Bacterial ancestry of the mitochondrial ATP exporter

Reviewer #1 (Public review):

Summary:

This paper tries to address an important outstanding issue, which is the evolutionary origin of the SLC25 family of mitochondrial carrier proteins, which are common to all eukaryotic life, with few exceptions. The authors have carried out phylogenetic analyses and DALI searches of AlphaFold databases of bacterial and archaeal membrane proteins. They identify two bacterial proteins, CysZ and YhiY, and they propose that they are progenitors of SLC25 family members. Whilst the paper addresses an interesting topic, the conclusions are not supported by the data and are not presented in an unbiased manner, as they highlight only features that provide some tentative support for the hypothesis. They do not address the large number sequence and structural properties that refute the hypothesis, such as the asymmetric vs three-fold pseudo-symmetric features, hexamer vs monomer, and the complete lack of any conserved motifs with similar functions. Any resemblances between CysZ/YhiY and mitochondrial carriers thus seem to be superficial and could well be coincidental, as they represent generic properties of membrane proteins rather than specific ones, indicative of an evolutionary relationship.

Strengths:

This paper explores the evolutionary origins of the SLC25 family of mitochondrial carrier proteins, which are found across nearly all eukaryotic organisms. They were likely to be present in the last common ancestor of all eukaryotes, around two billion years ago. The question is whether they are of bacterial, archeal or eukaryotic origin. The authors propose that two bacterial proteins, CysZ and YihY, may represent ancestral forms of these carriers, based on structural comparisons of models, a sequence motif, and phylogenetic analyses. While the research addresses an important and longstanding question, the presented evidence does not convincingly support their hypothesis.

Weaknesses:

A central concern is the reliance on structural similarity searches using predicted protein models, since these models are often built using known protein structures as templates, and thus these searches may produce misleading matches. The reported similarities between CysZ, YihY, and mitochondrial carriers are weak and fall within ranges expected for unrelated membrane proteins, which commonly share general structural features, such as helical bundles. Quantitative measures of similarity are low and do not support a shared evolutionary origin. The case for YhiY is extremely poor as neither structure nor sequence features support the claim. Importantly, the opening of the YihY is towards the membrane rather than the water phase, as is the case for carriers, indicating that it has a very different structure and function. The case for CysZ is somewhat better, as it is a helical bundle with two short helices somewhat resembling the matrix helices of mitochondrial carriers, and a short sequence PXDXXK that is part of one of the known sequence motifs of mitochondrial carriers, but this is where the similarities end.

Mitochondrial carriers have a distinctive threefold pseudo-symmetrical structure and a highly complex transport mechanism involving six structural elements. This paper's hypothesis does not explain how such a high level of threefold pseudo-symmetry could have evolved from entirely asymmetric proteins. To complicate matters further, CysZ is not functional as a monomer but forms a functional hexamer, which also explains why it has two half helices rather than two transmembrane helices. Thus, the hypothesis is that CysZ, which is an asymmetric protomer of a functional hexamer, has evolved into a three-fold pseudo-symmetric protein, which is functional as a monomer. A more convincing explanation is that the threefold pseudo-symmetrical structure arose from gene triplication and fusions, with later mutations introducing asymmetry to support diverse substrate binding. In support of this notion, mitochondrial carriers transporting large molecules, such as ATP, show more asymmetry, whereas those for small molecules remain nearly symmetrical. In general, the vast majority of transport proteins arose from gene duplications and fusions of the domains.

Although mitochondrial carriers have a similar sequence motif as found in CysZ (PXDXXK), their roles are very different. In mitochondrial carriers, this motif is located roughly in the middle of transmembrane helices H1, H3, and H5, where proline creates a pronounced kink, bringing the charged residues inward to form a salt-bridge network in the central water-filled cavity. The formation and disruption of this network is essential for the transport mechanism when switching between inward- and outward-open states. In CysZ, the motif is found at the end of a helix and in the following loop at the end of the transporter, with residues pointing outward toward the water phase. These residues are typical of membrane-water interface regions, where proline acts as a helix breaker and charged residues interact with the water phase. Thus, this motif in CysZ does not match the position or function seen in mitochondrial carriers, and its presence is likely to be coincidental, because these residues often occur in the water-membrane region. Importantly, none of the other important conserved three-fold symmetrical motifs of mitochondrial carriers is found in these bacterial proteins, such as the cytoplasmic network [YF][DE]xx[RK], cardiolipin binding sites, ER-links, and sequences of small amino acids, which are critical for its dynamic mechanism.

