Ditte Holm Andersen, Dipartimento di Biologia Evoluzionistica Speriemntale, Via Selmi 3, I-40126 Bologna, Italy. Tel.: +39/0512094172; fax: +39/051251208; e-mail: email@example.com
Maternal effects on progeny wing size and shape in a homozygous parthenogenetic strain of Drosophila mercatorum were investigated. The impact of external maternal factors (heat stress) and the impact of internal maternal factors (different maternal and grand maternal age) were studied. The offspring developed under identical environmental conditions, and due to lack of genetic variation any phenotypic difference among offspring could be ascribed to maternal effects. Wing size was estimated by centroid size, shape was analysed with the Procrustes geometric morphometric method and variation in landmark displacement was visualized by principal component analysis. Both kinds of maternal effects had a significant impact on progeny wing size and shape. Maternal heat stress led to the same pattern of response in size and shape among the progeny, with increased difference between the control group and progeny from heat stressed flies in both size and shape with increased maternal heat stress temperature. The effects of maternal age, however, led to different responses in size and shape between the different progeny groups. The observed variation in landmark displacements was similar, and in both cases mainly associated with shape differences of the posterior part of the wing. Finally, our results suggest that maternal effect has some evolutionary implications by altering the genetic correlations among traits, which can affect the response to selective pressures.
The phenotype of an individual largely reflects its genotype and the environmental factors experienced during development, but other factors can also contribute. Among these are maternal effects. Maternal effects are the product of cross-generational interactions between parental phenotype, parental environment and offspring genotype and are expressed as phenotypic differences in the offspring (Keena et al., 1998). The term ‘maternal effects’ therefore serves as a broad designation for the outcome of multiple cross-generational processes (Rossiter, 1996).
Maternal effects seem to be ubiquitous across a diversity of taxa and often cause a considerable proportion of the variance in many characters, especially those expressed early in development (Wolf et al., 2000). Its importance is not questionable, and there are experiments indicating that maternal effects can account for as much as 50–80% of the offspring variation in clutch size, individual size, survival and behavioural differences (Cheverud & Moore, 1994; McAdam et al., 2002; Pakkasmaa et al., 2003). Even though maternal effects can have a considerable effect on offspring phenotype, they have often been neglected due to the difficulties of quantification, as maternal effects can be hard to distinguish from environmental and genetic factors. In many cases, the form and function of maternal effects is not simply the accidental transmission of environmental information from one generation to the next, but maternal effects may have been shaped by natural selection (Mousseau & Fox, 1998). This can happen because the maternal environment provides a reliable indicator of the environmental conditions that the progeny will encounter. In such cases maternal effects may evolve as mechanisms for trans-generational phenotypic plasticity, whereby, in response to a predictive environment, a mother can change the type of eggs that she produces or can program a developmental switch in her offspring, which produces offspring prepared for the maternal environment (Fox et al., 1999).
Many aspects of the parental environment can influence the allocation of resources by parents to their offspring and the quality of parental care. Maternal effects can arise as a result of mothers being exposed to altered temperatures or changes in other environmental factors influencing the phenotype or fitness traits in the progeny. Gilchrist & Huey (2001) found that in Drosophila melanogaster, progeny from parents reared at 29 °C had a higher fitness than progeny from parents developed at lower temperatures, independent of offspring environment. Maternal effect can also result from the age of the parents (Parsons, 1964; Mousseau & Dingle, 1991). For example, offspring from old female insects are usually expected to be inferior relative to the offspring of young females (Mousseau & Dingle, 1991). Differences in grand maternal and maternal age are also known to affect egg-to-adult viability in D. serrata (Hercus & Hoffman, 2000) and increased maternal and paternal age in general produced offspring with decreased longevity in D. melanogaster (Priest et al., 2002).
