The most recent and refined version of the evolutionary worldview that was first presented in Evolutionís Arrow can be found in the 34 page document The Evolutionary Manifesto which is here
Chapter 9. Smarter Genes
How smart at evolving are autocatalytic sets? How good are they at discovering new adaptations and passing them on from generation to generation? Does evolution tune and hone the evolvability of autocatalytic sets?
We have seen that an autocatalytic set is a group of proteins in a watery environment that collectively reproduces itself. Each protein catalyses (manages) reactions that lead to the formation of other members of the set. Collectively this produces a proto metabolism in which other molecules (food) are managed by the proteins to reproduce the set.
But autocatalytic sets of proteins do not contain genes. How can sets evolve? How can the members of the set discover better ways to cooperate with the other members, and adapt their cooperation as the internal and external environment of the set changes? How can they do this in ways that will be passed on from generation to generation of sets, producing evolutionary adaptation?
Autocatalytic sets are able to reproduce in some circumstances. This is because they become more likely to break up into smaller sets as they increase in size. The new sets that are produced in this way will compete with each other for the food and other matter that they need for their survival and reproduction. If a change arises within a set, the changed set may prove to be more competitive than others. A changed set that does better will grow in size faster, reproduce more frequently, and tend to take over the population of sets. As a result, a change that makes a set more competitive will be established as an adaptation possessed by all members of the population. In this way, natural selection operating between sets will test the effectiveness of any change that arises within a set.
But evolution can occur in this way only if changes arise within sets, and if the changes can be passed on from generation to generation. How can changes of this type arise?
Perhaps the simplest way is if different parts of a particular set become physically separated, and if the parts happen by chance to contain different components. Any part which contains members that can reproduce collectively as an autocatalytic set will be able to survive and reproduce as a separate member of the population. If the ways in which a set differs from its parent make it more competitive, it can take over the population, producing a population of adapted sets.
It is conceivable that this simple change-and-test process could even evolve sets that are better at evolving. For example, consider what will happen if sets differ in their tendency to separate into parts. If a set has a tendency to separate into different parts too readily and too often, it risks breaking up effective arrangements and producing daughter sets that are not competitive. Alternatively, if a set never separates into parts that are different, it will not evolve. It risks being out-competed by those that do. Sets that balance these tendencies will be better at evolving, and will be favoured by evolution. All change-and-test mechanisms face this particular dilemma. They must strike a balance between preserving their accumulated discoveries by minimising changes, and boosting the search for new improvements by trying out changes more often. Finding the best balance between conservation and change is an old evolutionary problem.
Nevertheless this simple change-and-test process is very limited in its ability to explore the evolutionary potential of autocatalytic sets. It cannot produce a set that has new members that were not also members of the parental set. It is limited to trying out different combinations of the same members. It cannot discover a new protein that might contribute more to the efficient operation of the set than existing members.
However, there is another change-and-test process that enables new members to be tried out in an existing set. A new protein molecule may form by chance within a set. This might happen through the chance interaction of particular molecules that come together in the right positions and under the right conditions to form the protein molecule. Alternatively, a number of molecules of one or more new proteins might drift into the area occupied by the set. If these new proteins are then reproduced as part of the set, and if they improve its competitiveness, adaptive evolutionary change will have been achieved.
But the ability of this process to test out all the types of proteins that could possibly improve the competitiveness of the set is strictly limited. It can try out only those proteins that, when added to the set, will be reproduced as part of the set on an on-going basis. If a new protein arises in the set by chance, or if it drifts into the set, it will not survive for long unless its formation is catalysed by the set. For a new protein that is not already reproduced by the set, this will occur only in exceptional circumstances. It will occur only if the addition of the new protein to the set causes changes that result in the new protein being reproduced by the set. The catalytic activity of the new protein must set off changes in the set that eventually result in the formation of the new protein within the set.
This is a very restrictive condition. But only proteins that meet this narrow requirement can be tried out and established by this evolutionary mechanism. If a protein does not meet this condition, it cannot be discovered by the mechanism, no matter how much it might improve the competitiveness of the set. Of course, this is an instance of the barrier to the evolution of cooperation that we looked at earlier. The barrier limits the extent to which an evolutionary process is able to exploit the benefits of cooperative organisation. There is a close connection between this barrier and evolvability. All instances of the barrier can be seen as a limitation in the ability of the relevant evolutionary mechanism to discover useful cooperation. And the emergence of managed cooperative organisation and other arrangements that overcome the barrier can therefore be seen as improvements in evolvability.
The limited ability of autocatalytic sets to discover new proteins was overcome once RNA began to manage autocatalytic sets to form proto cells. The RNA had the ability to cause the formation of proteins that were not reproduced by the autocatalytic set itself. It could therefore try out proteins that could not be tried out by an autocatalytic set alone. RNA could discover and reproduce proteins that would improve the competitiveness of the proto cell, but would not be reproduced by the set in the absence of the RNA.
RNA was particularly suited to searching systematically through the great range of new possibilities that were opened up. Each RNA molecule includes a long sequence of four basic units. Each sequence of units will produce a different sequence of the basic units that make up proteins. So a different protein could be tried simply by a change in the sequence of the basic units in the RNA. Importantly, mutations that change these sequences do not alter the basic character of the RNA molecule, and do not interfere with its ability to reproduce. So there were no limits to the types of proteins that could be tried out by RNA management. Not only did RNA have the ability to produce new proteins, it also had the ability to try out changes easily and systematically in each of the proteins that it produced. The result was a significant advance in evolvability.
