Edited by: Jean-Michel Gibert, Centre National de la Recherche Scientifique (CNRS), France
Reviewed by: Vincent Debat, Muséum National d'Histoire Naturelle, France; François Mallard, INSERM U1024 Institut de biologie de l'Ecole Normale Supérieure, France
†Present Address: Shampa M. Ghosh, School of Biotechnology, Kalinga Institute of Industrial Technology (KIIT), Bhubaneswar, India
K. M. Satish, Department of Biotechnology, College of Agriculture, Navile, University of Agricultural and Horticultural Sciences, Shivamogga, India
Mohan Jayaram, Centre of Excellence in Genomics and Translational Medicine, University of Tartu, Tartu, Estonia
This article was submitted to Evolutionary Developmental Biology, a section of the journal Frontiers in Ecology and Evolution
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Canalization denotes the robustness of a trait against genetic or environmental perturbation. Plasticity, in contrast, indicates the environmental sensitivity of a trait. Stabilizing selection is often thought to increase the canalization of a trait, whereas directional selection is often thought to lead to decanalization. However, the relationship between selection, canalization, and plasticity remains largely unclear. Using experimental evolution, here, we ask whether long-term directional selection for reduced pre-adult development time in
For organisms developing in the wild, environmental variation at various spatio-temporal scales is ubiquitous; yet, the phenotypic expression of many traits shows a surprising degree of robustness (Félix and Barkoulas,
Canalization or robustness as a phenomenon is intertwined with the related phenomenon of phenotypic plasticity. Phenotypic plasticity reflects the environmental sensitivity of a trait. Traditionally, plasticity referred to the ability of a genotype to give rise to different phenotypes in different environments (Woltereck,
At a population level, both selection and canalization tend to limit deviations from an optimal phenotype (Eshel and Matessi,
Experimental evolution is a powerful approach for addressing questions in evolutionary biology (Garland and Rose,
Pre-adult development time had shown a strong response to selection with a 25% reduction in mean development time relative to ancestral controls after 100 generations of selection (Prasad et al.,
In view of this, we examined the extent to which direct and correlated responses to directional selection experienced under a stable environment are canalized, by investigating whether the selected populations exhibited canalization for pre-adult development time and/or survivorship across different rearing temperatures and egg densities, compared to their ancestral control populations.
Temperature and pre-adult density are known to affect both development time and survivorship in
In the faster developing populations, every generation, only the earliest 25% of the eclosing flies were selected and allowed to reproduce, as opposed to the ancestral control populations in which all flies irrespective of their development time were allowed to breed. Development time, therefore, is under strong selection in the faster developing populations in contrast to the controls in which the trait is not under conscious selection. Therefore, we expect development time to be more canalized in the selected populations compared to the controls, and hence show less variation, reflecting microenvironmental canalization. It is important to note, nonetheless, that the nature of selection on development time in our study populations is directional, and not stabilizing. However, in terms of relevance to fitness, the trait is of utmost importance in the faster developing populations unlike that of the controls, and hence we expect the trait to be more canalized in the former than the latter.
We also investigated the plasticity of development time in both sets of populations across different novel environments including conditions that could be moderately stressful (e.g., high rearing densities). The within environment variation for the trait in all assay environments was measured as well. Lower plasticity of development time
The other trait we investigated, i.e., pre-adult survivorship, had shown a strong correlated reduction in the faster developing populations compared to the controls in course of selection (Prasad et al.,
We used eight laboratory populations of
All populations were maintained on a discrete generation cycle at ~25°C, ~90% relative humidity and constant light, on banana-jaggery food. In both JB and FEJ populations, larvae were reared in 8 dram glass vials (9.5 cm height, 2.4 cm inner diameter) with ~6 mL food at a density of 60–80 larvae per vial, whereas eclosed adults were collected into Plexiglas cages (25 × 20 × 15 cm3) with abundant food, at breeding population sizes of about 1,500 flies. The JBs were maintained on a 3-week discrete generation cycle, and all eclosing adults were part of the breeding population. In JB populations, eggs were collected and dispensed in vials on the 21st day after the egg collection for the previous generation, corresponding to day 11 of adult life. FEJs were maintained under conditions similar to the JBs, except that 120 vials, rather than 40 vials, containing ~60–80 eggs were collected per population, and the vials were monitored for eclosion every 2 h after pupae darkened. As soon as the first 20–25% flies in each vial (12–15 flies) had enclosed, they were transferred into fresh cages containing food plates. These constituted the breeding adults. After 3 days of adult life, eggs were collected from FEJ cages to start the next generation. Thus, the FEJs were under strong primary selection to complete egg-to-adult development fast, and under secondary selection to be relatively fecund on day 3 of adult life.
