DEFICIENCY MAPPING OF THE GENOMIC REGIONS ASSOCIATED WITH EFFECTS ON DEVELOPMENTAL STABILITY IN DROSOPHILA MELANOGASTER

Authors


Abstract

Developmental stability is the tendency of morphological traits to resist the effects of developmental noise, and is commonly evaluated by examining fluctuating asymmetry (FA)—random deviations from perfect bilateral symmetry. Molecular mechanisms that control FA have been a long-standing topic of debate in the field of evolutionary biology and quantitative genetics. In this study, we mapped genomic regions associated with effects on the mean and FA of morphological traits, and characterized the trait specificity of those regions. A collection of isogenic deficiency strains established by the DrosDel project was used for deficiency mapping of genome regions associated with effects on FA. We screened 435 genome deficiencies or approximately 64.9% of the entire genome of Drosophila melanogaster to map the region that demonstrated a significant effect on FA of morphological traits. We found that 406 deficiencies significantly affected the mean of morphological traits, and 92 deficiencies increased FA. These results suggest that several genomic regions have the potential to affect developmental stability. They also suggest the possibility of the existence of trait-specific and trait-nonspecific mechanisms for stabilizing developmental processes. The new findings in this study could provide insight into the understanding of the genetic architecture underlying developmental stability.

It is important for organisms to be capable of producing a highly fit and replicable phenotype under environmental and genetic perturbation. Developmental stability is the tendency of morphological traits to resist the effect of developmental noise (Waddington 1957; Debat and David 2001; Klingenberg 2006). In theory, corrective mechanisms that buffer developmental noise and stabilize phenotypes evolve through stabilizing or fluctuating selection (Kawecki 2000). Artificial selection experiments on wing traits of Drosophila melanogaster revealed that phenotypic variance strongly increased under disruptive selection but decreased under stabilizing and fluctuating selection (Pelabon et al. 2010), thereby supporting the above theory. The processes responsible for developmental stability are largely unknown, but have been discussed extensively (Debat and David 2001; Meiklejohn and Hartl 2002; Klingenberg 2003b; Leamy and Klingenberg 2005). It has been suggested that a “molecular chaperone,” such as HSP90, which has been shown to buffer among genotype morphological variations in D. melanogaster and Arabidopsis thaliana (Rutherford and Lindquist 1998; Queitsch et al. 2002), is one possible corrective mechanism. Another possible mechanism is the architecture of genetic regulatory networks responsible for gene expression (Houchmandzadeh et al. 2002; Bergman and Siegal 2003). In fact, hubs in the gene regulatory network of yeast have been found to buffer environmental variation (Levy and Siegal 2008). These examples clearly show that there are molecular bases that buffer environmental and genetic perturbations and control phenotypic variation in organisms; however, it remains unclear whether they regulate developmental stability.

Developmental stability is commonly evaluated by examining fluctuating asymmetry (FA)—random deviations from perfect bilateral symmetry (Moller and Thornhill 1997; Milton et al. 2003; Debat et al. 2006; Kellermann et al. 2007). Because development can be viewed as a deterministic pathway around which there exists a random walk generated by external and internal disturbances (Emlen et al. 2006), the degree of deviation from the symmetry of otherwise perfectly symmetrical structures provides a marker of the degree of developmental stability. To examine whether FA of morphological traits is heritable, FA heritability has been estimated in a number of studies (Van Valen 1962; Palmer and Strobeck 1986; Moller and Thornhill 1997; Whitlock and Fowler 1997; Gangestad and Thornhill 1999; Fuller and Houle 2006). Some studies have found statistically significant additive genetic variations of FA (Scheiner et al. 1991); however, the general conclusion of those studies was that the heritability of FA is very low. Other than heritability estimation, molecular mechanisms controlling FA of morphological traits have been investigated in only a few studies until date. The effect of an HSP90 inhibitor, geldanamycin, and several mutation alleles of Hsp90 on FA of morphological traits has also been examined (Milton et al. 2003; Debat et al. 2006). In most experimental settings, the reduction of HSP90 activity did not affect FA, and Debat et al. (2006) concluded that although Hsp90 contributed to the buffering of phenotypic variation, it was not controlling the variation. Another well-studied stress protein gene, Hsp70 was examined for its effect on FA using two deficiency strains, but no significant effect was detected (Takahashi et al. 2011a). Recently, however, Takahashi et al. (2010) found that the suppression of the expression of three small Hsp genes—Hsp22, Hsp67Ba, and Hsp67Bc—resulted in increased FA of morphological traits, suggesting their involvement in stabilizing developmental processes. Other than Hsp genes, Scalloped wings (Scl) is suggested to interact with Rop-1 and affect asymmetry of bristle traits in an Australian sheep blowfly, Lucilia cuprina (Batterham et al. 1996; Davies et al. 1996). Until date, no other specific gene has been suggested to contribute to developmental stability; however, Breuker et al. (2006) showed that deficiencies in the genomic regions of D. melanogaster could potentially affect FA of morphological traits, thus suggesting the existence of a specific genetic architecture to control FA. Unfortunately, they assessed the effect of genetic factors on FA based on among-deficiency strains variation using 115 deficiency lines, but they did not evaluate the statistical significance of the each deficiency effect by comparing FA scores of each deficiency and the control. Using well-designed experiments, locating genomic regions responsible for FA would shed light on how FA and developmental stability are regulated in organisms.

