Introduction

The adaptive nature of chromosome inversions in Drosophila has been accepted since Dobzhansky's classic work (Wright and Dobzhansky, 1946), which first established that different chromosomal arrangements were under natural selection.

As no recombinants are produced between genes within different inversions – crossingover is suppressed within the inverted region in inversion heterozygotes – Dobzhansky (1970) believed that inversions could be considered supergenes, which accumulated coadapted alleles. Ford (1964) thought that including coadapted alleles in a chromosome inversion was the most effective way to ensure that they would segregate as a block, to produce only high fitness phenotypic variants. Ford also proposed that polymorphic adaptations occurred mainly through a genetic architecture in which genes are tightly linked so that they act as a single unit. Inversion polymorphisms in Drosophila have been associated with several fitness-related characters (Rodríguez et al, 1999), behaviour (Dahlgaard et al, 2001) and morphological traits (Izquierdo et al, 1991; Bitner-Mathé et al, 1995; Bertrán et al, 1998). Recently there has been a resurgence of interest in the significance of chromosomal inversions in evolution and speciation. Noor et al (2001) studied the genetic basis of hybrid sterility between Drosophila pseudoobscura and D. persimilis. They concluded that inversions could create linkage groups that cause sterility between hybridising taxa which would favour reproductive isolation. It has also been proposed that the reduction in recombination caused by chromosome inversions could act synergistically with isolation genes and increase the plausibility of certain models of speciation (Rieseberg, 2001).

There are presently about 2000 known species of Drosophila (Powell, 1997). Yet, only a few cases of colour polymorphism have been examined in this genus. One of the best studied cases is D. polymorpha. da Cunha (1949) identified one locus and two alleles (with no dominance) that determined the abdomen colour pattern polymorphism in this species. Heed and Blake (1963) later found another allele, and Martinez and Cordeiro (1970) discovered that modifiers segregated independently of the major locus. In D. kikkawai, abdominal colour polymorphism is determined by one locus and two alleles (Gibert et al, 1999), whereas in D. lebanonensis, the colour of the thorax is controlled by one locus (two alleles) and modifiers (Pipkin, 1962).

The tripunctata group is the second largest Neotropical group of Drosophila in number of species, with 56 described species (Vilela, 1992). It is particularly abundant in forested areas of southern Brazil during the winter (Saavedra et al, 1995). The group name tripunctata is derived from the presence of three dark spots on the last tergites of the abdomen. However, not all species show this trait. In some species, there is only a dark band on the last tergite instead of three spots, while in others there is intraspecific variation (Frota-Pessoa, 1954). This is the case in D. mediopunctata, a species distributed from southern South America to Central America. The abdominal colour pattern in D. mediopunctata varies considerably, from no spots to three dark spots, on the fourth, fifth and sixth abdominal tergites. In addition to number, these spots also vary in size. Males tend to have more and bigger spots than females (Frota-Pessoa, 1954).

D. mediopunctata has six pairs of chromosomes: five acrocentrics and a dot that does not undergo polytenisation. The X, second and fourth chromosomes are polymorphic for inversions. The second chromosome is the most polymorphic and its inversions can be divided into two groups according to the region as follows: there are eight inversions in the distal region (DA, DP, DS, DV, etc) and nine in the proximal region (PA0, PB0, PC0, PC1, etc) (Ananina et al, 2002). There is intense linkage disequilibrium between distal and proximal inversions, for example, DA is associated with PA0, with a D′ value of 0.98; DP is associated with PC0 (D′=0.97); and DS with PC0 (D′=0.95) (Peixoto and Klaczko, 1991).

In this paper, we present a genetic (chromosomal) analysis of the variation of the number of abdominal spots in D. mediopunctata. We show that this colour polymorphism is genetically determined mainly by one chromosome (the second chromosome). Examination of the influence of inversions of this chromosome on the character reveals a nonrandom association between inversions PA0 and PC0 and the number of spots. To our knowledge, this is the first time a conspicuous morphological polymorphism has been found to be associated with chromosome inversions.

