Founder-flush speciation in Drosophila pseudoobscura: a large-scale experiment.
Several processes leading to speciation have been postulated since the times of the dumbbell model, so many that some authors are claiming that a new unification of speciation theory is needed (e-g., Carson, 1985; Templeton, 1981, 1989). Speciation has been the subject of several recent symposia and collective volumes (e.g., Barigozi, 1982; Giddings et al., 1989; Otte and Endler, 1989). Numerous tnodels have postulated genetic drift as the critical factor in speciation (Mayr, 1963; Carson, 1970, 1975; Lande, 1980; Templeton, 1980a; Hedrick, 1981; Lyttle, 1989). Some models claim that speciation is often associated with founder events, i.e,., the derivation of a new population from one or very few founder individuals. Models of this kind include Mayr's (1963, 1982) theory of genetic revolutions triggered by founder effects, Carson's (1971, 1975, 1985) theory of, the founder-flushcrash, and Templeton's (1980a, 1981) theory of genetic trasilience.
Carson (1971) has pointed out that his founder-flush-crash model lends itself to experimental testing by means of suitably designed laboratory systems; he outlined an experimental protocol for the purpose. Several experimental tests have been carried out that follow Carson's protocol, with Drosophila pseudoobscura (Powell, 1978; Dodd and Powell, 1985), D. simulans (Ringo et al., 1985), and the housefly (Meffert and Bryant, 1991). In the present paper, we report the results of a large experiment designed to test Carson's founder-flush-crash model of speciation by mimicking in the laboratory to the extent possible the essential features of the model. The experimental design is largely similar to Powell's (1978) and Ringo et al.'s (1985), although it differs from theirs in important features. Our experimental organism is D. pseudoobscura, the same as in Powell's (1978; Dodd and Powell, 1985), but our results are substantially different.
Ancestral Populations. - Two ancestral populations of Drosophila pseudoobscura, BCA, from Bryce Canyon National Park, Utah, and MA, from Lake Zirahuen, Mexico, were established in June 1984. Several recently collected strains were generously provided by Professor Wyatt W. Anderson. The Bryce Canyon strains were chromosomally monomorphic. The Lake Zirahuen population, as well as the set of strains in our sample, is extremely polymorphic for inversion arrangements in the third chromosome (the only chromosome that is polymorphic in natural populations of D. pseudoobscura). In a sample of 464 wild individuals, 11 different chromosome arrangements were found at frequencies ranging from about 40% to less than 1% (W. W. Anderson, pers. comm.).
The BCA and MA populations were established by combining several isofemale lines and mass cultures as follows. Numerous adults captured in Bryce Canyon were used to initiate a mass culture and the remaining females to initiate 11 isofemale lines, represented as 1IBC, 2IBC, . . . ; 11 IBC. Eleven crosses were performed between the isofemale lines (11BC x 2IBC, 21BC x 3IBC .... ; 11 IBC x 1 IBC ); the adults used as founders of BCA were four couples from the F, progeny of each one of these 11 crosses and another four couples from each of the 11 reciprocal crosses, plus 14 additional couples from the mass culture. The BCA population was started, then, with 204 adult flies.
MA was established in a similar manner, using six isofemale lines (represented as 1 IM, 2IM, . . . ; 61M) and a mass culture (MM). Seven crosses were performed between these seven populations LIM9 x 2IM&, 2IM9 x 3IMC ...; 6IM x MM, MM x 1IM ). Eight couples were taken from the progenies of each of these crosses and the same number from the seven reciprocal crosses, plus 22 additional couples from the mass culture. Thus the MA population was started with 268 adult flies.
All BCA-forming lines carried the eye mutation orange, and the MA lines had the eye mutation sepia. These markers differentiate flies from the two populations and make it possible to detect contamination of one population by the other, or by any other D. pseudoobscura flies in the laboratory, which may be important in multi-year experiments like the present one. The markers were separately introduced in each BCA and MA line through six sequential backcrosses as follows. or females are crossed to wild-type males; [F.sub.1] females and males are intercrossed; [F.sub.2] or females are backcrossed to wild-type males and the cycle repeated. After the sixth backcross, or females and males were intercrossed to produce the mutant line. The procedure was the same for the se mutation, which is sex-linked in D. pseudoobscura. After six backcrosses, more than 98% of the genome is expected to be derived from the wild-type ancestor (Crow and Kimura, 1970).