Investigations trying to elucidate the impact of maternal conditions (internal or external) on the progeny have mainly been concentrated on how maternal effects affect offspring fitness traits. Maternal effects are therefore often measured as the difference in size of progeny characters, as bigger size is often related to increased fitness (McIntyre & Gooding et al., 2000; Sakwińska, 2004). This approach can, however, be too simple, as variation in shape rather than size will not be detected. The shape of morphological traits is in general assumed to be more stable than the size and under strong developmental constraints (Debat et al., 2003). Any changes in shape as a result of maternal effects would therefore stress the importance of maternal effects. The choice of method for characterizing morphological forms is very important. Conventionally, characterization has been done by analysing sets of linear distances measured on each specimen. In the past two decades, however, several new methods were developed that emphasize the geometry of a morphological structure and are based either on outline contours or the arrangement of landmark points (e.g. Bookstein, 1996; Dryden & Mardia, 1998). An ideal shape feature to use in the investigation of maternal effects on shape is the Drosophila wing, because the wing veins provide many morphologically well-defined landmarks, and Drosophila wing development is well understood (Stark et al., 1999; De Celis, 2003).
In this paper we consider the impact of maternal effects on the size and shape of the Drosophila wing using Procrustes analysis. To be able to ascribe any differences in progeny wing size and shape to maternal effects we used a homozygous parthenogeneic strain of D. mercatorum (Kramer & Templeton, 2001). The lack of genetic variance among progeny developed under identical environmental conditions allows us to exclude genetic and environmental factors, so that any phenotypic differences among the investigated individuals must be due to maternal effects. In order to evaluate the action of maternal effects we used two very different approaches. One is to evaluate maternal effects on progeny wing size and shape arising due to changes in external maternal conditions. An external factor known to be very important for insect distribution and abundance is temperature (Sørensen et al., 2001). It may therefore be expected that maternal heat stress exposure could affect the shape of offspring wings. We therefore investigated if exposing parental flies to increasing temperature stress has an impact on progeny phenotype. Another aim was to investigate maternal effects on offspring wing size and shape arising due to internal maternal conditions. This was done by investigating the maternal effects due to internal physiological changes associated with changes in maternal age. Age induced maternal effects may last for more than one generation (Hercus & Hoffmann, 2000), so we evaluated the effects of both maternal and grand maternal age. Lastly we would like to evaluate if these two very different origins of maternal effects (external and internal) differently affect the shape of the progeny wing, by acting on different parts of the wing depending on the kind of maternal effect.
Materials and methods
The investigation of internal and external maternal effects on offspring wing size and shape was in both experiments done by the use of a homozygous, parthenogenetic strain of D. mercatorum, belonging to stock Iv-23-0-Im (Kramer & Templeton, 2001). The flies reproduce by pronuclear duplication, which ensures complete homozygosity. This strain was established in 1990, and has been re-established from one female immediately before the start of our experiments, to exclude any genetic variance that might have arisen due to mutations. Previous to the experiments the flies used in both investigations were kept on instant Drosophila medium (Carolina Biological Supply, Burlington, NC, USA) at 25 °C and at a 12 : 12 h light : dark cycle. It is unfortunately impossible to completely exclude environmental variation among the progeny, but to minimize the environmental component as much as possible we measured, in both cases, only flies that hatched within the first 3 days, as these flies are unaffected by waste products experienced by the larvae in the media. The density of maternal flies was kept at 15 flies per vial. Neither chemical stress nor crowding is factors that affect shape or size in parthenogenetic D. mercatorum as the hatching rate is very low (1–2%) (Kramer & Templeton, 2001).
Maternal heat stress
After hatching, parental flies were transferred at random into new vials, each containing 15 flies. When the flies were 2-day old they were stressed in water-baths at 36, 37, 37.5, 38 °C or at a control temperature of 25 °C (water-bath accuracy is ±0.1 °C) for 30 min twice a day (at the same time of the day), 12 h apart, for a total of 4 days. Before each stress exposure, flies were transferred to empty vials with moistened stoppers to avoid desiccation stress during heat exposure. One hour after stress exposure, flies were returned to vials with medium. At day 6, 1 h after the last water-bath treatment, the surviving flies were transferred at random into new vials for egg laying, each vial containing 15 flies. The new vials ensured that all eggs were laid after the mothers had been exposed to the full stress period. Flies that hatched from these vials within the first 3 days were used for phenotypic measurements. For each temperature 250 flies were collected and measured.
