Body size and developmental temperature in Drosophila simulans: comparison of reaction norms with sympatric Drosophila melanogaster

Les normes de reaction de la taille du corps (aile et thorax) et du rapport aile/thorax ont ete analysees en fonction de la temperature de developpement par la methode des lignees isofemelles. Deux populations naturelles sympatriques francaises des especes soeurs Drosophila simulans et Drosophila melanogaster ont ete comparees. Les drosophiles ont ete elevees a sept temperatures constantes comprises entre 12 et 31°C, ce qui recouvre l'ensemble de la gamme des temperatures possibles pour ces deuz especes. La variabilite phenotypique entre les individus d'une meme lignee a ete analysee en utilisant les coefficients de variation, et la variabilite genetique en utilisant les coefficients de correlation intraclasse. La forme des courbes de reponse (ie, normes de reaction) a ete analysee par deux methodes: la variation des pentes et les ajustements polynomiaux. En accord avec la parente des deux especes, de nombreuses similitudes ont ete observees. En particulier les normes de reaction de l'aile et du thoras presentent un maximum a basse temperature, tandis que le rapport aile/thorax est une courbe sigmoide decroissante. De nombreuses differences ont aussi ete observees, parfois tres importantes. Au niveau phenotypique, D simulans est generalement plus variable que D melanogaster, tandis qu'au niveau genetique elle s'est averee en general moins variable. L'heritabilite varie avec la temperature, mais avec des modalites differentes dans chaque espece. Dans les deux especes, le dimorphisme sexuel (evalue par le rapport femelle/mâle) augmente avec la temperature, mais les valeurs et les courbes de reponse sont differentes. Les normes de reaction de l'aile et du thoras sont principalement differenciees par les TTMs (temperatures de taille maximale), avec des valeurs plus basses chez D simulans. Cette espece est egalement caracterisee par un rapport aile/thorax inferieur avec une TPI (temperature de point d'injection) plus elevee. Ces differences sont difficiles a interpreter. En effet, les variations de TTMs suggerent que D simulans pourrait etre plus resistante au froid que D melanogaster; en revanche le rapport aile/thorax plus faible de D simulans suggere une adaptation a la chaleur.

(Received 10 November 1995; accepted 7 July 1996) Summary -Reaction norms of two size-related traits (wing and thorax length) were analyzed in relation to growth temperature in a French natural population of Drosophila simulans, using the isofemale lines method. The wing/thorax ratio was also studied.
Data were compared to those of the sibling species Drosophila melanogaster from the same locality. Flies were reared at seven constant temperatures, representing the whole thermal range of the two species. Phenotypic and genetic variabilities were analyzed. For investigating the shape of the response curves (ie, reaction norms) two methods were used: analysis of slope variations and polynomial adjustments. As expected from the relatedness of the two species, many similarities were observed. Notably, the reaction norms of wing and thorax lengths exhibited a maximum at low temperature, while the wing/thorax ratio was a regularly decreasing sigmoid curve. Numerous and sometimes great differences were also observed. At the phenotypic level, D simulans was generally more variable, while at the genetic level, it was less variable than D melanogaster. Isofemale line heritabilities varied according to growth temperature, but with different patterns in the two species.
In both species, sexual dimorphism increased with temperature, but the average values and the response curves were different. The reaction norms of wing and thorax lengths were mainly characterized by different TMSs (temperatures of maximum size) with lower values in D simulans. This species was also characterized by a much lower wing/thorax ratio with a higher TIP (temperature of inflexion point). The possible adaptive significance of these variations remains unclear. Indeed, TMS variations suggest that D simulans could be more tolerant to cold than its sibling. On the other hand, the lower wing/thorax ratio of D simulans suggests a warm-adapted species. phenotypic plasticity / isofemale line / wing length / thorax length / wing/thorax ratio Résumé -Taille corporelle et température de développement chez Drosophila simulans : comparaison des normes de réaction avec l'espèce sympatrique Drosophila melanogaster. Les normes de réaction de la taille du corps (aile et thorax) et du rapport ailé/thorax ont été analysées en fonction de la température de développement par la méthode des lignées isofemelles. Deux populations naturelles sympatriques françaises des espèces sceurs Drosophila simulans et Drosophila melanogaster ont été comparées. Les drosophiles ont été élevées à sept températures constantes comprises entre 12 et 31 °C, ce qui recouvre l'ensemble de la gamme des températures possibles pour ces deux espèces. La variabilité phénotypique entre les individus d'une même lignée a été analysée en utilisant les coefficients de variation, et la variabilité génétique en utilisant les coefficients de corrélation intraclasse. La forme des courbes de réponse (ie, normes de réaction) INTRODUCTION Body size, which exhibits huge variations among living organisms, has long exerted a kind of fascination upon biologists. Size variations influence numerous biological traits, such as basal metabolism, duration of development or age at maturity (Reiss, 1989;Stearns, 1992;Charnov, 1993). Reciprocally, size is a target for natural selection and varies as a consequence of environmental pressures. For example, the old Bergman's rule describes, in numerous homeotherm species, an increase of size related to a colder environment. Finally size exhibits large variations between individuals of the same population, not only due to genetic differences but also due to phenotypic plasticity, related to different environmental conditions during development.
