by Merry Youle
…one would have to say that natural selection does not work as an engineer works. It works like a tinkerer – a tinkerer who does not know exactly what he is going to produce but uses whatever he finds around him…to produce some kind of workable object.
– François Jacob, 1977 (1)
Evolution skeptics point to complex biological features as evidence that the amazing intricacy of the biological world could not be the product of blind mutation. The de novo creation of such complexity, they argue, would require far too many coordinated mutational changes before the new form provided any fitness benefit. Thus, evolution simply cannot explain such apparent leaps. Evolutionists, beginning with Darwin, and continuing today, have posited that complex features do not require breathtaking leaps, but rather arise through stepwise modification of existing forms. Such random tinkerings, each requiring comparatively few mutations, occasionally generate a new form that is slightly more complex. For this process to proceed further, the new form must benefit the organism. Only then can it be maintained in the population long enough to serve as a stepping-stone to yet greater complexity. The probability of success is low, and many of these experiments will lead to evolutionary dead-ends. Does evolution proceed in this haphazard way? Can small random steps add up to an evolutionary leap?
Taking a leap in silico
Evolution experiments in silico say "yes." Here the evolving "organisms" are computer algorithms. Each digital organism possesses a heritable "genotype" that, when computationally expressed, yields a corresponding "phenotype." One such study by Lenski and colleagues (2) started with an ancestor with a circular genome of 50 "genes" that were expressed in sequence, generation after generation. Each gene was one of a set 26 executable instructions that included, among others, a copy instruction that carried out self-replication. No sex here, no horizontal gene transfer – only binary fission and vertical inheritance. To simulate evolution, both heritable variation and selection were introduced. Variation arose from random mutation during genome replication. These mutations included substitutions in which one instruction was randomly replaced by another, as well as indels in which one instruction was inserted or deleted. Each genotype displayed a different phenotype with a particular replication efficiency and computational metabolism. Like biological metabolism, computational metabolism required energy.
Some specific gene combinations, when expressed in the correct sequence, were able to perform one of nine logic functions of varying complexity. The ancestral genotype could perform none of the nine. Acquisition of even the simplest function by the ancestor required more than one mutation. The more complex the function, the more mutational steps required for even the most direct evolutionary route. Each organism was rewarded with energy packets for each function it performed during its "lifetime." The more complex the function, the more packets were awarded. Increased genome length also earned energy packets. These energy boosts, in turn, increased the organism's reproductive fitness. Over the course of hundreds or thousands of generations, driven by "selection" based on "reproductive fitness," a dominant organism evolved that was capable of performing all nine logic functions, including the most complex.
To explore what this in silico work can tell us about the evolution of complex features, the researchers focused on the acquisition of the most complex of the nine logic functions. Theoretically, the ancestor could gain this capability through a minimum of 16 mutations. As you would expect, the random paths taken by this simulated evolution were less direct; likewise, success was not guaranteed. However, that complex capability evolved in 23 out of 50 identical experiments. Among those 23, the most direct route comprised 51 steps; the longest one, 721 steps – a step being each time the offspring differed from the parent. Some of those steps were single mutations, some double or triple. The route was not a steady upward trajectory toward greater fitness; some steps were lateral, some were even deleterious relative to the parent. Conclusions? The probability of evolution of a beneficial complex logic function was high even when the route spanned several hundred small steps. Although this achievement was repeatable, it is extremely unlikely that it would ever arise by the same convoluted path twice.
An additional series of similar experiments tested the hypothesis that the intermediate steps en route to evolution of a complex function must themselves confer a fitness benefit. The researchers ran 50 experiments under "environmental conditions" that did not energetically reward any of the eight simpler functions. Energy packets would be given only for carrying out the most complex logic function. The result ? Zero out of 50 achieved that capability.
This in silico work provides supportive evidence that more complex features can evolve given the prior evolution of simpler intermediate features that are, themselves, beneficial. Nevertheless, it is quite a leap from in silico to in vivo. It would surely be more convincing to a skeptic if we could point to an in vivo example of a similar leap. Time to call on the phage !
A successful leap in vivo
Phage λ has been studied intensively for decades and decades, more than any other phage. The particular domesticated strain propagated in most labs lacks the side tail fibers of the ancestral wild-type phage, fibers that functioned in host recognition. As a result, this λ strain is absolutely dependent on the receptor binding protein (RBP) at the tip of its tail for adsorption to its E. coli host. That RBP recognizes and binds to a specific porin, LamB, located in the cell's outer membrane. Subsequently λ's chromosome crosses that first membrane through this porin. Because LamB is part of E. coli's maltose transport system, this frugal bacterium synthesizes it only when it needs to import maltose for its carbon and energy source. When E. coli is grown in the lab on glucose, LamB synthesis is repressed and consequently λ's adsorption rate is barely detectable. Throughout its long lab history, λ has never been known to adsorb to another receptor. Could it possibly do that? To answer this, researchers gave λ a strong incentive to take that leap, and then closely observed what happened.