Fifteen larvae were sampled from each of the six MA isofemale lines and the mass culture just before the start of our experiment; more than 50% (62/105) were heterokaryotypic for the third chromosome. No further analysis of chromosomal polymorphism was made.
Each ancestral population was maintained throughout the experiment by serial transfer: every week the adults recovered from four culture bottles were placed in a new culture bottle and the oldest one in the series was discarded. The minimum number of adult flies in a population was about 1,000 individuals.
Derived populations. - In December 1984, 27 populations were derived from BCA and 18 from MA (see Table 1). These populations were maintained according to a founder-flush-crash protocol. Each population was founded with n virgin pairs, with n = 1, 3, 5, 7, or 9. These populations were first allowed to grow "exponentially" for a few generations (flush phase) as follows. The n pairs laid eggs for a week in one culture bottle with standard Drosophila medium and for another week in a second bottle (half-pint culture bottles were used in the first few generations; 125-ml bottles thereafter). Fifty progeny flies were collected from each of these two bottles and evenly distributed among five new culture bottles, which thus had 20 flies each. After one week the 20 flies in each bottle were transferred to another bottle in a second set of 5 cultures, so that each population consisted of 10 cultures each started with 10 pairs of adult flies from the previous generation (see Fig. 1). The progenies (F, generation) were again distributed among five new bottles, but in a particular way: two flies were sampled from each of the 10 previous bottles of each set and these 20 flies were placed in a fresh culture. This was repeated five times, so that once again we had five cultures each with 20 flies that were transferred to five new bottles after one week. Additional generations of flush were prepared in a like manner. The density of 10 pairs of flies per culture was maintained throughout the flush phase, so that competition for limited resources would be slight or absent, as required by Carson's model.
[TABULAR DATA OMITTED]
After a certain number of flush generations (ranging from four to seven in different cycles, six on average), a crash was induced as follows. All flies (usually several hundred) emerging in any two nonconsecutive bottles (i.e., in two bottles that had different parents) were combined into a single culture bottle, left there for a week and then transferred to a second bottle for another week. This was done for all 10 bottles, so that all progeny flies from the last flush generation (typically, several thousand) were collected to become the parents of the crash generation. From the emerging progenies of each crash-generation culture, n virgin pairs were randomly chosen to start another founderflush-crash cycle. The value of n was kept constant throughout all cycles for any particular population.
The purpose of the experimental design is to reproduce to the extent feasible the, conditions postulated by Carson's founder-flush-crash model. Thus, each cycle starts with a founder event, where the number of founder pairs, n, ranges from one to nine. In the first flush generation, these pairs are allowed to produce many progeny in two bottles. During each of the following flush generations, only 10 pairs of adult egg-laying flies are present in each culture, so that competition for food and other resources is slight or absent. This allows recombinant genotypes, even some with low fitness, to survive along with nonrecombinants, so that they may in turn generate new recombinant genotypes through the various flush generations. Inbreeding (other than that attributable to the few founders) is kept in check by collecting the 20 parental flies in each culture equally from the 10 cultures of the previous generation. In the crash generation, resource competition is fierce, because several hundred parental flies are placed together in each culture bottle. This population crash is an essential feature of the model, because Carson postulates that it is the crash phase that selects new adaptive genotypes.
Harmonic mean estimates of effective population size and estimates of lost heterozygosity (Table 1; see Galiana et al., 1989) were calculated on the assumption that in a given generation the population size consisted of the 100 parental flies in the 10 bottles plus all the flies that would have been reared by the same protocol if none had been discarded in any previous generation of each flush-crash cycle.
Control Populations. - Two sets of inbred control populations were set up. One set consisted of six "endogamic" populations, three derived from each ancestral population, BCA and MA, and each kept for eight generations by brother x sister mating (one pair per generation) and thereafter by mass serial transfer. The other set consisted of six "prima" populations that were subjected to founder-flush-crash cycles similar to those used for the derived populations, except that the crash bottleneck consisted of only one pair and lasted three generations rather than one. (All flush cycles following the first one were accordingly two generations shorter in the "prima" than in all other populations.) In addition, we kept by mass culture the two ancestral populations, BCA and MA, as nonflush, nonbottleneck controls. (See Fig. 2 for a graphic representation of the scheme for maintaining the various populations.)
