There is grandeur in this view of life, with its several powers, having been originally breathed into a few forms or into one; and that, whilst this planet has gone cycling on according to the fixed law of gravity, from so simple a beginning endless forms most beautiful and most wonderful have been, and are being, evolved.
—Charles Darwin (1859)
After the beginning but a very long time ago, the world contained RNA but neither DNA nor protein. By mindless processes of natural selection, ineffective RNAs were degraded without issue and effective RNAs acted in ways that, directly or indirectly, promoted their own replication. Each successful RNA’s self-promoting effects were its raison d’être, its function or purpose in life. Some RNAs acted as catalysts to facilitate beneficial chemical reactions. Preferential copying of more efficient catalysts perpetuated RNAs that were able to discriminate preferred substrates from less useful molecules, and perpetuated variants that augmented their catalytic prowess by association with metal ions or chemical cofactors. Other RNAs responded to things in the world with choices of action. These choices were among the earliest expressions of meaning, but, from so simple a beginning, endless forms most beautiful and most hideous have been, and are being, evolved.
The primordial RNA world is long past, supplanted by more sophisticated forms of life, but direct descendants of some of its devices survive today in the “untranslated” regions of much longer messenger RNAs (mRNAs), where they control whether or not the “message” will be translated into protein. This chapter begins with a detour into biochemical minutiae to show how remarkably sophisticated feats of interpretation can be instantiated by allosteric macromolecules. Some readers may choose to skip ahead to subsequent sections that discuss the relation between the possible and the actual, and the importance of the arbitrary nature of the sign for the meanings of life.
An RNA that catalyzes a chemical reaction is called a ribozyme. Some ribozymes possess sequences that bind to specific small molecules with high specificity. These sequences are known as aptamers (from Latin aptus, “fitted”: Ellington and Szostak 1990). An aptamer’s binding partners are its ligands (from Latin ligandus, “fit to be bound”). The evolved fit of an aptamer for a ligand is a means whereby a ribozyme selects a thing from the environment for use in a chemical reaction. Once the ligand is bound, it participates in the chemical reaction. Aptamers are also used as sensors of things in the world that inform actions of an RNA without the ligand directly participating in a chemical reaction. Such an RNA, that functionally combines a sensor (aptamer) and an effector (expression platform), is called a riboswitch (Roth and Breaker 2009).
Ribozymes and riboswitches are molecular devices that cause a reaction if a ligand is present but not if the ligand is absent. A ribozyme is a tool used to effect an action, but a riboswitch is an interpreter that uses information in the choice of an action. The function of a ribozyme is to facilitate a reaction. The function of a riboswitch is to facilitate a reaction in the presence of ligand and to prevent the reaction in the absence of ligand. The distinction between a tool and interpreter is whether the “responses” to presence and absence of the ligand are both “intended,” that is, whether nonoccurrence of the reaction in the absence of the ligand has conferred a fitness benefit. For the ribozyme, the absence of a ligand is an obstacle to achieving its function, but the riboswitch “prefers” that the reaction not occur if ligands are absent. A ribozyme acts in the world, whereas a riboswitch interprets the world.
Modern riboswitches reside in noncoding sequences of mRNAs and control whether the mRNA is translated as an enzyme (a catalytic protein). Glucosamine-6-phosphate (GlcN6P) is a substrate for the construction of bacterial cell walls. Its synthesis is catalyzed by the enzyme GlmS encoded by glmS mRNAs. The upstream-untranslated regions of glmS mRNAs contain an aptamer that binds GlcN6P. When bound to GlcN6P, the aptamer gains catalytic activity to cleave its own mRNA at an adjacent site promoting degradation of the mRNA and preventing translation of GlmS enzyme (Klein and Ferré-D’Amaré 2006). By this means, the glmS riboswitch implements negative feedback control. GlmS is translated if GlcN6P is absent but not if GlcN6P is present (Collins et al. 2007). (This device has features of both a ribozyme and a riboswitch because GlcN6P directly participates in cleavage of the mRNA but both the ON and OFF states are functional.)
