In this essay I reply to what I consider to be the most important claim made by any critic of intelligent design: that direct experimental evidence has shown that evolution can indeed generate irreducibly complex biochemical systems. As I will show below, the claim is false.
Briefly, in his book Finding Darwin's God (Harper Collins, 1999) Kenneth Miller quite rightly says that a "true acid test" of Darwinism is to see if it could regenerate an irreducibly complex system that was knocked out using the tools of molecular biology. He then discusses work from the laboratory of Barry Hall of the University of Rochester on the lac operon of the bacterium E. coli. Miller strongly implies that natural selection pieced together the whole pathway in Hall's experiments, but in fact it only replaced one component (and even then it could only replace the component with a spare near-copy of the original component). When two or more components were deleted, or when the bacterium was cultured in the absence of an artificial chemical (called IPTG), no viable bacteria could be recovered. Just as irreducible complexity would predict, when several steps must be taken at once, natural selection is a poor way to proceed.
Since Miller calls this work the "acid test," that of course means that other examples he discusses in his book are not "acid tests"; they are at best indirect arguments. The more indirect the argument, the easier for Darwinists to overlook or conceal difficulties.
II. "A True Acid Test"
Brown University cell biologist Kenneth Miller has written a book recently defending Darwinism from a variety of critics, including me. In a chapter devoted to rebutting Darwin's Black Box, he marshals an array of examples which, he asserts, tell against claims of irreducible complexity. However, for all of his counterexamples I either disagree that he is dealing with irreducibly complex systems, disagree that he is focusing on the irreducibly complex aspects of a system, or disagree that his brief scenarios successfully answer the challenge of irreducible complexity (for an example, see my critique on this website of his blood clotting scenario). In this section I focus on his most serious claim--that an experiment has shown natural selection can construct an irreducibly complex system.
Professor Miller correctly states that "a true acid test" of the ability of Darwinism to deal with irreducible complexity would be to "[use] the tools of molecular genetics to wipe out an existing multipart system and then see if evolution can come to the rescue with a system to replace it." (Miller 1999, 145) Therefore the most important and novel part of Miller's rebuttal is his claim that experimental work in a bacterial system has actually succeeded in producing an irreducibly complex system by natural selection. In a section entitled "Parts is Parts," in which he discusses the careful work over the past quarter-century of Barry Hall of the University of Rochester on the experimental evolution of a lactose-utilizing system in E. coli, Miller excitedly remarks:
Think for a moment--if we were to happen upon the interlocking biochemical complexity of the reevolved lactose system, wouldn't we be impressed by the intelligence of its design? Lactose triggers a regulatory sequence that switches on the synthesis of an enzyme that then metabolizes lactose itself. The products of that successful lactose metabolism then activate the gene for the lac permease, which ensures a steady supply of lactose entering the cell. Irreducible complexity. What good would the permease be without the galactosidase? . . . No good, of course. By the very same logic applied by Michael Behe to other systems, therefore, we could conclude that the system had been designed. Except we know that it was not designed. We know it evolved because we watched it happen right in the laboratory! (Miller 1999, 146)
I will show this picture is grossly exaggerated.
Here is a brief description of how the lac operon functions. The lac operon of E. coli contains genes coding for several proteins which are involved in metabolism of the disaccharide lactose. One protein of the lac operon, called a permease, imports lactose through the otherwise-impermeable cell membrane. Another protein is an enzyme called B-galactosidase, which can hydrolyze the disaccharide to its two constituent monosaccharides, galactose and glucose, which the cell can then process further. Because lactose is rarely available in the environment, the bacterial cell switches off synthesis of the permease and B-galactosidase to conserve energy until lactose is available. The switch is controlled by another protein called a repressor, whose gene is located next to the operon. Ordinarily the repressor binds to the lac operon, shutting it off by physically interfering with expression of the operon. In the presence of the natural "inducer" allolactose (a by-product of lac B-galactosidase activity) or the artificial chemical inducer isopropylthiogalactoside (IPTG), however, the repressor binds to the inducer and releases the operon, allowing the lac operon enzymes to be synthesized by the cell.
When I first read this section of Miller's book I was quite impressed by the prospect that actual experiments--not theoretical, "just-so" stories--had produced a genuine, non-trivial counterexample to irreducible complexity. After going back to read Professor Hall's publications, however, I found that the situation was considerably different. Not only were Hall's results not what I expected based on Miller's description, in fact they fit most naturally within a framework of irreducible complexity and intelligent design. The same work that Miller points to as an example of Darwinian prowess I would cite as showing the limits of Darwinism and the need for design.
III. Adaptive Mutation
So what did Barry Hall actually do? To study bacterial evolution in the laboratory, in the mid 1970's Hall produced a strain of E. coli in which the gene for just the B-galactosidase of the lac operon was deleted. He later wrote:
All of the other functions for lactose metabolism, including lactose permease and the pathways for metabolism of glucose and galactose, the products of lactose hydrolysis, remain intact, thus re-acquisition of lactose utilization requires only the evolution of a new B-galactosidase function. (Hall 1999)
Thus, contrary to Miller's own criterion for "a true acid test," a multipart system was not "wiped out"--only one component of a multipart system was deleted.
