Larval growth rate is not a major determinant of adult wing shape and eyespot size in the seasonally polyphenic butterfly Melanitis leda

Study organism

M. leda is among the most widespread butterflies, distributed from Africa to Asia and Oceania (Latorre, 2018). It is able to use a wide variety of grass species (Poaceae) as host plant (Molleman, Halali & Kodandaramaiah, 2020b). M. leda displays seasonal phenotypic plasticity where wet-season forms have large and conspicuous eyespots, while dry-season forms have small eyespots (Fig. 2). Dry-season forms also have more falcate forewing tips and longer hindwing tails compared to the wet-season forms (Fig. 2; Brakefield, 1987; Halali et al., 2019). Field data suggest that this seasonal plasticity is to some extent mediated by temperature (Brakefield, 1987; Halali et al., 2021; Roy et al., 2021). Furthermore, the variation among dry-season form wing patterns of M. leda represents one of the most multi-faceted polymorphisms in wing coloration found in any animal (Ruiter & Brakefield, 1994).

General procedures and data collection

Three experiments were performed. The first experiment tested the effect of temperature on adult phenotype in a population from Ghana (Temperature Experiment). The second included effects of temperature and humidity using a population from South India (Temperature and Humidity Experiment). The third experiment tested for effect of host-plant species using the South-Indian population (Host Plant Experiment). While some methodological details differ, the data taken from these experiments are comparable.

In most cases, the parental butterflies were held in group cages so that eggs could not be attributed to particular mothers and were divided randomly over the treatments. When we did have eggs from specific mothers, we took special care to divide her offspring evenly over the treatments. In summary, caterpillars were reared on grasses growing in flower pots either from seeds or collected from the field (such as in Molleman, Halali & Kodandaramaiah, 2020b). The caterpillars were confined to the plants by placing the plants in fine-mesh sleeves. These sleeves were then kept in environmental test chambers that controlled temperature and humidity and had a 12/12 h light-dark schedule. Plants were watered every other day and were replaced once the majority of leaf material had been consumed or general plant quality had deteriorated. Once larvae reached the 5^th instar, sleeves were inspected every 24 h for prepupae and pupae. Prepupae and pupae were kept individually in 100 mL cylindrical transparent containers until eclosion. Prepupae were left attached to the plant whenever possible: using adhesive tape, a piece of the plant where the prepupa was attached was fixed to the side or lid of the container, so that the prepupae could hang freely while pupating. The prepupal stage never exceeded one day, so pupation date of prepupa was the following day. Absorbent paper was placed at the bottom of the container. Hardened pupae were gently removed from the plant material to be weighed and sexed, and then placed back inside the container. Containers were kept upside down with the absorbent paper on the lid so that pupae and eclosing butterflies were easy to observe through the transparent bottoms. The sides of each of these containers were lined with a strip of paper on which the eclosing adult could climb to enable it to expand its wings. Containers were checked daily for adult eclosions. Butterflies were frozen one day after eclosion.

Data were collected on development time from egg hatching to pupation (larval development time), and from pupation to eclosion (pupal time). The wings of the enclosed butterflies were removed and photographed or scanned. For the Temperature Experiment, both left and right wings were photographed on the dorsal side, and for the other experiments, one wing in a pair was imaged on the dorsal and the other on the ventral side. Eyespots, forewing tip, and hindwing tails were measured using an ImageJ macro (Schindelin et al., 2015) and averaged between left and right wings when appropriate. For this study, we focused on the largest eyespot on the forewing to represent eyespot size and the forewing tip to represent wing shape (Fig. 2). We used particular vein intersections as landmarks for our proxies for wing area and wingtip area (Fig. 2).

Temperature experiment, Ghanaian population

We collected eggs from 14 females from a stock population established based on 80 females from Bobiri, Ghana (Cites collection and export permit 012068 issued by the Wildlife Division, Ghana), and used at most the third generation reared in the laboratory. Sleeves contained 5–30 larvae, with the average being ca. 20, which were fed ca. two-week-old corn seedlings. These sleeves were placed within environmental test chambers at 75% relative humidity. Larvae were reared at five temperatures: 19 °C, 22 °C, 25 °C, 28 °C and 31 °C. Because of unexpected results, a follow-up experiment was carried out at three temperatures: 22 °C, 25 °C, and 28 °C (results in Appendix 1). Photographs were taken using a microscope imaging system (Leica DC200 digital camera with Leica MZ12 microscope).

Temperature and humidity experiment, Indian population

About fifty female M. leda butterflies were collected in Thiruvananthapuram, Kerala, South India, to provide eggs, and experiments were performed on three subsequent generations. Sleeves were set up with 15–17 caterpillars each, and were fed ca. two-week-old corn seedlings. Two environmental test chambers were used during subsequent experiments with combinations of four temperatures (19 °C, 21 °C, 27 °C, and 31 °C) and two levels of humidity (65% and 85% relative humidity). However, for technical reasons, a complete set of temperature and humidity combinations was not achieved so that the effect of humidity during rearing on adult phenotype could not be tested formally. The wings of the enclosed butterflies were scanned for phenotypic measurements (Konica Minolta, Bizhub 363, Osaka, Japan).