In the simplest conceptualization of the glmS riboswitch, a change in conformation couples one bit of information about the world to one degree of freedom in action. Whether or not GlcN6P is present determines whether or not GlmS is produced. But the glmS riboswitch of Bacillus subtilis is more subtle than this. Its aptamer binds to either GlcN6P or glucose-6-phosphate (G6P) but inhibits translation only when bound to GlcN6P (Watson and Fedor 2011). The logic of this added complexity is that G6P is a substrate used to synthesize GlcN6P. Therefore, competition between G6P and GlcN6P for binding to glmS aptamers renders the population of riboswitches within a cell sensitive to the ratio of substrate to product. Translation of GlmS is switched off only when its product is abundant and its substrate is scarce.
B-group vitamins are chemically related to the ribonucleotides from which RNA is synthesized and are essential, as cofactors of protein enzymes, for metabolic processes shared by all living things. They probably functioned as cofactors of ribozymes in the RNA world (Monteverde et al. 2017; White 1976). A major class of riboswitches possesses aptamers for thiamin pyrophosphate (TPP), the biologically active form of thiamin (vitamin B1). The aptamers exist as ensembles of rapidly interchanging states until binding of TPP stabilizes the “active” state. The transformation of the RNA energy landscape caused by binding of TPP results in a conformational change in the expression platform that mediates downstream effects on gene expression (Montange and Batey 2008; Winkler, Nahvi, and Breaker 2002). TPP riboswitches of prokaryotes directly regulate transcription or translation whereas those of eukaryotes regulate alternative splicing of mRNAs (Li and Breaker 2013; Wachter 2010; Wachter et al. 2007).
The thiM enzyme of Escherichia coli possesses binding sites for thiazole and adenosine triphosphate (ATP). This enzyme transfers a phosphate group from ATP to thiazole, creating thiazole phosphate from which TPP can be synthesized (Jurgensen, Begley, and Ealick 2009). thiM mRNA that encodes thiM contains a TPP aptamer that includes an anti-anti-Shine-Dalgarno (SD) sequence that is complementary to an anti-SD sequence of the expression platform that is complementary to an SD sequence immediately upstream of the translation start site of the mRNA. The SD sequence, when unbound to anti-SD, enables access of mRNA to the ribosome for translation into protein. Anti-anti-SD pairs with anti-SD in the absence of TPP allowing translation to proceed. Binding of TPP sequesters anti-anti-SD, allowing anti-SD to pair with SD, blocking translation (Winkler, Nahvi, and Breaker 2002). thiM is synthesized if TPP is sparse but not if TPP is abundant. But the double-negative logic—anti-SD (translation OFF) versus anti-anti-SD (translation ON)—hints that even this “simple” riboswitch performs more complex computations than use of one bit of information in binary choice.
Sometimes two riboswitches act in tandem. metE mRNAs of Bacillus clausii contain paired riboswitches for S-adenosyl methionine (SAM-e) and adenosylcobalamin (vitamin B12). Either riboswitch terminates transcription when bound to its ligand. Transcription proceeds if neither SAM nor B12 is present (Sudarsan et al. 2006). The metabolic rationale appears to be that metE enzyme produces methionine, a precursor of SAM, but the bacterium possesses a more efficient pathway of producing methionine that uses B12 as a cofactor. Therefore, metE is transcribed and translated as metE only if the cell is deficient for both SAM and B12 (Breaker 2008).
