IN A VARIABLE THERMAL ENVIRONMENT SELECTION FAVORS GREATER PLASTICITY OF CELL MEMBRANES IN DROSOPHILA MELANOGASTER

Authors


Abstract

Theory predicts that developmental plasticity, the capacity to change phenotypic trajectory during development, should evolve when the environment varies sufficiently among generations, owing to temporal (e.g., seasonal) variation or to migration among environments. We characterized the levels of cellular plasticity during development in populations of Drosophila melanogaster experimentally evolved for over three years in either constant or temporally variable thermal environments. We used two measures of the lipid composition of cell membranes as indices of physiological plasticity (a.k.a. acclimation): (1) change in the ratio of phosphatidylethanolamine (PE) to phosphatidylcholine (PC) and (2) change in lipid saturation (number of double bonds) in cool (16°C) relative to warm (25°C) developmental conditions. Flies evolved under variable environments had a greater capacity to acclimate the PE/PC ratio compared to flies evolved in constant environments, supporting the prediction that environments with high among-generation variance favor greater developmental plasticity. Our results are consistent with the selective advantage of a more environmentally sensitive allele that may have associated costs in constant environments.

How organisms maintain performance across heterogeneous environments remains a central question in evolutionary biology. Theory predicts that specialists will evolve in constant environments with performance maximized at the mean environment experienced by the population (Levins 1968; Lynch and Gabriel 1987; Futuyma and Moreno 1988; Gilchrist 1995). Antagonistic pleiotropy and mutation accumulation, however, can cause negative genetic correlations in fitness across variable environments that prevent the evolution of universally superior generalists that maximize fitness in every environment (Reboud and Bell 1997; Poisot et al. 2011). In contrast, variable environments should favor the evolution of generalists that maintain performance across environmental gradients, but at the cost of lower peak performance than specialists (Levins 1968; Lynch and Gabriel 1987; Gilchrist 1995). Plasticity, the capacity for a single genotype to express different character states in different environments, results in a reaction norm (Schmalhausen 1949). When genetic variation for this reaction norm exists within a population (VG×E), plastic life historical, behavioral, and physiological strategies that enable organisms to maintain performance and fitness across temporally and spatially heterogeneous landscapes can evolve.

The evolution of plasticity (i.e., optimal reaction norms) has been modeled using both quantitative genetic (Via and Lande 1985, 1987; de Jong 1990; Van Tienderen 1991; Gomulkiewicz and Kirkpatrick 1992; Gavrilets and Scheiner 1993; Van Tienderen 1997) and optimality (Gabriel and Lynch 1992; Gabriel et al. 2005) approaches. Quantitative genetic models indicate that plasticity emerges as a byproduct of selection toward different phenotypic optima favored in distinct environments, when character states across environments are partially genetically uncorrelated (Via and Lande 1985; Gomulkiewicz and Kirkpatrick 1992; Via 1993). Models of optimal plasticity predict that in fine-grained environments where the within-generation component of environmental variation exceeds the among-generation component, selection favors reversibly labile characters that enable an individual to dynamically match performance to the environment within their lifetime (Gabriel et al. 2005). In coarse-grained environments where the among-generation component of environmental variation exceeds the within-generation component, developmental (or irreversible) plasticity should evolve that maximizes performance at the mean environment experienced during development (Gabriel and Lynch 1992). Developmental plasticity enables specialization within a generation in environments that vary among generations.