In Drosophila, allometric relationships are not well documented, although important size variations exist between species (Ashburner, 1989). Several species including Drosophila melanogaster and Drosophila simulans exhibit genetic latitudinal clines with a larger size under colder climate (David et al, 1983;Capy et al, 1993), these clines presumably being linked to temperature. Laboratory experiments keeping strains at different temperatures for many generations have demonstrated genetic size variations over time, ie, smaller flies at high temperatures and bigger ones at low temperatures (Powell, 1974;Cavicchi et al, 1985;Partridge et al, 1994).
These observations remind one of Bergman's rule, although Drosophila is an ectotherm so that we do not know why it should be better to be larger in a colder climate Partridge et al, 1994).
In natural populations, adult size exhibits a huge variability, presumably related to variations in feeding and thermal conditions (Atkinson, 1979;David et al, 1980David et al, , 1983Coyne and Beecham, 1987;Imasheva et al, 1994;Partridge et al, 1994;Moreteau et al, 1995). This phenotypic plasticity cannot be considered as completely neutral. For example, a positive phenotypic correlation exists between size and fitness in nature (Boul6treau, 1978;Partridge et al, 1987). Moreover, Coyne and Beecham (1987) demonstrated that size variations were to some extent heritable in spite of a large environmental component due to plasticity. However, a positive phenotypic correlation between body size and adult fitness components, together with the existence of additive genetic variance for body size, does not necessarily lead to the conclusion that body size is the target of selection (Rausher, 1992).
Up to now, quantitative genetic variations among natural populations, including latitudinal clines, have generally been investigated at a single temperature (with the exception of Coyne and Beecham, 1987), most often 25 °C (David et al, 1983;David and Capy, 1988;Capy et al, 1993). On the other hand, natural selection, which is presumed to be responsible for the clines, acts at various temperatures in different localities and, in all cases, upon highly variable phenotypes. Moreover, temperature is the most important abiotic factor explaining geographic distribution and abundance of species in Drosophila (David et al, 1983;Parsons, 1983;Hoffmann and Parsons, 1991). Thus, for a better understanding of these problems, several temperatures must be investigated and compared. In other words, we have to investigate the relationship between developmental temperature and phenotypes, ie, the reaction norms of various traits.
Generally, authors who were interested in the genetics and evolution of reaction norms only considered two environments and consequently linear norms Lande, 1985, 1987;Lyman, 1989, 1991;De Jong, 1990;Scheiner, 1993a;Via, 1993). Gavrilets and Scheiner (1993) underlined, however, the necessity of studying nonlinear norms and proposed to model them using polynomial adjustments. Indeed, when a broad range of environments (eg, temperature) is investigated, norms of quantitative traits are as a rule nonlinear (David et al, 1983(David et al, , 1990Delpuech et al, 1995).
A recent controversy has developed concerning the genetics of plasticity. Various authors have considered that the mean value of a trait and the shape of the reaction norm should be distinguished. In other words, genes regulating the position of the curve (trait mean value genes) and genes regulating plasticity (shape genes) might coexist (Bradshaw, 1965;Lyman, 1989, 1991;Weber and Scheiner, 1992;Scheiner, 1993ab;Gavrilets and Scheiner, 1993). But this conception was criticized by Via (1993Via ( , 1994 who considered it an unnecessary complication, and recent papers have tried to reconcile these two approaches (Van Tienderen and Koelewijn, 1994;Via et al, 1995).
