Write a comment

Summary:

In this paper, the authors address the important question of how Aminoacyl-tRNA synthetases (AARSs) have evolved. A key attribute of AARSs is that they have specialized to transfer specific amino acids to their cognate tRNAs, with minimal cross-reactivity. Although there are two major classes of AARSs (Class I and II), they focus specifically on Class I AARSs (since they could not perform a stable phylogenetic analysis on Class II). To this end, they have employed structure based sequence alignment of HUP domains of different Class I AARSs, based on which they built phylogenetic trees and performed ancestral sequence reconstruction. They make interesting, but counterintuitive, observations on the evolutionary trajectory of AARSs in comparison to the timeline of emergence of amino acids themselves. Specifically, they note that AARSs which charge amino acids that emerged later in time appear as early branches in the phylogenetic tree and vice versa. They also observe that one of the AARS ancestor (Anc-all-minus) had a wide substrate binding pocket that did not confer amino acid sidechain selectivity, but rather selected for L-configuration ɑ-amino acids. Based on these results, the authors propose a new model of evolution of Class I AARSs called generalist-maintaining (GM), where the early ancestor with non-specific/generalist activity is maintained and as amino acids emerged later in time, became starting point for the evolution of specialized AARSs. Overall, the paper is concise and self-sufficient. The conclusions drawn by the authors are significant and well supported by data. There are a few minor points that we want to bring to the attention of authors, which could improve the manuscript further.

Minor points:

1. The authors discuss one set of ancestral reconstruction throughout the paper. Were there any alternatives generated by the software used? If yes, on what basis was this particular reconstruction chosen? Perhaps, if there are alternatives, the authors could build the ancestry based on them and see if it yields similar results. If it does, it could make the conclusions from this paper more robust.

2. The authors mention in a single sentence in Methods - “Ancestral states were inferred using codeml from the PAML package”. Since this is one of the most important steps in this work, some explanation about how this was done and any parameter choices or tuning can be useful.

3. On Page 7, they say “The anticodon binding domains are thought to have emerged later, in agreement with our analysis, which indicated that the anticodon domains of Class I AARSs relate to at least three separate evolutionary emergences (Table S1)”. The three emergences of anticodon domains is not clear from Table S1. If we are to go by the different H groups according to ECOD, there are only two in the anticodon binding domain column in Table S1. Some clarity on this will be helpful

4. Since the authors frequently compare the specialization of AARSs with the emergence of amino acids, a schematic showing the order of appearance of amino acids will be more illustrative than making the readers refer to multiple papers. Potentially using real amino acids in Figure 4 would be more clarifying than A,B,C,D,E?

5. The phylogenetic trees shown in Figure S4 are key to this work. The authors could make the Raxml tree (since this was used for ancestral reconstruction) a main figure to guide the reader about this important part of the paper.

6. The expansions of many abbreviations used throughout the paper haven’t been given (for example - HUP, HIGH, ECOD)

7. The importance of outgroups in ancestral reconstruction is remarked upon twice “due to the lack of a suitable outgroup, making ancestor reconstruction intractable.”... “Using Clade 2 as an outgroup, we succeed in the construction” - but would benefit from a sentence or two (and reference) explaining this necessity to readers from non-phylogenetic backgrounds.

8. The discussion of the structural/functional characteristics of the inferred ancestors could be expanded. For example:  “Further, the inferred Anc-All-minus pocket bears no hallmarks of a Val activating enzyme,” it is unclear which hallmarks you are referring to and what is different in the ancestor. Predicted AlphaFold2 structures of the ancestors might help here, especially aligned with substrate bound/docked structures of extant AARSs.

Ashraya Ravikumar and James Fraser (UCSF)

James Fraser had a long scientific relationship and personal friendship with the late Dan Tawfik and objectivity (which is always something in the eye of the beholder) may be particularly impacted here.

Competing interests

The author declares that they have no competing interests.