The Genetic Architecture of Drosophila Sensory Bristle Number

Corresponding author: Trudy F. C. Mackay, Campus Box 7614, North Carolina State University, Raleigh, NC 27695-7614., trudy_mackay{at}ncsu.edu (E-mail)

Communicating editor: J. B. WALSH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have mapped quantitative trait loci (QTL) for Drosophila mechanosensory bristle number in six recombinant isogenic line (RIL) mapping populations, each of which was derived from an isogenic chromosome extracted from a line selected for high or low, sternopleural or abdominal bristle number and an isogenic wild-type chromosome. All RILs were evaluated as male and female F₁ progeny of crosses to both the selected and the wild-type parental chromosomes at three developmental temperatures (18°, 25°, and 28°). QTL for bristle number were mapped separately for each chromosome, trait, and environment by linkage to roo transposable element marker loci, using composite interval mapping. A total of 53 QTL were detected, of which 33 affected sternopleural bristle number, 31 affected abdominal bristle number, and 11 affected both traits. The effects of most QTL were conditional on sex (27%), temperature (14%), or both sex and temperature (30%). Epistatic interactions between QTL were also common. While many QTL mapped to the same location as candidate bristle development loci, several QTL regions did not encompass obvious candidate genes. These features are germane to evolutionary models for the maintenance of genetic variation for quantitative traits, but complicate efforts to understand the molecular genetic basis of variation for complex traits.

SUSCEPTIBILITY to common human diseases and response to pharmacological therapy; traits of agronomic importance; morphological, physiological, and behavioral traits; and adaptive traits are all genetically complex. Phenotypic variation for these quantitative traits is attributable to the segregation of multiple quantitative trait loci (QTL) with individually small effects, whose expression is conditional on the environment. A comprehensive understanding of the "genetic architecture" of any quantitative trait would include the list of all genes affecting variation in the trait; estimates of their additive, dominance, epistatic, and pleiotropic effects and environmental sensitivities; and the molecular definition of QTL alleles.

Despite the importance of determining the genetic and environmental factors affecting variation in quantitative traits to medicine, agriculture, and an understanding of the evolutionary process, no quantitative trait is currently understood at this level of detail. The greatest chance of success is likely to come from the study of complex traits for which genes have been identified that are necessary for the production of the trait phenotype and in a model organism susceptible to genetic manipulation with genetic and genomic resources. One such system is the number of mechanosensory bristles in Drosophila melanogaster (MACKAY 1995 , MACKAY 1996 ). Drosophila bristles are classic quantitative traits. Bristle numbers have been used to check short-term (CLAYTON et al. 1957A ; FRANKHAM 1968 ) and long-term (CLAYTON and ROBERTSON 1957 ; JONES et al. 1968 ) selection theory and to assess the contribution of new mutations to quantitative variation (KEIGHTLEY et al. 1993 ), and they were the first traits for which QTL were localized by introgression (BREESE and MATHER 1957 ) and interval mapping combined with progeny testing (THODAY 1979 ; SHRIMPTON and ROBERTSON 1988 ). Further, bristles are external sensory organs of the peripheral nervous system (PNS), and many loci required for PNS development have been characterized (CAMPOS-ORTEGA 1993 ; JAN and JAN 1993 ; KANIA et al. 1995 ; SALZBERG et al. 1997 ).

The first step toward determining the genetic basis of variation for quantitative traits is a genome scan to localize and estimate effects of QTL by linkage to polymorphic molecular markers. Such studies have been facilitated in the past decade by the development of new statistical methods (LYNCH and WALSH 1998 ), the discovery of abundant polymorphic molecular markers, and the development of rapid methods to determine marker genotypes. Genome scans are limited to detecting QTL that vary between the parental strains used to construct the mapping populations. Therefore, recent studies to map QTL for Drosophila bristle number used parental chromosomes that were derived from 25 generations of divergent selection from a large outbred base population and hence captured most alleles that were segregating at intermediate frequency in nature (LONG et al. 1995 ; GURGANUS et al. 1999 ).

To further increase the precision of mapping, NUZHDIN et al. 1999 used an advanced intercross design (DARVASI 1998 ) to derive >250 recombinant isogenic lines (RILs) from the same selected chromosomes in the background of Samarkand, an unselected strain. NUZHDIN et al. 1999 scored the homozygous RILs for abdominal and sternopleural bristle number and cytological locations of roo transposable element markers and mapped at least 26 QTL for sensory bristle number. Here, we use the same RILs developed by NUZHDIN et al. 1999 to estimate homozygous, heterozygous, and epistatic effects of QTL for sensory bristle number in multiple-temperature environments. This study gives a comprehensive picture of the genetic architecture of Drosophila bristle number and reveals strong sex, environment, and background genotype dependency of QTL effects.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mapping populations:
Six populations of RILs were derived from isogenic X chromosomes selected for high (H) and low (L) sternopleural bristle number (GURGANUS et al. 1999 ) and from isogenic third chromosomes selected for H and L sternopleural (GURGANUS et al. 1999 ) and abdominal bristle number (LONG et al. 1995 ). Briefly, each of the six selected chromosomes was backcrossed for three generations to the wild-type highly inbred Samarkand (Sam; LYMAN et al. 1996 ) strain. After 5–10 further generations of random mating, individual isogenic chromosomes were extracted from each population and genotyped for insertion sites of roo elements (see below). Recombinant isogenic chromosomes were selected (the majority of the extracted chromosomes were nonrecombinant Sam). Although these lines had been backcrossed for a total of five generations to Sam, the background was further purified by three additional generations of Sam backcrosses. The six mapping populations (numbers of RILs) were C1, HST (46); C1, LST (44); C3, HST (48); C3, LST (34); C3, HAB (49); and C3, LAB (56), where C refers to the selected parental chromosome and ST and AB refer to the selected trait (sternopleural or abdominal bristle number, respectively). For further details of the origin of these lines, see NUZHDIN et al. 1999 .

The individuals scored for this experiment were the offspring of crosses of females from each of the RILs within a population to Sam (S) males and to males with the appropriate H or L selected parental chromosome (P) in the Sam background. The double backcross design is essentially a North Carolina III design for RILs (LEIPS and MACKAY 2000 ). It has the advantage that all three genotypes (SS, SP, and PP) are produced at bristle number QTL (SS and SP in the offspring of the cross to Sam, and PP and SP in the offspring of the cross to the selected parental line), enabling estimation of degrees of dominance of QTL.

Marker genotypes:
Cytogenetic insertion sites of roo transposable element were determined for each of the RILs by in situ hybridization of biotin-labeled roo DNA to polytene salivary chromosomes of third instar larvae, exactly as described previously (NUZHDIN et al. 1999 ). Markers were informative provided at least one recombination event was observed between adjacent markers. The informative markers and their map positions in centimorgans, estimated from observed recombination frequencies between adjacent markers, r, using the Kosambi map function, d = ¹/₄ln[(1 + 2r)/(1 - 2r)], where d is the distance between adjacent markers in morgans, are given for each population in Table 1. The first and last marker for each genotype map is the first and last marker where recombination is seen and represents the entire tip or end of the chromosome. All genetic maps were expanded in relation to the standard map (LINDSLEY and ZIMM 1992 ) as expected given multiple opportunities for recombination.

Bristle number phenotypes:
Two replicate vials of progeny from crosses between each RIL and both parental lines were reared at 18°, 25°, and 28°. Abdominal (total number of bristles on the fifth and sixth abdominal sternite of males and females, respectively) and sternopleural [total number of bristles on the left (L) and the right (R) sternopleural plates] bristle numbers were scored on 10 males and 10 females in both replicate vials for all RILs, for a total of 40 individuals scored per line per cross per temperature environment.

Phenotypic analyses:
Analysis of variance (ANOVA) was used to partition the variance in bristle traits among the RILs for each population into sources attributable to line (L, random), sex (S, fixed), temperature (T, fixed), and cross (C, fixed—Sam or selected parent) according to the model

where µ is the overall mean, R refers to replicate vials, E is the within-vial error variance, and parentheses represent the nesting of an effect. Reduced models were also run within temperatures and sexes, pooled over crosses. Tests of significance of F-ratios and estimates of variance components were computed using SAS procedures GLM and VARCOMP (SAS INSTITUTE 1988).