The great evolvability of RNA meant that there was advantage in RNA eventually taking over the production of all proteins within the proto cell. Its greater evolvability could be used to search systematically for adaptive improvements in each protein that it produced. But RNA had to proliferate within the protocell if it were to be able to take over the production and evolution of all proteins. This could not occur immediately. It could take place only once arrangements were developed to suppress destructive competition between the various RNA molecules that produced the different proteins. Until the suppression arrangements we discussed in Chapter 6 were in place, the proliferation of RNA in the cell would have produced only destructive competition.
But the suppression of competition between RNA molecules meant that no evolutionary change-and-test mechanism involving RNA could operate within the cell. If it were to operate, alternative RNA molecules would have to be produced and tested within the cell. RNA molecules would have to mutate, reproduce and compete within the cell. But the suppression arrangements would prevent this from occurring. Competition, and therefore evolution, was prevented within the cell. Instead, mutated RNA molecules could compete only through competition between the cells that contained them. If a mutated RNA molecule improved the competitive ability of the cell that contained it, the cell and the molecule could breed up and take over the population of cells. The change-and-test process operated only at the level of the cell. Changes arose in the RNA within cells, but were tested only through competition between cells.
In essence, this is the genetic evolutionary mechanism that produces evolution in all single celled and multicellular organisms. Some of the details have changed: in later cells, DNA took over from RNA as the ultimate level of management. And in multicellular organisms DNA is an internal distributed manager rather than an external manager as in cells. But the evolutionary mechanism is essentially the same in all these cases despite the differences: competition between mutations is suppressed within the organism during its life, and mutations are tested by competition between organisms.
In the remainder of this Chapter, we will take a close look at the evolvability of the genetic evolutionary mechanism. We will ask whether evolution has shaped the genetic systems of organisms to increase their ability to discover useful adaptations. Has evolution produced genetic systems that are smarter than if they used only random trial-and-error? A central issue here is whether genetic systems have evolved any capacity to anticipate the future. Do genetic systems target mutations at the types of environmental conditions that are likely to occur in the future? Does the pattern of genetic changes contain a higher proportion of changes that are more likely to match future evolutionary needs?
To see what such an ability might mean in more concrete terms, we return to our hypothetical example of a population of snow hares. The population faces a critical environmental challenge due to temperatures that fluctuate considerably across the generations. If the genetic system of the population could target its genetic changes, it would not produce changes randomly across all the genes of the organism. Instead, it would produce changes that were more likely to pay-off, given the types of environmental challenges faced by the population. It would target its genetic changes towards producing a variety of lengths and thickness of fur, increasing the chances that the population could adapt quickly to fluctuations in temperature. Changes would be made less often to genes in which any change was likely to be harmful, and less often to genes in which changes would not be relevant to likely environmental change.
A central dogma of evolutionary biology is that genetic mutation is blind to future evolutionary possibilities. On this view, genetic mutation is not targeted at all. Whether a genetic change proves to be useful in discovering a new adaptation or in adapting to environmental changes is completely a matter of chance. Mutation is random in relation to future adaptive possibilities.
But this dogma is not based on hard evidence. No biologist has ever gone out and collected the evidence that is needed to test the dogma. To do so thoroughly, a biologist would have to catalogue the genetic variation produced in a population, assess whether it arises randomly across all genes, and, if there is any bias, determine whether the bias is correlated with the adaptive challenges likely to be faced by the population.
Apart from the practical difficulties in gathering this evidence, doing so has not been given a high priority by biologists. Most have considered that there are strong theoretical reasons to accept the dogma that mutation is blind to adaptive challenges. First: there is no obvious mechanism that would enable genetic systems to target changes at future adaptive needs. How could the simple processes that produce mutations know anything about future evolutionary possibilities? Second, even though greater evolvability would be in the long-term interests of a population, it would not evolve unless it was also to the advantage of the individuals within the species. Smarter populations or species would be able to out-compete other species by adapting first when the environment changes, or by discovering new and better adaptations. But it is not enough that evolvability is good for the species. Unless it also continually benefits the individual genes that produce evolvability, these genes will be out-competed within the population, and greater evolvability will not evolve.
The great problem for genes that improve evolvability is that evolvability might not produce benefits continuously. The population might not encounter significant environmental challenges for many generations. When it does, any individuals that carry genes for greater evolvability can do better in evolutionary terms. These individuals are more likely to produce offspring who are better adapted to the new environmental conditions. But until the environment changes, individuals who carry genes for greater evolvability will not gain any benefit. In fact, if the arrangements that increase evolvability are costly, the individual will be disadvantaged.
This argument does not apply only to genes that improve evolvability. It also applies to the genes that establish the genetic change-and-test process itself. Foremost amongst these are genes that cause or allow mutations to arise in other genes. These genes will tend to be out-competed during periods in which there is no advantage to evolving. The cost of producing genetic changes will be a burden when there are no significant environmental challenges or other circumstances that make it worthwhile to try out genetic changes. There will be no pay-off for the change-and-test process if a population is well adapted to its environment, and if the environmental conditions are stable and unchanging. An individual will do better in these circumstances if it produces only offspring that are faithful copies of itself. Its young will be well adapted, like itself. An individual that instead produces some offspring that are changed from itself is producing maladapted young. Any change will be for the worse. All changes will be harmful.
For example, consider a hypothetical population of snow hares that contains a gene that produces mutations in other genes. The mutations influence the length and thickness of fur. An individual that carries this mutator gene will tend to produce some offspring with different types of fur. The mutator can do well if environmental temperatures change significantly every generation or so. The chances are that one of the offspring will have fur that is better for the new temperature. If so, it will out-compete other members of the population, including those that produced only offspring that were faithful copies of themselves. The mutated gene and the mutator gene that produced it can take over the population.