Central to Ferdinand de Saussure’s (1916, 1986) linguistics was the arbitrary nature of the sign, the lack of a necessary connection between signifier and signified. Allostery is a parallel concept in molecular biology in which binding at one site of a macromolecule causes functional change at another site (Goodey and Benkovic 2008; Monod and Jacob 1961). Allostery frees evolution by natural selection from stereochemical constraints because it enables physicochemically arbitrary associations of ligand and response. Jacques Monod, Jeanne-Pierre Changeux, and François Jacob’s eloquent conclusion, presented for proteins but relevant to RNAs, is worth quoting at length:
A regulatory allosteric protein therefore is to be considered as a specialized product of selective engineering, allowing an indirect interaction, positive or negative, to take place between metabolites which otherwise would not or even could not interact in any way, thus eventually bringing a particular reaction under the control of a chemically foreign or indifferent compound. In this way it is possible to understand how, by selection of adequate allosteric protein structures, any physiologically useful controlling connection between any pathways in a cell or any tissues in an organism may have become established. . . . By using certain proteins not only as catalysts or transporters but as molecular receivers and transducers of chemical signals, freedom is gained from otherwise insuperable chemical constraints, allowing selection to develop and interconnect the immensely complex circuitry of living organisms. (1963, 324–325)
In the context of riboswitches, conformational coupling of aptamers to expression platforms allows physicochemically arbitrary coupling of signal to response. Allostery enables evolutionary mix-and-match between aptamers and expression platforms. GlcN6P is not physicochemically necessary for cleavage by glmS ribozymes because a mutated ribozyme that differs at only three nucleotides cleaves its mRNA in the absence of GlcN6P (Lau and Ferré-D’Amaré 2013). Nor is there any physicochemical reason why glmS riboswitches could not reside in mRNAs that possess functions unrelated to the synthesis of GlcN6P. A glmS riboswitch, for example, could be substituted for the TPP riboswitch of thiM mRNA, rendering the synthesis of thiamin contingent on the presence or absence of GlcN6P rather than TPP. But, neither a ribozyme that degraded glmS mRNA in the absence of GlcN6P, nor an mRNA that blocked production of thiamin in the presence of GlcN6P, makes adaptive “sense.”
On an evolutionary timescale, the causal connections between informative inputs and meaningful outputs in the moment-to-moment functioning of biological interpreters are allosterically arbitrary but adaptively useful. Shifting shape makes sense.
Each increase in length from n to n + 1 nucleotides quadruples the number of possible RNA or DNA sequences (four nucleotides = 2 bits per nucleotide). Each increase in length from n to n + 1 amino acids increases the number of possible polypeptides twenty-fold (twenty amino acids ≈ 4.322 bits per amino acid). Sequences of n bits can convey 2n distinct messages. As a useful benchmark, the number of distinct strings of 300 bits approximates the number of elementary particles in the universe (Lloyd 2009). By this benchmark, the number of distinct 150-nucleotide RNAs and the number of distinct 70-amino acid proteins are of similar magnitude to the number of elementary particles. If you prefer a less-than-universal yardstick, a nonredundant pool of all possible RNA sequences 100 nucleotides in length would equal 1,013 times the mass of the Earth (Joyce 2002).
Most mRNAs are longer than 150 nucleotides, most proteins longer than 70 amino acids. An mRNA of the human IGF1R gene is more than 7,000 nucleotides in length and encodes a protein of more than 1,000 amino acids. The corresponding DNA sequence, including introns that are transcribed but spliced from the mature mRNA, is a staggering 316,000 nucleotides. The numbers of possible sequences of such lengths are hyperastronomic (Quine 1987, 224) or Vast (Dennett 1995, 109), but the human genome contains tens of thousands of protein-coding genes scattered as islands across uncharted oceans of noncoding sequence. Only a vanishingly small proportion of possible RNAs and possible proteins can have been explored within the extent and age of the knowable universe. Hyperastronomic spaces are “practically infinite,” in the sense that there is no practical difference between search in a hyperastronomic or infinite space. Such spaces are never exhausted or encompassed.
Although numbers of possible RNAs, DNAs, or proteins undergo rapid ascents to hyperastronomic magnitudes as their length increases, hyperastronomic numbers are not the domain of evolution by natural selection. The number of sequences that have ever existed is undoubtedly very large, but still an earthly number. For a bit of genetic information to have been transmitted from the deep past it must have made a difference between “life” and “death” many times, first to spread through the population of sequences from its origin by mutation in a single sequence, and then to be maintained in the population against genetic drift and the onslaught of new mutations. The sequences that persist are products of historical processes that could have been different.
If there were only one way to build an aptamer for a particular ligand, then a 100-nucleotide aptamer (some aptamers are shorter, some longer) could never have been found by natural selection during the age of this planet, but it is not unusual to find arbitrarily chosen chemical activities in pools of the order of 1012 random RNA sequences of length 30–200 nucleotides (Knight and Yarus 2003). This means that there must be many possible aptamers with similar properties and that not every nucleotide is constrained by function. Even so, extant highly selective aptamers are unlikely to have been spontaneously generated in their current form. Rather, imperfect fits of earlier versions of aptamers for their ligands are likely to have undergone subsequent refinement by natural selection acting on “random” mutational variation in the “vicinity” of already functional sequences.