QTL mapping:
Composite interval mapping (ZENG 1994 ), as implemented by the QTL Cartographer software (BASTEN et al. 1994 , BASTEN et al. 2000 ), was used to test the hypothesis that an interval flanked by two adjacent markers contains a QTL affecting the trait, while simultaneously controlling for the effects of chromosomally linked QTL by multiple regression on markers outside the test interval. These analyses were conducted on line means separately for abdominal and total sternopleural bristle number, sex, and temperature. The conditioning markers were chosen for each analysis by forward selection-backward elimination stepwise regression. A conditioning window size of 10 cM was used, such that only markers 10 cM away from the markers flanking the test interval were included in the model. The likelihood-ratio (LR) test statistic is -2 ln(L₀/L₁), where L₀/L₁ is the ratio of the likelihood under the null hypothesis (there is no QTL in the test interval) to the alternative hypothesis (these is a QTL in the test interval). The test statistic at a genome location is distributed as ² with 2 d.f. under the null hypothesis and was evaluated every centimorgan.

The significance level for each analysis to infer the presence of a QTL was determined by permutation. Empirical distributions of LR test statistics under the null hypothesis of no association between test intervals and trait values were obtained for each analysis by randomly permuting the trait data 1000 times and calculating the maximum LR statistic across all intervals for each permutation. LR statistics calculated from the original data that were exceeded by the permutation maximum LR statistics <50 times are significant at = 0.05 under the null hypothesis (CHURCHILL and DOERGE 1994 ; DOERGE and CHURCHILL 1996 ).

QTL effects:
Estimates of additive effects [a = (PP - SS)/2] and dominance deviations [d = SP - (SS + PP)/2] of each bristle number QTL within each sex, temperature, trait, and population are provided by the QTL mapping analyses (FALCONER and MACKAY 1996 ). To facilitate comparison across traits, environments, and populations, additive effects, within each trait and population, were scaled by dividing the effect by the square root of the variance component among lines from the reduced-model analyses by sex and temperature. The dominance deviations were scaled by the additive effects (d/a = degree of dominance).

To determine whether there were significant QTL x temperature and QTL x sex interactions, within a trait and population, the genotypes of the closest marker to each significant QTL peak, as determined by composite interval mapping, were used as categorical variables in ANOVA. A total of i separate analyses were carried out for each significant marker in turn, using the model

where µ is the overall mean, M_i are the significant markers, S refers to sex, T refers to temperature, E is the error variance, and M_x refers to the marker for which interactions are evaluated with all effects being fixed. Significance levels were determined by Bonferroni correction for multiple testing on the basis of the number of marker interaction tests (0.05/i) within each cross.

The significance of pairwise epistatic interactions between nonadjacent bristle number QTL detected by composite interval mapping analyses was evaluated by ANOVA on line means. A more sophisticated method (multiple-interval mapping) has been developed to identify epistasis among QTL (KAO et al. 1999 ) but this requires fitting many parameters simultaneously and is not applicable to the numbers of RILs in these mapping populations. Tests were carried out within each cross separately: Although each bristle number QTL could be either homozygous or heterozygous, both classes of homozygous (SS and PP) did not occur together within a line because of the breeding design used. For each interaction of interest, the genotype of each marker (SS, SP, or PP) closest to each significant QTL peak was used to evaluate the significance of the marker interactions on bristle number. For each test, the model

was fitted. It included µ as the overall mean; the main effects of sex (S), temperature (T), and all bristle number QTL (M_i) identified by the QTL mapping analyses within a bristle trait and population; the two-way interaction term (M_i x M_j) for the two focal marker genotypes; all possible interactions among the two focal markers, sex, and temperature; and the error variance (E). The three- and four-way interactions were used to investigate sex- and environment-specific epistatic interactions between pairs of QTL. Significance levels were determined by Bonferroni correction for multiple testing on the basis of the number of interaction tests within each cross.

The effects of significant two-locus interactions were estimated from the least-squares means of the four marker locus classes as [(₁₁ + ₂₂) - (₁₂ + ₂₁)], where the first subscript is 1 if the marker has a homozygous genotype for either parental strain and 2 if the marker has a heterozygous genotype, and the second subscript takes on the same values for the other marker in the interaction. With no epistasis between markers, this term should approach zero. Standard errors of the interaction effects were estimated as [MSI/(n₁₁ + n₂₂ + n₁₂ + n₂₁ - 4)]^1/2, where MSI is the interaction mean square for the ANOVA model and n is the number of lines in each of the four marker classes as described above.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Quantitative genetic analysis:
ANOVA of bristle number within each of the six RIL mapping populations showed highly significant variation among lines for abdominal and sternopleural bristle number in all populations, sexes, and environments (data not shown). The results of full ANOVA models partitioning the variance in bristle number within each mapping population into sources attributable to temperature, cross, sex, and line and all of the interactions of these main effects are given in Table 2 and Table 3. All populations also exhibited significant genotype x environment (V_LT), genotype x sex (V_LS), and/or genotype x cross (V_LC) interactions. Significant V_LT and V_LS terms mean the effects of bristle number genes vary according to temperature and sex, respectively (FALCONER and MACKAY 1996 ). A significant V_LC term indicates nonadditive (dominance and/or epistasis) effects of genes affecting bristle number. The proportion of the variation attributable to main effects and the various interactions for each population is shown for sternopleural bristle number in Fig 1 and abdominal bristle number in Fig 2. There was a great difference in the amount and type of variation seen among the populations for both bristle traits. Comparing the among-line component of genetic variation (V_L) to the total genetic variance, including interaction terms (V_L + V_LT + ... + V_TLCS), across both bristle traits and averaged over all six populations, showed that 30% of the total genetic variance in these populations was due to interactions (dominance, epistasis, sex, temperature). However, this ratio ranged from a low of 15% in C3, HST to a high of 34% in C3, LAB for sternopleural bristle number and from 22% in C1, LST to 57% in C3, LST for abdominal bristle number. Thus, the quantitative genetic analyses indicate that one should expect QTL mapping to reveal bristle number QTL with additive, dominance, and epistatic effects that can be conditional on sex and rearing environment.

View larger version (79K): In this window In a new window Download PPT slide   Figure 1. Partitioning of variance for sternopleural bristle number.

View larger version (78K): In this window In a new window Download PPT slide   Figure 2. Partitioning of variance for abdominal bristle number.

QTL mapping:
QTL affecting sternopleural and abdominal bristle number in each sex and temperature were mapped by linkage to molecular markers in each population using a composite interval mapping method (ZENG 1993 , ZENG 1994 ). The results are summarized in the supplementary table at http://www.genetics.org/supplemental/ and in Fig 3 Fig 4 Fig 5. The QTL that are reported are those for which the LR test statistic exceeded the permutation threshold. Two-LOD support intervals are given as 95% confidence intervals for the location of each QTL (LYNCH and WALSH 1998 ). If a QTL range overlapped the neighboring QTL range and had approximately the same effect, a single QTL at the position associated with the highest LR test statistic was assumed.

View larger version (23K): In this window In a new window Download PPT slide   Figure 3. QTL for the C1 population. The length of each chromosome is proportional to the total map length in centimorgans. Black circles represent roo transposable element markers for each population. Lines between chromosomes adjoin common markers between the populations. Lines between QTL represent epistatic interactions between QTL that were significant by Bonferroni correction. Lines through QTL represent the 2-LOD support interval. Red circles correspond to QTL with main effects only. Blue triangles correspond to QTL with main effects and significant QTL x sex interactions. Green triangles correspond to QTL with main effects and significant QTL x temperature interactions. (A) Sternopleural bristle number QTL. (B) Abdominal bristle number QTL.

View larger version (34K): In this window In a new window Download PPT slide   Figure 4. QTL for the C3 sternopleural population. See Fig 3 legend for explanation of symbols used. (A) Sternopleural bristle number QTL. (B) Abdominal bristle number QTL.

View larger version (35K): In this window In a new window Download PPT slide   Figure 5. QTL for the C3 abdominal population. See Fig 3 legend for explanation of symbols used. (A) Sternopleural bristle number QTL. (B) Abdominal bristle number QTL.

We mapped 144 QTL for sternopleural and abdominal bristle number, pooled over all six mapping populations, three temperature environments, and two sexes (supplementary table at http://www.genetics.org/supplemental/). Considering the additive and dominance effects of each QTL, the interaction of the markers with sex and temperature from analysis of variance, epistatic interactions between markers from analysis of variance, and the effects of the markers in the correlated trait led to the elimination of two QTL due to similarities in effects to neighboring QTL, giving a total of 142 QTL.

