Clinal variation and laboratory adaptation in the rainforest species Drosophila birchii for stress resistance, wing size, wing shape and development time


Ary Hoffmann, Center for Environmental Stress and Adaptation Research, La Trobe University, Bundoora, Victoria 3086, Australia.
Tel.: 61 394 792769; fax: 61 394 792361;


Clinal variation has been described in many invertebrates including drosophilids but usually over broad geographical gradients. Here we describe clinal variation in the rainforest species Drosophila birchii from Queensland, Australia, and potential confounding effects of laboratory adaptation. Clinal variation was detected for starvation and development time, but not for size or resistance to temperature extremes. Starvation resistance was higher at southern locations. Wing shape components were not associated with latitude although they did differ among populations. Time in laboratory culture did not influence wing size or heat knockdown resistance, but increased starvation resistance and decreased recovery time following a cold shock. Laboratory culture also increased development time and altered wing shape. The results indicate that clinal patterns can be detected in Drosophila over a relatively narrow geographical area. Laboratory adaptation is unlikely to have confounded the detection of geographical patterns.


Quantitative traits often vary among Drosophila populations, and this variation has sometimes been related to adaptation. In particular body size clines have been demonstrated in several species on a number of continents and related to temperature selection (James et al., 1995; Huey et al., 2000; Gockel et al., 2001). Populations from cold areas tend to have a relatively larger body size than those from warm areas. These differences match evolutionary shifts in population cages maintained at different temperatures for many years; cold conditions lead to populations with a larger body size than warmer conditions (Cavicchi et al., 1985; Partridge et al., 1994). Differences among populations for physiological traits may also be adaptive. In particular in D. melanogaster, clines in heat and cold resistance match predictions based on average temperature (Hoffmann et al., 2002a). However, the adaptive significance of many geographical patterns remains elusive, including clinal variation in wing shape as opposed to wing size (Gilchrist et al., 2000; Hoffmann et al., 2002b), and geographical variation in resistance to desiccation and starvation resistance which shows clinal variation in some species (Karan & Parkash, 1998) but not others (Robinson et al., 2000; Hoffmann et al., 2001a).

Most studies of clinal variation have focused on widespread Drosophila species that tend to be distributed continuously. These are exposed to the widest range of climatic conditions, making them the best candidates for divergence generated by climatic selection. However allozyme and molecular data often indicate high levels of gene flow in Drosophila species (e.g. Gockel et al., 2001; Pfeiler & Markow, 2001) that act to counter adaptive divergence. There is much less information on clinal variation in Drosophila species with distributions restricted by habitat, that may show limited movement among favourable habitat patches, and higher levels of adaptive divergence even when they are exposed to weaker climatic gradients (Shoemaker & Jaenike, 1997).

Here we test for clinal variation in several traits in Drosophila birchii, which is restricted to rainforest habitats in northern Australia and New Guinea. In previous work we showed that desiccation resistance exhibited a weak clinal pattern over a relatively short interval of a few hundred kilometres, increasing from northern to southern (and drier) sites (Hoffmann et al., 2003). To test for clinal patterns in other traits, we consider resistance to starvation and heat/cold temperature extremes as well as wing size and wing shape. Differences among populations are compared with those detected in more widespread species from the same region, including the sibling species D. serrata.

When examining clinal patterns, researchers often use stocks that have been held in the laboratory for different lengths of time, and/or stocks collected at the same time but maintained in the laboratory before establishing clinal patterns. Under these circumstances, clinal patterns can be obscured by adaptative changes in response to laboratory culture. Stocks held in the laboratory can evolve along evolutionary trajectories that differ depending on the genetic background of the population (Cohan & Hoffmann, 1989; Matos et al., 2002). Laboratory adaptation can also confound other types of comparisons such as those involving artificially selected lines (Harshman & Hoffmann, 2000). For these reasons, it is important to examine the effects of laboratory adaptation when investigating adaptive divergence among lines.