But if the temperature does not change over many generations, and if the population has the right fur for this temperature, the mutator will disadvantage any individual who carries it. The mutator will cause the individual to produce some offspring with different types of fur. The individual will be out-competed by others who produce only offspring with fur that is right for the unchanging environmental temperatures. It will be disadvantaged compared with individuals that do not try out anything different through their offspring.
Many evolutionary theorists argue along these lines that natural selection will favour zero mutation rates in most circumstances. They argue that organisms often experience stable environments for long periods, and do not face significant environmental changes every generation or so. During periods of stability, organisms would do better if they did not produce mutant offspring. Organisms would therefore suppress mutation if this could be done efficiently, they argue. They acknowledge that genetic mutations are continually produced in organisms, but suggest this is only because the arrangements needed to copy genes without mutations are too costly for the organism.
On this view, genetic evolution would have ended if it had ever discovered a cheap way to stop the copying errors that produce mutations. The genetic change-and-test mechanism exists not because it enables new adaptation to be discovered, but because mutations are unavoidable in practice.
However, this position is not intuitively attractive, and many biologists have searched for alternative explanations of how genetic evolvability might evolve. One approach is to look for ways in which the living or non-living environment of the species may be continually changing. If a population continually encounters adaptive challenges every generation or so, genes can be favoured that cause the production of some offspring that are genetically different. As we have seen, if the environmental temperatures met by a population of snow hares are continually varying from generation to generation, genes that cause individuals to produce some young with different types of fur could be very successful. They could do better than those that simply produce offspring with the same fur as the parents.
But biologists have had trouble showing that key features of the environment of organisms are in fact changing continuously in this way, every generation or so. The non-living environments of most species appear to remain stable for long periods, and the organisms appear to undergo very little evolutionary change during these periods.
The best candidate for a source of continual adaptive challenge is not the physical environment of a species, but other living organisms, particularly parasites. When a parasite discovers a better way to exploit a species, the species will benefit if it discovers a way to counter the change in the parasite. In turn, the parasite will benefit when it finds a way around the counter move, and so on, indefinitely. The result is an arms race of move and counter move that produces continual adaptive challenges for the parasite and the host species. Computer simulations have shown that these continual challenges can provide an evolutionary advantage for mechanisms that continually try out genetic changes in the search for new adaptations.
But this parasite theory is too narrow to explain most of the genetic changes that are continually produced by populations of organisms. The genetic changes produced generation after generation by organisms are not limited to changes that might be useful in combating parasites. Populations of organisms are continually testing changes in all aspects of the organism. This has been shown conclusively by experiments that subject organisms to artificial selection. Whenever animal breeders have set out to see if they can change a feature of an animal through artificial selection, they have generally met with success. Whatever features they look at, they find that genetic changes are being generated as the animals reproduce, and that these changes enable the animals to evolve in response to selection.
Another problem with the parasite theory is its prediction that genetic changes would cease to be produced if ever the arms race stopped. It is only the continual adaptive challenges produced by the arms race that provide a profit for evolvability. The parasite mechanism can explain the continual production of genetic changes only if the arms race never stops. But this condition is highly implausible. There is no good reason to expect that all species that continually produce genetic variety are continually engaged in evolutionary arms race with parasites.
So are most of the genetic changes produced by populations of organisms harmful but unavoidable? Do populations produce genetic changes largely because they are too costly to eradicate, not because they enable the population to evolve? Or can natural selection favour genes that produce evolvability? Can genes that cause genetic changes become established in a population because they improve evolvability, even though the population may experience long periods of environmental stability?
Theoretical work begun by American evolutionist Egbert Leigh in the early 1970's suggests that they can. He was able to show that in restricted circumstances, natural selection will favour genes that cause populations to continually try out mutations. Further work carried out by a number of theorists has extended his conclusions to a wider range of circumstances and genetic changes. This work has also shown that natural selection will favour genes that target their genetic changes so that the changes are more likely to meet future adaptive needs.
This new theoretical approach acknowledges that if a mutator gene is to be able to benefit from discovering useful mutations when the environment changes, it must be able to somehow survive periods when the environment is stable. The theory shows how a mutator can do this. It demonstrates that a mutator can always survive periods of stability if the rate at which it causes mutations is low enough. And this is the case even though all mutations produced during periods of stability are harmful because they change an organism that is already closely adapted to its environment.
The new theory begins by noting that individuals which carry the mutator will not always be disadvantaged in evolutionary terms. An individual will not be disadvantaged unless the mutator causes the individual to produce an offspring that carries a harmful mutation. When it does, the copy of the mutator gene in that offspring will be removed from the population. But until it causes a harmful mutation, each copy of the mutator will do as well as any other gene. The rate at which a mutator gene dies out of a population depends on the rate at which it produces harmful mutations. The lower the mutation rate, the longer a mutator can survive without discovering useful mutations. A mutator gene will not die out of a population until all of its copies have produced harmful mutations. And if the mutation rate is low enough, there will always be some copies of the mutator left in the population at the end of a period of stability. No matter how long the period of stability, a mutator can survive if its mutation rate is sufficiently low.
If a mutator can survive periods of environmental stability in this way, it can spread throughout the population when environmental conditions change and the mutator produces a successful mutation. When a copy of the mutator causes an individual to produce a useful mutation that takes over the population, the useful mutation will re-establish the mutator throughout the population. Provided the mutator and the mutation are closely linked genetically, all individuals in the population will eventually contain a copy of both. The mutator will hitch hike on the back of any successful mutation it produces.
However, this does not solve the mutators problem permanently. Every time that it re-establishes itself in the population by discovering a useful mutation, it then begins to die out again as it produces harmful mutations. To survive, it must again discover a useful mutation before it dies out. And to survive permanently in the population, the mutator must produce a useful mutation each and every time, before it dies out. Unless this condition is met indefinitely, a mutator will not survive continually in a population.