Additive and dominance effects:
Additive effects and dominance deviations of each QTL were estimated while fitting marker cofactors for the appropriate model in each analysis (BASTEN et al. 1994 , BASTEN et al. 2000 ). The distribution of additive QTL effects, scaled by the additive genetic standard deviation appropriate for each population/trait/sex/environment combination, is given in Fig 6. The effects are clearly exponentially distributed, with a few QTL of major effect and increasingly more with smaller effects, up to the limit of detection imposed by the sample size of the experiment. This observation is in accordance with the prediction of ROBERTSON 1967 and previous observations from high-resolution mapping of bristle number QTL (SHRIMPTON and ROBERTSON 1988 ).

View larger version (29K): In this window In a new window Download PPT slide   Figure 6. Distribution of additive QTL effects.

Observed exponential distributions of QTL effects could reflect an underlying exponential distribution of effects of the actual genes corresponding to the QTL. However, an underlying infinitesimal distribution of gene effects could also lead to an apparent exponential distribution of QTL effects, since overestimation of QTL effects could be an artifact of small sample size (BEAVIS 1994 ), linkage disequilibrium between QTL, low heritability, or epistatic interactions between QTL (BOST et al. 2001 ). Further, observed unequal distributions of QTL effects could reflect variation in the precision of estimating QTL positions.

Under the infinitesimal model, the magnitude of QTL effects would be directly proportional to the physical size of the interval containing the QTL. This hypothesis was tested by plotting the absolute values of QTL effects (scaled by the additive genetic standard deviation) against the genome size in kilobases. Cytological positions of roo element markers spanning the 95% confidence interval (2-LOD support interval) were noted, and the DNA content between the markers was calculated using the table of SORSA 1988 . These data are plotted in Fig 7. The product-moment correlation between QTL effect and the size of the genomic region to which the QTL mapped was 0.075. Clearly, these data do not support the infinitesimal model.

View larger version (15K): In this window In a new window Download PPT slide   Figure 7. Distribution of additive QTL effects based on the size of the QTL region.

The distribution of degrees of dominance of QTL is given in Fig 8. These span the range from strictly additive through complete dominance (recessivity), with most QTL partially dominant (recessive). There are a few cases where the value of the heterozygote lies outside the range of the homozygotes, which are most parsimoniously attributable to sampling error in estimating effects.

View larger version (20K): In this window In a new window Download PPT slide   Figure 8. Distribution of degrees of dominance of QTL.

The signs of the QTL effects were mostly in the expected direction for the selected trait, with some exceptions. In the C1, HST population, the Sam alleles had higher sternopleural bristle numbers for the two QTL mapped at 25° (the temperature at which selection was carried out). Possibly the divergence in sternopleural bristle number between the "high" and "low" selected X chromosomes was attributable to response for reduced bristle number only. There were also QTL with effects opposite to the direction of selection at 25°, in the C3, HST population, closely linked to QTL for high sternopleural bristle number relative to Sam on the selected chromosome.

QTL interactions with temperature and sex:
Many of the QTL significantly exceeded the permutation threshold in only one sex and environment, while others were significant in males and females or in more than one temperature. However, one cannot infer if there are significant QTL x sex or QTL x temperature interactions simply because a QTL was not significant in both sexes (all temperatures), since small differences in effect could lead to detection in one sex (environment) but not the other. Conversely, the presence of significant QTL across sexes (environments) does not mean there was no interaction, since effects could be of different magnitude or direction.

ANOVA was used to formally test for QTL x sex and QTL x temperature interactions, separately for each population and bristle trait. The ANOVA models included all markers closest to the peak LR for the significant QTL, plus the marker x sex, marker x temperature, and marker x sex x temperature interaction terms for the focal marker. Since a separate ANOVA was run for each focal marker in turn, the significance threshold for each trait was adjusted downward by a Bonferroni correction for the number of markers. Multiple-trait composite interval mapping (JIANG and ZENG 1995 ) is a more elegant method for formally testing interactions, but these models did not stabilize to global maxima for all populations, particularly those with many significant QTL. Presumably these models were overparameterized, given the number of RILs.

Summed over all populations, there were 42 QTL affecting sternopleural bristle number and 38 QTL affecting abdominal bristle number. These QTL exhibited a far greater number of interactions with sex (54/80 = 67.5%) and temperature (47/80 = 58.8%) than expected by chance, given a nominal 5% significance threshold. However, none of the three-way marker x sex x temperature interactions were even nominally significant. After Bonferroni corrections for multiple tests, 8/42 (19.0%) QTL for sternopleural bristle number exhibited significant QTL x sex interactions, and 14/42 (33.3%) exhibited significant QTL x temperature interactions (Table 4, Fig 9A). The QTL x sex interaction term was highly significant for all the abdominal bristle number QTL. A total of 21 (55.3%) of these QTL also exhibited significant interactions with temperature (Table 4, Fig 9B).

View larger version (20K): In this window In a new window Download PPT slide   Figure 9. Examples of (A) conditional neutrality for sternopleural bristle number in the C3, LST population and (B) antagonistic pleiotropy for abdominal bristle number in the C3, HAB population.

The nature of the interactions was assessed by comparing the stability of the genotypes for each QTL x sex and QTL x temperature interaction. GILLESPIE and TURELLI 1989 hypothesized, on the basis of a simple model of additive polygenic inheritance, that the variance of the phenotypes produced across environments by a multilocus genotype would decrease as the number of heterozygous loci increased. This genotype x environment interaction would then lead to the maintenance of genetic variation for a trait under stabilizing selection. These data afford an opportunity to test this hypothesis directly. The difference in bristle number between females and males was computed for the heterozygous genotypes, for each of the QTL exhibiting sex-specific effects (i.e., 8 sternopleural bristle number and 38 abdominal bristle number QTL). The heterozygotes were most stable in six cases; otherwise one or the other homozygote was the more stable genotype. Interestingly, the six cases where the heterozygote was the most stable were all in low sternopleural bristle number populations.

Since bristle number decreases with increasing temperatures, the slope of the regression of bristle number on temperature was computed for each of the three genotypes for the 35 QTL exhibiting significant interactions with temperature. Again, the heterozygote was rarely (5/35) the most stable genotype. While these data show GILLESPIE and TURELLI's (1989) hypothesis does not pertain in general to QTL affecting bristle number, there appear to be some QTL to which this mechanism could apply.

Each of the significant QTL x sex and QTL x temperature interactions was classified as exhibiting either conditional neutrality (genotypes having the same mean bristle number in one sex/environment but are different in the other sex/environment) or antagonistic pleiotropy (genotypes having opposite effects in different sexes/environments, leading to a crossing of reaction norms). Examples of conditional neutrality and antagonistic pleiotropy are given in Fig 9. For sternopleural bristle number, all of the QTL x sex interactions and 11 (78.6%) of the QTL x temperature interactions exhibited conditional neutrality. For abdominal bristle number, 22 (57.9%) of the QTL x sex interactions and 20 (95.2%) of the QTL x temperature interactions exhibited conditional neutrality.

In summary, interactions with sex and environment were common for QTL affecting both sternopleural and abdominal bristle number, but were more frequent for the latter trait. Interactions with environment generally showed conditional neutrality for both bristle traits. Interactions with sex showed conditional neutrality for sternopleural bristle number, but both conditional neutrality and antagonistic pleiotropy were observed for abdominal bristle number.

QTL x QTL interactions:
Detection of epistatic (QTL x QTL) interactions between genotypes at two loci was conducted using ANOVA models that fitted the effects of all significant QTL, a single pairwise interaction between QTL, and interactions of pairs of QTL with temperature and sex. The number of epistatic interactions that were significant at P = 0.05 greatly exceeded that expected by chance: 133/334 (39.8%) for sternopleural bristle number and 68/228 (29.8%) for abdominal bristle number. After Bonferroni correction, 30 significant interactions (9.0%) were affecting sternopleural bristle number and 25 significant interactions (11.0%) were affecting abdominal bristle number (Table 5, Fig 10). The QTL x QTL x sex term was significant only once for sternopleural bristle number and not at all for abdominal bristle number. None of the QTL x QTL x temperature terms and four-way QTL x QTL x sex x temperature interactions was significant for either bristle trait.