In this study, we (1) mapped genomic regions that had effects on the mean and FA of morphological traits, and (2) characterized the trait specificity of these regions. Because FA of morphological traits is most likely polygenic and sensitive to environmental perturbations (Leung et al. 2000), genetic background and environmental conditions must be strictly controlled. A collection of isogenic deficiency strains established by the DrosDel Project (Ryder et al. 2004; Ryder et al. 2007) is an ideal tool for genome-wide deficiency mapping of a polygenic trait such as FA. We screened 435 deficiencies that covered approximately 64.9% of the entire genome region of D. melanogaster to map the region having significant effects on FA of morphological traits. As a result, we found that 406 deficiencies significantly affected the mean of morphological traits, and 92 deficiencies increased FA. These results suggest that several genomic regions have the potential to affect developmental stability. They also suggest the possibility of the existence of trait-specific and trait-nonspecific mechanisms for stabilizing developmental processes. These new findings would provide new insights into the understanding of the genetic architecture underlying developmental stability.

Methods

FLIES

We obtained DrosDel isogenic deficiency strains from the Drosophila Genetic Resource Center in Kyoto, Japan and tested FA of morphological traits. The RS element-FLP system that was used to construct those deficiency strains enables the breakpoints of the deletions to be determined with single-base-pair resolution (Ryder et al. 2004). The control strain DSK001 is isogenic for the X, second, and third chromosomes, and was used to create RS-element inserted strains; therefore, except for deletions, the controls and all the deficiency strains have an isogenic background, thus providing an ideal tool by which genome regions involved in quantitative polygenic traits such as developmental stability can be screened. Because at least a few years had passed since the establishment of those strains, mutation accumulation might occur in both the control and the deficiency strains. The independently accumulated mutations in the control and deficiency strains may lead to a false positive effect of the deletions. However, given the spontaneous mutation rate (sum of the rates of the major source of mutations, single-nucleotide mutation rate and insertion rate) in D. melanogaster, 6.0 × 10−9 per base per generation (Haag-Liautard et al. 2007), and the effect of purifying selection and genetic drift to eliminate mutations from the population, we think that the number of mutations accumulated in those strains was small and negligible. In this study, we used 435 DrosDel deficiency strains that covered approximately 64.9% of the entire genome region, and 64.8% of the genes (Fig. 1, Appendix 1). We evaluated 11,418 flies for wing traits and 11,510 flies for bristle traits. More details of the deletion strains are available at the DrosDel website (http://www.drosdel.org.uk/).

Figure 1.

Distribution of 435 deletions on the second, third, and X chromosomes. Gray bars indicate chromosome arms and the gray circles at their tips indicate centromeres. Genome regions that have deletions are black, and black bars below each chromosome represent the location of each deletion.

