ECOSYSTEM SELECTION: THE PINNACLE OF GROUP SELECTION?
Although cultural evolution is seen by some social scientists as necessarily involving the highest level of selection (e.g., Plotkin and Odling-Smee, 1981), this standpoint is arguable because more than one species displays cultural evolution and hence any level inclusive of more than one species would therefore seem (to me, anyway) to be “higher”. This highest level is that of the community, as argued by D.S. Wilson (1976).
D.S. Wilson’s (1976) argument is couched in terms of interaction coefficient matrices, but he also provides a simple thought-experiment that makes the point. Consider two plant species that are identical in how they profit from the rest of the members of their community, but differ in how they affect them. In particular, consider the effect of species A being harmful to earthworms whereas species B is beneficial to them. Species B could therefore be described as a “good mixer” and species A a poor one. Within communities, there are no systematic effects altering the relative abundances of the two species, but one emerges when the persistence times of communities with differing proportions of species A and B are considered: the more species A individuals there are, the worse off are the earthworms, and the shorter is the persistence time of the community. Overall, given enough time, species B should replace species A because communities with high proportions of species B should persist longer than communities with high proportions of species A (and hence be more important sources of species for new communities).
By the criterion I introduced above, of searching for the lowest possible level at which models adequately describe the biology of the situation, Wilson’s model is truly one of selection at the ecosystem level, and not one of selection at the level of the species. If there were only one huge ecosystem, then the relative abundances of species A and B would depend only on stochastic effects. The occurrence of many communities with finite persistence times is needed before species B will necessarily replace species A; hence, the community is the level of selection. [Now we need ecologists to provide material, rather than thought, experiments on this process!]
Given the division of the living world into communities, Wilson’s (1976) model indicates that species will evolve not only competitive properties but also cooperative ones. There is a place for “good mixers” in the wider world of life, after all.
FROM ECOSYSTEM TO CELL: THE CHAOS WITHIN
I find it somehow pleasingly paradoxical that consideration of the highest level of selection leads naturally to consideration of the lowest: that within the genome.
The idea of selection, and for that matter genetic drift as well (Van Valen, 1983), occurring within the genome is at first odd, schooled as we all have been in the impartial precision of Mendelian meiotic mechanics. But there are sections of the genome that do not obey Mendel’s rules. These sections form a large part of the DNA, in some species a larger part than that that of the law-abiding genes we were taught 21bout that obey the principles worked out by the seer from Brno.
George Mikios, in his companion article, will say more about our current picture of the chaos within the cell that is the genome, so that only the briefest outline is needed here to make my points. The non-Mendelian sequences in the genome fall into two types:
Transposable genetic elements (“jumping genes”), are of two types:
Insertion sequences (ISs, which possess only the necessary flanking repeats to allow insertion, and the gene for the enzyme, transposase, necessary for copying themselves).
Transposons. (Tns, which are composed of insertion sequences plus additional genes, such as for antibiotic resistance, that may affect the biology of the organism as a whole).
Highly-repeated DNA. Various “families” of relatively short sequences of apparently non-coding DNA whi thousand.
Initially, the traditional view of selection as focussed at the level of the individual channelled thinking so that biologists racked their brains for some function for all that non-coding DNA. The first break came from Dawkins (1976), who wrote that “The simplest way to explain the surplus DNA is to suppose that it is a parasite, or at best a harmless but useless passenger...” carried along by the rest of the genome. Although attempts to explain the large amounts of non-coding DNA in terms of such functions as regulating coding DNA (Davidson and Britten, 1979) still appeared after Dawkins’s book, it was not long before biologists concerned with genome structure and function seriously considered the view expressed by Dawkins, with particularly influential papers being those of Doolittle and Sapienza (1980), Sapienza and Doolittle (1980), and Orgel and Crick (1980). From these papers, and from those that followed them, emerged the term “selfish DNA”, and an acceleration in efforts by population biologists to understand the dynamics of selection at the level of the genome.