Analysing plasticity leads to several related questions. What is the genetic basis of the reaction norms, and are there specific genes for their shape? What is the significance of the norm? Is it a consequence of internal constraints or is it adaptive, ie, shaped by natural selection?
It is generally recognized that, before developing a theory on the evolution of reaction norms, many more empirical data are needed, relating the norms with ecological adaptations and life history parameters. In this respect, it will be easier to compare different species (Harvey and Pagel, 1991) since a larger evolutionary time should have permitted a broader divergence of the norms, especially if they were shaped by natural selection. In this paper, we investigated the reaction norms of size traits of a natural population of D simulans from France, and compared the results with those obtained for the sibling D melanogaster from the same locality . We found similarities between the two species but, more interestingly, numerous significant differences. These differences demonstrate that, within a relatively short evolutionary time (about 2 million years) reaction norms have diverged. The possible adaptive significance of these variations is discussed.

MATERIALS AND METHODS
A D simulans population was collected in a vineyard in Pont de la Maye near Bordeaux (southern France). Variability of size according to temperature was analyzed, and compared to a population of D melanogaster collected in the same locality and previously studied .
The isofemale lines method was used. Wild living females were collected with banana traps and used to establish 20 isofemale lines, and ten of them were then randomly chosen. For each, ten pairs of the first laboratory generation were used as parents. They oviposited at room temperature (20 ! 2 °C) for about half a day. A rich feeding medium, based on killed yeast, was used for the development (David and Clavel, 1965). Such a food prevents crowding effects which could affect fly size. Density ranged between 100 and 200 eggs per vial. Vials with eggs were then transferred to one of seven experimental constant temperatures (12, 14, 17, 21, 25, 28, 31 °C). Measured flies thus correspond to the second laboratory generation. Such a procedure is a necessity for obtaining enough offspring (see Moreteau et al, 1995 for discussion). It also eliminates possible maternal effects and provides Hardy-Weinberg proportions within lines. ' From each line at each temperature, ten females and ten males were randomly taken. Their wing and thorax lengths were measured with a micrometer in a binocular microscope. Total wing length was measured from the articulation on the side of the thorax to the distal tip. Thorax was measured on a left side view, from the base of the neck to the tip of the scutellum. Analyses were made directly on measurements expressed in mm x 100, since a preliminary analysis with logtransformed data failed to show any scaling effect.
Statistical analyses and orthogonal polynomial adjustments were made with STATISTICA software (Statistica Statsoft Inc, 1993).

RESULTS
Variation of wing and thorax length: mean of the ten lines

Reaction norms
The response curves (fig 1) show that females are larger than males in both species and that D melanogaster is larger than D simulans. In both species, a maximum seems to exist at a low temperature. A steep decrease from this maximum is observed when temperature increases, and a shorter one when temperature decreases. In both species, significant differences exist between the reaction norms of wing and thorax. Finally D simulans seems to exhibit its maxima for both traits at lower temperatures than D melanogaster. This problem will be analyzed further.

Sources of variation
Variations were investigated simultaneously on the two traits in D sim!alans with MANOVA (table I). Sex and temperature are the main sources of variation. A highly significant line effect demonstrates their genetic heterogeneity. The temperatureline interaction, also highly significant, shows that the reaction norms of the different lines are not parallel but exhibit different shapes. Finally the sex-temperature interaction means that males do not react exactly as the females do. These results are similar to those obtained in D melanogaster , except that the sex-line interaction, which is not significant in D simulans, was significant in D melanogaster. These results are illustrated in figure 3 as a correlation between male and female t values. In D simulans, t values for the two traits can be divided into two groups: high values (= higher heritability) are observed at medium temperatures (21, 25, 28 °C) and low values at extreme temperatures (12, 14, 31 °C). Means of these two groups are 0.34 ! 0.03 and 0.12 ! 0.02 respectively and statistically different (Student's test, not shown). In D melanogaster, no temperature effect was observed for the wing, but a difference between high and low temperatures was observed for the thorax, with a higher genetic variability at high temperatures.
For the wing/thorax ratio (table IV), the general mean calculated on 14 observations is 0.27 ! 0.03, much lower than in D melanogaster (0.57 ! 0.02).