The phylogenetic relationship is also overstated, as there is no sequence similarity between these proteins other than that occurring because of similar biophysical properties, such as transmembrane helices. The authors suggest that a specific mitochondrial carrier represents the ancestral member of the family, but this conclusion appears to be inferred rather than rigorously demonstrated. Key aspects, such as tree rooting and taxon sampling, are not sufficiently addressed, weakening confidence in the evolutionary claims. Further, the selection of only a few bacterial and archaeal proteomes for analysis limits the study's scope. Broader searches would be necessary to support claims about conservation and ancestry. Independent sequence searches indicate that CysZ and YihY are not widely conserved in the bacterial groups most closely related to mitochondria, undermining the argument that they are plausible ancestors.

Overall, the presented similarities are superficial and can be explained by general features of membrane proteins rather than by specific adaptations to function. The hypothesis that CysZ and YihY are evolutionary precursors of mitochondrial carriers is not supported by the presented data.

Reviewer #2 (Public review):

Summary:

Here, the authors performed a phylogenetic analysis of mitochondrial ATP/ADP carrier (AAC) proteins. They also performed a structure-based screen for remote homologs, seeking to reveal their evolutionary origins. The authors claim that AACs are found at the root of their family tree, and through a structure-based homolog search protocol, identify putative prokaryotic homologs.

The proposed evolutionary history of AACs is bold and complicated, but the phylogenetic methodology and the way in which the tree is interpreted are incomplete and unconvincing. Further, the structure-based search strategy uses very relaxed cutoffs for fold similarity, which may be fine, but it does not clearly justify this decision. This is potentially very problematic, as I did not find the quantitative or qualitative assessments of fold similarity particularly compelling.

In summary, the authors have presented a bold and extremely interesting hypothesis for the evolution of these proteins, but there is insufficient support for their claims.

Strengths:

(1) The authors are presenting a very interesting hypothesis about the birth of these proteins, including that they may have undergone a radical rearrangement in their sequence at some point in evolution.

(2) The paper makes use of appropriate tools for structure-based homolog identification.

(3) Identification of a conserved sequence motif in these twilight zone proteins would be a rare and interesting occurrence, and could be consistent with their proposed homology.

Weaknesses:

(1) The phylogenetic analysis and its interpretations are incomplete. The authors regularly refer to the root of the tree, and its placement is given central importance. However, the methodology by which they selected the root is unexplained. This is notable, as the proposed root is curious and quite confusing. It implies that (at least) yeast and Paramecium AACs are independently paraphyletic. While certainly not impossible, this evokes quite a complicated evolutionary history. The taxonomy of this gene family, when rooted this way, does not seem to echo the phylogeny of species, suggesting an extremely complex history of duplication/loss and horizontal gene transfer, none of which the authors discuss in detail. Perhaps more clearly and specifically: I'm very surprised by the branching order at the root, where there are three independent branches of fungal proteins, followed by the excavate proteins in a monophyletic clade, followed by several independent branches of the Paramecium proteins. I very much expect incomplete lineage sorting at this evolutionary depth, but this seems extreme to the point that I question if it is accurately placed. More directly: this very much looks like an unrooted tree, presented radially.

(2) The Bayesian and ML trees seem quite incongruent, but this is not discussed. In fact, the text states that they "exhibit a similar tree topology." This is admittedly very difficult to assess without very carefully going over the tree, branch by branch, but there are nevertheless differences, the most obvious being paraphyly vs monophyly of taxon-specific AAC clades. Do the authors have any comments on this, and can they show some sort of consensus tree? How does this affect their interpretation?

(3) Presenting branch support as similarly-sized points makes it nearly impossible to actually judge the strength of support.

(4) The use of structure for remote homology detection is becoming increasingly popular, and in my opinion, is very powerful. But it is still much too early to be taken for granted. The methodology must be justified. Most importantly, the authors have not clearly described why they chose these quantitative cutoffs (I'm mostly thinking of the Dali Z-score cutoff, which here seems very low for a transmembrane protein of this size, as the Z-score is very dependent on alignment length). The authors reference categories defined by tool authors, but why a Z-score of 3, specifically? The same goes for TM scores. There are not yet any defined best practices, to my knowledge, so the authors should independently validate/justify their approach in some way and/or cite and discuss relevant literature (there have been a growing number of these screens using similar approaches in recent years).

