and Stubble in the third chromosome of Drosophila melanogaster can be modified ... zygous lethal and there is no crossing over in Drosophila males, this choice ...
GENETIC CHANGE OF RECOMBINATION VALUE IN DROSOPHILA MELANOGASTER. 11. SIMULATED NATURAL SELECTION1 MARGARET GALE KIDWELL
Diuision
of
Biological and Medical Sciences, Brown Uniuersity, Prouidence, Rhode Island 02912 Manuscript received June 1, 1971 Revised copy received November 9, 1971 ABSTRACT
Selection of GI-Sb coupling heterozygotes was carried out for more than one hundred generations commencing with six independent lines drawn from a common base population. Population sizes were eight, sixteen and forty-eight parents per generation. The effect of natural selection on recombination value was measured by sampling and testing females at varying intervals of time. There was a significant reduction in percentage recombination between GZ and Sb from fifteen to a level between five and ten in four out of six of the original lines. In most cases this reduction occurred rather rapidly after the initiation of the experiment. In the remaining two lines there was no significant decrease in recombination value; there was, however, a significant increase in a t least one subline of this group. The rapid rate of change of recombination value is most readily explained by the presence of a recombination modifying gene which is linked to the modified region. Genetic random drift was again shown to have an important effect on changes in recombination value in small ppulations. High recombination was almost completely recessive to low recombination in the one case examined. Lethal genes were fixed in sheltered regions of unmarked third chromosomes in five lines o r sublines. These results are discussed in relation to the mode of development of permanent heterozygosity in some species of plants.
I N the first of this series of papers (KIDWELL1972) it was shown that, in the presence of modifier genes, the recombination value between the genes Glued and Stubble in the third chromosome of Drosophila melanogaster can be modified rather quickly by artificial selection. It was also noted that genetic random drift plays an important role when the number of individuals selected is small. In the present paper the results of simulated natural selection will be presented. I n a mathematical study of the effect of selection of double heterozygotes, single and double homozygotes all being lethal, HALDANE (1962) stated that if the mean recombination value between the two loci for males and females is less than 0.25, selection will presumably tend to reduce the recombination value. This problem was studied in more detail by NEI (1967), who took into account the frequency change of modifier genes affecting recombination value. He showed that the frequency change of a modifier gene is largest when the recom111swork was supported by USPHS Research grant GM-17719 Genetics 70: 433-443 March 1972.
434
M. G. KIDWELL
bination value is around 0.10. He also showed that the rate of modification of linkage intensity by natural selection primarily depends on the degree of epistatic gene interaction in fitness and that the selection of double heterozygotes is one of the most favorable systems of selection for linkage modification. The present investigation was initiated to determine experimentally whether selection of double heterozygotes for two loci results in a reduction of recombination value. Here, no attempt was made artificially to reduce the recombination value, except to create a situation where all genotypes, other than double heterozygotes, are lethal. A s will be seen in the following, the results of the present experiment show at least qualitatively that natural selection is effective in reducing recombination value. Some of the properties of a recombination-modifying gene (or genes) ,which were discovered in the course of experiment, will also be presented. Natural selection similar to selection in the present experiment may occur in the process of evolution of so-called permanent heterozygosity, which is observed in some organisms such as Oenothera and Campanula persicifolia. MATERIALS A N D METHODS