But this condition can be met often in a typical population of organisms. Over long time frames, a population will become increasingly maladapted as environmental changes accumulate. All organisms become increasingly maladapted if they remain the same as time passes. No matter how stable an organisms living and non-living environment, it will eventually accumulate significant change, and features of the organism that were once adapted will no longer be. So as time passes, the success rate of mutations will get better and better. The less an organism is adapted to its living and non-living environment, the greater the likelihood that a random change in the organism will produce an improvement. As environmental changes accumulate, it therefore becomes increasingly likely that even random mutations will be useful. Mutators with a mutation rate that is low enough will be able to survive until the likelihood that they produce a useful mutation becomes a certainty.
If a population is adapted to its environment, all mutations are likely to be harmful. Many will be lethal because they damage the effective functioning of the organism. The remainder might produce organisms that can function, but they will be less adapted than non-mutants. However, even in a relatively stable environment, as environmental changes inevitably mount over many generations, and as the changes increasingly disadapt the organism, the chance that a mutation will produce an improvement increases. Provided a mutator is able to survive long enough, it will become increasingly likely that its mutations will be useful to an increasingly maladapted population.
To illustrate how such a mechanism can operate, we will look again at our hypothetical example of an evolving population of snow hares. Imagine that the environmental temperatures met by the population usually vary little over a period of 10,000 years, but that over a time scale of 50,000 years, significant change is likely to have accumulated. A population that is very well adapted to the temperatures at the beginning of a 10,000-year period will therefore generally be adapted at the end of the period. Mutations will be harmful during this period if they change features of the snow hare that are suited to the prevailing temperatures. So a mutator that produces a high rate of mutations that change these features will soon die out of the population.
Compare this with a mutator that instead causes an individual to produce a mutated offspring only once every 20,000 generations, on average. If this mutator was common in the population at the beginning of the period and there is one generation per year, the mutator is highly likely to be still common at the end of even the 50,000-year period. But a snow hare population that was adapted to environmental temperatures at the beginning of the 50,000-year period will be maladapted if the snow hares are unchanged at the end of the period. And there will be a much higher probability that a random mutation will produce a snow hare that is better adapted to the new environmental temperatures. If the mutator produces such a change, it will spread throughout the population again.
So mutator genes can survive in a population if their mutation rate is tuned to the rate of relevant environmental change. The environment does not have to be continually changing every generation or so to favour the evolution of this evolvability. A gene that produces evolvability can be successful if it operates over a time scale on which the environment is changing continually. And this will hold true no matter how slow the rate of change.
Natural selection can be expected to tune mutation rates to balance two opposing tendencies: first, a lower rate will improve the ability of a mutator to survive periods of little environmental change. Second, a higher rate will improve the ability of a mutator to discover useful mutations before another mutator does so. The higher the mutation rate caused by a mutator, the more likely it will discover the first useful mutation when the environment changes, provided it still exists within the population.
It can be expected that the optimum mutation rate will be different for different genes within the organism. The ideal mutation rate will be higher for genes that establish features of the organism that are affected by rapidly-changing environmental conditions. It will be much lower for genes that produce features that cope well with all but rare environmental events.
There is growing evidence from studies of bacteria that mutation rates do in fact vary for different genes, and that these differences are not random. Some genes are much more likely to mutate than others, and it is because there is evolutionary advantage in doing so. The higher rates have evolved because they improve evolvability.
For example, a particular group of genes in the parasitic Salmonella bacteria have been found to produce proteins on the surface of the bacterium. The organisms infected by these bacteria use the proteins to identify different types of bacteria. When the immune system of an organism has come up with a way of destroying a particular type of bacteria, it will be used against all bacteria with the surface protein common to that strain. But if the bacteria can change its surface protein, it can escape the defences of the host organism, until the host again locks onto its particular type of surface protein. Bacteria will do best if they can adapt to the continually-changing immune system of the host by continually searching for types of surface protein that are not recognised by the host organism. Researchers have found that genes that produce these proteins mutate far more rapidly than other bacterial genes that do not have to deal with aspects of the environment that are continually changing at such a high rate.
Natural selection will not only favour the production of mutation rates that vary across the genes of the organism. It will also favour the production of a targeted pattern of mutations that is more likely to include mutations that will meet future adaptive needs. A mutator will be more competitive if the mutations it produces include fewer that damage the individual, and more that change existing adaptations in ways that are likely to be useful as the environment changes. A mutator whose mutations are biased in this way is likely to do much better than a mutator whose mutations are random.
Mutators that are better at targeting their mutations will be discovered by the normal genetic change-and-test process. General mutation will produce a variety of mutators. Different mutators will produce mutations in different genes and in different parts of genes. Of these, the mutators whose pattern of mutations happens to do better at matching adaptive opportunities through time will out-compete those that are less effective.
For an illustration, we will revisit our hypothetical population of snow hares that experiences significant changes in environmental temperatures every few generations. We will consider what happens in the following circumstances: a particular mutator produces mutations randomly within a number of genes that are responsible for growing fur. Mutations in some parts of these genes change the length and thickness of the fur. When such a mutation produces a type of fur that adapts the hares to the prevailing temperature, the mutation and the mutator will spread through the population. But mutations in other parts of these genes are harmful. They produce hares with no fur, with damaged fur, or with only patches of fur. None of these mutations will ever be useful in the temperatures met by the population. So the mutator that produces mutations randomly across all parts of these genes will produce both useful and harmful mutations, depending in which parts of the genes the mutations arise. But despite producing some mutations that are always harmful, such a mutator can survive in the population, provided it is always able to eventually discover an adaptive mutation before it dies out.