The parental flies used in the experiment were collected immediately upon hatching and transferred at random to vials each containing 15 flies; the flies were transferred to fresh vials every third day. Two groups of offspring were used: (1) offspring from eggs laid by mothers 3–6-day old (Young mothers) and (2) offspring from eggs laid by mothers 15–18-day old (Old mothers). These two groups are referred to as Y and O, respectively, referring to the age of the mother.
The vials from the two age-classes gave rise to the F1 generation containing offspring from either Y or O mothers. The next generation was obtained in the same way as the F1 generation, giving rise to the F2 generation, classified as YY, YO, OY or OO (first letter referring to the age of the grandmother, second letter to the age of the mother). Only flies hatched within the first 3 days were used for size measurements. From each group 150 flies were collected for further measurements.
Measurements and Procrustes analysis
The wings of the flies from the two experiments were removed and placed in a droplet of lactic acid on a microscope slide and covered with a cover slip. The wings were measured using a camera attached to a dissecting microscope and a computer, and by the use of the software package ImageJ (Rasband, 2001). Five landmarks were used (A, B, C, D and E) (Fig. 1). To quantify possible measurement errors, we chose 11 parthenogenetic flies at random. For each of these flies all landmarks were measured 10 times and the within individual coefficient of variation for each mean was taken as an estimate of the measurement error, adding Haldane's (1955) correction for small sample size. Measurement errors were low for all landmarks ranging from 0.11 to 3.8% (mean = 0.36%).
After the landmark coordinates were recorded, the configurations were superimposed onto the overall mean configuration using the Procrustes generalized least-squares procedure (Rolf & Slice, 1990). The Procrustes method proceeds in several steps: (1) the landmark configurations of all wings of one body side are reflected to their mirror images, (2) all configurations are scaled to unit centroid size, (3) the centroids of the configurations are superimposed, for example, on the coordinates of the origin (0, 0). This step eliminates differences in position and (4) the configurations are rotated against each other around their centroid to achieve an optimal fit of corresponding landmarks. The best fit is defined as the rotation that minimizes the sum of squared deviations of the landmarks of all configurations from the corresponding landmarks of the overall consensus (mean) configuration. This last step eliminates variation in the orientation of the specimens.
Using landmark configurations as shape descriptor variables, multivariate statistics allows one to precisely and objectively characterize mean shapes and quantify shape variation. This variation may in turn be decomposed into independent components, which can be interpreted in terms of evolutionary processes (Debat et al., 2000).
The size of an individual wing was estimated by the centroid size of its landmarks configuration, i.e. the square root of the sum of squared distances between each landmark to the centroid (Slice et al., 1996). One-way anova was applied to the centroid size to test for differences in mean wing size among the offspring from mothers exposed to the different temperature treatments, or differences in mean wing size among offspring from the different maternal age groups. Post hoc comparisons were done by the use of Tukey's HSD test (Zar, 1999).
Overall shape variation
In order to test for differences in mean shape between offspring from the different maternal heat stress groups or differences in mean shape among offspring from the different maternal age groups a manova was applied to the Procrustes residuals. The transformation of Procrustes coordinates into Procrustes residuals was done by subtracting the mean shape (O'Higgins & Dryden, 1992). A canonical variates analysis (CVA) (multigroup discriminant analysis) was conducted for each of the two experiments. This approach allows optimal visualization of the relative position of the different samples in the multivariate statistical space by maximizing the among samples variation (Manly, 1986). To test whether differences in mean shape occur a permutation test for two multivariate groups was applied (2000 permutations) and the Mahalanobis distance was estimated, to test for differences between offspring from control mothers (25 °C) and offspring from the maternal heat stress groups or between offspring from YY mothers and offspring from the other maternal age groups, and a comparison between YO and OY offspring. Because there are four eigenvalues that are zero in Procrustes fit, generalized inverses must be used in the calculations of manova, CVA and Mahalanobis distance. We have used the software PAST (Hammer et al., 2001), that automatically uses the generalized inverses in these calculations.