Homeoviscous adaptation, the ability of cells to plastically alter the lipid composition of membranes when temperatures change, is a near ubiquitous acclimation response (Sinensky 1974; Hazel 1995; Hochochka and Somero 2002). In ectotherms, membrane function must be maintained in the face of changing environmental temperatures. Such changes impact membrane fluidity via effects on the motion of the fatty acyl chains of glycerophospholipids, the major components of biological membranes (Hazel 1995; Hochochka and Somero 2002). Low temperatures slow the motion of acyl chains and lead to a more rigid membrane. At higher temperatures, this motion increases, the membrane becomes more fluid, and eventually forms lipid aggregates. In response, the lipid composition of membranes changes to both adapt to the local thermal environment and to acclimate in response to thermal variation within the life span of an organism (Cossins and Prosser 1978; Hazel 1995; Hochochka and Somero 2002; Angilletta 2009). Drosophila membranes are composed primarily of the glycerophospholipids phosphatidylethanolamine (PE) and phosphatidylcholine (PC) (Jones et al. 1992). Changes in the PE/PC ratio and the degree of saturation of their fatty acyl chains (i.e., the number of double bonds) are predicted to maintain membrane fluidity across heterogeneous thermal environments by altering the strength of van der Waals forces within the lipid bilayer of membranes. Although Drosophila maintains a baseline capacity for developmental acclimation of membrane physiology (Ohtsu et al. 1998; Overgaard et al. 2008), whether populations of D. melanogaster experience selection on the capacity to developmentally acclimate the composition of membrane lipids as predicted by theory remains unknown.

Using replicated populations of D. melanogaster that have been experimentally evolved in constant versus a variable thermal environment, we show that populations evolved in environments with high among-generation variance in temperature have a greater capacity to acclimate the lipid composition of membranes than populations evolved in constant environments. We discuss what types of genetic effects may underlie the evolution of greater cellular plasticity in an environment with high among-generation variance in temperature.

Methods

EXPERIMENTALLY EVOLVED POPULATIONS

The experimentally evolved populations characterized in our experiment were generated by and are described in detail by Yeaman et al. (2010). Female D. melanogaster were captured in September 2005 from an orchard near Cawston in the Similkameen Valley, British Columbia, Canada. Descendents representing genetic variation from 298 isofemale lines derived from these females were used to found a large breeding population that was allowed to grow for six generations reaching a size of approximately 64,000 adults. After nine generations in the lab, experimental populations were created by allowing the breeding population to lay eggs in bottles. Eight bottles were distributed into each population cage and two cages were then randomly assigned to each of five replicate populations within each selective environment. Populations were placed in either 16°C or 25°C according to their selective environment.

In our experiment, we evaluated flies from five replicate populations within each of three of the selective environments described by Yeaman et al. (2010): (1) 16°C constant (C), (2) 25°C constant (H), and (3) a temporally variable environment where flies were moved between 25°C and 16°C every four weeks (T). In the T selective environment, the thermal environment varied more among than within generations enabling us to test the theoretical prediction that this type of environmental heterogeneity selects for greater developmental plasticity (Gabriel and Lynch 1992; Gabriel et al. 2005). Comparisons between the two constant selective environments test additional evolutionary predictions about physiological specialization. Each replicate population consisted of two population cages containing eight bottles with four bottles exchanged between paired cages every four weeks. Population sizes were between 4000 and 8000 adults per population, but on rare occasions as few as 800 adults were observed in a cage (Yeaman et al. 2010). A new generation began every two weeks for cages at 25°C and every four weeks for cages at 16°C. Thus, flies spent more generations at 25°C than at 16°C in the thermally variable environment. Flies in our experiment evolved in the selective environments for over three years with overlapping generations (32 generations at 16°C, 64 generations at 25°C, and an intermediate number of generations for the variable selective environment).

In August 2009, we sampled genotypes from each of five replicate populations within each of these three selective environments. For two generations, one virgin female was mated to one virgin male sibling within each line to establish isofemale lines that could be maintained in common environments without further evolution. Fifteen isofemale lines within each population were maintained in our lab in vials that contained standard Bloomington Drosophila cornmeal-yeast medium at 20.5°C on a 12:12 light cycle and transferred every three weeks for 27 months prior to this experiment. Establishing isofemale lines isogenizes the genome within each line, minimizes further laboratory evolution, and enables us to control for cross-generational effects.

