The evolution of replicators

Replicators of interest in chemistry, biology and culture are briefly surveyed from a conceptual point of view. Systems with limited heredity have only a limited evolutionary potential because the number of available types is too low. Chemical cycles, such as the formose reaction, are holistic replicators since replication is not based on the successive addition of modules. Replicator networks consisting of catalytic molecules (such as reflexively autocatalytic sets of proteins, or reproducing lipid vesicles) are hypothetical ensemble replicators, and their functioning rests on attractors of their dynamics. Ensemble replicators suffer from the paradox of specificity: while their abstract feasibility seems to require a high number of molecular types, the harmful effect of side reactions calls for a small system size. No satisfactory solution to this problem is known. Phenotypic replicators do not pass on their genotypes, only some aspects of the phenotype are transmitted. Phenotypic replicators with limited heredity include genetic membranes, prions and simple memetic systems. Memes in human culture are unlimited hereditary, phenotypic replicators, based on language. The typical path of evolution goes from limited to unlimited heredity, and from attractor-based to modular (digital) replicators.

The origin of the genetic code: amino acids as cofactors in an RNA world

The genetic code, understood as the specific assignment of amino acids to nucleotide triplets, might have preceded the existence of translation. Amino acids became utilized as cofactors by ribozymes in a metabolically complex RNA world. Specific charging ribozymes linked amino acids to corresponding RNA handles, which could basepair with different ribozymes, via an anticodon hairpin, and so deliver the cofactor to the ribozyme. Growing of the 'handle' into a presumptive tRNA was possible while function was retained and modified throughout. A stereochemical relation between some amino acids and cognate anticodons/codons is likely to have been important in the earliest assignments. Recent experimental findings, including selection for ribozymes catalyzing peptide-bond formation and those utilizing an amino acid cofactor, hold promise that scenarios of this major transition can be tested.

From replicators to reproducers: the first major transitions leading to life

A classification of replicators is proposed: life depends on replicators that can exist in an indefinitely large number of forms (unlimited heredity), and whose replication is modular rather than processive. The first template replicators would have increased at a rate less than exponential, because of self-inhibition arising from molecular complementarity. The result would be the survival of a varied population of replicators, rather than the victory of one type. This variability was important, because inaccurate copying meant that individual replicators were small (Eigen's paradox). The origin of cooperation between replicators, and the problem of molecular parasites, are discussed. Today, cooperation depends on cellular compartments, and on the linkage of genes on chromosomes, but we argue that at an earlier stage surface metabolism, in which replicators react only with neighbours, was important. The origin of translation and the genetic code is discussed. The essential step is the binding of amino acids to specific oligonucleotides. We suggest that this binding originated, not as a step in protein synthesis, but in the formation of coenzymes in a metabolically complex RNA world. Existing organisms are not replicators (that is, new individuals do not arise by copying), but reproducers that contain replicators. We outline Griesemer's concept of a reproducer, which brings out the essential role of development in evolution.

A classification of replicators and lambda-calculus models of biological organization

W. Fontana & L.W. Buss (Proc. Natn. Acad. Sci. U.S.A. 91, 757 (1994) and Bull. math. Biol. 56, 1 (1994) have put forward a scheme for a theory of biological organization based on the lambda-calculus. Their key innovation was to represent, with the aid of this calculus, a certain minimal chemistry. Although this idea is very promising, their concrete formulation could be improved if suggestions for the following items were incorporated: (i) a better coding of chemical reactions; (ii) a reinterpretation of the evolutionary behaviour of autocatalytic chemical networks; (iii) a better appreciation of morphological and genetic factors; (iv) a more complete embedding of the theory into the background of relevant earlier contributions. Confusion can be stopped by the application of a proper classification of replicators (important categories being: processive and modular, limited and unlimited hereditary replicators). Suggestions to facilitate improvement are made explicit in this paper. The most challenging task would be to model the transition from processive, limited hereditary replicators to modular replicators with limited and unlimited heredity.

The major evolutionary transitions

There is no theoretical reason to expect evolutionary lineages to increase in complexity with time, and no empirical evidence that they do so. Nevertheless, eukaryotic cells are more complex than prokaryotic ones, animals and plants are more complex than protists, and so on. This increase in complexity may have been achieved as a result of a series of major evolutionary transitions. These involved changes in the way information is stored and transmitted.

