Limited Evolutionary Potential


Sean D. Pitman M.D.

© November 2006 



Table of Contents














The Claim


Michael Behe, professor of biochemistry at Lehigh University, boldly claims that,  "Molecular evolution is not based on scientific authority.  There is no publication in the scientific literature in prestigious journals, specialty journals, or books that describe how molecular evolution of any real, complex, biochemical system either did occur or even might have occurred.  There are assertions that such evolution occurred, but absolutely none are supported by pertinent experiments or calculations." 1   



The Controversy


Since the publishing of Behe's book, Darwin's Black Box in 1996, a fair bit of controversy has arisen over such statements. Surprisingly, many evolutionary scientists seem to grudgingly agree with Behe, at least in some limited way.  For example, microbiologist James Shapiro of the University of Chicago declared in National Review that, "There are no detailed Darwinian accounts for the evolution of any fundamental biochemical or cellular system, only a variety of wishful speculations" (Shapiro 1996).  In Nature, University of Chicago evolutionary biologist, Jerry Coyne, noted that, "There is no doubt that the pathways described by Behe are dauntingly complex, and their evolution will be hard to unravel. . . . [W]e may forever be unable to envisage the first proto-pathways" (Coyne 1996).  In Trends in Ecology and Evolution Tom Cavalier-Smith, an evolutionary biologist at the University of British Columbia, wrote, "For none of the cases mentioned by Behe is there yet a comprehensive and detailed explanation of the probable steps in the evolution of the observed complexity. The problems have indeed been sorely neglected--though Behe repeatedly exaggerates this neglect with such hyperboles as 'an eerie and complete silence'" (Cavalier-Smith 1997).  Evolutionary biologist, Andrew Pomiankowski, agreed. In New Scientist, he challenged anyone to, "Pick up any biochemistry textbook, and you will find perhaps two or three references to evolution. Turn to one of these and you will be lucky to find anything better than 'evolution selects the fittest molecules for their biological function'" (Pomiankowski 1996). In American Scientist, Yale molecular biologist, Robert Dorit, suggested that, "In a narrow sense, Behe is correct when he argues that we do not yet fully understand the evolution of the flagellar motor or the blood clotting cascade" (Dorit 1997). 7



Observing the Evolution of a Complex System


Obviously though, there are many more scientists who passionately disagree with Behe's position.  These scientists argue quite strongly that the mechanism is known in detail and that it is observed in action all the time producing "irreducibly complex" systems with complex specified information.  Perhaps one of Behe's better opponents is Kenneth Miller (biologist, Brown University).  In his 1999 book, Finding Darwin's God, one of Miller's challenges of Behe's position includes a 1982 research study by professor Barry Hall, an biologist from the University of Rochester.  

What Hall did was indeed very interesting. He deleted a gene (lacZ) in a type of bacteria (E. coli) that makes a lactase enzyme (galactosidase).  This lactase enzyme converts a sugar called lactose into the sugars glucose and galactose.  E. coli then process glucose and galactose further to extract energy.  One might think that when Hall deleted the gene that codes for the lactase enzyme that these bacteria would never be able to use lactose for energy again.  However, when Hall exposed the mutant bacteria to lactose enriched growth media, that they quickly modified a different gene, which Hall named the "evolved Beta-galactosidase gene" (ebg), to produce a pretty good lactase enzyme.  This is interesting because the original gene product did not have the lactase function.  Only after a key random mutation was this genetic sequence able to produce a protein with the lactase function.2, 3

Behe counters by arguing that as far as the active sites of the lac and ebg Beta-galactosidase enzymes are concerned, that they are essentially the same with both being a part of a family of highly conserved Beta-galactosidases - identical at 13 of 15 active-site amino acid residues.  The two mutations in the ebg  Beta-galactosidase,  that increase its ability to hydrolyze lactose, change the two non-identical residues back to those of the other Beta-galactosidases.  So, before the evolution of the lactase ability of the ebg gene, its active site was already a near duplicate of other Beta-galactosidases.8

Even so, this really was quite an amazing experiment in that a novel enzymatic function, which was not present in the entire gene pool prior random mutation and natural selection, did in fact evolve in real time.  According to Miller and Hall, and many others quoting the same or similar experiments, such experiments give demonstrable proof of the proposed evolutionary mechanism in action.  Obviously then, Behe does not know what he is talking about . . . or does he? Consider that fairly often things are not quite as they would appear at first glance.



At First Glance


At first glance, Hall's experiment does indeed seem to be a "real time" example of improvement through mutation and natural selection.  Obviously then, this is evolution in action.  Certainly it is, but it might be a bit different from what one might expect.  

It is not a difficult thing to evolve a particular function if that function is only one point mutation (one step) away from realization. The odds that one particular point mutation will happen rapidly in an average sized bacterial colony are extremely good. A random step/mutation across such a small gap isn't a problem at all.  The real question is, what are the odds that such a potentially beneficial protein-based structure will actually exist in the potential of sequence space just one step away from what is currently present within a given genome?  It is this question that those like Hall and Miller and other evolutionary scientists do not address.  

However, if a small crack in the evolutionary sidewalk can be crossed, why can't these cracks be added up over time, one new little function at a time, until we end up with the awesome complexity and variations of life forms that we see today? After all, the theory of evolution is based on a fairly simple idea that random mutations create diversity while natural selection acts as a guide to select those changes that best aid the survival of that gene pool.  Thus, evolution, when it happens, is not truly a random process - although random events are involved in the process.  Natural selection guides the changes over time, adding them together, one by one, until the vast diversity of life forms that we see today is the result.  Certainly then, it should at least be technically possible to add even the smallest changes together over time to produce all the variety in life forms and functionally complex biofunctions that we see in the world today. Certainly this sounds like quite a reasonable possibility - at first glance. 