EXPERIMENTAL TREATMENTS

We evaluated the degree of cellular plasticity in isofemale lines from each of the 15 populations. Density was controlled for two generations prior to the experiment by only transferring five inseminated females into fresh vials. During the experimental generations, zero- to three-day-old mated females from each line were placed into yeasted food vials. Three days later, 20 females from these vials were split between two yeasted vials and allowed to lay eggs for 24 h at 20.5°C. Vials of eggs from each line were then randomly placed into either 16°C or 25°C constant thermal environments with a 12:12 h light cycle for development. We completed our experiment in four blocks with an equal number of lines from each population placed in each developmental treatment at the same time. In total, we reared offspring from five to seven isofemale lines from each of five replicate populations within each selective environment (C, H, and T) in two developmental environments (16°C and 25°C).

EXTRACTION, SEPARATION, AND QUANTIFICATION OF GLYCEROPHOSPHOLIPIDS

Lipids were extracted from two- to four-day-old male flies from each line after developing at 16°C and 25°C using established protocols (Folch et al. 1957; Kostal and Simek 1998). Pools of 20 male flies were weighed and homogenized using a Dounce pestle (Wheaton, Millville, NJ) in glass test tubes containing 0.5 mL of ice-cold chloroform–methanol extraction solution (2:1). Homogenates were centrifuged at 5000 g for 10 min at 4°C, and the supernatant was collected. This process was repeated on the pellet, and both supernatants were combined with 0.2 mL of 0.9% NaCl. Samples were vortexed for 5 sec and centrifuged one final time. The lower organic phase was collected, dried under a stream of nitrogen, and stored at –80°C pending analysis. This method extracts membrane lipids, as well as other lipids such as triacylglycerides (Hammad et al. 2011). Our analyses focus on glycerophospholipids known to be major components of the cell membrane.

We separated lipids by liquid chromatography followed by tandem mass spectrometry (LC-MS/MS) using a method optimized to quantify membrane lipids in D. melanogaster and described in detail by Hammad et al. (2011). The method uses internal PE and PC standards to account for biased LC-MS detection of glycerophospholipids with different head groups. Dried samples were reconstituted in 1 mL methanol and further diluted by a factor of 50 in methanol. The internal standards 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine and 1,2-dimyristoleoyl-sn-glycero-3-phosphocholine were added to the reconstituted samples with a final concentration of 400 pg/μl. Samples were analyzed using a Dionex 3000 Ultimate LC system (Dionex, Sunnyvale, CA) interfaced to a QTRAP 4000 triple quadrupole instrument (ABI Sciex, Foster City, CA). A 1-μl aliquot of the reconstituted lipid extract was injected onto a Kinetex C18 column (Phenomenex, Torrance, CA) (100 mm × 2.1 mm, 2.6-μm particle size) maintained at 25°C. Mobile phase A consisted of 10 mM ammonium acetate in H2O:methanol (10%:90%, v:v). Mobile phase B consisted of 10 mM ammonium acetate in isopropanol:methanol (50%:50%, v:v). The gradient conditions were 10% B to 65% B from 3 to 10 min followed by 3 min at 100% B.

The PE and PC glycerophospholipids that we quantified are listed in Table S1. For each sample, we evaluated the proportion (percent abundance) of each PE and PC glycerophospholipid in the total pool of PE and PC glycerophospholipids by converting analyte peak areas into mole fraction percentages (Overgaard et al. 2008). These percentages were corrected using the peak areas of the PE and PC internal standards in each sample. For each replicate line, we quantified the ratio of PE to PC and also the proportion of saturated glycerophospholipids in the pool of membrane lipids after development at 16°C and 25°C. The ratio of PE to PC (Ratio) is the proportion of the membrane composed of PE relative to the proportion of the membrane composed of PC. The amount of saturation was defined as the number of double bonds summed across the two acyl chains of each glycerophospholipid, and quantified as a proportion of the total membrane lipid pool (onebond, twobonds, threebonds, and fourbonds). The capacity for developmental acclimation of each line is the difference in the PE/PC ratio (ΔRatio) or in each of the saturation phenotypes (Δonebond, Δtwobond, Δthreebond, and Δfourbond) between the two developmental temperatures. Dividing by 9°C would give the slope of the reaction norm.

