I.  Examples of evolution by genetic drift

    A.  Dissolution of tristyly in Lythrum salicaria.

        1.  Natural history
 
            a.  The species Lythrum salicaria exhibits the phenomenon of tristyly
            b.  Species is native in Europe, but has become naturalized in Canada
            c.  Canadian populations range in size from a few tens of individuals to several hundreds

        2.  Maintenance of tristyly by natural selection (see Handout)

            a.  Frequency-dependent natural selection gives each morph an advantage when it is rare, and thus acts to
                 preserve each morph in the population.
            b.  At evolutionary equilibrium, all three morphs are equally common.
 
        3.  Expected effects of genetic drift

            a.  Genetic drift is expected to cause observed morph frequencies to deviate from the 1:1:1 ratio that is produced
                 by natural selection.
            b.   Because the effects of drift are greatest in small populations, these deviations are expected to be greatest
                  in small populations.
            c.   It is possible that, in very small populations, the deviations might be large enough to cause loss of one or
                  two morphs.
            d.   Genetics of tristyly suggests that S morph should be eliminated by drift more frequently than other morphs.

        4.  Observations

            a.  Surveyed 102 populations and estimated morph frequencies in each.
            b.  Estimated number of individuals in each population.

        5.  Results 
 
            a.  In all populations, morph frequencies differ from the 1:1:1 ratio
            b.  In some populations, one morph is missing.
            c.  Small populations more likely to have morph missing than large populations.
            d.  S morph more likely to be eliminated than other morphs.

        6.  Conclusions
 
            a.  Both natural selection and genetic drift influence the evolution of reproductive morphology.
            b.  As expected, genetic drift tends to be important only in small populations.

     B.  Genetic Drift in Italian populations

        1.  L. L. Cavalli-Sforza studied variation in blood-group frequencies among Italian populations in and around the
               city of Parma.

        2.  In the mountains above Parma, there are many small, isolated villages
.
           a.  In these villages, most matings take place with other individuals in the same village
           b.  Consequently, there is little migration from the outside, and each village is effectively a semi-isolated
                population in which gene frequencies can evolve more or less independently due to genetic drift.

        3.  In the plains around Parma, by contrast, the towns, including the city of Parma, are much larger, and there
             is more migration and intermarriage between towns.
 
        4.  Because genetic drift is most pronounced in small populations, expect to see its effects more pronounced in
             the mountain villages than in the plains areas.

        5.  In particular, expect genetic drift to cause gene frequencies to diverge to a greater extent among the mountain
             villages than among the plains towns.

        6.  Cavalli-Sforza and colleagues surveyed gene frequencies at the Rh, ABO, and MN blood-group loci, and
             estimated the degree to which populations in these regions differed from each other.

        7.  They found that there was much greater variation in gene frequencies among populations in the mountains,
             as is expected assuming genetic drift is responsible for these differences.

        8.  Conclusion:  genetic drift appears to cause variation in blood-group gene frequencies among small human populations.
 

II.  Interaction between natural selection and genetic drift

     A.  When drift predominates vs. when selection predominates: Theoretical results

        1.  Computer simulations (demonstrated in class)

            a.  W_AA = W_Aa = 0.99,  W_aa = 1.
            b.  initial gene frequencies equal 0.5 for both alleles
            c.  simulate up to 500 generations
            d.  Outcome depends on population size.
                --when N = 10, drift predominates, a fixed about half of time
                --when N = 100, neither drift nor selection predominates, a  lost an appreciable proportion of the time, but
                   fixed more often than lost.
                --when N = 500, selection predominates, a almost always fixed

        2.  Quantitative results: Whether drift or selection predominates in causing change in gene frequency depends
              on population size, N, and the magnitude of selection, s.

            a.  if 2Ns << 1, then genetic drift predominates
            b.  if 2Ns ~= 1, then neither drift nor selection predominates
            c.  if 2Ns >> 1, then selection predominates

        3.  Note that even in large populations, if  is small enough ( the fitness advantage of a-carrying individuals is weak
             enough), then even though W_AA < W_Aa < W_aa, genetic drift will predominate.  This means that the fate of most
             weakly advantageous mutations is not fixation by seleciton, but elimination by genetic drift (probability of
             fixation of a novel mutation is 1/2N, whereas probability of elimination is (2N-1)/2N ~= 1).

    B.  Drift vs. selection as cause of actual gene frequency change

        1.  Because all populations are finite, genetic drift always occurs.
        2.  For any observed change in gene frequency, therefore, the question of interest is not whether that change was
             caused primarily by drift or selection.
        3.  Method

            a.Compare observed change in gene frequency to the largest change reasonably expected to occur due to drift.
            b.  If observed change is larger, conclude something other than drift ( natural selection) must have been operating.

        4.  Example:  Cullenward and Ehrlich

            a.  Studied gene frequency change at an electrophoretically detectable locus (see p. 327 of text) in the
                 butterfly Euphydryas anicia in Colorado.
            b.  Estimated population size by mark-recapture techniques to be 50,000 - 100,000 individuals
            c.  In each of two years, obtained a large sample of individuals and scored their genotypes at the PGM locus.
            d.  Results for PGM:
                --year 1, freq. of PGM "fast" allele = 0.56
                --year 2, freq. of PGM "fast" allele = 0.44
                --therefore, Dp = 0.56 - 0.44 = 0.12
            e.  Calculate maximum  Dp that can reasonably expect to observe if only drift were operating
                 --  s = [pq/2N]^1/2 = [0.56 x 0.44/2 x 50,000]^1/2 = 0.0015
                --2s = 0.0030,  3s = 0.0045
                -- the observed |Dp|  is much greater than 3s, and thus would be expected to occur less than 1 time out of 100 if
                    genetic drift were the only process affecting gene frequency
                --Therefore, either (i) Cullenward and Ehrlich just happened to be around to observe a very rare event caused
                    by genetic drift, or (ii) some process other than genetic drift (i.e. natural selection) was operating to cause
                    the observed change in gene frequency.
                --Conclusion (ii) is more reasonable.
                --Therefore, conclude that the observed change in gene frequency at the PGM locus was due primarily to
                    natural selection.