Materials and methods

The colour polymorphism

The number of spots was counted on flies that were at least 4 days old, by which time the colour intensity had stabilised. The number of spots varied, with flies possessing no spots, one (on the sixth tergite), two spots (on the fifth and sixth tergites) and three spots (on the fourth, fifth and sixth tergites) (Frota-Pessoa, 1954). In flies with three spots, the spot on the fourth tergite sometimes merges with the posterior pigmented band. In these cases, the phenotype was scored as ‘3D’, but the entire analysis was carried out by considering this pattern as having three spots.

Strains used in the genetic analysis

The following strains were used:

  • ITC-29I: A strain selected for zero spots and brother–sister crossed for over 20 generations.

  • NA: A marker strain (Carvalho and Klaczko, 1993) carrying the visible mutations Δ (Delta), Im (Impar), cr (coral) and al (alfinete) on the second, third, fourth and fifth chromosomes, respectively, and sharing its X and Y chromosomes with strain CR27A.

  • CR27A: A marker strain carrying the dominant visible mutations Δ-5 (Delta-5) on the second chromosome, and the recessive mutations cb (cabernet), cr (coral) and al (alfinete) on the third, fourth and fifth chromosomes, respectively. This strain shares its X and Y chromosomes with strain CR27B.

  • CR27B: A marker strain carrying the visible mutations cb (cabernet), cr (coral) and al (alfinete) on the third, fourth and fifth chromosomes, respectively, and sharing its X and Y chromosomes with strain CR27A.

In strain ITC-29I, the zero spot phenotype was fixed whereas in the remaining strains the three spots phenotype was fixed.

Marker strains used to produce strains with the same genetic background

The following strains were used:

  • CR26A: A marker strain carrying the recessive mutations mt (merlot), cr (coral) and al (alfinete) on the second, fourth and fifth chromosomes, respectively. This strain shares its X and Y chromosomes with strains CR26B, CR26C and CR26J. The second chromosome karyotype is DV-PC0/DV-PC0.

  • CR26B: A marker strain carrying the dominant visible mutation Δ (Delta) on the second chromosome, and the recessive mutations mt (merlot), cr (coral) and al (alfinete) on the second, fourth and fifth chromosomes. The second chromosome karyotype is DV-PC0/DV-PC0.

  • CR26C: A marker strain carrying the dominant visible mutation Im (Impar) on the third chromosome, and the recessive mutations mt (merlot), cr (coral) and al (alfinete) on the second, fourth and fifth chromosomes, respectively. The second chromosome karyotype is DV-PC0/DV-PC0.

  • CR26J: A marker strain carrying the dominant visible mutation Δ-5 (Delta-5) on the second chromosome, and the recessive mutations mt (merlot), cr (coral) and al (alfinete) on the second, fourth and fifth chromosomes, respectively. The second chromosome karyotype is DA-PA0/DV-PC0.

Strains CR26A, CR26B, CR26C and CR26J share their sex chromosomes.

Strains with different second chromosomes on the same genetic background

To test whether there was a nonrandom association between second chromosome inversions and the phenotypes of dark abdominal spots, we produced strains that differed from each other on the second chromosome but shared the same genetic background. We used isofemale lines obtained from two collections made in Serra do Japi, SP, Brazil (23°11′S, 46°40′W) in July 1994 and May 1995. The females were crossed with marker strains (CR26A, CR26B, CR26C and CR26J) and the second chromosome was then made autozygous by inbreeding (Figure 1).

Figure 1
figure 1

Crosses made to obtain strains with different second chromosomes on the same genetic background. All original chromosomes from the isofemale line were replaced, except for the second chromosome, which was made autozygous. The chromosomes from the isofemale line are in bold.

