Reviewer's report 1
Anthony M. Poole, Stockholm University, Department of Molecular Biology & Functional Genomics, Stockholm, Sweden
This manuscript presents a highly speculative model for the origin of catalytic RNA via the prebiotic emergence of dsRNA, and includes a brief discussion of the RNA to DNA transition. I am not against speculation, but this area has been done to death, and in this particular case I do not feel the model sheds significant new light on the origin of life.
Author's response: This article describes a predictive framework, in which evolutionary steps are derived from the developmental sequences. As such, the model is not speculative, but indicates a sequence of functional intermediates in he evolution of life and was tested by providing a mechanistic basis for the scenario. Moreover, the model also incorporates the principles of complex system design, such as robustness, flexibility and driving forces, and clearly exemplifies how the steps fit in a view of evolution as an expanding molecular machine. As far as I know, no predictive framework for early genome evolution exists and no gradual, functionally continuous, steps have been published leading to the current configuration of Life as we know it. I am sure that the engineering framework and the scenario itself will be interesting to many and hope that it will give a impetus to different origin-of-life research.
I amended the article to stress the general setup of the article. I included the engineering methodology in the title and placed more emphasis on the functional continuity that is enforced by design-by-contract as a general framework for evolution. I also use more engineering terms related to complex system design including self-containedness and driving forces.
I will limit my criticism to three main points. Taking the latter stages first, the section concerning the RNA to DNA transition adds nothing new to discussions of the evolutionary origins of DNA. The author provides such a brief description of the RNA to DNA transition that it is not clear whether he favours the evolution of ribonucleotide reductases as a prerequisite to the origin of DNA. These enzymes perform the only known reaction leading to de novo deoxyribonucleotide synthesis (J Mol Evol 55:138 & 55:180). Furthermore, the growing consensus that the origin of DNA occurred very late in the evolution of cells (Biochimie 87:793; Mol Biol Evol 22:1444) is completely ignored. Explaining the origin of ribonucleotide reduction is non-trivial, as Forterre has recently argued (Biochimie 87:793). All the points regarding polymerase template and substrate specificity have been made in numerous other papers, and others have provided more in-depth discussion of the relevant experimental support for this. Entire papers have been devoted to specific parts of this complex transition; the current paper offers nothing new and is too patchy to be considered a concise summary of the existing body of work.
Author's response: I show that the RNA to DNA transition is functionally simple, because for the system it would make no difference whether the mRNA is derived from a dsRNA or a dsDNA template (cf. double-stranded RNA viruses that are transcribed by DNA-dependent RNA polymerases. The biochemical transition is not trivial but this is a general question for evolutionary science and I refer to the literature mentioned. For my article, it is necessary that deoxyribonucleotide (and thus the evolution of ribonucleotide reductases) would be present as a substrate. The exact timing would be irrelevant and even a gradual or late implementation would be possible.
As for the origin of polymerases, I do not know of any ribozyme counterpart or RNA constituent for any known polymerase, except for the RNA template of telomerase. Design-by-contract would predict that any RNA relics would be conserved in evolution in order to preserve existing interfaces, as is the case in transcription and translation. I therefore assume that RNA polymerase activity only evolved after protein translation, although RNA as a cofactor for abiotic ligation may be needed.
Reviewer's response #2: I think there is a risk that you may overstretch the predictive power you apportion to 'Design-by-contract'. In your paper, you say that there is no difficulty for dsRNA to be replaced by dsDNA, because this sort of change does not 'conceptually affect' any processes. We clearly see only partial preservation of RNA relics among modern cells (there are too few for us to see a complete RNA cell preserved within modern metabolism – see J Mol Evol 46:18), and some probable relics are not conserved in all three domains (making it harder to assign them to periods that date back to before the divergence of the three domains – see Curr Opin Genet Dev 9:672). More generally, replacements can occur without altering function, as exemplified by non-orthologous gene displacements (Trends Genet 12:334). These are important in the current case, because it would seem that this process must have happened in the origin of DNA genomes – perhaps more than once (Nucl. Acids Res. 27:3389; Mol. Biol. Evol. 22:1444). Given that interfaces remain intact in your scheme, but not necessarily the original functional units (i.e. DNA can replace RNA as genetic material), Design-by-contract cannot predict that anything as specific as RNA relics (i.e. a specific way of performing a function, tied to a specific type of chemistry) should be preserved, and therefore cannot be used to argue that ribozyme RNA polymerases never existed on account of them not currently existing. This point brings to mind Yarus' metaphorically charming 'Cheshire Cat' conjecture (FASEB J 7:31), and papers by Steitz which illustrate that a common two metal-ion mechanism for polymerization is at the basis of polymerization by unrelated polymerases, and is, chemically, within reach of catalytic RNA (e.g. Nature 391:231).
