Canalisation of Development of Scutellar Bristles in Drosophila by Control of the Scute Locus

CANALISATION OF DEVELOPMENT OF SCUTELLAR BRISTLES IN DROSOPHILA BY CONTROL OF THE SCUTE LOCUS

J. M. RENDEL, B. L. SHELDON AND D. E. FINLAY C.S.I.R.O., Diuision of Animal Genetics, North Ryde, New South Wales, Australia Received June 3, 1965

sex-linked mutant scute (sc), its noma1 allele, and a number of minor Tf:nes which can be shown to exist by selective breeding, form a developmental system which determines the number of bristles on the scutellum of Dro- sophila mlanogaster. In unselected stocks almost all flies grow four bristles when one or two + genes are present at the scute locus; that is, development is canalised at this level of phenotype. The average number of bristles in sc males in unselected stocks is less than one and in sc/sc females more than one, but there is considerable variation about these means. In stocks segregating for sc and +, selected for high bristle number, + flies may grow a mean of seven or eight bristles. There is considerable variation about these high means also. In high bristle stocks, sc males and sc/sc females average between three and four bristles, most individuals having four bristles but some having three, a few two and a very few having less than two. The scutellar phenotype is controlled at four bristles. If bristle number does not reach as high as four, it varies about some mean lower than four and is not controlled. If bristle number does not fall as low as four it varies about some mean higher than four and is not controlled. But a wide variety of genotypes, despite widely different en~ronmen~alconditi~ns, are con~olledto a level of exactly four bristles. If we write P = (p (M)where P is the phenotype, 111 is Make or that which forms or tends to form bristles and 9 is a function relating P and M, control at four bristles operates over a range of M values equivalent in extent to 60 of M.# in the scutellar system can be represented by a line for which the horizontal axis is M, and the vertical axis is P, with a slope of 45" to a plateau 6aM long followed by a slope of 45". P is 4 at a value of M anywhere under the plateau. Figure 1 shows such a curve relating P to 111. A similar curve based on fewer points was described by an equation containing six terrns in M (RENDEL1959b). Fuller discussion can be found elsewhere (RENDEL1962). M receives contributions from the genes at the scute locus and minor genes scattered over the genome. There is evidence suggesting that the major gene at the scute locus is controlled. A comparison of sc/sc, sc/-i- and +/+ shows that at all levels of gene background the difference between sc/sc and sc/+ is much greater than the difference between sc/+ and +/+. The second + gene contributes much less than the first and is under some control so that in the presence of +- it is less active than in the presence of sc. The suggestion is that the activity of the + gene is under negative feedback control. The more product the less activity until the gene at the scute locus is turned right off.

Genetics 52 : 1137-1 151 Decenther 1965. 1138 J. M. RENDEL et al.

0,l 1.2 2,3 3,d 4.5 5.6 PROBIT SCALE OF SCUTELLAR MAKE VALUES SHOWING BRISTLE THRESHOLDS

1 DIV I 1 STANDARD DEVIATION OF MAKE 1 FIGURE1.-The relationship between number of scutellar bristles and scutellar Make.

Four questions immediately arise: (1 ) Is the product which turns the gene off the product of the scute locus or the product of the scutellar developmental system as a whole? (2) Are the minor genes controlled? (3) Are the sc mutants controlled? (4)Why, if + is controlled, is s~/+ different from +/+? The experiment reported here was designed to provide evidence on these points. A careful examination of the effect of adding + or sc at different levels of the system as a whole should show us whether control of the gene at the scute locus is brought about through action of the product of the gene at this locus only, or by the product of all genes in the scutellar developmental system. In the former case, an addition would have an effect depending on the major gene to which it was added and be independent of the genetic background, high, medium or low; in the latter the effect of an addition would depend on both.