STATISTICAL ANALYSES

We tested three hypotheses. First, that developmental temperature will affect membrane lipid composition irrespective of selective environment, which would indicate a general acclimation response. Second, that populations evolved in the environment with high among-generation variance (T) will have a greater capacity to acclimate the lipid composition of membranes than those evolved in constant C and H environments (i.e., a steeper reaction norm or greater plasticity). Third, that flies evolved in constant C and H environments will have diverged in their overall lipid membrane composition, indicative of physiological specialization.

To test these hypotheses, we used mixed model analyses of variance using the nlme library in the software package R version 2.12.1 (Team 2008). We competed statistical models using Akaike information criterion (Burnham and Anderson 2002) to determine the best structure for random and fixed components using the statistical approach of Zuur et al. (2009). We used REML estimators to determine the optimal structure of the random component, followed by ML estimators to find the optimal structure for the fixed component. Then we used REML to estimate the parameters of the best model as determined in the previous steps. To test for acclimation responses, we evaluated the fixed effects of selective environment, development environment, block, their interactions, and the random effect of population on the Ratio, onebond, twobond, threebond, and fourbond phenotypes. To test for a greater capacity to acclimate, we evaluated the fixed effects of selective environment, block, the average weight of flies across development temperatures, their interactions, and the random effect of population on the ΔRatio, Δonebond, Δtwobond, Δthreebond, and Δfourbond phenotypes. To test for physiological specialization, we evaluated the fixed effects of the C and H selective environments, block, weight, their interactions, and the random effect of population on Ratio within developmental temperatures.

Results

PATTERNS OF DEVELOPMENTAL ACCLIMATION

Across selective environments, flies display a strong acclimation response to developmental temperature (Figs. 1A, 2; Table 1). Development at 16°C significantly increases the PE/PC ratio (Fig. 1A; Table 1). Developmental temperature also affects the degree of lipid saturation, with the exception of twobonds (Fig. 2; Table 1). Glycerophospholipids with less-saturated acyl chains (threebonds and fourbonds) make up a significantly greater percentage of membrane lipids after development at 16°C, whereas more saturated chains (onebond) make up a significantly greater percentage of membrane lipids after development at 25°C (Fig. 2). Both acclimation responses are in the direction predicted to maintain fluidity at each developmental temperature.

Figure 1.

The capacity to acclimate the phosphatidylethanolamine (PE)/phosphatidylcholine (PC) ratio is greater in flies evolved in a variable thermal environment. (A) Flies from all selective environments (T, triangles; C, squares; H, circles) strongly acclimate the PE/PC ratio in response to the developmental environment. (B) Populations evolved in the variable T environment have a greater capacity to acclimate this lipid ratio during development than flies evolved in constant C and H environments. The capacity to acclimate is calculated as the difference in the PE/PC ratio between developmental temperatures for each genetic line, and is our measure of developmental plasticity. Data are means ± SE for all genetic lines within each selective environment.

Figure 2.

The degree of saturation of fatty acyl chains acclimates in response to developmental temperature. The proportion of the membrane composed of glycerophospholipids with one (A), two (B), three (C), or four (D) double bonds is plotted as a function of the developmental environment. Asterisks denote significant effects of developmental environment (P < 0.0001). Data are for means ± SE for all genetic lines within T (triangles), C (squares), and H (circles) selective environments.

Table 1.  ANOVA evaluating the acclimation effect of development environment on cellular phenotypes.
Dependent variableFixed effect1num dfden df F-valueP-valueRandom effect2
  1. 1Only the fixed effects included in the best fitting model are reported.

  2. 2When the inclusion of a random effect significantly increased the fit of the model using AIC criterion.