As each FEJ population was derived from one JB population, selected and control populations bearing identical numerical subscripts were more related to each other than to other populations in the same selection regime. Consequently, control and selected populations with identical subscripts were treated as constituting random blocks in the statistical analyses. At the time of this study, the FEJs had undergone 295 generations of selection, and showed considerable evolutionary reductions in development time (~25%), dry weight (~50%), survivorship (~25%), and general level of activity (Prasad and Joshi,
Prior to assays, all eight populations were reared under a common (control JB type) regime for one complete generation in order to ameliorate non-genetic parental effects. The eggs of these flies, hereafter referred to as standardized flies, were then used for the various assays.
A fresh food plate was introduced into the cages and the standardized flies were allowed to lay eggs for 1 h. This plate was then replaced by a second food plate and an egg collection window of 1 h was provided. Eggs used for the assay were collected from the second round of egg lay and the first sets of plates were discarded. This was done to ensure that the eggs used for the assay were developmentally synchronized. Eggs were collected with a moistened paint-brush from the food-surface, counted under the microscope, and exact numbers of eggs for the assay were dispensed into vials with 6 mL of food at a density of 30, 70, or 300 eggs per vial and incubated at three different temperatures namely 18, 25, and 28°C. 25°C and 70 eggs per 6 mL of food represent the normal maintenance conditions for FEJ and JB populations. The egg density of 30 eggs per vial was considered in order to study development at density substantially lower than what the populations normally experienced. Three hundred eggs in 6 mL, on the other hand, represents substantial larval crowding for these
Eight vials were set up for each combination of temperature, density, selection regime and replicate block (total 72 treatments). In all, 576 vials (2 selection regimes × 4 replicate blocks × 3 temperatures × 3 densities × 8 vials) were thus set up for the experiment. The vials were monitored for the first eclosion and, thereafter, checked regularly at 4 h intervals and the number of eclosing flies was recorded. The observations were continued till no new fly eclosed for two consecutive days. The development time in hours was measured by subtracting the time of egg-lay from the time of eclosion. In 4 out of the 72 treatments (2 selection regimes × 4 replicate populations × 3 temperatures × 3 densities), one vial was not viable, possibly due to handling error. For these sets, only data for 7 vials were available. To maintain a balanced design for the statistical analyses, for all 72 treatments the data from one vial, chosen at random, were excluded from the analyses. Hence, data from a total of 504 vials were considered for the final analyses.
From the vials used for the development time assay, the total number of eclosed flies were counted, and the number was divided by the number of eggs (30, 70, 300) for the respective vial to obtain the pre-adult survivorship for any given vial. Similar to development time, survivorship data from a total of 504 vials were considered for analyses.
Being large and outbred populations, there is considerable variation for development time, especially within the JB, and once eclosion starts for a given population, it roughly continues for 24–48 h. A temperature of 25°C and density of ~70 eggs per vial represent the standard developmental condition of the JB and FEJ populations. Under such conditions, FEJ flies take 6–7 days to complete their development from egg to eclosion, whereas JB flies take 9–10 days to complete pre-adult development. For the development time data analysis, flies eclosing after 100 h (over 4 days) from the first eclosion for any given set (combination of particular selection regime, temperature, density and replicate population) were not included in the development time data analysis for 30 and 70 egg densities. For 300 egg density, flies eclosing after 200 h (over 8 days) from the first eclosion for any given set were not included in the development time analysis. The criteria for defining outliers, typically few in number and representing potentially pathological individual variants, was decided upon before the experiment, based on past experience of development time distributions in these populations. Ultimately, outlier removal had to be done for only 6 of the 72 treatment combinations and essentially implied the exclusion of one or two extremely late developing flies as outliers in any given vial. For survivorship data, all eclosed flies, including the outliers defined as above, were considered for the analysis as inclusion of these flies did not greatly increase the mean or variance for survivorship.
For both pre-adult development time and survivorship, we examined the mean trait value, and variation in trait values, in each of the eight populations in all nine combinations of rearing density and temperature. We quantified plasticity or macroenvironmental sensitivity of a trait as the relative change in its value from one assay environment to another. For a given selection regime, less plasticity of mean trait value across assay environments would indicate greater macroenvironmental canalization for a trait.