Full size image

Females from the isofemale line were initially crossed with males from strain CR26C (Figure 1). From the offspring, we selected males with phenotype Impar, which were heterozygous for the recessive markers mt, cr and al, and for the dominant marker Im. These males were individually crossed with females from strain CR26A and males with the phenotypes Impar, coral and alfinete were selected from the offspring. Thus, the fourth and fifth chromosomes from the isofemale line were replaced by the corresponding marked chromosomes. The selected males were crossed with females CR26A and the larvae from these crosses were karyotyped. If the karyotype was DA-PA0, males with the phenotypes Impar, coral and alfinete were crossed with females from strain CR26B. This strain carries mutations Δ and mt in a chromosome with a haplotype DV-PC0. Thus, in the progeny we are able to select double heterokaryotypes (DA-PA0/DV-PC0) to avoid recombination. If the karyotype was DP-PC0, DS-PC0 or DV-PC0, males with the same phenotype were crossed with females from strain CR26J. This strain carries mutations Δ-5 and mt marking the haplotype DA-PA0, and as, in the other cross, one may select double heterokaryotypes (eg DP-PC0/DA-PA0) avoiding recombination. In both cases, delta, alfinete and coral males and females of the offspring would form the next cross. The flies produced in this cross that were only coral and alfinete were used to start the new strain. If the second chromosome from the isofemale line was lethal, males and females delta were crossed and the strain was kept balanced against the Delta or Delta-5 mutation, which is also lethal.

Using this approach, we obtained strains that were differentiated in the second chromosome, but not in the other chromosomes, that is, they shared the same genetic background. Overall, 25 strains (14 PA0 and 13 PC0) were produced.

Genetic analysis

To carry out the genetic analysis of the colour polymorphism in D. mediopunctata, we used the same experimental strategy as Spicer (1991) when he analysed the differences in pigmentation between D. virilis and D. novamexicana. We first examined the effect of the sex chromosomes by reciprocal crosses, and then analysed the effects of the autosomes using visible mutants marking each autosome.

Detection of the effect of the sex chromosomes

Reciprocal crosses were carried out between strains ITC-29I and NA to detect the presence of genetic factors that could determine the colour polymorphism on the sex chromosomes. If the X or Y chromosome affected the character, then male offspring of one cross should be different from the males of the other. The female offspring served as a control and were expected to have the same distribution of phenotypes since they had the same autosomes and sex chromosomes in both reciprocal crosses. This would not happen if there was a maternal effect or a maternal inheritance.

Genetic (chromosomal) analysis

Females from strain ITC-29I (phenotype 0, for dark spots on the abdomen) were crossed with males from strain CR27A. This strain is fixed to phenotype 3 and has a recessive lethal mutation with a dominant visible effect (Delta-5, Δ-5), heterozygous on the second chromosome. The strain is homozygous for recessive mutations on the third, fourth and fifth chromosomes. Male offspring with phenotype Δ were backcrossed with strain CR27B. This strain carries the same mutations on the third, fourth and fifth chromosomes as CR27A, but is homozygous for the wild allele of Δ-5. The offspring were raised under controlled conditions of temperature (16.5°C) and density (20 larvae/vial). Since there is no crossingover in males of D. mediopunctata, it is possible to assess the influence of each of the marked chromosomes. If the distribution of phenotypes is independent of the genotype of the markers, then there is no influence of the chromosome; otherwise, the chromosome carries at least one factor that determines the character.

Strains carrying different inversions on the same background

First instar larvae were collected from each strain and 15 larvae per vial were kept at 20°C with culture medium (16% yeast, 2% sugar and 2.5% agar) until all adults had emerged.

The patterns of abdominal spots were analysed in only 10 strains for each karyotype, since the second chromosome of the remaining strains carried at least one lethal gene, which made it impossible to examine individuals autozygous for this chromosome. All strains carrying PA0 in the proximal region were DA in the distal region of the chromosome, whereas strains carrying PC0 carried DV, DS or DP in the distal region.

The statistical analysis was carried out using an ANOVA of the means of the number of spots per individual for each sex of each strain, to allow assessment of the effects of the second chromosome karyotype, sex and the interaction between these two factors. The advantage of using the means of each strain instead of the number of spots for each individual is that, as sample size increases, the means approach the normal distribution (Sokal and Rohlf, 1995).

