Developmental constraints on an adaptive plasticity: reaction norms of pigmentation in adult segments of Drosophila melanogaster
Abstract
SUMMARY Variation of dark pigmentation according to developmental temperature was investigated in two geographic populations (France and India) with the isofemale line technique (20 lines for each population). The response curves called the reaction norms, were established in females for seven different segments: the mesothorax and abdomen segments 2–7 (Abd 2–7). In all cases the response curves were non-linear and had to be described either by a quadratic convex polynomial for thorax and Abd 2–5, or by a cubic polynomial for Abd 6 and 7. Among abdomen segments, increasing antero-posterior gradients were observed for several traits, including average pigmentation, overall phenotypic plasticity, the temperature of minimum pigmentation, and the curvature parameter of quadratic norms. Genetic correlations between abdomen segments were high when adjacent segments were considered, but became nil when more distant segments were correlated, suggesting that different pigmentation genes are expressed in the anterior and the posterior part of the abdomen. Characteristic values of reaction norms provided information either on trait value (i.e., the extension of pigmentation) or on plasticity. Correlations between plasticity and pigmentation were generally low and non-significant, suggesting their genetic independence. The overall darker pigmentation which is observed at low temperatures is assumed to be an adaptive plasticity. However, the differences which are evidenced among segments reveal strong interactions with developmental genes. These interactions are less likely to be a consequence of natural selection and are better interpreted as developmental constraints. The reaction norms analysis reveals the complexity of these interactions and should help, in the future, in the identification of the responsible thermosensitive genes.
INTRODUCTION
Phenotypic plasticity is the capacity of a single genotype to produce different phenotypes in different environments. This general property of living organisms is receiving increasing attention among evolutionary biologists ( Via 1992, 1993, 1994; Scheiner 1993a,b; Schlichting and Pigliucci 1993, 1998; Van Tienderen and Koelewijn 1994; De Jong 1995; Gotthard and Nylin 1995; Via et al. 1995). In many cases, environmental conditions during development will modify the adult phenotype. In other words, plasticity may be described as a genotype–environment interaction as well as an interaction between development and environment. Moreover, if plasticity is a trait by itself, submitted to natural selection, it may have a specific genetic basis. A way to analyze the genetics of plasticity is to consider the response curve of a phenotype according to an environmental gradient, which is called the norm of reaction ( Woltereck 1909). In this respect, Drosophila may be considered as a model organism for analyzing the shape of the norms of reaction ( David et al. 1997, 2000) and for working out the possible occurrence of plasticity genes ( Via 1993; Scheiner 1993a; Via et al. 1995).
A major issue in plasticity investigation concerns the possible adaptive value of the phenotypic change. Numerous investigators have drawn attention to the danger of adaptive, ad hoc interpretations, pointing out that plasticity will be often neutral or even maladaptive ( Schlichting and Pigliucci 1998; Winn 1999). Adaptive interpretations are often based on indirect evidence and convergent comparative data among species. Such is the case for insect body pigmentation. In many species, body pigmentation varies in a regular way, darker phenotypes being observed in colder environments ( Watt 1969; David et al. 1985, 1990; Capy et al. 1988; Kingsolver and Wiernasz 1991a,b; Goulson 1994; Monteiro et al. 1994; Gibert et al. 1996; De Jong et al. 1996; Ottenheim et al. 1996; Munjal et al. 1997). Such variations are generally interpreted as physiological adaptations, according to the thermal budget hypothesis: being darker in cold conditions will favor the absorption of light radiations and improve the internal temperature; conversely, a light body color will prevent overheating in a warm environment ( Watt 1969; Gibson and Falls 1979; David et al. 1985, 1990; Capy et al. 1988; Kingsolver and Wiernasz 1991b; Goulson 1994; Gibert et al. 1996).
In D. melanogaster female, each abdomen segment exhibits a stripe of black pigment along the posterior margin of each tergite. The genes which are responsible of the chaetae pattern on each tergite and of the deposition of the black pigment are progressively unravelled ( Kopp and Duncan 1997; Kopp et al. 1997; Struhl et al. 1997) and further progress is expected in a short future. Developmental geneticists, however, do not pay much attention to the variability of pigment extension, which may be very pronounced in D. melanogaster. Such variations arise either from genetic polymorphism or from plasticity, and both mechanisms are now documented.