MATERIALS AND METHODS Six levels were chosen; + and sc1 were kept segregating in three lines, a high line, an unselected line (SW) and a low line. An extra sc or + could be added to a male carrying -t or sc in lines at three levels. The extra gene was added on a Y chromosome. The Y's were kindly given to us by PROFESSORM. M. GREEN who has described their manufacture (FRASERand GREEN 1964). He sent us three stocks containing translocated Y chromosomes. On one Y a few bands of the X containing the scute locus, the yellow locus, the achaete locus but not the white locus, carried the sc1 allele at the scute locus. The other two both carried a scf at the scute locus, one with y594 the other with y* at the yellow locus. We were thus able to compare + and sc males with sc/sc males, sc/+ males and +/+ males at three different levels of bristle number by backcrossing the SOY or sc+.Y chromosome into our three stocks. Males carrying the wanted Y chromosome were crossed with females from the stock into which the Y was to be introduced. Sons were again crossed to females from stock; sons were again taken and crossed to stock females until at least five cros.ses had been made by which time bristle numbers of females of the back- cross line were typical of females from the stock. All sons, with the exception of some nondisjunctional rarities, which can be recognized and ignored, will carry a Y and an X both having a scute locus and so will be diploid for this locus. The addition of + or sc to + or sc is then immediately mesaurable. The effect of substituting one or two + genes for one or both sc genes in females was also measurable at the three different levels. CONTROL OF DEVELOPMENT 1139

In addition to this set of comparisons, the translocation Y chromosomes were crossed into an attached-X line in order to see what +/+/+ and +/+/sc were like. Some sc/sc/sc females and males were also obtained as a result osf nondisjunction.

RESULTS

Tables 1 and 2 summarise most of the results as far as males are concerned. Raw distributions of bristle number in Tabk 1 have been converted into probits which are shown in brackets and the differences between different genotypes in probits have been tabulated in Table 2. Table 2 also shows the mean level of males of each line. The probit distance between the 0/1 cut-off, i.e. the threshold between 0 bristles and 1 bristle, and the 3/4 cut-off is fairly consistently 3.5~. The distance between the 3/4 and the 4/5 cut-off can be seen from Table 1 for a number of populations. The distance is not constant. It is 6u in -t males of the low line but decreases steadily as the mean of the population changes until in sc of the high line it is only about 30. There is a break between the 3/4 and the 4/5 cut-offs which we shall discuss later. For the purpose of drawing the graph in Figure 2 we have taken the 0/1,1/2 and 2/3 cut-offs as fixed points. The distances between these points are constant. All populations which can be related to these points have been so related. The width of the 4 class has been taken to be 5.8~ but the lines above the 3/4 cut-off have been related to the mean of the low line + population. The mean of this population is placed on the scale relative to the 3/4 cut-off and is taken to be 2.920 above this point. It is thus placed 2.77~above the sc population of the high line. The + low population is 3.12~below the 4/5 cut-off and the sc high population is 2.89 below the 4/5 cut-off so that if the 4/5 cut-off had been used to place the + low population it would have placed it just below the mean of the sc high population. There is a difference between the results of using the 3/5 and 4/5 cut-offs, because in sc high the lwidth of the 4 class

+ 10- 0:0-

0 , , I,.! ,',I, , , , , , , , 1 sc/. scl.. sc/. +/.+>. +>. LOW sw HIGH LDW SW HIGH RELATIVE LEVELS OF MAKE (M) OF BASE MALE GENOTYPES ( 1 DIV - 1 STANOARO DEVIATION OF MAKE ) FIGURE2.-The effect of adding an extra sc or + gene at different levels on the probit scale of Make values in males. TABLE 1

Frequency distributions of scutellar bristle number, and corresponding probit values (in brackets), in male genotypes of the main experiment