(5) The proposed homologs have very little quantitative structural similarity to the query structure, or to each other, as shown in Figure 3 (and hence my concerns about the methodology). Also, I did not find the structural alignments in the supplement or Figure 4 to be qualitatively compelling. They simply appear too different, and I cannot discard this qualitative assessment because the quantitative similarities are likewise very weak. It's not clear to me if this is because the folds are in fact different, or if my view of them is a presentation issue (perhaps it could be improved by visualizing more angles, or more carefully cartooning the similarities and differences).

(6) The authors point out that the alpha-helices are ordered differently in YihY and CysZ, and that their membrane orientation is flipped. Taken at face value, I would view this as evidence against homology. This could perhaps be more reasonably explained as convergent global fold similarity resulting from different underlying structures. However, the authors imply that this may be the result of the transposition of the sequences encoding these alpha helices, yet there is no convincing description or argument concerning when and how this could have occurred. I think this would be a deeply interesting phenomenon, but there is insufficient evidence and discussion to seriously consider whether or not it is homology or convergence.

(7) Following up on comment #5, the authors did perform a very interesting in silico experiment by transposing sequences to reorder the helices. They then note that structural similarity improved. This is very, very interesting, but without other evidence of homology between the transposed alpha helices, I do not think this disproves alternative hypotheses. Does any such evidence exist?

(8) The authors show in Figure 5E-F that sequence transposition flips the membrane orientation, such that YihY and CysZ have extracellular termini (which you would expect from homologs, I suppose). But it is just cartooned and not discussed. Is this computationally or experimentally supported?

(9) The putative presence of a conserved motif would be a very compelling piece of evidence consistent with homology. However, it is not clear to me in the text which proteins actually have the repeats - is it truly just CysZ? What does this mean for YihY? Further, what specifically is being proposed to be homologous? Is SLC25 repeat 2 proposed to be homologous to CysZ repeat 2 (and the same for 3 to 3)? If so, this would seem to have implications for the transposition hypothesis. The helix nomenclature (e.g., H1-6) suggests homology across the proteins (i.e, H1 is homologous to H1); however, wouldn't the presence of these conserved domains instead, for example, suggest homology between SLC H3 and CysZ H2? The authors' conclusions are not clear, and it is difficult to interpret what the implications are for assessing homology.

(10) The sequence retrieval methods are incomplete, so it is impossible to reproduce the searches or to judge their accuracy and scope. What were the E-value cutoffs and other settings used in the searches?

(11) The phylogenetic methods are incomplete. What substitution models were used, and how were they chosen? What branch support method was used? What were the stop conditions of the Bayesian analysis (e.g. did the authors monitor for convergence, and how)? How much of the Bayesian analysis was considered burn-in, if any? And echoing points 1 & 2 above, how were these phylogenies rooted?

(12) Throughout, there is a distinct lack of careful, evolutionarily informative language.

(i) In reference to the phylogeny, the authors frequently refer to "grouping," but it's not entirely clear what this means. Referring to clades and their branching order would be more informative.

(ii) The authors refer to the excavate branch as the "most ancient." Whether or not excavates most closely resemble LECA is somewhat irrelevant, because the branch itself is not the most ancient - it is equally as ancient as its sister branch, which may be all other eukaryotes.

(iii) Likewise, the authors refer to bacterial proteins as "the evolutionary ancestor of mitochondrial AACs," and state that "AAC emerged from the conserved sulfat transporter CysZ." But extant bacteria are not the ancestors of mitochondria - nor are extant proteins descended from other extant proteins. They are, perhaps more accurately, cousins.

(iv) The authors refer to AACs as "evolutionarily founder member of the SLC25 carrier family," but I'm not sure that has a clear evolutionary meaning, unless the authors mean to say that the common ancestor was more AAC-like than anything-else-like. Even if the rooting is accurate, a basal branch does not necessarily reflect the ancestral state.

Reviewer #3 (Public review):

Summary:

The most important weakness is that the authors have avoided the direct structural comparison of experimentally determined x-ray structures of AAC and CysZ. Instead, the comparisons are made through predicted membrane topologies and predicted structural models of protein homologs, which give rise to misleading results. Direct comparison of the X-ray structures of the ADP/ATP carrier and CysZ clearly shows that these proteins have very different folds. Therefore, flaws in the methods produce results that lead to the wrong conclusions, and the authors have not achieved their aims.

Weaknesses:

(1) Figure 2. There is something very strange about how the tree is drawn, given that S. cerevisiae AAC1, AAC2 and AAC3 share about 76-83% sequence identity but appear to be very diversified in the tree. The phylogenetic trees are only based on the sequences of three species. The authors should explain in much more detail how they made the phylogenetic trees to support their statement that all mitochondrial carriers have come from an ancient AAC.