Patterns of shape variation among individuals
To analyse and display the patterns of covariation in the positions of landmarks throughout the wing either as a consequence of maternal heat stress or maternal age, we used a principal component analysis (PCA) on the Procrustes coordinates, which has been used regularly in the context of shape analysis (Klingenberg & McIntyre, 1998; Debat et al., 2000; Klingenberg & Zaklan, 2000). This analysis extracts features of shape variation as a set of new shape variables (the principal components, PCs) that are uncorrelated with one another and successively account for maximal amounts of variation. It is possible to interpret them as independent features of variation that can be added together to make up the observed variation among samples. In geometric shape analysis, the PCs can be visualized graphically in direct relation to the landmark positions on the fly wing.
The anova on centroid size of the wings showed highly significant differences among the progeny from the different maternal heat stress groups (F4,1245 = 439.3, P < 0.001). Mean centroid size in the offspring increased with increasing maternal temperature exposure (Fig. 2). Post hoc comparisons revealed highly significant differences in centroid size between all maternal heat stress progeny groups (all P < 0.001) except between 37.5 and 38 °C where no difference in size was observed.
The anova applied to the centroid size of the wings of the progeny from the different maternal age groups was highly significant (F3,1196 = 122.9, P < 0.001) showing that maternal age had a significantly different effect on wing size. Post hoc comparisons revealed highly significant differences in centroid size in all cases except for offspring having OO and YO mothers where no differences were observed (see Fig. 3). The highest and lowest centroid sizes were observed in flies from the OY and the YY groups, respectively.
Overall shape variation
The manova on Procrustes residuals of the progeny from the maternal heat stress experiment, was found to be highly significant (Wilk's λ = 0.75, F40,9413 = 18.67, P < 0.001), suggesting that wing shape varies significantly among progeny from the different maternal heat stress groups. The CVA provided a discrimination of the samples in the plan defined by the two first canonical axes, which accounted for 59.39 and 28.85% of the among maternal temperature variance in shape, respectively (see Fig. 4). The combination of both axes permits a discrimination of the five samples whose positions showed an overall curved trajectory with increasing temperature. Canonical axis1 mainly contrasts the difference in shape of the offspring from control mothers (25 °C) from offspring of mothers stressed at 37.5 °C and canonical axis2 contrasts mainly the shape difference between offspring of mothers stressed at 36 and 37 °C from offspring of mothers stressed at 38 °C.
The multivariate permutation test for pair-wise comparisons of mean shape revealed in all cases a highly significant differences between offspring from mothers exposed to control temperature (25 °C) and offspring from temperature stressed mother (see Table 1). The Mahalanobis distance between control offspring and offspring from temperature stressed mothers increased with increasing maternal stress temperature up to 37.5 °C (see Table 1). The Mahalanobis distance between control offspring and offspring from maternal flies stressed at 38 °C decreased when compared to the Mahalanobis distance between control offspring and offspring from mothers stressed at 37.5 °C.
Table 1. Multivariate permutation test for differences in mean shape: (1) between the offspring from mothers exposed to a control temperature (25 °C) and offspring from temperature stressed mothers or (2) between offspring from mothers coming from different age groups.
Due to the large number of tests conducted we applied a Bonferroni correctiont (Rice, 1989). Following Miller (1981) we made a separate probability statement for each experiment (k = 4).
The values in the table are Mahalanobis distances.
*P < 0.01, **P < 0.001.
25 vs. 36 °C
25 vs. 37 °C
25 vs. 37.5 °C
25 vs. 38 °C
YY vs. YO
YY vs. OY
YY vs. OO
YO vs. OY
The manova on the Procrustes residuals of the progeny from the different maternal age groups was highly significant (Wilk's λ = 0.87, F30,3496 = 5.60, P < 0.001), suggesting that wing shape varies significantly among offspring from mothers and grandmothers with different age. The CVA provided discrimination between the samples by the first two canonical axes, which accounted for 51.92 and 31.37% of the among shape variation among the progeny coming from different maternal age groups, respectively (see Fig. 5). Canonical variates analysis revealed the highest similarity in shape between YY and OO progeny. Canonical axis1 mainly contrasts OO and YY progeny from YO progeny, and canonical axis2 mainly contrasts OO and YY progeny from the OY progeny.