But what will happen if a new mutator arises that has the same mutation rate, but produces only mutations that change the length and the thickness of the fur? The new mutator does not produce any mutations in the parts of the genes where mutations are always harmful. When the temperature changes, this new mutator will be more likely to find an adaptive mutation first, before the original mutator does so. The rate at which it produces useful mutations will be higher. When it is the first to produce an improvement, the new mutator will spread throughout the population. In contrast, the original mutator will have missed an opportunity to re-establish itself in the population. It will continue to produce mutations that are harmful, and if it continues to be beaten by the new mutator in the search for useful mutations, it will eventually die out. The mutator that is better at targeting its mutations will be favoured by natural selection. The smarter mutator will win.
In this way, the genetic evolutionary mechanism can be expected to improve its own evolvability. It will discover and establish mutators that target mutations at future adaptive needs. But this search process is very inefficient. It relies on costly trial-and-error to discover better mutators, and most of the alternative mutators that are tried will be inferior. Eventually, evolution might improve the ability of the genetic system to discover better mutators. The genetic system may learn to target mutations in mutators, increasing the chances that mutations will produce a better mutator. But again, it will take a lot of costly trial-and-error to develop this ability. There is obvious room for improvement here. A system that could target genetic changes without having to go through this expensive trial-and-error process would have a substantial evolutionary advantage. It would be clearly superior to the types of mutational systems we have considered to this point. The result would be a further significant advance in evolvability.
Evolution overcame this fundamental limitation in the evolvability of mutational systems when it discovered genetic recombination, which is part of the wider process of sexual reproduction. The evolution of recombination and sex arguably have been the most significant advance in the evolution of the evolvability of genetic systems. The great majority of complex single celled creatures and multicellular organisms now reproduce sexually and use recombination. We will see that the immense significance of recombination is that it is far more efficient and effective at producing genetic changes that are targeted at future adaptive needs. Sex and recombination are successful strategies for organisms because they make them smarter at evolving.
How does recombination do this? The genetic changes produced by recombination are better targeted because of the way in which the changes are generated. Recombination produces changes by putting together different combinations of existing genes. Unlike mutational systems, it does not produce changes in the genes themselves. As we have seen, most changes to genes that are already functioning well are likely to be harmful. They will disrupt the effective operation of the gene, and produce a damaged organism. Instead recombination generates changes by mixing existing genes together in different combinations. So the basic building blocks that it uses to produce new effects are genes that are tried and tested, and that already work well in the organism. It does not risk making changes to existing genes that are proven performers.
It works like this. The cells of most sexually-reproducing organisms contain two sets of genetic material. One set is inherited from each parent. Each set includes a number of chromosomes, and each chromosome is formed of a long string of genes. Because the organism has two sets of chromosomes, it has a pair of each different type of chromosome, one from each parent.
Each egg or sperm produced by the organism will have only one set of chromosomes. So when it fuses with another egg or sperm to produce a new organism, the new organism will have two sets of chromosomes. But the single set of chromosomes in each egg or sperm is not just a set selected from the chromosomes of the organism that produced the egg or sperm. Each chromosome in the set is new. It is a combination of the pair of chromosomes of its type in the organism. So each chromosome in the egg or sperm is a combination of the chromosomes inherited by the organism from its parents. Each is produced by a process called crossing over. Parts of the chromosome from one of the organisms parent are swapped for the same part of the matching chromosome from the other parent. This produces a new, full chromosome that is a combination of the chromosomes inherited from each of the organisms parents. The result is that the organism produces eggs or sperm that contain chromosomes with different combinations of genes to those in its own chromosomes or in its parents. It is likely that the new combinations of genes put together in this way will produce offspring that differ from their parents. Natural selection will test which of the new combinations are better adapted.
But what makes the new combinations produced by this process particularly effective is not just that they contain genes that have proven to be effective in the past. It is not just that recombination produces genetic changes without risking the wrecking of existing genes by mutation. The great advantage of recombination is that it generates new combinations that are likely to be advantageous in the future. It tends to produce changes that are shown by past experience to be likely to be adaptive in the future.
To see how it does this, we first have to look at the type of genes that accumulate in the genetic material and that are available for recombination. We will again use the hypothetical example of a population of snow hares in an environment where average temperatures fluctuate significantly every few generations. We will first consider how the genes that affect the length and fitness of fur are likely to evolve over time if the population adapts through the production and testing of mutations, rather than by recombination. Then we will consider what sort of genetic changes would be produced when recombination creates different combinations of these accumulated genes.
Any complex process in an organism will be the result of the action of many genes. So there are likely to be a number of genes that are involved in the production of fur. There are also likely to be many possible changes in these genes that will have an effect on the length and thickness of fur. When environmental temperatures change, the first mutation that arises that changes the length and thickness of fur in a direction that matches the new temperature will be established in the population. Each time the temperature changes, up or down, new mutations that modify the fur for higher or lower temperatures will be established in the population. It is possible that in some cases adaptation will be produced by a mutation that undoes a mutation that was established in the past. But, until many mutated genes that modify the fur accumulate, it will be more likely that the change will be produced by a new mutation in one of the many genes that influence fur production.
As a result, the snow hare population will accumulate a number of mutated genes that have had the effect in the past of increasing or decreasing the length and the thickness of fur to match temperature changes. The extraordinary effectiveness of recombination is that it produces changes by putting together different combinations of these genesdifferent combinations of genes that each tend to change the length or thickness of fur. Rather than change fur randomly in directions that have not been adaptive in the past, all the changes tried out by recombination will tend to increase or decrease the length or thickness of fur. This will prove to be a very effective strategy for producing genetic changes if environmental temperatures continue to change in the future, as they have in the past. In contrast, the pattern of genetic changes produced by untargeted mutation would be far less efficient at adapting the population. Many random mutations would be lethal, or would produce changes that had nothing to do with changing the length and thickness of fur. So recombination is superior because it continually recreates combinations that have proved effective at some time in the past. And it tends to produce new combinations that change fur in ways that have proved adaptive in the past.