There is good evidence for laboratory adaptation in D. melanogaster. Sgro & Partridge (2001) showed that lines set up from flies collected at the same location in different years varied for life history traits. Hoffmann et al. (2001b) subsequently used the same lines to demonstrate shifts in desiccation resistance due to laboratory culture. In. D. subobscura, evidence for laboratory adaptation has also been obtained from studies that have followed the same cultures for successive generations (Matos et al., 2000) although this design has also shown the absence of laboratory adaptation for heat resistance in D. melanogaster (Krebs et al., 2001). There is almost no information on laboratory adaptation in other species of Drosophila, particularly those with restricted distributions.

To examine laboratory adaptation in D. birchii, we modified the design of Sgro & Partridge (2001) by sampling flies repeatedly from several sites instead of one site, and by using isofemale lines instead of mass bred populations. In this design the stability of differences between geographical sites can be tested, and levels of variation within sites (among isofemale lines) can be compared with variation among sites. By examining the same traits as in the clinal comparison, we consider the potential of laboratory adaptation to influence geographical patterns, and we compare the speed of evolutionary changes in physiological traits vs. wing shape and wing size.



Drosophila birchii were collected along the eastern coast of Australia from the same four sites in March of 2000, 2001, and 2002. Between five and seven (usually six) isofemale lines were established with field females from each site and maintained in 300 mL bottles on potato medium at 19 °C under a 12 : 12 L : D cycle. The medium consisted of sugar (1.6% w/v), agar (3.2% w/v), dried yeast (3.2% w/v), potato (1.6% w/v), with 2% dihydrostreptomycin, 0.3% penicillin and 0.14% nipagin to counter microbial contamination. Once isofemale lines had been expanded over a couple of generations, each line was maintained at a census size of at least 300 individuals. In the 2002 collection, D. birchii were obtained from seven additional sites. Table 1 gives site locations and the number of isofemale lines established from each site, along with some climatic variables (9 am relative humidity, minimum/maximum temperatures of coldest/warmest months). Continent-wide surfaces for these variables were interpolated from weather station data (>30 years) with the program ANUCLIM (Houlder et al., 2000) using a 0.05° resolution digital elevation model (DEM) (Hutchinson & Dowling, 1991) and data for exact collection locations were then determined with ARCVIEW.

Table 1.  Collection sites for the lines used in the experiments, number of lines collected and climatic data based on 30-year averages.
LocationLatitudeNumber of isofemale lines collectedRelative humidity (%) Temperature (°C)*
200020012002Mean minimumMean maximum
  1. *Mean minimum of coldest month and mean maximum of hottest month.

Yeppoon22° 49′66771.912.329.5
Sarina21° 23′  673.411.230.2
Eungella21° 10′66673.28.029.6
Finch Hatton21° 04′66674.18.529.2
Proserpine20° 24′  673.812.031.3
Paluma18° 57′  673.511.029.6
Kirrama18° 11′  679.012.130.1
Lake Barrine17° 14′66679.111.128.7
Davies Creek17° 13′  680.010.827.7
Lake Placid16° 52′  676.914.330.3
Mossman16° 28′  575.614.830.2

Clinal variation was scored when the 2002 lines had been reared for five to seven generations in the laboratory. Flies for experiments were reared on potato medium; to control density of larvae, 300 adults were left in bottles for oviposition for 3 days and then removed. This ensured that larvae did not develop under crowded conditions (a maximum of 500 flies emerged from each bottle containing 300 mL of medium).

To examine laboratory adaptation, traits were scored on all lines from the four sites sampled repeatedly; at this stage, the 2000 lines had been in culture for 32 generations, the 2001 lines for 20 generations, and the 2002 lines for six to eight generations. Flies for experiments were obtained as described above. All experiments were carried out at 19 °C under a 12 : 12 L : D cycle.