Prezygotic Isolation Tests. - We performed multiple-choice mating tests by placing together in a mating chamber 48 flies (12 virgin pairs from each of two populations) and recording the matings by visual observation for 45 minutes. The virgin flies were kept in small vials with food for five to six days before introducing them into the mating chambers. At least four replications were done for each test.
The BC and M flies could be discriminated visually because of their different eye color. When flies from two BC or two M populations were tested in the mating chambers, we clipped the tip of each wing in two of the four sex-population combinations, one per sex, and two different combinations in each replicate. In the BC x BC and M x M experiments performed during the fifth and seventh cycles (see Results), 7,676 females and 7,638 males were clipped, and 7,838 females and 7,876 males were unclipped, for a total of 15,514 matings recorded. Differences attributable to clipping were not statistically significant.
The mating choice data can be analyzed by means of various assortative mating indices. Gilbert and Starmer (1985) recommend the homogeneity chi-square test (X). We have used this as well as the statistic Y (Ringo, 1987), defined by
[Mathematical Expressions Omitted]
where A and D are the numbers of homogamic matings and B and C the numbers of heterogamic matings. The significance of Y is tested by the statistic [X.sup.2] (Y), which is chi-square distributed with one degree of freedom (Fienberg, 1977):
[Mathematical Expressions Omitted]
About 75% of the flies mated in our tests. Gilbert and Starmer (1985) recommend that this fraction not be greater than about 50% to avoid interference of different mating propensities with assortative mating. In our data there are no detectable differences in mating propensity among strains. In any case, the index Y, based on the cross-product ratio AD/BC, is a margin-free association measure (Bishop et al., 1975); i.e., it reflects only assortative mating and not differences in mating propensities. Moreover, in our experiment both indices, X and Y, yield the same set of crosses that are significant at the 5% level. X and Y range from - 1 to + 1, so that negative and positive values stand for negative and positive assortative mating, respectively, and zero for random mating. (Neither B nor C were zero in any of our experiments, which would have made it necessary to give them arbitrary low numbers to avoid zero in the denominator.)
Postzygotic Isolation Tests. - Numerous crosses have been performed to test for hybrid sterility and other postzygotic isolation between the various populations. There is no meaningful evidence of postzygotic isolation between populations. The detailed results of these experiments will be reported elsewhere.
The number of possible two-population combinations to be tested for premating isolation is so large (1,081 crosses between the 47 derived populations for each flush-crash cycle, plus tests with the various sets of control populations) that not all can be done in each cycle. We have followed a gradual approach that eventually would include all possible pairwise combinations. Here we report data for four sets of experiments. In the first set (October 1986 and thereafter), which includes the tests corresponding to cycle 4, we tested the 2 ancestral and 10 derived populations (66 pairwise combinations), 1 for each bottleneck size and origin: BC2 and M3 (n = 1), BC7 and M7 (n = 3), BC 13 and M 13 (n = 5), BC 19 and M19 (n = 7), and BC25 and M25 (n = 9).
In early 1987, we started the second set of mating tests, i.e., those corresponding to the fifth founder-flush-crash cycle. In these experiments we tested all populations founded with one or three pairs of flies, plus the ancestral populations, for a total of 210 possible pairwise combinations.
The third set of mating tests was performed in October 1988, with populations from the seventh cycle. In this case we tested all derived populations with bottleneck sizes of 5, 7, and 9 pairs; 148 tests were performed out of a total of 325 possible pairwise combinations. We also carried out at this time all 52 possible crosses between the mentioned derived populations and the two ancestral ones. In addition, 15 mating experiments, involving populations MA, BCA, M3, BC2, M7, and BC7, that had been tested in the fourth and fifth cycles, were again performed.
The fourth set of experiments involved the endogamic and prima control populations (see Materials and Methods). The prima populations were tested after their seventh flush-crash cycle.
A total of 40,257 matings were recorded, of which 5,360 correspond to the first set, 14,520 to the second set, 17,681 to the third set, and 2,696 to the controls in the fourth set. (The detailed results are not herein given, but they are available upon request from the first author.)
Table 2 summarizes the results of the mating tests involving the control populations (ancestral, endogamic, and prima; see Materials and Methods). The BCA x MA tests (ancestral populations) yield in all cases nonsignificant deviations from random mating (Y = 0.081, 0.068, and 0. 135 for the fourth, fifth, and seventh cycles, respectively). There is no evidence of positive assortative mating between any two control populations, except that there is 1 (out of 15) statistically significant test between two prima populations (which are those subject to three generations of brother x sister matings in each flush-crash cycle). If we look at the whole set of experiments, out of 39 tests there is one case of positive (and two of negative) assortative mating significant at the 0.05 level, which is quite consistent with what can be expected by chance. We interpret these results as lack of evidence for the evolution of sexual isolation between the various control populations.