EXPERIMENTAL CONDITIONS

We screened genome regions that influence the mean and FA of morphological traits. Because most of the deficiencies examined in this study were homozygous and hemizygous lethal, deficiency-control heterozygotes (Df/+) were measured for morphological traits. One hundred eggs were collected from crosses between the control and deletion strains, and were introduced into a glass vial containing fly medium. For deficiency strains with a deletion on the second or third chromosome, we crossed females of the control strain with males of each deficiency strain to control maternal effects. For deficiency strains with a deletion on the X chromosome, we crossed males of the control strain with females of each deficiency strain because of their hemizygous lethality in males. The 1-L fly medium consisted of water (1000 mL), dried yeast (35.0 g), soy flour (20.0 g), cornmeal (73.0 g), agar (30.0 g), malt extract (46.3 g), and dextrose (75.0 g), and was boiled well. We added 13.8 mL of acid mix (412 mL propionic acid + 42.0 mL orthophosphoric acid in water, made in up to 1 L of solution) and 16.5 mL of nipagin (100 g methyl-p-hydroxybenzoate in 1 L of 90% ethanol). The eggs collected were then reared in incubators at 23°C under constant light. Emerging adults were genotyped (target genotype, Df/+; nontarget genotype, balancer/+), and preserved in 70% ethanol for further morphological measurement. Five replicate vials were set up for each deletion strain. To obtain control individuals (+/+), we collected 100 eggs from DSK001 and reared them as above.

MORPHOLOGY MEASUREMENT AND SHAPE ANALYSIS

Because metric and meristic traits may respond differently to a lack of developmental buffering (Kellermann et al. 2007), we evaluated the effect of each deletion on FA of both. For metric traits, we measured shape [wing shape (WS)] and size [centroid size (CS)] components of wing morphology, which have been used to assess developmental buffering in several previous studies (Milton et al. 2003; Breuker et al. 2006; Debat et al. 2006; Kellermann et al. 2007). For meristic traits, we scored the following five bristle traits: the number of sternopleural (SP), scutellar (SC), thoracic (TH), ocellar (OC), and orbital (OR) bristles, which have also been used to assess developmental buffering (Bubliy et al. 2000; Dworkin 2005; Milton et al. 2005). We sampled up to three males and three females of each of the deficiency heterozygote per replication and scored the five bristle traits on the right and left sides of each fly. We then removed the right and left wings, captured their images under a microscope, SZX16 (Olympus, Tokyo, Japan), using a CCD camera, DP25 (Olympus, Tokyo, Japan), and obtained x and y coordinates of eight landmarks (Fig. 2) with the tpsDIG2 program (http://life.bio.sunysb.edu/morph/). The procedure of the landmark acquisition from the wing photos was repeated twice to evaluate measurement error in the following analysis. Procrustes analysis was used to generalize least squares (Rohlf and Slice 1990; Bookstein 1991; Rohlf and Marcus 1993) with the “shapes” package in a statistical software R 2.8.1. Additional data manipulation and analyses were also done with R 2.8.1.

Figure 2.

Positions of eight landmarks on a wing.

To evaluate the effect of the deletions on mean WS, we calculated Procrustes distance of each WS from the mean WS of the whole dataset as a square root of the sum of squared distances between corresponding landmarks using Procrustes coordinates, and used it as an index of WS. We performed this calculation separately for female and male. We used principal component analysis (PCA) to reduce the dimension of the bristle traits, and used the first PC (Bristle PC1) to assess the collective effect of the deletions on mean bristle traits. To evaluate FA of WS, we calculated a univariate measure of FA devised by Klingenberg and Monteiro (2005) based on the idea of one-sample standard distance (Flury and Riedwyl 1986; Flury 1997), which is equivalent to the one-sample version of the Mahalanobis distance (Mardia et al. 1979). This measure of FA automatically provides a correction for directional asymmetry (DA) (Klingenberg and Monteiro 2005). For CS and bristle traits, FA was evaluated as |L−R|/(L+R)/2, where L indicates a trait value on the left side and R indicates that on the right side of the body. In our dataset, CS and some bristle traits showed significant effects of “side” when we checked DA with simple ANOVA. To adjust DA for those traits, we standardized the signed difference between sides to render the population mean zero before calculating the absolute difference. To collectively evaluate FA of bristle traits, we used a composite index of FA (CFA) proposed by Leung et al. (2000). This index [CFA2 in Leung et al. (2000)] is calculated by dividing individual bristle FA values by the average FA of each trait so that all traits contribute equally to the CFA measurement, and then summing the FA values across traits for each individual, thereby creating a composite FA score for each individual. Leung et al. (2000) showed that CFA2 was one of the best CFAs in their simulation. We used all bristle traits to calculate CFA2.

The relative amounts of DA, FA, and measurement error in WS variation were assessed using Procrustes ANOVA (Klingenberg and McIntyre 1998) with degrees of freedom under the isotropic model (Klingenberg et al. 2002). In this analysis, we included individual and side, and their interaction terms as independent variables, and added the sums of squares across landmarks and coordinates, assuming equal and isotropic variation at each landmark.