Host plant experiments, Indian population

The data presented here are based on the rearing of the fourth batch of larvae described in Molleman, Halali & Kodandaramaiah (2020b). In short, about fifty female M. leda butterflies were collected in Thiruvananthapuram to provide eggs. Larvae were then grown on potted plants belonging to 18 species of grass in an environmental test chamber with a constant temperature of 24 °C and a constant humidity of 69%. Photos of the wings of adults were taken with a Nikon D7000 camera under standard light inside a closed white styrofoam box with a shutter speed of 1/125 and an aperture of F14.

Data analysis

Relative eyespot size was measured by dividing the area of the M3 eyespot by a proxy of total wing area (Fig. 2). Wing shape was quantified as the area of the wing-tip triangle divided by the proxy for wing area (Fig. 2). Larval growth rate was calculated as (pupal mass)^1/3/larval development time), following Tammaru & Esperk (2007). We first explored the distribution of growth rate, relative eyespot size, and wing shape for both sexes using density plots. While larval growth rate approached a normal distribution, relative eyespot size and wing shape were biased towards very small numbers and this could be corrected by using the square-root transformation. For ease of use, we use ‘eyespot size’ as a shorthand for ‘square root of relative eyespot size’ and ‘wing shape’ as a shorthand for ‘square root of relative size of the forewing tip triangle’.

To generate provisional reaction norms, the growth rate, eyespot size and wing shape were plotted against treatment for all three experiments. For the Temperature and Humidity Experiment, data from a pilot experiment were included in the provisional reaction norm. These were second instar larva that had beenreared at 85% RH at either 21 °C or 27 °C, and sex was not determined. For the Host Plant Experiment, plants on which less than eight larvae were reared to adults were excluded, and plants were ordered according to larval growth rate.

To test if growth rate predicts wing traits within treatments, we implemented mixed models and performed model selection based on the Akaike Information Criterion corrected for sample size (AICc). In these models, the dependent variables were either the square root of relative eyespot size or of relative forewing tip size (wing shape). Predictors were larval growth rate, treatment (temperature, humidity, or host plant species), sex, wing area, and interactions between sex and growth rate and sex and treatment when the sample sized allowed it. Sleeve was included as random effect. We made sure that interactions between categorical predictors and growth rate were calculated correctly by subtracting the average growth rate from the measured growth rates in each experiment. To obtain more insight into relationships with body size and development time, we performed further analyses presented in Appendix 3. These include models that did not include wing area as a predictor, and models that replaced the predictor growth rate by pupal mass and development time and their interaction. All analyses were performed using the R packages lme4 (version 1.1.35.3; Bates et al., 2011) with lmerTest (version 3.1.3; Kuznetsova, Brockhoff & Christensen, 2015) to estimate degrees of freedom, and MuMIn (version 1.47.5; Barton & Barton, 2015) to perform model selection in R (version 4.4.0; R_Core_Team, 2023), and graphs were made using the R package ggplot2 (3.5.1; Wickham, 2016).

Cue use and reaction norms

We reared 364 individuals in 44 sleeves in the Temperature Experiment, 82 individuals in 18 sleeves in the Temperature and Humidity Experiment, and 260 individuals in 93 sleeves in the Host Plant Experiment (see sleeves numbers and sample sizes per treatment and sex in Appendix 2). Therefore, we report on a total of 706 butterflies that were reared from egg to adult.

Larval growth rate generally increased with temperature (Figs. 3A & 3B), although it was similar for those reared at 22 °C and 25 °C in the Temperature Experiment (Fig. 3A). The range of growth rates on different host-plant species was only slightly smaller than that generated by using different temperatures (Fig. 3C). Differences in growth rate between temperature treatments were mainly due to differences in development time, as at low temperature (19 °C) the development time was almost three times longer compared to that at high temperature (31 °C), while pupal mass was about one third higher when caterpillars are reared at lower temperatures (Appendix 3). In contrast, within treatments, most of the variation in growth rate was caused by variation in pupal mass, while development-time variation within treatments was modest compared to among treatment variation, especially at higher temperatures (Appendix 3). Across treatments, higher pupal mass was associated with longer development time, but within treatments, higher pupal mass was generally associated with shorter development (i.e., reaction norms for age and size at maturity had a negative slope). Pupal mass was strongly correlated with wing area, but this relationship differed between temperature treatments and the sexes (Appendix 3). Butterflies tended to be smaller when they were reared at higher temperatures (with high growth rate), while they tended to be larger on plants on which they grew faster.