The multivariate permutation test for pair-wise comparisons between offspring from the different maternal age groups revealed the lowest Mahalanobis distance among YY and OO offspring, the highest distance was found between progeny from YO and OY mothers (see Table 1).
Patterns of shape variation among individuals
The finding of low Mahalanobis distances among progeny groups from both experiments indicates that most of the observed variation in shape occurs within groups rather than between groups. The PCs therefore primarily describe the variation in landmark displacement within progeny groups.
The PCA showed that few PCs accounted for a relatively large amount of variation in both experiments. The patterns of variation corresponding to the first three PCs showed relations to the local arrangement of the veins (Fig. 6). The PC1 revealed the same kind of variation in landmark displacement in both experiments and the dominance of PC1 was linked to the large variability of landmark D, which was always associated with the fifth longitudinal vein (Fig. 6). The PC1 was, in both investigations, associated with a movement of landmarks B and C along the wing margins and a parallel movement of landmarks A and E along the second and third longitudinal veins (Fig. 6).
The PC2 was, in both experiments, also linked to a large variability of landmark D, which consisted of a contraction or expansion of IVR D (see Fig. 6). In both cases PC2 affected the distal landmarks B and C by a movement along the wing margin. The PC2 affected landmarks A and E differently in the two experiments. The variance in placement of landmarks A and B was directed along the second longitudinal vein and the anterior cross vein, respectively, among progeny from the different maternal heat stress groups. The movement of these two landmarks was directed along the anterior posterior axis for landmark A and along the third longitudinal vein for landmark E.
The PC3 was mainly related to variance of landmarks A and E, showing the same kind of movements in both experiments (Fig. 6). Differences in displacement of landmarks C and D was observed between the two experiments for PC3, resulting in an expansion or contraction of the wing due to maternal heat stress, but only a displacement along the wing margin due to differences in maternal age (Fig. 6).
Due to lack of genetic and environmental variability among the progeny the observed differences in offspring phenotype must be due to maternal nongenetic effects. Our results indicate that the external or internal factors experienced by the maternal flies in our two experiments have an impact on offspring phenotype, observed as differences in size (Figs 2 and 3) but also in shape (Figs 4 and 5).
The individual variance of shape due to the maternal effects described by the PCs was in both investigations related to a high variability of landmark D, and differences between the two experiments for individual variance in shape were only observed for landmarks A and E for PC2 and landmarks C and D for PC3. Since variation of the PCs is mainly determined by the within progeny groups variation rather than between groups, the intrinsic and extrinsic maternal effects were only responsible for a small part of the observed variation in landmark displacement in the progeny. Due to the observation of the same kind of shape variance in both experiments, mainly determined by the within group variation, the investigated maternal effects do therefore not seem to affect the variance in shape of the progeny or the two kinds of maternal effects may lead to the same kind of individual variance in shape.
The wing centroid size in the progeny increased with maternal heat stress temperature up to 37.5 °C, while at 38 °C no further increase in centroid size was observed (Fig. 2). The differences in mean wing shape found among the progeny belonging to the different maternal heat stress groups was visualized as a curved trajectory with increasing maternal stress temperature and the highest Mahalanobis distance for differences in shape, which was found between 25 °C control progeny and 37.5 °C. The changes in size and shape seem to follow the same pattern as a response to maternal heat stress, with the largest difference in shape and size found between 25 and 37.5 °C maternal heat stress progenies. The progeny, therefore, displayed an increasingly negative maternal effect on both size and shape with increased maternal heat stress. Wing size is usually associated with body size in Drosophila, meaning that the negative maternal effect could be associated with increased progeny size. An increase in body size is often positively associated with a number of fitness components (McCabe & Partridge, 1997; Reeve et al., 2000), so in this case the maternal effects passed on from the mothers to their offspring with increased maternal stress temperature could result in increased offspring fitness.