Author's response #2: I agree that it is difficult to exclude the possibility that ribozyme RNA polymerases existed, and their functional replacement with protein counterparts would indeed be in agreement with design-by-contract. In developing the hypothesis, I did assume that the RNA relics we observe are caused by the difficulties in replacing them due the need for functional continuity, in line with design-by-contract. Therefore, I would have expected to see at least some representatives of ribozyme-containing RNA polymerases, as we see in transcription and translation. The lack of any RNA component for RNA polymerases has been one of the main drivers for developing the dsRNA hypothesis and represents one of the test cases for the use of design-by-contract.
My next criticism is that the author ignores the considerable difficulties surrounding establishing plausible prebiotic syntheses for ribonucleotides. This amounts to what Joyce & Orgel have dubbed ' the molecular biologist's dream' – that these building blocks were readily synthesisable and available in sufficient amounts for the emergence of oligonucleotides (Joyce & Orgel Chapter 2 of The RNA world, 2nd edition. Gesteland, Cech & Atkins, eds. 1999 CSHL Press). The reasons why this is an unrealistic starting point for the origin of life have been expounded at great length, and, even considering current optimism surrounding possible prebiotic synthesis of ribose (e.g. Science 303:196), there is currently no reason to expect that abundant pools of nucleotides were available in prebiotic environments.
Author's response: The scenario assumes the presence of the building blocks of life, but recognizes that this is not trivial. However, current prebiotic research is mainly focused on the abiotic formation of a single-stranded RNA molecule, and the present theory focuses on the generation of dsRNA as a start which may require completely different environments and reactions. For instance, hybridization of nucleotides and formation of strands of dsRNA may actually scavenge and protect single nucleotides. RNA cofactors and special microenvironments may facilitate abiotic ligation and provide specific hybridization conditions. In other words, I think that without a specific scenario to test, abiotic chemistry cannot a priori conclude that availability of oligonucleotides is an unrealistic starting point.
We also seem to have a different definition of Life and its origin. My view on Life and its evolution is that of a self-contained system, forming basically a set of self-assembly instructions, and the origin of Life is about how this system formed. Anything that is not included in this system and part of the informational storage in the genome is in my view therefore a precondition for the origin of life, but it is not part of the system. Although the generation of nucleotides may not be simple, the essence of Life is our genome which is based on nucleotides and would therefore represent the starting point. If an abiotic generation of ribonucleotides is not possible and there are no functionally similar predecessors (e.g. preRNA), then an input from an external system seems the only other possibility. The dsRNA hypothesis for the origin of Life would not be different in such a case, and a discussion of such an environment beyond the scope of this article.
Finally, even if one elects to sidestep the 'prebiotic chemist's nightmare' and start at the 'molecular biologist's dream', for this kind of speculation to be useful, it needs to be backed up with experiments. The author should really specify some appropriate set of prebiotic conditions (temperature, salinity, oligonucleotide concentrations, etc.) and do experiments to show that, at minimum, dsRNA could be amplified under such conditions. Given that Dr. de Roos is arguing for a natural, non-enzymatic 'PCR-like' process driven by diurnal temperature changes, it would be straightforward to set up an experiment to examine whether such a system leads to ligation and subsequent amplification of dsRNAs from random RNA oligomers. That this system would lead to the emergence of catalytic RNA could likewise be experimentally tested. It is when one starts to entertain the experiments that one sees this model start to break down. For example, no explanation given as to why preformed ('prebiotic') RNAs would tend to prefer double-stranded intermolecular associations over intramolecular folding (given concentration of partners for dsRNA formation would presumably be low), nor why just catalytic ssRNAs would prefer folding into an active form rather than continuing to exist in the (according to the hypothesis presented) prevailing dsRNA state. Put simply, there is no reason why this should specifically hold for catalytic RNA as opposed to noncatalytic RNA – good folders are not necessarily good catalysts.
de Roos' model could be readily tested using existing experimental methods (a PCR machine for the thermal cycling required for non-enzymatic formation of dsRNA from oligonucleotides, and in vitro selection procedures to examine whether this process could lead to emergence of a simple catalytic RNA – here any catalytic RNA would suffice to convince me). This needs to be done before this idea can be taken seriously. My personal opinion is that if the author looks into the details of this more carefully, he will realise that the model as it stands is insufficiently detailed – and unrealistic.
Author's response: I recognized and referred to the possibility for in vitro testing of this hypothesis, and mentioned that in the case of DNA, by-products of PCR could arise in the control situation when only primers are present. Also, the partial transcription processes could be tested in vitro. I do not however present this as a research article, but as a hypothesis based on theoretical considerations, and am hopeful that scientists that have access to the research tools may try these experiments.
Folding of RNA is an important aspect and not trivial. I suggest that one of the driving forces towards catalytic RNA is folding since this will prevent rehybridization, but must be followed by natural selection for bioactive molecules. The sequence of the oligonucleotides determines folding, formation of dsRNA is dependent on temperature, salt concentration, sequence but short strand of ssRNA may also stabilize single-stranded regions. Therefore, I believe that there are enough possibilities for further research.
