One of the toughest problems for evolutionary biologists is to account for what might be called the problem of innovation — the appearance of a new beneficial biological feature that did not previously exist. Put another way, evolution by natural selection can be defined as “the survival of the fittest,” but evolution is unable to select for something that is not already there. It can only select among existing traits. When some new trait is required in order to survive, can evolution explain “the arrival of the fittest”? This clever turn of phrase comes from botanist Hugo de Vries (1848-1935), who was quoting a friend more than a century ago. Clearly the problem is not new.
Can this problem of innovation be solved? To answer this question it is necessary to start at the lowest level. Can something as seemingly simple as a new enzyme — one with a genuinely new function — evolve from something else? If you can’t evolve new enzymes, you can’t evolve even the simplest of cells, which are made up of hundreds of different proteins, many of them enzymes.
The work of Doug Axe and myself has centered on this question. What is possible for protein evolution by purely natural means? Not much so far. It can optimize an existing function if it is already there in substantial form, but it can’t generate new things, or even optimize a weak function that is not of the right design. It took years of work to reach that conclusion. Our new paper in the journal BIO-Complexity, “Model and Laboratory Demonstrations That Evolutionary Optimization Works Well Only if Preceded by Invention — Selection Itself Is Not Inventive,” outlines how the conclusion was reached,
… through studies in molecular biology and molecular evolution over the past fifteen years, beginning with a demonstration on two different enzymes that protein functions are much less tolerant of changes to their amino acid sequences than was commonly supposed. That result raised the question of just how small the target within sequence space is that would need to be hit for a new enzyme fold to be produced. The next project used limited randomization of the TEM-1 ?-lactamase to answer this question, resulting in a deeper challenge to the Darwinian view. Estimated to amount to a mere 1 part in 1074 of the sequence possibilities, the target corresponding to a new functional enzyme fold is far too small to be hit by any known evolutionary process. Building on this challenge, a full critique of the Darwinian explanation for protein folds was published in 2010.
By that point we had turned our attention to evolutionary invention on a smaller scale. Having argued that the evolutionary process cannot come up with new protein folds, we asked whether it is able to invent new functions for existing folds. To explore this, we conducted an extensive search for amino acid substitutions that would enable any of a variety of enzymes within the GABA-aminotransferase-like family to perform the action of one of their members — 8-amino-7-oxononanoate synthase. Using both rational and random experimental approaches in conjunction with a mathematical analysis of complex adaptation in bacterial populations we found evolutionary invention to be infeasible even on this small scale. Combining the fold-level results with results at the smaller scale, and considering all the relevant aspects of molecular evolution, we argued that the design interpretation of protein origins is much more plausible than the Darwinian interpretation.
Our previous work considered the difficulty of getting a new protein structure, or of adapting an old structure to a new function. But what if the starting point for an innovation already has a weak activity for the target function? This is the proposed means by which new proteins are thought to evolve — by optimization of pre-existing slight activities into full wild-type activities. Evolutionary biologists recognize the problem of innovation. Natural selection cannot select for something that is not already there. So they propose that some lucky pre-existing activity might become useful in some new context, and thus be selected for its beneficial activity. For example, a random sequence of DNA might happen to encode a random protein that just happens to provide a very weak activity the cell needs. Might it not be optimized to full function? Alternatively, an existing protein happens to have a novel enzymatic activity at very weak levels. Is it possible for that enzyme to be improved to wild-type levels?
We tested these two scenarios in two ways, in a virtual way using the program Stylus, and in the traditional experimental way, with actual proteins in the lab.
Stylus is a computer model that takes a DNA-like sequence and translates it into a two-dimensional protein-like shape, which can be scored for its similarity to an archetypal functional shape. (The shapes being used are Chinese Han characters. A brief description of the program is available here, here, here, and here.) The Stylus model world is analogous to the DNA to protein to function world of biology.
Using a random sequence whose product has very weak similarity to the target character as a starting point, Doug asked whether it could be evolved by random mutation and selection to produce a character that resembles the archetype that was its goal. Short answer — it could not. For details see the new paper describing the work.
Next he tested whether an already existing character with some weak similarity to the target could be evolved by mutation and selection to a proficient version of the target character. Once again, the answer was no. However, if the starting character was only six mutations away from optimization, it improved rapidly upon mutation and selection.
Finally, we tested both scenarios in the lab, first using a junk protein with weak activity against the antibiotic ampicillin, but without a properly folded enzymatic structure. It could not be improved by three rounds of random mutation and selection. In contrast, a weakly functional protein with a destabilized but properly folded structure could rapidly be optimized to wild-type levels and beyond.
Our conclusion? Unless the starting protein already exists as a functional fold of the right design, the protein’s activity cannot be optimized to wild-type levels. In other words, you’ll never get an innovation optimized, even with a pre-existing low level of the desired activity if the innovation is not already present in substantial form. By that I mean that the enzyme already has to be arranged to carry out the innovative function — its structure has to be of the right kind. Natural selection cannot create innovation, and it can’t even optimize pre-existing weak functions that are not of the right design to begin with.
The take-home message is summarized in the paper:
… it is the special arrangement of structural parts — and only this — that makes the exquisite performance of enzymes possible. Moreover, this seems to apply not just to enzymes but to everything else we recognize as having a specific function of any sophistication, whether from the living world or from the realm of human invention. What is true of enzymes is equally true of brains and eyes or of smartphones and sentences. We know that purposeful, intelligent action is necessary for achieving the special arrangement of letters that forms a coherent sentence or the special arrangement of materials that forms a smartphone, and the reasonableness of extending this knowledge to things like brains and eyes has been conceded even by people who believe that extension is ultimately mistaken. Geneticist Graham Bell, for example, acknowledged the point in these words:
A light bulb or a lathe are preconfigured in the mind, and constructed according to a plan. It is entirely reasonable to assume that beetles and daisies must be constructed after the same fashion, especially because they are much more complicated than anything that human ingenuity has so far managed to devise.
The title of Bell’s book — Selection: The Mechanism of Evolution — leaves no doubt as to his favored alternative explanation of living things, but in placing the above words on his opening page, Bell also made it clear what selection must explain. The success of evolutionary theory requires not just that selection have real measurable effects but, more importantly, that the production of remarkable things — things beyond human ingenuity — be among those effects.
To reiterate, it is not enough to have a weak level of starting activity available for evolution to work on. This proposed model for evolution fails for two reasons. First, one has to postulate that all necessary activities somehow came to be before they were needed, even if only to a small degree. Second, even if all the things necessary for life existed before they were needed, unless these activities could be improved to the levels we see today, those starting activities would be of no use.
…the suggestion that all the remarkable things in the living world pre-existed is relevant only if the versions that might have pre-existed really could have been honed by selection into the impressive versions we now see.
… we show here that invention is not superfluous. Although junk can be selected under some circumstances, the optimized endpoint after evolution has done all it can do is still junk. For the endpoint to be a highly refined functionally specific structure, the essential aspects of that structure have to be present from the outset.
To get a new enzyme, and by extension any sophisticated complex thing, requires that the new enzyme be already there, in essential form, before it can be optimized by selection. It has to be shaped by intelligent design before selection can act upon it. In other words, evolution is incapable of innovation — only intelligence can invent a sophisticated new cellular structure such as an enzyme, or an eye.