The dynamics of intra-genomic selection is a field of study of intense interest, both for experiment and theory. Furthermore, as stressed by Mikios (1982), and by Doolittle (1982), the difference between effect and function is especially important when considering both transposons and highly-repetitious DNA. Transposons, in particular, have numerous effects on the biology of the organisms in which they occur, but it is uncertairhas to whether or not these effects are the crucial agrents in determining the levels of representation of these elements.
A detailed examination of the large “selfish DNK literature is not feasible here, but a few remarks are worthwhile in the general context of levels of selection.
The first important fact is that even just an excess of DNA will be deleterious at the level of the individual through effects on such factors as cell cycle times and organ growth rates, although the strength of this factor will differ between organisms and between environments (Grime and Mowforth, 1982). Secondly, the evolutionary dynamics of transposable elements are markedly different from those of highly-repetitious DNA, and should be considered separately. Thirdly, because the dynamics of these two classes of non-Mendelian DNA interact, and highlyrepetitious DNA occurs only in eukaryotes, there are likely to be significant differences in the dynamics of transposable elements between prokaryotes and eukaryotes. Fourthly, the diploidy of most eukaryotes will also give rise to different dynamics relative to those of prokaryotes and of those eukaryotes which are usually haploid, because of the impact of diploidy in allowing storage of genetic variation.
Transposable elements generally insert only at specific sites (Finnegan et al., 1982), but these are usually very short. Because they are so short, the number of possible insertion sites is effectively infinite (e.g., for those cases in which the insertion site sequence is three bases long, there would be about 45 x 106 such sites in the human genome). Transposable elements are also excised from the genome with reasonable frequency; it is likely (but not certain) that this removal is caused by the genome’s recombination machinery (and is not under the direct control of the element itself).
Transposable elements cause mutations, because they disrupt the functioning of the sequence around their insertion site. Usually the mutation involves the inactivation of one or more genes, but activation of genes is also possible, as is major rearrangement of the genome (Finnegan et aL, 1982; Syvanen, 1985).
The similarity of the genome to a community of species can now be made more explicit. Traditional Mendelizing genes are relatively law-abiding members of the community of genes (could we call this the “endoblome”?). Transposable elements form a class of genes which have moved outside the tidy legal framework established by Mendel. Whereas insertion sequences do little but replicate themselves and cause mutations, transposons code additionally for functions potentially beneficial to the whole community of genes. Transposons are thus the equivalent of “good mixer” species at the ecosystem level of organization, despite their usually deleterious mutational effects.
Debate continues as to whether insertion sequences are maintained primarily by selection at the level of the individual (the complete community of genes) or at their own level (as “selfish DNK’), a debate exacerbated by the apparent inability of some participants to see that both kinds of selection will occur. Two further findings are relevant. In at least some cases, the rate of transposition of a transposon can be shown to be inversely proportional to the number of copies already present (Doolittle et aL, 1984). In general, a single transposon copy promotes its own transposition while inhibiting that of other copies (Syvanen, 1985). Syvanen (1985), a strong proponent of the importance of selection at the level of the individual or higher, argues that selection at the species level has promoted mechanisms to hold transposition in cheek until a new environment favors a new genetic makeup, which releases the transposons to cause the mutations necessary for adaptation. Doolittle et al. (1984) place the level of selection at that of the community of genes: “self-restraint” by a transposon type will prevent the genome silting up with copies and hence promote its greater longevity (i.e., transposons with selfrestraint are good mixers). But while both of these levels may be important, there is a third hypothesis, namely that this “self-restraint” is, instead, a competitive mechanism between transposons. This last hypothesis accords with the suspicion (Syvanen, 1985) that, in nature, there is considerable sequence divergence between copies of nominally the same transposable element.
There remains the distinction between effect and function. Chao et al. (1983) found that strains of the bacterium Escherichia coli “infected” with either of the transposons TnS and Tn10 tend in laboratory culture to outcompete strains lacking these elements. In the Tn10 case, the greater success of this strain occurred because of alteration to a specific site in the gpnome as a whole, and those cases in which the Tnl 0 strain lost in competition were also those in which this particular transposition event did not occur. Do these results necessarily indicate that transposable elements primarily are selected for at the level of the organism because of their production of mutations potentially increasing fitness at this level? No. While the competition experiments demonstrate that inter-bacterium competition is one component of the system, we still need to know the relative importance in nature of this selective pressure and those imposed by cross-infection and transposition.