The selection scheme consisted of choosing as parents only double heterozygotes, i.e., those individuals which carried GI and Sb in every generation. Because both these alleles are homozygous lethal and there is no crossing over in Drosophila males, this choice almost always ensured the selection in all generations of only non-recombinant gametes in the same phase as the first generation parents. The frequency of double crossovers in this interval is expected to be approximately 0.011 assuming no interference. Only a small percentage of double crossover gametes are, therefore, expected to have been included in this group of non-recombinants. Six lines were formed from the base population (KIDWELL1972) by random selection of GI-Sb coupling heterozygotes. Three different levels of population size were set up, each represented by two independent lines. 8A and 8B were maintained by four parents of each sex, 16A and 16B by eight parents of each sex, and 45A and 48B by 24 of each sex in every generation. The lines were maintained in discrete generations in half-pint milk bottles. As an insurance against loss, each line was split into two sublines of the same size, when the experiment had proceeded for 10 to 20 generations. The 8B1 subline was similarly duplicated after loss of the sister line occurred. Once formed, all sublines of the same line were maintained identically but independently. Changes of recombination over time were measured by taking a random sample of GI-Sb females from each line or subline at varying intervals up to 100 generations. Each female was mated with a wild-type male from the same line where possible and individually tested for recombination. Sample sizes for these tests varied between three and thirty in the early generations, the average being ten females per line. From generations 30 to 60, six to ten females were tested per sample and the time interval between tests was also less variable. A final test of between 10 and 20 females per subline was made after approximately 100 generations. RESULTS
Response to selection: The mean recombination values estimated at intervals over 100 generations are shown graphically in Figures 1,2 and 3 for population sizes N=8, N=16, and N=48 respectively. Sublines are differentiated by a number following the designation of the parent line. A lower case letter is used to differentiate sub-sublines where applicable. Considering the N=8 group, line 8A showed an initial steep fall in recombina-
435
SELECTION FOR RECOMBINATION CHANGE
1 0
50
100
Generations FIGURE 1.-Percent recombination between GZ and Sb in samples from populations of size N=8.
tion to a level below ten percent after five generations (Figure 1). After this initial drop there appeared to be no significant change in either subline, with the exception of the last sample from 8A1, at generation 50. Recombination value in the 8Bla subline also fell to a level between five and ten percent in 25 generations, after which there was no significant change. Line 8B2 was lost before generation 20.8Blb was lost between generations 50 and 60. In lines 16A1, 16A2, and 16B2 of the N=16 group recombination value showed a tendency to decrease as selection proceeded up to 60 generations (Figure 2). Although the linear regression of recombination value on generations was not statistically significant, the mean recombination value between generations 10 and 60 for the three lines was 11.5% which is significantly lower than
C
.-0
CI (P
20
’
C
a E 0
0
Q)
a L
Generations FIGURE 2.-Percent recombinationbetween GI and Sb in sampler from populations of size N=16.
436
M. G . RIDWELL
C
.-0
c,
m-
Q
E
a
E
8 8 CI
C Q)
f
a 0
50
loa
Generations FIGURE 3.-Percent
recombination between GI!and Sb in samples from populations of size N=48.
the initial value of 15%. 16A2 and 16B2 did not change significantly between generations 60 and 100. The recombination value in 16A1 appears to have increased between generations 60 and 100, but further tests are required to distinguished between genetic and environmental causes for this increase. In 16B1 there was a remarkable increase in recombination value from 10.3% to 24.0% between generations 20 and 50. The difference between 16B1 and 16332 was highly significant from generations 21 to 65 when 16B1 was accidentally lost. A more detailed discussion of the genetic differences between the 16B sublines will be given in a later section of this paper. In the N=48 group there was no significant change in the recombination value in 48A for the first 12 generations (Figure 3 ) . After