RatioDevelopmental environment118276.132<0.0001Population
 Block11811.4620.2281 
onebondDevelopmental environment1174262.14<0.0001 
 Block31746.552<0.0001 
 Selective environment21741.6590.1932 
 Developmental environment × Block31744.0280.0084 
twobondsDevelopmental environment11713.2810.0719 
 Block317144.869<0.0001 
 Selective environment21712.0310.1343 
 Selective environment × Block61712.6920.016 
threebondsDevelopmental environment117174.224<0.0001 
 Block317116.762<0.0001 
 Selective environment21712.421<0.0919 
 Selective environment × Block61712.4780.0253 
fourbondsDevelopmental environment117960.511<0.0001 
 Block317922.277<0.0001 

EVOLUTION OF CELLULAR PLASTICITY

Flies that evolved in the variable thermal (T) environment evolved a significantly greater capacity to acclimate the PE/PC ratio during development (Fig. 1B; Table 2) relative to flies that evolved in constant C (Tukey's HSD, P= 0.001) and H (P= 0.031) environments. Flies evolved in the C and H environments do not differ (P= 0.504) in the degree of cellular plasticity (Fig. 1B). Importantly, models that include the random effect of population do not provide a significantly better fit to the data. The lack of a population effect indicates a consistent phenotypic response to selection across replicate populations, which was robust to the random effects of genetic drift.

Table 2.  ANOVA evaluating the effect of the selective environment on the capacity to acclimate cellular phenotypes.
Dependent variableFixed effect1num dfden dfF-valueP-value
  1. 1Only the fixed effects included in the best fitting model are reported. Including the random effect of population did not improve the fit of the model for any of these traits.

ΔRatioSelective environment2867.1910.0013
 Block3863.4030.0213
ΔonebondSelective environment2864.0390.0211
 Block3869.06<0.0001
ΔtwobondsSelective environment2834.02130.0215
 Block3833.65140.0158
 Average weight1831.44580.2326
 Selective environment × Average weight2832.72090.0717
ΔthreebondsSelective environment2892.270.1093
ΔfourbondsSelective environment2892.46080.0912

The greater degree of cellular plasticity in flies that evolved in the T environment results from a significant reduction in the PE/PC ratio after development at 25°C when compared to flies from either the C or H selective environments (Fig. 1A). When reared at 16°C, flies from all selective environments have the same mean PE/PC ratio, with the best fitting model only including the effect of block (F= 2.342, P= 0.079). However, when flies develop at 25°C, there is a significant effect of selective environment on the PE/PC ratio (F= 4.648, P= 0.012), as well as an effect of block (F= 5.734, P= 0.001), with no evidence for an interaction. At 25°C, flies evolved in the T environment have a significantly lower PE/PC ratio than do flies that evolved in constant C (Tukey's HSD, P= 0.026) and H (P= 0.025) environments. Flies evolved in the variable thermal environment have a stronger capacity to counter the effects of warm temperatures on membrane fluidity by increasing membrane stability through a decreased contribution of PE to membrane lipid composition.

Because PE destabilizes the membrane, a higher PE/PC ratio in flies evolved in constant 16°C relative to constant 25°C environments independent of developmental temperature would indicate the evolution of physiological specialization. However, flies evolved in constant thermal environments do not differ in the PE/PC ratio when developed at 25°C (Tukey's HSD, P= 0.997, Fig. 1A), and selective environment did not affect the PE/PC ratio after development at 16°C (F= 0.4479, P= 0.646). Flies evolved in constant environments have similar capacities for acclimating the PE/PC ratio (Fig. 1B) and nearly identical PE/PC ratios within developmental temperatures. Thus, physiological specialization of the contributions of PE and PC to membrane lipid composition did not evolve in constant environments (Fig. 1A).

