Is Carl Woese losing a Kingdom?

Arlin Stoltzfus arlin at
Sun Sep 29 20:49:13 EST 1996

Dr. Radhey S. Gupta wrote:

>      The two main tenets of the current "three domain
> phylogenetic view" are:
> (i) Archaebacteria (or Archaea) constitutes a monophyletic domain
> which is completely distinct from the rest of the bacteria.

where 'monophyletic' = holophyletic.  There is some limited
evidence favoring paraphyly of the archaebacteria.

> (ii) Eukaryotic nuclear genome has directly descended from an
> Archaebacterial ancestor.

A common misunderstanding.  This is like saying that humans
evolved from chimpanzees.  Instead, the principle of common
ancestry suggests only that humans and chimpanzees evolved
from a common ancestor.  Only if archaebacteria are paraphyletic
would one be correct to say that the nuclear genome descended
from an archaebacterial ancestor.  And this is not the 
'conventional' view.

>      The question may be asked as to why should the chimeric
> model be preferred over direct descendance of eukaryotic cells
> from archaebacteria? Well, there are several reasons for doing
> so:

> (b)  It readily explains why certain characteristics of
>      eukaryotic cells (e.g. components of transcription and
>      translation machinery) are similar to archaebacteria,  while
>      others are clearly derived from eubacteria (e.g. ester-
>      linked straight chain membrane lipids, fatty acid
>      synthetase)

This pattern does favor a chimaeric model over any other model.
This is another misconception, one that dates back to the papers
of Zillig, et al.  Suppose that taxa A, B, and C have one of two 
states (x or y) for features 1-8:

	taxon: 	A	B	C
character #:
	1	x	x 	x
	2	x	x	y
	3	x	y	x
	4	x	y	y
	5	y	x	x
	6	y	x	y
	7	y	y	x
	8	y	y	y

Notice that these are all possible configurations for  
characters with two states.  In #2-#7, there are differences
between the taxa.  For instance, #2 could be x = has operons,
y = lacks operons, etc.  HOWEVER, THERE IS NO POSSIBLE TREE
THAT IS CONTRADICTED BY THESE DATA.  We could have the rooted tree
(A, (B, C)), or (B, (C, A)) or (C, (A, B)).  There is
absolutely *NO INCONGRUENCE* in the data that requires a
fusion hypothesis in order to be explained, because one simply 
*CANNOT HAVE* incongruence under these conditions.  For
instance, if A and B have operons and C lacks them, then, 
regardless of the tree, this could be the ancestral state, 
operons having been lost in taxon C [and for the tree (C, (A,B))
it could also be a derived state].

In order to have incongruence with non-polarized characters, 
one has to have four taxa.  The minimal display of incongruence 
would be this:

	taxon: 	A	B	C	D
character #:
	1	x	y 	x	y
	2	x	x	y	y

> (d)  It provides a plausible explanation for the enormous
>      structural difference between the eukaryotic and prokaryotic
>      cell types, and the absence of any species that are
>      intermediates in this transition. These observations cannot
>      be readily explained by simple evolutionary mechanisms.  By
>      contrast, a sudden and major evolutionary transition could
>      readily be explained by fusion and subsequent gene
>      assortment of two very different species.

No one ever said it was simple to explain the origin of eukaryotes.  
Please explain to us why fusion suddenly make complex constructive
evolution more likely than it otherwise would be?  As far as I 
can see, this is an utter _non sequitur_.

> (e)  The inferred time of divergence of eukaryotic species from
>      archaebacteria and the eubacteria (about 2 By ago) based on
>      genetic distances between different proteins sequences
>      (Doolittle et al , Science, 271, 470-477, 1996) and the
>      fossil records, can also be satisfactorily explained by the
>      chimeric model.

This in no way distinguishes a 'fusion' model from any other model.
Please give us an example of a model that is not consistent with
Doolittle's estimation, given its great uncertainty (e.g., if
the wrong saturation assumption was used, the date of divergence might
be 3-4 BYA instead of 2 BYA).  Besides, Doolittle based his 
calculation on an assumption that is not shared in your model,
namely the 'conventional' (eubacteria (archaebacteria, eukaryotes))
view of the tree of life, with all of those proteins being
representative of the same (non-chimaeric) tree!  This is a 
another _non sequitur_-- there is nothing in the 'model' you
have described that demands the occurence of a fusion event at 
a particular point in geologic time.  If true, the event could 
have happened 2 bya or 3 bya or 4 bya.  If you wish to postulate
a date for the event based on Doolittle's estimates, fine, it
doesn't really constrain the theory (because of the great uncertainty
in the estimates).  

> (f)  The phylogenies and signature sequences in various
>      eukaryotic genes also provide compelling evidence that all
>      extant eukaryotic species are monophyletic. This in turn

Monophyly of eukaryotes does not distinguish the fusion model 
from the 'conventional' view or any other model of which I am
aware.  Please give us an example of an alternative model
in which extant eukaryotes do not all derive from a eukarytoic
common ancestor.  Also, I question the basic logic of this
'implication' and its relevance to testing.  You argue that somehow 
fusion is a more likely explanation for the origin of eukaryotes,
because eukaryotes only arose once, and because fusion is rare.  
The logic of this argument is a bit like saying the Grand Canyon 
must be the landing strip for a Martian space ship, because there 
is only one Grand Canyon, therefore it must have a qualitatively
different cause from other canyons.  And why propose that a 
mutational event occurred only once?  If we never find a real 
biological example of it, this would hardly be evidence for 
point (f), since it would not allow us to distinguish a frequency of
occurrence of zero (i.e., it never happened) from a frequency of
once-only-in-the-history-of-the-universe.  On the other hand, 
if we *do* find real-life cases of the sort of 'fusion' that 
you are proposing, this would contradict point (f), which is 
absurd.  In fact, if these 'fusions' occur, we should find them.
A germaine result from population genetics is that most newly-arising 
favorable alleles are quickly lost due to random genetic drift.  
The chance that a given instance will succeed is on the order
of the selection coefficient (e.g., s = 0.005, then a 5/1000th
chance of making it; NB- the exact formula may differ, please
correct me if so!).  What this means is that, to the extent that 
we can categorize mutations into classes, a mutation representing
a particular class must generally happen many times before the
class of allele thus created becomes fixed.  

Arlin Stoltzfus
Department of Biochemistry
Dalhousie University
Halifax, Nova Scotia B3H 4H7 CANADA
(email) arlin at 
(phone) 902-494-3569 
(fax) 902-494-1355

More information about the Mol-evol mailing list