STATISTICAL ANALYSIS

The sample size used for the analyses was 4.9 ± 0.42 vials for both females and males, and 13.8 ± 2.1 individuals per strain for females and 14.2 ± 2.0 individuals for males. To evaluate the effect of deletions on the mean and FA of morphological traits, we conducted a pairwise comparison between +/+ and each Df/+ using analysis of variance (ANOVA). We used average scores at vial level in those analyses on mean and FA, and checked the normality of the distribution of vial-level average scores with Kolmogorov–Smirnov test. To adjust for the multiple tests for different deficiencies, we applied the Benjamini–Hochberg procedure to control the false discovery rate (FDR) (Benjamini and Hochberg 1995). The deviation from the normal distribution was assessed as significant if adjusted FDR P-values (i.e., the q-values) of the coefficient for genotype were smaller than 0.05. As a result, no significant deviation from normal distribution was detected for any measure in any deficiency strain. In the ANOVA, we used the mean or FA of each morphological trait as a dependent variable, and genotype (+/+ or Df/+) as an independent variable. Adjustment for the multiple tests was done with Benjamini–Hochberg procedure as above. In this study, we performed separate analyses for sexes because of the sex-specific fitness effect observed in a number of deficiencies. There were 67 deletions with sex-specific lethality or strong sex-specific fitness effect (with more than 30% difference in the mean survival rates between sexes) of 435 deficiencies used in our study (Appendix 3). For those deficiency strains, it is inappropriate to test the sex-specific effect using a statistical model with two-way factorial design with the interaction term between “sex” and “genotype.” To avoid the complication of the analyses, we did not evaluate the sex-specific effect of the deficiencies on mean and FA of the morphological traits.

Interpretation of the result based only on the statistical significance sometimes gives biased view of the result due to the variation in sample size and the existence of outliers. In this study, we calculated effect size (Cohen's d) for each deficiency to draw more robust conclusion from the result, and to make results from different tests comparable.

We also tested the correlation between mean and FA of the morphological traits with a simple correlation analysis. In this analysis, we used average scores at strain level for both mean and FA. Because of the relatively small number of tests, we applied Bonferroni correction to adjust P-value for multiple tests in a conservative way. All the analyses were done with the statistical software R 2.8.1.

SURVIVAL

The effect of the deletions on survival was assessed by comparing the survival of +/+ and Df/+. We paired the data from +/+ and each Df/+ and developed a generalized linear model with a logit link function and binomial distribution. We used the number of survived and dead individuals as a dependent variable and the genotype (+/+ or Df/+) as an independent variable. In this analysis, we again applied the Benjamini–Hochberg procedure to control the FDR. The correlation was assessed as significant if adjusted FDR P-values (i.e., the q-values) of the coefficient for genotype were smaller than 0.05. Because the relationship between the degree of FA and fitness has been a focus of discussion (Leung and Forbes 1996, 1997), we also examined the correlation between FA and survival rate using mean FA and survival of each deficiency heterozygote level. Again in this analysis, we applied Bonferroni correction to adjust P-value for multiple tests in a conservative way because the number of tests was relatively small. Separate analyses were performed for females and males. All the analyses were done with the statistical software R 2.8.1.

Results

MEASUREMENT ERROR

Procrustes ANOVA showed that the contribution of measurement error to overall shape variation was small (Table 1), and the effect of FA and DA was highly significant in all cases.

Table 1.  Procrustes ANOVA for the wing landmarks. Sums of squares (SS) and mean squares (MS) are in dimensionless units of Procrustes distance. The sums of squares are added over landmarks and coordinates, assuming that all landmarks have the same amount of isotropic variation.
  dfSSMSFP
FemaleIndividual71,92824,409,397339.3594.678<0.0001
 Side12352,21929,351.579404.598<0.0001
 Individual×Side71,9285,218,02472.5458.659<0.0001
 Measurement error143,8801,205,4708.378  
MaleIndividual65,07615,794,274242.7054.271<0.0001
 Side12321,08126756.726470.837<0.0001
 Individual×Side65,0763,698,13756.8287.011<0.0001
 Measurement error130,1761,055,0798.105  

