sternopleural selection line showed evidence of a segregating lethal with large effects on ... in the same direction for the same trait were crossed in all possible ...
Copyright 8 1995 by the Genetics Society of America
Polygenic Mutation in Drosophila melanogaster: Genetic Analysis of Selection Lines James D. Fry, Kimberly A. deRonde and Trudy F. C. Mackay Department of Genetics, North Carolina State University, Raleigh, North Carolina 27695-7614 Manuscript received May 31, 1994 Accepted for publication November 25, 1994
ABSTRACT We have conducted genetic analyses of 12 long-term selection lines of Drosophila melanogaster derived from a highly inbred base population, containing new mutations affecting abdominal and sternopleural bristle number. Biometric analysis of the number of effective factors differentiating the selected lines from the base inbred indicated that with the exception of the three lines selected for increased number of abdominal bristles, three or more mutations contributed to the responses of the selection lines. Analysis of the chromosomal distribution of effects revealed that mutations affecting abdominal bristle number occurred on ail three major chromosomes. In addition, Y-linked mutations with effects ranging from one to three bristles occurred in all three lines selected for decreased number of abdominal bristles, as wellas in one line selected for increased abdominal bristle number. Mutations affecting sternopleural bristle number were mainly on the X and third chromosomes. One abdominal and one sternopleural selection line showed evidence of a segregating lethal with large effects on bristle number. As an indirect test for allelism of mutations occurring in different selection lines, the three lines selected in the same direction for thesame trait were crossed in all possible combinations, and selection continued from the F2 hybrids. Responses of the hybrid lines usually did not exceed those of the most extreme parental lines, indicating that the responses of the parental lines may have been partly due to mutations at the same loci, although other interpretations are possible.
T
HE importance of spontaneous mutations in maintaining genetic variation for quantitative traits in natural populations depends on thegenetic properties of new mutations and the mechanism of natural selection (e.g., LANDE 1975, 1980; TURELLI1984,1985; LYNCHand HILL1986; COCKERHAM and TACHIDA 1987; HILL and KEIGHTLEY 1988;BARTON1990; KEIGHTLEY and HILL1990; KONDRASHOVand TURELLI 1992; GAVRILETS and DE JONG 1993) . Without information on ( 1) the numberof loci capable of mutating toaffect a given quantitative trait, ( 2 ) mutation rates at these loci, (3) homozygous and heterozygous effects of mutations on the trait and pleiotropic effects on fitness and (4)epistatic effects of mutations on the trait and on fitness, the validity of mutation-selection balance hypotheses for maintaining quantitative genetic variation is questionable. Current knowledge of the properties of new mutations affecting the model quantitative traits, abdominal and sternopleural bristle number in Drosophila melane gaster, is based on effects of P-elementinsertional mutations ( MACKAY et ul. 1992) and analysis of individual spontaneous mutations of large effect arising in artificially selected or inbred lines derived from a common isogenic base population (CABALLERO et al. 1991; SANTIAGO et al. 1992; LOPEZ and LOPEZ-FANJUL 1993a,b). For both categories of mutation, homozygous mutational Corresponding author: Trudy F. C. Mackay, Department of Genetics, Box 7614, North Carolina State University, Raleigh, NC 276957614. Genetics 139: 1293-1307 (March, 1995)
effects on bristle traits have high variance, are sometimes skewed, and are highly leptokurtic. Degrees of dominance are variable, with a tendency for mutations with large effects to be recessive. Mutations with large bristle effects also tend to be deleterious with respect to the viability component of fitness. Here we report the results of genetic analyses of D. melanogaster long-term replicate selection lines derived from a highly inbred base population, containing new mutations affecting abdominal and sternopleural bristle number (MACKAY et al. 1994). After -100 generations of mutation accumulation, replicate lines selected for high and low abdominal bristle number had diverged on average by 12 bristles, and those selected for high and low sternopleural bristle number had diverged by an average of 8 bristles. The minimum number of loci contributing to selection response was estimated from a biometric analysis of the number of effective factors (WRIGHT1968) differentiating each replicate selection line from the base inbred population and by determining the chromosomal distribution of effects contributing to selection response in each line (with the minimum number of loci given by the number of chromosomes with effects). The contributions of X-linked, Flinked and autosomal loci to the divergence in bristle number between the selection lines and the base inbred and of average dominance of X-linked and autosomal loci were estimated by generation means analysis (MATHER and JINKS 1971). Possible epistaticinteractions among muta-
1294
J. D. Fry, K A. deRonde and
tions were inferred from interchromosomal interaction effects and, more indirectly, from lack of fit to the additivedominance model used in thegeneration means analysis. To determine whether replicate lines had accumulated mutations at thesame loci, which would suggest high per locus mutation rates, selection was continued from crosses of lines selected in the same direction for the same trait. Pleiotropic effects of mutations affecting bristle number on viability were inferred indirectly from the frequencies of lethal chromosomes in the selection lines and heterozygouseffects of the lethal chromosomes on the bristle traits. MATERIALS AND METHODS Drosophila s t r h We have derived 12 lines containing spontaneous mutations affecting abdominal and sternopleural bristle number by long-term replicated divergentselection from ahighly inbred base D. mlanogasterpopulation ( MACKAY et al. 1994). All selection lines and crosses described below were reared in shell vials with 10 ml cornmeal-agar-molasses medium at 25", at a rearing density of 10 pairs per vial. Compound a_"d balancer chromosomes used were C(1)DX (abbreviated X X ) ; T(2;3) up" (abbreviated Xu) ; Basc (abbreviated B ) ; SM5, Cy (abbreviated Cy) and TM64 Tb (abbreviated T b ) ( LINDSLEY and ZIMM 1992). Abdominal and sternopleural bristle numbers were scored as described by MACKAY et al. (1994). Estimates of numbersof effective factors: The effective number of loci, ne, contributing to the difference in mean between two populations for a quantitative character, can be estimated by n, = D 2 / 8 VA, where Dis thedifference in mean between the two populations and VA the additive genetic variance segregatingin the cross (WRIGHT1968; LANDE 1981; FALCONER 1989; ZENG et al. 1990). This method assumes the two parent lines compared are fixed for all alleles increasing (decreasing) thevalue of the trait atall relevant loci, unlinked loci, equal allelic effects at all loci and additive inheritance. Violation of any of the first three assumptions leads to n, underestimating n, the true number of loci contributing to the difference in mean of the two parental lines. Violation of the additive inheritance assumption can upwardly bias ne (WRIGHT 1968; MATHER andJINKs 1971 ), but unless nonadditivity is extreme, this effect is likely to be overshadowed by the downward bias produced by violation of the first three assumptions (WRIGHT1968; ZENG et al. 1990). Therefore n, may be viewed as a rough estimate of the minimum number of loci differentiating two lines. We used this method to estimate the minimum number of loci differentiating each replicate selection line from the unselected inbred sublineused to found the 12 selection lines. The unselected linewas maintained contemporaneously with the selection lines by continuous full-sib mating. At selection generation 114, reciprocal crosses were made of each selection line replicate to the unselected inbred line. Reciprocal F, individuals were crossed inter se to give reciprocal F2 progeny, and four generationsof divergent selection for the bristle trait of the selected parent linewere carried out from the reciprocal F2 crosses, with the 10 most extreme pairs selected from 20 parents scored each generation. VA was estimated as Vp GML, where Vp is the mean phenotypic variance from the F2 and the four generations of upward and downward selection, and h h M Lis the realized heritability from divergent selection, calculated using the restricted maximum likelihood method of MACKAY et al. ( 1994). D was estimated
