Ancestral reconstruction of duplicated signaling proteins reveals the evolution of signaling specificity

Essential revisions:

1) Reviewer 3 raises concerns about the approach used in the present study for the ancestral sequence reconstruction and proposes an alternative approach. We recommend that the authors consider/discuss the approach proposed (see first weakness raised by reviewer #3). It is unclear if this approach will change the results significantly, and if so, if additional experiments would be needed. It is important that the authors clarify this point by responding to this letter, while preparing a revised submission. Depending on the results additional experiments might be needed. In the case that additional experiments are necessary, we invite you to respond with a proposed plan for the additional experiments required to revise the manuscript.

We appreciate this concern and have added additional analysis to address it. To ensure the concatenated phylogenies do not contain artifacts, we have built independent phylogenies based on HK-only and RR-only alignments. We find that the trees are largely concordant, with no changes in the membership of the EnvZ1/OmpR1 and EnvZ2/OmpR2 clades studied here, and only subtle changes in the placement of some of the outgrouping sequences. This is now shown in a new supplemental figure (Figure 2 —figure supplement 2) and referenced in the Results (first paragraph of “Ancestral protein reconstruction reveals early acquisition of paralog specificity”).

Furthermore, we have computationally reconstructed ancestors based on these HK- and RR-only phylogenies and find them to be highly similar to those based on our concatenated alignment, with no changes in the key residues identified here as important in diversification. Importantly, many of the positions that vary between the matched and individual phylogenies have already been mutated in our “alt-all” alternative ancestors. Here are the exact numbers for each of the major ancestors in the paper:

ancHK: differs at 11 positions, 8 already tested in alt-all ancHK1: differs at 23 positions, 17 already tested in alt-all ancHK2: differs at 12 positions, 11 already tested in alt-all ancRR: differs at 26 positions, 16 already tested in alt-all ancRR1: differs at 28 positions, 14 already tested in alt-all ancRR2: differs at 17 positions, 13 already tested in alt-all

Given that our “alt-all” ancestors incorporated between 29 and 43 point mutations and still behaved similarly, we believe it is unlikely that the relatively small number of outstanding changes are likely to produce significantly different specificity. We have added two additional supplemental figures (Figure 2 —figure supplements 4 and 5) to show the alignments for these HK-only and RR-only ancestors compared to our matched ancestors and the alt-all ancestors and which positions differ between them. We have also added a sentence to the Results section discussing the similarity between the sequences produced from these different methods of analysis (end of paragraph 1 of “Ancestral protein reconstruction reveals early acquisition of paralog specificity”).

Reviewer #2 (Recommendations for the authors):

1) The authors' propose (based on their data for EnvZ-OmpR) that "strong selection against crosstalk occurs immediately post-duplication followed by long periods of relative stasis." I agree with this interpretation, but feel it would be bolstered by more biological context. It would be helpful to include a description of the biological role of EnvZ-OmpR, and some discussion regarding the possible fitness costs of cross-talk (or conversely, fitness benefits conferred by evolving a new paralogous pair).

Added discussion of cross-talk to introduction.

2) The authors state that the extant EnvZ-OmpR pairs (EnvZ1-OmpR1 and EnvZ2-OmpR2) are only 50% identical to each other and that the ancestral sequences are only 50% identical to the extant sequences. However, I would appreciate more discussion of sequence identity early in the results section entitled 'Ancestral protein reconstruction reveals early acquisition of paralog specificity'. In particular, how identical are ancHK1-ancRR1 to ancHK2-ancRR2, and how identical are these two pairs to ancHK-ancRR?

Added this to the results section.

Reviewer #3 (Recommendations for the authors):

I have a number of suggestions on the presentation:

2. On Fig 3F and S4 the fold-preference color scale threw me off for a bit. I think this is because the mutation/arrow side of the graphic encodes HK/RR vertically and fold preference horizontally; the color bar encodes fold preference vertically. Maybe put color bar horizontal on the bottom?

We appreciate this suggestion and have changed this in Figure 3 and S4 (now Figure 3 – Figure Supplement 2).

4. The supports were illegible in Fig S2. In the main text, the "*" notation works for your aLRT. In the supplement, it would be helpful to see the raw values, at least for the main ancestors.

The raw values have been added to Fig S2 (now Figure 2 – Figure Supplement 1).

5. Related to #4, it would be helpful to report the mean posterior probabilities for the ancestors in the main text, rather than making readers dig through Fig S4.

This has been added to the results section.

Reviewer #4 (Recommendations for the authors):

2) Fig. 1C suggests that both paralogous systems are the subject to gene loss in a somewhat equal manner. However, in Fig. 2A (and Suppl. Fig. 2), the number of sequences in EnvZ2-OmpR2 cluster is twice smaller than in EnvZ1-OmpR1 cluster. How representative Cluster 2 is compared to Cluster 1? Was there a sporadic loss throughout alphaproteobacteria or is the system missing from a specific, rather large clade? The latter potentially might skew the comparison.

There was sporadic loss throughout the alpha-proteobacteria, with EnvZ2/OmpR2 lost more frequently than EnvZ1/OmpR1. The species shown in Figure 1C do not correspond 1:1 with the sequences used in Figure 2C, hence the discrepancy. However, highly similar sequences were removed from the phylogenetic analysis to ensure that the analysis was not skewed by oversampling of particular clades.

3) Suppl. Fig. 2: ancestors of EnvZ2-OmpR2 should be labeled as "ancHK2-ancRR2" (in green).

This has been corrected.