How can functional RNAs of thousands of nucleotides evolve? Most likely, natural selection first found functional sequences of much shorter length and then recombined these shorter sequences into longer sequences with more sophisticated functions (Lehman et al. 2011). The problem of selective search for novel functions among sequences of length 2n is much more manageable if it proceeds by recombination among already functional sequences of length n than by finding novel functions in a pool of random sequences of length 2n. Examples of innovation by recombination include the coupling of old aptamers to new expression platforms and the formation of complex computational devices by concatenation of simpler riboswitches. Such evolutionary bricolage, using materials already at hand (Jacob 1977), will leave traces of a hierarchical modular organization in complex structures. By such processes, riboswitches have been combined evolutionarily and experimentally into more complex circuits to perform a wide range of digital and analog computations (Breaker 2012; Etzel and Mörl 2017).
The number of possible RNA sequences of a thousand nucleotides is hyperastronomically large, and each of these sequences can exist in a hyperastronomic number of possible three-dimensional conformations. The complete energy landscape of an RNA encompasses all potential conformations of its linear sequence in timeless space. Over infinite time, an RNA would occupy every point in its energy landscape an infinite number of times with the frequency of transitions between conformations determined by the height of the energy barriers between them. But, in each infinitesimal moment of time, the RNA exists as some actual conformation. Life is not lived in infinite time and nothing happens in infinitesimal time. Between the infinite and the infinitesimal, the distinction between what could be and what is depends on timescale.
RNAs can be considered temporal ensembles of states, but the space of all possible conformations is much larger than could be actualized during the age of the universe for any RNA of appreciable length (Levinthal’s paradox: Plotkin and Onuchic 2002; Zwanzig, Szabo, and Bagchi 1992). No RNA of significant length lasts long enough to realize its full potential, but real RNAs fold into kinetically accessible structures in reasonable time. Folding proceeds hierarchically by rapid formation of local secondary structures followed by slower tertiary interactions among the rapidly folded elements (Brion and Westhoff 1997). The energy landscapes of functional RNAs have forms that funnel folding from high-energy states to low-energy functional states by multiple paths. These evolved mechanisms of self-directed assembly resolve Levinthal’s paradox (Dill 1999; Leopold, Montal, Onuchic 1992).
Conformations separated by negligible energy barriers interchange in nanoseconds. Such ephemeral states are sampled in proportion to their probability density on a timescale of milliseconds. One might say that actual states at the nanosecond timescale exist as superpositions of potential states at the millisecond time scale. Both timescales are involved in RNA function. Nanosecond fluctuations enable the ribozyme to “find” its substrate and stabilize an unstable transition state to enable a chemical reaction to occur in milliseconds. The timescale of catalysis is longer than the timescale of substrate “recognition” because catalysis requires passage of a higher energy barrier and the thermal fluctuations that overcome such barriers occur less frequently. Other conformations are sheltered behind even higher energy barriers and are sampled more rarely, some at much longer timescales than the half-life of the RNA.
Energy landscapes shift in response to things in the world reshaping the accessibility or stability of alternative states. One might say that a molecule’s experiences influence the realization of its potential. By such means, a ligand stabilizes one conformation of an aptamer out of an ensemble of rapidly interchanging conformations (Stoddard et al. 2010). At the timescale of the stabilized conformation, a potential structure is actualized by binding of the ligand. When a ligand stabilizes a conformation, the ligand selects the conformation (Csermely, Palotai, and Nussinov 2010). If the stabilized conformation has effects that contribute to the RNA being replicated, then the descendant sequences have been selected because of how the ancestral sequences responded to the ligand. In terms of mechanism, the ligand selects a conformation but, in terms of function, the sequence was selected because of how it responds to the ligand.
Time for you and time for me, and time yet for a hundred indecisions,
and for a hundred visions and revisions, before the taking of a toast and tea.
—T. S. Eliot
Different processes have different timescales. The selection of a conformation of a riboswitch by a ligand, and the selection of the riboswitch’s sequence by how it responds to the ligand, illustrate a separation of timescales between conformational changes in nanoseconds and evolutionary changes over hundreds to billions of years.