T. F. C. Mackay fromwhere $ the mean bristle number of each selection line replicate, was based on the scores from 20 males an! females from each of selection generations 113-115 and Z, the mean of the parental inbred, was based on abdominal and sternopleural bristle numbers scored on 100 individuals of each sex from two successive generations. Generationmeans analysis: Generationmeans analysis ( MATHER andJINKs 1971) was used to estimate the contributions of autosomal ( [ d,] ), X-linked ( [ d x ]), Y-linked ( [ 41 ) and cytoplasmic ( [ d,] ) effects to the difference in bristle number between the selected lines and the base inbred, as well as summed dominance deviations of autosomal ( [ h,] ) and X-linked ( [ h,] ) loci. The analysis is based on the mean bristle numbers of each parent line, of the reciprocal Fls and F2s (generated at G114 as described above) and of the backcrosses formed by crossing the reciprocal Fls to the parent lines in all possible combinations (SI X I, SI X S, I X SI, S X SI, IS X I, IS X S, I X IS, S X IS, where I and S are the inbred and selected strains, respectively, and SI and IS are the reciprocal Fls, with the first letter indicating the female and the second the male parent strain). The generationmeans analysis was performed by fitting the following regression models:
Y+ = [ ~ l + [ d , ~ l ~ ~ + [ d ~ I X ~ , + [ d ~ l X ~ ,
+ [ ha] X,, + [ h,] X,, + error
(for females),
x, = I r n I + [ d ~ l X 1 , + [ d , l X s , + [ d , l X ~ , +[h,]X4,+[4]X?2+error (formales), (1) Models for males and females were fitted separately. Klis the mean bristle number of flies of the given sex emerging from the j t h replicate vial belonging to the ith generation ( i = 114; ie., parental lines, reciprocal Fls and so o n ) . [ rn] is the grand mean, and the Xs are coefficients reflecting the expected contribution of the additive and dominanceeffects to the mean of the ith generation. For example, X,, is 1 for the selected parent, -1 for the inbredbase parent, and0 for both reciprocal Fls and F2s, which should have equal numbers of autosomal alleles from each parent line, whereas is 0 for each parent line and +1 for bothreciprocal Fls.The complete set of coefficients used is given in MATHER and JINM (1971, Table 101 ), with the exception of those for [ d,] and [ d,] ; these are +1 if the Y chromosome or cytoplasm came from the selected parent and -1 otherwise. The regression model ( 1) is based on the assumptions that the parent lines are homozygous for genes affecting bristle number, there are no epistatic interactions between bristle loci and genotypes in the segregating crosses ( F2 and backcrosses) have equal eggto-adult viability. It is not necessary to assume that the two parent lines are fixed for eitherall increasing or all decreasing alleles or that linkage is absent. For each of the 12 F,, F2 and backcross generations, there were two replicate vials with 30 individuals of each sex per vial scored for bristle number. For the parent inbred strain and @e sekcted lines, the vial means were those used to calculate Sand Zabove. Ordinary ( i e . , unweighted) least-squares regression was used to obtain parameter estimates and their standard errors. An Ptest for lack of fit of the regression model to the data was performed following NETER et al. ( 1990, p. 131). Chromosome extractions: To further characterize the chromosomal distribution of mutations contributing to the selection responses, balancer chromosomes were used to establish a set of sublines from each selection line (Figure 1) at selection generations 102-105; each subline contained a single first, second, and third chromosome derived from a single selection line male. In each subline at G4, + /Y;Cy/ +; Tb/+ males were crossed to B / + ; Cy/+; T b / + females
1295
Drosophila
GI
x x -
A
in
Xa
Mutation Polygenic of the double heterozygote class (Cy/+; Tb/+) minus the mean of each single homozygote class ( Cy/+; + /+ and + / Tb/+). Finally, three-way interaction effects were calculated as the following combination of means:
99
+:+
+;
(B/+orY;Cy/+;Tb/+) 8
Xa
8
C y ; Jb
G2
--
99
G3
f Cy Jb --
99
8
+1
+ (B/+orY;+/+;+/+) - (+/+orY;Cy/+;Tb/+) + ( + / + o r Y ; C y / + ; + / + ) + (+/+orY;+/+;Tb/+) - ( + / + o r Y ; + / + ; + / + ) . -
h 1
+zi
r G4
Tb ---
8
Cy
+1
+zi
99
h l
G5
8 "-
+ZI
Tb
+1 or
+Zl
hi
7
+I "-
Cy
Tb
+I Or
+z/
+3l
FIGURE1.-Crossing scheme for chromosomallocalizations of mutations affecting bristle number in the selection lines. "+" is a chromosome from the selection line; other symbols are described in the text. in a single vial, and bristle numbers were scoredon five progeny per sex of each of the resulting eight G5 genotypes. If genes on chromosome i contribute to the differentiation between the high and low lines selectedfor a particular bristle trait and assuming these genes are not wholly dominant in their effects on bristle number, then the mean bristlenumber of (or + / Y ) flies minus the mean bristle number of M i/+ (or B/Y) flies should be higher in the high lines than in the low lines (here, Mirepresents the dominant marker on the balancer for chromosome 2). These daerences ( chromosomal"effects")werecalculated for each subline not bearing lethals ( ie., in which all genotypic classes were present). For example, the second chromosome effect ina particular subline was estimated as the mean of all Cy' flies minus the mean of all Cy flies, with the means taken over sexes, B and B+ flies and Tb and Tb' flies. The mean effectfor a given line was then estimated as the average of the effects for all the sublines derived from that line, and the empirical standard error of effects among sublines was calculated. An extension of this method was used to calculate the means and standard errors of interchromosomal interaction effects. For example, the interaction between the second and third chromosomes in a particular subline was calculated as the mean of the double homozygote class ( + /+; + /+ ) plus the mean
+/+
- (B/+orY;Cy/+;+/+)
(B/+orY;+/+;Tb/+)
There are two major limitations of the above method for estimating chromosomal effects. First, because it cannot be assumed that chromosomal effects would have been zero in the base inbred, the method does not allow determination of whether a given chromosome contributed to the selection response in a given selection line but only whether the differentiation between the high and low lines was caused in part by that chromosome. Second, due to epistasis, a given chromosome could have a different effect in the genetic background provided by the balancer chromosomes than in the genetic background of the selection lines. Nonetheless,if all three lines selected upward for a given trait show greater estimated effects of a given chromosome than all three lines selected downward for the same trait, the most reasonable conclusion is that that chromosome contributed to the selection responses. In contrast, negative results for a given chromosome must be interpreted with caution. Absence of Cy' or Tb' flies from the progeny of a cross indicated that the particular subline carried a second or third chromosome lethal. Bristle numbers of the emerging genotypes were nonetheless counted. When two or more sublines from a given selection linewere found to have second or third chromosome lethals, the sublines were crossed to determine whether the lethals were allelic. Crossesbetweenreplicateselectionlines: To determine the degree to which replicate selection lineshad accumulated mutations affecting bristlenumber at different loci, all possible reciprocal crosses were made at selection generation 94 between replicate lines selected inthe same direction for the same trait (ie., 1 X 2, 2 X 1, 1 X 3, 3 X 1, 2 X 3, 3 X 2; the numbers refer to the three independent replicate selection lines), and directional selection was continued for nine further generations from the F2 of each cross.Under an additive model with different neutral mutations fixed in each replicate line, the final response of the cross should be equal to the sum of the parental responses. Ten pairs of parents were used for each cross. A sample of 20 FI individuals of each sex were scoredfor the appropriate bristle trait, and 10 pairs randomly mated togive F2 progeny. Selection was then continued from the F2of each cross in the same direction as the parental replicates for nine further generations, with the 10 most extreme pairs selected from20 pairs scored each generation. RESULTS
Estimates ofnumbers of effective factors: Estimates of the number of effective factors differentiating the selection lines from the unselected inbred, along with the parameter estimates used to derive them,are given in Tables 1 and 2. Estimates averaged 0.55 for the high abdominal bristle number lines, 5.0 for the low abdominal lines, 5.9 for the high sternopleural lines and 3.4 for the low sternopleural lines. Estimates are given for completeness for the high
J. D. Fry, K. A. deRonde T.and
1296
F. C. Mackay
TABLE 1 Divergent selection from the F2of crosses of G114 high (HA)and low (LA) abdominal bristle number selection lines to the inbred (I) base population Cross (female X male)
D
h&,,
(C. L.)
