STUDIES ON THE NATURE AND FUNCTION OF

Heredity (1974), 33 (3), 373-387 STUDIES ON THE NATURE AND FUNCTION OF POLYGENIC LOCI IN DROSOPHILA I. COMPARISON OF GENOMES FROM SELECTION LINES JAMES N. THOMPSON, Jr. Department of Genetics, University of Cambridge, Milton Road, Cambridge CB4 I XH Received 22.iii.74 SUMMARY Six mutants, which cause shortened veins or the formation of extra vein fragments, were selected for increased and for decreased amounts of total vein material. All lines responded readily to selection. F1 double heterozygotes from crosses between two different mutant lines, which had been selected for a decrease in total vein, often had a high frequency of flies with gaps in the normally wild type longitudinal veins. In these crosses, gaps were found to be localised to those veins which were affected by both original mutants. Similar crosses between two different mutant lines which had been selected for an increase in the total amount of vein, often had a high frequency of flies with fragments of extra vein material. Crosses between Long and Short lines having different mutants and crosses between unselected control lines generally produced wild type F1 double heterozygotes. The effects of Short and of Long selected polygenic backgrounds appear to be cumulative in the double heterozygotes; that.is, at least a proportion of the polygenic loci in each selected background modify mutant vein length indirectly by acting upon the steps in some common developmental process leading to the formation of vein material. 1. INTRODUCTION THE genetic architecture of a quantitative character is a reflection both of the level of genetic variation in natural populations and of the kind of selection which has acted upon the population in the past (Mather, 1966). But our knowledge of the precise relationships between genotype and phenotype is severely limited if we do not have a clear understanding of the ways in which the units of quantitative variation, either as individual factors or through interactions, produce their effects upon developmental processes. What role do polygenes play in development? Various authors have approached the problem of the development of quantitative characters (Maynard Smith and Sondhi, 1960; Sondhi, 1961, 1962; Rendel, 1967; Edwards, 1967a, b, 1970; Balal, 1972a, b; Hiliman et al., 1973; Rose and Hillman, 1973). Using polygene location techniques (Thoday, 1961; Milkman, 1970), attempts have been made to identify the developmental effects associated with the individual factors in a polygenic system (Spickett, 1963; Mohler and Swedberg, 1964; Milkman, 1970). One of the most interesting points to come out of these studies is that major changes in the expression of a quantitative character can arise from small, independent changes in a number of quite different developmental processes. Thus, Spickett (1963) found that an increase in the number of sternopleural chaetae in a high bristle number selection line of Drosophila 33/3 2.4 373

374 JAMES N. THOMPSON, Ja melanogaster could be traced to (I) a general increase in cell number, (2) an extension of the distribution of chaetae in the ventral section of the sternopleurum, and (3) by the retardation of the development of one large chaeta which resulted in the formation of several smaller chaetae in that region. In most systems, however, the action of polygenic variation is not clear and many general questions remain to be answered. The manipulation of environmental variables (Waddington, 1953) or the presence of a major gene mutation (see, for example, Dun and Fraser, 1959; Scharloo, 1962; Rendel et al., 1965) can be used to uncover some of the genetic variation present in wild populations but not usually expressed in the adult phenotype. In the present study, major gene mutations which affect the wing have been used to aid in discriminating among genotypes which contribute to quantitative variation in wing vein length. The wing of Drosophila melanogaster has long been recognised as a convenient system for the study of development and the effects of major gene mutations (see Waddington, 1939, 1940; Garcia-Bellido, 1972). The wing vein system, and specifically the polygenic loci which affect the expression of major mutants, is also suitable for the study of developmental effects for several reasons. First, the application of quantitative techniques requires that measurements of the phenotype are easily made and accurate. Vein length can be measured quite easily on mounted wings using a microscope eyepiece graticule. Second, vein mutants show a great amount of variation when they are outcrossed into a wild type genetic background, and vein length responds readily to artificial selection. Third, the histological development of the wing is easily studied and the development of the normal wing is well known (see Chen, 1929; Auerbach, 1936; Waddington, 1940). Thus, the histological effects of selected backgrounds and of individual polygenic loci are readily accessible to study. The developmental influence of polygenic loci can be studied at many different levels. One approach is to ask whether the polygenic modifiers which are selected through their effects upon one mutant have similar phenotypic effects upon other non-homologous mutants. That is, do the polygenic loci in selected genomes modify the expression of a specific mutant, or do they affect the expression of the mutant indirectly by modifying relevant stages in a common developmental process? Thompson and Thoday (1972) made a series of crosses among vein mutant selection lines at an early point in the selection programme and showed that modifiers of phenotypic expression could also act as modifiers of dominance. Their study, however, did not include crosses between Long and Short selected lines of different mutants, so that there was no means of determining whether penetrance of the mutant phenotype in the double heterozygote F1 flies was due to dominant modifiers in one line or to additive effects from both lines. This paper will first describe the origin and responses of the selection lines and then discuss the results of crosses in all combinations among selected lines at generation S-37. The results suggest that modifier effects are general and cumulative. Thus, at least a proportion of the modifiers in a selected genome appear to affect steps in a basic developmental process, rather than the products or function of a specific mutant locus. The general nature of the processes which might be affected by the selected modifiers will be discussed in the following paper (Thompson, 1 974b). The conclusion, that selected modifiers