Wing size and wing shape

To test for size variation among the lines, flies were collected from stock vials over a 2-day period and aged for 13 days (when reproduction in D. birchii reaches a peak). Flies were left for 12 h overnight to oviposit on laboratory medium containing treacle (9.8% w/v). Eggs were collected from this medium and transferred to 40 mL vials with the potato medium described above (10 eggs per vial, five vials per isofemale line). Vials were placed at 19 °C, 12 : 12 L : D. To score development time, eclosed flies were collected every 24 h and vials were scored until no new adults emerged within 48 h. Adult size was measured on a sample of emerged females (five per isofemale line, randomly selected from a sample pooled over vials) by taking the right wing from the flies and mounting this on a microscope slide between double sided tape and a coverslip. A digital photograph was taken of each wing using a Panasonic Super Dynamic II WV-CP460 video camera (Matsushita Communication Industrial Co. Ltd, Yokohama, Japan) attached to a Wild microscope. Images were landmarked with tpsDig ver. 1.2 written by F.J. Rohlf. Ten wing landmarks were placed mostly at vein intersections or terminations (Fig. 1) as outlined for D. serrata in Hoffmann & Shirriffs (2002). Wings from each experiment were landmarked in a randomized order. Wing size was scored as centroid size (the square root of the squared distance between each landmark and the wing centroid) and wing shape was analysed by the Procrustes technique as described below.

Figure 1.

Landmark positions on wing for Procrustes analysis and determination of wing centroid size.

Resistance to temperature extremes

Eclosing flies were collected over a 3-day period, aged for 4 days, sexed under CO2, and females aged for a further 5 days to reduce any adverse CO2 effects. A chill coma recovery assay (Gibert et al., 2001) was used to assess cold resistance because this measure has previously been shown to exhibit clinal variation in D. melanogaster (Hoffmann et al., 2002a) and D. serrata (Hallas et al., 2002). Nine females from each line (tested in three blocks of three females) were stressed. Females were placed individually in empty 40 mL vials at 1 °C for 2 h in a bath of ethylene glycol. Flies were left to recuperate at 19 °C and the length of their recovery time was recorded. Flies were monitored continuously and recovery time scored as the time taken for individuals to stand up.

Heat resistance was tested using a knockdown assay where flies were scored for the time taken to become incapacitated following exposure to a heat stress. This trait exhibits clinal variation in D. melanogaster (Hoffmann et al., 2002a). Nine females from each line were individually stressed (in three blocks of three flies) in 2 mL vials at 38.5 °C in a water bath. Flies were monitored continuously and the time taken for the female to stop moving any body part was recorded.

Resistance to desiccation and starvation

Eclosing flies were collected over 1 day and aged for 4 days before flies were sexed under CO2 and females were aged for a further 5 days to minimize any adverse CO2 effects.

Desiccation was tested following Hoffmann et al. (2001a). For each line, three replicates of 10 females were stressed in 20 mL vials. The flies were scored for survival every hour until at least half the flies had died. The LT50 of the females was then determined by linear interpolation.

To score starvation resistance, groups of 10 females were held in 40 mL vials, which were then inverted over a second vial containing cotton wool and water following Starmer et al. (1977). Flies in the vial were separated from the water with fine gauze and the two vials were sealed together with Parafilm®. The flies were scored for survival every hour until half the flies had died and the LT50 point was determined by linear interpolation.


Prior to analysing the cold and heat resistance data, individual values were corrected for any differences among blocks. For clinal variation, we initially examined differences among the locations for the traits using anovas, with the isofemale line term nested within the location term. By computing variance components for location and line, an indication of the relative magnitude of differences among locations vs. genetic variation within populations was obtained (Hoffmann et al., 2001a), although the differences among lines within populations include nonadditive genetic components as well as common environment and maternal components.

We also undertook regressions of latitude against isofemale line means to test for clinal patterns. As each isofemale line was established from a different field female, line means represent independent points for particular latitudes. Although clinal patterns in Drosophila can be nonlinear (James et al., 1995; Magiafoglou et al., 2002), only linear regressions are presented as the addition of nonlinear components did not significantly improve the fit of the regression equation for any of the variables.