Analysis of the shape of reaction norms: slope variations and derivative curves Wing and thorax For each isofemale line, length variation for a given temperature interval allows the calculation of a slope (ie, length variation per degree), by a linear intrapolation.
Repeating this process for successive intervals produces an empirical derivative of the reaction norm.
An ANOVA (not shown) was conducted on the slopes in D simulans. Results were similar for wing and thorax with a very significant temperature effect, demonstrating nonlinear norms. Contrarily to D melanogaster, there was no significant sex effect. No line effect was detected, as in the sibling species. In the two species a clear line-temperature interaction shows that derivative curves have different shapes among lines. Finally, a highly significant sex-temperature interaction is present, which was not found in D melanogaster.
Average curves and single line curves are given in figure 4, for wing in females only. In the two species, average curves ( fig 4a) show a progressive decrease from positive to negative values. These values are significantly lower at low temperature in D simulans and not significantly greater than zero. This means that the point where this derivative curve crosses the null line, which corresponds to the temperature of maximum size (TMS), is far less obvious in D simulans than in D melanogaster, especially for the thorax (see also fig 1). This observation is confirmed by the examination of the curves of different lines (fig 4b). Indeed in D simulans, wing length never reached the zero value in two lines, and for thorax length (not shown) the slope often crossed the null line several times. Hence in D simulans, a TMS can be calculated by using the average curves, but not for each isofemale line. Average curves point TMS values at 13.5 °C for wing and at 16 °C for thorax in D simulans, and at 16 and 19 °C respectively in D melanogaster. In other words TMS values appear to be lower in D simulans than in D melanogaster. For comparing the two traits, slopes were standardized and expressed as a percentage of the mean (curves not shown). With such a transformation , the amplitudes of variation for the two traits become similar. In D melanogaster the variation range was greater: the overall phenotypic plasticity seems to be less pronounced in D simulans.
Wing/thorax ratio Slopes of the wing/thorax ratio were calculated in the same way and an ANOVA (not shown) demonstrated a major effect of temperature, a low sex effect, no line effect but a significant line-temperature interaction. Average slope variations are illustrated in figure 4c for females. In the two species, average derivative curves are U-shaped indicating that the maximum phenotypic plasticity occurs at intermediate temperatures, and also that the wing/thorax ratio varies according to a decreasing sigmoid curve (see fig 2). A regular feature in D simulans is that the derivative curve is always above that of D melanogaster. Notably, at extreme temperatures, zero values correspond to the fact that the curve of the wing/thorax ratio was horizontal (see fig 2). Moreover, the overall amplitude of variation is larger in D simulans. Analysis of the shape of the reaction norms: polynomial adjustments Degree of polynomial adjustments After a theoretical study of linear norms, Gavrilets and Scheiner (1993) suggested that nonlinear norms should be adjusted to second degree polynomials, according to the formula P(t) = go + g l t + g 2 t 2 (if we are dealing with temperature, P(t) is the phenotype value at temperature t). The authors proposed for go, the intercept, a genetic significance fixing a basic value to the studied trait, while g l , the slope, could be a genetic parameter of adaptation to the environment, and g 2 a genetic parameter of curvature. A second degree polynomial implies that the derivative curve (ie, slope variation) is linear. Such was not the case for the three traits (see fig 4), so that at least a third degree adjustment should be used. Incomplete polynomials could also be used, for instance with no t 2 term. The validity of the various adjustments was assessed by adjusted R 2 values, a poor adjustment being characterized by a low adjusted R 2 . A third degree equation proved to be convenient for the wing/thorax ratio. For wing and thorax lengths, considering the similar shapes in the two species, we imposed a constraint on the adjustment, ie, the existence of a plausible TMS calculated by solving the equation P'(t) = 0. For third and fourth degrees, two or three solutions were obtained respectively, which needed to be checked to know which one corresponded to the overall maximum. Finally, for overall homogeneity, all the wing and thorax curves were adjusted to fourth degree polynomials, even those which were compatible with third degree polynomials. Also, similar adjustments were made with the data of D melanogaster to compare the two species. Such adjustments were not made in a previous paper .