POLYGENIC LOCI IN DROSOPHILA 375 act upon some developmental process and only affect the mutant phenotype indirectly, is supported by a series of chromosome substitution experiments reported by Thompson (1973) which showed that selected L4 vein modifiers affect L4 vein length in several different mutants and must, therefore, be considered as acting generally upon L4 vein development. 2. MATERIALS AND METHODS The following mutants, arranged by chromosome and map position, were used in this study: net (net, 2 0.0); short vein (shy, 2 3.8); plexus (px, 2 100.5); veiniet (ye, 3 0.2); radius incompletus (ri, 3 47.0); cubitus interruptus (ci, 4 0). Both px and net have fragments of extra venation, particularly along the margins and tips of the longitudinal veins. Shortened longitudinal veins are found in the other mutants. The L2 vein is shortened in ri; the L4 is shortened in ci; and terminal gaps are usually present in the L2, L3, L4 and L5 veins in shy and ye. These mutants were chosen because they have no known pleiotropic effects which might complicate the analysis and because they cover a wide range of possibly related vein modifications. For detailed descriptions, see Lindsley and Grell (1967) and Thompson (1974a). Selection lines and crosses were maintained at 25± 10 C. on an agarbase oatmeal and molasses medium seeded with a suspension of live yeast. Wing samples were made by removing the right wing from adult Drosophila and mounting them in DePeX mountant on a microscope slide. Vein lengths and total wing lengths were measured with an eyepiece graticule in a Zeiss binocular microscope at a magnification of 625 x. The graticule was aligned parallel to the vein, with the origin at the intersection of veins L2 and L3 or on a line perpendicular to this intersection. The length of the vein or wing was then recorded as the nearest whole number of graticule units from the origin to the end of the vein (1 unit = 0038 mm). For those wings in which the vein was made up of a number of small fragments, the fragments were measured individually and their lengths added. To correct for variations in wing size, vein lengths were standardised by expressing them as the ratio of vein length to total wing length. It should be noted that since the different veins intersect the wing margin at varying distances from the base line, a complete vein may be represented by a proportion of less than 10 (cf. vein lengths of Oregon wild type, table 1). When data are expressed as proportions, variances differ according to the means (Falconer, 1960), and a transformation is required to make them dependent only upon the sample size. An angular, or sine, transformation, in which each piece of data is transformed to the angle whose sine is the square root of the proportion (Fisher and Yates, 1957), was used for this purpose. Unless otherwise stated, then, "vein length" should be understood to mean "sine transformed ratio of vein length to wing length ". (i) Origin and maintenance of selection lines Artificial selection performs several functions in an investigation of the modifier background. First, it accumulates and makes homozygous those polygenic loci which affect the mutant phenotype in a particular way. Second, it provides some information on the nature of the vein system by

376 JAMES N. THOMPSON, JR showing the ways in which vein material is added to or eliminated from the wing and in demonstrating the extremes to which selection on the available genetic variation can alter the venation phenotype. Finally, accelerated responses and other aspects of the selection response can often suggest relationships among the modifier factors which may be useful in later analyses of the lines. TABLE 1 Relative vein length measurements in the original mutant stocks and in Oregon wild type Stock N L2 vein L3 vein L4 vein Orgeon 15 0623±0007 (0.784) 0900±0000 (1.000) 0807±0015 (0.974) Oregon 15 0619±0010 (0.778) 0900±0000 (1.000) 0797±0018 (0.968) ci 50 0723±0094 (0.908) ci 50 0 699±0 099 (0.882) ri 100 0420±0'044 (0.447) ri 100 0356±0011 (0.339) shy 50 0578±0023 (0.716) 0628±0011 (0.791) shy 50 0578±0022 (0.716) 0619±0011 (0.778) ye 50 0608±0O07 (0.762) 0679±0014 (0.858) 0397±0022 (0.408) ye 50 O597±0016 (0745) 0655±0015 (0828) 0379±O018 (0377) Mean and standard deviation are given for the transformed ratio of vein length to wing length (sin 'V(x)/lOO) with the retransformed mean in parentheses. The mutants shy, ye, ri and ci were selected for increased and for decreased vein length. Vein lengths in the original mutant lines are given in table 1. Before selection, these four mutants were outcrossed to a newly captured wild stock, Eversden- 14, to provide cultures with increased genetic variability. Reciprocal five-pair crosses were made, and the homozygous mutants were resegregated in the F2 generation and pooled to form a base population (table 2). TABLE 2 Relative vein length measurements in the base populations resegregated from an outcross to Eversden-14 Stock N L2 vein L3 vein L4 vein ci 50 0 552±0 144 (0.674) ci 3' 50 0488±0 129 (0566) ri 100 0428± 0065 (0461) ri 3' 100 0362±0035 (0348) shy 50 0584±0024 (0725) 0630±0028 (0794) shy 50 0563±0026 (0.692) 0884±0048 (0999) 0588±0'043 (0.731) ye 50 0609±0028 (0763) 0683±0027 (0.863) 0481 (0554) ye 50 0561 (0.689) 0669±0014 (0.846) 0484±005O (0.559) Mean and standard deviation are given for transformed ratios as in table 1. It is interesting to note in passing that the L4 veins in ci and ye were affected differently by outcrossing. L4 length decreased in ci, indicating that natural selection had acted to reduce mutant expression in the stock culture. In ye, on the other hand, outcrossing caused the L4 vein length to increase.

