Desiccation Resistance in Interspecific Drosophila Crosses: Genetic Interactions and Trait Correlations
Miriam J. Hercus, Ary A. Hoffmann


We used crosses between two closely related Drosophila species, Drosophila serrata and D. birchii, to examine the genetic basis of desiccation resistance and correlations between resistance, physiological traits, and life-history traits. D. serrata is more resistant to desiccation than D. birchii, and this may help to explain the broader geographical range of the former species. A comparison of F2's from reciprocal crosses indicated higher resistance levels when F2's originated from D. birchii mothers compared to D. serrata mothers. However, backcrosses had a resistance level similar to that of the parental species, suggesting an interaction between X-linked effects in D. serrata that reduce resistance and autosomal effects that increase resistance. Reciprocal differences persisted in hybrid lines set up from the different reciprocal crosses and tested at later generations. Increased desiccation resistance was associated with an increased body size in two sets of hybrid lines and in half-sib groups set up from the F4's after crossing the two species, but size associations were inconsistent in the F2's. None of the crosses provided evidence for a positive association between desiccation resistance and glycogen levels, or evidence for a tradeoff between desiccation resistance and early fecundity. However, fecundity was positively correlated with body size at both the genetic and phenotypic levels. This study illustrates how interspecific crosses may provide information on genetic interactions between traits following adaptive divergence, as well as on the genetic basis of the traits.

IN Drosophila and other organisms, a great deal of work has been done examining interactions among stress responses and life-history traits. Such interactions can limit the potential for adaptation to changing environments. Studies have employed a range of methods including correlated responses to selection (e.g., Hoffmann and Parsons 1989; Roseet al. 1992), phenotypic manipulation experiments (Leroiet al. 1994a), genetic engineering (Krebs and Feder 1997), and combinations of several approaches (e.g., Chippindaleet al. 1996).

There has been a degree of emphasis on desiccation resistance. This trait responds rapidly to artificial selection in Drosophila melanogaster (Hoffmann and Parsons 1989) but more slowly in other species (Blows and Hoffmann 1993; Hoffmann and Parsons 1993a). There is some evidence that variation in desiccation resistance within species is associated with body size (Parsons 1970), but other data suggest that glycogen reserves may be important (Service 1987; Graveset al. 1992; Gibbset al. 1997). Desiccation resistance may also be related to the rate of water loss (Hoffmann and Parsons 1989; Gibbset al. 1997), which may be mediated by changes in the mass-specific metabolic rate (Hoffmann and Parsons 1989), although apparent changes in metabolic rate could also reflect changes in the amount of nonmetabolizing mass (Djawdanet al. 1997).

Intraspecific variation in desiccation resistance may be linked to early fecundity and altered longevity. Lines with an increased level of resistance show a reduced early fecundity but an increased longevity (Roseet al. 1992; Hoffmann and Parsons 1993b). In addition, selection for increased senescence in D. melanogaster is associated with higher desiccation and starvation resistance (Serviceet al. 1985), although this is not always the case (Forceet al. 1995).

If associations among traits underlie long-term evolution and potentially constrain it, their impact should be evident on a longer time scale, restricting the way populations and species diverge from each other. Authors have suggested that it is difficult to study constraints over such a time frame because of the importance of genetic background on evolutionary outcomes (Leroiet al. 1994b). The roles of backgrounds and random factors may be large in evolution (Cohan and Hoffmann 1989; Travisanoet al. 1995) and cannot be easily considered by comparing species for different traits. Species comparisons are also fraught with problems because species do not represent independent data points for testing associations. Nevertheless, it is not clear if genetic interactions among traits detected at the intraspecific level will constrain long-term evolution. Other ways of investigating evolutionary constraints over a longer evolutionary time span would therefore be useful.

One way of examining genetic divergence on a longer time scale is to consider closely related species and hybrids derived from crosses between them. Hybrids are normally only used to investigate the genetic basis of trait differences among species by rearing interspecific crosses to the F1, F2, or backcross stages (e.g., Price and Boake 1995). These crosses can provide information on factors such as dominance, epistasis, sex linkage, and the number of genes controlling species differences. It is also possible to use hybrids to examine interactions among traits once genes are segregating within hybrid populations. Hybrid lines have been used both at an early stage of hybridization and also after allowing them time for assortment and segregation of the parental genomes, to test if patterns of genetic interactions evident within a species exist at the interspecific level (Berrigan and Hoffmann 1998; Blows 1998).