Toy models for simple forms of multicellularity, soma and germ

Simple models for the development, maintenance, and origins of primitive, one- or two-dimensional "toy organisms" are presented. They are similar to cellular automata with the added combination of internal degrees of freedom that are genetically programmed. The growing rules are implementable by cellular mechanics, biochemistry and genetics. The forms that are dealt with are: sexual unicells, filaments with fragmentation, cell doublets with spore formation, filaments with differentiation of soma and germ, and two-dimensional colonies with early segregation of somatic cells and germ line. The minimum models presented help to understand the feasibility of several important evolutionary transitions. An explanation, supported by the models, for the fact that all three multicellular kingdoms are primarily sexual, is offered by the observation that sexuality in unicells is an excellent preadaptation for development, for the former entails programmed differentiation of cell types (relying in part on the action of homeobox genes), use of cell surface molecules, programmed arrest of cell division, etc. Relevant examples of existing biological systems are also presented.

Co-operation and defection: playing the field in virus dynamics

A previous model (Szathmáry, 1992) is further developed for the dynamics of standard (V) and defective interfering (DI) viruses. The crucial retained element is the incorporation of population structure in the form of a complete distribution of cells infected by particles differing in number. New elements are: the non-linear shared benefit from the contribution of Vs to the group of viruses infecting the same cell (synergistic at low numbers, diminishing returns at high numbers, respectively); a dynamics for the total number of particles (V and DI); and the possibility of extinction if the frequency of Vs is small enough. In evolutionary genetical terms this is a frequency- and density-dependent evolutionary game. A crucial result is retained: coexistence of Vs and DIs is possible provided the multiplicity of infection (hence the size of the coinfection group) is large enough. Phase portraits and vector-field plots for a continuous-time and numerical solutions for a corresponding discrete-time case are presented. The latter reflect the basic features of serial, undiluted passage. The causes for two possible means of extinction (low initial frequency of Vs, and very high vigour of Vs) are revealed: the appearance of high amplitude fluctuations. Coexistence can apparently be ensured by stable points, periodic behaviour, or strange attractors. The paper clarifies the dynamical background of cycles found experimentally in V-DI systems.

Coding coenzyme handles: a hypothesis for the origin of the genetic code

The coding coenzyme handle hypothesis suggests that useful coding preceded translation. Early adapters, the ancestors of present-day anticodons, were charged with amino acids acting as coenzymes of ribozymes in a metabolically complex RNA world. The ancestral aminoacyl-adapter synthetases could have been similar to present-day self-splicing tRNA introns. A codon-anticodon-discriminator base complex embedded in these synthetases could have played an important role in amino acid recognition. Extension of the genetic code proceeded through the take-over of nonsense codons by novel amino acids, related to already coded ones either through precursor-product relationship or physicochemical similarity. The hypothesis is open for experimental tests.

The origin of chromosomes. I. Selection for linkage

A model is analysed of cells containing independently replicating genes, which segregate randomly when the cell divides. We follow the fate of a primitive chromosome, in which two genes are linked, so that they replicate and segregate together. Such a chromosome increases in frequency in the population provided that (i) the two genes act synergistically, so that a cell containing at least one copy of each grows faster than a cell lacking one or other, and (ii) the number of genes per cell is small. The increase in frequency occurs even if the chromosome takes twice as long as an individual gene to replicate, giving a two-fold selective disadvantage within a cell. The increase occurs because a gene that is linked does not run the risk of finding itself, in the next generation, in a cell that lacks its synergistic partner.

The evolution of chromosomes. II. Molecular mechanisms

It is likely that there was a phase in evolution when the genome consisted of unlinked RNA genes, separately replicated. In this paper, we suggest molecular mechanisms whereby (i) separate RNA genes could have become linked to form chromosomes, (ii) monocistronic transcripts could have been made from these chromosomes, and (iii) RNA was replaced by DNA as the genetic material.

A note on the reduction of the dynamics of multilocus diploid genetic systems with multiplicative fitness

The dynamics of any diploid multilocus genetic system with two alleles and uniform dynamics per locus can be deduced from that of the corresponding single-locus system by trinomial sampling to produce the frequencies of the different mutant classes, provided that fitness is multiplicative, and there is linkage equilibrium (allowing for physical linkage, however). The component processes that can be considered are mutation, gene conversion, gamete production (with recombination) and random mating, and selection. If mutations occur according to Bernoulli trials, outcrossing sexual systems have a genetic load exactly deducible from that of the single-locus system.