POLYGENIC LOCI IN DROSOPHILA 377 Vein length heritabilities were estimated in the newly resegregated mutant lines (table 3). Parents were selected from the base populations to begin eight independent selection lines for each mutant. In four lines, designated Long-I, II, III and IV, parents were selected in which the affected veins were the longest, i.e. in which the expression of the mutant was reduced. In four other lines, designated Short-I, II, III and IV, parents were selected in which the affected veins were the shortest, i.e. in which the mutant expression was increased. Selection was carried out on a fortnightly schedule. Two sets of five pairs of parents were selected from each line and placed in 4-inch tubes to mate for 24 hours before being transferred to bottles. These two cultures were identical in ancestry, and one bottle and the premating tubes therefore served as reserve cultures. TABLE 3 Vein length heritabilities estimated in the newly resegregated mutant stocks Mutant Vein N h2±s.d. ci L4 20 032±0l7 ri L2 29 062 008 she L2 30 078± 007 she L4 30 065±0l0 cc L2 29 08l±017 cc L3 29 058±026 cc L4 29 086±006 N is the number of single-pair cultures assayed. The primary objective of the selection programme for mutant expression was to obtain lines which were homozygous for polygenic modifiers causing an increase or decrease in vein length. Thus, only certain generations were sampled at the beginning of selection, and after the responses had slowed, the lines were sampled irregularly. It should also be mentioned that vein length is a truncated character in that the base of the wing (little or no vein, depending upon the vein structure) and the distal margin of the wing (full vein) are the effective limits to any phenotypic response, if one ignores fragments in the intervein areas. For this reason, the original directional selection pressure decreases as the phenotype approaches one or other extreme and eventually becomes stabilising selection for an extreme phenotype. The line ye Long-I was lost at generation S-26 through infertility. Since it was not available for analysis, it will not be considered further. Approximately one year after the initiation of the four primary sets of selection lines, four selection lines each of px and net were started. They also were outcrossed to Eversden- 14, which had been maintained in large populations by mass transfer. High-I and II were selected for increased amounts of vein material, while Low-I and II were selected for decreased amounts of extra vein material, i.e. selected to approach a wild type phenotype. Since the response of these lines is intimately associated with a separate study of the venation pattern, the px and net selection lines will be described in detail elsewhere. At the time of the crosses discussed below, however, the High and Low lines did not overlap in phenotype. 1'lies 33/3 2A2