View larger version (17K): In this window In a new window Download PPT slide   Figure 10. Examples of (A) diminishing epistasis for RI lines crossed to C3, HST and (B) synergistic epistasis for RI lines crossed to Sam for sternopleural bristle number in the C1, HST population.

The nature of the marker x marker interactions was investigated by determining whether the epistasis was "diminishing" or "synergistic" (MACKAY 2001 ), where diminishing refers to less-than-additive and synergistic to greater-than-additive epistatic interactions between two loci. A less-than-additive interaction means that the difference between the high and low genotypes at locus two is smaller in the more extreme genotype of locus one (Fig 10A), whereas the reverse is true for greater-than-additive interactions (Fig 10B). The nature of the epistatic interactions was evaluated only for the directly selected trait of each population. Summed over all populations, 26 significant epistatic interactions were affecting sternopleural bristle number, of which 22 (84.6%) were less than additive. There were 13 significant epistatic interactions affecting abdominal bristle number, of which 8 (61.5%) were greater than additive. Interestingly, one expects diminishing epistasis to occur between loci affecting traits under stabilizing selection, leading to genetic canalization (WAGNER et al. 1997 ).

Total number of QTL:
To estimate the minimum number of QTL affecting variation in sensory bristle number, we inferred which QTL mapped to the same location in different populations by comparing the marker maps and QTL locations. QTL were considered to be the same across populations if the peak LR coincided at the same marker and if the 2-LOD support intervals for the QTL overlapped. QTL that mapped to close but different cytological locations in two populations were considered to be the same if there were no markers between them and if they affected the same bristle trait (or at least one trait if a QTL was found for both bristle traits). According to these criteria, a minimum of 53 QTL affecting sensory bristle number were detected, 11 on the X chromosome and 42 on chromosome 3.

Three of the third-chromosome QTL that mapped to the same location were considered to be different because they affected different bristle traits. The QTL at 61B in the C3, HST population affected sternopleural bristle number and the one at 61A in the C3, HAB population affected abdominal bristle number; the QTL at 89A affected abdominal bristle number in the C3, LST population and sternopleural bristle number in the C3, LAB population; and the QTL at 91B affected sternopleural bristle number in the C3, HST population and abdominal bristle number in the C3, LAB and C3, HAB populations. However, LONG et al. 1998 ; LONG et al. 2000 showed that there were at least two different alleles at Delta and the achaete-scute complex segregating in nature, one of which affected abdominal bristle number, and the other of which affected sternopleural bristle number. Therefore, it is possible that these QTL should be considered to be the same, reducing the total minimum number to 50. Conversely, 3 sternopleural bristle number QTL (62D–E, 69F–70C, and 83A–83F) and 2 abdominal bristle number QTL (93D–93F and 98B–98C) were considered to be the same, although the cytological locations did not exactly coincide.

The number of QTL detected is a minimum because (1) QTL were not mapped on the second chromosome; (2) further recombination with a higher density of markers can separate linked QTL; (3) increasing the number of individuals scored per RIL enables detection of QTL with smaller effects; and (4) QTL that segregate only between the parental chromosomes used to establish the mapping populations can be mapped. Scaling the estimate of 53 QTL on the X and third chromosomes, which together represent 60% of the genome, to the whole genome thus yields a minimum of 88 bristle number QTL. The effects of points 3 and 4 above are minimized in this study, since sample sizes were large, and the selected chromosomes were likely to represent most alleles at intermediate frequency in nature, since they were derived from large selection lines in turn derived from a large base population (LONG et al. 1995 ; GURGANUS et al. 1999 ).

The locations and effects of the 53 QTL for sensory bristle number are listed in Table 6. The QTL effects show whether the QTL had a main effect only or interacted with sex and/or temperature.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In this study, QTL for mechanosensory bristle number have been mapped in six RIL mapping populations, each of which was derived from an isogenic chromosome extracted from a line selected for high or low, sternopleural or abdominal bristle number and an isogenic wild-type chromosome. All RILs were evaluated as male and female F₁ progeny of crosses to both the selected and the wild-type parental chromosomes at three developmental temperatures (18°, 25°, 28°). Previously, bristle number QTL were mapped in the homozygous RIL, reared under standard (25°) laboratory conditions (NUZHDIN et al. 1999 ). These data enable: (1) comparisons of the consistency of QTL locations in two independent experiments using the same mapping populations, (2) determination of the extent to which the same loci are associated with response to selection for increased and decreased values of the trait, (3) assessment of the extent to which correlated response to selection is attributable to linkage or pleiotropy, (4) evaluation of the genetic architecture of bristle number, and (5) insight as to potential mechanisms maintaining genetic variation for bristle number in nature.

Consistency:
A total of 70 QTL were detected at 25° in this study and that of NUZHDIN et al. 1999 , pooling across the separate analyses by bristle trait and sex. Of these, only 25 (36%) were common to both experiments (that is, the 2-LOD support intervals overlapped). A total of 20 QTL were unique to the experiment of NUZHDIN et al. 1999 ; however, 5 of these were also detected in this study at a different temperature. A total of 25 QTL were unique to this study; of these, six were significant by a ² criterion, but not the more stringent permutation threshold, in the study of NUZHDIN et al. 1999 . If the latter two categories are counted as a match, then 36 QTL (51%) are common to the two experiments. However, one could argue that a larger number of QTL ought to be detected in this study since the sample size was doubled.

Perhaps a fairer evaluation of the extent to which the same QTL were detected in both experiments is to ask what fraction of the total number of QTL found by NUZHDIN et al. 1999 was also detected at 25° in this study. NUZHDIN et al. 1999 mapped a total of 45 QTL, again pooling across the separate analyses by bristle trait and sex. Of these, 25 were also found in this study; and an additional 7 and 5 were found here in adjacent intervals and different temperatures, respectively. Thus, the locations of 56% of the QTL match exactly. If one includes those in adjacent intervals, the concordance is 71%, and if one also includes additional temperature treatments, the agreement between the two experiments rises to 82%.

The agreement between the two experiments is good, but there are clearly discrepancies. There are multiple potential, and nonmutually exclusive, explanations for these differences:

1. 1. The mapping populations were not identical. Several of the RILs were homozygous lethal and could not be scored by NUZHDIN et al. 1999 . However, these lines did produce viable F₁ progeny when crossed to the parental chromosomes and were included in this study. Thus, the sample size of this study was slightly increased, and the recombination maps were not identical.

2. Although the permutation test controls the overall Type I error, it does not eliminate false positives. Therefore, one expects some of the QTL in each study to be false positives.

3. Epistasis between QTL is not explicitly considered in the composite interval mapping algorithm. Extensive epistatic interactions between bristle number QTL were observed in this study; therefore, estimates of QTL locations and effects from the composite interval analyses, ignoring these interactions, are biased. Further, the genetic backgrounds of the two experiments differed, such that only additive x additive epistasis occurred in the populations of NUZHDIN et al. 1999 , whereas additive x additive, additive x dominant, dominant x additive, and dominant x dominant interactions potentially contributed to the epistatic interactions of this study.

4. QTL mapping is an exercise in model selection. QTL locations and estimates of effects can vary according to the marker cofactors that are fitted to control for the genetic background, even for the same data set (PASYUKOVA et al. 2000 ). Part of the observed difference could thus be due to differences between the models that were fitted in each case.

These issues highlight the importance of confirming QTL detected in such mapping experiments by independent methods.

Response to selection:
Comparisons across pairs of high and low selected populations revealed that QTL contributing to the response to selection for low bristle number are usually not the same as those contributing to selection response for high bristle number (Table 6). A total of 7 QTL affecting sternopleural bristle number were mapped in the C1, LST and C1, HST populations; of these, only 1 QTL (at 2A–2B) was common to the two populations. Of the 25 QTL for sternopleural bristle number mapped in the C3, LST and C3, HST populations, 3 QTL were common to both (at 69F–70C, 75B–75C, and 87A). Similarly, 2 abdominal bristle number QTL (at 77B–77E and 91B) out of a total of 18 were detected in the C3, LST and C3, HST populations. Thus, only 12% of QTL were associated with both high and low selection response.

