Dear All,
Some colleagues asked me post the responses of my question. Here they are.
Different messages are separated by a line of symbols. Thanks to all who
responded. -Cheers! -Jenny

#####

I think the key idea you are struggling towards is that of QTLs -
quantitative trait loci.

You can detect such loci by following the co- inheritance of DNA markers and
traits such as colour in crosses. You might even detect the loci that are
involved in the trait with much work.

You could do a literature with the term qtl in it.

For example on the scirus.com site. I had a first shot at it for you. See
here

http://www.scirus.com/srsapp/search?q=qtl+cross++lines+colour&ds=jnl&g=s&t=al
l

$$$$$

This is regarding your query on the evolution directory. This is an
interesting question! One would expect association (linkage
disequilibrium, LD) between the gene locus and the trait if the gene
indeed controls the trait. If the genetic architecture of the trait is not
too complex, this association might be detectable in natural hybrid
populations. However, there are other factors as well that can
cause LD in hybrid populations, and you would have to control for
these. E.g.: (1) In both experimental and natural hybrid
populations, the crossing of two species induces LD between
many or all loci in the genome, and these associations are
expected to decay in subsequent generations. (2) Persistent
genetic structure (population subdivision) in the hybrid population
also causes LD across the genome. (3) Selection that favors
particular gene combinations (epistatic selection) can do the same.
So one would have to control very carefully for these factors. If they
are all taken into account, it may be possible to interpret
associations between genes and traits in natural hybrid populations.

There is some recent literature on this topic, please let me know if
you need some references. I also worked on this topic myself a
little in the recent past, and I have a project running that addresses
these questions in tree hybrid zones.

#####

You'd need to be extremely cautious about making that assumption.  If the
hybrids are variable in their status (e.g. some F1s, some B1 to either
parental species, some advanced-generation intercrosses and introgressions),
one would expect to find associations between fruit color and almost any gene
in the genome.  For example, in this circumstance, red-fruited hybrids would
on average carry a higher proportion of the red-fruited species genome than
blue-fruited hybrids, so red fruit would show an association with almost any
allele that is at substantially higher frequency in the red-fruited species
than in the blue-fruited species.  Such associations would be strong evidence
for linkage only if (a) all the hybrids were of the same generation (e.g.
F2); or (b) the population were a hybrid swarm many generations old that has
undergone random mating among the hybrids and no mating with the parental
species.  In (a) you could still get association due to linkage with a
causative gene on the same chromosome, similar to linkage mapping.  In (b)
the region of linkage disequilibrium would be much smaller but would still
probably encompass more than a single gene.  Thus, even in these situations,
I'd only treat allele-phenotype associations as evidence for causation if the
gene were a functional candidate.

If the status of the hybrids is variable or unknown, it may still be possible
to test associations by controlling for ancestry; e.g. using Pritchard's
STRUCTURE program, but this would require scoring a genome-wide set of
markers to provide data on ancestry.  Anyway, good luck!!

$$$$$

Before making any inferences with regards the role of your candidate "red vs
blue" gene a degree of verification needs to be carried out. If you did find
sequence polymorphisms that cosegregate with fruit colour then this is an
indication that the gene controls the phenotype, however could be due to
genetic linkage. Tests can be carried out in a number of ways - QTL mapping
will show if the gene/trait are linked, however isnt concrete proof. You
could
grow up an F2 of, say, 10000 plants and (if single gene control) genotype the
2500 reds and 2500 blues to check the sequence polymorphism cosegregates (ie
bulk DNA from 50 reds, pcr and cut with an enzyme that cuts the 'red' gene.
anything that isnt cut is a 'blue gene' in a red plant. And the same for blue
plants/red genes.
Ultimatley introducing the blue gene into a red plant and vice versa is the
acid test.
An interesting paper showing strong correlation between DNA polymorphism and
phenotypic variation is Thornsberry et al (2001) Nature genetics 28:286
Also, it seems increasingly that developmental processes are controlled by
transcription factors and these TFs may not show sequence polymorphism (eg
John Doebleys work on tb1 in maize) and are instead up- or down-regulated.
Good luck,