Bristle number

~~ Male genotype 0 1 2 3 4 5 6 7 8 9 10 11 19 13 14 Low line sc/y 6738 26 ...... (7.67) ...... sc/sc.Y 59 214 1085 478 122 ...... (3.12) (3.92) (5.51) (6.54) ...... sc/y~gbsc+.Y ...... 4 1346 4 I ...... (2.25) (7.68) (8.19) ...... sc/y"cf.Y ...... 5 2208 33 ...... (2.15) (7.18) ...... sc+/Y ...... 14 8185 8 ...... (2.07) (8.09) ...... sc+/sc.Y ...... 2 924 7 ...... (2.14) (7.43) ...... scf/y59bsc+.Y ...... 1290 137 6 ...... (6.28) (7.64) ...... sc+/y"cf .Y ...... 2 2122 185 10 ...... (1.87) (6.38) (7.63) ...... Unselected (SW) line sc/Y 1124 343 125 5 ...... (5.54) (6.W (7.74) ...... sc/sc.Y 24 83 816 1129 1374 ...... (2.54) (3.14) (4.39) (5.25) ...... sc/y59bsc+.Y .. 1 2 4052 58 1 ...... (1.81) (7.19) (8.54) ...... sc/y*sc+ .Y .. .. 1 2042 185 26 3 ...... (1.65) (6.31) (7.23) (8.01) ...... TABLE 1-(Continued)

Bristle number Male Kenotvoe 0 1 -7 3 4 5 6 7 8 9 10 11 12 13 14

sc+/Y ...... 5 4135 41 ...... (1.96) (7.33) ...... sc+ /sc.Y .. 6 3672 54 5 ...... (2.05) (7.15) (8.01) ...... sc f / y59bscf .Y .. .. 5 3360 638 125 11 ...... (1.96) (5.89) (6.84) (7.78) ...... SC+/Y"C+.Y .. .. 3 1510 725 2.31 36 2 ...... (1.96) (5.26) (6.24) (7.16) (8.16) ...... High line sc/Y 1 22 492 1341 2346 7 1 ...... (1.46) (2.46) (3.84) (4.85) (7.89) (8.54) ...... sc/sc.ll 4 68 938 37 4 ...... (2.33) (3.51) (6.76) (7.67) ...... 1 24 89 182 155 97 26 2 1 (2.07) (3.29) (4.15) (5.03) (5.78) (6.64) (7.56) (7.93) ...... 19 50 152 157 112 19 3 ...... (3.21) (3.90) (4.83) (5.64) (6.72) (7.52) ...... sc+/Y 26 122 235 138 31 5 (3.32) (4.37) (5.49) (6.52) (7.37) sc f /sc.Y ...... 5 32 83 84 71 19 4 1 ...... (2.87) (3.84) (4.75) (5.47) (6.4.0) (7.13) (7.72) sc+/y59bsc+ .Y 2 8 33 97 169 73 27 17 6 .. 1 .. (2.4.0) (3.01) (3.71) (4.54) (5.56) (6.19) (6.59) (7.14) (7.83) . . .. sc+/yzsc+ .Y 7 34 73 214 129 48 14 4 1 .. (2.79) (3.58) (4.22) (5.32) (6.14) (6.80) (7.35) (7.89) . . .. 1142 J. M. RENDEL et al.

TABLE 2

Positions of means, on the probit scale, of the base male genotypes, and amounts of Make added to them by addition of an extra sc or an extra sc+ allele

sc/y +/Y Low line 2.67~below 0/1 cut-off 2.920 above 3/4 cut-off SW line 0.54~below 0/1 cut-off 0.73~above Low f High line 0.36~above 3/4 cut-off 4.07~above SW

Make added by addition of sc.Y to scfl Make added by addition of sc.Y to f/Y Low line 4.550 0.660 SW line 3.200 0.18~ High line 1.34~ 0.78~