VA
ne.
0.12 (0.03, 0.27) 0.21 (0.09,0.42) 0.12 (0.02, 0.27) 0.15 (0.06, 0.32) 0.19 (0.08, 0.38) 0.12 (0.03, 0.29)
0.36 0.61 0.29 0.59 0.73 0.35
0.99 0.60 1.02 0.50 0.07 0.14
0.43 (0.28, 0.64) 0.50 (0.35, 0.70) 0.44 (0.31, 0.63) 0.40 (0.27, 0.58) 0.32 (0.20, 0.50) 0.32 (0.20, 0.50)
3.57 3.41 2.73 2.37 2.31 3.52
6.16 6.45 5.08 5.85 3.74 2.45
VP
HA 1x1 1x1 2 x 1 1 x 2 3 x 13.85 1 x 3
1.70 1.54
3.18 2.83 2.53 3.83
0.63 2.96
LA 1x1 1 x 1 2 x 1 1 x 2 3 x 17.25 1 x 3
8.35- 13.26 6.83 6.16- 10.53 5.91 -8.31 10.91
D is the mean difference in bristle number between the base population inbred and the selection line (G113-Gl15 average), V, is the average phenotypic variance during four generationsof selection; h k M L(C. L.) are heritability estimates (95% confidence limits) from REML analyses of response to selection. Additive genetic variance (V,) is computed using h&ML and phenotypic variance. n, is the effective number of loci. abdominal lines, but in this case the use of WRIGHT’S estimator may not be warranted with so little divergence between theparental lines. Heritabilities estimated from the F2 of these crosses are appreciable, but divergence between theparental lines, averagedoverall three selection line replicates, is only a little over one bristle (Table 1) . It appears, therefore, that theassumption that the two parental lines are fixed for either all increasing or all decreasing alleles is violated for the high abdominal crosses. Generation means analysis:The contributions of autosomal, X-linked, Y-linked and cytoplasmic effects to
the difference in mean bristle number between the selected lines and the base inbred and summed dominance deviations of autosomal and X-linked genes were estimated by generation means analysis ( MATHER and JINKS 1971) . Cytoplasmic effects were never significant in males ( P > 0.1 ) and were significant in females (0.01 < P < 0.05) in only two lines, low sternopleural replicates 2 and 3. Estimated cytoplasmic effects ( [ d,] ) in these lines were-0.20 and -0.17, respectively.Given the small magnitude of these estimates and the number of significancetests performed, the reality of the cyto-
TABLE 2 Divergent selection from the F2of crosses of G114 high (HS) and low (LS) sternopleural bristle number selection lines to the inbred (I) base population
HS
3.78
1 1 2 1 3 1 LS 1 1 0.64 2 4.32 1 3 1
x x x x x x
1 1.53 1 1 2 1 1.54 3
3.01 5.96
1.74 3.91 3.44
4.54 2.04
x 11.17 0.48-2.13 0.47) 1.65 (0.17, x1 0.09 1.12 x0.49) 1 (0.21, -4.410.32 1.99 x 0.56 2 0.47) (0.18, 1.89 x 1 4.78 -1.76 0.08 0.25) 1.09(-0.02, 1.06 1.60 0.37 x 3
Parameters are as in Table 1.
0.21 (0.11, 0.38) 0.19 (0.09, 0.35) 2.82 0.40 (0.25, 0.62) 0.14 (0.06, 0.27) 0.17 (0.07, 0.33) 6.39 0.20 (0.09, 0.37)
0.33 0.34 1.58 0.48 0.26 0.40
3.48 3.38 9.31 9.95
0.29 (0.00, 0.26) 0.30 0.07 0.23 (0.11, 0.43)
0.10
5.45
Mutation Polygenic
in Drosophila
1297
TABLE 3 Results of generation means analysis of crosses between selected linesand the base inbred for bristle characters scored on males
Line
[&I
D
Abdominal bristles HA1 1.3 HA2 0.7 HA3 LA1 -12.9 LA2 -11.5 LA3 -9.4 Sternopleural bristles HS1 HS2 HS3 LSl -2.1 LS2 -4.6 LS3 -1.7
0.24 -0.41 0.29 -4.47 -2.96 -3.51
[d,l
(0.39) 0.30 (0.23) (0.29) 0.35 (0.18)* (0.32) -0.12 (0.19) (0.63)**** -0.92 (0.38)** (0.59)**** -0.40 (0.36) (0.66)**** 0.42 (0.40)
1.03 (0.21)**** 0.48 (0.15)*** 2.94 (0.35)**** -0.12 (0.25) 2.00 (0.44)**** 0.26 (0.32) -0.98 (0.14)**** 0.01 (0.10) -1.56 (0.15)**** -0.71 (0.11)**** -0.70 (0.17)**** -0.13 (0.12)
[ d,l
0.12 0.48 0.14 -0.65 -1.56 -1.08
(0.15) (0.11)**** (0.12) (0.24)** (0.23)**** (0.25)****
-
-
[h,l
R2
Fit (probability)
-0.27 (0.34) -0.16 (0.26) -0.26 (0.28) 0.81 (0.56) -0.16 (0.52) 1.30 (0.59)**
0.43*** 0.56**** 0.30* 0.95**** 0.95**** 0.91****
0.022 0.031 0.001 0.116 0.1 ) ; [ 41 was therefore left out of the final model for these lines to simplify the analysis. Estimates of the summed dominance deviations of autosomal genes ( [ ha]) weresignificantly different from zero in 8 of 24 cases, including for both sexes in low abdominal replicate 3 and low sternopleural replicate 1 (Tables 3 and 4). If a Bonferroni critical value of 0.05 / 24 = 0.002 is used to correct formultiple comparisons, none of the [h,] estimates are significant; by this criterion, the null hypothesis of additivity cannot be rejected. Two lines of evidence, however, argue against accepting the additive null hypothesis: 8 individually significant results out of 24 is considerably more than the 1.2 expected by chance andif the truevalue of [ ha] is zero in both sexes in all 12 lines, then the estimates should be randomly distributed about zero,without regard to the direction of selection of the line. This does not appear to be the case: the estimates are opposite in sign to the direction of selection ( i e . , negative for upward-selected lines and positive for downward-selected lines) in 21 of 24 cases. An exception to this pattern occurs in low abdominal replicate 2, in which both [ h a ] estimates are negative, with the estimate in females approaching significance. Although the estimates of [ h,] depend on theassumption of the generations means analysis of no selection in the F2 and backcross generations, the means of the F1 females were closer to that of base inbred females than to selected line females in all casesexcept low abdominal replicate 2 (data not shown), suggesting that the tendency of the [ ha] estimates to be opposite in sign to the direction of selection is not an artifact of violation of the noselection assumption. Assuming that the lack of signifi-
I 13 12 15
A
14
A
11-
I
A
10
-
A
A
8 A
4
9-
A
8-
A
7-
A
6 -
5
A I
l
l
9 0
L " L -
m
Low Abdominal Replicate 2
1
3 m
Low Abdominal Replicate 3
A
14 A A.
A A
A A
13
-
12 -
A A
A
A A
f A
11 -
A A
71
4
A
10
A
-
FIGURE2.-Evidence for Y-linked effects in lines selected for increased or decreased numbers of abdominal bristles. Each point represents the mean bristle number of 30 males reared in a single vial. Closed symbols represent males with Y chromosomes derived from the selection line, and open symbols represent males with Y chromosomes derived from the base inbred stock. "BGS" and "BC-I" refer to males formed by backcrossing reciprocal F, males to females from the selected line and inbred stock, respectively.