378 JAMES N. THOMPSON,Ja from the px and net Low lines occasionally had small fragments in the marginal, second or third posterior cells, though many individuals were wild type in appearance. Flies from the px and net High lines had large branches and fragments in all cells of the wing, and a large bend and forks in the distal section of the L4 vein. (ii) Crosses among selected lines At generation S-37, five-pair crosses among representative selection lines and wild type stocks were made in all combinations. The lines used in these crosses were: shy Long-Ill, shy Short-Ill, ye Long-IV, ye Short-TV, ri Long-Il, ri Short-TV, ci Long-IT and ci Short-I. Crosses to make single and double heterozygotes from unselected mutant lines served as controls. Reciprocal matings were made for each combination. Three cultures of each cross were assayed, and in each of these cultures 50 males and 50 females were examined for gaps in the longitudinal veins or fragments of extra vein material in one or both wings. In the discussion of these results, the three cultures of each cross have been pooled. Table 6 indicates whether there is significant heterogeneity associated with sex, culture, or between reciprocal crosses. 3. RESPONSES TO SELECTION Mean vein lengths in the base populations are given in table 2, and mean lengths in representative selection lines at generation S-37 are given in table 4. The lines described in table 4 are listed in section 2 (ii). Clearly, TABLE 4 Relative vein length measurements for representative selected lines at generation S-37 Selection line L2 vein L3 vein L4 vein ci Long 0782 0074 (0.958) ci Long 0798±0020 (0.968) ci Short 0323±0035 (0.285) ci Short o' 0333±0033 (0.301) ci Long 0578±0056 (0.716) ri Long 0528±0l00 (0.634) ci Short 03l0±0015 (0265) ri Short 0309±0010 (0.263) shy Long 0603±00l7 (0.754) 0 658±0 0l0 (0.832) shy Long 0604±00ll (0.756) 0653±00l5 (0.825) shy Short 0522±0O38 (0.624) 0697±00l2 (0.880) 0457±0014 (0.512) shy Short 0508±00l5 (0600) 0697±OO1O (0.880) 0436±0014 (0.475) ye Long 0636±0007 (0802) 073l (0.915) 0665±0013 (0.841) ye Long 0633±0008 (0.798) 0 731 (0.915) 0665±00l9 (0.841) ye Short 0466±0024 (0.528) 0 493±0039 (0.574) 0385±0 0l6 (0.387) ye Short 0405±0 015 (0421) 0 505±0 040 (0.595) 0378±0 017 (0.375) Mean and standard deviation are given for transformed ratios as in table 1. N 15. vein length has responded significantly to selection in all instances. Phenotypic differences between Long and Short lines at about generation S-9 are illustrated in Thompson and Thoday (1972).

The general selection responses were largely similar among the four parallel lines of each type. This is best illustrated by the Long and Short selection lines of cubitus interruptus (fig. 1). In all four ci Long lines, vein length increased rapidly during the first generations, and after S-l8 most 0.825 0.800 0.775 0.750 0.725 0.700 0.675 0.650-0.625 C) a) j 0.575 0.550 a) > 0.525 a) 0.500 0.475 0.450 0.425 0.400 0.375 0.350 0.325 0.300 POLYGENIC LOCI IN DROSOPHILA 379 2 6 8 1012141618202224 '36 38 Generation FIG. 1. Selection responses for L4 vein length in cubitus interruptus. Symbols mark the generations at which each Long line (solid lines) and each Short line (dashed lines) was sampled. flies had complete L4 veins. In the four ci Short lines, the rate of response was slower than that in the Long lines. By generation S-l 8, however, most flies had little or no L4 vein distal to the posterior crossvein. The slight differences among lines tend to reflect the occasional loss of vein at the base of the wing or curves in the posterior crossvein. The major exception to the observation that parallel lines responded in similar ways occurred in the ye Long lines. At about generation S-7 the

380 JAMES N. THOMPSON, Ji L3 and L4 veins began to increase abruptly in the ye Long-H line. Vein L2 was already complete. This increase was found to be the result of a spontaneous mutant which tended to suppress the expression of the veinlet mutant. The new mutant is located on chromosome II at about 285 cm and was named gravel (Thompson, l974a) because of an associated effect upon the facets of the compound eyes. There is a large sex difference in the expression of most longitudinal veins (table 2). The males usually have markedly shorter veins than the females. This sex difference, particularly in the L2 veins of ri, ye and shy, was usually maintained during selection, although it was obscured in some lines as the veins approached a phenotypic extreme. 4. Cosss AMONG SELECTION LINES OF DIFFERENT MUTANTS Flies which are heterozygous for two or more venation mutants often have vein defects at a low frequency (see for example, House, 1953, 1961). These defects are gaps or fragments of extra vein material, depending upon the phenotypes of the vein mutants in the lines, and the frequency with which defects appear is partly dependent upon the genetic background (Thompson and Thoday, 1972). The frequency of gap or fragment appearance can, therefore, provide one with a sensitive measurement of the similarity of modifier effects in various combinations of heterozygous selected genomes. By generation S-37, selection responses had generally reached a plateau and the variance in expression was low. From the sets of parallel lines, representative Long and Short lines of each mutant were chosen and crossed together in all combinations, including reciprocals. Results from three cultures of each cross have been pooled and summarised in table 5. TABLE 5 Summary of the phenotypes produced in crosses among selected lines. Gaps (G), extra vein fragments (F), wild type (0). Reciprocal crosses are represented only once unless they dffer, in whkh case a diagonal separates them, e.g. shy S x ri L3 above the diagonal, ri L x shy Sd' below the diagonal ye L ye S ri L,i S ci L ci S Oregon shvl F 0 F 0 0 0 0 shvs 0 G 0/G G 0 G 0 vel F/0 0 0 0 0 yes G G G/0 G G ril 0 0/F 0 ris G/0 G 0 cil. 0 cis - 0 The first thing which emerges from these results is that there is a high frequency of gaps in all crosses between Short selected lines of two different mutants. This confirms the observations reported by Thompson and Thoday (1972). The highest frequency is found in crosses between ye Short and shy Short. Both of these mutants tend to shorten the L2, L3, L4 and L5 veins. In the Short double heterozygotes, gaps appeared in one or more of these four veins in 321 out of 600 F1 flies (53.5 per cent) from the pooled reciprocal crosses (table 6).