For comparing trait variation across selection regimes and environments, we used the
Nested mixed-model analyses of variance (ANOVA) were performed on all data, using a completely randomized block design in which selection regime, temperature and density were treated as fixed factors, crossed among themselves, and with random blocks, representing both ancestral lineage and coincident handling of one matched pair of replicate selected and control populations during assays. Vial was treated as a random factor nested within the combinations of block and all three fixed factors. For analyzing across-environment variation in mean trait values, we used vial mean values for development time and vial values for pre-adult survivorship as the input data for the analysis. For comparing development time variation across individuals, the values of CV across individuals within each vial were used as input data for ANOVA. In addition, the across vial variability in mean development time for all combinations of block, selection regime, temperature, and density was also subjected to ANOVA. For survivorship, only vial means could be scored given the nature of the trait, and CV for each vial could not be calculated. Hence, only the across vial variability was used as a measure of survivorship variability for any population and treatment. All analyses were implemented using the software JMP (SAS institute) version 14. For
Trait variation observed in different combinations of temperature and density in the eight populations was also subjected to four-way mixed model ANOVA. Variation among individuals within a vial can reflect (i) genetic variation among individuals, (ii) developmental instability, which is hard to distinguish from (iii) variation due to possible ultra-microenvironmental heterogeneity within a vial, in addition to (iv) stochastic variation in trait measurement. Partitioning the within vial variation in phenotype among individuals into components caused by these four factors cannot be accomplished. However, in the context of our experimental design, if different combinations of the macroenvironmental factors, temperature, and density, affect the extent of within vial variation in phenotype among individuals, it can be cautiously interpreted as a change in developmental instability and/or ultra-microenvironmental sensitivity, given that the macroenvironmental change is unlikely to affect genetic variation or stochastic error. Incidentally, the one time the FEJ and JB populations were examined for developmental instability for any trait, no significant differences were observed (Shakarad et al.,
Across vial variation for development time reflects the effects of sampling, since there is within vial variation among individuals, overlaid by effects of microenvironmental heterogeneity, if any. There is evidence for such among vial microenvironmentally induced variation in life-history traits in
Data on pre-adult survivorship were treated exactly as described above for development time, except that mean pre-adult survivorship for each population was calculated by averaging the per vial pre-adult survivorship, followed by an arcsin square-root transformation. Since each vial yielded only one pre-adult survivorship value, only among vial variation was examined. The inference of macroenvironmental and microenvironmental sensitivity changes in populations subjected to different combinations of temperature and density was exactly the same as for development time.
Mean development time of FEJ is significantly less (~75%) than that of JB [
Mean pre-adult development time of FEJ and JB populations. The upper panel shows mean trait value plotted across temperature for each treatment density (e/V = eggs per vial). The lower panel shows mean trait value plotted across egg density for each treatment temperature. The error bars show 95% confidence intervals calculated from the variation among replicate populations in each combination of selection regime, temperature, and density.
Results of ANOVA on mean development time.
Selection | 1 | 5372.666 | <0.0001 |
Density | 2 | 219.1276 | <0.0001 |
Selection × density | 2 | 81.8427 | <0.0001 |
Temperature | 2 | 2215.963 | <0.0001 |
Selection × temperature | 2 | 1358.630 | <0.0001 |
Density × temperature | 4 | 11.3361 | 0.0005 |
Selection × density × temperature | 4 | 14.0372 | 0.0002 |
In both FEJ and JB, mean development time changes significantly across temperature for any given density but both JB and FEJ show similar changes across temperature. Mean development time in JB and FEJ increases by 68 and 67%, respectively, from 25 to 18°C, averaged across all densities (
To summarize, overall, mean development time is more plastic along the temperature axis than the density axis (
The standard deviation of (vial mean) development time across vials is significantly lower in FEJ than in JB [
Results of ANOVA on standard deviation of development time across vials.
Selection | 1 | 15.5816 | 0.0290 |
Density | 2 | 4.2555 | 0.0707 |
Selection × density | 2 | 2.9859 | 0.1259 |
Temperature | 2 | 25.9890 | 0.0011 |
Selection × temperature | 2 | 1.7682 | 0.2491 |
Density × temperature | 4 | 1.8152 | 0.1908 |
Selection × density × temperature | 4 | 0.8867 | 0.5009 |
Results of ANOVA on CV (coefficient of variation) of development time across vials.
