Keith Robison has a review of Michael Behe's book among the FAQs for
talk.origins. After reading it, I couldn't resist offering my own
review
of Keith's review. Here is Part 3 of my reply.
>What Good is An Antibody?
>Behe spends a whole chapter (Chapter 6), describing portions of the
>immune system. After going through a section describing the
>production of antibodies, which he claims is an "irreducibly complex"
>process and therefore could not be the product of evolution, he makes
>a very interesting statement. p.131-132:
>Antibodies are like toy darts: they harm no one. Like a "Condemned"
>sign posted on an old house or an orange "X" painted on a tree to be
>removed, antibodies are only signals to other systems to destroy the
>marked object.
>Behe then goes on to describe the complement system. And nothing
>else. Leaving us with the impression that antibodies and complement,
>each supposedly "irreducibly complex," require each other for
>function.
>The biological reality is quite different. Antibodies can function on
>their own. They can function in the absence of complement.
Keith then proceeds to explain how antibodies serve an immune
purpose without employing complement. This included the
neutralization of toxins, agglutination, and opsonization. And Keith is
correct. Behe is essentially correct if we restrict ourselves to
cellular
organisms, but Keith seems to be saying that the immune system
evolved first by producing antibodies and then complement evolved to
expand the reach of those antibodies. If this is true, then Behe is
mistaken in thinking of antibody-mediated immunity as an example of
IC. However, I would be curious as to whether there are any examples
of organisms that make vertebrate-like antibodies yet do not possess a
complement system?
But let's proceed to what I consider to be the more interesting problem.
>What of the antibodies themselves? How did they originate?
>It is, I believe, still an unsolved question. But again, Behe has left
>out all the interesting hints. And there are certainly interesting
>ones out there, hints which undermine Behe's claims.
Let's check out these interesting hints.
>For example, Behe makes a number of claims about the system which
>produces antibodies in mammals and how these systems are
>irreproducibly complex (Chapter 6; especially p.130-131). Among the
>key claims are:
>1. The system is tuned to produce hundreds of thousands of distinct
>antibodies, each with a distinct specificity. A system which
>produced only a few (or just 1) would be useless.
>2. The system could not function without the complex biochemical
>machinery which splices together antibody genes. In turn, this
>complex machine has no function in the absence of antibody genes.
>In the November 1996 issue of Scientific American there are two
>papers on the evolution of the immune system. Interestingly, these
>papers provide evidence which refutes both of Behe's claims.
"Refute" is a awfully strong word, don't y'think?
>First, some moths produce a protein called hemolin, which binds (in an
>apparently generic manner) to bacteria and assists in removing the
>bacteria from the hemolymph (the blood-like substance in insects). The
>protein sequence of hemolin reveals it to be a relative of vertebrate
>antibodies. So much for the hypothesis that a single antibody is not
>useful.
To "refute" Behe, Keith ignores Behe's definitions and instead relies on
equivocation. An antibody is defined as a protein that interacts with a
particular site (epitope) on an antigen and then facilitates clearance
of
that antigen by various mechanisms. Behe clearly defined it as such
when he said "each with a distinct specificity." Since hemolin is a
generic-binder, it is not an antibody. It may be roughly analogous but
that doesn't count. Keith finds an orange and thinks it refutes claims
about Behe's apples.
If we properly define "antibody," Behe is correct. A system which
produced only one antibody (one protein that recognized one epitope),
this is essentially useless given that any organism is likely to
encounter
millions of foreign epitopes. And that's going to make it difficult
imagining the evolution of antibodies.
Before getting to this in more detail, there are a couple of things
worth
noting about Keith's response.
First, it suffers from what I call "phylogenetic-sloppiness." Keith
claims
that the sequence "of hemolin reveals it to be a relative of vertebrate
antibodies." A more accurate way of putting this would be "the
sequence of hemolin is consistent with an interpretation of divergent
relationship." Of course, it could be consistent likewise with a
convergent relationship. But since I don't have the sequence, I can't
tell. But the reason I think Keith's explanation to be
"phylogenetically
sloppy" is because moths are very distantly related to vertebrates. A
more convincing candidate would be a lancelet. And I point this out
simply because what's going on in moths may have nothing to do with
vertebrate evolution.
