Reviewer's report 1
Stephan Beck
In their manuscript Joly and Rouillon report new evidence for the hypothesis that MHC class I genes undergo concerted evolution through gene conversion (e.g. non-homologous recombination). In support, they analysed 8 classical class I genes (termed class Ia) and 8 non-classical class I genes (termed class Ib) from primates and rodents with focus on three class Ib genes to which they refer to as CD94L family (human HLA-E, mouse H2-Qa1 and rat RT-BM1). The results are comprehensively discussed within the context of various relevant hypotheses, which adds a review-like flavour and greatly enhances the appeal of the manuscript.
Although some conclusions (and assumptions) are better supported than others, I only take issue with one particular point. Based on evidence I do not agree with, the authors assume the above mentioned CD94L family genes to represent orthologues and their many respective paralogues are not considered in subsequent analyses which may have affected some of the conclusions.
Author response:
Following this comment, and a suggestion made by Pierre Pontarotti on the phone, I have now modified the manuscript to remove the statement about 'the clear orthologous relationship of CD94L molecules within the primate or the rodent orders'. This is now replaced by 'There is very little room for doubt that all four primate CD94L genes descend from a common ancestral gene, and similarly for all four rodent CD94L genes'.
On page 12, for instance, the authors conclude that at least the alpha 3 domains of the CD94L genes have all undergone intra-species concerted evolution with their respective class Ia molecules but not one of the 22 informative positions is shared across species as one would expect for orthologues. The logic conclusion would have to be that gene conversion did not occur in the ancestral (e.g. pre 80 mya) CD94L genes studied here. This is unlikely, as gene conversion has been demonstrated to be a general mechanism clearly predating the species studied here.
Author response:
Thanks to the process of 'open refereeing', I have been able to discuss this point with Stephan over the phone directly. His comment sprouted from some slight misunderstanding, which has now been lifted.
The additional section (appended to main manuscript) does not really constitute a separate manuscript but adds further interesting points to the discussion and the key points could be summarized and included in the main manuscript.
Author response:
The solution to this has been to remove a sizeable portion of the discussion and to provide it as a clearly separate manuscript.
The paper itself now focuses on the demonstration that HLA-E and Qa1 are orthologues. It is now much shorter, easier to read, and the message is, I hope, much clearer.
The accompanying paper is now clearly labelled as 'hypothesis', and I have used it to regroup 4 topics of discussion touching on different aspects of MHC evolution that derive from the results obtained in the paper itself, but are not directly related to these results.
Reviewer's report 2
Lutz Walter
In this paper, Joly and Rouillon compare major histocompatibility complex (MHC) class I genes derived from human, non-human primates, and rodents. Based on multiple sequence alignments and phylogenetic tree reconstructions, the authors conclude that the MHC class I genes in these species are subject to concerted evolution by means of gene conversion.
One main point of criticism refers to the fact that not all known MHC class I genes of the species studied here are compared, and only a small extract from the full repertoire of class I genes was chosen for comparison. This may bias the interpretation of data. In this respect, it might be useful to concentrate on one or two species, e.g. the 'class I-rich species, mouse, rat, or rhesus monkey. In its current form, the paper contains data from a single mouse haplotype, but from several rat haplotypes. Thus, the data set should also be updated to allow the study of both inter- and intralocus gene conversion.
Author response:
One of the main challenges we faced when we started to do the work that would allow us to write this paper was not in terms of "How many sequences for MHC class I molecules can we collect and align ?". It was, in fact, exactly the reverse, i.e. :" With how few sequences can we proceed to demonstrate, beyond reasonable doubt, that MHC class I loci do undergo concerted evolution ?" All the figures presented in the paper were obtained from the one alignment we settled for in the end. Changing just one sequence in the list would require performing the whole study all over again, which would represent several weeks of tenuous work.
Although our observations lead us to discuss many aspects related to MHC evolution, evaluating the frequency of inter and intra-locus conversion events was not within the remit of this study. Regarding the choice of MHC sequences from several rat MHC haplotypes, and from a single mouse haplotype, we do not see why this should have any relevance to the type of work we have done, and to the conclusions we reach. Outside of particular situation where genes can co-evolve because they are closely linked (such as RT1-A and TAP), MHC haplotypes are, after all, relatively artificial sets of genes that happened to find themselves on the same chromosomal strand when inbred strains were generated.
Starting with many more sequences, we would have faced the following problems:
1) The alignment showed on Figure 1, which was used to generate the trees, could not have been provided within the manuscript. We also find that the clarity of figures containing trees degrades rapidly when these trees have too many branches.
2) The computer time required for calculation of the trees and of the ds/dn values grows exponentially with the number of sequences, and the time spent generating the alignments and the figures is also dependent on the number of sequences included.