Human experience has a characteristic timescale. A fully formed sensory experience develops in 100 to 200 milliseconds, a single conscious moment lasts 2 to 3 seconds (Tononi 2004), a life-span of three score years and ten clocks out at about 2 gigaseconds. By these measures, an elderly human has experienced on the order of a billion conscious moments and, by retrieval of stored memories, has conscious access to how things have changed over a period of 2 gigaseconds. We can investigate processes that take place at longer timescales than a human life (such as evolutionary changes in a riboswitch sequence) and shorter timescales than a conscious moment (such as conformational changes in the folding of a riboswitch), but these do not form part of our phenomenal experience.
Mosses in my garden are fixtures for a millipede but, on the timescale of growth, mosses explore space with complex behaviors of approach and avoidance. On this timescale of moss behavior, the movements of millipedes amid the moss are a potential field of the environment, and, at the timescale of millipede meandering, molecular motions within millipede mitochondria are a potential field of the milieu intérieur. Most questions we care to ask about the world have a characteristic timescale. States that change more slowly can be treated as invariant. States that change more rapidly can be treated as a potential field of possibilities.
From the perspective of a separation of short-term from long-term changes, what is considered short term depends on personal predilections of timescale. Psychologists view individual behavior as short term and individual development as long term, whereas Bergstrom and Rosvall (2011) view individual development as short term and genetic transmission between the generations as long term. Within evolutionary biology, a separation of timescales exists between equilibration of gene frequencies on timescales for which the introduction of new mutations can be ignored (short-term evolution) and changes in phenotype over much longer periods for which change is driven by the flux of mutations (long-term evolution) (Eshel 1996; Hammerstein 1996). For a physiologist, short-term evolution is so slow as to be safely ignored in experiments, whereas for a paleontologist, short-term evolution is so fast as to be a potential field of evolving lineages.
Once upon a time, Saussure’s (1916, 1986) school of structural linguistics distinguished a synchronic axis of simultaneity from a diachronic axis of succession. The synchronic axis concerned “relations between things which coexist, relations from which the passage of time is entirely excluded” (1986, 80). The diachronic axis concerned changes of language through time. “Everything is synchronic which relates to the static aspect of our science, and diachronic everything which concerns evolution” (1986, 81). For Saussure, the synchronic was an axis of ahistorical structure that existed outside of time, and the diachronic was the temporal axis of historical change.
Nothing happens outside of time. I have found two temporal interpretations of synchrony to be useful. In the simpler interpretation, synchrony refers to things that happen at faster timescales than diachronic change. On this interpretation, Saussure separates a synchronic axis of linguistic use in the here and now from a diachronic axis of linguistic change in the there and then. This interpretation parallels Mayr’s (1961) separation of proximate and ultimate explanations in biology and Bergstrom and Rosvall’s (2011) separation of a horizontal axis of development from a vertical axis of transmission. From this perspective, the causal connections between informative inputs and meaningful outputs of biological interpreters can be understood both as structural mechanisms, analogous to Mayr’s proximate cause or Saussure’s synchronic axis of simultaneity, or, as functional products of evolutionary history, analogous to Mayr’s ultimate cause or Saussure’s diachronic axis of succession.
In the second temporal interpretation of Saussure’s atemporal axis of simultaneity, the synchronic encompasses things that change at timescales both faster and slower than the diachronic timescale. On this interpretation, diachronic change occurs in the interplay of sameness and difference, where the synchronic present comprises both the fixity of sameness (things that change at longer timescales) and potential fields of difference (things that change at shorter timescales). On this view, the diachronic drama of human history is enacted against a synchronic backdrop of things that change more slowly than the historical past and more rapidly than the existential present.
A riboswitch’s function depends on diachronic information from its evolutionary past and synchronic information from its environmental present. Evolutionary information is instantiated in the RNA sequence and represents the retrospective degrees of freedom of the replicative lineage, what could have been, and accounts for the riboswitch’s repertoire of functional responses, what could be. Environmental information selects an actual response, what is, from these prospective degrees of freedom. From the synchronic perspective of mechanism, causal parity exists between the roles of ligand and aptamer in conformational change, but from the diachronic perspective of adaptive meaning, there is a grammatical separation of roles between the evolving RNA as subject and the unchanging ligand as object.
