Molecular origins and the null hypothesis: motifs from our Maker?
Vignette #1: The tRNA tag hypothesis and the origin of replication
Transfer RNA is one of the biochemical features held in common by all modern life forms, and it has been persuasively argued that tRNA is a remnant of the RNA world. However, given that translation did not, by definition, occur in an all-RNA world, the primordial role of tRNA is unclear. Weiner and Maizels (1, 2, 3) have suggested that tRNA may have served as a "tag" for signaling or initiating replication, much as it does in many modern viruses. While this is a plausible argument, it is no more plausible than many other explanations that could also be put forth for the origins of tRNA. For example, tRNAs could have, with equal plausibility, arisen in the context of a primitive translation apparatus for the preparation of non-catalytic peptides, such as the polyglutamate tails of folate cofactors. Or tRNAs could have served as carriers for amino acids involved in the biosynthesis of other cofactors, such as the glutamate tRNAs that currently act as the precursors for porphyrins in some organisms.
In an attempt to support the hypothesis that tRNA tags may have existed in the RNA world, Weiner and Maizels have suggested that modern, viral tRNA tags are direct descendents of those that may have existed in the RNA world. This hypothesis needs to be examined in light of the null hypothesis: that modern, viral tRNA tags were invented by opportunistic viruses at some time following the reign of ribozymes. I suggest that the null hypothesis has at least as much, if not more, logical force than the notion that tRNAs served as replication tags. Viruses frequently exploit cellular metabolism by interacting with, or altering, the functions of cellular metabolites or proteins. If a virus were to invent an RNA molecule that could be used to exploit cellular metabolism, then one of the most obvious targets would be the most abundant RNA molecule in the cell: tRNA. Moreover, if the tRNA tags on viruses were invented by viruses, rather than descending directly from a primordial RNA world, then it might be expected that the tags would be quite different in sequence and structure than the conserved and canonical cloverleaf of tRNA. This is, in fact, the case: many viral tags contain sequences and structures, such as pseudoknots, that have not, so far, been identified in the cellular world (4, 5). In addition, selection experiments have demonstrated the relative ease of generating RNA molecules that can interact with tRNA synthetases and other proteins that bind tRNA, and yet do not contain the sequences and structures that signify cellular tRNAs (6, 7; [ILLUSTRATION FOR FIGURE 1 OMITTED]). Although cellular tRNAs certainly perform more functions than RNAs that have merely been selected to bind proteins and thus may be subject to more sequence constraints, this is also true when cellular tRNAs and viral tags are compared. Thus, although the existence of a plausible explanation for the origins of tRNA would no doubt expand our knowledge of the RNA world and the origin of protein biosynthesis, the tRNA tag hypothesis cannot be claimed to be more likely than the null hypothesis: that viruses invented their tRNA tags.
Vignette #2: Arginine-binding sites and the origin of the genetic code
Anti-arginine aptamers have been selected from random-sequence populations (8, 9). Analyses of the functional sequences within these aptamers suggested that there was an unusual preponderance of sequence triplets that corresponded to arginine codons. This apparent statistical correlation was used to advance the hypothesis that direct interactions between arginine and an arginine codon were somehow involved in the evolution of the genetic code itself. The route by which an arginine codon may have become an aminoacylated tRNA is uncertain. Moreover, such a process seems mechanistically implausible; for example, it is unclear how an original, direct association between codon and amino acid could have become the rather indirect association between codon and amino acid that is mediated by an aminoacylated tRNA (10). Nevertheless, the notion that amino acid:RNA interactions drove the formation of the genetic code can be separately evaluated in comparison with the null hypothesis: that the apparent preponderance of amino acid:RNA interactions is either nonexistent or an anomaly. To ferret out the details behind the supposed correlation requires a deeper insight into how selection experiments are conducted. The anti-arginine aptamers that contained a preponderance of arginine codons were selected from an RNA population with a relatively short (25 residue) random-sequence tract. The selected sequences were predicted to form secondary structures that would include the constant sequence regions at the termini of the pool; in other words, the functional arginine-binding site was presented not in the context of random sequence, but in the context of an arbitrary, predetermined sequence. If the constant sequence, in fact, skewed the anti-arginine aptamers to include presumptive arginine codons, then other selections, in which the constant region did not play as large a role, should reveal this. In fact, subsequent selections for anti-arginine aptamers yielded binding species with affinities for arginine that were higher, with functional sequences and structures that were drawn solely from the original random-sequence tract, but with functional sequences containing no apparent predominance of arginine codons (11, 12). Thus, while it would be interesting to know whether the genetic code is more than the "frozen accident" that Crick hypothesized, selection experiments provide no evidentiary basis for such a speculation.