3) For the precise question we wanted to address, the only class Ib loci that were informative were those identified in at least two species. We also felt that it was best to restrict our analysis to those molecules for which a function had clearly been documented, and which had the same number of amino-acids as class Ia sequences (to avoid gaps in the alignment). When we embarked on this work, the only class Ib loci fulfilling these criteria were the CD94L and the murine M3 molecules. As far as we know, this is still true today.
page 16: the sister grouping of M3 and CD94L genes is due to a limited data set (see above) and does not reflect true phylogenetic relationship (and is not supported by bootstrapping). Furthermore, it contradicts data by Hurt et al. (2004) who studied the phylogenetic relationship of all rat and mouse class I genes;
Author response:
We are in complete agreement with the statement that the grouping of M3 and CD94L does not necessarily reflect phylogenetic relationship, and this despite a bootstrapping value of 63% (with the methods used for these comparisons, values above 60% are usually considered significant, and this is specified several times in the paper). Two alternative interpretations relating to this were (and still are) proposed in the manuscript. We actually pointed to this feature of the tree to underline our point of view that extreme caution must be exerted when carrying out phylogenetic analyses of members of multigene families that undergo extensive inter-genic exchanges.
page 16 and more: it is not obvious why the authors introduce a new abbreviation for 'residues outside the antigen recognition site (ROARS)' and do not use the widely accepted 'non-PBR';
Author response:
We chose to use ROARS because we think it sounds better than 'non-PBR', and also because not all class I molecules present peptides.
page 17, second paragraph: the authors should explain how homogenisation can be afforded in non-PBRs, particularly at those sites where PBRs and non-PBRs alternate. Is the degree of homology between the two sequences high enough to allow gene conversion to take place?
Author response:
What we witness here are signs that are very evocative of intra-species homogenisation, and gene conversion seems to be the most likely mechanism to explain this. We have no way of knowing when these events took place, and between what sequences (for example, some other genes, or pseudogenes, could have served as relay between certain sequences). Furthermore, although gene conversion is clearly favoured between homologous sequences, we are not aware of data documenting the minimal length of homologous sequences required for gene conversion to take place. Outside of the fact that this question seems to be way beyond the scope of our study, we therefore would have no way of addressing this question.
I would not recommend adding of the additional section into the manuscript, as the manuscript might become 'unreadable'. However, certain aspects of this "additional" discussion section might be included in the manuscript. Nevertheless, I would strongly recommend considerable shortening of the manuscript.
In my opinion, this paper should be published, but should be regarded as a 'hypothesis paper' as it contains many assumptions, which were not proven by experimental evidence, and it contains many review-like sections.
Author response:
As explained above, we have managed to comply to these slightly contradictory recommendations (i.e. including more points but shortening overall) by splitting the paper in two: One 'real' paper with the results, and one hypothesis paper.
Reviewer's report 3
Pierre Pontarotti
This article hypothesizes that the Peptide Binding Region of the mouse, rat and human Class I b, that presents the leader peptide from the Class I a molecules to natural killer cells, evolved from a common ancestor while the non PBR part evolved via gene conversion.
The arguments are based upon phylogenic analysis and upon the conserved location of these MHC class I b genes.
This contrasts with another hypothesis: the MHC class I genes are lineage specific, they come from a common ancestor which is different in the human and mouse lineage, (in other word class I gene from mouse and human are paralogues), and HLA E and H2Qa1 PBR evolved via convergent evolution...
In order to strengthen their hypothesis the authors should screen other mammalian lineages using ensembl data bases since some sequences of ensembl data base are not obligatory present in NCBI NR, especially those from canis, loxodanta, bos Taurus, canis Familirais and monodelphis (even if this species is out side of the eutherian group). If an "HLAE like" PBR orthologue is found in all these groups, the author hypothesis will be stronger supported.
Author response:
We would indeed have been very interested to identify MHC class I molecules with CD94L-like PBR outside of the rodent and primate genera. As was already indicated in the result section entitled "Certain residues are CD94L-specific, and others are homogenised within species" (on page 10 of the current manuscript), we have repeatedly tried to identify such molecules via several approaches in all the online databases available to us, and, as of 20 Dec 2005, we have not succeeded so far.
Second if the conversion of the non PBR HLA E like gene is an ongoing process, this could be seen at higher taxonomic level, for example at primate level by comparing human chimp and macaque MHC class I genes: more homogenization should be seen outside the "HLA E like" PBR than within the PBR.
Author response:
This is indeed exactly what we see, and this is discussed on page 18 of the manuscript (Comparison of this tree...... low support values).
Other comments:
Concerning the sentence page 14 L 11: Among the classI B ...years ago. I do not understand why the results confirm that CD94L molecules are much more evolutionary conserved than Class I a molecules.
Author response:
This was indeed confusing, and I have tried to clarify this point by writing the following sentence:
"The fact that the comparison of primate sequences strongly suggests that the four CD94L are orthologues, whereas this is much less clear for the corresponding class Ia sequences confirms previous reports that primate