Here we use hybrids to investigate the genetic basis of the difference in desiccation resistance between two (sibling) species, D. serrata and D. birchii. These species are endemic to Australia and Papua New Guinea (Ayala 1965). D. birchii has a very restricted distribution and is largely confined to tropical rainforests in northern Australia and New Guinea. The other species is more widespread and occurs in a range of habitats as far south as Sydney. The species can form hybrids (Ayala 1965) that are readily maintained as lines in the laboratory (Blows 1998). Hybrid flies have also been collected from the field (M. Blows, unpublished results; M.J.H., unpublished results). D. serrata is much more resistant to desiccation than D. birchii (Hoffmann 1991), which is consistent with the increased likelihood of D. serrata encountering desiccation stress in its much broader distribution.

We initially characterized desiccation resistance in a series of independent hybrid lines derived from reciprocal crosses between the species. Lines derived from the two reciprocal crosses differed for desiccation resistance in an unexpected manner. We investigated this difference further by generating F2 and backcross flies and by eventually setting up a second set of hybrid lines.

The two sets of hybrid lines were also used to investigate the physiological basis of genetic differences in resistance between the species. By examining several traits in the lines we tested if resistance was associated with glycogen levels and body size. We also used the hybrid lines to examine the association between desiccation resistance and fecundity, and the association between fecundity and size.

One limitation of this approach is that rapid evolution may take place in hybrid lines, selecting for particular genotypes and obscuring the effects of combinations of parental genes. This problem is likely to be exacerbated when hybrid lines have been cultured for many generations. For this reason, we tested associations among traits in hybrid lines maintained for 7 generations as well as lines maintained for longer (>16 generations). Moreover, we also examined trait associations in a different manner, involving a half-sib analysis undertaken 4 generations after hybridization. By mating each F4 male to several F4 females, we examined trait associations in half-sib families containing different combinations of parental genes.


Stocks: Two collections were used in this study. The first consisted of isofemale lines of D. serrata and D. birchii set up from females collected in April 1995 from the same sites in northeastern Queensland by Dr. M. Blows. These lines were combined to create mass-bred populations of the two species before 15 hybrid isofemale lines were generated for each reciprocal cross (see Berrigan and Hoffmann 1998; Blows 1998). The lines were used to look at desiccation resistance and other traits after being maintained for 17-20 generations to allow assortment of parental genes. Isofemale lines of each species were maintained alongside hybrid lines so that hybrid performance could be directly compared to that of the parental species.

The second set of lines was collected in January 1997 (by M.J.H.) from the same sites in northeastern Queensland as the first collection, and isofemale stocks of D. serrata and D. birchii were established from field females. Several hundred crosses were undertaken between isofemale lines of the two species in both directions. Five crosses in each direction between the species were successful and F1 progeny from these crosses were combined to produce two sets of hybrid F2's, one for each reciprocal cross. Backcrosses were also set up by combining an equivalent number of F1 males from the two reciprocal crosses and mating these to females of the isofemale lines that had been used in the successful interspecific crosses. Desiccation resistance and other traits were scored on the F2's, backcrosses, and parentals.

The F2's from each reciprocal cross were maintained as mass-bred populations of >1500 flies until the F4 stage. At this time, a half-sib analysis was undertaken and a set of hybrid lines was established. We set up 25 hybrid lines as isofemale lines from each mass-bred hybrid population. Desiccation resistance and other traits were scored on these lines after they had been independently maintained at a census size of >300 flies for another three generations. For the half-sib experiment, only flies from the D. birchii ♀ × D. serrata ♂ cross were used. Sixty males were each mated to five females and these female/male combinations produced the half-sib families, which were scored for a number of traits.