1299
Drosophila
in
Mutation Polygenic
cance of any single [ h a ]estimate when the conservative Bonferroni criterion is adopted may result from low power, the tendency for [ h,] estimates to be opposite in sign to the direction of selection is consistent with the conclusion that mutations a e c t i n g bristle number are often partially recessive in their effects. Estimates of summed dominance deviations of X-linked genes in females were significant in only one case (none after adopting the Bonferroni criterion) andshowed no consistent pattern (Table 4). The parameter estimates in Tables 3 and 4 are presented with the caveat that violation of the assumptions of the generation means analysis (see MATERIALS AND METHODS) could cause the estimates to be biased to an unknown extent. Violation of the assumptions may explain the significant lack of fit of the model in 14 of 24 datasets (Tables 3 and 4). Analysis of extracted sublines: For lines selected for abdominal bristles, estimates of chromosomal effects give evidence that all three major chromosomes contributed to the differentiation between the high and low lines (Figure 3 ) . In the low lines, which showed much greater selection responses than the high lines, the estimated effects for each chromosome were greatest for line 1, the line showing the greatest overall response, and smallest for line 3, the line showing the smallest response ( c t estimates of divergence in Table 1). The conclusion that theXchromosome contributed to thedifferentiation between the high and low abdominal lines contrasts somewhat with the results of the generation means analysis, which indicated the presence of Xeffects in only one low line and possibly one high line. For the sternopleural lines, only third chromosome effects showeda consistent difference between the high and low lines (Figure 4 ) . Estimated second chromosome effects for all six sternopleural lines were within one-half bristle of one another.Somewhat surprisingly, in high sternopleural replicate 2, the estimated X-chromosome effect was lower by more than one bristle than that for the other five lines, the opposite of what one would expect if Xchromosome mutations had contributed to the selection response in this line. This result could be explained by the occurrence of a second or third chromosome mutationin this line that enhanced the bristle-producing effect of the Basc balancer chromosome, which carries mutations atthe achaete-scute complex. (Note that the estimated X-chromosome effects are strongly negative for all sixsternopleural lines, indicating that Bflies always averaged more bristles than B+ flies). Thethree possible two-way interaction effects between the chromosomes, and the three-way interaction effect, were estimated for each extracted subline. As with the chromosomal main effects, it cannot be assumed that interactions would have been zero in the base inbred. However, if mutations contributing to the
ccc
m 1
t-a-i
m
Chromosome CM
+A+
I
1
I
I
I
I
-2.0
-1.5
-1.0
-0.5
0.0
0.5
m m
+A"
+"i t"+
2
ca"-l I
I
I
I
I
-1.5
-1.o
-0.5
0.0
0.5
w
+"+ k&
t"i
3
m
t-A+
I
I
I
I
I
I
-1.0
-0.5
0.0
0.5
1.0
1.5
Chromosome Effect
FIGURE3.-Estimates of chromosomal main effects inlines selected for high and low abdominal bristle number. Each point is the mean of the estimated chromosomal effects for the different extracted sublines derived from a given selection line replicate; barsgive the standarderrors of the means, calculated fromthe empirical standard deviationsamong sub lines. Filled symbols are for high lines and open symbols are for low lines. Replicates 1, 2 and 3 are represented by circles, squares and triangles, respectively. Sample sizes (number of sublines per selection line replicate) are as in Table 7.
Chromosome
w
CH
+A+
+"1
m-
w
I
I
-5.5
-5.0
I
I
I
I
--4.5
-3.5
4.0
-3.0
k 0 i
u3-
2
CM
u
I
I
I
-1.0
-0.5
0.0
HI
+-"--
3
1
-
0.5
w
0
rY
I
I
I
I
I
I
-3
-2
-1
0
1
2
Chromosome Effect
FIGURE4.-Estimates of chromosomal maineffects in lines selected for high and low sternopleural bristle number. Symbols are as in Figure 3.
selection responses interact with each other or with genes on the balancer chromosomes, then thedifferent lines selected for the same character should display different patterns of interaction. Analysis of variance on the estimated interaction effects shows highly significant variation among lines ( P < 0.01 ) in three cases:
J. D. Fry, K. A. deRonde and
1300
AB lines P = 0.007
Chromosomes
I
I
I
I
-1
-0.30
-0.1 5
0.00
0.15
0.30
AB lines P = 0.001
-0.4
-0.2
0.2
0.0
-
SB lines P = 0.0004
" Kti
2x3
F"
-0.75
-0.50
-0.25
0.00
0.25
Interaction Effect
FIGURE 5.-Estimates of interchromosomal interactioneffects for the cases in which significant among-line heterogeneity in estimatedinteractioneffects was present. "AB" and "SB" refer to abdominal and sternopleural bristles, respectively; other symbols are as in Figure 3. Significance levels are from one-way analyses of variance comparing the six lines.
the interactions between chromosomes 1 and 2 and between chromosomes 1 and 3 in the abdominal lines, and the interaction between chromosomes 2 and 3 in the sternopleural lines (Figure 5). In all three cases, two lines showed stronger mean interaction effects than the other four lines. The remaining two types ofinteraction in the abdominal lines and three types of interaction in the sternopleural lines did not show significant among-line variation ( P > 0.05). For these typesof interaction, the estimated interaction effectsof lines were within two standard errors of zero in all but 4 of 30 cases, where the standard error is calculated as the empirical standard error among sublines. Lethal chromosomes in the selection lines: The preceding analyses could only be performed on extracted sublines not containinglethal chromosomes. The numbers of sublines bearing second and third chromosome lethals are shown in Table 5. The high abdominal and low sternopleural lines had zero to three unique lethals in the 14-18 extracted sublines. In contrast, low abdominal replicate 3 and high sternopleural replicates 2 and 3 each showed from three to seven occurrences of one ormore lethal in 17-26 extracted sublines, with the other low abdominal and high sternopleural replicates having fewer lethals. Lethals from different selection line replicates always complemented one another. The relatively high frequency of noncomplement-
T. F. C. Mackay
ing lethals in three of the selection lines suggests that these lethals may have had heterozygous effects on bristle number, resulting in the maintenance of the lethal by a balance between artificial and natural selection. Two of the lines bearing high frequencies of particular lethals, low abdominal replicate 3 and high sternopleural replicate 2 showed behavior consistent with the presence of a segregating lethal with large heterozygous effects on bristle number: both lines showed unusually high phenotypic variances and rapid reversion of the selection response, accompanied by decline of the phenotypic variance, upon relaxation of selection (IMACKAY et al. 1994). A rough test for heterozygous effects of the lethal-bearing chromosomes on bristle number is to compare the mean bristle number of Cy/+; T b / + balancer heterozygotes from sublines bearing a particular lethal to that of balancer heterozygotes from sublines from the same selection line not bearing any lethals (Table 6). These comparisons give little evidence for heterozygous effects oflethal-bearing chromosomes in any selection line except high sternopleural replicate 3, which did not show behavior consistent with the presence of a lethal with heterozygous effect on bristle number. Two explanations for the failure to find heterozygous effects of lethals in low abdominal replicate 3 and high sternopleural replicate 2 can be considered. First, it is possible that the lethals in question have different effects in different genetic backgrounds, so that their effects on bristles are not expressed when they are made heterozygous against the balancer chromosomes. Second, the high phenotypic variances of these lines may have had nothingto do with the presence of segregating lethals, instead being caused by mutations causing developmental instability. One way to distinguish these two hypotheses is to compare the phenotypic means and variances of wild-type homozygotes ( + /+ or Y; /+; /+) from thenonlethal bearing extracted sublines to the means and variances of the original selection lines. If the former hypothesis is correct, then the means of extracted homozygotes should show reversion toward the unselected value, because lethal chromosomes would no longer be present; similarly, the phenotypic variances of theextracted homozygotes should be lower than that of the selected line. Under the latter hypothesis, there is no reason to expect either the means or the phenotypic variances to be different between the selection lines and the extracted homozygotes (except for possible Y and fourth chromosome effects, which could cause the means to differ; see below). The means and phenotypic variances of extracted homozygotes arecompared with the selection line means and variances in Table 7. The results give support for the presence of lethals with strong heterozygous effects on the selected trait in low abdominal replicate 3 and high sternopleural replicate 2. In both cases,