Genetic changes correspond to a large amount of variability between individuals from the same population, and a high heritability of the trait has been observed ( David et al. 1990; Gibert et al. 1998a). They are also evidenced by the occurrence of darker genotypes in colder countries often arranged according to latitudinal clines on different continents ( David et al. 1985; Munjal et al. 1997; Gibert et al. 1996, 1998b). Plastic variations are observed within the same population according to seasons ( Das 1995 and unpublished observations). A parallelism exists between genetical clines and plasticity variations: both are strongly correlated to ambient temperature, providing an indirect argument for an adaptive plasticity.
In D. melanogaster, complete development, from egg to adult, is possible between 12 and 32°C. We investigated the phenotypic plasticity over this thermal range, measuring pigmentation of females on seven different body segments (i.e., thoracic segment 2 and abdominal segments [Abd] 2–7). In all cases non-linear norms were observed and their shapes were adjusted to polynomials. The genetic variability of the shape of the reaction norms was analyzed with the isofemale lines (full sib families) procedure ( Capy et al. 1993). Moreover, we compared two natural populations (France and India) adapted to different climates and known to exhibit adaptive differences in their average pigmentation ( Gibert et al. 1998b).
We found in all segments a general tendency to produce darker phenotypes at low temperatures. Significant differences in plasticity and in the shape of the reaction norms were evidenced between all segments, which are likely to result from internal constraints that is from non-adaptive processes. Our observations might help to identify the developmental genes which specify the fate and structure of adult segments, but also react to environmental temperature and are responsible of the variable extension of the black pigment.
MATERIAL AND METHODS
1. Populations and experimental procedure
We collected wild D. melanogaster females in Rohtak (India) and in Bordeaux (Southern France) with banana traps, and we isolated these wild females in culture vials to establish isofemale lines. For each population, 20 lines were randomly taken and investigated. From each line, we used ten pairs of the first laboratory generation as parents. Oviposition took place at about 20°C for half a day in vials containing a killed yeast medium known to reduce crowding effects ( Karan et al. 1999). We then transferred vials with eggs at one of seven constant experimental temperatures (12, 14, 17, 21, 25, 28, and 31°C). On emergence, adults were transferred to fresh food and examined a few days later. From each line at each temperature, we randomly studied 10 females.
We repeated the same procedure on the next generation in order to analyze the genetic repeatability of phenotypic plasticity (see Gibert et al. 1998a. Results revealed a good repeatability of the pigmentation scores of each line over the two generations. In the present work, since we were mainly interested in comparing the shapes of the reaction norms in different segments, we pooled, for each line, the data of the two generations so that each mean value, at each developmental temperature, is based on the observation of 20 females.
2. Pigmentation scores
We analyzed pigmentation on the mesonotum (thoracic trident) and on abdominal segments (Abd 2–7) in females. Pigmentation was estimated by using phenotypic classes ranging from 0 (completely yellow) to 10 (completely dark) for abdominal segments, and from 0 (no visible trident) to 3 (dark trident) for thoracic pigmentation (see David et al. 1985, 1990 for details). A major problem in using phenotypic classes is the accuracy of the measurements and their repeatability over time. When a single observer estimates the pigmented areas, an excellent repeatability of the scores is obtained. For example, the correlation between family means of successive generations was superior to 0.80, demonstrating both a strong genetic component and also the validity of the method ( Gibert et al. 1998a). In various occasions ( David et al. 1985; Gibert et al. 1999; and unpublished data) the same flies were scored by two persons and high correlations were again obtained between observers. In the present work, two persons (P. G. and J. R. D.) made the observations, each on 10 lines from France and 10 lines from India, and differences between observers were small and never significant. For comparing thorax and abdominal pigmentation, we standardized their possible range of variation: the trident pigmentation score was multiplied by 3.33, so that its variability ranged also between 0 and 10.
3. Data analysis
We analyzed reaction norm shape by using a polynomial adjustment ( David et al. 1997; Gibert et al. 1998c). According to the shape, different kinds of models (linear, quadratic, cubic) were used and polynomials calculated. The polynomial coefficients were then used to calculate characteristic values of the norms providing information either on traits (pigmentation) or plasticity (see David et al. 1997. We also considered various genetic correlations either between pigmentation of different segments or between reaction norms parameters. Calculations were done using SAS and Statistica softwares ( S.A.S. 1985; Statistica 1997). For all estimates of genetic correlations we used the correlations between family means without correcting for family size ( Via 1984; Roff 1997; Gibert et al. 1998d).