Wing and thorax
Even with fourth degree polynomials, there were still some inadequate TMS values, for instance, 6.3 °C for a male wing. This often occurred from an abnormal value at a single temperature (= rearing accident?) which modified the adjustment equation and thus the TMS. Such cases represented six out of the 40 adjustments made on D simulans, but only two of them (for male thorax) deviated from a reasonable value.
Choosing a fourth power polynomial leads to much more heterogeneous g i parameters than an adjustment in t 2 . For instance, for females wing in D simulans, the ten go values ranged from 62 to 69 with the t 2 adjustment, and from &mdash;79 to +93 with the t 4 adjustment. A similar conclusion was obtained for all other parameters. Fortunately, calculation of critical points, such as TMS values, provided much less variable values, thus confirming previous observations on ovariole number (Delpuech et al, 1995).
In both sexes of D simulans thorax TMS values were generally higher than wing ones, as in D melanogaster. Also significantly higher values were demonstrated in females (ANOVA, not shown). In D simulans an overlap of TMSs of the two traits was observed, contrarily to D melanogaster. Mean values are given in table V and compared to those of D melanogaster. In all cases, TMS values are significantly higher for the latter species. Another striking species difference is the large dispersal among lines of D simulans contrasting with a better homogeneity in D melanogaster (see CVs in table V). Finally, in all cases, values of males and females of the same line were positively correlated, suggesting that they provide, at least in part, the same genetic information.
As in David et al (1994), values of both sexes were averaged for each trait. A scatter plot of wing and thorax TMS values (fig 5) clearly contrasted the two species. Interestingly, a positive correlation is found in D melanogaster while a non-significant but negative correlation is found for the eight lines of D simulans (excluding two lines with aberrant TMS for male thorax). The between-line heterogeneity seems to be mainly due to thoracic variations.
Taking all values into consideration, average curves were also adjusted to the fourth degree and gave TMS values of 13.5 °C (females) and 12.4 °C (males) for the wing, and of 16.1 °C (females) and 13.2 °C (males) for the thorax. These values are lower than in D melanogaster (respectively 15.6 and 14.8 °C for wing, 19.2 and 17.6 °C for thorax). They are close to the mean values of the ten lines given in table V and thus characterize the species. Interestingly, the g i parameters of the average curves were similar to the mean values of the g i of the ten lines.
Wing/thorax ratio The g i parameters of the third degree polynomial were very variable; CVs ranged between 16.5 and 40% for the four female coefficients (mean CV = 28%) and between 22 and 65% (mean CV = 46%) for males. Curves were then characterized by their temperature of inflexion point (TIP), ie, the temperature where the second derivative becomes null.
One line posed a problem in both sexes (aberrant inflexion point value because of a hyperbolic rather than sigmoid shape) and was excluded. TIPs (fig 6) ranged between 19.9 and 22.8 °C (mean: 21.1 ! 0.3 °C) in females and between 19.9 and 21.3 °C (mean: 20.6 ! 0.2 °C) in males. There was neither line nor sex effect (ANOVA, not shown).
In D melanogaster, the same adjustments produced far more variable g i coefficients: mean CV of 69% in females and 92% in males, ie, more than twice as large as in D simulans. This also resulted in a much greater dispersal of the TIP values of the different lines (see fig 6). Also the TIPs were on the average significantly lower (ANOVA, not shown) in D melanogaster than in D simulans: 19.0 f 0.9 °C in females and 16.9 t 1.2 °C in males.
A final observation was that for a given temperature, the ratio of the polynomially adjusted wing value to the polynomially adjusted thorax value was the same as the polynomially adjusted wing/thorax ratio.

DISCUSSION AND CONCLUSION
Our results need to be discussed from two different points of view: a methodological approach for the description of reaction norms, and the comparative evolutionary biology of the two sibling species. A major, still unsolved problem, will be to decide which species is better adapted to a warmer environment.
How should empirical reaction norms be investigated?
In Drosophila, genetic plasticity of quantitative traits such as wing and thorax length was first investigated over two environments Lyman, 1989, 1991;Weber and Scheiner, 1992;Scheiner, 1993a) and a linear model was used. When a broad range of environmental conditions is used, as such was the case here, most reaction norms are, however, nonlinear (David et al, 1983Gavrilets and Scheiner, 1993) and this raises a major problem: what is the best way to describe and analyse the shape of the curve? Factors of variation can be identified with ANOVA or MANOVA, as well as numerous interactions which demonstrate, for example, that the norms significantly differ among isofemale lines from the same population. More precise analyses are however needed for describing the norms, and two kinds of methods may be used: slope variations and mathematical adjustments.