A Closer Look


Most descriptions of Hall's experiments end with E. coli evolving the lactase function back again.  This is very interesting because Hall's actual experiments did not end there.  After his initial success, Hall wondered if any other genes would be able to evolve the lactase function.  So, he deleted the ebg gene as well as the lacZ genes to test this hypothesis. And, something most interesting happened - nothing.  No new gene or portion of DNA evolved the lactase function despite tens of thousands of generations of time, a huge population size, high selection pressure, and a high mutation rate.  Now that is just fascinating . . .

In order to understand what happened, lets consider Hall's experiment in just a little more detail. Behe summarizes Hall's methods pretty well in the following description:


Without Beta-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 Beta-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) 8


In short, Hall did in fact delete the lacZ gene (as well as other lac genes with other related functions) in E. coli bacteria.  These mutant bacteria then evolved the ability to use lactose over a very short period of time in a non-lethal lactose enriched environment.  They were able to do this with the use of a very fortuitous "spare tire" gene (ebg) that, with a single point mutation, was able to achieve enough lactase activity to give the cell a selective survival/reproductive advantage (a second mutation was also required in the promoter region, but this will be discussed in more detail below). 



What are the Odds?


What are the odds?!  That is the real question here.  What is the likelihood that some portion of the collective E. coli genome in a particular colony of 10 billion would be so close to producing a protein structure with a particular function, or any other beneficial function at an equivalent level of functional complexity?  How often would this happen? - on average?  Would different types of functions require different minimum sequence size or degrees of specificity with regard the the specific arrangement of the amino acid residue "parts"?  If so, would greater minimum size and/or specificity requirements result in differences in the likely gap distance between what exists in a genome and what might be beneficial if it were ever found via a random search through the vast potential of sequence space?

In order to begin answering this question, it might be a good idea to take a closer look at the actual genetic sequences involved with this experiment.  The lacZ gene is quite long.  It consists of approximately 3,500 base pairs in the DNA molecule.  This gene codes for a protein that is also fairly long for proteins (~ 1,000 amino acids).  This protein then combines with three other identical proteins to form a large "tetramer" protein of approximately 4,000 amino acids (see illustration).4  The complexity of this lacZ gene would seem to be quite evident.  The level of size and apparent complexity of the ebg gene is similar.  

Appearances can be deceiving though.  Maybe lactose hydrolysis is not really as complicated as the size of this gene makes it appear? As it turns out, a BLAST search through the known protein databases quickly shows that the smallest known functional lactase enzyme in any creature is about 380 amino acid residues in size.  Some of these residues also seem to carry with them a fair degree of sequence specificity.  However, many of the residue positions can change, and some of them can change dramatically, without a significant loss of lactase function.  But, overall, the changeability of the lactase enzyme is at least moderately limited.  Some have suggested to me that there are around 10100 potential lactase enzymes in all of sequence space made up of chains of proteins containing 380 residues. Although 10100 does sound like an absolutely huge number (only 1080 total atoms in the entire universe), it is actually rather tiny when compared to the total size of sequence space (~10494 possible combinations for a string of just 380 amino acid residues).  With this ratio, for every 1 lactase enzyme, there are about 10394 non-lactase sequences.




A Game of Checkers


The potential of "sequence space" can be visualized as a gigantic checkerboard.  Each square on the checkerboard represents a different amino acid residue sequence.  Each member of a population can only occupy one square at a time (though any one square may be occupied by many individuals at any one time).  A limited population simply cannot cover all the potential squares on the checkerboard at any given moment of time.  With each mutation to an individual, it changes squares.  If any one individual comes across a beneficial sequence square, like a residues sequence with the lactase enzyme function while in a lactose rich environment, that individual and its offspring will tend to stay on that square because of the selective advantage given by that square in a lactose rich environment.  This advantage will be translated into increased population numbers that are on and immediately around that particular square of the checkerboard.  Eventually, in a steady state population, the entire population will be around that one square because they will all be descendants of the original individual that first came across that beneficial square on the checkerboard.  

The problem is that not every square is beneficial.  Depending upon the level of functional complexity in question, most squares are completely neutral for survival and many more are detrimental.  So, in traveling from one beneficial square to another beneficial square at the same level or even a higher enzymatic level of complexity, an open ocean of non-beneficial sequences will have to be crossed.  The problem with this open ocean is that during this voyage over neutral waters, natural selection cannot direct the process at all.  Nature is blind to such voyages and so the process becomes purely random.  In fact, this voyage has a mathematical name called "random walk" which does in fact occur in real life (i.e., Kimura's Neutral Theory of Evolution).

The interesting thing about random walk is that with each doubling of the distance length to the average beneficial sequence square on our checkerboard, the time involved increases exponentially.  For example, if the average random walk required for a particular colony of bacteria to achieve a particular level of complexity required 5 neutral steps or changes in DNA, the total number of options or potential spaces on our checkerboard between the starting point and a new "winning" square would be 4 (remember that there are four potential bases that can fill one given location in a string of DNA) to the power of 5, or 1,024 squares. So, the random walk would not simply take 5 steps in a straight line to the new beneficial square.  Not at all.  The random walk would wander randomly or blindly around 1,024 squares, taking far more steps, on average, to find the one square out of 1,024 that is actually selectable as "beneficial" to those that are searching the sequence space of that checkerboard. Depending on our population's size and mutation rate, we could estimate an average time required to reach all of these squares at least once beginning at a given starting point.  Obviously though, the bigger the population and the higher the mutation rate, the faster random walk could reach all of the squares.  