Selection | 1 | 6.6807 | 0.0814 |
Density | 2 | 2.4583 | 0.1660 |
Selection × density | 2 | 1.6794 | 0.2635 |
Temperature | 2 | 0.2419 | 0.7924 |
Selection × temperature | 2 | 0.1603 | 0.8554 |
Density × temperature | 4 | 1.7471 | 0.2044 |
Selection × density × temperature | 4 | 0.4901 | 0.7432 |
Coefficient of variation (CV) of mean pre-adult development time of FEJ and JB populations across vials. The upper panel shows CV of the trait plotted across temperature for each treatment density (e/V = eggs per vial). The lower panel shows CV of the trait plotted across egg density for each treatment temperature. The error bars show 95% confidence intervals calculated from the variation among replicate populations in each combination of selection regime, temperature, and density.
Overall, the among individual within vial CV in development is significantly less in FEJ compared to JB [
Results of ANOVA on CV of development time across individuals, within vials.
Selection | 1 | 16.7120 | 0.0265 |
Density | 2 | 397.0883 | <0.0001 |
Selection × density | 2 | 24.9087 | 0.0012 |
Temperature | 2 | 25.8835 | 0.0011 |
Selection × temperature | 2 | 0.7155 | 0.5264 |
Density × temperature | 4 | 11.8821 | 0.0004 |
Selection × density × temperature | 4 | 0.3931 | 0.8097 |
Results of ANOVA for mean survivorship.
Selection | 1 | 133.7525 | 0.0014 |
Density | 2 | 135.7491 | <0.0001 |
Selection × density | 2 | 6.6117 | 0.0304 |
Temperature | 2 | 0.9151 | 0.4499 |
Selection × temperature | 2 | 5.6709 | 0.0414 |
Density × temperature | 4 | 1.3094 | 0.3213 |
Selection × density × temperature | 4 | 0.7017 | 0.6056 |
Coefficient of variation (CV) of pre-adult development time of FEJ and JB populations across individuals within a vial. The upper panel shows CV of the trait plotted across temperature for each treatment density (e/V = eggs per vial). The lower panel shows CV of the trait plotted across egg density for each treatment temperature. The error bars show 95% confidence intervals calculated from the variation among replicate populations in each combination of selection regime, temperature, and density.
The pattern of macroenvironmental effects on mean pre-adult survivorship in the FEJ and JB is not as consistent as in the case of development time. Overall, mean survivorship in FEJ was less than in JB [
Mean pre-adult survivorship of FEJ and JB populations. The upper panel shows mean trait value plotted across temperature for each treatment density (e/V = eggs per vial). The lower panel shows mean trait value plotted across egg density for each treatment temperature. The error bars show 95% confidence intervals calculated from the variation among replicate populations in each combination of selection regime, temperature, and density.
Mean survivorship in both JB and FEJ is significantly reduced at a density of 300 eggs per vial compared to either of the two lower densities (
CV of survivorship across vials is consistently and substantially higher than that of development time for both JB and FEJ populations (
Coefficient of variation (CV) of pre-adult survivorship of FEJ and JB populations across vials. The upper panel shows CV of the trait plotted across temperature for each treatment density (e/V = eggs per vial). The lower panel shows CV of the trait plotted across egg density for each treatment temperature. The error bars show 95% confidence intervals calculated from the variation among replicate populations in each combination of selection regime, temperature, and density.
Plot of vial values of survivorship (y-axis) vs. vial mean values of development time (x-axis) for FEJ and JB populations across all treatments. Each circle, or triangle, represents value or mean obtained from one vial.
Results of ANOVA for standard deviation of survivorship across vials.
Selection | 1 | 0.4120 | 0.5666 |
Density | 2 | 44.5447 | 0.0003 |
Selection × density | 2 | 5.4391 | 0.0449 |
Temperature | 2 | 0.2040 | 0.8209 |
Selection × temperature | 2 | 0.5221 | 0.6180 |
Density × temperature | 4 | 0.2232 | 0.9202 |
Selection × density × temperature | 4 | 0.7778 | 0.5606 |
Results of ANOVA for CV of survivorship across vials.
Selection | 1 | 1.5481 | 0.3018 |
Density | 2 | 16.3390 | 0.0037 |
Selection × density | 2 | 7.0492 | 0.0266 |
Temperature | 2 | 0.0853 | 0.9193 |
Selection × temperature | 2 | 1.1776 | 0.3703 |
Density × temperature | 4 | 0.0821 | 0.9864 |
Selection × density × temperature | 4 | 0.9375 | 0.4750 |
Thus, CV of pre-adult survivorship across vials is affected more strongly by changes in density than temperature (
The structure of our experiment permits us to make several pair-wise comparisons of the effects of long-term directional selection on the sensitivity of fitness-related traits to micro- and macroenvironmental effects. Before drawing these contrasts, and discussing some of the implications for our understanding of the relationship between selection and canalization, we first briefly summarize the patterns seen in our results when we examined pre-adult development time and pre-adult survivorship in the selected populations (FEJ) and their ancestral controls (JB). It should also be noted that though both development time and survival to adulthood are fitness components, it is development time that has the higher correlation with fitness in the FEJ selection regime, with pre-adult survivorship being a secondary fitness correlate.