To assess clinal changes in wing shape, a Procrustes analysis was first undertaken on the wing landmarks. This approach has previously been used to investigate clinal variation in Drosophila (Gilchrist et al., 2000; Hoffmann & Shirriffs, 2002). Procrustes residuals contain all the shape information from wing landmarks after adjusting for wing size. Procrustes analysis was undertaken with a generalized least squares approach using the IMP set of programs written by H.D. Sheets (Canisius College: The Procrustes residuals were used to test for the effects of location and isofemale line by undertaking separate anovas on landmark coordinates and then adding the sum of squares as outlined in Klingenberg & McIntyre (1998). The degrees of freedom for these summed values are obtained by multiplying the degrees of freedom in the standard anova by twice the number of landmarks minus four. The significance of F ratios was tested by conventional parametric tests. To determine the relative importance of the different landmarks in the changes in shape, we undertook anovas on landmark coordinates, and summed mean squares of the x- and y-analyses of each landmark (Klingenberg & McIntyre, 1998; Badyaev & Foresman, 2000). Variance components for each factor in the analysis were then computed and expressed as a percentage. Note that these components are intended to give a general sense of the strength of the effects of different landmarks; they cannot be analysed by themselves because Procrustes landmarks are not independent of each other.

As well as examining the effects of location and line on shape, we also tested for clinal changes in shape using a multivariate approach as outlined in Monteiro (1999). The TpsReg program written by F.J. Rohlf was used for this analysis. Mean Procrustes residuals of the isofemale strains were computed and this was followed by a multivariate regression of shape on latitude. This analysis involves computation of functions that describe the way the shape of an object is changed in tangent space: these consist of global or uniform changes in shape as well as localized changes (partial warps). Latitude was considered as a linear variable in this analysis. The significance of the multivariate regression model involving uniform and local changes, as well as a model involving only localized changes, was determined from an F ratio test as outlined in Monteiro (1999).

For the laboratory adaptation comparisons, we undertook nested anovas with three factors (year, location and isofemale line nested within year and location) to assess changes in stress resistance variables and wing shape. Differences between means were further examined with Tukey B post hoc tests. For the wing landmarks, we obtained Procrustes residuals as described above and undertook Procrustes anovas to examine the relative importance of the different factors. We also examined the contribution of the landmarks to the different factors by computing variance components for landmarks separately as described above.


Clinal variation

For the stress traits, anovas showed significant differences among locations for resistance to starvation, but not for resistance to the temperature extremes (Table 2). Locations also differed for desiccation resistance as tested previously (Hoffmann et al., 2003) and the anova on this trait is included in Table 2 for comparison. There were significant differences among isofemale lines for all traits except for heat resistance. Variance components indicated that the location effect was relatively larger for starvation resistance than for desiccation resistance. Line differences were relatively larger for resistance to starvation and desiccation than for the other traits. Regression analyses using isofemale line means indicated significant associations between latitude and starvation resistance, but not the other two traits (Table 3). Desiccation resistance was also associated with latitude (Table 3: Hoffmann et al., 2003). The latitude association remained significant when computed from location means (analysis not shown). Starvation resistance increased at higher latitudes (Fig. 2), as did desiccation resistance (Hoffmann et al., 2003).

Table 2. anovas testing differences among locations and nested isofemale lines in the clinal experiment. Desiccation resistance was tested previously (Hoffmann et al., 2003).
TraitTermd.f.Mean squareF% Variance
  1. *P < 0.01, **P < 0.001, ***P < 0.0001.