Crosses between strains with the same genetic background

Since autozygosis for an entire chromosome is quite unlikely in nature, we planned crosses between strains with the same karyotype, so that the resulting offspring would simultaneously be homokaryotypic and allozygous (two alleles of independent origin) for the proximal inversions of the second chromosome. The crosses were chosen randomly among all possibilities, with the condition that each strain would participate in only two crosses. In this way, the number of crosses was equal to the number of strains available and we did not inflate the sample size.

Since it was not possible to make all of the crosses at once, they were made in sets of five. One of these crosses was repeated each time a new set of crosses was made, in order to confirm that there were no changes in the environmental conditions during the experiment.

Crosses were made with males from one strain and females from another strain. Reciprocal crosses were not carried out. Again, first instar larvae were collected from the offspring of each cross and placed in vials containing 10 ml of culture medium. These larvae were kept under four conditions: 12 or 96 larvae per vial, at 20°C, and 12 or 96 larvae per vial, at 16.5°C.

The effect of each of these variables (karyotype, temperature, density and sex) on the colour polymorphism was tested by ANOVA, in a manner analogous to the analysis of the strains.

Results

Genetic analysis

Effect of the sex chromosomes

There were no significant differences between the results of the reciprocal crosses between strains NA (pure 3) and ITC-29I (pure 0). Therefore, sex chromosomes had no significant effect (Table 1).

Table 1 Numbers of flies from each phenotype in reciprocal crosses
Full size table

Autosomes

Genetic analysis revealed that the second chromosome was the most important chromosome in determining the colour polymorphism. Figure 2 shows that the phenotypes changed considerably, depending on the origin of the second chromosome (compare the lower and upper halves of each graph in the figure). In addition, ANOVA (Table 2) revealed significant effects of sex and the fifth chromosome; the third chromosome exerted an effect that bordered on significance (P=0.088). ANOVA detected no significant effects for any of the interactions (not shown).

Figure 2
figure 2

Percentages of the patterns of colour polymorphism based on genetic analysis showing the effects of each chromosome on males and females separately. In each chart, the effect of the second chromosome becomes evident when the upper half (where the second chromosome is a heterozygote from a pure 0 strain and a pure 3) is compared with the lower half (where the second chromosome is a homozygote from a pure 3 strain).

Full size image
Table 2 ANOVA for the data used to assess the effects of each autosome and sex on the number of abdominal spots
Full size table

Nonrandom association between karyotype and colour pattern

Strains with different second chromosomes on the same genetic background

Figure 3 shows that strains in which the karyotype was PA0 usually had fewer spots than those in which the karyotype was PC0. The difference between the two karyotypes was significant, as was the difference between sexes. There was no significant interaction between these two factors (Table 3).

Figure 3
figure 3

Percentages of the patterns of colour polymorphism in strains with different second chromosomes on the same genetic background. In each pair of bars, the upper bar indicates the proportions of the phenotypes among the females and the lower bar refers to males. Each strain is identified by a letter on the left side of the corresponding bars.

Full size image
Table 3 Results of the ANOVA applied to the average number of spots per individual of each strain to assess the effects of karyotype, sex, strain (nested by karyotype) and interaction between genotype and sex
Full size table

Crosses

The karyotype of the second chromosome had a significant effect on the phenotype of the colour polymorphism among the offspring of crosses between strains with the same genetic background for this chromosome (Table 4). Figure 4 shows that crosses between strains carrying PA0 tended to produce offspring with fewer abdominal spots than crosses between strains carrying PC0.

Table 4 Results of the ANOVA applied to the average number of spots in each cross, to assess the effects of karyotype, density, temperature, sex, cross (nested by karyotype) and all possible interactions
Full size table
Figure 4
figure 4

Patterns of colour polymorphism for crosses with different second chromosomes on the same genetic background. The data are for flies raised in 20°C at density of 12 larvae per vial. In each pair of bars, the upper bar indicates the proportions of the phenotypes among the females and the lower bar refers to males. The letters on the left of the bars identify each cross, corresponding to the strains in Figure 3 (for example, N × E is the cross between strains N and E from Figure 3).