All laboratory stocks were maintained at 25° under continuous light, on an agar-dead yeast-sugar medium in two standard culture bottles per line at a population size of >150 individuals per bottle. Flies were tipped every 4 days and generation time was 3-4 wk. Reciprocal hybrid lines are designated as either B♀S♂ (from female D. birchii × male D. serrata) or S♀B♂ (from female D. serrata × male D. birchii). Backcrosses are designated as B1 (D. birchii female × hybrid male) or B2 (D. serrata female × hybrid male).

Traits and measurements: Individuals used in experiments were reared at controlled density (50 pairs of adults) and food conditions. This allowed a comparison of the groups over time and collections.

Desiccation resistance: This trait was measured only on females at 8% RH at 25°. Flies (0-24 hr posteclosion) were transferred to bottles without yeast and aged at 25° for 4 days. Sexing was performed using CO2 and flies were left to age for a further 3 days before the experiment was performed. Forty females from each of the two F2's, backcrosses, and parentals were assessed for individual knockdown time. Flies were transferred without CO2 into empty vials (75 × 10 mm) and covered with gauze. Empty vials were held at 100% humidity before being placed onto a rack, which was placed into a desiccator containing silica gel. Individual knockdown time was recorded every half-hour, beginning 6 hr after vials had been set up. For the half-sib analysis, desiccation resistance was also scored on individuals tested in the same way, and 5 females were tested for each family.

For the two hybrid-line comparisons, desiccation resistance was measured on five replicates of 10 females per line. Flies were tipped into empty 30-ml glass vials (100 × 25 mm), covered with gauze and placed into desiccators containing a layer of silica (indicator) gel. The number of flies dead per vial was scored every hour. The time taken for the first five flies in each vial to die was recorded, giving an LT50 score for each vial.

Fecundity: After eclosion and before mating, flies were transferred to vials without live yeast and males and females were aged separately at 25° for 4 days. Males and females were paired after this time in 30-ml vials with yeast and tipped to fresh vials for 2 days before testing fecundity. The pairs were placed into empty glass 30-ml vials containing a plastic spoon filled with 2 ml of treacle medium. The medium was dyed green with food color to count eggs clearly. A live yeast paste (10%) was spread over the surface of the medium. The spoons were replaced every 24 hr, for 7 days. Fecundity was assessed by scoring the number of eggs laid daily, to give total egg production over 7 days. Fecundity was scored in this way on 40 females for each F2, backcross, and parental line. For the half-sib analysis, fecundity was assessed for 2 females per family in the same way.

Fecundity was also measured for the hybrid-line comparisons. For each line, 70 females and 70 males were placed into two empty bottles and eggs were collected from watch glasses with 2 ml of treacle medium (as above) inverted over the bottles for 12 hr. Eggs were transferred to 30-ml vials with 13 ml medium; 20 eggs were placed into two replicate vials per line. Fecundity was scored on 10 females per line, taken randomly from the vials.

Body size: Body size was measured as wing length. Wings were removed with fine forceps and mounted on a glass slide using double-sided tape. A glass coverslip (22 × 22 mm) was placed over them. The length of the third longitudinal vein from its intersection with the anterior cross vein, to the wing tip was measured (×40) using an image analysis program (VideoTrace). Wing length is highly correlated with other measures of body size including dry weight and thorax length (Robertson 1957).

Wing-length measurements were made on the individual female flies assessed for desiccation or fecundity from the F2 generation, enabling phenotypic correlations with size to be computed. For the half-sib analysis, wings were measured on 7 of the females per family used for measuring fecundity and desiccation resistance. This enabled phenotypic as well as genetic correlations with size to be evaluated. For the hybrid line comparisons, wing length was measured on 10 females per line.

Glycogen content: This was assessed following Clarke and Keith (1988) and Blows and Hoffmann (1993). For the half-sib analysis, three replicates of 10 females per line were collected into 30-ml vials with 10 ml medium and aged at 25° for 7 days. Wet weight measurements were made after this time, the flies were frozen at –80° until the glycogen assay was performed. Flies were ground by hand in 1.2 ml homogenizing buffer (0.01 m KH2PO4, 1 mm EDTA, pH 7.4). A total of 25 μl of homogenate was added to 5 ml of test reagent [Sigma (St. Louis) preparation PGO enzyme (Cat. No. 51°0-A) with 0.1 units/ml amyloglucosidase added] and this was then incubated at 25° for 45 min. Optical density was read at 450 nm using a Bio-Rad (Richmond, CA) microplate reader (model 450) and the concentration of glycogen was determined by a glycogen standard. The amount of glycogen was expressed as a proportion of wet weight.