subdivision, recombination value declined to about 8% in 30 generations in 48A1, while in 48A2 it remained unchanged for 60 generations with a mean of 15.5% (P == 0.05). The recombination value in 48A2, however, declined to about the eight percent level between generations 60 and 100. The difference between 48A1 and 48A2 was statistically significant at generations 30 to 60 (P = 0.01) but not at generation 100. There was no significant difference between the 48B sublines at any time. Recombination value fell from approximately 15% to 8% between generations 7 and 30. I n generation 18 the number of parents was drastically reduced in 48B2 due to the failure of an incubator thermostat. Only in the N=8 group were lines or sublines lost due to low fertility or fecundity. Such an event is indicated by a V in Figure 1. The high proportion lost in this group was probably due to inbreeding depression. As a rough guide to the amount of inbreeding expected with variable population size, the inbreeding coefficients for generations 10, 20, 50, and 100 were computed using the formula Ft == I - (1 - 1/2N)$ where t is time in generations and N is the number of flies mated per generation. Values of F, for N=8, N=16, and N=48 are shown in Table 1. The effective
SELECTION FOR RECOMBINATION CHANGE
43 7
TABLE 1 Inbreeding coeficienis computed for three population sizes N Generation
8
16
48
10 20 50 100
0.48 0.73 0.96 1 .oo
0.27 0.47
0.10 0.19 0.41
0.86 0.96
0.65
number of parents is likely to have been less than N , the actual number of parents. On this basis the inbreeding coefficient is expected to have been underestimated. On the other hand, these calculations do not apply to loci on most if not all of the third chromosome where heterozygosity was artificially maintained in the selected region, and to a lesser extent in adjacent regions, due to linkage disequilibrium. I n summary, the results of recombination testing in the GI-Sb interval indicate that in four out of six of the original lines, 8A, 8B, 48A, and 48B, recombination value was reduced to a level between five and ten percent. With one exception, 48A2, reduction occurred fairly rapidly in the early generations. I n the remaining two lines, 16A and 16B, there is some evidence that a smaller reduction in recombination occurred during at least some of the periods under test. I n at least one of the four sublines (16B1) ,which were formed from the original lines under discussion, a rapid increase in recombination occurred. Since a marked interval of the third chromosome was kept heterozygous in all generations and the population size was small, lethal or semilethal genes were expected to be fixed in this interval in both the marked and unmarked third chromosomes. Only fixation in the unmarked chromosome, however, could be detected in this experiment. Fixation of lethal genes in fact occurred, before subdivision, in lines 8A and 48B and after subdivision, in 48A2, 8Blc and 16B2b. I n these lines the approximate time of fixation is indicated by an L in Figures 1, 2, and 3. The fixation was recognized by an absence or paucity of wild-type flies in one or both sexes. The fixation of lethal genes in the unmarked third chromosome appeared to be unrelated to recombination change. Genetic analysis of a recombination-modif ying gene: As noted earlier (Figure 2), the 16B1 subline exhibited an exceptional and highly significant increase in recombination value between generations 20 and 50 compared with its sister subline 16B2. One of the most likely explanations, discussed more fully later, is the random fixation of a high recombination-modifying allele, either a new mutation o r an allele coming from the base population. The rapid change of recombination value suggested that this is due to a single modifier gene (or closely linked gene loci). Therefore, the genetic properties of this gene were studied in comparison with its sister line 16B2. a ) At generation 53, reciprocal crosses were made between randomly selected G1-Sb females and wild-tvDe males from the 16B1 and 16B2 sublines. Samdes of
438
M. G . KIDWELL
TABLE 2 Mean percent recombination of 16Bl and l 6 B 2 sublines and reciprocal crosses between them Parentage of tested females
1GB1 X 1GB2
Recombination n
14.8 2827
1GB2 X 16B1
14.8 1438
Control 16B1 X 1GB1
20.1 1596
Control 1GB2 X 1GB2
14.4 1591
n = number of individuals examined.