Without B-galactosidase, Hall's cells could not grow when cultured on a medium containing only lactose as a carbon source. However, when grown on a plate that also included alternative, useable nutrients, bacterial colonies could be established. When the other nutrients were exhausted the colonies stopped growing. However, Hall noticed that after several days to several weeks, hyphae grew on some of the colonies. Upon isolating cells from the hyphae, Hall saw that they frequently had two mutations, one of which was in a gene for a protein he called "evolved B-galactosidase," ("ebg") which allowed it to metabolize lactose efficiently. (Despite considerable efforts by Hall to determine it, the natural function of ebg remains unknown) (Hall 1999). The ebg gene is located in another operon, distant from the lac operon, and is under the control of its own repressor protein. The second mutation Hall found was always in the gene for the ebg repressor protein, which caused the repressor to bind lactose with sufficient strength to de-repress the ebg operon.
The fact that there were two separate mutations in different genes--neither of which by itself allowed cell growth (Hall 1982a)--startled Hall, who knew that the odds against the mutations appearing randomly and independently were prohibitive (Hall 1982b). Hall's results and similar results from other laboratories led to research in the area dubbed "adaptive mutations." (Cairns 1998; Foster 1999; Hall 1998; McFadden and Al Khalili 1999; Shapiro 1997) As Hall later wrote,
Adaptive mutations are mutations that occur in nondividing or slowly dividing cells during prolonged nonlethal selection, and that appear to be specific to the challenge of the selection in the sense that the only mutations that arise are those that provide a growth advantage to the cell. The issue of the specificity has been controversial because it violates our most basic assumptions about the randomness of mutations with respect to their effect on the cell. (Hall 1997)
The mechanism(s) of adaptive mutation are currently unknown. While they are being sorted out, it is misleading to cite results of processes which "violate our most basic assumptions about the randomness of mutations" to argue for Darwinian evolution, as Miller does.
IV. A Nearly-Identical Active Site
The nature of adaptive mutation aside, a strong reason to consider the lac/ebg results quite modest is that the ebg proteins--both the repressor and B-galactosidase--are homologous to the E. coli lac proteins and overlap the proteins in activity. Both of the unmutated ebg proteins already bind lactose. Binding of lactose even to the unmutated ebg repressor induces a 100-fold increase in synthesis of the ebg operon. (Hall 1982a) Even the unmutated ebg B-galactosidase can hydrolyze lactose at a level of about 10% that of a "Class II" mutant B-galactosidase that supports cell growth. (Hall 1999) These activities are not sufficient to permit growth of E. coli on lactose, but they already are present. The mutations reported by Hall simply enhance pre-existing activities of the proteins. In a recent paper (Hall 1999) Professor Hall pointed out that both the lac and ebg B-galactosidase enzymes are part of a family of highly-conserved B-galactosidases, identical at 13 of 15 active site amino acid residues, which apparently diverged by gene duplication more than two billion years ago. The two mutations in ebg B-galactosidase that increase its ability to hydrolyze lactose change two nonidentical residues back to those of other B-galactosidases in ebg's phylogenetic class, so that their active sites are identical. Thus--before any experiments were done--the ebg active site was already a near-duplicate of other B-galactosidases, and only became more active by becoming a complete duplicate. Significantly, by phylogenetic analysis Hall concluded that those two mutations are the only ones in E. coli that confer the ability to hydrolyze lactose.
The phylogenetic evidence indicates that either Asp-92 and Cys/Trp-977 are the only acceptable amino acids at those positions, or that all of the single base substitutions that might be on the pathway to other amino acid replacements at those sites are so deleterious that they constitute a deep selective valley that has not been traversed in the 2 billion years since those proteins diverged from a common ancestor. (Hall 1999)
Such results hardly support extravagant claims for the creativeness of Darwinian processes.
V. Caveats Unmentioned
A critical caveat not mentioned by Kenneth Miller is that the mutants that were initially isolated would be unable to use lactose in the wild--they required the artificial inducer IPTG to be present in the growth medium. The reason is that a permease is required to bring lactose into the cell. However, ebg only has a B-galactosidase activity, not a permease activity, so the experimental system had to rely on the pre-existing lac permease. Since the lac operon is repressed in the absence of either allolactose or IPTG, Hall decided to include the artificial inducer in all media up to this point so that the cells could grow. Thus the system was being artificially supported by intelligent intervention. Hall clearly wrote:
At this point it is important to discuss the use of IPTG in these studies. Unless otherwise indicated, IPTG is always included in media containing lactose or other B-galactoside sugars. The sole function of the IPTG is to induce synthesis of the lactose permease, and thus to deliver lactose to the inside of the cell. Neither the constitutive nor the inducible evolved strains grew on lactose in the absence of IPTG. (Hall 1982b)
With further growth and selection, Hall isolated secondary mutants with improved B-galactosidase activity. These mutants all had the same two changes (mentioned above) at positions 92 and 977 of ebg B-galactosidase. Hall discovered that, in addition to hydrolyzing lactose, the double mutants could also synthesize some allolactose, just as the homologous lac B-galactosidase can do, allowing them to induce expression of the lac operon without further need of IPTG. Critically again, however, the lac permease induced by the action of the double mutant ebg is a pre-existing protein, part of the original lac operon, and was not produced in the experiment by the selection procedures. In the absence of that required component, the bacteria cannot use lactose.