POLYGENIC LOCI IN DROSOPHILA 381 The lowest gap frequency was found in the F1 double heterozygote from crosses between ri Short and ci Short. These lines might be supposed a priori to be the least similar in their selected modifier systems, since ci had been selected solely upon the expression of the L4 vein, while ri had been selected for L2 vein expression. TABLE 6 Occurrence of gaps (G) in double heterozygotes of some crosses among Short and Long selection lines of shy, ye, ri and ci Cross Female offspring Male offspring A Female Male + G + G " * P5 P0 Pr shos yes 65 85 89 61 ** n.s. * ye S shy S 48 102 77 73 shvs ril 150 0 150 0 n.s. ril shus 150 0 122 28 *** n.s. shy S ii S 145 5 65 85 "" n.s. ris shos 142 8 112 38 *** n.s. shy S ci S 143 7 132 18 * * cis shvs 135 15 85 65 "u yes ril 149 1 147 3 n.s. 1 ** ril ye S 150 0 138 12 n.s. yes ri S 135 15 110 40 * ris ye S 119 31 103 47 * * yes ci L 150 0 146 4 n.s. cil ye S 150 0 150 0 n.s. yes ci S 124 26 119 31 n.s. ** cis ye S 120 30 69 81 n.s. yes Oregon 142 8 136 14 n.s. * Oregon ye S 133 17 125 25 n.s. n.s. ri S ci L 150 0 144 6 n.s. ci L ri S 150 0 150 0 n.s. ri S ci S 146 4 147 3 n.s. 1 * cis ris 147 3 136 14 ** n.s. x' significance levels are given for comparisons between sexes within a cross (Ps) and for heterogeneity among the three cultures within each reciprocal (P0). Sexes and cultures were pooled in a test of the difference between reciprocals (Pr). * 005>P>001. ** 0 01>P>0 001. P<0001. In the heterozygotes from other crosses between Short selected lines, it was found that the gaps usually appear in the vein which is affected by both of the original mutants. For example, both shy and ri shorten the L2 vein. Thus, in the cross between shy Short and ri Short, 136 out of 600 F1 flies have gaps in the L2 vein. Similarly, the mutants shy and ci shorten the L4 vein. In crosses between shy Short and ci Short, most of the 168 affected flies had gaps in the L4 vein. Vein specificity was also observed in the other crosses, although L5 gaps were common in the offspring from all crosses involving ye Short. The L5 vein is completely absent distal to the posterior crossvein in the yeshort I **

382 JAMES N. THOMPSON, JR selection lines, and L5 gaps were even observed in the F1 from crosses to Oregon wild type (table 6). The second main observation is that double heterozygotes from crosses between lines selected for longer veins are either wild type in appearance or have fragments of extra vein material (tables 5 and 7). These fragments usually occur in the marginal cell between the L2 vein and the anterior margin of the wing. The proportion of flies with extra vein material was highest in the crosses between ri Long and shy Long. TABLE 7 The occurrence of fragments of extra vein material (F) in double heterozygotes of some crosses among Short and Long selection lines of shy, ye, ri and ci. Cross Female offspring Male offspring A A A -' /- ' r \ Female Male + F + F P5 Pc Pr 51w L ye L 130 20 143 7 ** * 1 *** vel shvl 142 8 150 0 n.s. shy L ri L 49 101 67 83 * n.s. ri L shy L 105 45 127 23 ** n.s. ft vel ril 147 3 145 5 n.s. ri L ye L 150 0 150 0 n.s. ri L ci S 150 0 150 0 n.s. ci S ri L 137 13 136 14 n.s. n.s. ft Data are arranged as in table 6. Defects were also observed in five crosses between a selected Short line of one mutant and a Long line of another mutant. In four types of crosses gaps were produced in the F1 flies. These were the crosses between ri Long females and shy Short males, ye Short females and ci Long males, ri Short females and ci Long males, and the reciprocal crosses between ri Long and ye Short. In all instances, however, the frequency of vein defects was significantly lower than in the respective crosses between Short selected lines (table 8). TABLE 8 Comparison of vein gap frequency in crosses among Short lines (S x S) with that in crosses between Long and Short lines (L x S). Reciprocal crosses are pooled. (The Long line is given first in column 1) Cross SxS LxS A (L) (S) + Gaps + Gaps X2i cixve 432 168 596 4 l825 <0001 rixve 467 133 84 16 1049 <0.001 rix shy 464 136 572 28 824 <0001 cixri 576 24 593 7 96 O0l>P>00O1 Fragments of extra venation were observed in the cross between ci Short females and ri Long males, though not between ci Long and ri Long. This is not necessarily surprising, however, since the ci Long line was not directly selected for L2 vein modifiers and could carry a suppressor of the ri Long effect. In combination with other genetic backgrounds, the ri Long males produced a large proportion off1 flies with vein fragments (table 7). There is a significant difference between the reciprocal crosses of each