These principles can be expected to apply to any feature of an organism that must adapt genetically to some changing feature of the environment. The population will accumulate a number of genes that each will have changed the feature in a direction that produced adaptation at some time in the past. Recombination will produce changes by putting together different combinations of these genes. Each new combination is likely to change the feature in ways that have produced adaptation in the past. As a result, if future environmental changes are similar to past changes, recombination will be far more successful than mutation at targeting genetic changes at future adaptive needs. And the pattern of genetic changes that are tried out by a population using recombination will be far from random in relation to future adaptive needs.
The rate at which a population tries out new combinations of genes, and the content of the new combinations will be affected by a number of aspects of the recombination process. The rate and content will be influenced by: the frequency of crossing over between chromosomes when sperm and eggs are produced (if there is no swapping there will be no new combinations, and the chromosomes from the parents will be passed on to all offspring unchanged); the proportion of the chromosomes that are exchanged; and the location of different genes on the chromosomes (genes that are closely linked on the same chromosome are less likely to be separated and recombined by crossing over).
Because these aspects of the recombination process are themselves controlled genetically, they are evolvable. So natural selection can tune them to optimise the rate and content of the genetic changes that are produced through time. As a result, we can expect that the rate at which recombination produces changes in particular features of the organism will be tuned to the rate at which environmental changes affect the adaptedness of the organism. And we can expect that the content of the changes will be tuned so that changes are targeted more accurately at the types of adaptive opportunities that repeatedly confront the population.
Consistent with these expectations, the rate of recombination has been found to vary widely across different regions of the chromosomes of organisms that have been studied. And artificial selection in the laboratory has been able to readily change these rates of recombination. But little work has yet been done on assessing whether these differences in recombination rates are adaptive as a result of their ability to control the rate of production of genetic changes.
There are other ways in which evolution can improve the evolvability of organisms by modifying the pattern of genetic changes that they produce. It can do this by changing other factors that influence the genetic composition of offspring. For example, evolution can program organisms to select mates that are likely to contain useful genes, or to mate with non-relatives so that they produce young that differ more from themselves.
In all these ways, evolution can improve the evolvability of cells and multicellular organisms by shaping the pattern of genetic changes that they produce through time. Selection can tune the rate at which genetic changes are produced, and target them at likely future adaptive needs.
But it is not only by shaping the pattern of genetic changes that evolvability can be improved. The evolvability of a population of organisms depends on their ability to produce offspring with changes that are more likely to meet adaptive needs. But whether offspring will be better adapted will not be determined solely by the nature of the genetic changes that the population generates. It will also depend on what effects these genetic changes have on the way the offspring develop and function. It will depend on the way in which the genetic changes interact with the existing features of the organism to produce actual change in the organism. And this will depend on the way the organism is structured and organised.
From this we can see that selection can shape the pattern of changes produced by a population in a number of ways: it can modify the pattern of genetic changes, or it can change the organisation of the organism, or it can do both. If evolution were unable for some reason to modify the pattern of genetic changes, it would still be able to shape the pattern of actual changes that were produced by genetic changes. It would do this by changing the organisation of the organism so as to modify the effect of the genetic changes.
An example will make this clearer. Complex multicellular organisms have physiological systems that enable them to adapt to changes in their internal and external environment. These systems adapt the organism to changes that would otherwise disrupt its efficient functioning. The physiological and other adaptive systems also enable the organism to adapt to internal and external changes that occur as it develops from an egg into a fully-grown organism. Again, in the absence of these adaptive systems, the changes could damage the organism, and disrupt its proper development.
These internal adaptive systems can also enable the developing organism to survive genetic changes that would otherwise be lethal to the organism that carried them. This enables an organism to make fewer costly errors in the search for genetic adaptation. For example, consider two organisms that produced exactly the same pattern of genetic changes in their offspring: the organism with the better internal adaptive systems would produce fewer offspring that die or malfunction as a result of the genetic changes. The organism would take fewer errors to discover a particular adaptation. It would be better at evolving.
In recent years, various theorists have pointed to a number of other ways in which an organism could be organised to improve its evolvability. For example, if each of an organisms internal functions are organised into a separate module or compartment within the organism, evolution can explore changes in a particular module or compartment without necessarily disrupting the functioning of the rest of the organism. Redundancy of function can also contribute to greater evolvability: if more than one component of an organism can perform a particular function, evolution can explore changes in one of the components without disrupting that function. This is particularly significant in the organisation of the genetic material. If there are multiple copies of a particular gene, changes in one of the copies can be made without loss of function.
One final example: evolution could readily explore new structures in organisms that are constructed out of basic building blocks that can be combined together in different ways to produce a wide variety of structures. Small changes in the type, numbers and placement of building blocks could produce a diversity of structures. The recombination process would particularly lend itself to producing these changes. There is growing evidence that many key features of organisms are, in fact, constructed in this way.
But there is a fundamental problem with arguing that selection for improved evolvability is responsible for significant features of the way organisms are organised. If these features provide a pay-off only while a population is adapting to environmental changes, how do they justify their cost during periods of environmental stability? As we have seen, arrangements that produce genetic changes can survive these periods be minimising their cost to the organism. If the rate of mutation or recombination is sufficiently low, they can survive until the environment changes and they can provide an adaptive benefit. But there is no equivalent way in which a significant feature of the organism itself can minimise its cost to the organism during periods when it does not provide any benefit.
