PHENOTYPIC PLASTICITY IN COLONIZING POPULATIONS OF Drosophila subobscura

The phenotypic plasticity of some quantitative traits of two colonizing populations of Drosophila subobscura (Davis and Eureka, California) was studied. Temperature effects and the effect of rearing in the laboratory were studied. Laboratory rearing during four generations at 18oC significantly increased the wing and tibial length. This increase was similar to that obtained when the flies were reared at 13oC during two generations.The low temperature environment can be considered more stressful for females than for males, as shown by the increase of phenotypic variance. The two populations analyzed had great phenotypic plasticity in spite of the genetic bottleneck during the colonization event. Our study shows that keeping flies for a relatively short time in the laboratory significantly changes some quantitative traits, emphasizing the need to analyze flies immediately after collecting them in order to obtain reliable estimates for the analysis of these traits in natural populations.


INTRODUCTION
In general, flies from cooler regions tend to be larger than flies from warmer regions.Both Bergmann's and Allen's rules express the relationship between body surface and environmental temperature, but these rules may or may not concern heat conservation in poikilotherms (Ray, 1960).Size seems to be genetically controlled.Populations maintained in the laboratory at different temperatures diverge genetically with respect to body size (Anderson, 1966(Anderson, , 1973;;Powell, 1974;Cavicci et al., 1989).Furthermore, Drosophila flies from humid tropical and temperate zones grown at different temperatures show a similar trend in body size and weight phenotypic differentiation (Ray, 1960).Powell (1974) suggested that some populations show genetic variation for body size for reasons other than temperature adaptations and it is only in the artificial, laboratory environment that temperature is the selective force on this variance.On the other hand, phenotypic variance has been shown to increase in stressful environments (Burla and Taylor, 1982;Barker and Krebs, 1995), suggesting that this phenomenon could contribute to the increase observed in genetic variation in marginal environments, with more rapid evolution during periods of special stress.
Phenotypic plasticity is exhibited as environmentally mediated change in the phenotype (Via et al., 1995).Although the study of phenotypic responses has a long history recently there has been a new interest in this phenomenon at both theoretical and experimental levels (review in Via, 1994).Laboratory conditions constitute a completely new environment for the flies, in which genetic variation for body size could be expressed in a different way from the wild.Drosophila subobscura is a Palearctic species distributed all over Europe (Krimbas, 1993).This species was detected for the first time in Puerto Montt (Chile) in February 1978 (Brncic et al., 1981) and subsequently it has spread very quickly all over the country.It was also detected in North America in summer 1982 (Beckenbach and Prevosti, 1986) and has become established all over the Western Pacific coast, from British Columbia to Ojai (California) (Prevosti et al., 1989).We examined two colonizing populations to ascertain whether the genetic bottleneck that took place during the founder event (Prevosti et al., 1983;Ayala et al., 1989;Mestres et al.;1990, Balanyà et al., 1994) ) had any influence on the phenotypic plasticity of this species.

MATERIAL AND METHODS
Two American populations of D. subobscura, Davis and Eureka (California) were analyzed.The population of Davis (38º 32'N, 121º 46'W) is located in the Californian Central Valley (altitude 18 m) and has extreme weather conditions (the range of temperatures is between -6.1ºC and 43.9ºC, annual average 16.2ºC, annual rain 345 mm).Eureka (40º 48'N, 124º 10'W) is a coastal locality (altitude 18 m) in Northern California (the range of temperatures is between -2.8ºC and 27.8ºC, annual average 12.2ºC; annual rain 803 mm).Six quantitative traits were studied in males: wing length was measured along longitudinal vein IV, divided into two segments, L 1 (from the base of the fourth longitudinal vein to the posterior cross vein) and L 2 (from the posterior cross vein to the extreme of the media, according to Robertson and Reeve (1952) and Prevosti (1955)); wing width (W) from the extreme of the V vein to the costal border, running perpendicular to the third vein (Figure 1); tibial length (TL), and two meristic traits: number of teeth of the proximal (PC) and distal (DC) sex combs (on the right leg).In females only the continuous variables L 1 , L 2 , W and TL were analyzed.
The experimental procedure for each population

