Referee 4's Review:

**********
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.
**********

Line 105 and table S1.
The relationship logic used here may be valid for genes with a single functional domain. However, many Adeh genes appear to have two or more important functional domains. I am not convinced that use of the "highest top non-paralogous similarities" is statistically valid as the best estimate of closest relationship when there are several functional domains.

Line 124.
It is not obvious how phenetic trees can be quantified for the purpose of combining them with phylogenetic relationships.

Figure 2.
The figure is too small to be read, and it is not clear why this figure is presented. What does it show? I recommend that table S1 be presented in place of Figure 2 in the main text, and Figure 2 be made supplementary.

Line 144.
It would seem that the Brownian motion model could only be consistent with a particular phylogeny. Which phylogenetic hypothesis does the model rule out, and with what confidence?

Line 158.
It would have been much more convincing for this argument to show that Mx8 made plaques on, or lysogenized Adeh, or injected its DNA into Adeh.

Line 160.
Isn't the estimate of 4% the most conservative? What % of Adeh genes have their highest Blast hit to an outgroup species in Figure 1?

Line 233.
While it is true that Bdellovibrio lack type IV pili, nevertheless its pili are likely to be retractile. Indeed, Bdellovibrio has flagella as well as pili, like Adeh. A complete phenetic comparison of Bdellovibrio with Adeh.would be interesting.

Line 242etc.
The point is not clear. Either the clustering with acc and rib genes is an historical accident forged by the need for the pil genes to have an active promoter, or there was a metabolic selective advantage for the clustering. It appears that the authors prefer selection. To render that preference plausible, a description of a possible metabolic scenario should be presented, or the inability to come up with one should be acknowledged. Moreover, the description of the phylogeny seems contrary to the analysis of aerobic vs anaerobic matabolism in line 390etc.

Line 258 to 261.
One cannot conclude that there is a requirement for chemotaxis simply because there are several clusters of adaptive two component systems. The logic here is the same as that in line 242 except applied to a need to explain why several copies were selected.

Line 271.
In what way is mglA associated with dif, adjacent, or liked? Do they share a promoter?

Line 273.
The proposition lacks evidence. No evidence other than the anecdote is presented that M.xanthus has less proximity among correlated functions than Adeh.

Line 301.
E coli, Proteus, and Vibrio use flagella for swarming, and translocating across a moist agar surface, while M.xanthus uses A and S motility for its swarming. In the context of the phylogenies that are presented, how can myxobacterial replacement of swimming with gliding motility be explained?

Line 372-374.
Is there evidence from % GC that HGT explains the RDase?

Line 382-438.
The argument for convergent evolution presented here seems to be contrary to the line of argument concerning signal transduction on line 255 etc above.

Line 488 and much of the Materials and Methods.
Heavy use is made of statistical methods, mainly in the computer programs listed in Methods. Values of the critical parameters employed in those programs should be given there so the results can be reproduced by other investigators. Not all parameters need to be given; just those used to support the major conclusions.