Analysis of slope variations was used by David et al (1990) for demonstrating different pigmentation norms in successive abdominal segments. This method can be of general use for comparing different traits or species, and significant differences may be easily demonstrated. Also the overall shape of the reaction norm may be inferred from the shape of its derivative. In D simulans, and contrarily to D melanogaster , this method was not satisfactory (problems in TMS values determination) and the shapes of the curves had to be studied by mathematical adjustments.
An adjustment to a mathematical model should be a better method but numerous equations could be chosen. In the present case there was no a priori reason for guiding the choice and thus we used a general method, ie, a polynomial adjustment, as suggested by Gavrilets and Scheiner (1993) and Via et al (1995). Because of the great variability among the polynomial coefficients of various lines, it appeared difficult to give them a genetic sense, contrarily to what has been suggested (Gavrilets and Scheiner, 1993). These parameters are, however, conveniently used for calculating critical points of the curves, especially the temperature of maximum size (TMS) for wing and thorax lengths or the temperature of inflexion point (TIP) for the wing/thorax ratio. Reaction norms appear to be better characterized by these points, which are less variable and seem to have a biological significance, and presumably also a genetic basis. In this respect, we found that TMS values of males and females of the same line were positively correlated in both species and, among lines, thorax and wing TMS values were also correlated in D melanogaster . Interestingly in D melanogaster, calculating the TMS values either by considering slope variations or with polynomial adjustments provided similar results. In D simulans, fourth power polynomials had to be used instead of quadratic ones for a better characterization of TMS values. But even in that case, the adjustment could not be performed for some isofemale lines. This may reflect either true genetic peculiarities of these lines or some experimental imprecisions. This problem needs further investigation, for example, by analyzing the same line over two successive generations.
Similarities between the two species Similarities between closely related species are expected because of phylogenetic constraints and also from a possible similarity of their ecological niches (Harvey and Pagel, 1991). In the present study, numerous similarities were observed, which are briefly summarized below.
In the two species females are larger than males, and this could be a general result in most Drosophila. The female/male ratio gives similar data for wing and thorax and could be considered as a good measure of sexual dimorphism. This dimorphism is a phenotypically plastic trait with minimum values at low temperatures in both species.
Reaction norms of the three characters (wing and thorax length and wing/thorax ratio) are nonlinear and present the same sources of variation. Wing and thorax both exhibit a maximum at low temperature. The response of the wing/thorax ratio to temperature is a sigmoid decreasing curve, similar for both sexes.
In all cases, coefficients of intraclass correlation (t) were significantly greater than zero, demonstrating (Hoffmann and Parsons, 1988) a high heritability of the traits. Moreover a regular line-temperature interaction indicates significant genetic variations in the shapes of reaction norms among isofemale lines.
The within-line CVs varied with temperature in all cases, with maxima at extreme temperatures. This is likely due to an increase of the developmental noise under stressful conditions. In both species, the wing/thorax ratio is less variable (lower CVs) than the traits themselves. This is due to the fact that wing and thorax variations are correlated at the individual level.
Differences between the two sibling species Numerous and important differences were found between the two species. These differences demonstrate that canalization during development is not very strong so that the investigated traits could diverge, either as a consequence of drift or of ecological adaptation.
As already known from numerous observations (see Capy et al, 1993) D simulans is a smaller species. We may argue that speciation was accompanied by size gene variations, determining the position of the reaction norms on the Y axis.