RESULTS
1. Average reaction norms in different segments
Average curves for thoracic pigmentation on the mesonotum and for Abd 2–7 are shown in Fig. 1. For each segment, data were submitted to ANOVA (Table 1) and in each case, we observed a major effect of temperature. Genetic differences among isofemale lines of the same population are also evidenced, as already described in previous publications ( David et al. 1990; Gibert et al. 1998a). In the present paper, isofemale lines are mainly considered as biological repetitions, allowing to calculate confidence intervals along average curves (see Fig. 1). A significant difference is also found between populations, and the population × temperature interaction is significant only for the trident and the last two abdomen segments. This interaction means that the shapes of the reaction norms are not identical in the two populations.
Average reaction norms of pigmentation for the mesothorax and for abdomen segments 2–7, according to growth temperature. For each population (France and India) values are the mean of 20 isofemale lines. Vertical bars: 95% confidence interval.
| df | Trident | Abd. 2 | Abd. 3 | Abd. 4 | Abd. 5 | Abd. 6 | Abd. 7 | |
|---|---|---|---|---|---|---|---|---|
| Population (1) | 1 | 109.59 *** | 19.52 *** | 11.69 *** | 20.29 *** | 28.91 *** | 15.96 *** | 8.46 *** |
| Temperature (2) | 6 | 369.17 *** | 124.49 *** | 344.33 *** | 507.14 *** | 760.10 *** | 1446.77 *** | 2237.46 *** |
| Line (3) | 38 | 7.30 *** | 5.53 *** | 6.90 *** | 5.53 *** | 5.60 *** | 5.79 *** | 3.04 *** |
| 1*2 | 6 | 10.91 *** | 0.18 ns | 1.10 ns | 2.27 * | 1.04 ns | 14.46 *** | 5.19 *** |
| Error | 228 |
- *** p < 0.001; **p < 0.01; *p < 0.5; ns p > 0.5.
- df, degree of freedom.
- Family means, each based on 10 females, were used for calculations. The line factor was nested in population. Values of F parameters (variance ratio) are given.
2. Change in mean pigmentation and in phenotypic plasticity along the antero-posterior axis
Fig. 1 clearly shows that some segments are less pigmented than others. We quantified this observation by calculating for each segment and each isofemale line, an average pigmentation over the seven experimental temperatures. As shown in Fig. 2A, we observe a general trend of increasing darkness from Abd 2–Abd 6, then a slight decrease in Abd 7. Indian flies appear also regularly lighter than French ones and the major difference between the two populations is found for the thoracic trident.
(A) Variation of mean pigmentation (averaged over seven temperatures) along the antero-posterior axis. (B) Variation of phenotypic plasticity (defined by the variance of pigmentation among temperatures) along the antero-posterior axis. Vertical bars: 95% confidence interval.
We also tried to quantify the magnitude of phenotypic changes for each segment (i.e., its phenotypic plasticity), and several quantitative indices could be considered. In the present case, since phenotypic variations are constrained between 0 and 10, we used, as in a previous paper ( Gibert et al. 1999), the variance between values obtained at different temperatures. Under our conditions, the variance is constrained between 0 and 27.8. As seen in Fig. 2B, plasticity was almost nil in Abd 2 and then increased rapidly backward, with a maximum in Abd 7. There was no consistent difference between the two populations. A fairly high plasticity was also observed in the thorax, with a significantly higher value in the French population.
3. The shape of the reaction norms: empirical derivatives
As in previous papers ( David et al. 1997; Gibert et al. 1998c), the reaction norms were adjusted to polynomials, and a major problem was to choose their degree. The best practical method ( David et al. 1997; Gibert et al. 1998c) seems to calculate, for each segment, an empirical derivative, and then analyze the variation of the derivative in relation with temperature. This procedure is illustrated in Fig. 3 for the thorax and Abd 7. For each temperature interval, a slope value is calculated and assigned to the middle of the interval. With seven experimental temperatures, six slope values are available for each isofemale line.
Analysis of the shape of the reaction norms by calculating an empirical derivative. A linear slope is calculated between pigmentation score of two successive temperatures. (A) Results obtained in 40 isofemale lines for pigmentation of thorax or segment 7. (B) Average curves (with confidence intervals) for French and Indian flies. For thoracic pigmentation the derivative exhibits a monotonous linear increase so that the pigmentation of the thoracic trident (Fig. 1) can be described by a quadratic function. For abdominal segment 7, the derivatives are convex, non-linear curves, so that the reaction norms (Fig. 1) need a cubic, sigmoid adjustment.
