In their very interesting work on the molecular evolution of MXB (also known as MX2) in primates, Mitchell et al. identify the N-terminal portion as a major target of positive selection. This is in contrast with our previous report indicating that, in eutherian mammals, most positively selected sites are located in MX2 loop 4 (L4) (Sironi et al., http://mbe.oxfordjournals...). Mitchell et al. suggest that our data are affected by two problems: 1) the use of highly divergent sequences, leading to dS saturation and an increased rate of false positives, and 2) the failure to account for recombination together with the inclusion of pseudogenized MX2 sequences. Moreover the authors state that it is unclear which species we included in our work.
We thus revisited the analyses we previously reported (Sironi et al., http://mbe.oxfordjournals...).
First of all, we acknowledge that Supplementary Table S1 (with the species list) was incorrect and the species did not correspond to those we included in the analysis (we are in the process of amending this with the Journal). Indeed, the species we analyzed are those reported in the tree in figure 1. Thus, rodents were not included, nor was cat. Based on Mitchell's data, on those published by Braun et al. (http://www.pnas.org/conte...), and on careful inspection of all sequences, we only included two species with pseudogenized MX2 genes: the dolphin and killer whale. At the time of our analysis the MX2 sequence for the killer whale was only partially available and we imputed one nonsense codon to sequencing errors. All other species we analyzed have intact MX2 coding sequences. In our original analysis, we had also screened the alignment using GARD and detected no recombination breakpoint.
In order to assess the effect of including the pseudogenized MX2 sequences, and to exclude the possibility of undetected recombination or gene conversion events in the N-terminal portion, as described in Mitchell et al., we have now rerun the evolutionary analyses. Specifically, we focused on an alignment portion corresponding to codons 500-715 (comprising part of the stalk domain, L4 and the C-terminal BSE) and removed the killer whale and dolphin sequences. We obtained statistical evidence of positive selection using the M1/M2 and M7/M8 tests. In our previous analysis we had identified 7 selected sites in this region. Six of them were confirmed by BEB (from model M8, codon frequency model: F3x4) in this revised dataset and five were also detected by MEME (BEB and MEME sites: 518S, 600L, 613F, 617E, 635L, amino acids and numbering refer to the human sequence). We thus consider that the presence of recombination and the erroneous inclusion of two non-functional MX2 gene sequences were not responsible for our detection of positive selection in the L4 region. In fact, after accounting for these effects and under the most conservative scenario (i.e. only considering sites detected by both BEB and MEME), we still detect 3 positively selected sites in the 42 amino acid-long loop 4, confirming that this region represents a preferential selection target in mammals.
With respect to the issue of dS saturation, the total tree length in our original analysis was 7.1 (whole gene) and it amounts to 8.15 in this new analysis (C-terminal fragment). Based on the simulation analyses presented by Anisimova et al. (http://mbe.oxfordjournals...), Bayes prediction using model M8 is accurate for trees of similar length (tree length = 8.44 in their simulations); indeed, Anisimova and coworkers conclude that Bayes prediction is tolerant of multiple substitutions at the same site. Thus, we do not expect a high false positive rate in the analysis of mammalian MX2 genes. Also, we combined two different methods to call a positively selected site, as MEME was used in addition to BEB. The mean pairwise divergence in the MX2 alignment was 0.32; Murrell and coworkers (http://journals.plos.org/...) indicated that this level of divergence is expected to result in a rate of false positive MEME sites around 5%. This rate is acceptable, especially if the results are combined with BEB analysis.
We thus believe that positive selection did indeed shape diversity in the L4 region in mammals and that the results we previously reported are not artefactual.

Manuela Sironi, Rachele Cagliani, Diego Forni, Mara Biasin, Mario Clerici