[TABULAR DATA OMITTED]
All other mating tests can be sorted out into two types: (1) those involving one ancestral and one derived population and (2) those between derived populations. A summary of the results conceming the first type of tests is given in Table 3. Out of a total of 118 tests that involved one ancestral and one derived population, 8 (7.2%) exhibit significant positive assortative mating, 6 (5.1%) at the 0.05 level, and 2 at the 0.01 level, and 5(4.2%) yield significant negative assortative mating at the 0.05 significance level (see Table 3 and Fig. 3). The difference between the cases of positive and negative assortative mating is not significant ([X.sup.2] = 0.31, df = 1, P > 0.5). All cases but one of positive assortative mating involve derived populations with larger bottlenecks, but if we pool the data for n = 1-3 and for n = 5-9, this trend is not statistically significant ([X.sup.2] = 4.12, df = 2, P > 0.1). It should be noted that the two significant tests at the 1% level both involve population BC22 (with either BCA or MA).
With respect to the ancestral-derived relationship, our data (the individual crosses as well as all data sets combined; see Table 4) are almost perfectly symmetrical and thus do not give support to Kaneshiro's (1980) hypothesis that ancestral females are more discriminating than derived females (see Table 4). Our results differ in this respect from Dodd and Powell's (1985, Table 2; also in Powell, 1989, Table 11.2).
Table 4. Lack of mating asymmetry between ancestral and derived populations. (A) Number of matings for all tests; (B) matings for the two tests involving population BC22, which yields significant assortative mating in both cases. Female Female (A) (B) Ancestral BC22 Male Ancestral Derived Male Ancestral 2,358 2,141 Ancestral 57 31 Derived 2,114 2,308 BC22 32 56
The Y values for tests between derived populations are summarized in Figure 3 and Table 5. Out of a total of 370 tests, 23 show significant positive assortative mating at the 0.05 level and 13 at the 0.01 level; 3 show significant negative assortative mating at the 0.05 level and 1 at 0.01. The data for all three cycles (4, 5, and 7) may be pooled ([x.sup.2] = 5.4 5, df = 4, P > 0. 10). The number of positive cases (36) is greater than the number (4) of negative ones ([x.sup.2] = 24.02, df = 1, P < 0.005; see also Fig. 3). There are more positive cases (22, including 2 out of 6 in cycle 4) between populations with small bottlenecks (n = 1-3) than with large (n= 5-9) bottlenecks (9); and there is a significant departure from homogeneity when comparing positive with negative cases in the two groups of data ([x.sup.2] = 3.85 with continuity correction, P < 0.05). There is over-representation of populations derived from one or the other ancestral population in these significant crosses.
[TABULAR DATA OMITTED]
When numerous simultaneous tests are performed, as is the case in Table 5, a fraction of them are expected to be statistically significant just by chance. The method of sequential Bonferroni comparisons can be used as a conservative test (Rice, 1989). Table 5 shows that k = 45 tests were made in cycle 4. If we use a significance level of [alpha] = 0.05, then [alpha]/k = 0.00111. The test involving populations M19 x BC7 yields Y = 0. 414, [x.sup.2] = 14.68 with 1 df, P = 0.00012, which is statistically significant. (The homogeneity test for these two populations is X = 0.411, [x.sup.2] =15.68, P = 0.00008, which is also significant.) The test M7 x BC25 yields Y = 0.284, [x.sup.] = 8.67, P = 0.00323, which is greater than [alpha]/(k - 1) = 0.00114. In cycle 5, the test M7 x BC 10 yields Y = 0.5 27, [x.sup.2] = 12.87, P = 0.00033, greater than [alpha]/k (0.05/ 171 = 0.00029). However, X = 0.497, [x.sup.2] = 14.64, P = 0.00025, which is statistically significant. In cycle 7, k = 154; M 1 5 x BC14 yields Y = 0.343, [x.sup.2] = 9.08, P = 0.00331, which is precisely the significance level for independent tests, as is the case here [1 - (1 - [alpha]).sup.1/(1+k)] = 0.00331]. In conclusion, the strongly conservative sequential technique of Bonferroni for multiple tests shows three significant cases of sexual isolation between derived populations, one in each of cycles 4, 5, and 7.