+
+
1301
Polygenic Mutation in Drosophila
TABLE 5 Lethal chromosomes extracted from the selection lines
Chromosome trait
Selected
abdominal
High Low abdominal
1 2 3 1
N
2
17 18 14 18
0 1 (1) 0 0
3
1 (1) 2 (1, 1) 0
High sternopleural Low sternopleural ~~
~
Nis the numberof sublines established from each selection line replicate. Given in the table is the number of independent lethals found in each selection line and in parentheses the numberof occurrences of each. zygous that would be expected under this hypothesis. The extracted sublines bore Y chromosomes derived from the attached-X stock (Figure 1) rather than the selection line, but this also cannot explain the upward shift in bristle number, because the shift is observed in females aswellas males (albeit the shifts are slightly greater in males in all cases, possibly reflecting Y-chromosome effects). An alternative possibility is that the upward shifts of the mean in these lines were caused at least in part by fourth chromosome effects. The fourth chromosome was not controlled in these crosses, and the extractedsublines would havehad each fourth chromosome derived from one of the balancer stocks with independent probability 3 of 4, assuming no selection. Crosses between selection lines: To determine the degree to which replicate selection lines had accumu-
the mean of the extracted homozygotes shows a large shift toward the unselected value, and the phenotypic variances of the extracted homozygotes are dramatically lower than those of the selection lines. No other lines show a similar decrease in the phenotypic variance between the selection line and theextracted homozygotes. One curious result of the data in Table 7 is that the mean bristle number of extracted homozygotes from two lines not showing high frequencies of particular lethals, low abdominal replicates 1 and 2, alsoshow large shifts toward the unselected values. This result is not likely to be explainedby elimination of lethals with heterozygous effects decreasing bristle number, because the frequencies of lethals in these lines were low (Table 5 ) and the lines do not show the decrease in phenotypic variance upon making chromosomes homo-
TABLE 6 Heterozygous effects of lethal-bearing second and third chromosomes extracted from the selection lines
Lethals
Line Chromosome 2 lethals LA3 HS1 23.65 HS3 Chromosome 3 lethals LA2 LA3 LA3 23.41 HS2 22.73 HS3
P 14.88 ? 0.13 (2) 22.23 t 0.23 (2) t 0.33 (3)
13.87 ? 0.34 (13) 22.16 t 0.20 (14) 22.29 ? 0.41 (6)
0.86 0.46 0.04
14.60 i. 0.80 15.14 t 0.61 14.64 i. 0.77 t 0.72 5 0.34
13.81 t 0.39 (19) 13.87 2 0.34 (13) 13.87 t 0.34 (13) 23.32 ? 0.79 (6) 22.29 ? 0.41 (6)
0.73 0.97 0.84 0.47 0.21
(2) (6) (4) (4) (7)
Values are mean bristle numberof Cy/+;Tb/+ flies, averaged overB and B+ genotypes, sexes and extracted sublines 2 the SE among extracted sublines, with number of sublines in parentheses. Sublines bearing noncomplementing lethals (ie., lethals found at least twice in the chromosomes extracted from the given selection line) are comparedwith sublines not bearing lethals.P gives the result of one-tailed t tests for the hypothesis that the heterozygotes for lethal-bearing chromosomes had more extreme bristle number, in the direction that the line was selected, than heterozygotes for nonlethal-bearing chromosomes.
J. D. Fry, K. A. deRonde and
1302
T. F. C. Mackay
TABLE 7 Means and variances for bristle number in the selected lines compared with means and variances of sublines homozygous for chromosomes extracted from the same line
~
HA1 HA2 HA3 LA1
LA2 LA3 HSl HS2 HS3 LSl LS2 LS3
15 15 14 17 19 11 14 5 6 14 16 15
17.9 18.0 17.5 3.1 5.2 8.2 20.4 23.1 22.6 14.1 11.5 13.9
17.7 16.6 15.8 8.3 10.8 12.9 19.0 18.8 21.5 14.2 10.5 14.7
2.83 3.23 2.73 3.20 6.36 13.22 3.83 7.18 3.66 1.48 1.48 1.22
3.04 2.19 3.39 5.66 5.36 6.52 2.83 1.38 6.07 1.30 0.99 1.61
3.70 3.30 3.16 2.70 6.29 10.76 2.55 9.49 3.79 0.76 1.43 0.95
18.9 19.0 16.9 7.1 10.0 13.5 18.9 19.7 22.8 14.8 11.3 15.2
18.2 19.3 16.7 3.2 4.8 9.2 19.6 23.0 23.1 14.4 12.3 14.1
4.75 3.42 4.23 5.55 7.43 5.09 2.04 1.80 5.07 1.32 3.54 2.1 1
N is the number of sublines homozygous for single chromosomes 1, 2 and 3 derived from the selection line; ~ l n , - l f l , is the mean bristle number in the selection line from generations 102 through 105; V,f12-lfl, is the average phenotypic variance during the same period; X, is the mean bristle number of the homozygous sublines; VH is the average phenotypic variance of the homozygoussublines.Generation 102-105 meansandvariancesarebased on 20 flies of each sex scoredper generation; homozygote means and variances are based on 5 flies of each sex scored per subline.
lated mutations affecting bristle number at different loci, reciprocal crosses were made between all pairs of lines selected in the same direction for the same trait, and directional selection was continued for nine further generations from the F2 of each cross. Under an additive model with different neutral mutations fixed in each replicate line,the final response of the cross should be equal to the sum of the parental responses. A variety of factors could confoundthis prediction, however, including epistatic interactions between bristle mutations and limitation of the selection responses of the crosses by deleterious effects of the bristle mutations, finite population size or simply insufficient time. The selection responses of the synthetic lines, as well as those of the pure lines during the same period, are shown in Figures 6 and 7. Given that the original unselected population had a mean of -15 abdominal and 16 sternopleural bristles ( MACKAY et al. 1994),it is clear that in no case did the response of a synthetic line approach that expected from the sum of the parental responses. In most cases, the response of the synthetic line after nine generations did not even exceed that of the more extreme parental line. Despite the lack of additional response, the realized heritabilities from response to selection in thesynthetic lines were in most cases greater than realized heritabilities from response to selection in each replicate selection line during the same period. Realized heritability from response to selection from generations 93 to 103 averaged over allthree replicate high abdominal bristle lines was 0.005 ( MACKAY et al. 1994), compared with 0.05 averaged over the six synthetics (Table 8 ) . For the low abdominal bristle number lines, the analogous
I
"
.-In
F
l
F
2
1
2
3
4
5
6
7
8
9
I"
C .-
5
n
10
-
86 -
42-
o
!
l F
I
l
,
l
F
2
1
l 2
l 3
l 4
l 5
, 6
, 7
l 8
l
9
Generation
FIGURE6.-Selection responses of lines formedby crossing the lines selectedfor increased or decreased numbersof abdominal bristles at generation94 and continuingto select in the same direction as the parentlines. (Top) Upward-selected lines; (bottom) downward-selected lines. The pure selection lines are represented by solid symbols and the synthetic linesby open symbols. Amongthe pure lines, replicates 1, 2 and 3 are identified by circles and solid lines, squares and dashed lines, and triangles and dotted lines, respectively. a synthetic line are identified The female and male parents of by the symbol type and line type, respectively. TheF, generation in the synthetic lineswas contemporaneouswith generation 95 of selection in the pure lines ( MACKAY et al. 1994).