Miller's prose ("Irreducible complexity. What good would the permease be without the galactosidase?") (Miller 1999, 146) obscures the facts that most of the system was already in place when the experiments began, that the system was carried through nonviable states by inclusion of IPTG, and that the system will not function without pre-existing components. In contrast to Miller, Hall himself is cautious and clear about the implications of his results.
The mutations described above have been deliberately selected in the laboratory as a model for the way biochemical pathways might evolve so that they are appropriately organized with respect to both the cell and its environment. It is reasonable to ask whether this model might have any relationship to the real world outside the laboratory. If it is assumed that the selection is strictly for lactose utilization, then a growth advantage exists only when all three mutations are present simultaneously. (Hall 1982a)
Hall is nonetheless optimistic.
Any one of the mutations alone could well be neutral (it is unlikely that any would be disadvantageous); but neutral mutations do enter populations by random chance events, and are fixed by a chance process termed genetic drift. (Hall 1982a)
However, if a mutation is not selected, the probability of its being fixed in a population is independent of the probability of the next mutation. Such a system is irreducibly complex, requiring several steps to be taken independently of each other before having selective value. If three mutations are required before there is any selective value, then the cumulative probability starts to become very small indeed, even considering the size of bacterial populations. In the present case Hall argued that a small selective value might accrue after the second mutation (in the ebg repressor). (Hall 1982a) However, I find his rationale unconvincing and having little experimental support. Furthermore, Professor Hall does not discuss the implications of the requirement for the preexisting lac permease gene.
Miller ends the section in his typical emphatic style: (1)
"No doubt about it--the evolution of biochemical systems, even complex multipart ones, is explicable in terms of evolution. Behe is wrong." (Miller 1999, 147)
I disagree. Leaving aside the still-murky area of adaptive mutation, the admirably-careful work of Hall involved a series of micromutations stitched together by intelligent intervention. He showed that the activity of a deleted enzyme could be replaced only by mutations to a second, homologous protein with a nearly-identical active site; and only if the second repressor already bound lactose; and only if the system were also artificially supported by inclusion of IPTG; and only if the system were also allowed to use a preexisting permease. Such results are exactly what one expects of irreducible complexity requiring intelligent intervention, and of limited capabilities for Darwinian processes.
Cairns, J. (1998). Mutation and cancer: the antecedents to our studies of adaptive mutation. Genetics 148, 1433-1440.
Foster, P. L. (1999). Mechanisms of stationary phase mutation: a decade of adaptive mutation. Annual Review of Genetics 33, 57-88.
Hall, B. G. (1982a). Evolution of a regulated operon in the laboratory. Genetics 101, 335-344.
Hall, B. G. (1982b). Evolution on a Petri dish: The evolved ? B-galactosidase system as a model for studying acquisitive evolution in the laboratory. In Evolutionary Biology. (M. K. Hecht, B. Wallace, and G. T. Prance, Eds.) pp. 85-150. (Plenum Press: New York.)
Hall, B. G. (1997). On the specificity of adaptive mutations. Genetics 145, 39-44.
Hall, B. G. (1998). Adaptive mutagenesis: a process that generates almost exclusively beneficial mutations. Genetics 102-103, 109-125.
Hall, B. G. (1999). Experimental evolution of ebg enzyme provides clues about the evolution of catalysis and to evolutionary potential. FEMS Microbiology Letters 174, 1-8.
McFadden, J. and Al Khalili, J. (1999). A quantum mechanical model of adaptive mutation. Biosystems 50, 203-211.
Miller, K. R. (1999). Finding Darwin's God: a scientist's search for common ground between God and evolution. (Cliff Street Books: New York.)
Shapiro, J. A. (1997). Genome organization, natural genetic engineering and adaptive mutation. Trends in Genetics 13, 98-104.
1. Miller's prose is often exaggerated and sometimes borders on the bombastic. Perhaps he uses such a relentlessly emphatic style in the hope of overwhelming readers through the sheer force of his words. Perhaps he just has a much-larger-than-average share of self-confidence. Fortunately, in this section on the "acid test," experiments exist to show that his prose is bluster. Let me be blunt--Miller always writes (or speaks) with the utmost confidence, even when experiments show him to be quite wrong. I would caution readers of his work not to be swayed by his tone, whose confidence never wavers even when the evidence does.