In terms of the macroenvironmental effects of rearing temperature and density on mean development time in the FEJ and JB, temperature had greater effects than density on mean development time (
Like temperature, larval density is a major macroenvironmental variable affecting pre-adult development time in
It is important to note that the effect of temperature on individual variation for development time, and how the degree of variation changed from one thermal environment to another, varied widely across replicate populations. Thus, neither did the thermal plasticity of development time evolve in FEJs compared to JBs, nor was the extent of trait variation across temperature consistent from one replicate population to another within a selection regime. All of these indicate an overall lack of canalization for development time against macroenvironmental temperature changes in our study populations. This is in contrast with density, where changes in trait variation across treatments were consistent across blocks, and also, FEJ mean development time is more canalized against high density than JBs.
Regarding the sensitivity of FEJ and JB development time to microenvironmental variation
Turning now to sensitivity of the pre-adult survivorship, we observe that macroenvironmental effects of rearing temperature and density on pre-adult survivorship, a correlated response to selection, in the FEJ and JB reveal a pattern different from that seen for pre-adult development time, the trait under direct selection in FEJ populations (
For pre-adult survivorship variation
In the FEJ the nature of selection for development time was directional, at least for the first few hundred generations, although the selection response did eventually plateau (A. Joshi, pers. obs.). While stabilizing selection has been predicted to result in canalization of a trait, directional selection is often thought to lead to decanalization (Schmalhausen,
Pre-adult survivorship, a trait that is not under direct selection in our populations but has undergone correlated reduction alongside development time in the faster developing populations, did not show any clear trends for macro- or microenvironmental canalization against either temperature and density.
Overall, then, our results underscore that whether or not canalization results from directional selection may often depend on the trait in question, the environmental factor under study, the definition of canalization, and specific details of the ecology and physiology of the populations studied. Some other recent findings also suggest that directional and disruptive selection may not always lead to decanalization as predicted by theory (Hansen et al.,
In our study, we focused on classic life-history traits that are fitness correlates. Historically, though, studies of canalization have typically focused on morphological traits. Waddington (
Our study is exploratory in nature and it sheds light on a number of aspects of the relationship between selection, canalization and plasticity. We demonstrate that (a) a quantitative trait (development time) can evolve changes in its mean value without any change in its plasticity (e.g., thermal plasticity), (b) canalization can evolve for a life-history trait (development time) in laboratory populations within a span of few hundred generations, (c) a trait may show canalization both at macro- and microenvironmetal levels due to same underlying mechanisms, (d) contrary to expectations, directional selection can lead to environmental canalization (against developmental density or crowding) of a life history trait, as a byproduct of the evolved changes in the life history and physiology of the population, and, (e) a trait can be canalized for one environmental factor (density) but may not be canalized for another (temperature). The last two points show that canalization, like many other biological phenomena could be context specific, and one needs to be cautious before drawing generalized conclusions about the causes and manifestations of canalization. Canalization needs to be viewed as a phenomenon controlling trait variation across different hierarchical levels in biology, starting from cellular molecules to complex traits. However, both proximate and ultimate causes of canalization remain poorly understood, and a concerted effort of theoretical and empirical studies are needed to understand this fascinating phenomenon. And last but not the least, this study also demonstrates that experimental evolution can be a powerful tool to understand the intricacies of the evolutionary process, with particular focus on trait variation.
The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.
SG and AJ conceived of the experiments. SG, KS, and MJ carried out the experiments. SG did the data analysis. SG and AJ wrote the manuscript, and both SG and AJ carried out subsequent editing of the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank Archana Mohan for handling data, Ananda T. for help in the experiment and N. Rajanna and M. Manjesh for general help in the laboratory. SG was supported by a DST WOS A fellowship [SR/WOS-A/LS-1179/2015(G)] from the Department of Science & Technology, Government of India, during the preparation of the manuscipt, and by a fellowship from JNCASR during the experimental work. KS thanks Council of Scientific and Industrial Research, Government of India, for a senior research fellowship. The experimental work was supported in parts by funds from the Department of Science and Technology, Government of India, and the preparation of the manuscript by a J.C. Bose National Fellowship of the Science and Engineering Research Board, Government of India, to AJ. We thank the reviewers for comments on earlier drafts of this manuscript.