The dispersal of the slope values among the 40 isofemale lines is shown Fig. 3A, and the average derivative curves are given in Fig. 3B. For the thorax, ANOVA evidenced a regular increase of the slope value according to developmental temperature, with no difference between the two populations. The data were further analyzed by a linearity test ( Dagnélie 1975), and no significant departure from a linear regression was evidenced. The same conclusion was reached for segments Abd 2–5. A linear increase of the derivative implies that, for describing the reaction norm of pigmentation, a quadratic polynomial has to be used.
Linearity was rejected for segments 6 and 7, and in these cases, convex derivatives were observed with maximum (still negative) values at extreme low and high temperatures (see Fig. 3). A non-linear derivative implies that the reaction norms for these segments must be described by a third degree function.
Derivative analyses thus lead to the conclusion that all anterior segments, from thorax to Abd 5, exhibit quadratic convex reaction norm. A transition occurs after Abd 5 since the last two segments exhibit not only a higher plasticity, but also reaction norms with a decreasing sigmoid shape.
4. Characteristic values of the reaction norms
Polynomial adjustments are best used for calculating, for a given curve, a few characteristic values. For a quadratic curve, these are a curvature, g2 parameter and the coordinates of the minimum (or maximum) (see David et al. 1997; Gibert et al. 1998c. The minimum pigmentation (minP) characterizes the trait investigated, in that case the pigmentation of each segment in each isofemale line. Curvature (g2) and Temperature of minimum Pigmentation (TminP) characterize the reactivity of the trait to developmental temperature and thus describe its plasticity.
With a cubic adjustment (Abd 6 and Abd 7), things are more complicated since many more characteristic values can be calculated ( David et al. 1997) such as the coordinates of the inflexion point, the coordinates of a maximum and of a minimum, and the slope at the inflexion point (K). Three of these values characterize the trait and four characterize its plasticity. Mean characteristic values for all segments in the two populations are given in Table 2. For the five quadratic norms, the trait values (minP) are significantly different and less in India than in France, as could be expected from curves in Fig. 1. Plasticity characters (TminP and g2) on the other hand, are never different, except for thoracic trident. For the sigmoid norms, three significant differences between India and France are observed in Abd 6: two concern pigmentation and one plasticity (K, the slope at inflexion point). For the Abd 7, differences are less and generally non significant.
| Trait | Characteristic points | France | India | F(1,38) |
|---|---|---|---|---|
| Thoracic | minP | 1.893 ± 0.282 | −0.251 ± 0.089 | 52.65 *** |
| trident | TminP | 24.732 ± 0.218 | 25.818 ± 0.199 | 13.53 *** |
| g2 | 0.047 ± 0.002 | 0.031 ± 0.002 | 36.62 *** | |
| Abd. 2 | minP | 1.598 ± 0.062 | 1.242 ± 0.068 | 14.89 *** |
| TminP | 24.302 ± 0.298 | 24.498 ± 0.332 | 0.19 ns | |
| g2 | 0.010 ± 0.001 | 0.009 ± 0.001 | 0.12 ns | |
| Abd. 3 | minP | 2.350 ± 0.071 | 1.971 ± 0.067 | 15.21 *** |
| TminP | 26.253 ± 0.275 | 26.681 ± 0.579 | 0.45 ns | |
| g2 | 0.011 ± 0.001 | 0.011 ± 0.001 | 0.001 ns | |
| Abd. 4 | minP | 2.456 ± 0.066 | 2.160 ± 0.067 | 9.94 ** |
| TminP | 27.628 ± 0.519 | 27.682 ± 0.412 | 0.07 ns | |
| g2 | 0.013 ± 0.001 | 0.012 ± 0.001 | 2.66 ns | |
| Abd. 5 | minP | 2.319 ± 0.069 | 1.700 ± 0.122 | 19.51 *** |
| TminP | 30.967 ± 0.737 | 30.798 ± 0.735 | 0.03 ns | |
| g2 | 0.016 ± 0.001 | 0.016 ± 0.001 | 0.13 ns | |
| Abd. 6 | Pip | 5.832 ± 0.133 | 5.065 ± 0.103 | 21.28 *** |
| maxP | 9.696 ± 0.123 | 9.792 ± 0.095 | 0.39 ns | |
| minP | 1.968 ± 0.261 | 0.338 ± 0.191 | 26.12 *** | |
| Tip | 23.001 ± 0.285 | 23.147 ± 0.337 | 0.11 ns | |