Heat knockdownLocation1121.5271.4360.7
Isofemale line5314.9870.8250.0
Error51818.162 99.3
Cold recoveryLocation1186.0940.6120.0
Isofemale line43148.7071.511*3.3
Error51998.415 96.7
Desiccation resistanceLocation1118.1592.198*8.2
Isofemale line548.2612.928***44.7
Error1302.822 47.0
Starvation resistanceLocation111077.5622.896**13.5
Isofemale line52379.7332.506***30.4
Error118151.517 56.1
Development (males)Location10953.1632.849**11.1
Isofemale line54338.1311.951***14.8
Error240173.341 74.1
Development (females)Location101290.5904.766***16.3
Isofemale line54272.0441.650**9.9
Error251164.866 73.8
Wing size (×10−2)Location103.3021.6623.6
Isofemale line551.9872.072***17.2
Error2620.959 79.2
Shape (×10−5)Location16011.9721.415*** 
Isofemale line8808.4642.320*** 
Table 3.  Linear regression analysis on the effects of latitude on isofemale line means for the traits. The clinal analysis for desiccation resistance was presented previously (Hoffmann et al., 2003).
Development time (females)1.6840.5120.1463.2900.002
Development time (males)0.9200.4070.0672.2660.024
Wing size−0.0020.0040.0020.3870.700
Figure 2.

Clinal patterns for starvation resistance and development time. These traits showed significant differences among locations. Location means and standard errors are based on isofemale line means.

Development time of both sexes differed significantly among locations, and differences among lines were also evident for these traits (Table 2). Variance components indicate that location differences were similar in magnitude to those for starvation and desiccation resistance. Regressions indicate significant associations between latitude and development time of both sexes (Table 3). Development times were relatively faster for lines collected from the more northerly locations, although lines collected from around 21°S also showed a rapid development time (Fig. 2). For wing size, there were differences among the lines but not locations (Table 2), and no latitudinal trend (Table 3).

Procrustes anovas indicated significant differences among locations and isofemale lines for shape (Table 2). A variance component analysis showed that the largest differences among locations were for landmarks 5, 6, 7, 8 and 10 (Table 4). In a discriminant analysis comparing the 11 locations, there was a significant difference among locations (Wilks’ lambda = 0.011, inline image = 231.51, P < 0.001) and there were four canonical variates each accounting for more than 10% of the variation (27, 20, 14 and 13% in decreasing importance). To examine clinal patterns, a multivariate regression of shape on latitude was undertaken. When partial warps and uniform components were considered, the regression of shape variables on latitude was not significant (F = 1.74, d.f. = 16, 49, P = 0.071). With the uniform component excluded, the regression was also not significant (F = 1.61, d.f. = 14, 51, P = 0.103). These analyses provide only suggestive evidence for an association between shape and latitude, although there were shape differences among the populations.

Table 4.  Variance components (%) for terms in the Procrustes anova when separated for the different landmarks in the clinal comparison and the laboratory adaptation experiment.
Clinal comparison
 Isofemale line71.960.866.361.736.840.430.737.750.447.5
Laboratory adaptation experiment
 Location by year1.
 Isofemale line32.325.831.426.323.620.427.527.325.529.7

Laboratory adaptation

anovas (Table 5) indicate no effect of time in laboratory culture on heat knockdown, male development time, or wing size, but a significant effect of this factor on cold recovery and particularly on starvation resistance. Means (Fig. 3) indicate that recovery time decreased under laboratory culture, reflecting an increase in cold resistance. Starvation resistance increased under laboratory culture. There were significant differences among locations for starvation resistance, reflecting higher resistance to this stress in the more southern locations compared with Lake Barrine for both traits (Fig. 3). These differences among the locations are consistent for the isofemale lines from all three collection years, reflecting trait stability despite changes in this trait under laboratory culture. They also match clinal patterns determined from the more extensive collections (Fig. 2). For female development time, there was an interaction between location and time in culture (Table 5). Means indicate that flies from the 2000 cultures had a slower development time than those from the other years, except in the case of Yeppoon where the recent 2002 lines had a relatively slower development time (Fig. 3). The relatively longer development time of the Yeppoon lines matches the geographical pattern in the more extensive clinal survey (Fig. 2).