Finally, I recommend the interested reader to consider two papers by Paul & Joyce – the first reports a self-replicating ligase ribozyme, the second is a general review of this area (PNAS 99:12733 and Curr Opin Chem Biol 8:634). Paul & Joyce's experimental results do not solve the problems surrounding the origin of catalytic RNA, but they do provide an important piece of the puzzle, and illustrate how these problems are being productively addressed through experiment.
Author's response: I generally read these articles with the following questions in mind. Is a framework applied that can be used to unravel the different steps in evolution? What mechanistic steps have been proposed to explain the origin of dsDNA as an information carrier and ssRNA as a catalytic molecule? Is there any physical evidence for the existence of a self-replicating replicase? How can functional continuity be guaranteed when moving from a single-stranded genome to a double-stranded genome? What are the driving forces for early evolution and how are the proposed different components reconciled in one genome? I believe that these questions should be addressed first because they are at the basis of an understanding of Life.
I think the efforts in trying to make a self-replicating ligase ribozyme may be a dead-end for explaining the evolution of life. The theoretical approach used in my article not only identifies alternative solutions, it also argues against scenarios that violate the used engineering rules. The incorporation of a dual function in one molecule will not provide the necessary flexibility for the individual evolution of the different functionalities. It seems impossible to move from a single-stranded genome to a double-stranded while maintaining functional continuity. A reliance on external independent systems would not be in line with a required self-contained character of a complex system as evolution. Thus, before focusing on making a self-replicating replicase, we should be sure that it leads to an explanation of Life where these problems are sufficiently addressed to form a coherent mechanistic series of events.
Reviewer's report 2
Eugene V. Koonin, National Center for Biotechnology Information, NIH, Bethesda, Maryland, United States
It seems to me that the central point of this proposal on the origin of first replicating systems is replication by thermal cycling (convectional PCR), possibly, with the involvement of ribozyme ligases, an activity that is, indeed, fairly easily selected in ribozymes. I think that this could have been made clearer in the subtract.
Author's response: Thermal cycling was not supposed to be the central point of the article, but is used as an example how the predicted sequence of functional events can be implemented mechanistically and biochemically. The article predicts a sequence of events based on an engineering paradigm in which dsRNA as an informational carrier preceded catalytic ssRNA and served as a template for the generation of ssRNA. I have made this more clear in the article. See also comments to dr. Poole's review.
Given the difficulties with the selection of highly processive ribozyme polymerases correctly pointed out by the author, this is an interesting idea. Incidentally, this possibility has been mentioned before, albeit in passing (Koonin, Martin, Trends Genet. 2005 Dec;21(12):647–54). This being said, the hypothesis is not developed into a coherent scenario for the origin of replicating ensembles of RNA or DNA molecules, with or without protein involvement. The difficulties with ribozyme polymerases are genuine, but it is unclear whether or not thermal cycling is capable of providing a viable alternative replication mechanisms and even less clear how the transition from this hypothetical primordial replication mechanism to a replication mechanism would occur. Further, I believe that the ease of the transition from dsRNA to dsDNA that the author finds so attractive is illusory given the major structural differences between dsRNA and dsDNA.
Author's response: The main point of my article is the functional continuous scenario that is presented. As indicated in the article, the proposed transition from abiotic to biotic (enzymatic) replication is functionally similar, since the input and output are the same in both cases (dsRNA in, ssRNA out). I explained this in more detail in the comments to dr. Poole's review.
The major structural differences between dsRNA and dsDNA do not prevent incorporation of ribonucleotides instead of deoxynucleotides by polymerases, hybrid strands to be able to form, or viral dsRNA to be transcribed by DNA-dependent RNA polymerases. Thus, it seems that the structural differences between dsRNA and dsDNA may have drastic effects on for instance stability, but preserve functional continuity. This is an important aspect in my theory: later modification of the informational carrier cannot affect existing functions.
Thus, although thermal cycling might have been one of the processes involved in the early evolution of life, I do not see any substantial explanatory power in the present hypothesis. The possibility of actual RNA replication by non-enzymatic thermal cycling is worth investigating experimentally, so, inasmuch as the present manuscript brings attention to such experiments, I find it useful. However, this is where it stops as there are no specific testable predictions here, and no coherent scenario for any of the stages in the early evolution of life.
Author's response: As explained in the comments to dr. Poole's review, I use a predictive engineering framework that can be tested by investigating whether it is biochemically and biophysically feasible. I believe that the presented scenario is coherent in functional and in engineering terms. The mechanistic scenario can be tested in vitro as the figures give clear indications for real experiments, although I expect several modifications may be needed for the mechanistic scenario of the dsRNA-first scenario. As the article uses an engineering approach to model evolution, evolution could also be used as a test case to build an evolvable (software) system based on the same paradigms.
Reviewer's report 3
EugeneShakhnovich, Harvard University, Cambridge, MA, United States
This reviewer provided no comments for publication.