Glycogen content was also measured for both the hybrid line comparisons. Three replicates of 40 females were assessed per line.

Data analysis: Data from the F2's, parental populations, and backcrosses were compared with analyses of variance (ANOVAs). For all traits, data were normally distributed and there was no correlation between the means and variances. Post hoc analysis (Tukey B) was used to determine which of these groups differed from each other. For the F2's, Pearson's correlations were computed to look for associations between wing length and desiccation, and between wing length and fecundity.

The F4 data comprised a standard half-sib design, albeit for a hybrid population rather than a mass-bred population of conspecifics. Data were first checked for normality and a lack of association between family means and variances. They were analyzed using nested ANOVAs, the female term being nested within the male term. Only data for progeny from males mated with at least three females were used. Significance of the male term provides an indication of additive genetic effects separating the species, while other genetic effects and common environmental effects contribute to the female term. Phenotypic correlations were computed where more than one measurement was available from the same fly. This was possible only in the case of correlations involving wing length. Other trait associations were examined by obtaining trait means for each family and averaging them to provide a mean score for each male parent. These values were then used to examine genetic correlations between traits. We focused only on a restricted set of correlations to test specific hypotheses based on intraspecific patterns. These were correlations between size and both desiccation resistance and fecundity, as well as the correlations between desiccation and both fecundity and glycogen.

Hybrid F7 data were analyzed by ANOVA, with the replicate lines nested within the reciprocal cross group (B♀S♂ and S♀B♂) from which they had been derived. Parental data were not available for these lines. However, for the other set of hybrid lines tested at the F17-F20 stage, parental data were available because isofemale lines of D. serrata and D. birchii had been maintained the same way as the hybrid lines. We could therefore test for differences between the species as well as differences between hybrid lines derived from the two reciprocal crosses. For both sets of hybrid lines we used mean values of the traits to test for associations among them.

Correlations performed using line means may confound error covariance within the lines with the true among-line covariance (Via 1984). This can lead to biased estimates and statistical tests that are too liberal. An adjustment for error covariance was therefore made in cases where both traits had been measured on the same individuals. This adjustment involved subtracting the error covariance from the among-line covariance. Because the adjustment led to coefficients that did not differ from those based on means, only the latter are presented.


We first examine the genetic basis of the different traits by considering data collected at the F2 stage and data from the two sets of hybrid lines. This is followed by results from the half-sib analysis and, finally, the correlations among traits from all four approaches.

View this table:

ANOVAs (mean squares) comparing F2's and hybrid lines derived from the reciprocal crosses

Desiccation resistance: F2 stage: There was a significant difference between the parental species, reciprocal F2's, and two backcrosses (F(5,227) = 13.91, P < 0.001). Differences between reciprocal F2's are also significant (Table 1). Means and standard errors indicate two groups (Figure 1) confirmed by post hoc tests (Tukey B). The first of these consists of D. birchii, the backcross to D. birchii females, and the S♀B♂ F2's. The second more resistant group consists of D. serrata, the backcross to D. serrata females, and the B♀S♂ F2's. The difference between D. serrata and D. birchii is expected on the basis of previous data (Hoffmann 1991). However, the difference between reciprocal hybrids was surprising. If desiccation resistance had been influenced by a maternal effect, the S♀B♂ F2's should have been more resistant than F2's from the reciprocal cross. On the other hand, a negative maternal effect as suggested by the F2 data was not evident in the backcross generation. The backcrosses were set up with an equal number of males from the S♀B♂ and B♀S♂ F1's crossed to parental females, so an overriding negative maternal effect should have resulted in resistance levels in the backcrosses lower than those of the parental species.

F7 hybrid lines: The ANOVA showed a significant difference between lines derived from the reciprocal crosses (Table 1) in a direction consistent with the F2 results (Figure 1), in that lines derived from the S♀B♂ cross had relatively lower levels of resistance. In addition, the ANOVA indicates significant differences among the hybrid lines within the reciprocal cross groups.