A riboswitch’s energy landscape is its form. Some conformations have occurred sufficiently frequently to have been subject to natural selection, whereas other conformations have been sampled sufficiently infrequently to have been causally irrelevant for understanding the riboswitch’s function. The evolved energy landscape, shaped by natural selection, takes advantage of aspects of form that are sensitive to small perturbations (butterfly effects) and aspects that are insensitive to large perturbations (bathtub effects). By these means a riboswitch can be insensitive to “irrelevant” perturbations but acutely responsive to “relevant” inputs. Its ligand flips the system from one basin of attraction to another.
Consider a paradigmatic riboswitch comprising an aptamer coupled to an expression platform. The aptamer undergoes reversible fluctuations (vacillations) among potential conformations until an actual conformation is stabilized by its ligand. The ligand-induced stabilization of the aptamer causes an allosteric shift in the expression platform that causes an irreversible chemical reaction such as termination of transcription. The evolved structures of aptamer and expression platform are respectively uncertain and undecided until detection of the ligand by the aptamer causes an irreversible reaction mediated by the expression platform. The riboswitch has instantiated a reasoned choice in which a decision (irreversible action) occurs for a reason (detection of the ligand). The allosteric mechanism that couples input to output, that converts vacillation to decisive action, is physicochemically arbitrary but makes adaptive sense.
Let us concede to advocates of reductionist mechanism that everything that exists in the world can be considered an outcome of nothing but efficient and material causes. Structural biologists can explain the mechanism of a TPP riboswitch at an atomic level. They can describe how binding of the ligand stabilizes a transient conformation of the aptamer that is selected from among an ensemble of rapidly interchanging possible conformations. But this explanation of synchronic mechanism (how it works) leaves unanswered the historical how come? and the teleological what for?
Extant TPP aptamers are not products of recent spontaneous generation. Rather, an RNA sequence that bound TPP was discovered four billion years ago, give or take a few million years, and has been the progenitor of all existing TPP aptamers. In this four-billion-year history there has been both a continuity of genetic transmission and a continuity of TPP binding to aptamers. A full account of how come? in terms of efficient and material causes is unattainable, and the details, if such an account were possible, would be of little consequence. Such an account would need to keep track not only of the survival and reproduction of all the ancestors of current aptamers but also of the fate of mutant aptamers that left no descendants in the struggle for existence because, in an ecologically constrained world, the demise of the losers is a necessary part of a causal account of the success of the winners. TPP riboswitches enhance metabolic efficiency because they shut down synthesis of thiamine when thiamine is not needed. The proximate causes of death of organisms that lack this ability will have been many and varied, resulting from idiosyncratic combinations of circumstances in which slight differences in metabolic efficiency made the difference between life and death. The difference that made the difference in all these life-or-death outcomes was the difference between binding and not binding to TPP.
All TPP aptamers are descendants of an ancestral aptamer that evolved in the RNA world before the origin of DNA and proteins. These highly conserved structures are now associated with diverse expression platforms that regulate thiamin metabolism in bacteria, archaea, and eukaryotes (Duesterberg et al. 2015; Winkler, Nahvi, and Breaker 2002). An RNA sequence that recognized TPP was discovered by selective search more than two billion years ago and its descendants have persisted ever since, despite genetic drift and the constant introduction of nonfunctional variants by mutation. The evolutionary maintenance of TPP aptamers is explained by their affordances, the aptamers’ aptnesses, in particular the useful handle an aptamer provides for functional engagement with TPP. I will make the deliberately provocative claim that everything other than this affordance is causally irrelevant to understanding why TPP aptamers remain billions of years after their origin. TPP aptamers exist and persist for the sake of binding TPP.
Numerous and varied are the objections that have been advanced against the theory of selection . . . to the opposition of our own day, which contends that selection cannot create but only reject, and which fails to see that precisely through this rejection its creative efficacy is asserted.