The incentive was indeed strong, and prolonged. Meyer and colleagues (3) cultured λ together with E. coli for 28 days in a medium that contained glucose, but no maltose. Every day they transferred 1% of the culture into fresh medium. E. coli grows well under such conditions, but λ barely persists. A small phage population (1 virion per 102 to 103 cells) subsists by infecting the small number of bacteria that spontaneously synthesize LamB. This placed the phage under very strong selection pressure to evolve a workaround. This λ did – by quickly and repeatedly evolving the ability to use a different receptor. Quickly here means within 12 days on average under these experimental conditions. Repeatability was demonstrated by λ's success in 24 out of 96 replicate experiments. If this was so easy for the phage, you might be wondering why no one had observed this previously. The answer: anyone testing λ's ability to use other receptors would have ended their experiments after hours, not days or weeks. In other words, researchers had asked what were the receptor-binding capabilities of their phage stock, not what abilities could the phage evolve under extreme duress.
Did all 24 winners find the same solution? All did evolve to use the same receptor, porin OmpF. To a researcher predicting λ's new receptor choice, OmpF would have seemed the most likely candidate because this porin is the one most similar structurally to LamB. Moreover, unlike LamB and most porins, OmpF is synthesized constitutively, thus always available to the phage. Genome sequencing demonstrated that all 24 winners had also evolved essentially the same mechanism; they all possessed a set of four mutations located within the same four narrow regions within the RBP gene.
Granted, this change of receptor was a small leap that involved only one phage protein, but it was still a leap. All four mutations were required before the phage was able to use OmpF. This suggests that all four must have occurred simultaneously. However, the probability of this happening within 12 days in these small phage populations was so low as to defy even phage ingenuity. Furthermore, this was not the achievement of one exceptional phage, a singular event not to be repeated. Rather it was achieved by 24 out of 96 replicate cultures. Something else must have been going on to account for this rapid and consistent success.
To explore this puzzle, the researchers took a closer look at the ongoing interactions between phage and host over the course of these experiments. Faced with long-term maltose-free conditions, the phage population was desperate to exploit whatever LamB was made by the bacteria. The host, having no need for maltose but being subjected to persistent phage attack, evolved tighter repression of LamB synthesis in all of the cultures. This further reduced the minimal amount of LamB synthesized, and thereby decreased the number of receptors exposed on the cell surface. Now a phage colliding with a bacterium was less likely to locate a receptor. This, in turn, increased the pressure on the phage to convert as many of those lucky hits as possible into successful binding and chromosome delivery.
The researchers found that the phage achieved improved adsorption to LamB by mutations in the RBP gene before they had acquired the ability to use OmpF. The evidence ? All of these RBP mutations were nonsynonymous changes, which indicated that they were under positive selection. They all clustered at specific points in the RBP structure that interacted closely with LamB. Lastly, very similar mutations were found repeatedly among the successful phage lineages. Therefore, the researchers concluded that λ had not made one giant leap to OmpF binding, but rather had progressed by a series of mutational steps that themselves conferred immediate benefit while the phage was still dependent on LamB as its receptor.
This begs another question. If it was this easy, with beneficial stepping-stones all along the way, why did only one in four phage lineages succeed ? One possible reason considered was that some of the mutations that yielded an immediate benefit also hindered evolution of OmpF-binding capability. No evidence was found for this in further experiments. Instead, the researchers found that the phage switch to OmpF had been blocked by host mutations that arose during the experiment. These included mutations that altered two other proteins in the E. coli maltose transport system, ManY and ManZ. These proteins form maltose transport channels through the cell membrane, channels λ also uses for passage of its chromosome from the periplasm into the cytoplasm. Thus, these host mutations obstructed λ infection no matter which receptor was used; they even drove the phage to extinction in some cultures. Thus, whether λ made the leap to OmpF or failed was dependent not only on the phage, but also on the dynamic co-evolution of its host.
Is there a take-home message here ? Both of these experimental approaches support the hypothesis that evolutionary leaps are, in actuality, a sequence of small, beneficial steps – but caution ! The inviting stepping-stones-across-a-burbling-brook analogy is misleading. Imagine stepping blindly is every direction; imagine stepping-stones that continually shift or disappear entirely. Even the destination shore doesn't stay put. Environmental conditions fluctuate. Co-evolution is forever moving the goalposts. The possibilities are endless. The outcome of any step is uncertain, but here we are, and the evolutionary dance continues.
Merry Youle is a phageophilic microbiology writer/editor and STC blogger emerita living on the Big Island of Hawaii. Her books include the recently-published Thinking Like a Phage and the 2015 title, Life in Our Phage World. This post is excerpted from her book in progress, Phage in Community.
(3) Meyer JR, Dobias DT, Weitz JS, Barrick JE, Quick RT, and Lenski RE. 2012. Repeatability and contingency in the evolution of a key innovation in phage lambda. Science, 335 (6067), 428 – 432. PMID 22282803 (Paywall; Open Access PDF here)