F17-20 hybrid lines: Resistance levels were higher than in the F7 line comparison. This may have reflected minor changes in the temperature at which desiccation was undertaken. Females from D. serrata isofemale lines were more resistant to desiccation than those from D. birchii lines, as expected (Figure 1), and this difference was significant in a nested ANOVA (F(1,6) = 25.70, P < 0.01). The hybrid line results are similar to the F7 line results, with a significant effect of reciprocal cross origin and a significant difference among lines (Table 1). Overall, hybrid line means were more similar to D. serrata than D. birchii although there was some overlap with both species.

Fecundity: F2 stage: There was no difference between the reciprocal hybrids for this trait (Table 1) although fecundity did differ between the parental, F2, and backcross classes (F(5,234) = 3.75, P < 0.01). This was due to the higher fecundity of the F2's and backcrosses compared to the parental species (Figure 2). Post hoc tests indicated significant differences between two groups, one consisting of the two parentals and the other consisting of the S♀B♂ F2's and the B2 backcross. The remaining crosses did not differ significantly from either group. The higher fecundity in the crosses compared to parentals suggests heterosis.

F7 hybrid lines: There was no significant difference between the reciprocal groups, but there were differences among the hybrid lines (Table 1). Fecundity scores were higher than in the F2 stage comparisons, but it is not clear if this is due to laboratory adaptation or differences in environmental conditions.

F17-20 hybrid lines: Results were similar to those for the F7 hybrid lines, with significant differences among the lines but not between the reciprocal groups (Table 1). The D. serrata isofemale lines tended to lay more eggs than the D. birchii lines (Figure 2), and this difference was significant by nested ANOVA (F(1,6) = 7.43, P < 0.05). The hybrid lines tended to have a higher fecundity than D. birchii females but did not exceed that of the D. serrata lines, indicating that heterosis effects were no longer evident.

Figure 1.

—Mean desiccation resistance measured as individual knockdown time in F2, parental, and backcross populations. Error bars represent standard error values. The dashed line indicates where the mean F2 hybrids and backcross values should fall if the additive-dominance model explains the variance observed. Also shown are desiccation resistance measures of mean LT50 for F7 and F17-20 isofemale lines.

Wing length: F2 stage: Differences among parental, F2, and backcross groups were significant (F(5,225) = 16.84, P < 0.001). D. serrata wings were shorter than those of D. birchii (Figure 3) and this difference was significant (F(1,72) = 22.17, P < 0.001). F2's from the reciprocal crosses differed significantly (Table 1). Wings of F2's from the S♀B♂ cross were longer than those of the reciprocal cross. Post hoc tests indicate three groups. The first consists of D. birchii and the B1 cross, the second consists of the S♀B♂ F2's and the B2 backcross, and the third consists of the B♀S♂ F2's. D. serrata did not differ significantly from either of the last two groups. As for desiccation resistance, hybrids from the reciprocals therefore behaved in an unexpected manner.

Figure 2.

—Mean fecundity measured as total number of eggs laid over 7 days for F2, parental, and backcross populations. Error bars represent standard error values. The dashed line indicates where the F2 and backcross mean values should fall if the additive-dominance model explains the variance observed. Also shown are mean fecundity measurements for F7 and F17-20 isofemale lines.

Wing lengths were independently measured in two other groups of flies from each cross at this stage, mainly because of the surprising result that D. birchii flies were larger than those from D. serrata (we had previously found the opposite trend in other stocks). These additional comparisons (data not shown) confirmed that D. birchii was larger than D. serrata. Moreover, they also confirmed the size difference between the reciprocal crosses.

Figure 3.

—Mean wing length measured on F2, parental, and backcross populations. Error bars represent standard error values. The dashed line indicates where the F2 and backcross mean values should fall if the additive-dominance model explains the variance observed. Also shown are mean wing lengths for F7 and F17-20 isofemale lines.