(2) There are at least three and seven X-ray structures of CysZ (with about 43% sequence identity to the E. coli homolog) and AAC, respectively, deposited in the Protein Data Bank. Therefore, there is no need for the approach using predicted structures as described in the manuscript. It is clear from direct comparison of the CysZ and AAC structures that they have very different folds, i.e. lengths of the transmembrane helices, their orientation and packing. CysZ has been suggested to form dimers or trimers of dimers (eLife 2018;7:e27829), with each protomer formed by two long transmembrane helices and four short helices that do not cross the membrane totally. Thus, CysZ has a different membrane topology and oligomeric state than AAC (monomer with six transmembrane helices). CysZ is therefore rightfully classified in a separate 3D domain fold from mitochondrial carriers in various protein family and domain databases.

(3) In the 3D structures of CysZ, the conserved QYXDYPXDNHK motif is involved in a network of hydrogen bonds and salt bridges thought to hold the helical bundle together (eLife 2018;7:e27829). This motif is similar to PX[DE]XX[KR], a part of the signature motif, typical of mitochondrial carriers, which is repeated three times in the sequences and forms a three-fold pseudo-symmetrical salt bridge network of the so-called matrix gate that opens and closes during the transport cycle. Therefore, although this single motif in CysZ is similar to those of mitochondrial carriers, it is not found in a similar structural context to those in AAC structures.

(4) It appears odd that the sulfate transporter CysZ should be more similar to nucleotide-transporting AAC than any of the other mitochondrial carriers, of which some transport sulfate.

(5) The alphafold model of YihY is not very similar to either the crystal structures of CysZ or AAC.

(6) The authors are relying too much on the TM-score results. The values of 0.5-0.6 between AAC and CysZ or YihY probably reflect that they contain six main helices. However, as noted in point 2, they have very different folds.

Author response:

Thank you for your decision letter with the public review and the recommendations. While we are delighted that the referees feel the work is addressing an outstanding and important issue, they have raised concerns regarding the strength of the support. We will address all the concerns in full in a revised manuscript in the due course. Please find below a couple of general points regarding the referees’ concerns and a proposal as to how we plan to address them.

(1) The idea of the manuscript is to present a plausible solution for a long-standing question in the field of mitochondrial biology and evolution. The fact that the identified solution to the origin of AAC transporters is a remote structural homolog (as you will see in our later detailed response that it is better than any other sequence/structure available till date) is to be expected. If the actual similarities were any better than what we have identified (with a special case of circular permutation), they could have been identified by other simpler structural homology search methodologies.

(2) A recurrent and strong disagreement of the reviewers on the findings presented in this manuscript is rooted on the fact that the structural and sequence relatedness between AAC and CysZ detected in this work are so weak that they can be co-incidental and not an actual evolutionary link. Based on the above, we now searched carefully in all available structural databases such as SCOP, CATH, ECOD etc. whether the above fold link has been noted by others independently. We notice that in the ECOD (Evolutionary Classification of Protein Domains) database only AAC and CysZ are grouped together under a single Possible homology group (X) called ‘Mitochondrial ADP/ATP carrier-like’. The ECOD database contains hierarchical classification of protein domains organized according to their evolutionary relationships and the server is maintained by Prof. Nick Grishin at The University of Texas Southwestern Medical Center.

Link to ECOD database: http://prodata.swmed.edu/ecod/index_af2_pdb.php

Reference: Cheng H, Schaeffer RD, Liao Y, Kinch LN, Pei J, et al. (2014) ECOD: An Evolutionary Classification of Protein Domains. PLOS Computational Biology 10(12): e1003926. https://doi.org/10.1371/journal.pcbi.1003926

Therefore, our study and the independent findings of the ECOD database team together offers greater confidence on the proposed remote evolutionary relationship between AAC and CysZ, and that the structural and sequence similarity we report in the manuscript are not a mere co-incidence. We will also incorporate the details of possible evolutionary relationship between AAC and CysZ identified in the ECOD database in the revised version of manuscript.

(3) One point we would like to stress is that considering all the similarities identified, it very unlikely falls into the class of ‘convergent evolution’. We will make this point explicit in the revised version.

(4) Lastly, while we totally agree that the similarities are in the twilight zone, considering the importance of the problem, we feel that our work would induce researchers from the field of protein design to attempt possible interconversion of the two distantly related transporters thus providing an experimental rationale for the evolution of these transporters.