| k | −0.583 ± 0.022 | −0.727 ± 0.022 | 20.84 *** | |
| TmaxP | 13.022 ± 0.255 | 13.271 ± 0.314 | 0.37 ns | |
| TminP | 32.980 ± 0.491 | 33.024 ± 0.547 | 0.003 ns | |
| Abd. 7 | Pip | 4.762 ± 0.066 | 4.672 ± 0.046 | 3.77 ns |
| maxP | 9.826 ± 0.119 | 9.516 ± 0.085 | 4.44 * | |
| minP | −0.302 ± 0.063 | −0.172 ± 0.025 | 3.65 ns | |
| Tip | 21.808 ± 0.185 | 21.225 ± 0.237 | 1.25 ns | |
| k | −0.902 ± 0.031 | −0.815 ± 0.015 | 6.38 * | |
| TmaxP | 13.239 ± 0.215 | 12.253 ± 0.347 | 5.84 * | |
| TminP | 30.377 ± 0.382 | 30.197 ± 0.219 | 0.17 ns |
- minP, minimum Pigmentation; TminP, Temperature of minimum Pigmentation; g2 curvature parameter; Pip, Pigmentation at the inflexion point; max P, maximum Pigmentation; Tip, Temperature at the inflexion point; K, slope at the inflexion point; TmaxP, Temperature of maximum Pigmentation.
Table 2 permits also a comparison of the characteristics values across segments. Two cases are illustrated in Fig. 4, and they both concern plasticity. The temperature of minimum pigmentation, which can be estimated with quadratic and cubic norms, shows a quite regular increase along the antero-posterior axis (Fig. 4A) from a value of 25.3°C in the thorax up to 33°C in Abd 6, and a slight decrease in Abd 7. The shape of this curve is fairly similar to that of average pigmentation (Fig. 2A), although its biological significance is completely different. The curvature (g2) coefficient was estimated only in quadratic norms (Fig. 4B): a greater value means that the trait is more plastic. A high plasticity is observed for the thoracic pigmentation, more pronounced in France than in India. Curvatures are much less in the abdomen segments. ANOVA applied to the abdomen data evidenced a significant increase toward posterior segments, but no difference between populations, so that a single g2 value is shown on the graph. Notice that the variations of g2(Fig. 4B), which reflect variations in plasticity, are quite analogous to those of average pigmentation shown in Fig. 2B, which also concern plasticity.
Variation of plasticity characteristics along the antero-posterior axis. (A) The temperature of minimum value increases toward the end of the abdomen. (B) The curvature of the quadratic norms is maximum for thoracic pigmentation; a slight but significant increase is observed for abdomen segments 2–5, but no difference among populations.
5. Genetic correlations among pigmentation of different segments
Correlating the isofemale line mean values of two different segments at a given temperature estimates the genetic correlation between the two traits ( Gibert et al. 1998d). All pairwise combinations were calculated and, with seven temperatures, seven segments and two populations, 294 such coefficients should be available, with each based on 20 observations. In fact, six of them could not be calculated, due to a lack of genetic variability in Abd 7 of the Bordeaux population grown at 31°C. The whole data set was submitted to ANOVA after a z transformation. This analysis (Table 3) failed to reveal any significant population effect but a slight temperature effect ( p = 0.023). This effect is shown in Fig. 5: average correlations are less in the middle of the thermal range and increase slightly under extreme conditions.
| df | MS | F | |
|---|---|---|---|
| Population (1) | 1 | 0.093 | 2.31 ns |
| Segment (2) | 20 | 1.069 | 26.48 *** |
| Temperature (3) | 6 | 0.103 | 2.56 * |
| 1 * 2 | 20 | 0.271 | 6.70 *** |
| 1 * 3 | 6 | 0.036 | 0.88 ns |
| 2 * 3 | 120 | 0.049 | 1.21 ns |
| Error | 114 | 0.040 |
- *** p < 0.001; **p < 0.01; *p < 0.5; ns p > 0.5.
- df, degree of freedom; MS, mean square; F, variance ratio.
Relationship between genetic correlations among segments and growth temperature in the two populations. The difference between populations is non-significant. The temperature effect is, however, significant, with lesser values in the middle of the thermal range.
