Referee 1's review (Jan Drake):

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication, the manuscript has been revised in light of these comments and to address other editorial requirements.
*********

I soon found that I lacked sufficient experience in computational biology to judge the technical soundness of this paper, although I was able to comprehend most of the arguments and certainly the central conclusions. One key conclusion is that certain large-scale aspects of vertebrate DNA sequence, including the isochors that vary by GC content, can be understood on the hypothesis of mutation pressures conditional on base nearest neighbors. These pressures might have evolved because they could reduce the impact of missense mutations. Another key conclusion is that this mutation pressure applies approximately equally to coding, intronic, and mostly non-genic regions, so that it operates strongly and independently of codon preferences and GC contents.

Strong nearest-neighbor effects on base-specific mutation rates are well documented both observationally and experimentally, as are effects over longer distances, but their extensive impacts on DNA primary structure was news to this reviewer, and was quite exciting. The authors' speculations on molecular mechanisms are inoffensive. They point out the currently huge difficulties of testing their conclusions, but I do not find that to be a drawback to publication, which I believe to be well warranted according to the PLoS ONE criteria but with the limitation of my own minimal competence in this area.

The authors might be interested in this snippet: Walter, C. A., et al., 1998 Mutation frequency declines during spermatogenesis in young mice but increases in old mice. Proc. Natl. Acad. Sci. USA 95: 10015-10019. The article reports (using a mouse transgene) that somatic mutation frequencies are 8-fold higher than frequencies in sperm cells.