The observed phenotypic differences among the progeny due to temperature stress of the mothers are a combination of environmental covariance between parents and offspring due to direct effects of the temperature stress affecting the oocytes in the maternal flies and the effects mediated through maternal physiology. The direct effects of temperature on the oocytes could be viewed as behavioural maternal effects that occur in nature when maternal flies choose to stay in places with high temperature (e.g. sun exposed surfaces). The distinction between the impact of the direct effects of temperature on the oocytes and the maternal physiologically mediated effects on the phenotype of the progeny is not possible in this investigation.
In our investigation, offspring from the OY group displayed the largest centroid size in contrast to the flies from the YY group, in which the smallest centroid size was observed. In both the OY and the YY groups the flies have a young mother (Fig. 3). The differences in size must therefore be due to maternal effects from the grandmother leading to differences in the centroid size found between these two groups. The offspring from the OO group was equal in centroid size to YO, larger than the YY, but smaller than the OY flies. Because of the old maternal age in the OO and YO groups; the mother is less affected by the age of the grandmother, since maternal effect is believed to have the largest impact early in life (Fox, 1997; Wolf et al., 2000). This indicates that the grandmother's role for offspring centroid size is insignificant when the mother is old, but plays a role when the mother is young. When considering shape, however, the flies displayed a very different scenario. In contrast with the maternal effects resulting from maternal heat stress, maternal effects arising from maternal age acted differently on size and shape. We found that the progeny groups that were the most similar in shape were offspring from YY and OO mothers (Table 1 and Fig. 5). It seems that the maternal effects creating the differences in shape are more related to the difference in age between the grandmother and the mother than to the actual age of the maternal flies. Our experiment does not indicate that maternal effects are cumulative in their action on the progeny wing shape phenotype, because progeny that have both an old grandmother and an old mother do not differ more in shape from YY progeny than progeny having only one old mother. Rather it seems as though the impact of maternal age on progeny wing shape, by having an old grandmother, is in some way counteracted by also having an old mother, making the YY and OO progeny the most similar in shape. The biological meaning of the similarity in wing shape between progeny from the OO and the YY groups and the processes responsible for the observed similarity need to be studied further. The scope of our paper was to investigate if and how intrinsic and extrinsic maternal effects affected size and shape. Further studies should be performed, including the use of several independent parthenogenetic strains and other organisms, to investigate if the shape similarities between the OO and the YY groups are a general phenomenon, and other approaches should be applied in order to elucidate which factors lead to the observed similarities.
In both investigations the phenotypic differentiations are mainly due to differences in placement of landmark D. Maternal age and maternal heat stress are two very distinct kinds of stress, leading to the same kind of individual variation in wing shape in the progeny. The shape variation was mostly associated with IVR D, related to the width of the wing. Maternal effects acted differently on the shape of the wing when compared to shape variation as a result of direct exposure of individuals to different developmental temperatures which was mainly associated with variation of landmark C associated with the length of the wing (Debat et al., 2003).
The observed change of shape produced by heat-stress or aging also has some evolutionary implications, as a change in shape implies a change of the correlations between traits and the correlation matrix among several traits. The estimation of covariance matrices has long been a central part of evolutionary quantitative genetics (e.g. Roff, 1997; Lynch & Walsh, 1998) since the response to selection depends on the patterns of genetic and phenotypic variation, represented by the additive genetic and phenotypic covariance matrices (Cheverud, 1984). The estimation of genetic variance–covariance matrices from phenotypic data is, however, difficult, as nonadditive effects may interfere. We have avoided these difficulties in our investigation by using a monoclonal strain, which means that any difference between two phenotypic matrices implies that maternal effect also can alter the genetic correlations among traits (reducing or increasing it) and therefore altering the speed of response to selective pressures.
We would like to thank Christian Klingenberg for many useful comments and valuable help, and one anonymous reviewer for critical comments on the MS. We are grateful to Anni Røgilds for laboratory work. The investigation was supported by grants from the Danish Natural Sciences Research Council (642-01-0087) to Ditte Holm Andersen and by grants from the Danish Natural Sciences Research Council (21-01-0526 and 21-03-0125) and the Marie Curie Fellowship of the European Community Host Development program under contract number HPMD-CT-2000-00009 to Cino Pertoldi.