F7 hybrid lines: Reciprocal hybrids differed significantly, and a line-within-group effect was evident (Table 1). However, unlike in the F2's, females from lines derived from the S♀B♂ cross were smaller than those derived from the reciprocal cross (Figure 3). Unfortunately, we were unable to determine if species rankings as well as hybrid-line rankings had changed because lines of parental species maintained in an identical fashion were not available. Nevertheless, we did find that females from a mass-bred line of D. serrata were larger than those from a mass-bred D. birchii line maintained in the laboratory for around the same time as the F7 hybrid lines.

F17-20 lines: D. serrata lines tended to be larger than the D. birchii lines (Figure 3) and this difference was significant by a nested ANOVA (F(1,6) = 33.45, P < 0.001). In agreement with the F7 line results, there was a significant difference between reciprocals (Table 1) and S♀B♂ were smaller than those from the reciprocal cross (Figure 3). The S♀B♂ hybrids were therefore more similar to D. birchii than to D. serrata as in the case of desiccation resistance.

Figure 4.

—Glycogen content expressed as mean proportion wet weight for F7 and F17-20 isofemale lines.

Glycogen content: F7 stage: The glycogen content of the females was expressed as a proportion of wet weight and arcsin transformed before analysis. The mean glycogen content of the lines ranged between 3.5 and 5.5% of wet weight (Figure 4), which is similar to values given for this species in Blows and Hoffmann (1993). Reciprocal hybrid lines did not differ significantly from one another and there were no differences among the lines (Table 1).

F17-20 stage: Mean glycogen content of the lines was somewhat higher than in the F7 comparison and ranged between 4.5 and 7% of wet weight (Figure 4). There was no significant difference between reciprocals, but line differences were evident (Table 2).

Half-sib analysis: ANOVA results are shown in Table 2. For fecundity and glycogen, there were no differences among the families. For desiccation resistance, nested female effects were significant, while both male and female effects were evident for the wing-length data. These findings suggest genetic differences among families for wing length and desiccation resistance but not the other traits. For size, additive effects appear important as suggested by the significant male component. We also computed narrow and broad sense heritabilities where possible following Roff (1997). These need to be interpreted cautiously because they are heritabilities based on a hybrid population. The narrow sense heritability was 11% for desiccation, 12% for fecundity, and 43% for wing length, while the broad sense heritabilities for these traits were 60, 12, and 43%, respectively. In obtaining these estimates, we have assigned a value of zero to VD when this component was negative.

View this table:

Results of the half-sib analysis undertaken at the F4 stage

Correlations between trait values: Desiccation resistance and wing length: The phenotypic correlation at the F2 stage was not significant (Table 3), suggesting that these traits were independent. However, the phenotypic correlation between these traits for the half-sib data was highly significant and positive. An indication of the genetic correlation was obtained from the half-sibs and the hybrid-line comparisons. All three estimates were positive, and two of them were significant. Genes increasing desiccation resistance therefore appear to be positively correlated with wing length. For the hybridline data, positive correlations (not shown) were also evident when lines derived from the two reciprocals were considered separately.

View this table:

Correlations between traits in the F2's, half sibs, and two sets of hybrid line comparisons

Fecundity and wing length: The phenotypic correlation between the fecundity and size measurements of F2 individuals was highly significant and positive (Table 3), suggesting that large females had a higher fecundity. The phenotypic and genetic (sire) correlations from the half-sib analysis were also positive and significant, as were both the hybrid-line correlations. A size-fecundity relationship was therefore evident at the genetic and phenotypic levels.

Desiccation resistance and glycogen content: One of the correlations (for the F7 lines) was significant and negative (Table 3), but the other correlations were positive and not significant. Hence there is no evidence that glycogen content is positively associated with desiccation resistance in these lines.

Desiccation resistance and fecundity: There was a significant positive association between desiccation resistance and fecundity in the F7 hybrid lines (Table 3). However, this was not evident in the other two comparisons. Overall, there is no evidence that increased desiccation resistance is associated with a decrease in fecundity.

Because both fecundity and desiccation resistance were positively correlated with wing length, we also investigated the association between these variables in a partial correlation analysis once wing-length effects had been excluded. This had little effect on the correlations between resistance and fecundity for the half-sibs and the F17-F20 hybrids, while the positive correlation evident at the F7 stage was reduced to 0.20 (P = 0.12).