F, GI-Sb female progeny from both crosses were tested for recombination, alongside control females from two sublines. The results are shown in Table 2. There were no significant differences among the two crosses and the 16B2 control. The 16B1 control was significantly higher than the other three groups (P = .Ol). These results suggest that the increased recombination in 16B1 is dependent on a recessive modifier unlinked with the Gl-Sb markers. b) Further tests were designed to localize the recombination-modifying gene. Recombination in the GI-Sb region was measured after replacing one or both of either the first or second chromosomes in the 16B1 and 16B2 sublines. Multiple marked chromosomes were inserted by a marked inversion technique. The first chromosomes carried the markers yellow, y 2(1-0.0) ; crossveinless, cu (1-1 3.7) ; vermilion, u (1-33.0) ; forked, f (1-56.7). The second chromosomes were drawn from a marker stock obtained from Dr. P. T. IVES. They carried the markers aristaless, a1 (2-0.01) ; clot, CI (2-16.5) ; black, b (2-48.5) ; curved, c (2-75.5) ; speck, sp2 (2-107.0). The first and second marked chromosomes are denoted by and respectively, in Table 3, which gives the percent recombination between GI and Sb in eight different genetic backgrounds. All eight sets were tested within a period between generations 57 and 63. A significant difference in recombination value was found between 16B1 and 16B2 in sets 1,3 and 4. In set 2, where both second chromosomes were replaced, 16B1 and 16B2 showed similar recombination. These results suggest that the presence of at least one 16B1 second chromosome is essential for the increased recombination in this subline. The results of attempts to localize the effective regions within chromosome I1 were inconclusive. c) Additional tests of recombination were made in females which were heterozygous for the multiple marked second chromosome “a”. Recombination in the pericentric interval b-c was significantly higher in the 16B1 subline than in the 16B2 subline. There was no significant difference between the two sublines in the other three intervals measured. Thus recombination change was restricted to the centromeric region and was in the same direction in both chromosomes I1 and 111. Lly7?
&Ca77,
DISCUSSION
The results of the present experiment support the conclusion from the artificial selection experiment (KIDWELL1972) that random genetic drift plays an
439
SELECTION FOR RECOMBINATION C H A N G E
TABLE 3
Percent recombination between G1 and Sb in the I6BI and I6B2 sublines with four diflerent genetic backgrounds Set
Genotype of tested female
B,
B,
GZSb(B,)
B2
B2
a
B,
B,
n
Percent recombination
235
22.98
316
11.71
2066
13.36
1124
13.34
1466
19.92
2989
12.01
22.38
B, B, GlSb(B,) - - ____ B, B, B,
B, - (gen61) B,
1981
10.35
B, and B, denote a chromosome from the 16B1 and 16B2 sublines respectively. n = number of progeny examined.
important role in changing recombination frequency in small populations. In the artificial selection experiment, selection for low recombination was not effective despite the demonstration of genetic variability in both directions in the base population; selection for high recombination was very effective. The patterns of recombination frequency changes in the 16A and 16B sublines in the present experiment have strong similarities with the artificial selection results. Here again no reduction in recombination could be demonstrated statistically, but there was a general tendency towards lower recombination. Also, in at least one subline there was a rapid increase in recombination value against the direction in which natural selection was operating. This indicates that a high-recombination gene or genes were segregating in the population. The increase in their frequency is most readily explained by random genetic drift. It is not clear whether the high-recombination genes were present in the base population or were new mutations. The results of recombination tests in four out of six of the original lines do dem-
440
M. G . KIDWELL
onstrate that natural selection is efi'ective in reducing recombination value. In the absence of any evidence to the contrary, it is assumed that this reduction is achieved by the selection of one or more recombination-modifying genes. Apparently modifiers resulting in a reduction of recombination to approximately the eight percent level were present in the base population. Presumably the fixation of such alleles in a population precluded any further change in recombination value. The rate of reduction in recombination value observed in the N=8 and N=48 lines was in general rather rapid. With a balanced lethal system, NEI (1967) showed by numerical computation that the modifier gene frequency changes from 0.01 to 0.99 in about 190 generations when the recombination values of the three possible genotypes M M , M m , and mm for a pair of modifier genes ( M and m ) are 0.0,0.05, and 0.1, respectively. I n the present case