A highly significant effect was observed according to the kind of correlated segments. These values, averaged over temperatures, are presented in Table 4. Data for French and Indian populations are separated because ANOVA showed a strong interaction component between populations and correlated segments. An inspection of Table 4 shows a general tendency for the coefficients to be higher when adjacent segments are correlated. Also, correlations involving the thoracic pigmentation are generally low and non-significant. We investigated further this problem by plotting the correlations against the physical distance between segments. More precisely, we considered the intervals between segments, a value of zero being given when adjacent segments were correlated. We also considered separately the correlations involving only abdominal segments and those involving both thoracic and abdomen pigmentation. The results are shown Fig. 6. Correlations involving the thorax are always low and not influenced by the distance. Moreover, as shown in Table 4, they are significantly higher than zero in the French population (average r = 0.205 ± 0.030) but not in the Indian one (r = 0.027 ± 0.028). Results are completely different for correlations among abdominal segments: a sharp, linear decrease is observed when more distant segments are correlated, from a positive r = 0.70 value for adjacent segments down to zero for an interval of three segments.
| Thorax | Abd. 2 | Abd. 3 | Abd. 4 | Abd. 5 | Abd. 6 | Abd. 7 | |
|---|---|---|---|---|---|---|---|
| Thorax | 0.06 ± 0.078 | –0.03 ± 0.083 | 0.01 ± 0.060 | 0.15 ± 0.061 | –0.01 ± 0.064 | –0.02 ± 0.032 | |
| Abd. 2 | 0.16 ± 0.066 | 0.77 ± 0.063 | 0.70 ± 0.081 | 0.52 ± 0.091 | 0.35 ± 0.117 | 0.03 ± 0.084 | |
| Abd. 3 | 0.18 ± 0.056 | 0.71 ± 0.045 | 0.93 ± 0.011 | 0.64 ± 0.077 | 0.57 ± 0.083 | 0.02 ± 0.100 | |
| Abd. 4 | 0.24 ± 0.077 | 0.46 ± 0.066 | 0.85 ± 0.040 | 0.73 ± 0.074 | 0.60 ± 0.088 | 0.03 ± 0.084 | |
| Abd. 5 | 0.34 ± 0.076 | 0.06 ± 0.088 | 0.43 ± 0.053 | 0.66 ± 0.038 | 0.66 ± 0.087 | 0.11 ± 0.083 | |
| Abd. 6 | 0.19 ± 0.109 | –0.23 ± 0.090 | 0.03 ± 0.093 | 0.25 ± 0.090 | 0.67 ± 0.084 | 0.38 ± 0.129 | |
| Abd. 7 | 0.12 ± 0.119 | –0.22 ± 0.088 | 0.05 ± 0.095 | 0.27 ± 0.078 | 0.47 ± 0.107 | 0.67 ± 0.134 |
Variation of the genetic correlations as a function of the interval between the segments which are correlated. Correlations among abdominal segments decrease sharply when more distant segments are compared. Correlations implying the thoracic trident are always very low and do not exhibit a distance effect.
6. Genetic correlations between pigmentation and plasticity in the same segment
A general problem in phenotypic plasticity studies is to know if genes controlling the phenotype and genes controlling plasticity are the same ( Scheiner 1993a; Via 1993). A possible approach is to consider genetic correlations between these two properties. Since mean values were generally different in the French and Indian flies, correlations were always calculated within each population to avoid possible artifacts.
In the present work, several ways for estimating pigmentation and plasticity were available. A first possibility, shown in Fig. 2, was to consider the mean pigmentation of each line and to estimate its plasticity by the variance of the trait over temperatures. We correlated these two values for each segment and each population. Among 14 values, only one was superior to the significance threshold (Bonferroni correction, Rice 1989, but four were negative, although non-significantly so. The overall mean (r = 0.187) was positive but very low. We may conclude that mean pigmentation and plasticity were on average not correlated.