Genetic basis of desiccation resistance: The difference in resistance between D. serrata and D. birchii appears to have a complex genetic basis. All the comparisons we made indicate that there is an unexpected difference between the reciprocal F2's and between the hybrid lines set up from the reciprocal crosses. Hybrids derived from crosses between D. serrata females and D. birchii males have a lower level of resistance than those from the reciprocal cross. This difference persists even when hybrid lines have been cultured separately for several generations. It is therefore not a maternal effect carried over for one or two generations.

How can this reciprocal difference be explained? One possibility is a cytoplasmic factor in D. serrata decreasing desiccation resistance. Another possibility is that there are genes on the D. serrata X chromosome decreasing resistance. Both effects would have to be masked by interactions with D. serrata autosomal genes to account for the (relatively) high level of resistance in the parental species. Nevertheless one problem with an X-linked model is that the F2's from the B♀S♂ cross were so similar to D. serrata (and vice versa for the reciprocal cross and D. birchii). The reciprocal hybrid F2's might have been expected to converge to some extent because some of the F2's from the reciprocals would have had the same X-chromosome constitution.

Either the X-linked model or the cytoplasmic model could at least partly explain the backcross results. In our backcrosses, F1 males from each reciprocal cross were mated in equal numbers to females from each parental species. This led to backcrosses with resistance levels similar to the respective species to which they were crossed. The high level of resistance in the D. serrata backcross could be explained if there were recessive or codominant D. serrata genes that masked negative cytoplasmic effects or negative X-linked effects (and vice versa for D. birchii).

A difference between reciprocals was evident for wing length as well as desiccation resistance. However, here patterns are complicated by an apparent shift in the relative size of D. serrata and D. birchii. In the F2 comparison, D. birchii was larger than D. serrata, but this was not the case when the F17-F20 hybrid comparison (and probably the F7 hybrid comparison) was undertaken. This switch may reflect adaptation to the laboratory or genotype-environment interaction due to changes in environmental conditions. However, the latter appears unlikely because the same density of flies, rearing medium, and temperature were used in rearing the lines.

Assuming that a change in the relative wing length of the species had taken place by the F7 stage, the difference between the reciprocals then matches the pattern for desiccation resistance, with the S♀B♂ F2's and hybrids tending toward D. birchii and vice versa for the B♀S♂ F2's and hybrids. This suggests that genetic interactions influence interspecific variation in wing length as well as desiccation resistance. Moreover, the same pattern among the hybrid lines has been observed for heat resistance: D. serrata is more resistant than D. birchii, but hybrid lines derived from the B♀S♂ cross tend to be more resistant than those from the reciprocal cross (Berrigan and Hoffmann 1998). Nevertheless there are differences between these species that do not follow this pattern. For instance, we have found that D. serrata females mature sexually before D. birchii. When measured on the F17-20 hybrid lines, this trait showed no difference between the reciprocal hybrid lines, and the same pattern was evident in the F7 lines.

Turning to the remaining traits, fecundity showed heterotic effects at the F2 stage although this was not evident in the F17-20 lines. The absence of heterosis in the hybrids may reflect evolutionary changes in the hybrid lines and/or in the species lines.

How do these findings compare to those from other interspecific crosses among Drosophila species? F1 hybrids and backcross populations from interspecific Drosophila crosses tend to be intermediate between parental species for morphological traits (Val 1977) and acid resistance (Amlouet al. 1997). Although differences between reciprocal crosses can occur (Val 1977), we have not encountered examples where hybrids are closer to the male parental species rather than the female parental species in the reciprocal cross. The only suggestion for such an effect comes from Carson et al. (1994) who found that a hybrid population from a cross between D. silvestris females and D. heteroneura males evolved toward a head shape similar to D. heteroneura, whereas there was no change in the head shape of a line from the reciprocal cross. However, Val (1977) showed that directly after crossing, hybrids between these species tended to be similar to their maternal parent.

Heterosis has previously been reported for fitness traits in intraspecific crosses of D. melanogaster populations (Ehiobu and Goddard 1990). We are unaware of interspecific Drosophila crosses where fecundity or other fitness traits have been scored. However, heterosis has been detected for fitness traits in interspecific crosses of other animals (e.g., Wanget al. 1994; Basavarajuet al. 1995) and in plants (Li and Wu 1996).