Two possible scenarios could explain these results. If the distribution of allelic effects is symmetrical, then these data are consistent with low frequencies of QTL alleles with high and low effects on bristle number, such that the probability of sampling both the high allele in the high selection line and the low allele in the low line is low. Alternatively, it is possible that the distribution of allelic effects is asymmetrical at the majority of loci contributing to selection response, such that there exists a common allele with either a high or a low effect on bristle number relative to other wild-type alleles segregating at the locus. The latter hypothesis is consistent with experiments showing that response to selection from single-pair bottleneck populations was little diminished relative to response from large base populations, indicating that most QTL alleles are at intermediate frequency in nature (ROBERTSON 1968 ; FRANKHAM 1980 ). These hypotheses are not mutually exclusive: LONG et al. 2000 showed that both low and intermediate frequency molecular polymorphisms at the achaete-scute complex were associated with phenotypic variation in abdominal bristle number.

Correlated responses:
Of the 53 QTL, 13 (24.5%, Table 6) affected both direct and correlated responses to selection. In these populations, therefore, approximately three-quarters of the QTL contributing to correlated selection response did so by hitchhiking along with linked selected loci. The remaining one-quarter of the QTL for which direct and correlated responses mapped to the same region could be attributable to pleiotropy. However, pleiotropy cannot be distinguished from linkage as causing a genetic correlation until a single molecular polymorphism is shown to affect both traits. As noted above, two separate molecular polymorphisms in Delta (LONG et al. 1998 ) and the achaete-scute complex (LONG et al. 2000 ) affected sternopleural and abdominal bristle number. The polymorphic sites were in each case at intermediate frequency and in linkage equilibrium; therefore, correlated selection response attributable to each locus could vary, depending on the initial gene frequencies of each polymorphic site in the base population. These observations go a considerable way toward explaining why correlated responses to selection for abdominal and sternopleural bristle number are notoriously difficult to predict, given estimates of genetic correlations in the base population, and poorly replicable (e.g., CLAYTON et al. 1957B ).

Genetic architecture of bristle number:
Early experiments to map QTL using the same selected chromosomes as this study concluded that a few QTL with moderately large effects might account for the majority of the divergence among the selected chromosomes (LONG et al. 1995 ; GURGANUS et al. 1999 ). The picture emerging from this study and that of NUZHDIN et al. 1999 is rather more complex, with potentially >80 QTL in the whole genome affecting the two bristle traits. This inference is more in line with the estimate of 18 QTL for sternopleural bristle number on chromosome 3 reached by SHRIMPTON and ROBERTSON 1988 . Part of the reason for the increase in total number of QTL is the increased opportunity for recombination afforded by the advanced generation intercross design used here. The other factor is the extent to which QTL are environment specific. Of the 43 QTL for sternopleural bristle number, 14 (32.6%) exhibited genotype-by-environment interaction (GEI), and 11 were detected in only one of the three temperature environments. Of the 38 QTL for abdominal bristle number, 21 (55.3%) exhibited GEI, of which 20 were detected in only one environment. This raises the specter of how many more QTL might be detected if the animals were reared in additional environments and the question of what QTL are responsible for genetic variation in bristle number in the range of environments we call "nature."

Bristle number QTL also exhibit strong sex-specific effects (genotype-by-sex interaction, GSI). This phenomenon is particularly evident for abdominal bristle number, where all QTL had sex-specific effects: 22 of the 38 QTL (57.9%) were detected in only one of the two sexes, and the remainder had opposite effects in males and females. Only 8 of the sternopleural bristle number QTL exhibited GSI; all were detected in only one sex. FRANKHAM 1968 first observed autosomal sex-specific effects for abdominal bristle number; later studies have repeatedly documented this phenomenon for naturally occurring variation (LYMAN and MACKAY 1998 ), spontaneous mutations (MACKAY et al. 1995 ), P-element-induced mutations (MACKAY et al. 1992 ; LYMAN et al. 1996 ), QTL (LONG et al. 1995 ; GURGANUS et al. 1998 , GURGANUS et al. 1999 ; NUZHDIN et al. 1999 ; this study), and even molecular polymorphisms in candidate genes associated with naturally occurring variation in bristle number (LONG et al. 1998 , LONG et al. 2000 ).

The picture painted by decades of biometrical analyses is of sensory bristle numbers as archetypical quantitative traits, with predominantly additive and little nonadditive (dominance and epistatic) genetic variance (FALCONER and MACKAY 1996 ). This study shows that this perception is inaccurate and likely a consequence of the full- and half-sib designs typically used to estimate quantitative genetic parameters in outbred populations. These designs have little power to detect dominance and epistatic variance (FALCONER and MACKAY 1996 ). When more sophisticated methods are employed, such as the North Carolina III design used here, there is ample evidence for nonadditive effects.

In particular, epistatic interactions between bristle number QTL are large and pervasive. Further, the magnitude of epistasis was likely underestimated because a conservative Bonferroni correction was used to declare significant interactions and because interactions were tested only between markers closest to QTL that were themselves significant. The single-marker analysis always underestimates effects of both QTL and interactions between QTL by an amount that is a function of the distance of the QTL from the marker (FALCONER and MACKAY 1996 ). QTL that themselves did not reach the permutation significance threshold could also interact with significant QTL or with each other. Although it is in principle possible to test for all n(n - 1)/2 interactions between the n markers in any mapping population, in practice this is not feasible since adjusting the significance threshold for the large number of tests means that only exceptionally strong interactions can be detected. Unfortunately, if epistasis is present but unaccounted for, estimates of QTL main effects are biased and must be viewed with caution. True multiple-interval mapping methods that simultaneously fit main effects and interactions (KAO et al. 1999 ) solve this problem, but determining which of many equally likely models fit the data presents another difficult problem (ZENG et al. 1999 ).

Maintenance of quantitative genetic variation:
The number of Drosophila sensory bristles is thought to be under strong stabilizing selection for an intermediate optimum in nature, because mean bristle numbers are relatively stable among natural populations. Efforts to deduce the relationship of bristle numbers to fitness in the laboratory have, however, reached contradictory conclusions, with experiments supporting strong (KEARSEY and BARNES 1970 ; LINNEY et al. 1971 ; NUZHDIN et al. 1995 ), moderate (CLAYTON et al. 1957A ; LATTER and ROBERTSON 1962 ; GARCIA-DORADO and GONZALEZ 1996 ), or very weak stabilizing selection (MACKAY 1985 ).

If bristle numbers are indeed under strong stabilizing selection, it is a mystery why so much genetic variation for these traits persists in the face of selection, which eliminates variation. Of course, some fraction of genetic variance must be due to this mutation-selection balance. However, estimates of the mutational variance for bristle number are too low by an order of magnitude to account for observed levels of segregating variation (BARTON and TURELLI 1989 ; BARTON 1990 ; KEIGHTLEY et al. 1993 ). Further, models for the maintenance of variation by mutation-selection balance predict that equilibrium levels of genetic variance are reached when the mutant allele is at low frequency (BARTON and TURELLI 1989 ; BARTON 1990 ). Common molecular polymorphisms in candidate genes are associated with variation in bristle number (LONG et al. 1998 , LONG et al. 2000 ; LYMAN et al. 1999 ). It is likely that these polymorphisms are associated with a causal variant that is itself at intermediate frequency and not with rare variants with large effects (LONG et al. 1998 ). Since common polymorphisms are not consistent with mutation-selection balance, additional mechanisms must be operating to maintain the observed levels of segregating variation.

GILLESPIE and TURELLI 1989 proposed that intermediate frequency polymorphisms could be maintained for a trait under stabilizing selection by GEI. Their model assumes stabilizing selection for a single phenotype that is optimal in all environments, constant fitnesses across environments, alleles with different additive effects on the trait in different environments, and a high variance of allelic effects across environments relative to the mean difference in allelic effects. Under this model, heterozygotes will have lower phenotypic variance than will homozygotes, and hence higher mean fitness. The mean bristle number of heterozygotes was more stable across environments than that of homozygotes for 5 of the 35 QTL x temperature interactions and 6 of the 46 QTL x sex interactions observed in this study. However, in all of these cases, the mean difference in QTL effects was greater than the variance in effects across environments. GILLESPIE and TURELLI's (1989) model does not, therefore, appear to be a viable explanation for maintaining genetic variation in bristle number. This inference must, however, be tempered by the caveat that a fair test of this model requires that means and variances of QTL effects must be measured in whatever range of environments is relevant to maintaining the observed variation in nature, which is unknown.