Table 5.  Nested anovas testing the effects of year, location and isofemale line on traits in the laboratory adaptation experiment. For effects other than error, the first degree of freedom refers to all anovas apart from shape, the second refers to that for the Procrustes anova on shape. The anova on desiccation resistance was presented previously (Hoffmann et al., 2003).
TraitYear (d.f. = 2/32)Location (d.f. = 3/48)Year by location (d.f. = 6/96)Isofemale line (d.f. = 60/960)Error (d.f.)
  1. *P < 0.01, **P < 0.001, ***P < 0.00001.

Heat knockdown17.98111.7978.9738.8057.075 (571)
Cold recovery877.261*143.690215.600215.605*143.356 (576)
Desiccation resistance20.49276.501***11.54113.004***4.203 (141)
Starvation resistance4684.088***4390.882***417.238351.521***112.316 (145)
Development time (females)488.507128.427565.542*446.312***197.420 (273)
Development time (males)348.634117.944260.261342.185***176.990 (274)
Wing size (×10−2)2.0712.1954.8882.939***1.215 (279)
Shape (×10−5)18.159***16.681***11.018**7.829***2.556 (4464)
Figure 3.

Changes in starvation resistance, cold resistance and female development time when lines from four locations were cultured in the laboratory for different lengths of time. Collection year and location means and standard errors based on isofemale line means are shown.

We also examined changes in the variance among the isofemale line means because of culture time and population, to test for convergence in the phenotype exhibited by the isofemale lines over time. Kolmogorov–Smirnov tests (not presented) were initially undertaken to confirm that variances were normally distributed for the traits. For starvation resistance, there was a significant effect of time in laboratory culture (F = 52.91, d.f. = 2, 12, P < 0.001) and population (F = 7.18, d.f. = 3, 12, P < 0.05) on variances. As time in culture increased, the variance among the isofemale lines was reduced, while variances were consistently higher for Lake Barrine compared with the other populations (Fig. 4). The change in variances with laboratory culture was in the opposite direction to the change in means, suggesting convergence among the lines rather than an effect of shifts in trait scores on variances. For male development time, there was a significant difference among populations (F = 5.79, d.f. = 3, 12, P < 0.05) and a marginally nonsignificant effect of culture time (F = 4.04, d.f. = 2, 12, P = 0.077). Variances were higher in Lake Barrine and in the recently collected lines (Fig. 4), following the patterns for starvation resistance. There were no significant differences among populations or culture times for the other traits.

Figure 4.

Variances among strain means for strains from four locations cultured in the laboratory for different lengths of time. Data for starvation resistance and male development time are shown.

The Procrustes anova undertaken to examine changes in shape because of laboratory adaptation showed significant effects of time in laboratory culture, location, the interaction between these variables, and isofemale line (Table 5). The landmarks contributing most to these effects were identified from variance components. Landmarks 3, 4, 7 and 8 were associated with the location factor, whereas changes in culture were associated with landmarks 5, 6 and 7, and the interaction term was associated with landmarks 4 and 5. Plots of Procrustes means for the different years (Fig. 5a) suggest a shift in the position of landmarks 5 and 6 near the basal stalk, with these landmarks moving closer together during laboratory culture. This change occurred soon after flies were established in culture, as the same change was evident in the flies from 2001 and 2000. For the Procrustes plot with data averaged across the locations, there was a difference between Lake Barrine and the other locations for the relative width of the outer wing margin between landmarks 2/3 and 4 (Fig. 5b). The variance component because of isofemale lines was relatively consistent across all 10 landmarks, suggesting similar levels of genetic variation in shape in different parts of the wing.

Figure 5.

Differences in wing shape between groups of flies from the laboratory adaptation experiment. Mean x and y Procrustes residuals (measured in dimensionless units) of the different landmark positions are plotted. Wing orientation is the same as in Fig. 1. The position of two of the landmarks is indicated to facilitate comparison. (a) Comparison of flies held in the laboratory since 2000, 2001 or 2002. (b) Comparison of flies originating from different locations.