Table 6 summarizes the values of Y for tests that involve each particular derived population combined with all other derived populations. We have made these tests by accumulating the sexual isolation data in two-by-two contingency tables, where there are two population entries: one is for a particular derived population and the second for all other derived populations with which it has been tested in a given cycle. Twenty-three out of the 45 derived populations exhibit positive (sometimes high) significant values of assortative mating in one or more cycles, and there are no statistically significant negative values. These results indicate a tendency for populations to diverge towards sexual isolation: chance deviations would tend to be compensated when data from the various pairwise combinations are accumulated. Eleven Y values from Table 6 remain statistically significant using the sequential Bonferroni technique: BC7, BC1O, BC13, BC22, BC30, M3 (cycles 4 and 5), M7, M8, M10, and M14.
[TABULAR DATA OMITTED]
Populations M3 and BC7 deserve particular attention. The pooled data for these two populations show significant assortative mating in the two cycles (4 and 5) in which they were tested with several other populations (Table 6); moreover, when the two are tested together, which was done in all three cycles, the Y values are significantly positive in all three cases (see Table 7). This at face value indicates that positive assortative mating may persist through the generations. Table 7 shows all pairwise combinations tested three times. Notice that these populations are not a random sample of the total set, but rather involve only 4 of the 45 derived populations, i.e., BC2 and M7 in addition to M3 and BC7, and that the set of tests was conceived precisely to ascertain the specificity and persistence of the isolation between M3 and BC7. In addition to the three M3 x BC7 significant cases of positive assortative mating, there are four more significant combinations, but two are positive and two negative, and none of these shows stability through time.
[TABULAR DATA OMITTED]
We have used the cumulative data in Table 6 to test the effects of geographic origin and of bottleneck size. The BCA populations (which are chromosomally monomorphic) do not show a higher incidence of assortative mating [x.sup.2] = 0. 50, or [x.sup.2] = 0.19 with continuity correction, P > 0. I 0). However, there are significantly more cases of assortative mating for populations with small (n = I and 3) than large (n = 5-9) bottlenecks [x.sup.2] = 6.3, or [x.sup.2]= 5.30 with continuity correction, P < 0.05).
[TABULAR DATA OMITTED]
Only three populations exhibit a high fraction (> 25%) of significant positive assortative mating in the tests with other populations: M3 (n = 1), BC7 (n = 3), and M14 (n = 5), all of which exhibit significant assortative mating for the cumulative data (Table 6). These three populations are involved in 19 of the 44 significant positive individual tests in the whole data set (Tables 3 and 5).
Bottlenecks may seriously alter the genetic variance of a population (Bryant et al., 1986a, 1986b; Carson and Wisotzkey, 1989). The founder-flush-crash model of speciation (Carson, 1968, 1971, 1975; Carson and Kaneshiro, 1976; Carson and Templeton, 1984) postulates that during the flush phase of the cycle there are opportunities for extensive recombination in the absence of strong resource competition and thus opportunities for creation of novel genotypes. According to the model, the crash phase sorts out the resulting genotypes selecting those well fit to succeed in the competition for resources, many of which may be novel genotypes. The sequence of genetic events thought to occur during a founder-flush-crash cycle is also thought to enhance the divergence of a colonizing population relative to its ancestral population or to other colonizing populations (Carson, 1975; Carson and Templeton, 1984).
We have succeeded in observing the evolution of assortative mating after only several generations and several cycles of a founder-flush-crash protocol. There is a clear bias towards positive assortative mating between some particular population pairs (Tables 3 and 5; Fig. 3) as well as in the pooled data (Table 6). There are derived populations (M3, BC7, M14) that repeatedly manifest positive assortative mating with respect to other populations, and there is one particular pair (M3 and BC7) that exhibits assortative mating already in the fourth-cycle tests that has persisted throughout the fifth- and seventh-cycle tests - more than 20 additional generations. Several statistical tests encompassing the statements just summarized are significant.