Variation for traits under stabilizing selection can also be maintained by epistasis. Stabilizing selection favors reduced phenotypic variation for the selected traits, through environmental and genetic canalization, where the former refers to insensitivity of the phenotype to environmental perturbations and the latter to insensitivity to genetic perturbations (mutations; WAGNER et al. 1997 ). Genetic canalization arises from diminishing epistatic interactions among QTL affecting the canalized trait (WAGNER et al. 1997 ). Diminishing epistasis was observed for 22 of the 26 (84.5%) significant epistatic interactions between QTL affecting sternopleural bristle number, and 5 of the 13 (38.5%) interactions between QTL affecting abdominal bristle number. GIMELFARB 1989 has proposed an even more general model of epistasis leading to maintenance of variation for traits under stabilizing selection. Thus epistatic gene action cannot be ruled out as an explanation for the twin observations of large amounts of segregating variation and apparently strong selection.

Future prospects:
QTL mapping is only the first step toward understanding the genetic basis of any complex trait. Ultimately, one wants to know what genetic loci correspond to the QTL (quantitative trait genes, QTG) and, further, what molecular polymorphisms at these genes cause the quantitative variation in phenotypes. Initial QTL mapping experiments utilizing the same selected chromosomes examined in this study noted that the QTL regions typically encompassed candidate genes affecting peripheral nervous system development (LONG et al. 1995 ; GURGANUS et al. 1998 , GURGANUS et al. 1999 ). Quantitative complementation tests of QTL alleles to mutant and wild-type alleles of the candidate genes often (but not always) implicated allelic variation at the candidate gene, which was associated with the quantitative variation in phenotypes (i.e., the candidate genes were QTG; LONG et al. 1996 ; MACKAY and FRY 1996 ; GURGANUS et al. 1999 ).

The intervals to which bristle number QTL were mapped in this study also typically contained candidate genes with known effects on bristle number development or with bristle number mutant phenotypes (Table 6). However, several QTL regions do not contain obvious candidate genes. Deficiency complementation mapping to narrow the QTL intervals to regions containing small numbers of loci (PASYUKOVA et al. 2000 ), followed by complementation tests for all genes for which mutant alleles are available in all the QTL regions (MACKAY 2001 ), should succeed in identifying candidate QTG for further study. Although environment-specific QTL effects have rarely, if ever, been considered in fine-scale mapping studies, it is clear from these results that it is necessary to do so. This map-based approach is likely not only to confirm that some "obvious" candidate genes harbor naturally occurring variation for bristle number while others do not, but also to identify genes with hitherto unknown pleiotropic effects on bristle number and to elucidate the function of predicted genes.

Since epistatic interactions confound the interpretation of complementation tests to deficiencies and mutations, candidate QTG so identified must be confirmed by independent methods, such as linkage disequilibrium mapping to associate molecular polymorphisms in the candidate genes with phenotypic variation in bristle number (LONG et al. 1998 ; LYMAN and MACKAY 1998 ; LYMAN et al. 1999 ; LONG et al. 2000 ). It will be of particular interest to determine which sites (and presumably regulatory motifs) are associated with sex-, environment-, and genotype-specific expression of QTGs. Polymorphisms with female-specific effects on abdominal bristle number have been detected in noncoding regions of Delta (LONG et al. 1998 ) and the achaete-scute complex (LONG et al. 2000 ), but the causal variants have yet to be identified.

Bristles as model quantitative traits:
Lessons learned about the genetic architecture of Drosophila bristle number extend not only to other Drosophila quantitative traits, but also to quantitative traits in other species, including common and complex human diseases. QTL for bristle number detected in initial and low-resolution mapping studies tend to fractionate into multiple QTL on further investigation. The same phenomenon has been observed when QTL for Drosophila life span were fine mapped by deficiency complementation (PASYUKOVA et al. 2000 ). Sex-specific effects of autosomal QTL were first observed for Drosophila bristle number (see above). Subsequent studies of other traits, including olfactory behavior (ANHOLT et al. 1996 ; MACKAY et al. 1996 ), life span (NUZHDIN et al. 1997 ; LEIPS and MACKAY 2000 ; PASYUKOVA et al. 2000 ; VIEIRA et al. 2000 ), and fitness (WAYNE et al. 2001 ), indicate this is a common feature of Drosophila quantitative traits, even at the level of the transcriptome (JIN et al. 2001 ). Similarly, environment-specific QTL effects have also been observed for Drosophila longevity (LEIPS and MACKAY 2000 ; VIEIRA et al. 2000 ) and fitness (FRY et al. 1996 ); and there are epistatic interactions between life span QTL (LEIPS and MACKAY 2000 ).

It is not likely that moderately large numbers of QTL, some of which have sex-, environment-, and genotype-specific effects, are a peculiar feature of Drosophila complex traits. However, with the exception of sex-specific effects, detecting environment- and genotype-dependent QTL effects is difficult in organisms that are not amenable to sophisticated genetic manipulation (e.g., humans). The practical consequence of lack of control over genotype and environment in studies of human diseases is that the marginal effects of QTL, averaged over all genotypes and environments, may be quite small, necessitating rather large sample sizes to detect individual QTL alleles. Lessons learned from Drosophila bristles can and should be used to guide experimental design in other systems.

	ACKNOWLEDGMENTS

We thank F. Lawrence, S. Heinsohn, and B. Hackett for help with the flies. Thanks to J. Leips, C. Basten, and Z-B. Zeng for discussion of various aspects of this experiment. This work was supported by National Institutes of Health grant GM 45146 to T.F.C.M. This is a publication of the W. M. Keck Center for Behavioral Biology.

Manuscript received March 8, 2002; Accepted for publication September 3, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ANHOLT, R. R. H., R. F. LYMAN, and T. F. C. MACKAY, 1996  Effects of single P element insertions on olfactory behavior in Drosophila melanogaster.. Genetics 143:293-301.[Abstract]

BARTON, N. H., 1990  Pleiotropic models of quantitative variation. Genetics 124:773-782.[Abstract]

BARTON, N. H. and M. TURELLI, 1989  Evolutionary quantitative genetics: How little do we know? Annu. Rev. Genet. 23:337-370.[Medline]

BASTEN, C. J., B. S. WEIR and Z-B. ZENG, 1994 Z-map—a QTL cartographer, pp. 65–66 in Proceedings of the 5th World Congress on Genetics Applied to Livestock Production: Computing Stategies and Software, Vol. 22, edited by C. SMITH, J. S. GAVORA, B. BENKEL, J. CHESNAIS, W. FAIRFULL et al. The Organizing Committee, 5th World Congress on Genetics Applied to Livestock Production, Guelph, Ontario, Canada.

BASTEN, C. J., B. S. WEIR and Z-B. ZENG, 2000 QTL Cartographer, Version 1.14. Department of Statistics, North Carolina State University, Raleigh, NC.

BEAVIS, W. D., 1994 The power and deceit of quantitative variation: lessons from comparative QTL studies, pp. 252–268 in 49th Annual Corn and Sorghum Research Conference. American Seed Trade Association, Washington, DC.

BOST, B., D. DE VIENNE, F. HOSPITAL, L. MOREAU, and C. DILLMANN, 2001  Genetic and nongenetic bases for the L-shaped distribution of quantitative trait loci effects. Genetics 157:1773-1787.[Abstract/Free Full Text]

BREESE, E. L. and K. MATHER, 1957  The organisation of polygenic activity within a chromosome in Drosophila. I. Hair characters. Heredity 11:373-395.

CAMPOS-ORTEGA, J. A., 1993 Early neurogenesis in Drosophila melanogaster, pp. 1091–1129 in The Development of Drosophila melanogaster, Vol. 2, edited by M. BATE and A. MARTINEZ ARIAS. Cold Spring Harbor Laboratory Press, Plainview, NY.

CHURCHILL, G. A. and R. W. DOERGE, 1994  Empirical threshold values for quantitative trait mapping. Genetics 138:963-971.[Abstract]

CLAYTON, G. A. and A. ROBERTSON, 1957  An experimental check on quantitative genetical theory. II. The long-term effects of selection. Genetics 55:152-170.

CLAYTON, G. A., J. A. MORRIS, and A. ROBERTSON, 1957a  An experimental check on quantitative genetical theory. I. Short-term responses to selection. Genetics 55:131-151.

CLAYTON, G. A., G. R. KNIGHT, J. A. MORRIS, and A. ROBERTSON, 1957b  An experimental check on quantitative genetical theory. III. Correlated responses. Genetics 55:171-180.