there was no recombination in males, but only coupling-phase double heterozygotes were selected every generation. Therefore, the modification of recombination value can be more rapid, as will be seen in the following. Assume that recombination in the Gl-Sb region is under the control of a single autosomal locus with two alleles, M and m, unlinked to the marker region. Let r l , r2,r3 be the recombination values between Gl and Sb for female genotypes MM,Mm, and mm, respectively. Since only non-recombinant genotypes are selected in each generation, the fitnesses for M M , M m , and mm in females are proportional to l-rl, 1-r2, and l--r3, respectively. In males fitness is the same for all genotypes, since no recombination occurs in this sex. The exact formula for the change in gene frequency in this case is rather complicated, but WRIGHT (1969) shows that if the selective differences between the sexes are not great, a close approximation may be obtained by using the mean genotype fitnesses and mean gen? frequencies for males acd females and treating them as if there were no sex differences. Thus we have Genotypes Fitness Frequency
Mm
MM 1-r1/2 P2
mm 1-r3/2 qZ
1-r2/2 2Pq
where p (=l-q) is the mean gene frequency of M for males and females. The amount of change in gene frequency per generation is given by A p = p q C p ( r 2 - - r l )+ q ( r 8 - r 2 ) 1
where W = 1 - 1/2 (perl -t2pqr2
/
(2E)
+ qzr3).The mean recombination value is
-I = p>r1+ epqr, + q2r3
Therefore, the amount of change in recombination value per generation is approximately given by AT = (dF/dp) A p = -pq[(r2 - r d p
+
(13
- r2)qI2/
The values of A? for three different levels of dominance, three different initial
SELECTION FOR RECOMBINATION CHANGE
44 1
gene frequencies and for two sets of recombination values were computed. The true recombination values for each genotype are not known in the present experiment. However, on the basis of the observed data, two sets of r values were selected. In the first set rl = 0.08 and r3= 0.15. In the second set rI = 0.08 and and r3 = 0.22. These two sets were chosen to represent approximately the minimum and maximum differences respectively for the difference in recombination between the two homozygotes. Maximum values of LGare attaiued when the initial frequency of p is .5, .25 and .75 for the cases of no dominance, M dominant, and M recessive, respectively. Substituting rl = 0.08 and r3 = 0.15 in the formula for A3 gives a range of values from -.00006 to -.00055. The values of A7 for rl = 0.08 and r3 = 0.22 are approximately four times greater than the values computed for the first set (-.00024 to -.0023). I n the present experiment the mean reduction in recombination value over the first ten generations for all three population sizes was approximately three percent. This is of the order of five to ten times larger than the values computed for A7 with r1 = 0.08 and r3 = 0.15. The observed rate of change in recombination does however come close to the maximum computed value for A; when the difference in recombination value between the two homozygous genotypes is assumed to be .14. The close linkage of a modifier gene with the region modified is expected to result in a more rapid change in recombination value than when the modifier locus and modified region segregate independently (NEI 1969). This possibility may be an explanation of the rather rapid reduction in recombination in the present experiment, but there is no direct evidence for such linkage. I n fact, the evidence presented in the case of a high-recombination modifier (16B1) suggests that the modifier locus and the modified region are unlinked. However, there may be other loci which were primarily concerned with the reduction in recombination value. The results of reciprocal crosses between 16B1 and 16B2 for testing dominance are fairly consistent with those from the high and low artificial selection line crosses (KIDWELL1972). In the latter case female progeny of the F, high and low line crosses showed a recombination value closer to that of the low line than the high line but reciprocal crosses differed significantly in recombination value. In the 16B tests recessiveness of the enhancing allele was shown to be almost complete. There was no difference in recombination value between reciprocal crosses. These results differ from those of CHINNICI(1971) who found that high X low interline crosses indicated codominance. The conclusions regarding the effect on recombination value of substituting marked first and second chromosomes should be considered tentative due to the small numbers of females tested. However, it is of some interest that recombination in the unselected centromeric region of the second chromosome was significantly changed in the same direction as that in the third chromosome. The possibility exists that centromeric regions are under a different type of recombination control than other parts of the genome. This control, in any case,
442
M.
G.