Make added by addition of sc+,Y to sc/Y Make added by addition of sc+.Y to +/Y ys%2+.Y y%c+.Y ys9bsci.Y y3cC.Y Low line 8.4~ 8.50 1.810 1.710 SW line 6.90 7.10 1.44u 2.070. High line 4.5u 4.7u 1.86~ 2.220 is 3u less than it is in + low. The sc high population is lower than the + low population below 4 but higher above 4. The variance of the high line is greater than that of the other two. It follows that the yard stick, uM,is underestimating distances in the high line by comparison with those in other lines in which uM is less. Correction proportional to differences in uM has been made in the high line when placing the mean of this line on the scale. The reason for differences in variance will be discussed later. The horizontal axis in Figure 2 is the level of Make of the base genotypes to which an extra gene is added. It is measured in uM. The vertical axis is the effect of adding the extra gene, also measured in uM.The effects of the two different sc+.Y translocations run close together. They add about 8 or 9 oarat the lowest sc level of M. The amount they add falls steeply to 1.7 at a level of M corresponding to + low, after which the amount they add remains constant. The effect of adding sc is similar; sc added to low sc contributes 4.5. The amount falls steadily to 0.6 at which level it stays fairly constant. There can be no doubt that at low levels of M both sc and + contribute more than they do at high levels of M, and that beyond the point where M is so high that 99% of the population have four bristles or more the contribution of the additional gene remains constant. This is interpreted as showing (1 ) that control of the scute locus is exercised by product of the developmental system as a whole, (2) that the control system cannot completely shut off or exclude the extra gene at the scute locus on the Y from contributing. The results of substituting + for sc in females are essentially similar to and are interpreted in the same way as the results of adding + genes in the males. Dis- tributions of bristle numbers of the three genotypes at the three levels and the corresponding probit values are given in Table 3. The differences in probits between the genotypes being compared are tabulated in Table 4 and shown graphically in Figure 3. CONTROL OF DEVELOPMENT 1143 1144 J. M. RENDEL et al.

s%c 0-1s%c 1.2 2.3 %c3.4 %,% 4.5%c 0 0-1 1.2 2.3 3.4 4.5 0 I1111' I1 ;Ll111' II'I' 1111 1

""/,, ~ SXC ~~ %c %c%c %c LOW sw HIGH LOW SW HIGH RELATIVE LEVELS OF MAKE (MI OF BASE FEMALE GENOTYPES

I 1 DIV i1 STANDARD DEVIATION OF MAKE ) FIGURE3.-The effect of substituting + for sc at different levels on the probit scale of Make values in females.

Data obtained from crosses of the translocation stocks to attached-X females, up to and including the Flogeneration in each case, are given in Table 5. Addi- tions of sc or + in males have similar, if rather smaller, effects to the male results already given. The addition of sc in attached-X females to give +/+/scY has a detectable effect, though smaller than the effect of its addition to + in males. Similarly the addition of + in females to give +/+/sc+.Y has a smaller effect than its addition to + in males. Comparisons involving sc/sc/scY females and sc/scY/scY males in the low background, derived from nondisjunctional events, are given in Table 6. Addition of sc to sc/sc in females has about loless effect than sc added to sc in males. The effect of addition of sc to sc/scY in males to give sc/scY/scY can be assessed only approximately, because the population of males in the 2nd Series (ii) was a mixture of sc/scY and sc/scY/scY, thoretically in equal proportions but prob- ably having a shortage of the latter genotype. Comparison of this population with the sc/scY males of 2nd Series (i), permissible because of the close agree-

TABLE 4 Positions of means, on the probit scale, of the base female genotypes, and amounts of Make added by substituting the + allele for the sc allele

sc/sc sc+/sc Low line 1.070 below 0/1 cut-off 1.980 above 3/4 cut-off SW line 0.560 above 0/1 cut-off 2.950 above 3/4 cut-off High line 3.600 above 0/1 cut-off 0.810 above 4/5 cut-off

Make added by substituting Make added by substituting + for sc in sc/sc + for sc in sc+/sc Low line 6.6~ 0.90 S W line 5.80 1.247 High line 3.3u 1.1u CONTROL OF DEVELOPMENT 1145

TABLE 5

Frequency distributions of scutellar bristle number, and corresponding probit values, in genotypes obtained from crosses of sc.Y or sc+.Y stocks to an Attached-X stock (sc+/sc+), with amounts of Make added by addition of sc or sc+ alleles