DARVASI, A., 1998  Experimental strategies for the genetic dissection of complex traits in animal models. Nat. Genet. 18:19-24.[Medline]

DOERGE, R. W. and G. A. CHURCHILL, 1996  Permutation tests for multiple loci affecting a quantitative character. Genetics 142:285-294.[Abstract]

FALCONER, D. S., and T. F. C. MACKAY, 1996 Introduction to Quantitative Genetics, Ed. 4. Longman, Essex, England.

FRANKHAM, R., 1968  Sex and selection for a quantitative character in Drosophila. I. Single-sex selection. Aust. J. Biol. Sci. 21:1215-1223.[Medline]

FRANKHAM, R., 1980 The founder effect and response to artificial selection in Drosphila, pp. 87–90 in Selection Experiments in Laboratory and Domestic Animals, edited by A. ROBERTSON. Commonwealth Agricultural Bureaux, Slough, UK.

FRY, J. D., S. L. HEINSOHN, and T. F. C. MACKAY, 1996  The contribution of new mutations to genotype-environment interaction for fitness in Drosophila melanogaster. Evolution 50:2316-2327.

GARCÍA-DORADO, A. and J. A. GONZÁLEZ, 1996  Stabilizing selection detected for bristle number in Drosophila melanogaster.. Evolution 50:1573-1578.

GILLESPIE, J. H. and M. TURELLI, 1989  Genotype-environment interactions and the maintenance of polygenic variation. Genetics 121:129-138.[Abstract/Free Full Text]

GIMELFARB, A., 1989  Genotypic variation for a quantitative character maintained under stabilizing selection without mutations: epistasis. Genetics 123:217-227.[Abstract/Free Full Text]

GURGANUS, M. C., J. D. FRY, S. V. NUZHDIN, E. G. PASYUKOVA, and R. L. LYMAN et al., 1998  Genotype-environment interaction at quantitative trait loci affecting sensory bristle number in Drosophila melanogaster.. Genetics 149:1883-1898.[Abstract/Free Full Text]

GURGANUS, M. C., S. V. NUZHDIN, J. W. LEIPS, and T. F. C. MACKAY, 1999  High-resolution mapping of quantitative trait loci for sternopleural bristle number in Drosophila melanogaster.. Genetics 152:1585-1604.[Abstract/Free Full Text]

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. BATE and A. MARTINEZ ARIAS. Cold Spring Harbor Laboratory Press, Plainview, NY.

JIANG, C. and Z-B. ZENG, 1995  Multiple trait analysis of genetic mapping for quantitative trait loci. Genetics 140:1111-1127.[Abstract]

JIN, W., R. M. RILEY, R. D. WOLFINGER, K. P. WHITE, and G. PASSADOR-GURGEL et al., 2001  The contributions of sex, genotype and age to transcriptional variance in Drosophila melanogaster. Nat. Genet. 29:389-395.[Medline]

JONES, L. P., R. FRANKHAM, and J. S. F. BARKER, 1968  The effect of population size and selection intensity in selection for a quantitative character in Drosophila. II. Long-term response to selection. Genet. Res. 12:249-266.[Medline]

KANIA, A., A. SALZBERG, M. BHAT, D. DEVELYN, and Y. C. HE et al., 1995  P-element mutations affecting embryonic peripheral nervous system development in Drosophila melanogaster.. Genetics 139:1663-1678.[Abstract]

KAO, C-H., Z-B. ZENG, and R. D. TEASDALE, 1999  Multiple interval mapping for quantitative trait loci. Genetics 152:1203-1216.[Abstract/Free Full Text]

KEARSEY, M. J. and B. W. BARNES, 1970  Variation for metrical characters in Drosophila populations. II. Natural selection. Heredity 25:11-21.[Medline]

KEIGHTLEY, P. D., T. F. C. MACKAY, and A. CABALLERO, 1993  Accounting for bias in estimates of the rate of polygenic mutations. Proc. R. Soc. Lond. Ser. B 253:291-296.[Medline]

LATTER, B. D. H. and A. ROBERTSON, 1962  The effects of inbreeding and artificial selection on reproductive fitness. Genet. Res. 3:110-138.

LEIPS, J. W. and T. F. C. MACKAY, 2000  Quantitative trait loci for life span in Drosophila melanogaster: interactions with genetic background and larval density. Genetics 155:1773-1788.[Abstract/Free Full Text]

LINDSLEY, D. L., and G. G. ZIMM, 1992 The Genome of Drosophila melanogaster. Academic Press, San Diego.

LINNEY, R., B. W. BARNES, and M. J. KEARSEY, 1971  Variation for metrical character in Drosophila populations. III. The nature of selection. Heredity 27:163-174.[Medline]

LONG, A. D., S. L. MULLANEY, L. A. REID, J. D. FRY, and C. H. LANGLEY et al., 1995  High resolution mapping of genetic factors affecting abdominal bristle number in Drosophila melanogaster.. Genetics 139:1273-1291.[Abstract]

LONG, A. D., S. L. MULLANEY, T. F. C. MACKAY, and C. H. LANGLEY, 1996  Genetic interactions between naturally occurring alleles at quantitative trait loci and mutant alleles at candidate loci affecting bristle number in Drosophila melanogaster.. Genetics 144:1497-1510.[Abstract]

LONG, A. D., R. F. LYMAN, C. H. LANGLEY, and T. F. C. MACKAY, 1998  Two sites in the Delta gene region contribute to naturally occurring variation in bristle number in Drosophila melanogaster.. Genetics 149:999-1017.[Abstract/Free Full Text]

LONG, A. D., R. F. LYMAN, A. H. MORGAN, C. H. LANGLEY, and T. F. C. MACKAY, 2000  Both naturally occurring insertions of transposable elements and intermediate frequency polymorphisms at the achaete-scute complex are associated with variation in bristle number in Drosophila melanogaster.. Genetics 154:1255-1269.[Abstract/Free Full Text]

LYMAN, R. F. and T. F. C. MACKAY, 1998  Candidate quantitative trait loci and naturally occurring phenotypic variation for bristle number in Drosophila melanogaster: the Delta-Hairless gene region. Genetics 149:983-998.[Abstract/Free Full Text]

LYMAN, R. F., F. LAWRENCE, S. V. NUZHDIN, and T. F. C. MACKAY, 1996  Effects of single P-element insertions on bristle number and viability in Drosophila melanogaster. Genetics 143:277-292.[Abstract]

LYMAN, R. F., C. LAI, and T. F. C. MACKAY, 1999  Linkage disequilibrium mapping of molecular polymorphisms at the scabrous locus associated with naturally occurring variation in bristle number in Drosophila melanogaster. Genet. Res. 74:303-311.[Medline]

LYNCH, M., and B. WALSH, 1998 Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland, MA.

MACKAY, T. F. C., 1985  Transposable element-induced response to artificial selection in Drosophila melanogaster. Genetics 111:351-374.[Abstract/Free Full Text]

MACKAY, T. F. C., 1995  The genetic basis of quantitative variation: number of sensory bristles of Drosophila melanogaster as a model system. Trends Genet. 11:464-470.[Medline]

MACKAY, T. F. C., 1996  The nature of quantitative genetic variation revisited: lessons from Drosophila bristles. BioEssays 18:113-121.[Medline]

MACKAY, T. F. C., 2001  Quantitative trait loci in Drosophila. Nat. Rev. Genet. 2:11-20.[Medline]

MACKAY, T. F. C. and J. D. FRY, 1996  Polygenic mutations in Drosophila melanogaster: genetic interactions between selection lines and candidate quantitative trait loci. Genetics 144:671-688.[Abstract]

MACKAY, T. F. C., R. F. LYMAN, and M. S. JACKSON, 1992  Effects of P element insertions on quantitative traits in Drosophila melanogaster. Genetics 130:315-332.[Abstract]

MACKAY, T. F. C., R. F. LYMAN, and W. G. HILL, 1995  Polygenic mutation in Drosophila melanogaster: non-linear divergence among unselected strains. Genetics 139:849-859.[Abstract]

MACKAY, T. F. C., J. B. HACKETT, R. F. LYMAN, M. L WAYNE, and R. R. H. ANHOLT, 1996  Quantitative genetic variation of odor-guided behavior in a natural population in Drosophila melanogaster. Genetics 144:727-735.[Abstract]

NUZHDIN, S. V., J. D. FRY, and T. F. C. MACKAY, 1995  Polygenic mutation in Drosophila melanogaster: the causal relationship of bristle number to fitness. Genetics 139:861-872.[Abstract]

NUZHDIN, S. V., E. G. PASYUKOVA, C. L. DILDA, Z-B. ZENG, and T. F. C. MACKAY, 1997  Sex-specific quantitative trait loci affecting longevity. Proc. Natl. Acad. Sci. USA 94:9734-9739.[Abstract/Free Full Text]

NUZHDIN, S. V., C. L. DILDA, and T. F. C. MACKAY, 1999  The genetic architecture of selection response: inferences from fine-scale mapping of bristle number quantitative trait loci in Drosophila melanogaster. Genetics 153:1317-1331.[Abstract/Free Full Text]

PASYUKOVA, E. G., C. VIEIRA, and T. F. C. MACKAY, 2000  Deficiency mapping of quantitative trait loci affecting longevity in Drosophila melanogaster. Genetics 156:1129-1146.[Abstract/Free Full Text]

ROBERTSON, A., 1967 The nature of quantitative genetic variation, pp. 265–280 in Heritage From Mendel, edited by A. BRINK. The University of Wisconsin Press, Madison, WI.