For instance, if we started out with a population of 1 trillion bacteria and if all of these bacteria started out on one square on our checkerboard, it would take around 65,000 generations (if each individual sequence of a given length was mutated at least once in each generation) for them to reach equilibrium over all the squares of the checkerboard - kind of like a tall column of sand being let loose at one location and then rapidly flowing and spreading itself in all directions to other locations.  At equilibrium (i.e., when the pile of sand becomes perfectly "flat" or evenly distributed over all surfaces), about 0.098% of the bacteria will be on each one of the 1,024 squares of the checkerboard.  Even though 0.098% does not really seem like a big number, it actually works out to be around 9.8 billion out of a population of 1 trillion.  In other words, after about 65,000 generations, there would be an average of 9.8 billion bacteria covering each one of the 1,024 squares on our checkerboard of potential space.

So obviously, a gap of 5 neutral mutations would not be a problem for a population of 1 trillion bacteria to cross in relatively short order.  But, what happens if we double the gap to 10?  A gap of 10 neutral mutations/steps would create a checkerboard with over 1 million squares of potential space (1,048,576 to be exact).  At equilibrium, our population of 1 trillion would have only 953,674 individuals on each of the squares instead of the 98 billion it had when the gap averaged only 5 steps wide.  Doubling the gap again to 20 steps makes our checkerboard grow a million fold to just over 1 trillion squares of potential space (1,099,511,627,776).  Now, our population of 1 trillion would average a bit less than one member of the population on any one square at any given point in time.  I think the trend is obvious by now, but just for kicks, doubling the gap again to 40 steps increases the size of our checkerboard a trillion fold to just over 1 trillion trillion squares.  Now, at equilibrium, each one of the members of our population of one trillion are surrounded, on average, by one trillion empty squares that they have to search out all by themselves.  




Average Time


Take a population of bacteria the size of all the bacteria that currently exist on the entire Earth - about 1e30 bacteria.  Let's say that this steady state population produces a new generation at a rate of 20 minutes and has a mutation rate of 1e-8 per codon position - given a genome per bacterium of 10 million codons.  How long would it take such a population to find a new beneficial function at the level of 1,000 fairly specified residues?

Well, first we have to calculate the likely gap size.  Using an average between the calculations of Yockey and Sauer, the ratio of potential beneficial vs. non-beneficial for 100aa systems is about 1e-40.13,14,15  This creates a ratio for a 1,000aa system of about 1e-40(1000/100) = 1e-400. So, the average gap size between potentially beneficial sequences at this level would be about 308 residue differences -  i.e., 20308 = 1e400. 

At his point, one can calculate the Poisson distribution curve to determine the odds that any particular gap would exist (given the average gap of 308 residue differences). Although extremely unlikely given the Poisson distribution, let's say that our colony has a few closer sequences that just aren't "average" - close enough to be only 50 specific residue changes away from at least one beneficial function at this level of minimum size and specificity.  How long would it take to get just 50 specific residue changes?

A gap of 50 specific residue differences from a given 1,000aa sequence means that each of these sequences is surrounded by 1e65 non-beneficial options.  But, we have 1e30 bacteria with 1e7 codons each.  For arguments sake, lets say that each bacterium has 1e5 sequences of 1,000 codons that are within 50 residue changes of success.  This gives us a total population of 1e35 starting points that are within 50 changes of success.

Now, how long will it take to get these 50 needed changes in at least one bacterium in our population?  After equilibrium of distribution randomly through sequence space is realized, each one of our starting point sequences must search through a sequence space of 1e65/1e35 = ~1e30 sequences, on average, before success will be realized.  With a mutation rate of 1e-8 per codon per generation our 1,000-codon sequence will get mutated once every 1e5 generations.  With a generation time of 20 minutes one mutational step takes about 2,000,000 minutes or approximately 4 years. So, with one random walk step every 4 years, it would take 1e30 * 4 = 4e30 years to achieve success - on average (i.e., trillions upon trillions of years).


Evolution Stalling Out


With each doubling of the likely neutral gap, the average time required for "success" increases exponentially.  Of course, one way to reduce the average required time is to increase the population's size exponentially.  This does help for a while, but very quickly the required size of the population becomes impractical for any environment to support and further evolution simply stalls out, in an exponential manner, with attempts to reach higher and higher levels of functional complexity.  Interestingly enough, Barry Hall discusses this very problem:


Given a gene of 1000 base pairs there are over 1034 sequences that differ from the wild-type sequences by 10 or fewer mutations. Not only can we not explore all of those possible variants, life itself has barely had sufficient time to explore all of those possibilities. The mass of the earth's oceans is about 1.4 x 1024g. Even if living cells constituted a 10-4 of the mass of the oceans, given about 1012 bacterial cells per gram, a reproduction rate of about 1 cell generation per day and a mutation rate of about 10-9 per cell generation and 4 billion years of life there has been sufficient time to explore only 1.6 x 1034 variants of a single 1000-bp sequence. However, evolution does not proceed by exploring all possible variants but by incorporating single mutations, selecting the fittest of those variants, expanding the population of the fittest variants, and incorporating additional single changes.