heritabilities are -0.004 from direct selection within lines and 0.12 from response of the crosses, and for high sternopleural bristle number lines heritability
Mutation Polygenic
24 -
&
22-
LI
E
2 2
20-
15
3
a, -
14
E
a,
13 12
F
l
F
2
1
2
3
4
5
6
7
8
9
Generation
FIGURE7.-Selection responses of lines formedby crossing the lines selected forincreased or decreased numbers of sternopleural bristles at generation 94 and continuing to select in the same directionastheparentlines.Symbolsareas in
Figure 6. from selection within lines averaged 0.06 compared with 0.13 from the crosses. However, averageheritabilities from direct selection within lines (0.08) and from crosses (0.07) of low sternopleural bristle number lines were similar. The synthetic lines formed by crossing low sternopleural replicates 1 and 3 were exceptional in showing no evidence of response to continued selection in either reciprocal cross (Figure 7, Table 8). The combination of greater realized heritabilities in a synthetic line than in the parentallines, along with the failure of the synthetic line to show a greater selection response than the more extreme parent, would be expected if one parental line was fixed for a subset of the mutations carried by the other. Alternatively, it is possible that the different lines are fixed for nonallelic mutations, but these either show duplicate-type interactions or are closely linked and failed to recombine during the nine generations of continued selection. It is also possible that insufficient time had elapsed for the responses of the selected lines to exceed those of the parents. DISCUSSION
Replicate selection lines derived from an inbredbase population have diverged by an average of 12 abdominal and 8 sternopleural bristles from the accumulation of new mutations over 125 generations (MACKAY et al. 1994). For each of the 12 selection lines, the following genetic analyses were performed. (1) The number of effective factors contributing to divergence from the inbred base population was estimated using WRIGHT’S method (WRIGHT1968; ZENG et al. 1990). ( 2 ) Generation means analysis ( MATHER and JINKS 1971) of paren-
in Drosophila
1303
tal, reciprocal F1 and F2 and backcross generations was used to estimate additive autosomal, X-linked, Flinked and cytoplasmic effects, as wellas average degree of dominance of autosomal and X-linked loci causing divergence of the selection lines from the base inbred. ( 3 ) Chromosomes were extracted from the selection lines and their main effects and interaction effects on bristle number estimated. Lethal-bearing chromosomes were sorted into complementation groups and heterozygous effects of high frequency lethal chromosomes on bristle number were estimated. ( 4 ) Replicate lines selected in the same direction for the same trait were crossed and selection continued for nine further generations from the FP.These analyses give information on numbers, chromosomal locations, departures from additivity, pleiotropic effects on fitness and allelism of new spontaneous mutations affecting bristle number. Genetic properties of spontaneous bristlemutations Numbers of effectivefactors: The numbers of effective factors differentiating the lines selected for high abdominal bristle number from the base population inbred averaged less than one. However, the low divergence between the parental lines (only 1.3 bristles, on average) coupled with the appreciable heritabilities in the F2 of the crosses suggests that the assumption that the parental lines were fixed for either all increasing or all decreasing alleles may have been violated for these crosses. For the other selection lines, those with the least response averaged three factors, whereas estimates of ne for lines with larger responses ranged from 5 to 8. Because ne is likely to be downwardly biased ( ZENG et al. 1990), these estimates can be viewedas rough minimum estimates of the number of loci differentiating the lines. Chromosomal locations: Mutations affecting bristle number have accumulated on all major chromosomes, and possibly also on the small fourth chromosome in some lines. The results of both generationmeans analysis and chromosome extraction show that the largest effects for both traits are autosomal. Autosomal mutations affecting abdominal bristle number were on both chromosomes 2 and 3. Autosomal mutations affecting sternopleural bristle number occurredon chromosome 3,but no evidence for such mutations on chromosome 2 was found. It is possible that this reflects the limitations of the method used, however (see MATERIALS AND METHODS). For example, chromosome 2 mutations may have contributed to the responses of the sternopleural lines, but if these were either dominant or interacted in particular ways withthe balancer chromosomes, their effects would not have been detected. Chromosome 4 was ignored when chromosomes were extracted from the selection lines. Sublines from all three low abdominal bristle selection lines that were isogenic for the X, second and third chromosomes had
J. D. Fry, K A. deRonde and
1304
C.T. F.
Mackay
TABLE 8 Selection responsesof lines formed by crossing the selected linesat generation 94 and continuing to select in the same direction as the parent lines
linesSternopleurallines Abdominal Cross VP hLML(C. L.) (female X male)
VP
hLML
(C. L.)
High 1 2 1 3 2 3
x x x x x x
2 1 3 1 3 2
3.77 3.53 3.05 2.86 3.60 3.03
x x x x x x
2 1 3 1 3 2
4.88 4.17 8.46 8.31 12.90 7.26
(0.08, 0.188.90 0.02 (-0.03, 0.13) 0.35) 8.47 0.04 (-0.03, 0.18) (0.05, 0.135.77 0.02 (-0.04, 0.13) 0.27) 4.71 0.07 (-0.01, 0.21) 8.66 0.07 (0.00, 0.20) 7.05 0.08 (0.00, 0.22)
0.42) (0.14, 0.25 0.26) (0.03, 0.11 0.050.17) (-0.01, 0.06 (0.00, 0.18)
Low 1 2 1 3 2 3
0.18 (0.07, 0.35) -0.07 (-0.11, 0.03) 0.17 (0.08, 0.32) 0.21(0.11,0.36) 0.03 (-0.02, 0.14) 0.18 (0.08, 0.35)
0.071.31 1.43 0.14) (-0.06, 0.010.94 0.12) (-0.07, -0.01 0.97 0.041.20 0.251.48
(-0.01, 0.19) 0.06 (-0.01, 0.21)
(-0.03, 0.19) 0.44) (0.12,
Phenotypic variance (V,) is calculated averaged over the nine generations of selection. Other symbols are
as in Tabie 1. higher mean abdominalbristle numbers than theoriginal selection lines (by 5 bristles, on average); in only one case could this difference be attributed to a lethal in the selection line, suggesting the possibility that the other two lines had accumulated fourth chromosome mutations with large abdominal bristle effects.Two other possible explanations for the shifts in the means of the extracted sublines are cytoplasmic effects (because theextracted sublines derived their cytoplasm from one of the balancerstocks) and double crossovers with the balancer chromosomes, which are known to take place at an enhanced rate when three balancers are used simultaneously. The generation means analysis gave no evidence for cytoplasmic effects in the abdominal lines, however. Double crossovers withthe balancers could conceivably explain the shifts if a single locus with a large effect on bristle number was present in a region having a high rate of double crossovers. The chromosomal analysis implicated mutations at X-linked lociin the high-low divergence of the abdominal bristle selection lines, but the generation means analysis only detected additive X-linked effects in both sexes in one low abdominal line.The latter analysis may have less powerto detectX-linked effectsin segregating generations if they have deleterious fitness effects, however. The chromosomal analysis gave no evidence that X-linked mutations were involved in the high-low divergence of the sternopleural lines, but in this case the results may have been confounded by the use of the Busc balancer, which contains alleles of the achuete-scute complex with large positive effects on bristle number. In contrast,thegenerationmeans analysisgaveevidence for X-linked mutations in one low sternopleural replicate. Finally,Y-linked mutations of about -1 to