Full size image

Other effects

Temperature had a significant effect and was the source of most of the variation (Table 4), leading to flies with more spots at lower temperatures. The effect of sex was also significant so that males tended to have more spots than females (Tables 5 and 6). In addition, there were significant interactions between karyotype and temperature, and between sex and temperature, but we did not detect any significant effect of density (Table 4).

Table 5 Average number of spots per offspring in crosses between strains PA0, in relation to temperature and larval density
Full size table
Table 6 Average number of spots offspring for crosses between strains PC0, in relation to temperature and larval density
Full size table

Control cross

The phenotype of the cross used as the control was particularly favourable because it was intermediate at 20°C. This cross produced offspring in which neither of the extreme phenotypes (0 and 3) was the most common, thus allowing us to detect any displacement of the phenotype caused by environmental changes.

The use of a control cross ensured that we would detect any change in the environmental variables during the course of the experiment. This was indeed seen: significant differences between the offspring of control crosses conducted on different dates were observed (ANOVA, not shown). Thus, the crosses were reanalysed, using the logarithm of the ratios between the mean of each cross and the mean of the corresponding control cross, considering the date, sex, density and temperature. In this way, the means of the crosses were standardised to the control, in order to eliminate all environmental effects, including the influence of the factors mentioned above. This second analysis (not shown) reduced the effects of temperature and density, which became nonsignificant whereas the effect of sex remained. The interactions between karyotype and temperature, and sex and temperature also persisted and a new, marginally significant (P=0.049) interaction (between density and sex) appeared.

The effect of karyotype remained highly significant and became the factor with the highest influence on the character (F=48.52; P=0.001). Thus, the mathematical correction diminished the environmental effects and emphasised the difference between the two karyotypes.

Discussion

Genetic analysis

Chromosomal analysis is useful for identifying the sources of genetic variation in a character among chromosomes. This classic approach has been used in the genetic analysis of hybrid sterility between D. pseudoobscura and D. persimilis by Dobzhansky (1936). He concluded that the inheritance of the character was polygenic, with at least one locus on each chromosome. An almost identical genetic analysis to the one used here revealed the existence of several autosomal suppressors of sex ratio in D. mediopunctata, with at least one locus in each major autosome (Carvalho and Klaczko, 1993).

Several studies have investigated the genetics of differences in abdominal pigmentation between closely related species. Spicer (1991) – applying the same experimental design used here – showed that four autosomes and the sex chromosomes had significant effects on the abdominal colour differences between D. virilis and D. novamexicana, showing that at least five loci are involved. Hollocher et al (2000) studied two species of the D. dunni subgroup and analysed three areas of the abdomen individually. They showed that the species differences were due to both X chromosomes and autosomes, as well as maternal and paternal effects, depending on the abdominal area. Another genetic analysis between two closely related species – D. yakuba and D. santomea – revealed that at least three genes (each located on a different chromosome) in females and five genes in males affected the pigmentation differences between the two species (Llopart et al, 2002b). In addition, the X chromosome was responsible for nearly 90% of the pigmentation differences. The difference between D. americana and D. novamexicana is also determined by at least one locus in the X chromosome and a moderate number of genes located in the autosomes (Wittkopp et al, 2003).

Colour polymorphisms have also been analysed in other groups, such as the butterfly Bicyclus anynana that has been extensively studied (for a review see Brakefield, 2003). This butterfly presents seasonal variation with respect to wing colour pattern (Brakefield and Reitsma, 1991). A genetic analysis of this trait concluded that at least five genes are involved in its determination (Wijngaarden and Brakefield, 2000).

As shown here, the colour polymorphism in D. mediopunctata was determined by genes localised on the second and fifth chromosomes. The influence of the second chromosome was much higher than for any of the other chromosomes, suggesting the existence of a major locus or several minor effect loci on this chromosome.