POLYGENIC LOCI IN DROSOPHILA 383 type (tables 6 and 7). This suggests that there may be sex-linked modifiers, maternal, or cytoplasmic effects associated with the lines. No vein defects were found in any of the control crosses between unselected mutant stocks. In the crosses among the ye, shy, ri and ci lines, the results are, therefore, quite consistent. In crosses between two mutant lines which had been selected for a decrease in the length of a particular vein, the F double heterozygotes tend to have gaps in that vein. When two lines are selected for an increase in vein length, i.e. selected for an increase in total vein material, the double heterozygotes often have fragments of extra venation. TABLE 9 Occurrence of extra vein material in hetero zygotes from crosses among px High (H), px Low (L), net High, net Low, shy Long (Lg), shy Short (S), ye Long, and ye Short. Cross Female offspring Male offspring neth x pxh 0 150 0 150 neth x pxl 20 130 79 71 netl x pxh 0 150 0 150 netl x pxl 146 4 150 0 shy Lg x px H 0 150 2 148 shy Lg x px L 49 101 114 36 shy S x px H 134 16 149 1 shvs x pxl 150 0 150 0 shy Lg x net H 0 150 0 150 shy Lg x net L 56 94 67 83 shvs x neth 105 45 142 8 shvs x netl 150 0 150 0 velg x pxh 89 61 146 4 ye Lg x px L 150 0 150 0 yes x pxh 148 2 150 0 yes x pxl 150 0 150 0 velg x neth 0 150 0 150 velg x netl 150 0 150 0 yes x neth 141 9 150 0 yes x netl 150 0 150 0 Flies from crosses between Short and Long lines containing different mutants and between comparable unselected lines are usually wild type. Similar relationships between selected modifier backgrounds are also observed in the final series of crosses (table 9). When flies from the px and net lines, which had been selected for an increase or a decrease in total vein material, are crossed to flies from the shy and ye selection lines, vein fragments occur with high frequencies in those crosses in which the parental lines had been selected for an increase in the amount of vein. In crosses in which the parental lines had been selected for a decrease in the total amount of vein, the frequency of fragments in the F1 double heterozygotes was decreased or zero. 5. Dxscussxor The mutants in these selection lines represent a wide variety of abnormal venation phenotypes. All mutant lines responded readily to selection in

384 JAMES N. THOMPSON, JR each of the two directions: for a polygenic background which reduced the ability to form vein material (e.g. ye Short, px Low) and for a polygenic background which increased the fly's production of vein (e.g. ye Long, px High). General selection responses were largely similar among the four parallel lines of each type. The rapid L4 vein responses in most lines, and especially in ci (fig. 1), suggest that only a small number of polygenic modifiers affect the expression of this vein. Indeed, the rapid responses of all veins in the four mutants are consistent with the hypothesis that a relatively small number of polygenic loci account for the majority of the response to artificial selection. A small number of effective factors was also found to be sufficient to account for a large proportion of the responses of certain sternopleural chaetae lines studied by Thoday and his colleagues (Thoday et al., 1964; Spickett and Thoday, 1966; reviewed by Lee and Parsons, 1968; see also Bateman, 1959; Milkman, 1965). When two lines containing different vein mutants, which had been selected for decreased vein-forming ability, were crossed the decreased veinforming ability was expressed as a high frequency of gaps in the F1 fly wings. Conversely, when two comparable lines selected for increased vein-forming ability were crossed, the increase was often found to result in F1 flies with a high frequency of extra vein fragments. Control crosses were usually wild type in appearance. If the modifiers accumulated by selection simply affected the products or function of the single mutant locus through which their effects were originally observed, the frequency of vein defects in Short x Short double heterozygotes would be expected to be no higher than in Short x Long double heterozygotes. But the effects of the selected genomes appear to be cumulative. The results imply, therefore, that at least a proportion of the polygenic loci in each selected background modify mutant vein length indirectly by acting upon steps in a common developmental process which is basic to the formation of the vein system. The set of modifiers in one selected Short genome supplements the action of the modifiers in another Short line genome. Indeed, the same loci may be acting in both mutant lines. In the Short double heterozygotes, then, gaps appear in the vein or veins determined by the process which is modified by both sets of polygenes. In the control crosses and in crosses between Short and Long selected lines, the dissimilar modifier actions presumably compensate for each other. Others have also studied the ways in which a single set of polygenes can modify the expression of several related characters. This has usually been done by the measurement of correlated responses in selection lines (Mather, 1943; Clayton et al., 1957; Haskell, 1959). The measurement of correlations, however, is directly relevant to the question of the generality of polygene action only when the experiments have been designed to test whether or not the correlations are due to linkage. Thus, Davies and Workman (1971) not only measured the correlations between abdominal and sternopleural bristle numbers when each was modified separately by selection, but also actively selected these bristle characters in opposite directions in the same line to test the importance of linkage. They were able to separate the phenotypic effects and show that polygene specificities were different for the two bristle systems and that linkage played an important part in the correlated responses. This conclusion was supported