In higher eukaryotes such as ourselves and Drosophila, diploidy leads to large stores of genetic variation impossible in prokaryotes, and hence to a much lesser reliance on mutation to provide an immediate response to selection in eukaryotes. Furthermore, many, and in some species most, transpositions will not result in mutations because they will occur within non-coding DNA. Furthermore, the much greater occurrence of genetic recombination through sexual processes in higher eukaryotes will probably, in combination with the two factors listed above, lead to a much greater significance of infection as against interstrain competition as a form of selection acting on transposable elements. This is not to deny that transposons can cause mutations in eukaryotes (Spradling and Rubin, 1981; Engels, 1983; Syvanen, 1985), or that such mutations do give rise to an increase in response to selection (Mackay, 1985), only that such effects may be less important than selection within the genome for transposition rate. The theoretical work already done (e.g., Charlesworth and Charlesworth, 1983; Kaplan et al., 1985; Ohta, 1985) needs to be extended, the phylogenetic distributions of transposable elements further elucidated (e.g., Brookfield et al., 1984; Hunt et al., 1984), and, especially, much more work done on the occurrence and effects of these elements in natural populations (e.g., Montgomery and Langley, 1983; Mackay, 1985)
Highly-repeated DNA sequences lack the sophistication of transposable elements, being simply the same (or very similar) sequences repeated thousands or hundreds of thousands of times. Insertion sequences may be derived from the breakup of transposons, and transposons from retroviruses (Flavell, 1984), but highly-repeated DNAs must have a different origin. It seems that, in some cases at least, they result from DNA fragments being made from RNA transcripts (a reversal of the normal cell functioning), with these fragments then being more or less randomly inserted into the chromosome again (Ullu and Tschud, 1984; Sharp, 1983; Brown, 1984; Rogers, 1986; — the process of the reintegration of DNA copies made from RNA, while well-documented, remains even more mysterious than most molecular evolutionary events).
While there is sufficient difference between the various copies of highly-repeated DNA “families” elements to indicate the workings of quite old-fashioned mutation (Mikios, 1982, 1985), various people have been sufficiently impressed with the overall similarity of such DNA sequences within species as against between them to coin the term “concerted evolution” to describe this apparent correction of sequences. Dover in particular (e.g., Dover, 1982; Dover et al., 1982) has suggested that this process may be important in speciation (through reducing compatibility of chromosomes in the meiosis of hybrids), and, although many cases of speciation are known where this factor is absent (such as in several Hawaiian drosophilids), it remains an exciting possibility. More important for our theme is the suggestion (Hickey, 1982; Dover, 1984) that such sequences could have significant fitness effects at the level of the individual.
Fig. 5. Unequal crossing-over as a mechanism promoting change in copy number or “correction” of repeated sequences. R and R’ are different members of the same highly-repeated DNA sequence family, and repeated rounds of unequal crossing-over, accompanied by genetic drift, lead to the loss of one or the other. Generation of chromosomes with increased numbers of copies may (a) or may not (b) result.
Much interest has centered on the mechanism of “concerted evolution”. One factor is unequal crossing-over, resulting from an uneven alignment before recombination of DNA sequences which have several copies of a sequence. Unequal crossing-over leads to changes in the number of copies and through drift to strings of identical copies (Figure 5). While unequal crossing-over will thus have the effect sought, attention has shifted to gene conversion as more likely to be a strong enough force to explain the observed levels of similarity (Dover, 1982; Nagylaki and Petes, 1982). Gene conversion is a consequence of normal meiotic processes in which repair mechanisms occasionally lead to a departure from 1:1 ratios in the gametes produced by a heterozygote. For example, a meiosis in a heterozygote A/A’ may yield three A and one A’ gamete. Selection in this context would occur by the A allele being copied at a higher rate than the A’ allele, through, for example, having a stronger affinity for the DNA synthesizing enzyme.