Eyespot size had a bimodal distribution across all data, but was unimodal within treatments (Appendix 2, Figs. 4A & 4B). Eyespot size increased with increasing rearing temperature in both populations (Figs. 4A & 4B). However, the relationship was highly non-linear in the Ghanaian population with eyespot size being smaller when reared at 25 °C than at 27 °C (Fig. 3A). Similar results were found in the follow-up experiment designed to verify this non-linear pattern (Appendix 1). Notably, the sexes showed qualitatively similar temperature reaction norms for eyespot size (Figs. 4A, 4B) that were nevertheless significantly different in the Temperature Experiment (Table 1, interaction between temperature and sex). At intermediate rearing temperatures, individuals with both wet- and dry-season form eyespots were produced, with a few intermediates (22 °C, 25 °C Fig. 3A, 21 °C, 27 °C Fig. 4B). Higher humidity appeared to result in more wet-season form individuals at intermediate temperatures (Fig. 4B), but note that these are pilot data starting from 2^nd and 3^rd instar caterpillars, and these data are not included in further analyses. Average eyespot size also varied with host-plant species (Fig. 4C), but the range was narrower as full dry-season phenotypes were not produced in this batch (other batches without quantitative data on wing phenotypes did include dry-season phenotypes). Larger wings were associated with relatively smaller eyespots, indicating that wet-season form butterflies tend to be smaller than dry-season form ones (see Appendix 3 for an illustration of this relationship and graphical and statistical analyses with development time and pupal mass as predictors of butterfly phenotype).

In contrast to eyespot size, wing shape did not show a bimodal distribution in the Indian population, as a large number of individuals showed intermediate wing shapes (Fig. 5, Appendix 2). Otherwise, results for wing shape were similar to those for eyespot size: increasing temperature caused more wet-season phenotypes (less falcate wing tips), and there was an effect of host plant (Fig. 5), with a significant interaction between temperature and sex in the Temperature Experiment (Table 1). Larger wings were also associated with more falcate wing tips (see Appendix 3 for an illustration of this relationship and graphical and statistical analyses with development time and pupal mass as predictors of butterfly phenotype). While eyespot size and wing shape were correlated with each other, this relationship was not particularly strong (Fig. 6), with some butterflies combining large eyespots with falcate wing tips and vice versa (Fig. 7).

Growth rate influencing eyespot size

At higher temperatures, larval growth was faster and adults tended to show more wet-season phenotypes (larger eyespots) in both the Ghanaian population (Fig. 8A, Table 1) and in the Indian population (Fig. 8B, Table 1). Similarly, higher growth rates were associated with larger eyespots within treatments (Fig. 8C, Table 1). However, within treatments, the relationship between growth rate and eyespot size was not very strong, with much scatter and in some treatments opposite trends (Fig. 8, see for statistics of individual regression lines Appendix 2). The relationship also differed between the sexes (Table 1), and in the Host Plant Experiment this relationship was much stronger among males (Fig. 8C). Moreover, at intermediate temperatures, there were clear differences in adult phenotype between temperature treatments that produced a similar range of growth rates (22 °C and 25 °C in Fig. 3A, and 21 °C and 27 °C in Fig. 8B). In the host plant experiment however, when growth rate was included in the model, host plant was not selected in the model and was not a significant predictor, indicating that the effect of host plant on eyespot size can be explained by its effect on growth rate. Wing area was a significant predictor of relative eyespot size in all three experiments (Table 1A) and was selected into the final models (Table 1B). When wing area was included into the models, growth rate remained a significant predictor of relative eyespot size (Table 1; model results without wing area are given in Appendix 3, Table A3.2).

Growth rate influencing wing shape

In all experiments, lower larval growth rate was associated with more falcate wings in both sexes across and within treatments (Fig. 9, Table 1). However, at intermediate temperatures, there were clear differences in adult phenotype between temperature treatments that produced a similar range of growth rates (22 °C and 25 °C in Fig. 4A, and 21 °C and 27 °C in Fig. 9B), and there was considerable scatter and some treatments showed opposite trends (see Appendix 1 for statistics of individual regression lines). The relationship was stronger among males than among females, particularly in the Host Plant Experiment (Table 1, Fig. 9C). Males tended to have less falcate wings when they were wet-season forms than females, so that the range of wing shapes was wider in males (most clearly seen in Figs. 9A & 9C). Overall, both eyespot size and wing shape were significantly but weakly related to larval growth rate within treatments (Table 1), the relationships appearing stronger for wing shape and among males (Table 1, Appendix 2). Notably, growth rate was a particularly poor predictor of wing phenotype among the 22 °C and 25 °C treatments of the Temperature Experiment. Nevertheless, in the host plant experiment, host plant was a significant predictor, but was not selected in the model when growth rate was included, indicating that the effect of host plant on wing shape size can to a large extent be explained by its effect on growth rate. Wing area was a significant predictor of relative eyespot size in the Temperature Experiment and the Temperature and Humidity Experiment (Table 1A) but was not selected into final models (Table 1B). When wing area was included, growth rate remained a significant predictor of wing shape (model results without wing area are given in Appendix 3, Table A3.2).