an alternative to the traditional monophyletic insight (2 groups of LTR retroelements) suggested by prior RT, RNAse H and INT analyses. Onto understand the two opposing scenarios, we performed phylogenetic analyses based on the RT, RNAse H and INT and found consistency with the traditional perspective of 2 separate LTR retroelement groups using the RT and RNAse H ([13-15]). Analysis of the INT revealed different perspectives depending on the NJ or parsimony method used in the analysis (see Methods). While the NJ method supports the 2 LTR retroelement groups, the parsimony method splits the Retroviridae into 2 branches not supported by bootstrap (data not shown). This is because our model of INT alignment covers the 3 subdomains described in the amino acidic architecture of a conventional INT domain. The traditional core used for inferring INT phylogenies is common to all INTs in general, and includes 2 of these sub-domains; the conserved zinc finger "HHCC" binding motif [38] at the N-terminus, and the central sub-domain containing the conserved D-D-E trait [39,40]. The C-terminal sub-domain of all INTs is usually dismissed from analysis because it is less preserved than the other 2 sub-domains. In Ty3/Gypsy and RetroviridaeINTs, it is definite that this sub-domain is a small trait called GPY/F module, which was probably recruited modularly during evolution [14]. The module name refers to the strongly preserved GPY/F amino acid motif [14], which will be referred to as the canonical motif throughout the rest of this paper. Indeed, the GPY/F module appears to be responsible of the signal discrepancy in phylogenetic analyses (INT parsimony tree performed without this module is in agreement with the NJ analysis, data not shown). From that point, we investigated the GPY/F module in relation to the 3 Retroviridae classes. The module, seen from this viewpoint, shows a number of protein isoforms based on GPY/F motif polymorphisms. With rare exceptions, the modules of class I INTs usually preserve the canonical motif, while the modules of classes II and III exhibit other variants (Figure (Figure7A).7A). Here, classes II and III do not make an intrinsic phenotypic distinction, each genus exhibiting a particular variant of the motif within these 2 classes. The modules coded by Ty3/Gypsy LTR retroelements delineate similar perspective. As shown in Figure Figure7B,7B, while the canonical motif is practically predominant in the modules of Ty3/Gypsy elements of plants and fungi, the modules of bilateria Ty3/Gypsy LTR retroelements are rich in motif polymorphisms (canonical motif included). This indicates that Ty3/Gypsy LTR retroelements described in bilateria organisms exhibit greater GPY/F motif variability than their Ty3/Gypsy counterparts of plants and fungi, and strongly suggests a number of transitions from the canonical motif toward other states during evolution. This scenario is not completely consistent with gag and PR perspectives; for instance, while Micropia/Mdg3 modules preserve the canonical motif, the different Retroviridae genera belonging to class II exhibit different motif polymorphisms. Nevertheless, the GPY/F module relates the Retroviridae class I with Ty3/Gypsy LTR retroelements of plants and fungi through the common preservation of the canonical motif, while classes II and III can be related with bilateria Ty3/Gypsy LTR retroelements by an increase of the motif variability. In fact, the whole module of class I INTs appears to be more similar to those preserved by the INTs of chromoviruses (Figure (Figure8A)8A) and Athila and Tat clades (Figure (Figure8B)8B) than to those of classes II and III. Alignment between classes I and II reveals a dramatic loss of sequence information by class II during evolution (Figure (Figure8C).8C). The module carried by spumaretroviral INTs is similar to that of class I, but they greatly differ in the motif (Figure (Figure8D).8D). That is, spumaretroviral modules lost the GPY/F motif, substituting it with a highly diverged KT/SP motif. Again, this outlines an intriguing parallelism between spumaretroviruses and Ty3/Gypsy errantiviruses because the modules of these 2 LTR retroelement lineages are qualitatively similar (Figure (Figure8D).8D). Moreover, Ty3/Gypsy errantiviruses also lost their GPY/F motif during evolution. Therefore, whatever the INT function involving the GPY/F module coded by the Retroviridae class I, this class appears to be a molecular fossil preserving GPY/F module phenotypes that were predominant among Ty3/Gypsy ancestors, predating the split between plants fungi and animals. In contrast, Retroviridae classes II and III maintain a number of module isoforms more recently emerged during evolution.