The second possibility was to analyze the correlations among the characteristic values which are shown in Table 2. This was done separately for quadratic and cubic norms. For quadratic norms (thorax and Abd 2–5) three characteristic values were available and the mean correlations are given below.
minP·TminP: r = 0.179 ± 0.111 ns n = 10 coefficients
minP·g2: r = –0.375 ± 0.093 ** n = 10 coefficients
TminP·g2: r = –0.606±0.096 *** n = 10 coefficients
The minimum pigmentation is not correlated to the temperature of minimum pigmentation, but negatively so to g2. The two plasticity parameters, g2 and TminP, are strongly but negatively correlated. These observations mean that a stronger curvature accompanies a lower minimum pigmentation. It is interesting to notice that these genetic trends, which are observed when comparisons are made between lines for the same segment, may be in opposition with the results obtained between different segments. For example, for Abd 2–5, we found a progressive increase of both curvature and TminP along the antero-posterior axis (Fig. 4), that is an overall positive correlation.
Correlations were more complex for the sigmoid norms of segments 6 and 7. Since seven different characteristic values were calculated (Table 1), 21 different coefficients were available for each population and segment. These coefficients fall into three groups: between pigmentation characteristics (three coefficients), between plasticity characteristics (six) and between pigmentation and plasticity (12). Among pigmentation characteristics, Pip was positively correlated with Pmax and Pmin (average r = 0.635 ± 0.104, n = 8 coefficients) while the correlation between Pmax and Pmin was nil (r = –0.055 ± 0.067, n = 4). Among plasticity characteristics, Tip was positively correlated with Tmax and Tmin (r = 0.698 ± 0.084, n = 8). This correlation is explained by horizontal shifts of the norms along the temperature axis. K, the slope at the inflexion point, was negatively correlated to Tmax only (r = 0.670 ± 0.033, n = 4). Other correlations were not consistent and generally non-significant. For the correlations between pigmentation and plasticity, the coefficients were on average less (overall mean was –0.043). Among 48 values (12 values for each segment and population), 25 were significantly different from 0 (after a Bonferoni correction). Consistent results among populations and segments were observed in two cases only, between TminP and minP (r = –0.69 ± 0.084, n = 4) or Pip (r = –0.73 ± 0.081, n = 4).
DISCUSSION AND CONCLUSIONS
Reaction norms of adult pigmentation were previously analyzed in D. melanogaster females, but in smaller samples (10 isofemale lines) and only for the last three abdomen segments ( David et al. 1990; Gibert et al. 1996). Extant data provide information for six abdomen segments. Moreover, a much bigger sample size (two populations; 40 lines; 5600 females; 39,200 measures) has permitted a more precise analysis of the shape of reaction norms and of the phenotypic and genetic variability of pigmentation according to body segment.
All reaction norms proved to be non-linear, and similar shapes have been found in both populations. Their analysis has evidenced numerous differences between segments, often organized along an antero-posterior gradient. The antero-posterior axis of the body plan is established during early embryogenesis, in relation with the bithorax gene complex ( Lawrence 1992; Lewis 1992; Gehring 1998) and exists in all insect groups. This structure is highly conserved and must be a very strong developmental constraint. Plasticity differences among segments are difficult to explain in terms of specific adaptations and are more easily considered as a consequence of non-adaptive internal constraints. In this respect, we clearly distinguished all anterior segments (thorax and Abd 2–5) with quadratic convex norms, from the last two segments (Abd 6 and Abd 7) with sigmoid decreasing norms. The quadratic norms imply that dark pigmentation increases at very high temperatures, above TminP. This phenomenon is not much pronounced except for thoracic trident (see Fig. 1) and cannot be fitted to the thermal budget adaptive hypothesis (see INTRODUCTION). Constraints instead of adaptive changes may also be envisaged when regular variations are observed along the antero-posterior axis. Such variations have been found for average pigmentation (Fig. 2A) and various aspects of phenotypic plasticity 2, 4. In all cases, the reactivity of the thorax is quite different from that of the abdomen, and also the difference between the two populations is more pronounced for the thorax.
The occurrence of a stripe of black pigment at the posterior edge of each abdomen tergite in female, and of a trident pattern on the thorax, is specific to D. melanogaster, D. simulans, and a few other related species ( Capy et al. 1993; Gibert et al. 1996; unpublished observations). This pigmentation pattern may be considered as a recently evolved trait which implies specific but still unknown pigmentation genes. We recently investigated ( Gibert et al. 1999) pigmentation and developmental temperature in D. kikkawai, a quite distant relative of D. melanogaster, although included in the same taxonomic group. D. kikkawai is polymorphic for a major diallelic pigmentation locus expressed only at the tip of the female's abdomen. This species exhibited a very low level of plasticity with respect to temperature, showing that both pigmentation and plasticity can evolve independently and rapidly across species. Female pigmentation in D. melanogaster is less strongly canalized by developmental constraints, and thus may remain polymorphic and able to respond rapidly to natural selection. We have been able to analyze this variability in two ways, comparing two geographic populations and considering the intrapopulational variability.