For heat resistance, both the absence of geographical variation and lack of variation among the isofemale lines suggest that this trait as measured by knockdown time has a low heritability in D. birchii. The trait shows clinal variation in D. melanogaster, but over a much larger geographical gradient (Hoffmann et al., 2002a). Heat resistance in D. melanogaster can be selected and exhibits heritable variation (McColl et al., 1996; Bubli et al., 1998; Gilchrist & Huey, 1999) and heat resistance can also be selected in other species (Sørensen et al., 1999). The absence of variation in this trait in D. birchii therefore contrasts with results from the more widespread species. Heat resistance in D. birchii may show, like desiccation resistance in this species (Hoffmann et al., 2003), a particularly low level of genetic variation, although the lack of variation at the geographical level may reflect similar levels of heat stress in the different populations.

In contrast, there was variation among isofemale lines, among geographical locations and/or among laboratory culture populations for resistance to starvation and cold. Variation in resistance to these stresses could reflect energy reserves in the flies. In D. melanogaster, lipids have been related to both starvation (Chippindale et al., 1996; Harshman et al., 1999) and cold resistance (Ohtsu et al., 1998). The increase in resistance to these stresses under laboratory culture could reflect an increase in fat or other reserves. When flies are confined to vials during laboratory culture with minimal activity, there is the potential for fat and glycogen reserves to accumulate and for flies to become lethargic (Harshman & Hoffmann, 2000). Because changes in desiccation resistance are often related to mechanisms other than lipid levels, such as cuticular hydrocarbons and patterns of ventilation (Gibbs et al., 1997; Hoffmann & Harshman, 1999), a shift in levels of energy reserves may not affect resistance to this stress (or the wing traits).

Latitudinal variation in starvation and desiccation resistance may reflect differences in habitat conditions. Relative humidity declines with latitude of the collection sites (Table 1, r = −0.80, P < 0.01). Low humidity may lead to selection for increased desiccation resistance in D. birchii, particularly as this species is relatively sensitive to desiccation stress compared with D. serrata and other more widespread species. Habitat features that select for altered patterns of adult starvation resistance are unknown, although starvation resistance may be favoured in areas where food availability declines.

How steep are the clinal changes in starvation resistance, desiccation resistance and development time compared with those found in other species? One way of comparing clinal steepness is to compute means of populations at the ends of the cline and compare the shift in mean to the distance between the populations. Because traits are usually measured in different ways in different studies, the shift in a trait along a cline can be expressed as a percentage rather than an absolute value. In D. serrata and D. melanogaster from eastern Australia, there are no clinal patterns for desiccation and starvation resistance (Hoffmann et al., 2001a; Hallas et al., 2002) so these comparisons cannot be made. In D. kikkawai from India, there is a change of 1.5% per degree latitude for starvation resistance when expressed relative to the most sensitive population (computed from highest and lowest values for populations in Karan & Parkash, 1998). Estimates for other drosophilids in the same region vary from 3.9 to 6.6% (data from Karan et al., 1998). This compares to a value of around 9% for D. birchii when the most southern isolated population at Yeppoon is excluded. For desiccation resistance in India in several drosophilids, estimates range from 1.3 to 4.7% (data from Karan et al., 1998; Karan & Parkash, 1998), compared with around 8% for D. birchii. For development time, D. melanogaster exhibits a change of around 0.1% per degree latitude along the east coast of Australia (data from James & Partridge, 1995), compared with around 1% in D. birchii. These comparisons suggest that latitudinal clinal changes in D. birchii are steep relative to clinal changes in other drosophilids.