ROBERTSON, A., 1968 The spectrum of genetic variation, pp. 5–16 in Population Biology and Evolution, edited by R. C. LEWONTIN. Syracuse University Press, Syracuse, NY.

SALZBERG, A., S. N. PROKOPENKO, Y. C. HE, P. TSAI, and M. PÁL et al., 1997  P-element insertion alleles of essential genes on the third chromosome of Drosophila melanogaster: mutations affecting embryonic PNS development. Genetics 147:1723-1741.[Abstract]

SAS INSTITUTE, 1988 SAS / STAT User's Guide, Release 6.12 Edition. SAS Institute, Cary, NC.

SHRIMPTON, A. E. and A. ROBERTSON, 1988  The isolation of polygenic factors controlling bristle score in Drosophila melanogaster. III. Distribution of third chromosome bristle effect within chromosome sections. Genetics 118:445-459.[Abstract/Free Full Text]

SORSA, V., 1988 Chromosome Maps of Drosophila, Vol. II. CRC Press, Boca Raton, FL.

THODAY, J. M., 1979 Polygene mapping: uses and limitations, pp. 219–233 in Quantitative Genetic Variation, edited by J. N. THOMPSON and J. M. THODAY. Academic Press, New York.

VIEIRA, C., E. G. PASYUKOVA, Z-B. ZENG, J. B. HACKETT, and R. F. LYMAN et al., 2000  Genotype-environment interaction for quantitative trait loci affecting life span in Drosophila melanogaster. Genetics 154:213-227.[Abstract/Free Full Text]

WAGNER, G., G. BOOTH, and H. BAGHERI-CHAICHIAN, 1997  A population genetic theory of canalization. Evolution 51:329-347.

WAYNE, M. L., J. B. HACKETT, C. L. DILDA, S. V. NUZHDIN, and E. G. PASYUKOVA et al., 2001  Quantitative trait locus mapping of fitness-related traits in Drosophila melanogaster. Genet. Res. 77:107-116.[Medline]

ZENG, Z-B., 1993  Theoretical basis of precision mapping of quantitative trait loci. Proc. Natl. Acad. Sci. USA 90:10972-10976.[Abstract/Free Full Text]

ZENG, Z-B., 1994  Precision mapping of quantitative trait loci. Genetics 136:1457-1468.[Abstract]

ZENG, Z-B., C-H. KAO, and C. J. BASTEN, 1999  Estimating the genetic architecture of quantitative traits. Genet. Res. 74:279-289.[Medline]

This article has been cited by other articles:

				J. Flint and T. F.C. Mackay Genetic architecture of quantitative traits in mice, flies, and humans Genome Res., May 1, 2009; 19(5): 723 - 733. [Abstract] [Full Text] [PDF]

				J. A. Kerns, E. J. Cargill, L. A. Clark, S. I. Candille, T. G. Berryere, M. Olivier, G. Lust, R. J. Todhunter, S. M. Schmutz, K. E. Murphy, et al. Linkage and Segregation Analysis of Black and Brindle Coat Color in Domestic Dogs Genetics, July 1, 2007; 176(3): 1679 - 1689. [Abstract] [Full Text] [PDF]

				S. J. Macdonald and A. D. Long Joint Estimates of Quantitative Trait Locus Effect and Frequency Using Synthetic Recombinant Populations of Drosophila melanogaster Genetics, June 1, 2007; 176(2): 1261 - 1281. [Abstract] [Full Text] [PDF]

				D. Sambandan, A. Yamamoto, J.-J. Fanara, T. F. C. Mackay, and R. R. H. Anholt Dynamic Genetic Interactions Determine Odor-Guided Behavior in Drosophila melanogaster Genetics, November 1, 2006; 174(3): 1349 - 1363. [Abstract] [Full Text] [PDF]

				S. J. Macdonald, T. Pastinen, and A. D. Long The Effect of Polymorphisms in the Enhancer of split Gene Complex on Bristle Number Variation in a Large Wild-Caught Cohort of Drosophila melanogaster Genetics, December 1, 2005; 171(4): 1741 - 1756. [Abstract] [Full Text] [PDF]

				M. A. Carbone, A. Llopart, M. deAngelis, J. A. Coyne, and T. F. C. Mackay Quantitative Trait Loci Affecting the Difference in Pigmentation Between Drosophila yakuba and D. santomea Genetics, September 1, 2005; 171(1): 211 - 225. [Abstract] [Full Text] [PDF]

				T. F. C. Mackay, R. F. Lyman, and F. Lawrence Polygenic Mutation in Drosophila melanogaster: Mapping Spontaneous Mutations Affecting Sensory Bristle Number Genetics, August 1, 2005; 170(4): 1723 - 1735. [Abstract] [Full Text] [PDF]

				T. F.C Mackay and R. F Lyman Drosophila bristles and the nature of quantitative genetic variation Phil Trans R Soc B, July 29, 2005; 360(1459): 1513 - 1527. [Abstract] [Full Text] [PDF]

				R. C. Fuller, C. F. Baer, and J. Travis How and When Selection Experiments Might Actually be Useful Integr. Comp. Biol., June 1, 2005; 45(3): 391 - 404. [Abstract] [Full Text] [PDF]

				A F Wright Neurogenetics II: complex disorders J. Neurol. Neurosurg. Psychiatry, May 1, 2005; 76(5): 623 - 631. [Abstract] [Full Text] [PDF]

				C. Zeyl The Number of Mutations Selected During Adaptation in a Laboratory Population of Saccharomyces cerevisiae Genetics, April 1, 2005; 169(4): 1825 - 1831. [Abstract] [Full Text] [PDF]

				S. J. Macdonald and A. D. Long Identifying Signatures of Selection at the Enhancer of split Neurogenic Gene Complex in Drosophila Mol. Biol. Evol., March 1, 2005; 22(3): 607 - 619. [Abstract] [Full Text] [PDF]

				J. Hermisson and G. P. Wagner The Population Genetic Theory of Hidden Variation and Genetic Robustness Genetics, December 1, 2004; 168(4): 2271 - 2284. [Abstract] [Full Text] [PDF]

				A. L. Caicedo, J. R. Stinchcombe, K. M. Olsen, J. Schmitt, and M. D. Purugganan Epistatic interaction between Arabidopsis FRI and FLC flowering time genes generates a latitudinal cline in a life history trait PNAS, November 2, 2004; 101(44): 15670 - 15675. [Abstract] [Full Text] [PDF]

				A. J. Moehring and T. F. C. Mackay The Quantitative Genetic Basis of Male Mating Behavior in Drosophila melanogaster Genetics, July 1, 2004; 167(3): 1249 - 1263. [Abstract] [Full Text] [PDF]

				T. F. C. Mackay Methods for Genetic Dissection of Complex Traits Sci. Aging Knowl. Environ., April 28, 2004; 2004(17): pe17 - pe17. [Abstract] [Full Text]

				M. Turelli and N. H. Barton Polygenic Variation Maintained by Balancing Selection: Pleiotropy, Sex-Dependent Allelic Effects and G x E Interactions Genetics, February 1, 2004; 166(2): 1053 - 1079. [Abstract] [Full Text] [PDF]