EFFECT OF DELETIONS ON THE MEAN AND FA OF MORPHOLOGICAL TRAITS

As a result of deficiency mapping of genomic regions associated with effects on mean morphological traits, we found 270, 185, and 313 regions with significant effects on mean WS, mean CS, and mean Bristle PC1, respectively (Fig. 3, Appendix 1). A majority of the effect of the deletions were found only in one sex (66.7% for mean WS, 65.4% for mean CS, and 69.6% for mean Bristle PC1), and the number of the regions with significant effect was biased toward females in mean WS and CS, and was biased toward males in Bristle PC1 (Fig. 3, Appendix 1). The distribution of such regions was scattered over all chromosomes (Fig. 3, Appendix 1). As for FA of morphological traits, 89 and 5 regions had significant effects on WS FA and Bristle CFA, respectively, and the number of the regions with significant effect was biased toward males in both cases (76.4% for WS FA and 100% for bristle CFA; Fig. 3, Appendix 1). Overlapping deletions tended to have consistent effect, but their effect was sometimes detected in different sexes among overlapping deletions (Fig. 3, Appendices 1 and 2). Pairwise correlation analyses for FAs of the morphological traits using the entire dataset showed a significant positive correlation between WS FA and CS FA in both female and male (P < 0.001 after Bonferroni correction) (Table 2). The frequency distributions of the effect size of the deletions on FA are shown in Figure 4. Large effects of the deletions (effect size larger than 4.0) were observed only in WS FA in female and Bristle CFA in male. FA of morphological traits of deficiencies with top 5% effect size and the corresponding mean trait scores are shown in Figure 5. All the deficiencies with a significant effect on FA increased the FA value compared to that of controls in both females and males (Fig. 5, Appendix 2). Among them, only Df(2L)ED228 had a significant effect on FAs of multiple morphological traits (WS FA and Bristle CFA) (Appendix 2). There were several deficiencies that had significant effects on both the mean and FA of a certain morphological trait: 37 deficiencies significantly affected the mean and FA of WS in females and 26 affected these in males, four deficiencies had a significant effect on the mean and FA of bristle trait in males (Fig. 5, Appendix 2). Summary of the deletions with top 5% effect size on WS FA in female or male is given in Table 3. Most of the deficiencies with top 5% effect size on WS FA encompassed one or more genes that were suggested to be involved in wing morphogenesis in previous microarray studies or databases such as The Interactive Fly (Table 3). Butler et al. (2003) found 94 genes showing a highly restricted expression pattern in a wing disc, and 17 of them were encompassed by 14 deficiencies with top 5% effect size on WS FA (Table 3). Ren et al. (2005) found 1335 genes changing their expression significantly during wing morphogenesis and differentiation, and 221 of them were encompassed by 35 deficiencies with top 5% effect size on WS FA (Table 3). The Interactive Fly listed 271 genes as those involved in wing morphogenesis, and 37 of them were encompassed by 16 deficiencies with effects on FA of wing morphologies (Table 3).

Figure 3.

Distribution of deficiencies with significant effects on the mean or fluctuating asymmetry (FA) of wing shape (WS), centroid size (CS), and bristles on the second, third, and X chromosomes. Genome regions covered by deficiencies are black and bars for deficiencies with significant heat sensitivity are colored based on sex specificity (a significant effect only in females is indicated with red, a significant effect only in males is indicated with blue, and a significant effect in both females and males is indicated with purple).

Table 2.  Correlation coefficients for pairwise correlation among WS FA, CS FA, and bristle FA.
  WS FACS FA Bristle FA
  1. ***P < 0.001.

FemaleWS FA-0.365*** 0.023
 CS FA -−0.078
 Bristle FA  -
MaleWS FA-0.273***−0.014
 CS FA -−0.041
 Bristle FA   
Figure 4.

The frequency distribution of the effect size of the deletions on WS FA, CS FA, and Bristle CFA in female and male.

Figure 5.

The mean and FA of WS, CS, and bristles of deficiency heterozygotes with significant effects on FA of those traits for females and males. Error bars represent standard errors. *q < 0.05, **q < 0.001, ***q < 0.0001.