—August Weismann (1896)
Darwinism has often been criticized as claiming that things of value can be produced by purely random processes. The Princeton theologian Charles Hodge (1878, 52) rejected Darwin’s reliance on the “gradual accumulation of unintended variations of structure and instinct”:
In like manner we may suppose a man to sit down to account for the origin and contents of the Bible, assuming as his “working hypothesis,” that it is not the product of mind either human or divine, but that it was made by a type-setting machine worked by steam, and picking out type hap-hazard. In this way in a thousand years one sentence might be produced, in another thousand a second, and in ten thousand more, the two might get together in the right position. Thus in the course of ‘millions of years’ the Bible might have been produced, with all its historical details, all its elevated truths, all its devout and sublime poetry, and above all with the delineation of the character of Christ, the ιδεα τϖν ιδεϖν [idea of ideas], the ideal of majesty and loveliness, before which the whole world, believing and unbelieving, perforce bows down in reverence. And when reason has sufficiently subdued the imagination to admit all this, then by the same theory we may account for all the books in all languages in all the libraries in the world. Thus we should have Darwinism applied in the sphere of literature. This is the theory which we are told is to sweep away Christianity and the Church! (Hodge 1871, 61)
Hodge, and many similar critics of natural selection, interpreted Darwin as assigning creativity to randomness. They could not see how a mindless process could generate order from randomness. Only a mind could be creative.
Hodge’s type-setting machine could have been unproductively employed in the printing press of Jorge Luis Borges’s (2000) Total Library, the repository of all possible books mechanically produced by randomly stringing together letters and punctuation marks. Somewhere on the shelves of that library would exist every book that has been written, every book that will be written, and every book that could be written. It would contain a copy of this book and every one of its rejected drafts. Somewhere on its shelves, tantalizingly out of reach, would be a version of this book that would convince all its readers of my intended meaning. But the library is completely useless.
Everything would be in its blind volumes . . . but for every sensible line or accurate fact there would be millions of meaningless cacophonies, verbal farragoes, and babblings. Everything: but all the generations of mankind could pass before the dizzying shelves . . . ever reward them with a tolerable page. (Borges 2000, 216)
For all practical purposes, nothing makes sense in a Total Library. Some principle of selection is required to find value in randomly generated texts.
A different reason, perhaps the same reason in different guise, for rejecting natural selection as creative is to view selection as a purely negative process that merely eliminates variation generated by other means that are the true source of creativity. A sample of quotations will give the flavor:
The function of natural selection is selection and not creation. It has nothing to do with the formation of new variation. It merely decides whether it is to survive or be eliminated. (Punnett 1913, 143)
[Natural selection] is essentially a negative substitute for teleology: it accounts for the disappearance only and not for the emergence of forms—it suppresses and does not create. (Jonas 1966, 51)
My problem is that some Darwinists are inclined to attribute creative powers to what they call natural selection. . . . The only creative element in evolution is the activity of living organisms. (Popper 1986, 119)
Where does adaptive change come from? A trivial but sometimes overlooked point is that it never comes from natural selection . . . natural selection cannot create anything. (Dupré 2017, 5)
Most of these critics ascribe innovation to the mutational processes that generate the variation that is uncreatively accepted or rejected by natural selection. They implicitly ascribe creativity to the authors of the books in Borges’s Total Library. The view that mutation is the source of meaning misunderstands meaning. Mutation is nonmeaning. At the very beginning, in the origin of difference, is nonsense.
[Natural selection] is far more “creative” than the pruning of a tree, to which it has sometimes been compared, and more creative even than the whittling of wood from a block to form an image which, among an infinite number of other potential images, had, in a sense, lain latent within the block. If this is not actual creation, then no sculptor creates his statues, and no poet, in selecting his particular words out of an almost infinite number of possible combinations creates his verses.
—Hermann Muller (1949)
Consider the token tree of all existing TPP aptamers and trace its branches back, by converging paths, to their last common RNA ancestor, the urtoken. The immediate predecessor of an extant aptamer will be a DNA sequence; then this DNA sequence will have exclusively DNA ancestors, until one gets back to a world of exclusively RNA ancestors. The RNA urtoken undoubtedly already had a high affinity for TPP, but as we trace its ancestry further back in time we would eventually come to sequences with less and less affinity for TPP, until we finally came to a stem-token with no appreciable affinity (figure 13.1).