Hall goes on to explain:  If the evolved sequence differs from the wild-type at n sites there are n possible first-step amino acid replacement mutants. Each of those single mutants can be created by site directed mutagenesis and the effect on fitness determined by competition experiments. The best (fittest) of those amino acid replacements can be chosen and the n31 possible second-step mutants created, the fittest double mutant chosen, the n32 possible third-step mutations introduced, etc. The effect of this exercise is to mimic an evolutionary pathway in which the fittest single mutant is fixed into the population, that population expands, the fittest double mutant arises and is fixed into the population, etc. Orr has recently shown on theoretical grounds that adaptive evolution is expected to proceed in exactly this fashion in which the first mutations to be fixed are those that have the greatest positive effect. 9


Of course, all of Hall's single steps here are functionally beneficial. What happens if there are true gaps in function? What happens to evolution? Hall continues: 


If that sequence involves six amino acid replacements, we might find that after introducing three replacements, each of which further improves fitness, none of the three remaining replacements improves fitness. Assuming that the final six-mutant sequence is significantly fitter than the triple mutant, that result means that two, or perhaps even three, of the remaining substitutions must be introduced simultaneously to further improve fitness. This simultaneous occurrence of two or more specific mutations is obviously highly unlikely, but what about the possibility that one of the two mutations will arise, by selectively neutral, but be fixed into the population by drift? Were that to occur the second mutation would quickly be incorporated by selection. The probability that a newly arisen neutral mutation will be fixed into the population is the reciprocal of the population size. When populations are large enough that the probability of the occurrence of the mutation is very high, e.g., the population size approaches the reciprocal of the spontaneous mutation rate, then the probability of the fixation is very low. Although neutral variants arise constantly, it is very unlikely that the particular neutral variant we require will be fixed into the population. Thus, unless each of the mutations confers a selective advantage relative to its parent, it is unlikely that the final six-mutant sequence would evolve naturally. In the example above we would conclude that the evolutionary potential may well be limited to the triple mutant. 9


So, even Hall admits that very small neutral gaps of three mutations might be enough to "limit" further evolution. Thus, such functions that are isolated from other functions by neutral gaps might be quite difficult for a theory based on the mechanism of random mutation and function-based "natural" selection to explain.  



Humpty Dumpty Had a Great Fall


What is also most interesting is that the same bacteria that couldn't evolve a relatively simple single protein enzymatic function, like lactase (i.e., Hall's double mutant E. coli), would quickly evolve resistance to any modern antibiotic in short order via random mutation and natural selection.  Why then is it so easy to evolve antibiotic resistance but not a beneficial enzyme starting with what is currently available in the collective genome?  Well, perhaps it is because functions are not created equal.  Some functions are extremely simple while others are vastly more complex (i.e., in both the minimum length requirement of a coded sequence or number of required parts as well as the minimum degree of specified arrangement of the sequence or types of parts).  

De novo antibiotic resistance is one of the most simple functions around.  The reason for this can be found in the specificity of a given antibiotic for a particular target within a bacterium.   Because of this high specificity, there is a very high ratio of potentially "beneficial" changes compared to "non-beneficial" changes that can in fact interfere with such specific interactions of molecules.  The neutral gap involved in finding at least one of these very common potential interfering sequences is quite small indeed.  Perhaps, in some cases, a majority of mutations would result in interference with such a specific interaction.  Obviously then, when an interfering mutation does come along that blocks or completely destroys the interaction of an antibiotic with its specified target, the antibiotic resistance function is evolved.  It is much like the breaking of Humpty Dumpty when he fell off the wall.  He broke easily because there are so many ways he could be "broken".  However, it wasn't so easy to put him back together again because the vast majority of potential ways to arrange his parts just won't "work" - and that's the catch for the theory of evolution.  

Functions that are based on the destruction or interference with other pre-formed functions or interactions are universally very simple to evolve.  This is mathematically predictable and in real life it does in fact occur quite commonly.  Antibiotic resistance evolves very rapidly in real life even without access to such complex antibiotic enzymes like penicillinase (which has never been shown to evolve in real time by the way).  A few other types of drug resistance, such as chloroquine resistance, have taken a bit longer to evolve in real life (though never in laboratory conditions), but even this type of evolution works on the same basis of interference with a preformed function.

Chloroquine resistance (CQR) seems to require at least a few mutations (as many as 6, but perhaps only two mutations), before at least some beneficial level of resistance can be realized. It turns out that several of the mutations seen in CQR are selectively advantageous once the initial two or three are realized.  Resistance to both mefloquine and chloroquine is achieved via the blocking of or interference with a pre-established interaction of these drugs with specific target proteins. 

Again, the prediction that such interferences are relatively easy to achieve holds up in these cases as well. No new protein or enzyme is evolved in these cases. The only thing that happens is a disruption of a specific interaction that was pre-established.  It is just a different way of breaking Humpty Dumpty, but still relatively easy to do with large populations. This evidence fits very well with my predictions for the required time needed to evolve functions at this very low level of complexity (more of the details of chloroquine resistance in the reference section below).10



Climbing the Ladder of Functional Complexity


However, as one moves up the ladder of complexity to those functions that are not based on interference with pre-established functions (single protein enzymes like lactase, nylonase, penicillinase, etc), the relative number of such sequences is dramatically reduced.  This makes the evolution of such functions much more difficult and, in real life, there are far fewer examples at this level of functional complexity.   Certainly the evolution of lactase, nylonase, and several other single protein enzymes that have been demonstrated in real life are definite examples of higher-order evolution in action when compared to the extremely simple function of antibiotic resistance and other such functions.  However, there are far fewer examples at this level, relatively speaking, And, there are very interesting limitations at this level - as illustrated in Hall's experiments with lactase evolution in E. coli.   

Evidence for this position can be found in the fact that all bacteria can and do rapidly evolve antibiotic resistance to any antibiotic that is brought their way.  However, only a very few of them can evolve new enzymatic activities, such as a relatively simple lactase function provided by any one of literally trillions upon trillions of potential lactase enzymes dispersed through sequence space.  

Of course, there are even higher levels of complexity that involve multiple proteins all working together at the same time in order for a particular function to be realized.  A classic example of such a level of function can be found in bacterial motility systems.  All bacterial motility systems are dependent upon the simultaneous action of many different proteins all working together in harmony in specific orientation with each other.  A common example of such a motility system is the flagellar system of motility.  The flagellar system requires around 50 genes to construct and regulate the eubacterial flagellum and around around 18-20 fairly specified proteins (averaging around 300 residues each), to form the actual motor-switch-shaft-propeller complex.