Table 3.  Deficiencies that showed significant effect on WS FA or CS FA, and corresponding putative genes involved in wing morphogenesis found in Butler et al. (2003), Ren et al. (2005), and in a database. The Interactive Fly.
ChromosomeDeficientNo. of genes deletedButler et al. 2003Ren et al. 2005The Interactive Fly
2LDf(2L)ED381 Su(H), ZnT35C, CG42313Cul-3
 Df(2L)ED77841 Vha68-2bun, kekl, nub, spict
 Df(2L)ED780S49 Vhan68-2bun, kekl, nub, spict
 Df(2L)ED119665konncm, CG5783, CG10211, CG10283,CG10348, CG10383 
 Df(2L)ED131750CG9338Fs(2)Ket, CG9323, CG9336, CG3168ik2, vari
 Df(2L)ED14542   
 Df(2L)ED28092 galectin 
 Df(2L)ED465184 E23, NTPase, Mad, okr, pgant2, CG3523, CG9663, CG17224, CG17261, CG17264Mad
2RDf(2R)ED1735101Lcpl, CG30359Cyp4el, Kermit, Obp44a, Pbp49, CG2121, CG8693, CG8701, CG11198, CG17977,CG30372Opitx
 Df(2R)ED243674 Flo, Gpo-1, Zasp52, CG8192, CG8249, CG12964, CG30080, CG34365Poxn
 Df(2R)ED361089 elk, GstE5, GstE6, GstE9, IM2, Mapmoduln, Muc55B, CG14492, CG14500, CG18536fs, sbb
 Df(2R)ED3683126 abba, edl,GstE5, GstE6, GstE9, IM2, slim, SP2637, CG5224, CG5493, CG15080, CG15093, CG18536, CG18609sbb
 Df(2R)ED395255CG3624wrapper, CG4610, CG6758, CG11275Rtfl
 Df(2R)ED904557 Cpr49Ah 
3LDf(3L)ED207120CG9192Cctl, LysX, LysS, mwh, nerfin-1, Klp61F, l(3)02640, Psa, Ptp61F, trio, CG2469, CG9129, CG9134, CG9149, CG9184, CG12003, CG13907emc, Racl, rho
 Df(3L)ED22885CG17549CdGAPr, fon, lectin-37Da, msbll, Pax, CG10026, CG10237, CG10268, CG13081, CG13082, CG17349, CG17549, CG17567brat
 Df(3L)ED23051CG11370Ddxl, laza, CG11438, CG11370, CG11367, CG14448 
 Df(3L)ED417781 nerfin-1, Reg-2, trio, CG13894, CG13907ban, emc, hipk
 Df(3L)ED4341119ImpL2Akh, dyl, Eip63F-1, Faa, ida, nab, NT1, Syx17, Tie, VhaM9.7-1, CG1309, CG1311, CG1319, CG1136, CG7447, CG11353, CG14995, CG15005nab, scrt, wit
 Df(3L)ED442182 bol, Fhos, Hsp22, Hsp26, Hsp27, Hsp67Bc, PGRP-LA, PGRP-LC, UGP, CG5026, CG5280dally, Doc1
 Df(3L)ED4606144CG13044, CG13053Cpr72Ea, Dab, Lasp, nxf2, CG4925, CG4998, CG9705, CG12272, CG13025, CG13031, CG13042, CG13050, CG13053, CG13056, CG13064, CG13065, CG13067, CG13071Abl, argos
 Df(3L)ED468573CG13023Cpr73D, Fit2. TpnC73F, CG7724, CG11905, CG13027tap
 Df(3L)ED485886 in, Ir76b, NPFR76F, RhoBTB, polo, CG7365gig, Su(z)12
3RDf(3R)ED510096 Nep2, opa, rpk, Syt14, CG1124, cg2016, CG12001, CG14636, CG14654, CG34357 
 Df(3R)ED551145 Ugt86Da, CG17726, CG17721, CG31388Tctp
 Df(3R)ED566033 Neu3, Rrp6, VhaPPA1-1, CG7530, CG14860 
 Df(3R)ED566235 Neu3, Rrp6, VhaPPA1-1, CG7530, CG14860 
 Df(3R)ED56886 Tm2, CG14866 
 Df(3R)ED580785htl, sir,PP2A-B’, Vha 100-2, CG7126, CG7988repo,14-3-2epsilon
 Df(3R)ED591145 Cyp12a4, dnk, endoA, gukh, gwl, CG6026, CG14299 
 Df(3R)ED605824C15CG15497, CG31176E2f, InR
 Df(3R)ED610345 Rassf, CG1387, CG17119, CG17121hh
 Df(3R)ED105568 VhaPPA1-1 
 Df(3R)ED1055625 Rpb7, CG12241 
 Df(3R)ED1310244Obp99aBub3, Obp99a, Obp99c, CG1969, CG7582, CG7593, CG7598, CG15506Dr
XDf(1)ED682960GripSpx, Tsp5D, CG3726, CG3842, CG4666, CG5921, CG5928, CG15896sqh

SURVIVAL

Mean survival from egg to adult was significantly reduced in 252 deficiencies in females and 157 deficiencies in males (Appendix 3). Among 48 deficiencies that significantly increased FA of any morphological trait in females, 33 showed a reduction in survival with q-values <0.05, whereas 36 of the 67 deficiencies with significant effects on FA of any morphological trait in males showed a reduction in survival with q-values <0.05 (Appendix 3). The correlation between FA and survival was significant only in WS FA in both females and males after Bonferroni correction (P < 0.001 for both sexes) (Fig. 6).