Now start with this stem-token and follow its descendants forward in time. The branches of this tree will have been heavily pruned by natural selection. Most mutations would have little effect on affinity for TPP or would reduce affinity, but mutations that reduced affinity would preferentially be found on branches that were cut out of the tree. There will be one unique path that leads forward from the stem-token to the urtoken and that replays forward the backward “tape of ancestry.” The rare mutations that increased affinity for TPP would preferentially be found on this path. If one had pruned the token tree at random, ever so many times, one would never have obtained the sequence of the urtoken by mutation because of the hyperastronomic size of the sequence space. But the trick is turned when the tree is pruned by the environment, because new random mutations now occur on branches that had already been nonrandomly selected because they possessed mutations for greater aptness.
The mutations that occur on the path from stem-token to urtoken exibit a trend toward increasing affinity for TPP. How can one account for directional mutation toward greater aptness? The answer is simple. Mutation is a locally random process with respect to aptness, but the series of mutations found on successful paths proceed haltingly in the direction of greater aptness. The directionality comes from the environment that selects, not from the mutational process that generates branches of varying aptness. Hermann Muller, who was to receive the 1946 Nobel Prize in Physiology or Medicine for his discovery of X-ray mutagenesis, expressed the implications this way: “[It is] the peculiar power of multiplication of mutant forms which turns this trick of converting accident into order, by making such very extraordinary combinations of accidents possible as could not otherwise occur” (1929, 498).
Many considerations come into play when pruning a tree. The aptamer was shaped by what was rejected as much as by what was retained. The selective process eliminated unwanted associations with similar ligands that did not work as well or that spoiled the effect. In the evolved fit of a TPP aptamer for its ligand, many parts must work together to conform to the form of the ligand. The aptamer’s text has been subject to selection on all of its disparate effects. Some parts must do triple duty in molding the form to the ligand, in coupling the aptamer to the expression platform, and in interacting with other aptamers. Natural selection is a poet who tries the mutations in search of a bon mot. Riboswitches, genes, and organisms are the poetry of life. They mean many things at once. Perhaps this is the ultimate cause of disagreement within evolutionary theory. There are many ways to interpret a poem.
The power of Selection, whether exercised by man, or brought into play under nature through the struggle for existence and the consequent survival of the fittest, absolutely depends on the variability of organic beings. Without variability nothing can be effected.
—Charles Darwin (1883)
The origin of meaning can be ascribed to natural selection sorting meaningful from meaningless mutations. Differential copying preserves variants of value and gives directedness to the sequence of mutations in evolutionarily successful lineages. This is a process that separates gold from dross. Muller addressed the standard argument from improbability head on:
In beings without the property of multiplication of variations, and its corollary, natural selection, any such incredible combination of accidents as ourselves would have been totally impossible of occurrence within the limits of practically any number of universes. We are thus really justified in feeling that we could not have fallen together by any accident of inanimate nature. But, given the power of multiplication of variations resident in “living” things, due to their genes, and all this is changed, and we are enabled to enjoy the benefits—such as they may be—of being the select of the select, such as it would have taken a surpassingly vast number of worlds to search through, before our match could be found anywhere by the ordinary processes of chance. (1929, 504)
Ronald Fisher (1934) similarly criticized theories that ascribed the “effective guidance of the evolutionary process to the agencies which cause mutation.” He acknowledged mutation “as a condition which renders evolution possible” but, when he came to locate “in time and place the creative causation to which effects of especial importance are to be ascribed” he found it to be “in the interaction of organism and environment—in the myriad biographies of living things—that the effective causes of evolutionary change must be located.” Systems of mutable replication exhibit “spontaneous creativity” in their selected responses to the environment. Fisher later returned to this theme:
Just where does the theory of natural selection place the creative causes which shape evolutionary change? In the actual life of living things; in their contacts and conflicts with their environments, with the outer world as it is to them; in their unconscious efforts to grow, or their more conscious efforts to move. Especially in the vital drama of the success or failure of each of their enterprises. (1950, 17)
He further wrote: “Living things themselves are the chief architects of the Creative activity” via their “willing and striving” (intentions) and their “doing or dying” (actions). “It is not the mere will but its actual sequel in the real world, its success or failure that is alone effective” (19). It is the engagement of living things with their environment, mediated by the very real consequences of death or survival, that has shaped living things.
Weep not that the world changes—did it keep
A stable, changeless state, ’twere cause indeed to weep.
—from “Mutation” by William Cullen Bryant (1794–1878)