Do deleterious mutations act synergistically? Metabolic control theory provides a partial answer

Metabolic control theory is used to derive conditions under which two deleterious mutations affecting the dynamics of a metabolic pathway act synergistically. It is found that two mutations tend to act mostly synergistically when they reduce the activity of the same enzyme. If the two mutations affect different enzymes, the conclusion depends on the way that fitness is determined by aspects of the pathway. The cases analyzed are: selection for (1) maximal flux, (2) maximal equilibrium concentration (pool size) of an intermediate, (3) optimal flux, (4) optimal pool size. The respective types of epistasis found are: (1) antagonistic, (2) partly synergistic, (3-4) synergism is likely to predominate over antagonism. This results in somewhat different predictions concerning the effect of metabolic mutations on fitness in prokaryotes and eukaryotes. The fact that bacteria are largely clonal but have often a mosaic gene structure is consistent with expectations from the model.

Viral sex, levels of selection, and the origin of life

Several issues in Chao's related paper J. theor. Biol. (1991, 153, 229-246) are revisited. It is argued that mixes of segments from different viral coinfection groups cannot be regarded as sex, unless one is willing to accept that these groups are replicators and individuals. But, because selection in coinfection groups is dynamically analogous to that in trait groups in structured demes, one should also regard these latter groups as replicators. This approach is unacceptable since the groups in question have irregular ploidies, an unfixed number of parents, and no rules analogous to those of meiosis. It is emphasized, however, that the effective presence of neighbour-modulated fitness can ensure dynamical coexistence of covirus segments, even if the equal net reproduction rate within groups is not warranted. It seems that during the origin of coviruses from complete viruses, a higher-level evolutionary unit has become disintegrated, whereas during the origin of life a higher-level unit, the protocell, has emerged from lower-level ones, i.e. unlinked, replicating genes. These two gene-level systems are not homologous, but analogous. Although it is true that the resistance to parasites and the need to avoid a mutational collapse of the genome are likely to have called for some compartmentation in precellular stages of evolution, no clear demonstration, that the proposed mechanisms (the compartmentalized hypercycle and the stochastic corrector model) do in fact solve the error threshold problem, exists. Neither has a plausible mode of protocellular sex been suggested.

A statistical test of hypotheses on the organization and origin of the genetic code

Theories of the origin of the genetic code assign different weights to amino acid properties such as polarity and precursor-product relationship. Previous statistical work on the origin of the genetic code has produced controversial results. We analyze relationships between various amino acid and tRNA properties by one and the same statistical method. It is shown that polarities as well as precursor-product relationships are both likely to have been important in shaping the genetic code, together with codon swapping that left protein sequences intact.

Natural selection and dynamical coexistence of defective and complementing virus segments

Defective interfering (DI) particles are known to coexist with wildtype viruses under high multiplicity of infection. The complementing segments of coviruses (multiparticle, segmented viruses) coexist under similar conditions. In all cases, within-cell reproductive advantage to one of the segments is rather common. This fact, and the observation that DI particles are parasites, whereas covirus segments are mutualists, call for a non-trivial model of stable dynamical coexistence. The methodical novelty is the application of the structured deme model to virus dynamics. It assumes that biochemical ("ecological") interactions occur among segments within a coinfection group, established through random infection of the cells, and there is complete mixing of the various types emerging from all the coinfection groups (cells) in the virus pool between two infections. Through the application of the model, analytic results on the coexistence of virus segments are obtained for the following cases: virus-DI particle, virus-DI particle-resistant virus, covirus pair, virus-covirus.

What is the optimum size for the genetic alphabet?

An important question in biology is why the genetic alphabet is made of just two base pairs (G.C and A.T). This is particularly interesting because of the recent demonstration [Piccirilli, J. A., Krauch, T., Moroney, S. E. & Benner, S. A. (1990) Nature (London) 343, 33-37] that the alphabet can in principle be larger. It is possible to explain the size of the present genetic alphabet as a frozen character state that was an evolutionary optimum in an RNA world when nucleic acids functioned both for storing genetic information and for expressing information as enzymatically active RNA molecules--i.e., ribozymes. A previous model [Szathmáry, E. (1991) Proc. R. Soc. London Ser. B 245, 91-99] has described the principle of this approach. The present paper confirms and extends these results by showing explicitly the ways in which copying fidelity and metabolic efficiency change with the size of the genetic alphabet.