Even though the X chromosome seems to play an important role in the determination of the abdominal pigmentation differences between the studied species of Drosophila, we did not detect its influence on the colour polymorphism in D. mediopunctata.

Caveat

It is important to bear in mind that the genetic analysis we carried out can only detect differences between the strains analysed. Thus, there is the possibility that loci (located in other chromosomes), which contribute to the determination of this character, were not detected because the strains used in our experiments carried the same alleles (or alleles for equivalent effects). To minimise this limitation, we used strains selected for the maximum difference between them.

Temperature, sex and density

The effect of temperature was very intense since it increased the number of spots at the lower temperature and was responsible for most of the variation in the results of crosses between strains with the same background. In D. melanogaster and D. simulans, Gibert et al (1996) also found a strong temperature effect on abdominal pigmentation, similar to that which we observed in D. mediopunctata. In D. melanogaster and D. simulans, the influence of temperature was similar on the trident pigmentation on the thorax (Capy et al, 1988). In D. simulans, the trident is evident only in flies raised at low temperatures. Although there is little phenotypic plasticity for pigmentation in D. kikkawai, a more intense pigmentation is seen in flies raised at lower temperatures (Gibert et al, 1999). Increased pigmentation as a result of lower developmental temperatures is often interpreted as an adaptation for thermoregulation (Capy et al, 1988; Gibert et al, 1996, 1999): darker flies absorb more radiation and heat faster, which is advantageous in a cold environment. However, according to Stevenson (1985), any radiant heat that is absorbed by an insect as small as a Drosophila will quickly be lost by convection. Based on this, Crill et al (1996) suggested a number of alternative explanations, including nonadaptive linkage with other genes that are adaptive at low temperature, strengthening of melanised cuticles that would help compensate for larger body size, increased protection against UV radiation and concealment from visual predators when the insect tries to raise its temperature by sitting on a dark object. Therefore, the association between developmental temperature and pigmentation in Drosophila needs further investigation.

The effect of sex was always significant with males tending to have more spots than females. These results are in agreement with those of Frota-Pessoa (1954), who observed that D. mediopunctata males usually had more spots, which also tended to be darker than in females. A significant interaction between sex and temperature was observed, suggesting that males and females have different reaction norms. Kopp et al (2000) showed that differences in the regulation of the bric-a-brac (bab) gene are responsible for the sexual dimorphism in the abdominal pigmentation of D. melanogaster. They also found that males discriminate against mutant females with a male pigmentation pattern – darker abdomen – but females do not show any preference between normal or mutant (ie with a female abdominal pigmentation) males, which led the authors to suggest that the less pigmented abdomen in females may be maintained by sexual selection. However Llopart et al (2002a) argued that this mechanism is unlikely to work under natural conditions (see also Stern, 2000).

Density was not an important factor determining the colour polymorphism. Freire-Maia (1964) suggested that the frequencies of the abdominal colouration phenotypes in D. kikkawai depend upon the conditions of the culture medium and larval density. This does not seem to be the case with D. mediopunctata.

Chromosome inversions

Since the second chromosome, which is the most polymorphic in D. mediopunctata (Peixoto and Klaczko, 1991; Ananina et al, 2002), showed greater influence on the character, we examined the correlations between inversions and the number of abdominal spots, that is, whether different second chromosome karyotypes had different effects on the colour polymorphism phenotype.

Dobzhansky (1970) believed that the suppression of recombination caused by chromosome inversions would be advantageous if the inverted region carried a supergene, such that inversions that kept coadapted gene complexes together would be favoured by natural selection. Evidence for polymorphism conditioned by supergenes has been found in land snails (Murray and Clarke, 1976; Jones et al, 1977), ladybird Adalia bipunctata (Marples et al, 1993) and butterflies (Mallet et al, 1990; Gordon and Smith, 1998).