Sexual dimorphism presented different reaction norms in the two species. It is unfortunate that we do not have a convenient evolutionary theory for sexual dimorphism in organisms like Drosophila (Charnov, 1993). Heritability of size traits was different in the two species, contrarily to what was found by Capy et al (1994) in a broad survey of numerous populations reared at a single temperature (25 °C). In our study of two sympatric populations, D melanogaster appeared on the average more variable than D simulans. In both species variations of isofemale line heritabilities were observed according to developmental temperature, but with different patterns for different traits. These differences are difficult to interpret, and many more comparative studies should be undertaken. At the within-line level, phenotypic variability exhibits a major environmental component (Falconer, 1989) and thus reflects in some way the reactivity of individuals to minor variations in culture vials (eg, food desiccation or larval competition). This reactivity may be estimated by considering the CVs. D simulans appeared more variable than D melanogaster for thorax length over the whole temperature range and for the other two traits at high temperatures only. These results are somewhat surprising, because phenotypic variability was previously found to be similar in the two species (Capy et al, 1994). A problem remains: are these results general to the species or specific to the studied populations? A major difference between the two species concerns their TMS values, which are much lower in D simulans than in D melanogaster, with a translation toward the left in D simulans. As the thermal ranges are about the same in the two species (Cohet et al, 1980, and this work), it was more difficult to calculate TMS values in D simul d ns. A careful analysis showed that, besides the translation, the shapes of the norms were somewhat different in the two species. Within species, a significant line-temperature interaction demonstrates genetic variations in the curve shapes. Finally, the heterogeneity of TMS values between lines is larger in D simulans than in D melanogaster, in spite of a lower genetic variability within each temperature in the former species. These observations argue in favor of a genetic regulation of the reaction norm shape.
A last but major difference between the two species concerns the wing/thorax ratio which is much smaller in D simulans and presents higher TIPs.
All these differences support a general trend: the more the two species are compared, the more they appear different (see Capy et al, 1993Capy et al, , 1994 for discussion and references).
Reaction norms and the thermal adaptation of the two species Since we investigated the effects of developmental temperature, we must ask the question: is one species better adapted to a colder or warmer climate? Answering this question is difficult, since we have conflicting observations.
Although the thermal laboratory ranges are similar (12-31 °C) in the two species , ecological surveys (Louis, 1983) have shown that D simulans is generally more abundant than D melanogaster in warm temperate and subtropical regions, while it is rare or even absent in cold regions where D melanogaster is still present. These observations lead to the classical interpretation that D simulans is less tolerant to cold than D melanogaster (Parsons, 1983). So our results are surprising. Indeed, even if the biological meaning of a TMS is not clearly established, we expect that a maximum should be related to some optimum (Parker and Maynard-Smith, 1990;Gabriel and Lynch, 1992;Stearns, 1992). Could we suppose, then, that D simulans is more adapted to cold than D melanogaster, contrarily to what was believed up to now, and that reaction norms indicate the direction of adaptation? In fact, this hypothesis is not unlikely. Indeed, from an ecological point of view, D melanogaster enters human buildings where it is protected during winter, whereas this is not the case for D simulans (Rouault and David, 1982).
So the latter will suffer lower temperatures than D melanogaster during winter, and hence will be selected for cold tolerance. Two other arguments support this hypothesis. Firstly, in D melanogaster, males reared at 12 or 13 °C are sterile, whereas this is not the case in D simulans (David, unpublished observations).
Other observations suggest however a reverse interpretation. The wing/thorax ratio, which is inversely proportional to wing loading and wing beat frequency (P6tavy et al, 1992, 1996) decreases with temperature, presumably in relation with a better muscular efficiency at higher temperature (Reed et al, 1942). In other words, a low wing/thorax ratio could indicate a warm adapted phenotype, and D melanogaster. Moreover the TIP, which corresponds to a maximum of plasticity, is higher in D simulans. Even if the possible relationship between the TIP and the optimum flight temperature remains to be investigated, this could support the hypothesis of a better adaptation of D simulans to a warmer environment. Molecular studies at the within-population level have shown that D simulans was generally more polymorphic than D melanogaster (Aquadro et al, 1988;Begun and Aquadro, 1991;Aquadro, 1992). To explain this observation, the former authors suggested that the population effective number is higher in D simulans, due to a higher migration rate and a better dispersal capacity. In this respect the lower wing/thorax ratio in D simulans could be more a dispersal adaptation than a thermal adaptation. However, in spite of numerous studies (Brodsky, 1994) we do not know what is the best strategy for dispersal, ie, high speed correlated with high wing loading and relatively short flight duration, or vice versa. In conclusion, the two sibling species which are increasingly investigated as a model for evolutionary studies, appear very different when more thoroughly analyzed, and interpretations are difficult. Concerning the evolution of reaction norms and their possible relationship with thermal adaptation, further comparative studies are needed, either on geographic populations of the two sibling species and on other Drosophila species clearly adapted to warm or cold climates.