We found genetic differences between French and Indian flies with consistently a lighter average pigmentation in the population of India which lives under warmer conditions. Such differences are likely to reflect a climatic adaptation according to the thermal budget hypothesis ( Gibert et al. 1998b). The biggest difference has been observed for the thorax, and it might be argued that thoracic pigmentation is a privileged target for natural selection, related to the activity of flight muscles. In this respect, we may recall that latitudinal clines of pigmentation are well documented in D. melanogaster, but for thoracic trident only ( David et al. 1985; Munjal et al. 1997), and they also exist in the sibling D. simulans ( Capy et al. 1988).
The within-population variability has been documented in previous investigations. An excellent genetic repeatability of the pigmentation scores was observed when the same isofemale lines were measured in successive generations ( Gibert et al. 1998a). Also, for all segments a high heritability was observed ( David et al. 1990; Gibert et al. 1998a ). In the present paper, since we were more interested by the interaction between pigmentation and the body plan, we focused our analysis on various kinds of genetic correlations.
The thoracic pigmentation exhibited, on average, only a very small positive correlation with that of the abdomen. This agrees with the fact that specific genes determine the pigmented trident pattern on the mesonotum (see David et al. 1985; Capy et al. 1988. Among abdomen segments, however, we reached a clear-cut conclusion that the correlation decreases in a linear way, from 0.70 to 0 when more distant segments are compared. Interestingly, a high correlation is always found between any set of adjacent segments, whether anterior, median, or posterior. But the anterior (2 or 3) segments are absolutely independent of posterior ones (6 or 7). A lack of correlation between quantitative traits is generally considered as an indication that the two traits are determined by different genes ( Falconer 1989; Roff 1997). In other words, different genes seem to control the dark pigmentation seen at the posterior edge of each segment, and the segment position interferes with the activity of such genes. A similar process has been clearly shown in D. kikkawai ( Gibert et al. 1999) where the dark pigmentation allele at a major locus is highly expressed in the posterior segments but progressively repressed in the anterior ones. Such a phenomenon may also occur in D. melanogaster even if, in that species, the genetic basis of abdomen pigmentation, which appears as a polygenic trait, is presumably more complex. Further investigations, (e.g., directional selection experiments) might help to untangle these interactions.
The genetic processes which specify the identity of each abdomen segment in D. melanogaster are progressively unravelled by developmental genetic analyses. Recent investigations on adult segments ( Kopp and Duncan 1997; Kopp et al. 1997; Struhl et al. 1997; Kopp, personal communication) have shown that several genes are expressed in the different compartments of each tergite and control their antero-posterior pattern which is evidenced by the distribution and orientation of micro- and macrochaetae.
The deposition of black melanin in the extracellular matrix of the tergite ectoderm is mediated by a transcription factor (optomotor blind, omb), and a higher activity results in a broader black stripe at the posterior margin of each segment ( Kopp et al. 1997). The activity of omb seems to be regulated at least in two ways: by the hedgehog signaling pathway, but presumably also by homeotic genes Abdominal A and Abdominal B (Kopp, personal communication).
A major, still unresolved problem in phenotypic plasticity studies concerns the possible occurrence of specific plasticity genes regulating the activity of trait genes ( Via 1993; Scheiner 1993a; Via et al. 1995). We may hypothesise that pigmentation genes, similar to the major abdomen pigmentation locus found in D. kikkawai, also exist in D. melanogaster. In that species, the genes are likely to be more numerous (see the lack of correlation between distant segments) and their expression is highly sensitive to temperature. A major problem remains to identify the target of environmental temperature. This may be the pigmentation genes themselves, but also the upstream regulatory genes, like omb. In the latter case, omb would be a plasticity gene. The analysis of the complex interactions between developmental genes, pigmentation genes and temperature will help to understand the realisation of the adult phenotype as a compromise between constraints and adaptations. In this respect, evolutionary genetics meets developmental genetics, with complementary approaches.
Acknowledgments
This work was supported by the Indo-French Centre for the Promotion of Advanced Research (IFCPAR, Contract 1103.1). We thank Dr. Artyom Kopp for helpful comments on an earlier draft.