Vignette #3: Anti-aminoglycoside aptamers and the origin of the ribosome
Aminoglycoside antibiotics have long been known to bind to ribosomal RNA (rRNA). More recently, some aminoglycoside antibiotics have been shown to bind to ribozymes. This has led some researchers to suggest that the aminoglycoside antibiotic-binding sites on the ribosome are related to the aminoglycoside antibiotic-binding sites on ribozymes, and thus that ribozymes and ribosomal RNA are somehow historically related (13). As with the previous vignettes, the plausibility of this argument turns on the competing plausibility of the null hypothesis: that the occurrence of aminoglycoside-binding sites on ribosomal RNA and ribozymes is coincidental. In essence, these competing hypotheses can be reduced to the question of whether there are few or many ways for RNA to bind aminoglycosides. To answer this question, Lato et al. (14) selected anti-aminoglycoside aptamers with affinities and specificities for the aminoglycosides lividomycin and kanamycin that are similar to those observed for rRNA. When the selected populations were characterized, it was discovered that there were literally millions and perhaps billions of unique ways to form aminoglycoside-binding sites. In other words, it was unclear whether a functional relationship (aminoglycoside binding) was, in fact, an accurate predictor of a historic relationship. Rather, the functional relationship could have readily existed even if the aminoglycoside-binding sites on ribosomal RNA and ribozymes were quite independently derived. Overall, this latter interpretation is also most consistent with the hypothesis that aminoglycosides were invented by bacteria to bind rRNA, rather than vice versa. To the extent that aminoglycosides evolved as biochemical weapons and were targeted to the translation apparatuses of many different organisms, then many different RNAs, including non-ribosomal RNAs such as ribozymes, should be able to bind to them.
These three vignettes highlight a disturbing trend in modern thinking about abiogenesis and the evolution of metabolism. We can now bring increasingly powerful techniques and methods to bear on experimental problems that affect our understanding of the origins of life. However, the results of these experiments are often interpreted as being proof of a particular scenario, rather than as general support for a more vague view of our origins. While it would be extremely exciting to establish that the RNA world evolved according to a particular scenario, it is unfortunately unlikely that sequence minutiae have remained unchanged and unscathed following a several billion year journey through multiple different types of organisms and biochemistries. Much of the comparative biochemistry of the RNA world was lost at the transition to protein-based catalysis, and again later during the ascendance of the progenote. Thus, just as paleontologists and anthropologists are cautious in interpreting an incomplete fossil record, molecular biologists should be similarly circumspect in interpreting a molecular fossil record that continues to undergo change and is open to many interpretations. In many instances, despite the exuberance of the Discussion sections of papers, we will likely have to forever remain agnostics about the true nature of the RNA world.
1. Weiner, A. M., and N. Maizels. 1987. 3[prime] terminal tRNA-like structures tag genomic RNA molecules for replication: implications for the origin of protein synthesis. Proc. Natl. Acad. Sci. USA 84: 7383-7387.
2. Maizels, N., and A. M. Weiner. 1993. The genomic tag hypothesis: modern viruses are molecular fossils of ancient strategies for genomic replication. Pp. 577-602 in The RNA World, R. F. Gesteland and J. F. Atkins, eds. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York.
3. Maizels, N., and A. M. Weiner. 1994. Phylogeny from function: evidence from the molecular fossil record that tRNA originated in replication, not translation. Proc. Natl. Acad. Sci. USA 91: 67296734.