A theoretical test of the DNA repair hypothesis for the maintenance of sex in eukaryotes

The DNA repair hypothesis for the maintenance of sex states that recombination is necessary for the repair of double-strand DNA damage. In a closed (mitotic) genetic system crossing-over generates homozygosity. This reduces fitness if deleterious recessive alleles become expressed. Thus, outcrossing is required to restore heterozygosity destroyed by recombination. The repair hypothesis is tested by comparing outcrossing sexuality with a hypothetical parthenogenic strategy (the Prudent Reparator) which destroys as little heterozygosity during repair as possible. In the Prudent Reparator, repair of double-strand DNA damage results in a small amount of homozygosity due to gene conversion only, since this process does not render outside markers homozygous. Diploidy, deleterious recessives, multiplicative fitness and linkage equilibrium in mutation-selection balance are assumed. The average fitness of this population increases, and complementation (i.e. masking of recessives in heterozygous form) decreases with the rate of damage per locus. The equilibrium fitness of the Prudent Reparator can be well above that of the sexual population. A lower complementation ability of parthenogens may not be an impenetrable barrier to their successful establishment if the invader's genome is relatively uncontaminated by mutant alleles: there are always such genotypes in the sexual population. Thus, the Prudent Reparator could solve the problem of repairing damage as well as that of invading an existing outcrossing population. As we do not see this strategy widely adopted instead of sexuality, the repair hypothesis is likely to miss some essential feature of the evolution of sex.

Four letters in the genetic alphabet: a frozen evolutionary optimum?

Piccirilli et al. (Nature, Lond. 343, 33-37 (1990)) have shown experimentally that the replicatable introduction of new base pairs into the genetic alphabet is chemically feasible. The fact that our current genetic alphabet uses only two base pairs can be explained provided that this basic feature of organisms became fixed in an RNA world utilizing ribozymes rather than protein enzymes. The fitness of such ribo-organisms is determined by two factors: replication fidelity and overall catalytic efficiency (basic metabolic or growth rate). Replication fidelity is shown to decrease roughly exponentially, and catalytic efficiency is shown to increase with diminishing returns, with the number of letters for a fixed genome length; hence their product, i.e. fitness, gives rise to a set of values with an optimum. Under a wide range of parameter values the optimum rests at two base pairs. The chemical identity of the particular choice in our genetic alphabet can also be rationalized. This optimum is considered frozen, as currently the dominant catalysts are proteins rather than RNAs.

Sub-exponential growth and coexistence of non-enzymatically replicating templates

The experimentally reported kinetic behaviour (sub-exponential but supra-linear growth) of non-enzymatic template replication is incorporated into a simple model of template competition. Sub-exponential growth is shown to lead to coexistence invariably. Thus coexistence of different non-enzymatically replicating sequences is predicted. This type of coexistence could have been important in maintaining a sufficient diversity of RNA modules used later to build functional molecules such as ribozymes. Experimental tests of this theoretical prediction are possible.

Group selection of early replicators and the origin of life

A major problem of the origin of life has been that of information integration. As Eigen (1971) has shown, a mutant distribution of RNAs replicating without the aid of a replicase cannot integrate sufficient information for the functioning of a higher-level unit utilizing several types of encoded enzymes. He proposed the hypercycle model to bridge this gap in prebiology. It can be shown by a nonlinear game model, incorporating mutation of a hypercycle, that the selection properties of hypercycles make them inefficient information integrators as they cannot compete favourably with all kinds of less efficient information carriers or mutationally coupled hypercycles. The stochastic corrector model is presented as an alternative resolution of Eigen's paradox. It assumes that replicative templates are competing within replicative compartments, whose selective values depend on the internal template composition via a catalytic acid in replication and "metabolism". The dynamics of template replication are analyzed by numerical simulation of master equations. Due to the stochasticity in replication and compartment fission the best compartment types recur. An Eigen equation at the compartment level is set up and calculated. Even selfish template mutants cannot destroy the system though they make it less efficient. The genetic information of templates is evaluated at both levels, and the higher (compartment) level successfully constrains the lower (template) one. Compartmentation together with stochastic effects is sufficient to integrate information dispersed in competitive replicators. Compartment selection is considered to be group selection of replicators. Implications for the origin of life are discussed.

Early evolution of microtubules and undulipodia

A critique of both autogeneous and symbiotic hypotheses for the origin of microtubules and cilia and eukaryotic flagella (undulipodia) is presented. It is proposed that spirochetes provided the ancient eukaryotic cell with microtubules twice; cytoplasmic microtubules originated from phagocytosed spirochetes whereas axopodial tubules of undulipodia were transformed from ectosymbiotic spirochetes. A role in transport for microtubules in spirochetes together with a detailed scenario by which free-living spirochetes attached as ectosymbionts and subsequently differentiated into undulipodia is outlined. A mechanism for the continuity of motility in the form of "training" of the novel microtubular axoneme by the ancient spirochete motility apparatus is proposed. Transitional states (missing links) are unlikely to have survived. Constraints regarding the nature of the host cell are discussed. A corresponding flowchart of the early evolution of eukaryotes is presented in which plastids and mitochondria are polyphyletic in their origins.