-3 bristles occurred in all three low abdominal bristle selection replicates and of about 1 bristle in one high abdominal replicate. That the significant Y-linked effects are not artifacts of violation of any of the assumptions of the generation means analysis is attested to by the results presented in Figure 2, in which males differing only in the origin of their Yare compared. The results of the chromosomal analyses generally agree with the conclusions from WRIGHT’Smethod in giving evidence that two or more mutations contributed to the responses of most lines.This result, and the occurrence of X- and Y-linked mutations in several of the lines, contrasts with the conclusions of CABALLERO et al. (1991 ) and LOPEZand LOPEZ-FANJUL (1993a,b), who selected an isogenic line for increased and decreased numbers of abdominal bristles. These authors concluded from thepattern of selection responses thatthe responses in most lines were caused by a single mutation, and comparing the means of reciprocal F1 males gave no evidence for sex-linked effects.These differences between our results and theirs could have a variety of explanations, including the shorter duration of their experiment (47generations) and differences in thebase stocks used. It is alsopossible that chromosomal analysis or other biometric analyses would haveproduced evidence for additional mutations in some of their lines. Departures from additiviq Although the generation means analysis did not allow formal rejection of the hypothesis that all new mutations affecting bristle number areadditive in theireffects, there was a tendency for estimates of summed autosomal dominance deviations ( [ h a ]) to be of opposite sign to the direction of selection. Although the [ h a ]estimates do not give information about the degreeof dominance of individual loci,
+
in
Mutation Polygenic
they do suggest that some mutations affecting bristle number were partially recessive, for the following reason. If we assign values - d, hand d to the nonmutant homozygote, heterozygote and mutant homozygote at a locus, then the degreeof dominance k of a mutation is defined as h/d, with a negative value of k indicating a departure from additivity in thedirection of recessivity. Because [ha] is defined as the sum of h valuesover loci, it can be written as X a dl, where the sum is taken over loci. Formutations that were fixed in the selection lines, the dl will mostly be of the same sign as the direction of selection; therefore [ha]values that are opposite in sign to the direction of selection imply that some k are negative. Furthermore, loci with large d will have the largest influence on [ha]; therefore one interpretation for the tendency of [ ha] estimates to be of opposite sign to the direction of selection is that mutations of large effect on bristle number areoften partly recessive. This would be consistent with previous work on both spontaneous (SANTIAGO et al. 1992; LOPEZand LOPEZFANJUL 199313)and Pelement-induced (MACKAY et al. 1992) mutations affecting abdominal and sternopleural bristle number. The chromosomal analysis indicated that the selection lines showed significantly different patterns of interchromosomal interaction from each other, with two lines in each of the abdominallines and the sternopleural lines showing stronger interactions thanthe remaining four lines. One interpretation of this result is that the two lines showing stronger interactions had accumulated mutations that interacted with each other or with genes on thebalancer chromosomes. More indirect evidence for interactions between bristle mutations comes from the lack of fit shown by many lines to the model used for the generation means analysis, which assumes absence of epistasis; however, the lack of fit could also be explained by differential survival ofgenotypes in the F2 and backcross generations. Pleiotropiceffects on fitness: Two lines (low abdominal replicate 3 and high sternopleural replicate 2 ) apparently each contained a segregating lethal with large heterozygous effects on bristle number. This conclusion is based on three pieces of evidence. ( 1) Each line had extremely high phenotypic variance and showed rapid reversion of response and decline in phenotypic variance when selection was relaxed ( MACKAY et al. 1994). ( 2 ) Each line showed a large shift in the mean toward the unselected value and a large decline in phenotypic variance when chromosomes from the line were made homozygous. ( 3 ) Each line had one or more segregating lethals in relatively high frequency. A fourth piece of evidence was contradictory: the lethal chromosomes extracted from the lines did not have detectable bristle effects when made heterozygous against balancer chromosomes. We believe that the most plausible explanation for this set of results is that a lethal in each of the lines had a heterozygous effect on the selected trait in
+
Drosophila
1305
the genetic background of the lines, but these effects were not expressed in the different genetic background provided by the balancers. Any alternative explanation must be able to account for observations ( 1) and ( 2 ) above. Oneother line (high sternopleural replicate 3) showed a high frequency of lethals, and one of these lethals showed an apparent heterozygous effect of about 1.4 bristles when tested against the balancer (Table 6 ) . This line showed no otherbehavior that one might expect of a line carrying a lethal with heterozygous effects on the selected trait, however, such as a decline in phenotypic variance when chromosomes were made homozygous or reversion of selection response upon relaxation of selection (MACKAY et al. 1994). Thefailure to observe these signatures of a lethal in this selection line could reflect the relatively small effect of the lethal in question; alternatively, the apparent effect of the lethal mayhave been caused by a chance association withselected mutations or simply by a statistical type I error. LOPEZand LOPEZ-FANJUL ( 1993a,b) found thatof 25 lines derived from an isogenic base that responded to selection for increased or decreased numbers of abdominal bristles, a lethal contributed to the response in seven lines. If the above conclusion regarding the presence of a lethal with heterozygous effects on abdominal bristles in low abdominal replicate 3 isaccepted, then alethal contributed to the response in one of six of our abdominal bristle selection lines. These proportions are not significantly different. Allelism of bristle mutations: Evidence that spontaneous mutations arising in replicate lines selected in the same direction for the same trait occur at different loci is mixed. From the generation means analysis, no two replicate lines have the same pattern of significant additive autosomal, additiveX-linked or autosomal dominance effects, suggesting that the replicate lines vary at some loci. However,four of sixabdominal bristle number lines have accumulated Ylinked mutations, which, given the small size of this chromosome, could well be at the same locus. Synthetic lines formed by crossing one pair of lines selected for low numbers of sternopleural bristles did not show any detectable response to continued selection, suggesting that these lines may have been fixed for mutations at the same loci. However, this result could also be explained by epistatic interactions among bristle mutations contributed by the different lines. In most other cases, synthetic lines showed more rapid responses to selection than the parent lines, as measured by realized heritabilities, but still did not exceed the more extreme parent after nine generations. This result could have a variety of explanations, including the selection lines having mutations at overlapping sets of loci, epistatic interactions or simply insufficient time for greater responses to have occurred. Regardlessof theexplanation, our results
+
1306
J. D. Fry, K. A. deRonde and T. F. C. Mackay
provide an interesting contrastwith those of CABALLERO et al. (1991), who found that synthetic lines formed by crossing selected lines derived from an isogenic base clearly exceeded the response of the parent lines after 6-15 generations of continued selection for low abdominal bristle number. Finally, we note that interline contamination is not a likely explanation for the presence of allelicmutations in different lines in our experiment; each linehas a unique patternof new P-element insertions ( MACKAY et al. 1994), suggesting that nosuch contamination occurred. We cannot rule out the possibility that some loci affecting bristle number have high mutation rates in these lines (including especially a locus or loci on the Y; see below). The initial base population was an inbred subline of Harwich, a strong Pstrain in the P-M hybrid dysgenesis system. Many P-element transpositions and excisions have occurred in the selection lines derived from it (MACKAY et al. 1994). It is not inconceivable that Pelements in the inbredHanvich strain situated by chance near loci affecting bristle number have induced mutations at these loci at a high rate by local transpositions and excisions. Nonetheless, it should not be assumed that the strong P nature of the strain used in this experiment is responsible for anyof the results reported here.