Bristle number Genotype 3 4 5 6 7 8 9 10

sc +/se + /SPY 2773 201 14 11 (6.46) (7.58) (8.43) sc+ /sc+ /Y 3 4047 223 13 (1.81) (6.60) (7.75) sc+ /scY 3 3995 60 5 .. (1.81) (7.14) (8.04) sc+/Y 8 3832 51 I (2.14) (7.21) (8.43) se f/sc f / yjQbscf.Y 1 2501 862 221 28 10 2 (1.57) (5.50) (6.46) (7.29) (7.72) (8.24) sc+/sc+ /Y 4043 371 67 3 1 (6.29) (7.15) (8.12) (8.54) sc+/y59bsc-+.Y 2 2437 628 173 20 10 1 (1.76) (5.66) (6.54) (7.35) (7.71) (8.43) sd/Y 5 3798 111 13 4 (1.99) (6.84) (7.64) (8.09) sc + /se+ y2sef .Y 1 2890 379 44 2 (1.57) (6.13) (7.20) (8.24) se+/sc+/Y 3 4095 43 (1.81) (7.31) sc f/y%f .Y 5 3594 84.9 166 6 1 ( 1.94.) (5.77) (6.78) (7.97) (8.54) sc+ /Y 4 4529 24 (1.88) (7.56)

~ Amount of Make added by addition of sc+.Y Amount of Make added by addition of SCY To sc+/sc+ To se/Y

To rc+/sc+ To sc+/y V9b r' V9b Y' 0.160 0.26O 0.80 1.20 I.& l.%J ment between the female scores in (i) and (ii), shows the mean of the mixed population to be about 0.4~higher. Hence the estimate that sc/scY/sc.Y males are about 1.OO higher than sc/scY. These results and those from attached-X crosses presented above are interpreted in the same way as the main body of male data. The general consistency of the results reinforces the conclusions already stated, namely (1) that the gene at the scute locus has less effect the higher the level of total activity of the developmental system; control of major gene activity is exercised to fix the level of the total product of the developmental system, mjor gene -k polygenic background + environment; and (2) that each addition of an extra gene at the scute locus has some effect; the major gene can never be reduced to zero activity by the control system. The differences between males and females when an extra scute locus is c- c P TABLE 6 0, Frequency distributions of scutellar bristle numbers, and corresponding probit values, in genotypes arising from nondisjunction in the Low line containing sc.Y or sc+.Y translocation, with differencesin amount of Make due to addition of an extra sc allele

Females Males Bristle number Bristle number Genotype 0 1 2 3 4 Genotype 0 1 2 3 4 1st Series sc/sc 590 160 83 2 624 36 12 .. .. (5.54) (6.27) (7.82) .. (6.47) (7.10) ...... sc/sc/sc.Y 15 35 154 33 7 12 41 156 58 17 4 (3.46) (4.18) (5.98) (6.90) .. (3.28) (4.11) (5.63) (6.56) . . 8 2nd Series id (9 sc/sc 500 288 15 1 5 .. sc/Y 820 67 15 .. .. U M (5.07) (5.97) (7.56) .. .. (6.34) (7.13) ...... r sc/sc/scY 2 41 515 131 33 sc/scY .. 23 408 175 55 2 R (2.23) (3.44) (5.75) (6.69) .. . (3.19) (5.39) (6.38) . . h (ii) Mixed population sc/scY and sc/sc 447 232 115 .. .. sc/scY/scY 8 182 133 82 (5.16) (6.06) .. .. (2.94.) (4.92) (5.83) . sc/sc/scY 1 33 377 108 23 ...... (2.09) (3.47) (5.70) (6.72) ......

Amount of Make added by addition of sc To sc/sc To sc/u To sc/sc,u 1st Series 2.w 3.1~ ... 2nd Series ...... (i) 2.k 3.9u ... (ii) 2.80 ... 1.0~(approx.) CONTROL OF DEVELOPMENT 1147 added may well be associated with the proportion of total number of X chromosomes to number of scute loci. Presumably the X chromosome carries its share of genes controlling the scute locus; the available evidence indicates that the control genotype is polygenic (SHELDON,RENDEL and FINLAY1964; RENDEL, SHELDONand FINLAY1965). Two X chromosomes might be expected to contribute more effectively to control of three genes at the scute locus, as in females, than one X chromosome to control of two genes, as in males, the extra gene in each case being in a very small bit of X chromosome attached to a Y chromosome.