4. Haenni, A. L., S. Joshi, and F. Chapeville. 1982. tRNA-like structures in the genomes of RNA viruses. Prog. Nucleic Acid Res. Mol. Biol. 27: 85-104.
5. Guerrier-Takada, C., A. van Belkum, C. W. Pleij, and S. Altman. 1988. Novel reactions of RNase P with a tRNA-like structure in turnip yellow mosaic virus RNA. Cell 53: 267-272.
6. Peterson, E. T., J. Blank, M. Sprinzl, and O. C. Uhlenbeck. 1993. Selection for active E. coli tRNA (Phe) variants from a randomized library using two proteins. EMBO J. 12: 2959-2967.
7. Peterson, E. T., T. Pan, J. Coleman, and O. C. Uhlenbeck. 1994. In vitro selection of small RNAs that bind to Escherichia coli phenylalanyl-tRNA synthetase. J. Mol. Biol. 242: 186-192.
8. Connell, G. J., M. Illangesekare, and M. Yarus. 1993. Three small ribooligonucleotides with specific arginine sites. Biochemistry 32: 5497-5502.
9. Connell, G. J., and M. Yarus. 1994. RNAs with dual specificity and dual RNAs with similar specificity. Science 262:1137-1141.
10. Hirao, I., and A. D. Ellington. 1995. Re-creating the RNA world. Curr. Biol. 5: 1017-1022.
11. Famulok, M. 1994. Molecular recognition of amino acids by RNA-aptamers: an L-citrulline binding RNA motif and its evolution into an L-arginine hinder. J. Am. Chem. Soc. 116: 1698-1706.
12. Osborne, S. E., and A.D. Ellington. 1997. Nucleic acid selection and the challenge of combinatorial chemistry. Chem. Rev. 97: 349370.
13. Schroeder, R., B. Streicher, and H. Wank. 1993. Splice-site selection and decoding: Are they related? Science 260: 1443-1444.
14. Lato, S. M., A. R. Boles, and A. D. Ellington. 1995. In vitro selection of RNA lectins: using combinatorial chemistry to interpret ribozyme evolution. Chem. Biol. 2: 291-303.
LANDWEBER: Why did you analyze the pool of cyclic aptamers after only four rounds of selection? Isn't that going to force the result that you will find a lot of sequence diversity? If you went further, just a few would have actually survived.
ELLINGTON: We selected the pool until it had characteristics, binding constants, similar to those of the ribosome. We wanted to ask, "What was the universe of sequences that could bind aminoglycosides like the ribosome?" That pool, containing a million to a billion sequences, was composed of sequences that could all bind like the ribosome, with about the same dissociation constant. If we had pushed that pool further, we could have worked it down to one sequence. That's what people who work on in vitro selection do. However, we wanted to ask the evolutionary question, rather than the molecular recognition question, for that particular set of results.
YARUS: Regarding the idea of codons in amino acid binding sites, I recently counted these up, and there are, in sites published and unpublished, 13 codons at 11 sites. I think the hypothesis of a connection between these and the genetic code is still alive. My general comment is that randomness being captured by evolution is not an unknown biological scenario, it is precisely the Darwinian scenario. Finding that randomness can generate patterns that are later found to be selected during evolution is usually counted in favor of the relevance of those structures, rather than against it, as you seem to propose.
ELLINGTON: The point I am trying to make is that, given a large number of potential scenarios, how does one choose from amongst them? All we have to look at are the modern descendants, which may or may not be related to what occurred in antiquity. To the extent that there were many other possibilities, we must be agnostics about whether the modern structures had any of our wonderful hypothesized roles in antiquity. I'm not an atheist about this. I am an agnostic, trying to encourage other people to be agnostic about models.
YARUS: It strikes me that there are things that fail your test that cannot be generated by randomness, and must therefore be eliminated as evolutionary material. I feel quite differently. Rather than finding that oligos are precise, haven't you found that there might be a sort of a Grisham's Law, where largely inaccurate results drive out the accurate things? In other words, have you done competitions?
ELLINGTON: No, but we would like to do that with more delimited pools; for example, where one of the positions is ribose versus xylose. We are going to do competitions to see whether there could be the selective purification that I have hypothesized.