Figure 6.

Correlation between FA of WS, CS, and bristles and survival for females and males.

Discussion

In this study, our genome-wide deficiency mapping, for the first time, located 92 genomic regions affecting FA of morphological traits of D. melanogaster. Although a large number of genomic regions showed significant effect on all the mean traits (270 for WS, 185 for CS, and 313 for Bristle PC1), the number of the genomic regions that affected FA differed strikingly among traits (89 for WS, 0 for CS and 5 for Bristle CFA). Distribution of the effect size on FA of each deletion also suggested that effect of the deletions on WS FA was larger than on CS FA and Bristle CFA. Part of the reason for this difference would be that WS can vary in more flexible ways compared to simple traits such as size of body parts or number of bristles, and be more informative (Debat et al. 2006).

These genome regions with significant effect did not encompass Hsp90 of which potential to affect developmental stability was suggested in the previous studies. Because deletion of Hsp90 resulted in heterozygous lethality in the DrosDel deficiency kit (K. H. Takahashi, unpubl. data), the effect of Hsp90 on FA of morphological traits could not be examined in this system. The deficiency, Df(3L)ED4421, that encompasses four small Hsp genes of which effect on WS FA and Bristle CFA was detected only in male using RNAi (Takahashi et al. 2010), influenced only WS FA in female in this study. A possible reason for not being able to repeat the result of the earlier study may be the lower survival rate of the Df(3L)ED4421/+ males because of the deletion of multiple important genes for survival with sex-specific effect. Another possibility is the insufficient reduction of the expression of the small Hsp genes because of the deletions. Although we did not examine the expression level of individual genes in this study, we expect that the expression level of the genes encompassed by the deficiency would be half as much in Df/+ individuals compared with that in +/+ individuals. The putative 50% reduction in the expression level may not be enough to hinder the normal function of all the genes. More severe knockdown of the expression would have been necessary to detect the effect of small Hsp genes on developmental stability.

In our study, the majority of the genomic regions associated with the effects on the mean and FA of morphological traits had significant effect on only one sex. Some of those sex-specific detections of the effect were due to the sex-specific lethality or strong sex-specific fitness effect of the genomic deletions, and some others might be the sex-specific effect of the deletion on the mean and/or FA of morphological traits. Such sex-specific effects have been known for a number of quantitative trait loci for mean WS or wing-vein position (Birdsall et al. 2000; Zimmerman et al. 2000; Weber et al. 2001), and for mean bristle number in D. melanogaster (Dilda and Mackay 2002; Norga et al. 2003; Mackay and Lyman 2005), indicating the sex-specific regulation of the developmental process of such traits. On the other hand, little is known about the sex-specific regulation of FA. So far, three small Hsp genes, Hsp22, Hsp67Ba, and Hsp67Bc (Takahashi et al. 2010), and 38 genomic regions found in this study have been shown to affect FA of wing or bristle traits only in male. FA of certain morphological traits has been considered as the target of sexual selection in many organisms, although publication bias in favor of the hypothetical connection between them has been pointed out (Moller and Pomiankowski 1993; Moller and Thornhill 1998; Polak 2008). In Drosophila, higher FA of wing and bristle traits in males have been suggested to be associated with reduced mating success (Markow et al. 1996; Polak and Taylor 2007; Polak 2008). The sex-specific detection of the effect of the genomic regions on FA of those traits found in the current study might reflect the result of sexual selection. The relationship between sex specificity of the effect of the genomic regions and sexual selection would be an important topic for additional research.