This reminds me of the classic debate about the evolution of the eye.
We are often told that all those eye-spots in flatworms are somehow
relevant to the question. But are they? Lancelots don't have eyes or
eye-spots! Eyes did not evolve in one smooth continuum from flatworm
to vertebrate.
Anyway, let's stick to the topic. The second thing worth noting about
Keith's example is that it poses an interesting dilemma. When did
hemolin evolve? If it appeared hundreds of millions of years ago, why
in the world have bacteria not evolved ways to evade this dumb
protein and render it essentially useless?
Enough of the small matters. Let's get back to the evolution of
antibodies. How did all these specific-binders evolve? As Behe notes,
a
genome with one (or even a dozen) specific binders is likely to be no
use. So perhaps we would have to start with something like this generic
binder. Then undoubtedly, the magic wand of "gene duplication" would
be invoked so that now the organism would have two generic binders.
Then, some lucky mutations start experimenting on one copy and make
it more specific. Well, there are two problems at this point. First,
the
specific-binder would have to recognize something not seen by the
generic binder. But if the generic binder is worth its selective-self,
how
likely is this? But if the specific binder and generic binder see the
same
thing, there is no selection at work shaping the sequence-specific
binding of the wanna-be-specific binder. The other problem is that this
such an
evolutionary process is unlikely to be effective. If we were to compare
the rate of gene duplication, mutation, and the fixation of these
changes
in a population of multicellular critters against the rate at which
bacteria evolve resistance, the multicellular critters are just too
slow.
That is, even if a specific binder appears and imparts immunity to a
strain/species of particular bacteria, it is likely those bacteria would
evade the specific binder through mutation and resistance before the
second specific binder could be added to the genome.
As if this wasn't bad enough, let's imagine our organism evolves a
rather large family of antibody genes. This would represent millions of
years of selective evolution. Another duplication/deletion event
occurs.
That is, one branch loses a number of antibody genes. But let's say
it's
no problem because those specific antibodies recognize specific
pathogens who are no longer are in contact with the population. Thus,
the deletion spreads through the population. Then, the population
becomes infected by a pathogen they used to be immune to. But since
they lost the antibody genes, they succumb. All that selective
evolution
over those millions of years gets flushed because of one bad cross-over.
:(
Then there's one final problem. What happened to the generic binder?
I suppose it could be lost in a deletion that spreads through the
population, but that would be disadvantageous unless the sum total of
specific binders recognize the same number of epitopes as the generic
binder. So how many epitopes does this generic binder recognize? If
they are few, this makes it more dispensable, but then it also becomes
hard to explain the selective pressure behind the generic binder's
existence given the myriad of pathogenic surfaces the organism will see
in a lifetime.
If the molecule is very plastic and recognizes a huge number of
epitopes, it's hard to imagine the selective pressure behind the
evolution of so many specific binders. To select so many specific
binders, you need proteins that see what the generic binder doesn't see.
But the more specific binders you create that bind to things not seen by
your generic binder, the more your generic binder ends up being truly a
"specific" binder! And then when we're back to wondering about the
selective advantage behind this "generic" binder?
In other words, it's hard to see how you mix, in an evolutionary sense,
a
generic-binding system with a specific-binding system. The generic
system might need only one gene, but then it makes it hard to select for
specific binders. And the specific system needs *much* more than one
gene to be selectively advantageous, meaning you can't start with one
specific binder.
Thus, when Keith appeals to hemolin, he may be actually strengthening
Behe's point. Hemolin would demonstrate an unrelated, alternative
immune strategy - that of generic binding. Vertebrates demonstrate
the opposite strategy - that of specific binding. But what we don't
find
is a generic binder working alongside a whole class of specific binders.
>The second article describes the antibody genes of sharks. In humans,
>an antibody gene is assembled by mixing-and-matching various DNA
>segments, which are all found lined up on a chromosome. In individual
>immune system cells, antibody genes are assembled according to the
>following schematic: Genome: V1-V2-V3-V4-D1-D2-D3-J1-J2-J3-C
>Possible Antibodies: V1-D3-J1-C C is a region Constant between
>antibodies, whereas the V, D, and J segments (Variable, Diversity, and
>Joining) are each drawn from large pools of segments.
