Detection of positive selection

Jonah Thomas JEThomas at ix.netcom.com
Mon Sep 11 14:08:23 EST 1995


In <430vhg$jb0 at mserv1.dl.ac.uk> "JAMES O. MCINERNEY" <James.Mcinerney at UCG.IE> 
writes: 

>My question is: when can you detect instances of positive darwinian selection
>(PDS)?

>I'll elaborate...A gene duplication occurs (or speciation) and the
>conditions are conducive for positive selection (lets not get technical, the
>reason for positive selection is unimportant).  Amino acid-altering
>mutations begin to occur at a faster rate than is expected (calculations based
>on orthologous genes in other organisms, or some such measurement).

>So at some stage, our gene of interest has fine-tuned its new function and now
>is under the same kinds of selective constraints as most genes (Ks values are
>perhaps 2-5 times more frequent than Ka values).

>Will I always be able to detect this event?  Can I only detect it while it is
>happening?  Will it become masked after a certain period of time?

I'm no angel, I'll rush in.

The earliest literature I know about this comes from Cold Spring
Harbor, around 1951.  One by Szilard and someone else, the other by Day
et al, I think.  They grew bacteria over many generations under obvious
selection pressure and got mutants.  There's been a whole lot of work
since then, and the earlier among the seminal papers reference those
two.

First, they could detect the spread of each mutation because neutral
mutations were reduced.  When the favorable mutation began to dominate
then the variability of the population went down, and most of the
neutral mutants got washed out because most of them weren't in cells
that also had favorable mutations.  That trick works for large
populations, when neutral mutations are known that are linked to the
selected ones.  It might work OK with an asexual yeast culture.

The bacterial mutations seemed to happen OK without a gene duplication.
There are lots of genes that only get used sometimes.  If you modify a
protein for fucose metabolism when there isn't any fucose around, you
won't need the original version until you get some fucose, maybe a lot
of generations down the road.  Even then, the new function might be
more important.

Somebody (Barry Hall?) did experiments finding E coli that could grow
on allolactose and other peculiar sugars.  Some of them resulted in
mutations in a gene whose "normal" function was not known.  I suspected
from Hall's data that it might have something to do with
N-acetyl-glucosamine or something related, but I dunno.  I haven't done
a recent lit search on that, either, it might be known by now.

OK, they do tend to get gene duplications too.  Generally the first
step is to derepress an existing gene.  Something that doesn't work
well, so the successful cell makes more of it.  Then there might be
some steps where amino acid substitutions result in improved function.
Then with continued selection there are some gene duplication steps.
Produce twice as much enzyme and you can get just about twice the
productivity.  There's a metabolic cost, but that's slight in early
duplication steps.  How many amino acid substitutions will give twice
the productivity?  So the big deal seems to be to control the amounts
of each protein present, while coming up with improved procedures is
secondary.  That shouldn't be surprising to people who've seen human
management teams at work.  8-)

OK, can you measure any of this by amino acid substitution rates?  I
tend to doubt it.  Maybe a lot of substitutions have small effect.  And
most of that effect will be conservative.  If you get an improved
protein, and if the regulation results in the right amount of activity
instead of too much of a good thing, then the effect will mainly be to
produce less of that protein and save a little metabolic cost.  A small
metabolic effect might usually translate to an even smaller selective
effect, which would take a _long_ time to spread.  Meanwhile the
occasional large-effect mutation might take over fast and push the
small-effect ones back to start.  Active selection might result in a
few changes due to large effects.  No selection might result in a lot
of changes, most of them present at low frequency.  Conservative
selection might tend to block the changes.  So given a new selective
force, the response might be first a regulatory mutation, then a few
changes, then perhaps some gene duplications or other regulatory
changes, and finally nothing observable.

All of this is tentative theory derived from old research.  It's based
on prokaryotes.  I hope there's something useful in it.




More information about the Mol-evol mailing list