POLYGENIC LOCI IN DROSOPHILA 385 by Davies (1971) who, by locating many of the bristle number genes, demonstrated that the genes affecting sternopleural bristle number were distinct from the genes affecting abdominal bristle number In another study, Fraser (1968) showed that the modifier systems of the bristle mutants scute and extravert were independent. It is probably unlikely that the numerous examples of cumulative modifier action from independent vein length selection lines would all be traceable to the linkage of mutant-specific modifier factors, although linkage cannot be ruled out at the present level of examination of these genomes. Thompson (1973) has recently shown that single heterozygous whole chromosomes carrying L4 vein modifiers from one selection line have qualitatively similar effects upon the expression of the L4 vein in a second vein mutant. The polygene location technique of Thoday (1961) will be used to see whether this is also true of much smaller regions within a selected thromosome. Histological similarities (Waddington, 1940) and frequent interactions among wing vein mutants (e.g. House, 1953) suggest that the primary actions of these mutants are intimately related. Such a close developmental relationship may have facilitated the measurement of modifier similarities in this system, although Scharloo (1964) reported that he had observed little similarity in the modifiers affecting L4 vein length in Hairless and cubitus interruptus Dominant in Drosophila melanogaster. It is concluded that general, as well as specific, modifiers can be present in selection lines. It is hoped that the lines described in this paper may be used to provide some understanding of the ways in which these two groups of polygenic loci produce their phenotypic effects and, thus, throw new light on the functions of quantitative genetic variation in natural populations. Acknowledgments. I would like to thank Professor J. M. Thoday for his helpful discussions and comments on the manuscript. During this work I was supported by a Marshall Scholarship. 6. REFERENCES AUERBAcH, c. 1936. The development of the legs, wings, and halteres in wild type and some mutant strains of Drosophila melanogaster. Trans. Roy. Soc. Edin., 58, 787-815. BATEMAN, K. o. 1959. The genetic assimilation of four venation phenocopies. 3. Genet., 56, 443-474. BALAL, M. s. 1972a. Developmental genetic studies on some quantitative characters of rice, Oryza sativa L. I. Variation and correlation in length of different organs of rice. Egypt. 3. Genet. Cytol., 1, 85-91. BALAL, M. s. 1972b. Developmental genetic studies on some quantitative characters of rice, Oryza sativa L. II. Degree of pleiotropic action of polygenes among different organs of rice. Egypt. 3. Genet. Cytol., 1, 149-156. CHEN, T. i. 1929. On the development of imaginal buds in normal and mutant Drosophila melanogaster. 3. Morph. Physiol., 47, 135-199. CLAYTON, G. A., KNIGHT, 0. R., MORRIS, J. A., AND ROBERTSON, A. 1957. An experimental check on quantitative genetical theory. III. Correlated responses. 3. Genet., 55, 171-180. DAVIES, K. w. 1971. The genetic relationship of two quantitative characters in Drosophila melanogaster. II. Location of the effects Genetics, 69, 363-375. DAVIES, R. w., AND WORKMAN, s'. L. 1971. The genetic relationship of two quantitative characters in Drosophila melanogaster. I. Responses to selection and whole chromosome analysis. Genetics, 69, 353-361. DUN, R. B., AND FRASER, A. s. 1959. Selection for an invariant character, vibrissa number in the house mouse. Austr. 3. Biol. Sci., 12, 506-523.