ABSTRACT
The phenotypic plasticity of some quantitative traits of two colonizing populations of Drosophila subobscura (Davis and Eureka, California) was studied.Temperature effects and the effect of rearing in the laboratory were studied.Laboratory rearing during four generations at 18ºC significantly increased the wing and tibial length.This increase was similar to that obtained when the flies were reared at 13ºC during two generations.The low temperature environment can be considered more stressful for females than for males, as shown by the increase of phenotypic variance.The two populations analyzed had great phenotypic plasticity in spite of the genetic bottleneck during the colonization event.Our study shows that keeping flies for a relatively short time in the laboratory significantly changes some quantitative traits, emphasizing the need to analyze flies immediately after collecting them in order to obtain reliable estimates for the analysis of these traits in natural populations.
Departament de Genètica, Facultat de Biologia, Universitat de Barcelona,Av. Diagonal,645,08071 Barcelona,Spain.Send correspondence to L.S. Fax: +34-3-411-0969.E-mail lluis@ porthos.bio.ub.es was as follows: 10 wild females were placed in individual rearing vials at 18ºC.All the F 1 offspring from each vial were put into plastic chambers measuring 13 x 9 x 6 cm (one plastic chamber for each vial) to obtain eggs.One day later, 100 eggs were collected from each of the 10 chambers and placed in individual vials (100 eggs per vial) in order to prevent larval competition and kept at 18ºC.The same procedure was repeated with another independent sample of 10 wild females but the vials were keat at 13ºC.For measuring the quantitative traits 10 F 2 males and 10 F 2 females were selected at random from each vial.Thus, 100 F 2 males and 100 F 2 females were analyzed for each temperature (18º and 13ºC).All these adult flies were preserved in glycerine-alcohol (1:2) until the measurement of the quantitative traits.The remaining F 2 offspring kept at 18ºC were maintained for two generations at the same temperature (100 eggs were chosen at random from each chamber in each generation to prevent larval competition), and the same procedure for the measurement of quantitative traits was carried out in the F 4 generation.The measurements were made under a compound microscope at 50X magnification with an ocular micrometer.Wing length and width and tibial length were recorded to the nearest unit of the micrometer scale, which corresponded to 0.029 mm.The data were statistically analyzed by ANOVAS for each variable and by three one-way multivariate analyses of variance (MANOVA), considering as factors the temperature (13º and 18ºC), the population (Davis and Eureka) and the number of generations in the laboratory at 18ºC (F 2 and F 4 ), respectively.In all cases sexes were analyzed separately due to the significant sexual differences in the variables analyzed.A canonical analysis to study the differences among the six characterized groups (Eureka 18ºC F 2 , Davis 18ºC F 2 , Eureka 13ºC F 2 , Davis 13ºC F 2 , Eureka 18ºC F 4 and Davis 18ºC F 4 ) was also performed.

RESULTS
The mean values for the continuous variables L 1 , L 2 , W and TL (Tables I and II) were very similar for the  L1 and L2: Segments from the base of the fourth longitudinal vein to the posterior cross vein and from the posterior cross vein to the extreme of the media, respectively; W = wing width from the extreme of the V vein to the costal border, running perpendicular to the third vein; TL = tibial length.13ºC F 2 and 18ºC F 4 groups, in both sexes.Furthermore, the 13ºC environment could be classified as more stressful for the females, as shown by the higher values of the standard deviation in this sex.The differences between the standard deviation at 13º and 18ºC were significant at the 0.05 level for the females in all cases (Table III).
A comparison was made between F 2 and F 4 females from Davis reared at 18ºC to determine the effect of rearing in the laboratory (Table IV).For the multivariate case, there was homogeneity of the variance-covariance matrices (P = 0.174) and the difference between the mean vectors was significant (F = 34.78 with 4 and 195 d.f.).The results were similar for the males from Davis and for both sexes in Eureka.
A comparison between F 2 males from Eureka at 13º and 18ºC was made to determine the effect of temperature (Table V).Although in this case the Box-M test of homogeneity of the variance-covariance matrices is significant (P = 0.011), the use of the MANOVA procedure is still justified because most of the signs of the correlation coefficients coincide.A decrease of 5ºC provoked a clear increase in the variances (Table III), which could reflect a loss of homeostasis.Furthermore, the mean vectors were also significantly different.This result is in agreement with those obtained by other authors (Ray, 1960;Anderson, 1966Anderson, , 1973;;Sokoloff, 1966;Powell, 1974).The corresponding results for females and for both sexes in Davis were equivalent.The effect of the population (Davis versus Eureka) was determined for F 4 males at 18ºC (Table VI).The variance-covariance matrices were homogeneous (P = 0.139) and the difference between the mean vectors was significant, but when considering the analysis for each variable separately, some of the groups did not differ significantly (P = 0.158 for variable L 1 ; P = 0.850 for L 2 ; P = 0.399 for W and P = 0.194 for TL).
A canonical analysis was made to determine the similarities between groups (Table VII, Figure 2).Although there was no homogeneity among the variance-covariance matrices, the elements of these matrices have, in general, the same sign, which justifies the application of the method (Cuadras, 1991).The first two canonical axes explain 89.5 and 97.91% of the total variance in males and females, respectively, which is more than sufficient for the bi-di- For abbreviations see Tables I and II.For abbreviations see Tables I and II.For abbreviations see Table I.
mensional representation of the characterized groups.The 18ºC F 2 group from both populations clearly separates from the other two groups (18ºC F 4 and 13ºC F 2 ), both in males and females.