Can this difficulty be overcome in cases where improved evolvability helps the species as a whole to adapt and survive? Species that are better at evolving because they have features that improve evolvability can be expected to out-compete other similar species. They are likely to spread and produce new species at the expense of less evolvable organisms. But there is a problem. The longer-term advantage to the species will not maintain the features during periods when they fail to provide enough benefits to cover their costs. During these periods, organisms that have invested heavily in these features will be disadvantaged compared with those that have not. So the features will be maintained within the species only if they have other effects that are favoured by evolution. If this is the case, they will not owe their existence within the species to their contribution to improved evolvability. And changes in the features that improve evolvability will not be favoured unless they also have other advantages. But features that enhance evolvability will contribute to the success of the species.
Of course, this fundamental problem does not arise if populations are continually adapting genetically to their living and non-living environment. If this is the case, there will be a continual pay-off for features and processes that contribute to this adaptation. However, as we saw earlier, the environments of most animals are usually not thought to be continually changing in any significant way. Many species exist without apparent change for long periods in environments that seem stable. But, as we shall see now, this does not rule out entirely the possibility that highly-evolvable organisms are continually and profitably adapting to small changes in their living and non-living environment.
The living and non-living environment of any organism will always be seen to be changing if it is looked at on a scale that is fine enough. Examined closely enough, everything is continually changing, everything is in a state of flux. Temperatures, wind, humidity, and the intensity of sunlight change from hour to hour and day to day. The characteristics of food organisms, parasites, predators and other members of the species are also changing continually. So if a population of organisms were good enough at evolving, it might be able to find profitable ways to adapt to these continuous environmental changes.
This point is clearly demonstrated by complex multicellular organisms such as ourselves. Our heart rate, blood pressure, breathing, metabolic rate, and many other features of our bodies are being adapted continually to small-scale environmental changes. And the pay-off from this continual adaptation is apparently sufficient to justify the considerable investments made by our bodies in the systems that produce this adaptation.
As genetic evolvability improved throughout evolutionary history, populations can be expected to have got better and better at adapting profitably to finer and finer environmental changes. But are highly-evolved genetic systems that use recombination and sexual reproduction good enough at evolving to continually find profitable ways to adapt? Are they continually producing highly-targeted genetic changes that will adapt the organism to the small-scale environmental changes that all populations experience continually, no matter how stable their environment appears? If they are, organisational features that can improve evolvability might continually be able to provide a sufficient pay-off. They may be favoured continually by natural selection at the level of the individual within the population, and not contribute only to the success of the species.
I think that most populations of organisms are adapting continually in this way. If we could view a greatly sped up movie of a population of organisms over many generations, the population would appear to be adapting as continually and effectively as does a complex organism during its life. The population would appear alive in its own right, continually trying out highly-targeted changes, fluidly discovering new adaptations as conditions change, revising adaptations as necessary, and so on. But my view is little more than an intuitive guess at this time. Our current state of knowledge is a long way from enabling us to decide this issue conclusively.
* * * * *
In this Chapter we have looked at the ways in which evolution can be expected to have improved the evolvability of the genetic evolutionary mechanism. And we have seen that the genetic mechanism is much smarter than has commonly been thought by most evolutionists during this century. The pattern of genetic changes that are tried out by genetic systems, particularly those that utilise recombination, can be expected to be far smarter than random.
However, it is also easy to see that the evolvability of the genetic change-and-test mechanism is limited in a number of ways. As an evolutionary mechanism, it could be much improved. Its limitations leave substantial potential for further improvements in evolvability. And, as we shall see in the next two Chapters, this potential has progressively driven the evolution of new and better evolutionary mechanisms, and will continue to do so.
An obvious limitation of the genetic mechanism is that its capacity to learn is restricted. The targeting of genetic changes involves a form of learning. However, when organisms such as ourselves discover an adaptation that is effective in particular circumstances, we can learn to produce it again whenever those circumstances arise again. This is the ultimate in targeting. Trial-and-error is completely eliminated. But a genetic system cannot do this. It has no way of sensing the environment to distinguish one set of circumstances from another. So it cannot learn that particular adaptations are useful in particular environmental conditions. As a result, the genetic mechanism is unable to respond to specific environmental events by immediately producing a genetic adaptation that it has learnt is effective in those circumstances.
Nor can the genetic mechanism construct complex models of how its environment is likely to change through time, and use these to determine what genetic changes it should produce. Unlike us, it cannot use internal models to plan ahead, to test possible changes before they are tried out in practice, or to maintain adaptations that have no immediate benefits but will pay off in the future. A genetic system can miss a major beneficial adaptation by a single mutation, and never know. And it is unable to use a model of the direction of evolution to guide its search for better evolutionary adaptation.
But it was not these limitations of the genetic mechanism that immediately drove the progressive evolution of new adaptive mechanisms. The key limitation that did this was the inability of the genetic mechanism to adapt the organism during its life. As we have seen, a genetic change-and-test mechanism is unable to operate within the organism. This is because destructive competition must be suppressed within the genetic managers of single cells and the genetic managers of multicellular organisms. If genetic changes could arise and compete within the genetic managers during the life of the organism, destructive competition would undermine cooperation. Evolution therefore favoured arrangements that suppressed the possibility of this competition. But without competition between genetic alternatives, new and better adaptations could not be discovered within the organism. Genetic management could not provide a change-and-test process to adapt the organism during its life.