In our study, we found a significant correlation between FAs of WS and CS in both females and males. This result corresponded with those of previous studies in which the correlation of FAs of shape and size of wings was observed in D. melanogaster (Breuker et al. 2006; Takahashi et al. 2010), and suggested that genetic architecture underlining developmental stability of WS and CS overlaps, at least partially. In contrast to this overall correlation, only a few genomic regions had a significant effect on FA of multiple traits. This clear discrepancy between the general tendency in the entire dataset and individual genomic regions might be because a large number of genomic regions that have minor effects affect both WS FA and CS FA whereas genomic regions that have major effects, in general, are trait specific. This result provides support to the idea that there are multiple buffering mechanisms with trait-specific effects (Takahashi et al. 2010). This also supports the findings from Takahashi et al. (2011b) where multiple genomic deletions were found to have significant effect on the temporal canalization of developmental period, but no significant correlation between the degrees of temporal canalization and FA of WS was observed. This indicates that canalization of life-history trait such as developmental period and developmental stability of a morphological trait are regulated by independent mechanisms. Furthermore, the predominance of the genomic regions with trait-specific effects over the ones with general effects may suggest that developmental stability is governed in a hierarchical manner where a few general stabilizing mechanisms buffer global developmental noise and a number of trait-specific stabilizing mechanisms buffer local developmental noise. This hypothesis could be tested by detailed analysis of the interaction of the individual genomic regions.

Some of the genomic regions associated with the effects on FA of morphological traits found in this study also had a significant effect on the mean of traits whereas some did not, thereby indicating that two types of genomic regions with effect on FA; those involved in the process of both morphogenesis and stabilization of the mean, and those involved only in the stabilization of the developmental process. Because each deficiency encompasses an average of approximately 63 genes, it is still unclear whether a single gene affected both FA and mean trait. In the case of yeast and Caenorhabditis elegans, mutations in developmental networks have been shown to expose otherwise buffered stochastic variability in gene expression, leading to increased phenotypic variation (Levy and Siegal 2008; Raj et al. 2010). Most of the genomic regions associated with the effects on WS FA encompassed genes that were supposed to be involved in wing morphogenesis or genes that show spatially and temporally wing-disc-specific expression patterns during development. These genes could be candidates for the stabilization of developmental processes of wings. A further search for candidate genes within the specific genomic regions would reveal novel genes for developmental stability and provide a comprehensive picture of how developmental stability is regulated in insects.

FA has been used as a popular tool for estimating the quality and fitness of individuals and populations (Leung and Forbes 1996; Clarke 1998; Moller and Thornhill 1998). Although studies have reported inconsistent relationships between FA and fitness measured under environmental stress (Bjorksten et al. 2000) and inbreeding (Fowler and Whitlock 1994; Vollestad et al. 1999) along with the possibility of publication bias (Palmer 1999), some studies suggested a strong FA–fitness relationship in particular traits (Moller and Pomiankowski 1993; Hill 1995). In our study, only WS FA showed a significantly negative but relatively weak correlation with survival whereas FA of other traits showed no correlations. The trait-specific effect of the genetic perturbation by the deficiencies on FA supports the idea that FA is not necessarily a general bioindicator of stress and fitness (Lens et al. 2002). Further understanding of the genetic link between FA and fitness would provide a new insight into the long-standing argument of the use of FA as a bioindicator of stress and fitness.

FA of morphological traits can be affected not only by the error correction mechanism of developmental processes such as molecular chaperones, but also by nonlinear dynamics of developmental process itself (Klingenberg and Nijhout 1999; Klingenberg 2003a). In the current study, whether the genomic deletions might inhibit the function of the error correction mechanisms or they increased the sensitivity of developmental processes directly is still unclear. Our finding that a lot of morphogenic genes were encompassed by the deletions with strong effect on FA may support the latter possibility. Future examination of the effect of the individual genes in the candidate genomic regions would be necessary to understand how developmental stability is maintained in organisms.

In our study, and for the first time, we mapped the genome regions associated with effects on developmental stability. We identified multiple genome regions with a significant effect on FA of WS and bristle traits, and characterized their trait specificity. This novel finding may contribute to a better understanding of the mechanism by which the development noises are buffered and phenotypes are stabilized.


Associate Editor: C. Klingenberg

ACKNOWLEDGMENTS

We thank M. Yongmin Yi, Y. Takahashi, H. Saeki, M. Nagano, and I. Yamane for their assistance with the morphological measurements. We also thank Dr. I. Dworkin for helpful comments on the manuscript. This work was financially supported by Special Coordination Funds for Promoting Sciences and Technology of The Ministry of Education, Sport, Culture, Science and Technology of Japan, and by the Sumitomo Foundation to KHT.

Ancillary