The nature of loci affecting bristle number FRANKHAM (1980) selected replicate lines for high and low abdominal bristle number for 100 generations from an inbred base population. The pattern of selection response was not unlike that observed for theselection lines studied here, with greater response to selection forreducedthanfor increased numbers of abdominal bristles. Genetic analysesof FRANKHAM'S lines showed selection responses in independent replicates were due to deletions of tandem repeats at the rDNA(bobbed, bb) locus on the X chromosome; males did not show the bb phenotype because the Y chromosome had wild-type rDNA copy number ( FRANKHAM et al. 1978;FRANKHAM 1980). X-chromosome bbmutations were caused by nonhomologous recombination at the bb loci on the X and Y chromosomes ( GILLINGS et al. 1987; FRANKHAM 1988). A reasonable explanation for the Y-linked mutations observed in all three low and one high abdominal bristle number selection lines reported here is that these mutations are also at the bb locus on the Y chromosome. It is not likely that these mutationsarose by nonhomologousrecombination with the X chromosome, because such events generate lethal Ychromosome deficiencies. Nonetheless, the fact that Y mutations arose in four of the six abdominal lines suggests that some mechanism generates a relatively high rate of Flinked mutations with effects on bristle number in these lines (whether atbobbed or some other locus).Even if we make the conservative assump-
tion that Y-linked mutations aroseonly once in the lines where Y effects were observed, the estimated mutation rate is ( 4 mutations)/ ( (31 males scored per generation, on average) (113 generations of selection) ( 6 abdominal lines)) or 1.9 X 10 p 4 . This is comparable with the frequency with which X-Y exchanges occur ( FRANKHAM et al. 1980; MADDERN 1981) . Mutations at other bristle loci knownfrom major mutations affecting bristle number may have occurred in the selection lines. Many ofthese loci havefundamental roles in nervous system and sense organ (bristles are adult sense organs) development and manifest strong phenotypic and molecular interactions (reviewed by CAMPOS-ORTEGA 1993;JANand JAN1993). Preliminary deficiency mapping indicates high sternopleural bristle replicates 2 and 3 both have mutations at the extramacrochaetae (emc) locus onchromosome 3 (T. F. C. MACKAY, unpublished data). Loci affecting bristle development are found on all major chromosome arms, so interactions of spontaneous mutations at these loci could explain the observed interchromosomal epistatic effects. This hypothesis can be tested by complementing the selection lines with mutations and deficiencies of candidate bristle loci. This work was supported by grants GM-45344 and GM-45146 from the National Institutes of Health.
LITERATURE CITED BARTON,N. H., 1990 Pleiotropic models of quantitative variation. Genetics 124: 773-782. CABALLERO,A,,M. A. TORO and C. L~PEZ-FANJUI., 1991 The response to artificial selection from new mutations in Drosophila mdanogaster. Genetics 128: 89-102. CAMPO.S-~RTF:'E(:A, J. A,, 1993 Early neurogenesis in Drosophila melanogaster, pp. 1091- 1129 in The Deoelopment of Drosophila melanogastm, Vol. 2, edited by M. BATE and A. MARTINEZ ARIAS. Cold Spring Harbor Laboratory Press, Plainview, N Y . COCKERHAM, C. C., and H. TACHIDA, 1987 Evolution and maintenance of quantitative genetic variation by mutations. Proc. Natl. Acad. Sci. USA 8 4 : 6205-6209. FAI.CONER, D. S., 1989 Introduction to Quantitative Genetics. Longman, New York. FRANKHAM, R.,1980 Origin of genetic variation in selection lines, pp. 56-68 in Selection Expm'ments in I,aboruatq and Donmtic Animals, edited by A. ROBERTSON. Commonwealth Agricultural Bureaux, Slough, UK. FRANKHAM, R., 1988 Exchanges in the rRNA multigene family as a source of genetic variation, pp. 236-242 in Proceedings o f t h e 2nd International Conferenre on Quuantitatiue Genetics, edited by B. S. WEIR,E. J. EISEN,M. M. GOODMAN and G. NAMKOONG. Sinauer, Sunderland, MA. FRANKHAM, R., D. A. BRISCOEand R. K. NURTHEN,1978 Unequal crossing over at therRNA locus as a source of quantitative genetic variation. Nature 272: 80-81. FRANKHAM, R., D. A. BRISCOEand R. K. NURTHEN, 1980 Unequal crossing over at the rRNA tandon as a source of quantitative genetic variation in Drosophila. Genetics 95: 727-742. GAVRIIETS, S., and G. DE JONG, 1993 Pleiotropic models ofpolygenic variation, stabilizing selection and epistasis. Genetics 134 609625. GII.I.INGS,M. R., R. FRANKHAM, J. SPEIRSand M. WHAI.I.EY, 1987 xYexchange and coevolution of Xand YrDNAarrays in Drosophila melanogaster. Genetics 1 1 6 241-251. HILI,,W. G., and P. D.KEIGHTLEY, 1988 Interrelations of mutation, population size, artificial and naturalselection, pp. 57-70 in
Polygenic Mutation in Drosophila Proceedings of the 2nd International Conference on Quantitativr Genetand G. ics, edited by B. S. WEIR,E. J. EISEN,M. M. GOODMAN NAMKOONG. Sinauer, Sunderland, MA. JAN, Y. N., and L. Y. JAN,1993 The peripheral nervous system, pp. 1207-1244 in The Development of Drosophila melanogaster, Vol. 2, edited by M. BATEand A. MARTINEZ ARIAS. Cold Spring Harbor Laboratory Press, Plainview, NY. KEIGHTLEY, P. D., and W.G. HILI., 1990 Variation maintained in quantitative traits with mutation-selectionbalance:pleiotropic sideeffects on fitness traits. Proc. R. Soc. Lond. B 2 4 2 95-100. KONDRASHOV, A. S., and M. TURELLI,1992 Deleteriousmutations, apparent stabilizing selection and the maintenance of quantitative variation. Genetics 132: 603-618. LANDE, R., 1975 The maintenance of genetic variability by mutation in a polygenic character with linked loci. Genet. Res. 26: 221235. LANDF., R., 1980 The genetic covariance between characters maintained by pleiotropic mutation. Genetics 94: 203-215. LANDE,R., 1981 The minimum number of genes contributing to quantitative variation between and within populations. Genetics 99: 541-553. LINDSLEY, D. L., and G. G. ZIMM,1992 The Genome of Drosophila melanogaster. Academic Press, San Diego, CA. LOPEZ,M., and C. LOPEZ-FANJUI., 1993a Spontaneous mutation for a quantitative trait in Drosophila melanogaster. I. Response to artificial selection. Genet. Res. 61: 107-116. L ~ P E ZM., , and C. LOPEZ-FANJUL., 199313 Spontaneous mutation for a quantitative trait in Drosophila melanogaster. 11. Distribution of mutant effects on the trait and fitness. Genet. Res. 61: 117-126. LYNCH,M.,and W. G. HILI., 1986 Phenotypic evolution by neutral mutation. Evolution 40: 915-935.
1307
T. F. C., R. F. LYMAN and M. S. JACKSON, 1992 Effects of P-element insertionson quantitative trait?in Drosophila melanogaster. Genetics 130: 315-332. MACKAY, T. F. C., J. D. FRY,R. F. LYMAN and S. V. NUZHDIN, 1994 Polygenic mutation in Drosophila mlanogaster: estimates from response to selection of inbred strains. Genetics 136: 937-951. MADDERN, R. H., 1981 Exchange between the ribosomal RNA genes of the X and Y chromosomes of Drosophila melanogaster. Genet. Res. 38: 1-7. MATHER, K., andJ. L.JINKS,1971 Biometrical Genetics. Cornell University Press, Ithaca, W. NETER,J., W.WASSERMANand M. H. KUTNER,1990 Applied Linear Statistical Models. Irwin, Homewood, IL. E., J. ALBORNOZ, A. DOMINGUEZ, M. A. TOROand C. L ~ P E Z SANTIAGO, FANJUL,1992 The distribution of mutations on quantitative traits and fitness in Drosophilamelanogaster. Genetics 132: 771781. TUREI.I.I, M., 1984 Heritable genetic variation via mutation-selection balance: Lerch’s zeta meets the abdominal bristle. Theor. Popul. Biol. 25: 138-193. TUREI.I.I,M., 1985 Effects of pleiotropy on predictions concerning mutation-selection balancefor polygenic traits. Genetics 11 1: 165-195. WRIGHT,S., 1968 Evolution and the Genetics of Populations. Vol. 1. Genetic and Biometric Foundations. University of Chicago Press, Chicago. C. C. COCKERHAM, 1990 How informative ZENG,Z.-B., D. HOULE and is Wright’s estimator of the number of genes affecting a quantitative character? Genetics 126: 235-247. MACKAY,
Communicating editor: P. D.
KEIGHTI.EY
