AFLP reliable for study of evolution?

[david baum]

I worry that Lai-Lai and others may think I do extensive AFLP work.  While
a LITTLE is happening in the lab I should caution that my comments are
based on "theoretical" knowledge.

Lai-Lai Teo wrote:
>I understand that some AFLP marker are dominant, whereas others are
>co-dominant.  The easiest way is to score the presence-absence of the band
>(presuming all the markers are dominant), but I remember I read somewhere
>that some markers are codominant.  I don't really understand the author
>determine whether an AFLP marker is dominant or codominant.

In principle, you could sequence the band in question and then use that
information to design site specific primers.  Those specific primers might
be found to amplify a product in some individuals and nothing in others
(dominant) or they might amplify products of two lengths with some
individuals (heterozygotes) having both bands (codominant).  My
understanding from the literature is that most AFLP markers are co-dominant
resulting from the loss or gain of restriction sites rather than changes in
the length of the amplified product.

>If it does
>exist, I wonder if the two codominant markers are inherited from the same
>ancestor (with later insertion of deletion of one to several nucleotides)?
>If so, does it mean that co-dominant markers is better in describing the
>ancestry of the plant of interest?

Really you are asking whether the specific molecular difference between the
two character-states (presence/absence or length 1 vs. length 2) resulted
from a single mutational event.  The answer is that for an individual
character you cannot tell for sure (but if it is a complex change such as
an indel you may believe a priori that this could only have happened once).
But you can pool characters, make some assumptions about the evolutionary
process (e.g., rate of generation of new character-states is quite low;
characters are independent...etc.), and then estimate the phylogenetic
history for the genomes in question.  This phylogenetic estimate can, in
turn, shed light on the homology or homoplasy of particular characters (but
note, apparent homoplasy need not be due to parallel mutation  but could
reflect genealogical discordance, i.e., reticulate gene histories).
>
Lai-Lai wrote:
>I use AFLP to determine the parentage of wild hybrids.  I haven't came
>across any publication that use AFLP for this purpose (maybe there are some
>of them which are beyond my accession), and thus am not very sure whether it
>is suitable for this purpose.  I assume that the AFLP marker of both parents
>will be inherited additively in the hybrid, as does RAPD marker which is
>commonly used to study hybridization.  I prefer AFLP over RAPD because of
>it's high repeatability.

Are your putative hybrids homoploid or polyploid?  Do you a large enough
sample of the putative parents to know if a marker is fixed?  If the
markers in question are fixed in the putative parental species (i.e., all
parents are homozygotes) then all markers should be additive in
allopolyploid hybrid populations and in homoploid F1 hybrids.  In F2's and
later generations of homoploid hybrids, markers can be lost, however, and
this loss may not be random (see Loren Riesberg's work on Helianthus).  In
that case the question becomes more statistical and I haven't really
thought through what you would do (and if hybridization is not recent you
would also need to allow for some de novo variation to arise).  Sorry I
can't be more specific.


David