DISCUSSION

The mechanism of control which may be at work is suggested by the results in this field in microorganisms. Following the summary given by JACOB and MONOD(1961), we can suppose the scute locus to be held inactive by conjugation at a critical site with a molecule which is removed when gene substrate or its analogue enters the cell. Removal of this suppressor by substrate liberates the gene which becomes active until, as the result of gene activity, suppressor once more becomes available, either through removal of substrate used in manufacture of gene product or through stimulation of suppressor manufacture elsewhere in the genome by a gene product. There will be competition between suppressor and gene substrate for the gene site so long as both substrate and suppressor are present in the cell. As the temperature sensitive period of whatever the scute gene does is perhaps some 12 hours around the 85th hour, of larval life at 25"C, the time during which the scute gene is active is likely to be taken up to a large extent by the time during which suppressor and gene substrate are present together. As long as these two substances compete, a gene at the major Iocus must have some activity and this could account for two genes always contributing more to M than one. The effect with sc present is less than the effect with + present merely because sc is less effective than +. We must assume that the minor genes either operate on the same substrate as the scute locus, which seems highly likely, or stimulate the manufacture of the same suppressor. Their contribution to M is clearly taken account of when the level at which the major gene is controlled is determined. There is no evidence that the minor genes themselves are controlled. It is true that SW-Low and High-SW, two differences due to minor genes, are smaller at the + level than at the sc level but this can be attributed to control of the gene at the scute locus which will be tighter at higher levels of gene activity. In high-line populations, in which the + genes are completely controlled out, variation is greatest, strongly suggesting that minor genes are not controlled. It is possible, though not proven, that we have here a qualitative dis- tinction between MATHER'Soligogenes and polygenes which is quite objective. Oligogenes are those genes controlled with respect to the developmental processes under consideration, polygenes are those which, if controlled at all, are controlled with respect to some other developmental process. We suggest later that some selection may decontrol the major gene and so increase expression and variability. Although this is true, the minor genes which distinguish our low, medium and 1148 J. M. RENDEL et al. high lines from each other, are not acting in this way. They are acting in a way which enhances control over the major gene so that it contributes less at high levels and in the high line. Control of the major gene at the scute locus has been introduced by selective breeding to operate at the 2-bristle level. A population homozygous for sc is so tightly controlled that over 90% of all flies have two bristles (RENDEL,SHELDON and FINLAY1965). This effect was brought about by two main periods of response to selection each lasting some 15 generations. This indicates several genes are involved, and it appears that what controls the major gene at the scute locus is the product of a set of regulator genes. This conclusion is supported by the way this population and a wild-type population responded to inbreeding (SHELDON, RENDELand FINLAY1964). Control of the wild-type allele remained at the non- inbred level or increased in intensity in half the inbred stocks and was reduced in the other half to varying extents. In summary, the following genetic elements have to be distinguished: the major gene which is controlled, the regulator genes whose product does the controlling, and minor genes which support the major gene but which are not controlled. There are several other points in the data to be explained. We may consider together the difference between sc and -I- genotypes in the probit width of the 4 class; the difference between populations of a given major genotype in low lines and high lines, in width of the 4 class; and the difference in variance between populations in the high line and those in other lines. The differences of width of the 4 class in sc and + genotypes are clear from Table 1. The decline of width of the 4 class with increase in mean amount of M is shown as follows:

Line Genotype Width oj 4 class

Low 6.04. U 5.29 U 4.51 U

sw 5.46 U 5.10 U 3.65 U

The greater variance of populations with a high M can be seen from the width of classes common to high and low populations. The 5 class in the low line is 1.25~~in SW 1.0~~and in high lines about 0.80. In general in the high lines, classes above 5 are more numerous and narrower than those below 4. These three observations are to be expected on the assumptions that only the major gene is controlled, that control is with respect to total gene activity of the whole system, and that control starts at the lowest level becoming more intense as background activity increases. By the time most of a population has five or more bristles the gene at the scute locus will be shut off to the maximum extent in all individuals. Below this level maximum shut-off will not have been reached in all individuals, and increases in background will be accompanied by decreases in major gene activity and vice versa. Variability below 4 will be damped down CONTROL OF DEVELOPMENT 1149 by buffering in and out on the gene at the scute locus. This is a consequence of the shutting-off process starting at the beginning and being a gradual process. In low lines contribution from the major gene is maximal, and in low individuals of low lines it is greater than in high individuals. This is true of all lines until the level is so high that the major gene is turned off all the time in all individuals. At'this point variance will be highest since all activity of the major gene that can be controlled out has been controlled out. As level of minor-gene activity increases and the contribution from the major gene decreases, the width of the 4 class shrinks. When major-gene activity is maximum, i.e. about 8 to 9~,the distance between sc/Y low and +/Y low, the + gene is turned off in all individuals as a level sufficient for four bristles is reached. In the low line very few individuals reach a level of total activity sufficient for five bristles; total activity is made up of minor-gene activity added to a contribution from the + gene some of which is uncontrollable. The 4 class in low +/Y measures about 60; this represents + gene activity which can be controlled out as minor-gene activity increases. As level of minor-gene activity increases beyond this point, more and more individuals in the population reach five or more bristles because there is an increase in the fraction of the population in which minor-gene activity added to the irreducible contribution from the 4- gene is sufficient for five bristles. The 4-class width reduces to 5,4,3 etc. as shown above and finally to less than 2a in High sc/sc+.Y. At and above this level, population means are so high that the 4-class width is usually not measurable. This suggests that the irreducible minimum can be less than 2u. As the sc allele is less powerful than the +, it has less which, by being turned off, can be made to act as the control mechanism at the 4-bristle level. This is why the 4-class width in sc flies measures only 30 compared with the maximum of 60 in + flies and why high sc flies show more variance below 4 than + flies do. Even if the sc gene could be completely controlled out it would have less buffering ability than +, but in fact it too has an irreducible minimum activity, about 0.6~ in sc+/scY males. Variations in position of the 3/4 cut-off at high mean levels of M, and the fact that it is still measurable owing to the occasional occurence of a fly with only three bristles, are assumed to be determined by minor fluctuations in the precise degree to which control on the major gene cuts the gene off. In individuals with lolw minor-gene activity relative to the rest of the population, the + gene must be shut off occasionally before it can contribute enough to bring the total activity to the 4-bristle level. This is more likely to happen in low than in high lines. The general picture is one of considerable reserve activity of the + gene available to individuals of most populations above the low + population. There will presumably be some lag period between gene activation and major- gene suppression which will account for the irreducible minimum of activity contributed by every addition of a major gene. According to the above interpretation, 4 in the equation P = 4 (M),represented in Figure 1, is determined by the manner in which the major gene is controlled. This study enables us to progress from an indirect statistical view of 1150 J. M. RENDEL et al. a canalised character to one in which the broad lines of the genetic mechanism seem identifiable. The effects of substituting + for sc in sc/sc or +/sc females are less than the addition of 4- to sc/Y or +/Y in males, partly because one is a substitution and the other an addition, the sc gene itself always having some activity. In part we are also dealing with the difference between the ratio of one scute locus to one X chromosome and two scute loci to one X, as indicated previously. However, the explanation may not be as simple as this, since comparison between the male and female sets of data have certain other features which are not obviously con- sistent with the model proposed here. For instance, addition of sc to sc/Y in the low line adds 4.6 units, bringing the mean of sc/scY to about the middle of the 2 zone on the male scale. Yet low sc/sc females are 1.1 below the 0/1 cut-off, or some 3 units below sc/scY males, if the male and female scales are identical. Even if allowance is made for a postulated difference of 4-1.7 between the female and male zero-zone widths (RENDEL1963), the sc/sc females are still below expectation, as control of the sc gene at this level should hardly enter the picture. Perhaps the low result in low sc/sc females comes about because the positive increment due to the second sc gene is partly discounted by the additional low modifiers on the second X chromosome. To take a similar example, sc/scY males are about 0.3~above sc/sc/scY females, taking the male and female scales as identical. Some doubt arises as to whether this apparent reversal of expectation is explained adequately by a tighter control on the sc genes in the females for reasons associated with a different ratio of scute locus to total X. FRASERand GREEN( 1964), using the same Y chromosomes, obtained much the same results as we did which they have interpreted somewhat differently. A find- ing quite definitely different from our results was the lack of correlation they found between the scores of sc and + segregants in some of the selection lines. Their lines were selected as wild-type lines and later had the sc gene introduced into them by backcrossing, while our high and low lines were segregating for sc and + throughout and were selected largely on the sc segregants. It is possible that high selection in wild type has been, at least in some lines, partly for lack of control of the 4- gene. If this has come about, increase in bristle number in 4- would not necessarily be reflected in sc flies to anything like the same extent. If selection has picked out 4- genes which are poorly controlled, sc genotypes will not be affected at all. If it is the control system which is affected, sc genotypes will be affected but in proportion to the activity of the + and sc genes, unless the scute allele used is a mutated operator which appears not to be so in sc'. DUNand FRASER(1958, 1959) working with the Tabby gene in mice put forward the hypothesis that canalisation of the number of secondary vibrissae was disrupted by Tabby, +/Y and +/+ being more strongly canalised than +/Ta, which was more strongly canalised than Ta/Y. This hypothesis was re- jected by RENDEL(1959a) on the ground that the sc genotype in stocks selected for high bristle number was canalised at 4 bristles, similar to unselected 4- genotypes, and the f genotype at 8 bristles was completely uncanalised, so that it was the phenotype 4-bristles that was canalised regardless of the gene at the major CONTROL OF DEVELOPMENT 1151 locus. The present work has made it possible to reconcile canalisation at the phenotypic level of 4 bristles with control of the developmental system through the agency of the scute locus.