The flagellum is in fact a biochemical machine that does very much resemble something a human would design.  There is the helical filament (propeller), the hook (universal joint), the rod (drive shaft), the S-P ring (bushing around the rod - in gram negative bacteria), the SMC ring complex, and the "motor" which includes the stator and the rotor. The entire assembly is hollow, including the actual filament. The rotor, hook and filament are made of different helical proteins that self assemble to form hollow, cylindrical structures (in the case of the filament, the cylinder is helical so that it acts as a "screw propeller" when it rotates.  Also, many eubacteria can switch the direction of rotation of the propeller (and hence the direction of travel) and the "switch" mechanism appears to be part of the motor complex (More detail is listed below in the reference section as well as in another essay of mine dealing specifically with the flagellum). 1, 11



Minimum Number and Specificity of Parts


Without a minimum number of parts being present in the proper "specified" order and orientation, the function of motility could not be realized, even a little bit. Many, like biologist Kenneth Miller, argue that such multi-part systems of function are made up of less complex sub-systems of function that have other functions within the cell - meaning that they are not truly "irreducibly complex".  The argument is made that several of the flagellar sub-structural proteins and even systems of protein parts have homologues in other independently functional cellular systems.  One well-known example is a secretory system called the "Type III protein secretion system" or "TTSS".  Interestingly enough, some of the parts used in the secretory systems of some species are nearly identical to some of the parts used in the flagellar motility system (More detail in reference section and flagellum paper).11 Of course, the object of such arguments is to suggest that as long as all the needed sub-parts are there, that a beneficial apparatus of higher functional complexity, like the flagellar motility system, will obviously self-assemble, eventually, when needed.  

There are several potential problems with this hypothesis however.  Perhaps the most obvious problem is the fact that no such demonstrations of the evolution of multiprotein systems of function have ever been observed or even theorized in a falsifiable way when they require more than a few hundred fairly specified amino acids working together at the same time (i.e., the multiprotein flagellar system of motility requires a minimum of several thousand specifically arranged amino acid residues, working together at the same time).  Lower levels of functional complexity, that involve interference with pre-established functions (antibiotic resistance) or that are based on single protein enzymes (lactase, nylonase, etc), have been shown to spontaneously evolve.  However, no function at a higher level of complexity, that involves multiple proteins totaling more than a few hundred fairly specified residues, has ever been shown to evolve in real time.  There isn't a single published example.  It is here that Behe is correct in saying that no such system of function has been or likely can be explained through evolutionary mechanisms of "random mutation combined with natural selection".  

Also, just because all the necessary parts are available, in close proximity, to form a potentially beneficial system does not mean that the parts will "know" how to spontaneously self-assemble to form such a beneficial system if each intermediate step is not also more "beneficial" than that which came before.  And, we know that the intermediate steps are not all beneficial when in comes to the functional systems of living things.  In fact, we know that the large majority of all potential changes to both functional as well as non-functional DNA are neutral at best and, if functional, are almost always detrimental.  This becomes more and more true at higher and higher levels of functional complexity due to the exponential growth of neutral gaps with each step up the ladder of functional complexity. 



All Functions are "Irreducibly Complex"


The fact is that all cellular functions are irreducibly complex in that all of them require a minimum number of parts in a particular order or orientation.  I go beyond what Behe proposes and make the suggestion that even single-protein enzymes are irreducibly complex.  A minimum number of parts in the form of amino acid residues are required for them to have their particular functions.  The lactase function cannot be realized in even the smallest degree with a string of only 5 or 10 or even 100 residues of any arrangement.  Also, not only is a minimum number of parts required for the lactase function to be realized, but the parts themselves, once they are available in the proper number, must be assembled in the proper order and three-dimensional orientation.  Brought together randomly, the residues, if left to themselves, do not know how to self-assemble themselves to form a much of anything as far as a functional system that even comes close to the level of complexity of a even a relatively simple function like a lactase enzyme.  And yet, their specified assembly and ultimate order is vital to function. 

Of course, such relatively simply systems, though truly irreducibly complex, have evolved.  This is because the sequence space at such relatively low levels of functional complexity is fairly dense.  It is fairly easy to come across new beneficial sequences  if the density of potentially beneficial sequences in sequence space is relatively high.  This density does in fact get higher and higher at lower and lower levels of functional complexity - in an exponential manner.  

It is much like moving between 3-letter words in the English language system.  Since the ratio of meaningful vs. meaningless 3-letter words in the English language is somewhere around 1:18, one can randomly find a new meaningful and even beneficial 3-letter word via single random letter changes/mutations in relatively short order.  This is not true for those ideas/functions/meanings that require more and more letters.  For example, the ratio of meaningful vs. meaningless 7-letter words and combinations of smaller words equaling 7-letters is far far lower at about 1 in 250,000.  It is therefore just a bit harder to evolve between 7-letter words, one mutation at a time, than it was to evolve between 3-letter words owing to the exponential decline in the ratio of meaningful vs. meaningless sequences.  

The same thing is true for the evolution of codes, information systems, and systems of function in living things as it is for non-living things (i.e., computer systems etc).  The parts of these codes and systems of function, if brought together randomly, simply do not have enough meaningful information to do much of anything. So, how are they brought together in living things to form such high level functional order?