The large scale of the experiment makes it unlikely that the overall pattern of bias towards positive assortative mating is a chance occurrence. The results of our experiments, which attempt to model Carson's founder-flush-crash theory to the extent feasible, may then be seen as supporting Carson's model to the extent that they show that the conditions of the model may lead to the evolution of ethological isolation. However, significant ethological isolation between individual populations is observed in only a few of the very many cases tested, although in a few more than is expected by chance; and, with one exception, ethological isolation between particular populations does not persist through the various cycles. Thus, our results substantiate Barton and Charlesworth's (1984) position that "although founder effects may cause speciation under sufficiently stringent conditions" (p. 133), "[t]here are no empirical ... grounds for supposing that rapid evolutionary divergence usually takes place in extremely small populations" (p. 157; see also Charlesworth and Smith, 1982).
The results of our experiment differ in important ways from the previous founder-flush-crash study carried out with D. pseudoobscura (Powell, 1978, 1989; Dodd and Powell, 1985). The most significant differences concern the frequency and the magnitude of cases of positive assortative mating, which are much larger in Powell's (1978) than in our experiment, and their persistence over time, which is again conspicuously greater in the previous experiment (Powell, 1978; Dodd and Powell, 1985). Powell (1978, Table 4) tested 36 two-population combinations involving eight derived and one "original" (O) population, and observed 31% significant chi-squares for positive (and none for negative) assortative mating; if we exclude the tests involving the O population (one significant case out of eight tests), the incidence of significant positive assortment is 36%. In contrast, we have observed only 9.7% cases of positive assortative mating between pairs of derived populations (Table 5) and 6.8% between ancestral and derived populations (Table 3); and we have also observed significant cases of negative assortative mating in both kinds of test. When Powell (1978, Table 6) again tested pairs of populations that had earlier exhibited positive assortative mating, he observed four significant cases in five tests. The values of Y (which we have calculated for his data in Table 4 and which happen in these particular cases to be numerically nearly identical to those reported by him using a different statistic) range 0.28-0.74. In contrast, we have observed only one instance of persistent assortative mating (the pair M3 and BC7; see Table 7) and our Y values are lower (Table 7; Fig. 3). Some pairs of Powell's populations still exhibited positive assortative mating several years later, although they had been maintained by mass culture in the intervening years (Dodd and Powell, 1985).
We have no obvious explanation for these differences between Powell's and our experiments. One difference concerns the magnitude of the experiments: ours involves many more populations and many more tests than Powell's. But Powell's deviations from randomness are much too large (both in the number of instances of assortative mating and the magnitude of the assortment) to be attributable to chance. The experimental design was different in some important respects; but we attempted to meet the postulates of Carson's model more accurately. Perhaps the relevant difference is the origin of the founding populations. Powell started with a polyhybrid population (derived by intercrossing four geographically distant natural populations), whereas our experimental populations each derive from only one natural population (one or another of two populations). If this is the explanation for the different results, ours would be more relevant to Carson's model, because a polyhybrid origin is not appropriate for Carson's founder-flush-crash model, as pointed out by Charlesworth et al. (1982), Barton and Charlesworth (1984) and Barton (1989).
Our results are quantitatively more similar to those of Ringo et al. (1985; Ringo, 1987) than to Powell's. As in the present experiment, Ringo et al.'s yielded little and erratic assortative mating. Sexual isolation occurred in 12 out of 216 tests (Ringo et al., 1985; Table 4), of which 2/48 were between one ancestral and one derived population and 8/168 between two derived populations. The significance of this low incidence of assortative mating has been discounted on the grounds that D. simulans is a cosmopolitan species and, hence, not a good choice to test the founder-flush-crash model (Templeton, 1980a, 1980b). This reservation does not apply to our experiment.
Positive assortative mating appears also as a rare event in a founder-flush experiment with another cosmopolitan species, Musca domestica (Meffert and Bryant, 1991). Six founder-flush populations were studied, two for each of three bottleneck sizes (n = 1, 4, and 16). Tests after five founder-flush cycles gave two significant cases of positive and one of negative assortative mating. Fifteen population pairs were tested, nine between derived populations and six between one derived and one ancestral population. The experiments yielding positive assortative mating were two separate tests involving the same two populations (one with n = 1, the other with n = 4), which suggests that assortative mating was not a chance event in this case (Meffert and Bryant, 1991). The assortative index between the two populations was Y = 0.21, P < 0.01 (calculated by us from Table 2 in Meffert and Bryant, 1991, with the data for the two tests combined).