The inability of the genetic mechanism to adapt organisms continually during their life meant that there was enormous potential for the evolution of new mechanisms that could do this. This drove the evolution of new change-and-test mechanisms within the organism. Initially these were not evolutionary mechanisms. The adaptations that they produced were not passed from generation to generation. But, as we shall see, in humans these internal adaptive mechanisms have evolved into new evolutionary mechanisms that have overcome many of the limitations of the genetic mechanism. And we will see that they have to evolve further in the future to overcome other limitations and to improve the evolvability of humans as individuals and collectively.
. Bagley, R. J. and Farmer, J. D., (1991) Spontaneous emergence of a metabolism. In: Artificial Life II. (Langton, C. et al, eds.) New York: Addison and Wesley.
. Kauffman, S. A. (1993) The Origins of Order: Self-organisation and selection in evolution. New York: Oxford University Press.
. Stewart, J. E. (1995) Metaevolution. Journal of Social and Evolutionary Systems 18: 113-147.
. Wilson, D. S. and Sober, E. (1989) Reviving the Superorganism. Journal of Theoretical Biology 136: 337-356.
. For example, see Dawkins, R. (1986) The Blind Watchmaker. London: Longmans.
. Williams, G. C. (1966) Adaptation and natural selection. A critique of some current evolutionary thought. Princeton University Press: Princeton.
. Karlin, S. and McGregor, J. (1974) Towards a theory of the evolution of modifier genes. Theoretical Population Biology 5: 59-103; and Lieberman, U. and Feldman, M. W. (1986) Modifiers of mutation rate: a general reduction principle. Theoretical Population Biology 30: 125-142.
. Provided, of course, the mutator and the mutated gene are sufficiently closely linked on the same chromosome for the mutator to share in the success of the mutated gene.
. Leigh, E. G. (1973) The evolution of mutation rates. Genetics 73 (Supplement): 1-18; and Kondrashov, A. S. (1988) Deleterious mutations and the evolution of sexual reproduction. Nature 336: 435-440.
. Ghiselin, M. T. (1974) The economy of nature and the evolution of sex. University of California Press: Berkeley, California; Williams, G. C. (1975) Sex and evolution. Princeton University Press: Princeton, New Jersey; and Hamilton, W. D. (1980) Sex versus non-sex versus parasite. Oikos 35: 282-290.
. Jaenike, J. (1978) An hypothesis to account for the maintenance of sex within populations. Evolutionary Theory 3: 191-4; and Hamilton, W. D. (1980) Sex versus non-sex versus parasite. Oikos 35: 282-290.
. Hamilton, W. D., Axelrod, R. and Tanese, R. (1990) Sexual reproduction as an adaptation to resist parasites. Proc. Nat. Acad. Sci. USA 87: 3566-3573.
. Stewart, J. E. (1993). The maintenance of sex. Evolutionary Theory. 10: 195-202.
. Gillespie, J. H. and Turelli, M. (1989) Genotype-environment interactions and the maintenance of polygenic variation. Genetics 121: 129-138.
. Stewart: The maintenance of sex. op. cit.
. Leigh, E. G. (1970) Natural selection and mutability. Am. Nat. 104: 301-305; and Leigh: The evolution of mutation rates. op. cit.
. Ishii, K., Matsuda, H., Iwasa, Y. and Sasaki, A. (1989) Evolutionary stable mutation rate in a periodically changing environment. Genetics 121: 163-174; and Stewart, J. E. (1997) The Evolution of Genetic Cognition. Journal of Social and Evolutionary Systems 20: 53-73.
. Stewart: The Evolution of Genetic Cognition. op. cit.
 Of course, a mutator could also survive a period of stability if it produced no mutations during the period, and if it commenced producing mutations only once it was switched on by relevant environmental change. There is little clear evidence yet that such mechanisms have evolved, but it is early days. See Radman, M. (1999) Enzymes of evolutionary change. Nature 401: 866-9.
. Stewart: The Evolution of Genetic Cognition. op. cit.
. Moxon, R. E. and Thaler, D. S. (1997) The Tinkerers Evolving Tool-box. Nature 387: 659-662; and Brookes, M. (1998) Day of the Mutators. New Scientist 14 February 1998: 38-42. For references to other examples, see Pennisi, E. (1998) Molecular evolutionhow the genome readies itself for evolution. Science. 281: 1131-1134.
. Stewart: The Evolution of Genetic Cognition. op. cit.
. Stewart: The Maintenance of Sex. op. cit.; and Stewart: The Evolution of Genetic Cognition. op. cit.
. Stewart: The Evolution of Genetic Cognition. op. cit.
. For a fuller discussion of how these factors can influence the rate at which variation is produced and its content see Stewart: The Evolution of Genetic Cognition. op. cit.; and Mather, K. (1943) Polygenic inheritance and natural selection. Biological Reviews 18: 32-64.
. Evolution will favour controller genes that are linked closely enough to their effects to capture the benefits they create.
. Brooks, L. D. (1988) The evolution of recombination rates. In: The evolution of sex (Michod, R. E. and Levin, B. eds.) Sunderland Mass.: Sinauer; Lichten, M. and Goldman, A. S. H. (1995) Meiotic recombination hotspots. Annu. Rev. Genetics. 29: 423-444: and Pennisi: Molecular evolutionhow the genome readies itself for evolution. op. cit.
. Wagner, P. W. and Altenberg, L. (1996) Complex adaptations and the evolution of evolvability. Evolution 50: 967-976.
. Gerhart, J. and M. Kirschner (1997) Cells, Embryos and Evolution. New York: Blackwell Science Inc.
. Wagner and Altenberg: Complex adaptations and the evolution of evolvability. op. cit.
. Dawkins, R. (1989) The evolution of evolvability. In: Artificial Life. (Langton, C. ed.) New York: Addison and Wesley; and Gerhart and Kirschner: Cells, Embryos and Evolution. op. cit.
. Dawkins: The evolution of evolvability. op. cit.