Correlations among traits at the interspecific and intraspecific levels: In one case there was a good match between the correlations at the interspecific level and those at the intraspecific level. We found that wing length was positively correlated with fecundity when both line means and individual female measures were considered. This result was consistent for the F2's, halfsib families, and hybrid lines. It matches intraspecific data from numerous ectotherms, which indicate that body size and fecundity should be correlated (Roff 1992). An association between size and fecundity has also been reported in D. melanogaster (e.g., Robertson 1957). Interactions between these traits in Drosophila therefore extend beyond the intraspecific level.

In another case we found a correlation at the interspecific level that was not evident in intraspecific studies. Our data indicated that wing length was significantly positively correlated with desiccation resistance in all comparisons except those involving the F2's. We might expect these two traits to be correlated because bigger flies have a lower surface area:volume ratio, making them potentially more tolerant of stresses like desiccation. At the intraspecific level, however, Hoffmann and Parsons (1989, 1993a) found no correlated response in body size (wing length, wing width) in D. melanogaster and D. simulans lines selected for desiccation resistance, while Blows and Hoffmann (1993) also failed to find a correlated response in size (weight) following selection for desiccation resistance in D. serrata. The only exception is a study by Parsons (1970) on desiccation resistance in two populations of D. melanogaster; he found those isofemale lines with larger body sizes lost less water than smaller lines, resulting in a greater resistance to desiccation.

Despite the consistent size-resistance association in our comparisons, size variation does not necessarily account for a large proportion of the variance in desiccation resistance. Correlation coefficients tended to be in the low to intermediate range. Moreover, D. birchii was larger than D. serrata in the F2 comparisons, even though D. serrata had a higher desiccation resistance as in all other comparisons. Finally, F2's from the B♀S♂ cross were relatively more resistant to desiccation even though these were smaller than those of the reciprocal cross.

Finally, the physiological traits suggest a case where associations apparent at the intraspecific level do not extend to the species level. We found that glycogen content was not associated with desiccation resistance, in contrast to intraspecific data collected on D. melanogaster (Service 1987; Graveset al. 1992; Chippindaleet al. 1996). There was also no evidence that glycogen levels led to tradeoffs between early fecundity and desiccation resistance as argued by several workers (Service 1987; Serviceet al. 1988; Hoffmann and Parsons 1989; Graveset al. 1992) on the basis of D. melanogaster lines selected for postponed senescence or altered levels of stress resistance. Effects of glycogen levels on stress resistance may depend on the species being considered (Blows and Hoffmann 1993) or differ between the inter- and intraspecific levels.

Why are some correlations evident at the different levels whereas others are not? Three possibilities come to mind. First, there may be no genetic variation at the interspecific level for some of the traits. The sib analysis suggests that genetic variance for some traits was low, particularly fecundity and glycogen content. Second, interactions among some traits may be more flexible at the evolutionary level than others. Thus, although genetic correlations are evident at the intraspecific level, these do not determine long-term patterns of evolutionary change. This may apply to physiological traits, whereas interactions involving morphological traits may be more entrenched (Schluter 1996). Finally, because differences between species are generally larger than those within species, different patterns of correlations may become evident. For instance, the size differences between the species could have been large enough to influence desiccation resistance, whereas those within species may not have been, leading to different correlation patterns.

In summary, the genetic basis of differences in desiccation resistance between D. serrata and D. birchii appears complex and different from the genetic basis of intraspecific variation. In addition, patterns of interactions among traits may differ within and between species. It is possible that some associations evident at the intraspecific level are not found at higher levels, raising doubts as to whether the same factors limit evolutionary change at these levels. This raises the general issue of whether comparative studies on trait interactions using species (e.g., Huey and Kingsolver 1993) are expected to produce related patterns of correlations.


We thank M. Blows for generating one set of hybrid lines and T. Bjorksten and M. Robinson for technical assistance. D. Berrigan provided useful discussions at the early stage of this work. This work was supported by a grant from the Australian Research Council.


  • Communicating editor: L. Partridge

  • Received June 22, 1998.
  • Accepted January 4, 1999.


View Abstract