SUMMARY Three stocks in which the number of scutellar bristles was high, medium and low were kept segregating for scl. An extra scute locus carrying scl or sc+ was added to all of these on a Y chromosome to which the tip of an X had been translocated. An extra scute locus carrying the + gene tended to increase bristle number more in genotypes such as sc and sc sc, with low activity at the scute locus, than in genotypes such as + and ++, and more in stocks at a low level of background activity than in stocks at a high level. It was concluded that control of phenotype (canalisation) was exercised through control of the gene at the scute locus by a product of the developmental system as a whole. The minor genes responsible for the variable background were not controlled with respect to the scutellar phenotype, if at all, since with a + or ++ genotype in the high-line stock an extra + gene added a minimal amount to bristle number, yet variability of bristle number due to minor genes reached its maximum at this level.

LITERATURE CITED

DUN,R. B., and A. S. FRASER,1958 Selection for an invariant character-"vibrissa number"-

in the house mouse. Nature 181: 1018-1019. ~ 1959 Selection for an invariant character, vibrissa number, in the house mouse. Australian J. Biol. Sci. 12: 506-523. FRASER,A., and M. M. GREEN,1964 Variation of scutellar bristles in Drosophila. 111. Sex- dimorphism. Genetics 50: 351-362. JACOB, F., and J. MONOD,1961 Genetic regulatory mechanisms in the synthesis of proteins. J. Mol. Biol. 3: 318-356. RENDEL,J. M., 1959a Canalisation of the scute phenotype. Evolution 13: 425-439. - 1959b Variation and dominance at the scute locus in Drosophila melanogaster. Australian J. Biol. Sci. 12: 524-533. - 1962 The relationship between gene and phenotype. J. Theoret. Biol. 2: 296-308. - 1963 Correlation between the number of scutellar and abdominal bristles in Drosophila melanogaster. Genetics 48: 39148. RENDEL,J. M., B. L. SHELDON,and D. E. FINLAY,1965 Selection for canalisation of the scute phenotype. 11. Am. Naturalist (in press). SHELDON,B. L., J. M. RENDEL,and D. E. FINLAY,1964 The effect of homozygosity on developmental stability. Genetics 49: 471-484.