Limited Evolutionary Potential


    And yet, despite these many problems, professors Hall and Miller and many other scientists like them would have us believe that the evolution of even more complex functions than single protein enzymes is still a relatively simple or at least a doable process given a few million or even billion years.  Such conclusions might be a bit premature to say the least since many of Hall's mutant E. coli seemed to have more than a little difficulty evolving just one relatively simple single-protein enzymatic function.  Hall himself described these strains as having "limited evolutionary potential." 3  Hall noted that with both the lacZ and the ebg genes missing, E. coli bacteria cannot evolve lactase ability at all despite his own efforts and those of several others, such as J. H. Campbell, to test for and observe such evolution over the course of many years (since 1973) totaling hundreds of thousands of bacterial generations.6  

         Hall did seem to realize somewhat of the implications of discovering that only one mutation was needed to "evolve" efficient lactase activity in lacZ negative E. coli strains.  In his paper he said, "The realization that a single mutation in ebgA [ebg = evolved b-galactosidase gene] was sufficient to convert ebg0 enzyme into an efficient lactase was therefore disappointing." 3  The problem, as Hall himself pointed out, is that there are mutations that do not yield changes in protein function toward anything useful to the cell.  The proteins that result from these mutations might in fact be useful to another organism somewhere in the universe, but for the particular organism that they have evolved in, they are either neutral in function or nonfunctional . . . or, even worse, detrimental in function.  

         No cell or organism or even an entire gene pool has an infinite vocabulary.  All living things have limited individual vocabularies.  Out of the huge number of possibilities for different kinds of proteins of a given length, any one individual cell or gene pool of cells "recognizes" or can use only a small fraction of them in a beneficial way (and this fraction gets exponentially smaller as the level of complexity increases).  Therefore, some functions are in fact out of statistical reach for that particular cell, or gene pool of cells, as well as their offspring because they do not recognize, as beneficial, any change in the functions of intermediary proteins along the way toward those sequences that would in fact be beneficial.  Such neutral evolution looses the guidance of natural selection as a driving force.   Hall describes such evolutionarily-challenged bacterial strains as having "limited evolutionary potential."  I propose that every living creature has very limited evolutionary potential.  



Bridges Between Old and New Functions


        Some might argue that some of the changes described by Hall did in fact cross bridges of non-function.  This is true.  Hall described the crossing a nonfunctional bridge that was two mutations wide.  Hall found even this challenge statistically unlikely, but it did in fact happen experimentally.  The statistical problem Hall describes is that for each genetic change in function, a change in regulator function is also needed.  A regulator is needed to control the production of the regulated gene.  Without regulation, genetic products are not advantageous and will be selected against in later generations.  The needed lactase regulator change also required, in this case, a single point mutation that was in line with the single lactase gene point mutation.  Independently, the statistical odds of the needed genetic mutation happening in a given bacterium, according to Hall, was 2 x 10-9 and the statistical odds for the needed mutation in the regulator region of that gene is 1 x 10-8 (at best).  The generation time for Hall's bacterial cultures averaged 6 hours and the average number of bacteria that were being studied at any given steady state was at best 1010 cells.  According to Hall's own statistical calculations, the average time required for both of these needed mutations to occur in any one of his bacterial colonies was on the order of 100,000 years. 

         It seems though that Hall's math was a bit off.  According to the statistics of random walk in a population of 10 billion, a gap of two mutations (only 16 squares to cover on our checkerboard) would be crossed in short order - and it was crossed in short order.  According to Hall, colonies containing both of these mutations were isolated in as little as nine days.  However, because of Hall's calculations (based on the requirement of full fixation of the first "correct" mutation in the colony before a gain of the second "correct" mutation) and estimates of much more time to achieve such a crossing, Hall concluded that, "under some conditions spontaneous mutations are not independent events."  He went on to say that this is "heresy, I am aware." 3

        Because of his statistical calculations, Hall was forced to conclude that mutations are not always random events but that sometimes point mutations occur in tandem at a higher rate than random chance alone would allow.   Of course there was no logical explanation for this assumption, and yet Hall assumed that was is in fact what was happening.  He felt forced to conclude that nature seems to know what it wants ahead of time and manipulates mutations without the aid of any functional advantage.  Truly, this is scientific heresy as Hall indicates.  It is basically a statement of magic.  Of course, Hall was wrong.  Random walk in such a large bacterial population can easily cross a gap of 2 or 3 or several more mutations in very short order, but what is there that explains the existence of gaps that average hundreds, thousands, or even millions of mutations wide?  Magic? - or intelligent design?



Statistical Road Kill


     Miller and Hall have failed to defeat Behe's argument of irreducible complexity for one simple reason that Behe describes as statistical "road kill."  As previously described, the road-kill problem is the problem of gaps - gaps of neutral function.  Single point mutations quickly come to gaps of neutral or even non-function that require multiple mutations to cross.  At this point, evolution is stuck.  And yet, we know that these gaps have in fact been crossed somehow, but how?  Obviously something or someone has crossed them at some point, because genes do in fact exist on the other side of these gaps.  The gaps themselves exclude natural selection as a force that can fly evolution over to the other side since natural selection is dependent upon the detection of functional phenotypic change.  Without natural selection, the crossing of these gaps via any naturalistic process remains a mystery.  Random mutation, by itself, tends toward homogeny, not the increase of meaningful genetic information.  Therefore, without guidance, random chance fails as a creative force of new high-level systems of function because random chance eventually creates homogeny (i.e., goop).