Gene conversion could be a significant evolutionary force only if one allele is systematically favored over the other. A survey by Lamb (1984) shows that such biases in conversion occur often, which then validates the burgeoning literature on the evolutionary importance of the phenomenon (e.g., Ohta and Dover, 1984; Lamb, 1985; Walsh, 1985). Of particular interest is the case where fitness at the level of the individual is affected by the number of copies of A’ it has, rather than of A. Ohta and Dover (1984) conclude that, under appropriate relative strengths of selection and conversion bias, selection could be ineffective in preventing the spread of A, even if it is deleterious to individuals, because all members of the population at any one time are very similar with respect to number of A’ copies. By contrast, Walsh (1985) concludes that selection at the individual level would usually prevail, if present.
Clearly, the jury is still out in the matter of the evolutionary dynamics of highly-repeated DNA sequences. There are two further considerations. Firstly, efforts to find effects of highly-repeated DNA at the organism level are relatively scanty, but efforts to find these effects by varying the amounts of such DNA have failed to do so (Mikios, 1982, 1985). Secondly, gene conversion can only systematically change the total copy number of a family to the extent that non-homologous pairing can occur, and this seems likely to be small. However, because unequal crossing~over lacks any apparent capacity for bias, it would seem a poor candidate for a mechanism to produce very large numbers of copies of particular sequences.
The mechanism for producing many copies of a sequence therefore seems most likely to be reverse copying from RNA. There are two models for how this may occur. In one model, a great many of the repeats observed have the ability not only to be transcribed and then reverse-transcribed but also to be then reintegrated into the genome. There is evidence that many repeated sequences contain recognition sites for the appropriate enzymes, leading Rogers (1985) to stress this model. The other possibility is that there is, for each family, a parent sequence that has a normal organismic function, which is then copied via processes not yet understood into many sites on the genome. Some genes are “amplified” by the production of extra-chromosomal DNA copies at times during development of high demand for their products (Sang, 1984), and these may be particularly susceptible to this process. Brown (1984) stresses this model for Alu sequences because of the discovery that they are stripped-down versions of a functional gene and Rogers (1986) believes that the LINE sequences (1ong interspersed elements” several kilobases long) may be copies of an (unknown) parent sequence. It does not seem to have been generally noticed that most of the original basis for proposing “concerted evolution” is removed under the parentsequence model, because the concert can then be orchestrated by changes in one or at least very few coding sequences (Brown, 1984). Given that DNA is also often excised by mechanisms at present poorly understood, the picture then is one of constant insertion of new sequences from RNA copies of certain sequences whose messages lend themselves to it, and their removal by the enzyme products of what might be called “sheriff” genes (Rothstein and Barash, 1983). As the coding genes evolve, so also changes the spate of reverse-transcribed copies, giving rise to apparent “concerted evolution”. In the known cases of this phenomenon, the reverse-transcribed copies are sometimes themselves transcribed but, it appears, never translated. It seems reasonable to suggest that a copy of a coding gene which produced function-coding copies in abundance would thereby unbalance the cell’s metabolism, screening out such cases, and perhaps even selecting for alleles of the coding locus which can only give rise to nontranslatable reversed transcripts!
It seems worthwhile to note that phylogenetic analysis can be used to estimate the relative importance of the two factors proposed for producing highly-repeated sequences. Sequencing information can be used to produce phylogenetic trees of the sequences in any one genome. Under the parent-copy model, the resulting tree would consist of a single long stem with unbranched shoots, whereas under the autonomous-replication model the tree would be “shrubby” due to the repeated branching of the shoots. There are general statistical methods exist for testing the shape of trees (e.g., Penny and Hendy, 1985; Crozier et al, 1986), and these could then be used to test the relative difference of observed dendrograms from purely “shrub-like” or “tree-like” forms. A simple test of this nature indicates that Alu elements probably can transpose: Economou-Pachnis and Tsichlis (1985) found that an apparently new Alu copy has higher similarity to other Alu copies than to the functional gene believed to be the ultimate ancestor for this gene family.