#####

I would say that your first suggested approach, essentially a crossing study
that correlates the presence of a particular allele with a particular trait,
would provide stronger evidence than the second, strictly correlative,
approach.  In the first case, if you find only sequence A in plants with
blue fruits, only sequence B in plants with red fruits, and both sequence A
and sequence B in plants with mosaic fruits, then, assuming A and B are
segregating at the same locus, you have evidence that this locus is linked
to fruit color.  (The mosaic result actually makes your question a little
confusing because it suggests something other than simple codominance,
perhaps a cell lineage effect of some sort)  Such results would not prove
that the locus under scrutiny controls fruit color, only that it is linked
to a locus that does.  If the locus is in a pigmentation pathway, then you
could strengthen your argument, but it would not be proof.  I would also
recommend that you not rely on natural variation in a hybrid zone, and
perform the crosses yourself, in the lab.

I would be wary about using the latter, strictly correlative, approach.
There are two problems here.  First, you cannot eliminate the effect of
phylogeny.  For example, if blue fruited plants were all derived from a
single common ancestor that had blue fruits, and red fruited plants were
derived from a different common ancestor, one would predict that any other
loci that distinguished the blue fruited ancestor from the red fruited
ancestor would "segregate" (it is not really segregation, it is historical
or phylogenetic linkage, not chromosomal linkage) with fruit color in
nature, regardless of whether the locus has any effect on fruit color.

Secondly, for intraspecific studies, given that few individuals and many
variable nucleotide positions are sampled, it is not very hard to find a
statistically significant correlation between sequence polymorphisms and
traits, even when you can be pretty sure that the locus under investigation
does not control the trait of interest.  I have been misled by such sampling
errors on several occasions in my own work.  Further sampling of individuals
proved that the statistical significance found with fewer individuals was
spurious.

Best of luck,

$$$$$

If the phenotypes are basically Mendelian and the phenotype/haplotype
associations are perfect, then you would have a pretty strong case.
Associations based on population structure in the hybrids would be
incomplete; you would find some plants with the "blue" haplotype producing
red fruits, etc.  I'm curious, though, about the mosaic phenotype and the
recombinant haplotype.  In the mosaic phenotype, do you have some individual
fruits that are red and some that are blue on the same plant? That seems odd
unless there is some sort of random inactivation of one allele, or unless the
color differences are in individual seeds and the embryo/endosperm genotype
is determining the color.  As for the recombinant haplotype, having
recombination within the gene should be very rare, so I suspect you are
getting PCR-generated recombinants.  This would only occur in heterozygotes.

#####

-----Original Message-----

From: evoldir@evol.biology.mcmaster.ca
[mailto:evoldir@evol.biology.mcmaster.ca]
Sent: Tuesday, March 08, 2005 1:32 AM
To: jenny_xiang@ncsu.edu
Subject: Other: linking phenotypes and gene sequences

Dear All,
I have a question about gene variation in relationship to phenotype diversity
that needs your help to clarify:

If two species are different in phenotype (say fruit color red vs, blue) and
they hybridize in nature. The hybrids show differences in fruit color
phenotypes  (say some have red fruits, some have blue and some have a mosaic
of red and blue). For a gene contributes to the color difference, do we
expect to see an association or correlation between the fruit phenotypes and
gene sequences in the hybrid populations? Whereas  no such correlations
should be observed for a gene that has nothing to do with the fruit color?
Is this correct?

If you observe a correlation between the phenotypes and the sequences of a
gene regulates the pathways of the pigments synthesis, could you infer that
difference in the gene is potentially linked to the difference in the
phenotypes?

Your thoughts and opinions about this would be very helpful and highly
appreciated.
Thanks!
Sincerely,
Jenny Xiang

----

Jenny Xiang <jenny_xiang@ncsu.edu>

Jenny Xiang <jenny_xiang@ncsu.edu>