DISCUSSION
In general, the response of body size to temperature is considered to be adaptive and due to natural selection.The developmental system responding to the growth environment could be a phenomenon of adaptive plasticity (Schmalhausen, 1949;Bradshaw, 1965;Gomulkiewicz and Kirkpatrick, 1992).The increase in body size and cell size resulting from development at low temperature has also been considered a case of adaptive phenotypic plasticity (Partridge et al., 1994).David et al. (1994) also detected a response of wing and thorax lengths to temperature but found significant variations between lines and significant line-temperature interactions, demonstrating different norms of reaction among the various lines.Alto-gether, in the present study the main conclusions that can be drawn from the MANOVA results and the canonical analyses of the characterized groups are the following: temperature, although important, is not the only factor that explains the phenotypic differentiation of Drosophila flies kept in the laboratory.The size of the flies reared in the laboratory at 18ºC for four generations was equivalent to the size of the flies reared at 13ºC for only two generations.The response was much clearer for continuous variables.Meristic variables (number of teeth of the sex combs) did not differentiate appreciably.The multivariate mean tests were always significant, whether we consider laboratory rearing, temperature or the population effects.This is expected due to the sensitivity of the multivariate techniques and to the sample size.On the other hand, the univariate tests show similarity between the Davis and Eureka populations.As pointed out above, the colonization of the American continent by D. subobscura is a recent phenomenon.The colonizing populations are very much alike genetically (Prevosti et al., 1988 and1989;Ayala et al., 1989;Mestres et al., 1990Mestres et al., , 1992Mestres et al., , 1995;;Balanyà et al., 1994).This genetic similarity could explain the resemblance between Eureka and Davis populations in terms of quantitative traits: the multifactorial genotype controlling these traits would not have differentiated significantly since the colonization took place, in spite of the environmental differences between these two localities (Pascual et al., 1993).The two populations analyzed still show great phenotypic plasticity in spite of the genetic bottleneck during the founder event (Prevosti et al., 1989).This result is in agreement with the empirical evidence obtained from Drosophila and housefly populations, supported by several theoretical models (Bryant et al. 1986;Goodnight, 1987;Lewin, 1987;Carson, 1990) indicating that genetic variance available to selection may actually increase following a population bottleneck.
Finally, analysis of the correlation of quantitative traits with environmental factors has been widely used to detect natural selection in the wild (Ford, 1975;Endler, 1986), and some data are available on the existence of latitudinal clines for quantitative traits in Palearctic populations of D. subobscura (Prevosti, 1955, Misra andReeve, 1964;Pfriem, 1983;Pegueroles et al., 1995).Nevertheless, studies of this kind rely heavily on all samples being reared under the same laboratory conditions prior to measurement.As it is clearly shown in our analysis of the effect of rearing in the laboratory, all samples should be measured immediately after being collected, or at least be kept in the laboratory for the same period of time, to get reliable estimates of the existence of natural selection acting on quantitative traits in natural populations.

ACKNOWLEDGMENTS
Research supported by Grant PB86-0014 of the Dirección General de Investigación Científica y Técnica (Spain).We are grate-

Figure 1 -
Figure 1 -Continuous variables measured in the wing of D. subobscura (L 1 , L 2 , and W).See text for details.

Table I -
Mean value (M) and standard error (SE) of the continuous variables L1, L2, W and TL, for the Drosophila subobscura females of each group -Eureka 18º F 2 , Davis 18º F 2 , Eureka 13º F 2 , Davis 13º F 2 , Eureka 18º F 4 and Davis 18º F 4 (100 individuals measured in each group).

Table II -
Mean value (M) and standard error (SE) of the continuous variables L1, L2, W and TL and the meristic variables DC and PC, for the Drosophila subobscura males of each characterized group -Eureka 18º F 2 , Davis 18º F 2 , Eureka 13º F 2 , Davis 13º F 2 , Eureka 18º F 4 and Davis 18º F 4 (100 individuals measured in each group).
DC and PC = Distal and proximal sex combs.For other abbreviations see legend to TableI.

Table V -
Effect of temperature on the quantitative variables L 1 , L 2 , W and TL and meristic variables PC and DC with the statistical values (F),

Table IV -
Effect of rearing in the laboratory on the quantitative variables L 1 , L 2 , W and TL with the statistical values (F), degrees of freedom (d.f.) and significance levels.Comparison of Davis 18ºC F 2 and Davis 18ºC F 4 populations (200 Drosophila subobscura females measured).

Table III -
Results of F-test for the comparisons of variances between 18ºC (F 2 and F 4 ) and 13ºC (F 2 ) for Drosophila subobscura females in both populations (Davis and Eureka).

Table VII -
Results of the canonical analysis of the characterized groups (see legend to Figure2).
ful to Dr. C. Arenas (Dept.d' Estadística, Universitat de Barcelona) for her valuable suggestions.We also thank Mr. Robin Rycroft (S.A.L., Universitat de Barcelona) for the English correction.