KIDWELL
appears to have the capacity for rather rapid change. Perhaps a hierarchial control system may act in this region with a few genes controlling the centromeric regions en bloc. Changes of recombination in small subintervals of the region could be under the control of a larger number of genes with smaller effects. I n the present state of knowledge, it may be misleading to assume that the type of control exhibited in one region of the genome is necessarily exhibited in all regions. The phenomenon of permanent heterozygosity of part of the genome is not 1967). There uncommon, especially in plants, such as Oenothera (see CARSON appear to be a t least two essential conditions for such a state to evolve, namely, establishment of reciprocal translocation or inversion chromosomes in a population and the development of a balanced lethal system. Establishment of translocation chromosomes with a high frequency in a population appears to be highly dependent on population size (WRIGHT 1941; CARSON 1967) as well as on the frequency of alternate chromosome segregation at meiosis. If the frequency of such translocations becomes high, then natural selection is expected to reduce recombination occurring in the regions between the centromeres and the breakpoints of the translocations. This is because recombination occurring in these regions leads to gametes with deficient and duplicate chromosomes, particularly when the meiotic chromosome configuration is of the zigzag type, In this case natural selection for reducing recombination would be very similar to the selection scheme in the present experiment. Reduction of recombination in the regions between the centromere and the breakpoints of chromosomes seems to be important for developing a balanced lethal system, because the probability of fixation of lethal genes in a sheltered chromosome is highly dependent on the recombination value between a lethal locus and the locus which maintains the chromosome heterozygous (NEI 1970). In the present experiment five lethal genes were fixed on the wild-type chromosome in a relatively short period of time. In the case of translocation heterozygotes the probability of fixation of lethal genes would be much smaller than these observations, since the translocation heterozygotes will segregate homozygotes both for the translocation and original chromosomes before a balanced lethal system is established. Nevertheless, the evolution of permanent heterozygosity i n Oenothera and other plants seems to have occurred in the same way as the establishment of lethal genes in the present experiment. Note added in proof: Examinations of salivary gland chromosomes were kindly at North Carolina State University. No strucmade by Mr. OSAMUYAMAGUCHI tural abnormalities were found except in the case of two 8B1 sublines. The same unique inversion of a small median segment of the left arm of chromosome I11 was observed in both sublines. The conclusions that recombination change was due to recombination modifying genes are therefore essentially supported. The initial phase of these experiments was suggested by Dr. J. F. KIDWELLin whose laboratory the work was carried out. Dr. M. NEI made many helpful suggestions in the preparation of the manuscript. The encouragement and advice of both these people is gratefully acknowl-
edged.
SELECTION FOR RECOMBINATION CHANGE
443
LITERATURE CITED
CARSON, H. L., 1967 Permanent heterozygosity, pp. 143-168. In: Euolutionary Biology 1. Edited by TH. DOBZHANSKY, M. K. HECHT and W. C. STEERE.Appleton-Century-Crofts, New York. CHINNICI,J. P., 1971 Modification of recombination frequency in Drosophila. I. Selection for increased and decreased crossing over. Genetics 69: 71-83.
HALDANE, J. B. S., 1962 The selection of double heterozygotes. J. Genet. 58: 1%-128. KIDWELL,M. G., 1972 Genetic change of recombination value in Drosophila melanogaster. I. Artificial selection for high and low recombination and some properties of recombinationmodifying genes. Genetics 70 : 41 9-432. NEI, M., 1967 Modification of linkage intensity by natural selection. Genetics 57: 625-641. 1969 Linkage modification and sex difference in recombination. Genetics 63: 681-, 699. --, 1970 Accumulation of nonfunctional genes on shelterd chromosomes. Am. Naturalist 104: 31 1-322.
WRIGHT, S., 1941 On the probability of fixation of reciprocal translocations. Am. Naturalist 75: 513-522. -, 1969 Evolution and the Genetics of Populations. Vol. 2, The Theory of Gene Frequency. University of Chicago Press, Chicago.