        However, if it is still difficult to see that neutral genetic gaps significantly limit the powers of Darwinian evolution, consider the Shakespearean phrase, "Methinks it is like a weasel" used by Richard Dawkins to illustrate the power of natural selection.12  Try and change any one letter or space and still have the sentence remain meaningful as well as beneficial in a given situation/environment.  You might change it to read, "Me thinks it is like a weasel" then, "He thinks it is like a weasel."  It still makes sense and it means something different.  Aha!  Evolution in action.  Keep going though.  How far can you evolve this sentence where each and every character change remains meaningful much less beneficial?  You might mutate it to, "She thinks it is like a weasel" then, "She thinks it is like a teasel" then, "She thinks it is like a tease" then, "He thinks it is like a tease" then… well… you see, it is getting quite difficult to keep evolving unique as well as meaningful phrases one mutation at a time. We run into evolutionary dead ends really fast.   

         The same problem happens with genetic "evolution."  The genetic blueprints of living things are written according to very specific rules of "grammar."  The order or sequencing of genes is very important to function.  Not just any order will do.  The "spelling" of genetic words also matters.  The cell will not recognize just any spelling as beneficial in a given environment.  So, to go from one functional genetic phrase to another uniquely functional genetic phrase in a higher level of functional complexity might be a bit of a problem if such changes require the crossing of even a few neutral or detrimental steps of random walk.

         Each mutation that does not cause a beneficial change in function (i.e., neutral or detrimental) is one lane in the statistical highway that Behe describes.  The blind turtle of evolution must cross this highway to reach the new beneficial function.  With each additional lane added to the highway, the average time needed for the blind turtle to make it across increases exponentially.  Each lane that is added skyrockets the average needed time for success until 4 or 5 billion years pales into the distance of trillions upon trillions of years. 



Given Enough Time Anything is Possible


         There are those who say that evolution is improbable, but that time makes the improbable - probable.  Time becomes the savior of evolution.  This might be true except for one small problem. Evolution needs more time than the history of the universe, or even millions upon billions upon trillions of universes have to offer - on average.  How high do the odds have to go before we suspect that all this just isn't the result of any non-deliberate process?  How many times would the same person be able to win the California Lottery in a row before one could reasonably suspect the possibility of deliberate cheating?



  1. Behe, Michael J. Darwin's Black Box, The Free Press, 1996.

  2. Miller, Kenneth R., Finding Darwin’s God, HarperCollins Publishers, 1999.

  3. B.G. Hall, Evolution on a Petri Dish.  The Evolved Beta-Galactosidase System as a Model for Studying Acquisitive Evolution in the Laboratory, Evolutionary Biology, 15(1982): 85-150.

  4.  Lewin, Benjamin, Genes V, Oxford University Press, 1994.

  5. Levinson, Warren E. et al., Medical Microbiology and Immunology 3rd Ed., Appleton & Lange, 1994.

  6. Campbel, J.H., Lengyel, J., and Langridge, J., 1973, Evolution of a second gene for B-galactosidase in Escherichia coli, proc. Natl. Acad. Sci. USA 70:1841-1845.

  7. Behe, Michael J., Irreducible Complexity and the Evolutionary Literature: Response to Critics, Discovery Institute, July 2001. (

  8. Behe, Michael J., "A True Acid Test" - Response to Kenneth Miller, Discovery Institute, May 2002. (

  9. Hall, Barry G., Toward and Understanding of Evolutionary Potential, FEMS Microbiology Letters 178 (1999) 1-6, June 1999. (

  10. Jane MR Carlton, David A Fidock, Abdoulaye Djimdé,Christopher V Plowe and Thomas E Wellems, Conservation of a novel vacuolar transporter in Plasmodium species and its central role in chloroquine resistance of P. falciparum, Current Opinion in Microbiology, 2001. (


    In human red blood cells, Plasmodium falciparum (the malaria parasite) supports its growth by taking up host cell cytoplasm in an acidic digestive food vacuole. Toxic heme, in its hematin form, is released in the vacuole by hemoglobin digestion and is crystallized into innocuous hemozoin, or "malaria pigment". Chloroquine (CQ) interferes with this process by complexing with hemozoin. This complex prevents hematin from crystallizing into the innocuous hemozoin form. The "toxic" effects of free hematin are caused by hematin's ability to increase membrane permeability which lead to cell lysis and death. Hematin also is known to inhibit parasite enzymes. 

    Chloroquine resistant strains of P. falciparum show a reduced accumulation of CQ in the digestive vacuole. The genetic mutations associated with this reduced accumulation have been isolated to the PfCRT (P. falciparum chloroquine resistance transporter) gene. The gene contains 13 exons that cover 3.1 kb. The PfCRT gene product is a 423 amino acyl ten-transmembrane channel or transporter protein that catalyzes chloroquine flux and H+ equilibrium across the digestive vacuole membrane. Many different point mutations have been isolated in resistant CQR strains of malaria (M74I, N75E, K76T, A220S, Q271E, N326S, I356T, and R371I). Of these, only the K76T and the A220S mutations are shared in common between resistant malaria strains on various affected continents of Asia, Africa, and S. America. The K76T mutation in particular seems to be the most important marker of CQR. What is interesting is that the K76T mutation is never seen by itself, but is always associated with a few other point mutations. However, in some CQ resistance strains the K76T mutation is absent. One such strain is the "106/1" strain that has the K76I mutation instead. This strain has six of the other point mutations, but is has the K76I instead of the K76T mutation at position 76. Even without the K76T mutation the 106/1 strain does have a fairly high level of CQR. However, the level of resistance is not as high as those strains that do have the K76T mutation. Interestingly enough, Fidock et al., performed an experiment with the 106/1 strain where stepwise CQ pressure was added to the population. The result was a fairly rapid change at position 76 from the K76I to the more resistant K76T mutation.