IBBA: I would like to present another null hypothesis. The RNAs you select, that bind synthetases, don't look like tRNAs because I believe that most of them weren't actually functioning tRNAs. They're not actually charged.
ELLINGTON: That is correct. In order to rescue the tRNA tag hypothesis, one must hypothesize that there are a very limited number of RNA-like molecules that could actually function as tRNAs and be charged. The tRNA tags don't actually function as tRNAs in the sense that they go into the ribosome; they are just charged. A testable prediction of the tag hypothesis is that there should be very few charged tRNA structures. If there are very few potential charged tRNA structures derived from randomness, one could conclude that virus tags are, in fact, somewhat unique.
WEINER: You can ask that question in a different way. If some motif is easy to invent, that could actually lead to its perpetuation. For example, if two molecules interact and one of them is easy to invent, then others can arise which also interact with the same partner. Co-evolution allows the repeated selection of a particular motif. Your question is, "Could something like that arise de novo?" That is different from an interactive system, where one part of it may have a molecular memory of the other part, allowing more such parts to arise. We would never say that there is a linear physical descent from the beginning to today. We would say that this kind of structure was invented by biology and then perpetuated by co-evolution.
ELLINGTON: I agree. The real question is, "What is the information content of the interaction?" If the information content is relatively limited, either in a direct selection or in a co-evolutionary perspective, and if molecules must have historically been funneled a certain way, then that's support for a particular hypothesis. I'm trying to point out whether there are many ways to generate these sort of tags. If these tags just happen to be lying around because a virus invents a tRNA-like molecule, tRNA is the most prevalent RNA molecule in the cell, and it could have been so four years ago or four billion years ago. If there is no way to distinguish between those possibilities, then the aspects of the tag hypothesis are less predictive than they could be.
MAIZELS: Have you tested whether these little RNAs that bind tyrosyl synthetase also bind any of the other enzymes involved in tRNA metabolism, including RNaseP and CCA-adding enzyme?
ELLINGTON: That is a similar sort of question.
MAIZELS: No, it is not a similar question. If you are only looking at a single trait, you are not really asking the question that you are stating, tRNAs have multiple properties. You can't say that you have assayed for a property of tRNA when you haven't assayed for the whole spectrum of properties of tRNAs. I think this is also true of looking for the arginine recognition. You are turning this into an extremely simplified and reductionist approach and pretending that this gives you an answer, which it doesn't.
ELLINGTON: Perhaps I should modify my statement to say, it is relatively easy to generate molecules that have a particular portion of tRNA-like function, which is the ability to bind to a synthetase. It is currently unknown what the universe of answers is for RNA molecules that can both bind tRNA, recharge and interact with RNaseP.
MAIZELS: ... and interact with about four other enzymes with tRNA-like molecules.
ELLINGTON: Olke Uhlenbeck (1) has conducted experiments looking at binding both to synthetase and to EFTu and has found that there were a fairly wide range of sequences that could do so. Tao Pan (2) reported a large variety of sequences that can interact with ribonucleaseP. We haven't done the sequential experiments, but whenever it has been done individually, or even in just a single sequential (EFTu and tRNA synthetase) way, the answer remains the same.
Rather than tRNA molecules being so delimited in function that they must be an ancient relic, there seem to be too many sequences that allow function; so I cannot believe that RNA tags are anything other than modern viral inventions.
1. Peterson, E. T., J. Black, M. Sprinzl, and O. Uhlenbeck. 1993. Selection for active E. coli tRNAPhe variants from a randomized library using two proteins. EMBO J. 12: 2959-2967.
2. Pan, T. 1995. Novel RNA substrates for the ribozyme from Bacillus subtilis ribonuclease P identified by in vitro selection. Biochemistry 34: 8458-8464.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||includes discussion; Evolution: A Molecular Point of View|
|Author:||Ellington, Andrew D.|
|Publication:||The Biological Bulletin|
|Date:||Jun 1, 1999|
|Previous Article:||Prebiotic synthesis on minerals: bridging the prebiotic and RNA worlds.|
|Next Article:||Self-capping RNA catalysts derived from selection-amplification.|