The clinal variation in wing shape suggests that shape does not vary with latitude in D. birchii. In D. serrata, there was a latitudinal change in the relative width of the outer wing margin (Hoffmann & Shirriffs, 2002); northern locations had wider outer wings relative to southern locations, although this pattern was confounded by a previously unknown cryptic species in the most northern collections (Schiffer et al., 2003). There was also a decrease in wing length relative to wing width in the northern locations (Hoffmann & Shirriffs, 2002), a pattern that was not confounded by the presence of the cryptic species. These changes in wing shape in D. serrata contrast with the absence of significant latitudinal patterns in D. birchii. We also directly examined the relative width of the outer wing margin and ratio of wing length to width in D. birchii and found no latitudinal pattern (analysis not shown).

The shape analysis suggests that wing shape was altered during laboratory culture, although wing size in D. birchii was constant. Although the effect of laboratory culture on wing shape has not previously been examined, other studies have demonstrated abundant heritable variation for wing shape in Drosophila (Bitner-Mathe & Klaczko, 1999; Weber et al., 1999), consistent with the variation we detected among D. birchii isofemale lines. Wing shape may therefore evolve readily in response to laboratory culture or other factors. Variation in wing shape can be related to flight performance (Kölliker-Ott et al., 2003). However, as flight performance is not under direct selection in laboratory conditions because flies are maintained in small containers, the shape changes presumably reflect indirect effects of selection on other traits. The changes in some traits under laboratory culture highlight the importance of assessing laboratory adaptation when comparing Drosophila for geographical variation, particularly for stress resistance and life history traits.

The design used by Sgro & Partridge (2001) and Hoffmann et al. (2001b) to test for laboratory adaptation has been criticized by Matos & Avelar (2001). These authors argued that when collections are made from the same location repeatedly, lines may show a different trajectory leading to different evolutionary endpoints. The current study differs from the others in two ways; flies were held and tested as isofemale lines, and were collected from several locations rather than one location, whereas they were obtained from one location and maintained as mass-bred populations in the other studies. As the D. birchii isofemale lines were initiated with the offspring of single females, these lines were partially inbred, and less likely to evolve than more genetically variable mass bred cultures. Nevertheless, with the exception of female development time, changes in responses to culture conditions were consistent for lines from the different locations. For development time, genetic background effects may have influenced the outcome of selection, as the Yeppoon location behaved differently than the other locations, although there may also have been changes in the genetic constitution of the Yeppoon population over time.

Heat resistance and wing size did not change with laboratory culture. These traits are also stable with laboratory culture in D. melanogaster. For wing size, clinal patterns persisted even when lines had been cultured in the laboratory for several years (Gockel et al., 2001). For heat resistance, resistance levels did not change as lines were cultured for increasing generations in the laboratory (Krebs et al., 2001). However in the case of D. birchii, a low heritability for heat resistance might be responsible for the absence of significant changes in the laboratory rather than an absence of selection. In D. melanogaster, development time evolved under laboratory conditions (Sgro & Partridge, 2001), in contrast to the situation in D. birchii where no consistent changes were detected.

In conclusion, geographical variation in traits was evident in D. birchii although this species occupies a limited geographical range. In D. serrata, clinal patterns have also been detected over short distances in the southern range of this species, but patterns were not stable (Magiafoglou et al., 2002). The presence of consistent clinal patterns in D. birchii for desiccation and starvation resistance might reflect low levels of gene flow in this species compared with D. serrata, which occurs over a range of habitats rather than being restricted to rainforest habitats that are often disjunct. The restricted distribution of D. birchii may also account for the steepness of the latitudinal clines in this species relative to other drosophilids. The shifts with laboratory culture for starvation resistance and cold resistance as well as wing shape indicate genetic variation for these traits in D. birchii populations, and suggest caution when interpreting selection responses for these or correlated traits and when deducing geographical patterns from lines maintained in the laboratory for many generations.


This research was supported by the Australian Research Council via their Special Research Centre program. We are grateful to Michael Kearney for running the ARCVIEW/ANUCLIM programs to determine climatic variables at collection sites. Chris Klingenberg and an anonymous reviewer provided valuable comments on an earlier version of this paper.