386 JAMES N. THOMPSON, Ja EDWARDS, K. J. R. 1967a. Developmental genetics of leaf formation in Lolium. I. Basic patterns of leaf development in L. muitlorum and L. perenne. Genet. Res., Camb., 9, 233-245. EDWARDS, K. J. K. l967b. Developmental genetics of leaf formation in Lolium. 2. Analysis of selection lines. Genet. Res., Camb., 9, 247-257. EDWARDS, K. j. R. 1970. Developmental genetics of leaf formation in Lolium. 3. Inheritance of a developmental complex. Genet. Res., Camb., 16, 17-28. FALCONER, D. s. 1960. Introduction to Quantitative Genetics. Oliver and Boyd, London. FISHER K. A., AND YATES, F. 1957. Statistical Tables for Biological, Agricultural, and Medical Research. Oliver and Boyd, London. FRASER, A. s. 1968. Specificity of modifiers of scute and extravert expression. Genetics, 60, 179. GARCIA-BELLIDO, A. 1972. Pattern formation in imaginal disks. In: H. Ursprung and R. Nothinger (eds.), The Biology of Imaginal Disks, Results and Problems in Cell Differentiation, Vol. 5, pp. 59-91. Springer-Verlag, Berlin. IIASKELL, G. 1959. Further evidence against pleiotropic gene action in correlated responses to selection. Genetica, 30, 140-151. HILLMAN, R., SHAFER, S. J., AND SANG, J. Ii. 1973. The effects of inhibition of protein synthesis on the phenotype of Abnormal abdomen. Genet. Res., Camb., 21, 229-238. HOusE, V. L. 1953. The interaction of mutants affecting venation in Drosophila melanogaster. I. Interaction of Hairless, engrailed, and cubitus interruptus. Genetics, 38, 199-215. HOUsE, v. L. 1961. Mutant effects in multiple heterozygotes of recessive venation mutants in Drosophila melanogaster. Genetics, 46, 871. LEE, B. T. 0., AND PARSONS, P. A. 1968. Selection, prediction, and response. Biol. Rev., 43, 139-174. LINDSLEY, D. L., AND GRELL, K. H. 1967. Genetic Variations of Drosophila inelanogaster. Carnegie Institution of Washington Publ. No. 627. MATHER, K. 1943. Polygenic inheritance and natural selection. Biol. Rev., 18, 32-64. MArHER, ic. 1966. Variability and selection. Proc. Roy. Soc., Lond. B, 164, 328-340. MAYNARD SMITH, J., AND SONDHI, K. C. 1960. The genetics of a pattern. Genetics, 45, 1039-1050. MILKMAN, R. D. 1965. The genetic basis of natural variation. VII. The individuality of polygenic combinations in Drosophila. Genetics, 52, 789-799. MILKMAN, R. 1970. The genetic basis of natural variation in Drosophila melanogaster. Ads. Genet., 15, 55-114. MOHLER, 3. D., AND SWEDBERG, o. s. 1964. Wing vein development in crossveinless-like strains of Drosophila melanogaster. Genetics, 50, 1403-1419. RENDEL, j. M. 1967. Canalisation and Gene Control. Logos Press, London. RENDEL, J. M., SHELDON, B. L., AND FINLAY, D. E. 1965. Canalisation of development of scutellar bristles in Drosophila by control of the scute locus. Genetics, 52, 1137-1151. ROSE, R. W., AND HILLMAN, R. 1973. Multiple gene control of the trna aminoacylating system in Drosophila melanogaster. Genet. Res., Camb., 21, 239-245. SCHARLOO, w. 1962. The influence of selection and temperature on a mutant character (cid) in Drosophila melanogaster. Arch. Neerl. Zool., 14, 431-512. SCHARLOO, w. 1964. Mutant expression and canalization. J'Tature, Lond., 203, 1095-1096. SONDrn, Sc.. 1961. Developmental barriers in a selection experiment. Isfature, Lond., 189, 249-250. SONDRI, K. C. 1962. The evolution of a pattern. Evolution, 16, 186.191. SpICKET'r, s. o. 1963. Genetic and developmental studies of a quantitative character. J'Iature, Land., 199, 870-873. spickxrr, s. G., AND THODAY, j. M. 1966. Regular responses to selection. 3. Interaction between located polygenes. Genet. Res., Camb., 7, 96-121. THODAY, J. M. 1961. Location of polygenes. Xature, Land., 191, 368-3 70. THODAY, 3. M., GIBSON, J. B., AND SPIC5CErr, S. G. 1964. Regular responses to selection. 2. Recombination and accelerated response. Genet. Res., Camb., 5, 1-19. THOMPsON, j. N., JR. 1973. General and specific effects of modifiers of mutant expression. Genet. Res., Camb., 22, 211-215. THOMPSON, j. N., JR. 1974a. A new suppressor of veinlet and comments on two other vein mutants. Drosophila Information Service 50, 59. THOMPsON, j. N., JR. l974b. Studies on the nature and function of polygenic loci in Drosophila. II. The subthreshold wing vein pattern revealed in selection experiments. Heredity, 33, 389-401.

POLYGENIC LOCI IN DROSOPHILA 387 THOMPSON, j. N., JR., AND ThODAY, J. M. 1972. Modification of dominance by selection in the homozygote. Heredity, 29, 285-292. WADDINOTON, C. H. 1939. Preliminary notes on the development of the wings in normal and mutant strains of Drosophila. Proc..Watl. Acad. Sci., U.S.A., 25, 299-307. WADDINGTON, C. H. 1940. The genetic control of wing development in Drosophila. J. Genet., 41, 75-139. WADDINGTON, C. H. 1953. Genetic assimilation of an acquired character. Evolution, 7, 118-126.