Since chromosome inversions can behave exactly like supergenes, as predicted by Ford (1964), the existence of polymorphisms would be expected to be determined by different chromosome inversions. Several studies have demonstrated the influence of Drosophila chromosome inversions on morphological characters such as wing size and shape in D. mediopunctata (Bitner-Mathé et al, 1995), extra bristles in D. melanogaster (Izquierdo et al, 1991), size-related traits in D. buzzati (Bertrán et al, 1998) and more recently in traits showing clinal variation (Weeks et al, 2002; Calboli et al, 2003). Our results show that the colour polymorphism in D. mediopunctata is associated with chromosome inversions. Karyotype PA0 was associated with phenotypes with fewer abdominal spots, whereas PC0 was associated with phenotypes with more spots.

Owing to the strong linkage disequilibrium between proximal and distal inversions on the second chromosome, PA0 was mainly linked to DA, whereas PC0 was mainly linked to DP, DV and DS (Peixoto and Klaczko, 1991). Owing to this, the strains analysed in this work were DA-PA0, DV-PC0, DS-PC0 and DP-PC0. Hence, it is possible that genes determining the colour polymorphism are also located in the distal region of this chromosome, and that the conclusions drawn here can be extended to the inversions in the distal region.

There was a significant interaction between karyotype and temperature, suggesting that the gene complexes within each karyotype had different reaction norms related to temperature.

Our results agree with two other data sets already found for this species. First, inversion frequencies in D. mediopunctata change during the year so that PA0 tends to be more frequent in colder months than in warmer months, whereas PC0 undergoes the opposite changes (Klaczko, 1995). These data suggest that PA0 is more adapted to lower temperatures and PC0 to higher temperatures. Second, we have circumstantial evidence (unpublished data) that, in natural populations, light genotypes increase in frequency with low temperatures, while dark genotypes decrease – although, due to plasticity, the average number of spots increases, that is, flies are phenotypically darker. Since PA0 determines phenotypes with fewer spots and this karyotype increases in frequency in colder months, an increase in the frequency of genotypes leading to phenotypes with fewer spots is also expected in colder months. Thus, a comprehensive picture – even if speculative – emerges for the patterns of variation in the inversions and the colour polymorphism. Since PA0 seems to be adapted to lower temperatures, and PC0 seems to be favoured by higher temperatures, it is possible that these inversions have undergone the accumulation of genes adapted for each of these temperature conditions. So, if the genes that determine abdominal colour patterns are located in these inversions, natural selection would favour an association in which the linked alleles would be more adapted to the temperatures in which each karyotype is the more frequent. That is, our results go exactly in the direction of the predictions of Ford (1964) that one should find an association of factors or genes determining a polymorphism forming a supergene, and that these genes would be kept together through a chromosome inversion, which would be the most efficient way to do it. This interpretation, however, remains speculative until further research confirms that the genes controlling the polymorphism are really within the inversions and that more than one gene is actually involved, hence representing a supergene. In addition, it is also necessary to investigate the adaptive significance of this colour polymorphism and whether it is under some type of balancing selection, thus helping to maintain the inversion polymorphism.

Moreover, it is necessary to study in more detail the finding of increased frequency of light genotypes with low temperatures in nature. This situation is the opposite of what is found for other species of Drosophila, which is commonly explained as an adaptive character related to the thermoregulation (the thermal budget hypothesis). For example, latitudinal clines in which genetically darker flies are found in higher latitudes were observed in the abdominal pigmentation of D. melanogaster (Das et al, 1994) and for the trident colouration of both D. melanogaster and D. simulans (Capy et al, 1988). Apparently, the variation in abdominal pigmentation in D. mediopunctata in nature can not be explained by the thermal budget hypothesis. This would lend some support for the ideas of Crill et al (1996). It opens the possibility of a number of alternative explanations, including linkage disequilibrium, which needs further investigation to be adequately tested.

We are presently working to answer these questions. Nevertheless, this is the first time an association between a conspicuous morphological polymorphism and chromosome inversions has been described.