In our experiment, a majority of the tests yielding significant positive assortment are between pairs of derived populations rather than between ancestral and derived populations. This was also the case for Powell (1978). Barton and Charlesworth (1984) have suggested that this finding is more consistent with Wills' (1977) model - where the critical parameter is inbreeding rather than founder effects as such - than with Carson's model. If this were correct, ethological isolation would be also expected between inbred populations, such as the "prima" (F > 0.9; Table 1) and "endogamic" (F > 0.5) populations in our experiment, or the ones in Powell (1978) and Powell and Morton (1979), which is not the case. We carried out relatively few assortative-mating tests between our endogamic and prima lines, in part because these showed signs of inbreeding depression (low viability), which made it difficult to collect virgins for the tests.
Ringo et al. (1985) point out that reproductive isolation appeared in their experiment primarily between populations - one of which was ancestral, the other derived. They argue that, instead, one might expect isolation to evolve more readily between populations derived from different natural populations. This is what we have observed in our experiment where the one persistent case of sexual isolation involves populations M3 and BC7, which are derived from different natural populations. If the departure of each derived population from the ethological norm of the ancestral population is unpredictable (as Meffert and Bryant, 1991, have elegantly shown), one would expect to find more frequent, and more intense isolation between pairs of derived populations than between ancestral and derived populations. If the evolution of the ethological norm is unpredictable, it also follows that assortative mating might be nontransitive. This is what we have occasionally observed; for instance, we have observed significant assortative mating between populations BC33 and BC32, as well as between BC32 and M7, but none between BC33 and M7. The significance of this observation is, however, doubtful, because the scarcity of significant instances of assortative mating makes any conclusion about pattern of evolution tentative at best.
The number of individuals at the bottleneck may be an important parameter (see Fig. 3) that was not tested in earlier Drosophila experiments. As noted, it seems that populations with more founders (n [is greater than or equal to] 5 pairs) are less likely to evolve mating preferences under the founder-flush-crash protocol. However, the instances of assortative mating are too few, in spite of the large scale of experiment, for drawing any definitive conclusion. The matter would deserve additional testing, but the enormous amount of work involved makes it very difficult to carry out experiments of the needed magnitude. The one general conclusion of our experiment is that the founder-flush-crash protocol yields sexual isolation only as a rare (and erratic) event, which makes it difficult to quantify the contribution of the different variables involved.
Barigozzi, C. (ed.) 1982. Mechanisms of Speciation. Alan R. Liss, Inc., N.Y., USA. Barton, N. H. 1989. Founder effect speciation, pp. 229-256. In D. Otte and J. A. Endler (eds.), Speciation and Its Consequences. Sinauer, Sunderland, MA USA. Barton, N. H., and B. Charlesworth. 1984. Founder effects, genetic revolutions, and speciation. Annu. Rev. Ecol. Syst. 15:133-164. Bishop, Y. M. M., S. E. Fienberg, and P. W. Holland. 1975. Discrete Multivariate Analysis. MIT Press, Cambridge, MA USA. Bryant, E. H., L. M. Combs, and S. A. McCommas. 1986a. Morphometric differentiation among experimental lines of the housefly in relation to a bottleneck. Genetics 114:1213-1223. BRYANT, E. H., S. A. McCommas, and L. M. Combs. 1986b. The effect of an experimental bottleneck upon quantitative genetic variation in the housefly. Genetics 114:1191-1211. Carson, H. L. 1968. The population flush and its genetic consequences, pp. 123-127. In R. C. Lewontin (ed.), Population Biology and Evolution. Syracuse University Press, NY USA. _____. 1970. Chromosome tracers of the origin of species. Science 168:1414-1418. _____. 1971. Speciation and the founder principle. Stadler Genet. Symp. 3:51-70. _____. 1975. The genetics of speciation at the diploid level. Am. Nat. 109:83-92. _____. 