Keith forgot the L exon which is in front of the V segment. The L exon
encodes a signal peptide for transport into the ER. Of course, all this
means that the antibody itself may be a candidate for IC. Since it is
composed of V, D, and J segments, we may ask if antibodies lacking any
one of these segments is useful. And to this we could add the interplay
between the H and L chains.
>In sharks, however, the arrangement in the genome is: V5-D5-J5-C
>V6-D6-J6-C V7-D7-J7-C
>(Note: the numbers are just used as markers; they don't signify
>anything else).
>That is, in sharks the genes are already assembled, and these genes
>are arranged in long strings of such assembled genes (tandem arrays).
>Molecular genetic processes which can generate such tandem arrays
>are well-known.
This is very interesting. And since Keith likes to cite places where
evolution makes predictions, I should point out that this arrangement
was not predicted. In 1989, Christopher Wills (University of
California)
wrote a book entitled, "The Wisdom of the Genes: New Pathways in
Evolution." In talking about the evolution of antibodies (he has a
whole
chapter), Wills states:
"A simple but wasteful way to generate all the millions of antibodies in
our final line of defense would have be to have one gene in our genome
for each type of antibody. But it would be much less wasteful to carry
a
fairly small number of genes that can be mixed and matched to make
the complete antibody. We now know that the evolution of the immune
system has taken the second route. There are a couple of excellent
reasons why this is so."
Wills then goes through the reasons and his case is pretty convincing.
So then we are left with the enigma of the shark genome. Evolution has
a hard time explaining it, as Wills noted in 1989, for there are severe
"difficulties evolving a "one gene, one antibody" system."
>Also note that these same processes could take the shark
>arrangement and generate: V5-V6-D6-J6-C V7-D7-J7-C by a single step
>--which looks a little like the human case. Further such deletions,
>especially working in conjunction with re-expansions by tandem
>duplication, could easily generate the arrangement seen in humans. Of
>course, this arrangement is not useful without the splicing machinery,
>but the shark example clearly disproves Behe's claim that antibody
>diversity requires the splicing machinery.
I suspect Behe had in mind the mammalian system that is so-well
studied. The shark example would only disprove Behe *if* you could
show the shark arrangement was indeed the ancestral state to the
mammalian state.
And boy, are there problems if that is Keith's position. Let's consider
his cross-over state. Keith says "it looks a little like the human
case."
Er, yeah. And those rocks on Mars look a little like a human face.
Keith
claims this state is useless without the splicing machinery and indeed
it
is. But it is actually likely to be disadvantageous. When the V5
segment is fused to the V6 segment, you have a 2/3's chance of a
frameshift. And even if there is no frameshift, what would be
transcribed is a V5-V6-J6-C antibody. And that is probably a useless
antibody. After all, if you could join different V segments to get a
working antibody, you would expect this to have been incorporated as a
mechanism to generate diversity. But this strategy is not used by
mammals. In fact, are there any antibody genes in the shark genome
that *do* show this arrangement?
Finally, Keith makes it seem as if it was *easy* to segregate all the V
segments from the J segments from the C segments. And all this
happened while a specific splicing machinery evolved. Really? That's a
whole lot of specific crossin' over y'got goin' on there. And what
happens
when you first separate V5 from J5 in the same genome? You lose your
antibody UNLESS the splicing machinery is in place. So you better
account for that splicing machinery, which is a story in-of-itself.
[snip]
>What features of the mammalian immune system are missing in other
>vertebrates? We see already that the shark immune system is laid out
>on a similar, but distinctly different plan. What sort of immune
>system do non-vertebrate chordates possess? I don't know offhand,
>but it probably tells us something interesting about immune system
>function. Any missing parts would certainly undermine claims of
>"irreducible complexity".
Let me get this straight? Because sharks don't need splicing machinery,
it's not really part of an IC system in mammals?? Don't buy it. Simply
finding a simpler system someplace else doesn't mean anything unless
you can show that the simpler system did indeed give rise to the more
complex system.