    The results of such observations suggest that the K76T mutation is not selectively advantageous by itself. the A220S may fulfill a particular requirement in the development of CQR since this mutation has consistently been found to accompany the K76T mutation in CQR parasites from the different New and Old World foci. "The suggestion that K76T cannot occur in the absence of other PfCRT point mutations may also explain the slow genesis of CQ resistance in the field as well as the difficulties that have been experienced with attempts to select CQ resistance in the laboratory."11

  11. Ian Musgrave, Evolution of the Bacterial Flagella, Personal Website, created in 2000. (

  12. Dawkins, Richard. The Blind Watchmaker, 1987.

  13. Yockey, H.P., J Theor Biol, p. 91, 1981

  14. Yockey, H.P., "Information Theory and Molecular Biology", Cambridge University Press, 1992

  15. Sauer, R.T. , James U Bowie, John F.R. Olson, and Wendall A. Lim, 1989, 'Proceedings of the National Academy of Science's USA 86, 2152-2156. and 1990, March 16, Science, 247; and, Olson and R.T. Sauer, 'Proteins: Structure, Function and Genetics', 7:306 - 316, 1990.






The Flagellum:


Propeller: The Filament (propellor) is composed of the proteins FlaA and FlaB.  Deletion of FlaB doesn't seem to do hinder motility too much and deletion of FlaA results in trucated flagella which still produce some motility, and have some motility.


Hook/universal joint: This is formed by FlgE proteins and possible a few others.  There is very limited sequence similarity between the FlgE's of Salmonella sp. and Heliobacter sp. (33% identity).  


Drive Shaft: The rod (driveshaft) is composed of a complex of the proteins FlgG, FlgB, FlgC, FlgF and FliJ, P, Q, R. The M ring is formed from FliF proteins.  The L and P rings are composed of proteins FlgH, FlgI, but these can be absent or present without significant effects on flagella function.


Motor: The motor consists of a rotor (the part that spins) and the stator (the part that does the spinning). The motor is largely contained by the C ring motor complex. This is formed from FliG, FliM, FliN proteins which form the rotor/switch apparatus. The stator is formed from either MotA and B in some species, and PomA, PomB, MotX and MotY in others .


Stator: Mot A and B forms a proton pump which provides the power of the motor, MotB also serves to anchor the motor to the cell. Deletion of Mot A or B paralyses the cell. FliG and FliM proteins are also involved.


In Rhodobacter sphaeroides, the genes equivalent to MotA and MotB have only 19% sequence identity. The Rhodobacter motor doesn't switch as does the other motors, but turns on and off, and re-orientation is via Brownian motion. 


In Vibro species, the motor is a sodium, rather than proton pump, composed of pomA, pomB, MotX and MotY.  The MotX and Y proteins are unrelated to MotA or B, but PomA seems to be related to MotA from R. spheroides, and R. spheroides.  MotA can partially restore swimming in PomA paralyzed mutants.


The rotor: The FliG protein is involved in torque generation.  It turns the proton gradient into rotational motion in a poorly understood way. It may also have a role in switching. FliM is also involved in switching, but probably not in torque generation. FliN is probably not involved directly in either torque generation or switching, and may be a stabilizing protein. In Bacillus species it is replaced by the protein FliY, which resembles a fusion between FliM and FliN.


Homologues:  It is true that homology studies have shown that many of the flagellar proteins are related to parts of the type III protein secretion system.  Some of these parts are even identical in some bacterial species. FliN is homologous to the Spa0 and HrcQ protein export proteins in the Salmonella and Pseudomonas respectively. FliP,Q,R and F proteins are homologous to HrcR,S,T and J proteins of the Pseudomonas Hrp type III secretory system, and indeed have homologues in most type III secretory systems. 

The type III secretory system forms a "rivet" structure identical to the rod and SMC ring complex of the flagellum.  Furthermore, the switching/torque generation system (FliG,FliN/Y), has homologues in virtually every type III secretory system examined so far.  Proteins exported by this system are shunted through the hollow SMC ring and through the rod to the outside of the cell.  In flagellum assembly, flagellins and hook proteins are shunted to the outside of the cell via the rod and ring complex. The proteins attach to the outer rim of the rod and self assemble into a tubular structure that will become the hook and filament.  The flagellar proteins then pass through this tube as it grows. However, there is no apparent homologue of the motor (MotAB) in type III secretory systems.  Several of the type III secretory systems have tubular structures attached to the rod. The Hrp secretion system forms basal ring/rod system with a pilus that strongly resembles the flagellar system. It is not clear if the Hrp pilus has any relation to the flagellar filament. However, E. coli has a filamentous structure attached to one of its type III secretory systems which has significant similarity to the flagellar filament.



. Home Page                                                                           . Truth, the Scientific Method, and Evolution   

. Methinks it is Like a Weasel                                                 . The Cat and the Hat - The Evolution of Code   

. Maquiziliducks - The Language of Evolution             . Defining Evolution    

. The God of the Gaps                                                           . Rube Goldberg Machines  

. Evolving the Irreducible                                                     . Gregor Mendel  

. Natural Selection                                                                  . Computer Evolution  

. The Chicken or the Egg                                                         . Antibiotic Resistance  

. The Immune System                                                            . Pseudogenes  

. Genetic Phylogeny                                                                . Fossils and DNA  

. DNA Mutation Rates                                                            . Donkeys, Horses, Mules and Evolution  

. The Fossil Record                                                                . The Geologic Column  

.  Early Man                                                                                . The Human Eye  

. Carbon 14 and Tree Ring Dating                                     . Radiometric Dating  

 . Amino Acid Racemization Dating                   . The Steppingstone Problem

.  Quotes from Scientists                                                           . Ancient Ice

 . Meaningful Information                                                          . The Flagellum

 . Harlen Bretz                                   . Milankovitch Cycles




Search this site or the web powered by FreeFind

Site search Web search








Since June 1, 2002





Locations of visitors to this page




 Free Web Stats in real-time !