1985. Unification of speciation theory in plants and animals. Syst. Bot. 10:380-390. Carson, H. L., and K. Y. Kaneshiro. 1976. Drosophila of Hawaii: Systematics and ecological genetics. Annu. Rev. Ecol. Syst. 7:311-346. Carson, H. L., and A. R. Templeton. 1984. Genetic revolutions relation to speciation phenomena: The founding of new populations. Annu. Rev. Ecol. Syst. 15-97-131. Carson, H.L., and R.G. Wisotzkey. 1989. Increase in genetic variance following a population bottleneck. Am. Nat. 134:668-673. Charlesworth, B., R. Lande, and M. Slatkin. 1982. A neo-Darwinian commentary on macroevolution. Evolution 36:474-498. Charlesworth, B., and D. B. Smith. 1982. A computer model of founder effect speciation. Genet. Res. 39:227-236. Crow, J. F., and M. Kimura. 1970. An Introduction to Population Genetics Theory. Harper and Row, N.Y., USA. Dodd, D. M. B., and J. R. Powell. 1985. Founder-effect speciation: An update of experimental results with Drosophila. Evolution 39:1388-1392. Fienberg, S. E. 1977. The Analysis of Cross-Classified Categorical Data. MIT Press, Cambridge, MA USA. Galiana, A., F. J. Ayala, and A. Moya. 1989. Flush-crash experiments in Drosophila, pp. 58-73. In A. Fontdevila (ed.), Evolutionary Biology of Transient Unstable Populations. Springer, Berlin, Germany. Giddings, L. V., K. Y. Kaneshiro, and W. W. Anderson. (eds.) 1989. Genetics, Speciation and the Founder Principle. Oxford University Press, UK. Gilbert, D. G., and W. T. Starmer. 1985. Statistics of sexual isolation. Evolution 39:1380-1383. Hedrick, P. W. 1981. The establishment of chromosomal variants. Evolution 35:322-332. Kaneshiro, K. Y. 1980. Sexual isolation, speciation and the direction of evolution. Evolution 34:437-444. Lande, R. 1980. Models of speciation by sexual selection on polygenic traits. Proc. Natl. Acad. Sci. USA 78:3721-3725. Lyttle, T. W. 1989. Is there a role for meiotic drive in karyotype evolution?, pp. 149-164. In L. V. Giddings, K. Y. Kaneshiro, and W. W. Anderson (eds.), Genetics, Speciation and the Founder Principle. Oxford University Press, UK. Mayr, E. 1963. Animal Species and Evolution. Harvard University Press, Cambridge, MA USA. _____. 1982. Processes of speciation in animals, pp. 1-19. In C. Barigozzi (ed.), Mechanisms of Speciation. Alan R. Liss, Inc., N.Y., USA. Meffert, L. M., and E. H. Bryant. 1991. Mating propensity and courtship behavior in serially bottlenecked lines of the housefly. Evolution 45:293-306. Otte, D., and J. A. Endler. (eds.) 1989. Speciation and Its Consequences. Sinauer, Sunderland, MA USA. Powell, J. R. 1978. The founder-flush speciation theory: An experimental approach. Evolution 32: 465-474. _____. 1989. The effects of flounder-flush cycles on ethological isolation in laboratory populations of Drosophila, pp. 239-251. In L. V. Giddings, K. Y. Kaneshiro, and W. W. Anderson (eds.), Genetics, Speciation and the Founder Principle. Oxford University Press, UK. Powell, J. R., and L. Morton. 1979. Inbreeding and mating patterns in Drosophila pseudoobscura. Behav. Genet. 9:425-429. Rice, W. A. 1989. Analyzing tables of statistical tests. Evolution 43:223-225. Ringo, J. M. 1987. The effect of successive founder events on mating propensity of Drosophila, pp. 79-88. In M. Huettel (ed.), Evolutionary Genetics of Invertebrate Behaviour. Plenum, N.Y., USA. Ringo, J. M., D. Wood, R. Rockwell, and H. Dowse. 1985. An experiment testing two hypotheses of speciation. Am. Nat. 126:642-661. Templeton, A. R. 1980a. The theory of speciation via the founder principle. Genetics 94:1011-1038. _____. 1980b. Modes of speciation and inferences based on genetic distances. Evolution 34:719-729. _____. 1981. Mechanisms of speciation: A population genetic approach. Annu. Rev. Ecol. Syst. 12: 23-48. _____. 1989. The meaning of species and speciation: A genetic perspective, pp. 3-27. In D. Otte and J. A. Endler (eds.), Speciation and Its Consequences. Sinauer, Sunderland, MA USA. Wills, C. J. 1977. A mechanism for rapid allopatric speciation. Am. Nat. 111:603-605.
|Printer friendly Cite/link Email Feedback|
|Author:||Galiana, Agusti; Moya, Andres; Ayala, Francisco J.|
|Date:||Apr 1, 1993|
|Previous Article:||Sexual selection in the barn swallow Hirundo rustica. III. Female tail ornaments.|
|Next Article:||Evolutionary effects of selection on age at reproduction in larval and adult Drosophila melanogaster.|