Experimentally evolved populations differ in their capacity to acclimate the degree of lipid saturation, but flies evolved in the T environment do not generally have a higher capacity to acclimate this trait (Fig. 2; Table 2). Selective environment did affect the capacity to acclimate the relative abundance of lipids with a single double bond (Fig. 2A; Table 2). However, flies evolved in the T environment do not differ from those that evolved in H (Tukey's HSD, P= 0.62) or C (P= 0.174) environments. Rather, flies evolved in C and H environments differ from one another (P= 0.018). This was the only indication of physiological specialization in the data. All selective environments had the same capacity to acclimate the abundance of the highly unsaturated lipids containing three and four double bonds (Figs. 2C and D; Table 2). Thus, flies from variable and constant environments did not generally diverge in their capacity to acclimate the saturation of acyl chains.

Discussion

EVOLUTION OF CELLULAR PLASTICITY

We have shown that flies evolved in an environment with high among-generation variance have a greater degree of cellular plasticity than those that evolved in constant environments. Across all selective environments, membrane lipids acclimate via a decrease in the PE/PC ratio at warmer developmental temperatures and via changes to lipid saturation. These results support physiological predictions regarding the maintenance of membrane fluidity (Sinensky 1974; Hazel 1995). However, flies evolved in an environment with high among-generation variance in temperature have a higher capacity to acclimate the PE/PC ratio during development relative to flies evolved in constant environments. The steeper reaction norm exhibited by the flies evolved in a variable environment results primarily from a greater ability to decrease the PE/PC ratio during development at 25°C. In contrast, experimentally evolved flies do not differ in their capacity to acclimate lipid saturation, indicating that these two components of membrane lipid composition have different genetic architectures.

Our results complement recent findings that developmental plasticity has a selective advantage in a spatially heterogeneous environment. Populations of the common frog Rana temporaria that experience high migration between environments that differ in the probability of desiccation have evolved a higher degree of plasticity for developmental time (Lind et al. 2011). In contrast, previous studies of physiological plasticity during development in natural populations do not support the predictions of the theory (Gunderson et al. 2000; Cunningham and Read 2003; Ayrinhac et al. 2004; Hoffmann et al. 2005; Eggert et al. 2006; Cooper et al. 2010). The ability to predict when acclimation will evolve hinges on the estimation of the within- and among-generation components of environmental variation experienced by natural populations across evolutionary timescales (Cooper et al. 2010). Our attempts will fail if the selective environment is not properly defined, as might be the case if behavior buffers environmental variation. We have overcome this challenge by using experimental populations (sensu Bull and Wang 2010) for which we know the amount and type of variation in temperature experienced by the evolving populations. Our results demonstrate that genetic constraints do not preclude the evolution of optimal reaction norms for the PE/PC ratio that presumably canalizes cell function across developmental environments.

ADAPTATION AND ACCLIMATION OF CELLULAR MEMBRANES

Although the capacity to acclimate the PE/PC ratio diverged between flies evolved in constant versus variable thermal environments, flies in constant environments did not evolve physiological specialization of membrane lipid composition. Flies evolved in the constant C and H environments have remarkably similar levels of PE to PC in each developmental environment, in addition to identical reaction norms across developmental temperatures. Thus, even though individual flies have the capacity to acclimate the PE/PC ratio in response to developmental temperature, populations of flies evolved for more than three years at constant 16°C (32 generations) and 25°C (64 generations) have not diverged in the baseline proportions of PE and PC in the membrane.

The lack of divergence that we observed in the PE/PC ratio between the experimentally evolved C and H populations indicates that either genetic variation for changes in the baseline proportions of PE and PC did not exist in the ancestral population, selection was not intense enough to elicit a response in this amount of time, or other aspects of membrane function have diverged between these populations. We also evaluated how changes in lipid saturation might have diverged among warm- and cool-selected populations, but did not find general patterns for specialization of this trait in constant environments. Other factors (e.g., abundance of cholesterol) can affect the stability of biological membranes, and pathways involved in the regulation of these factors could have been targeted by selection. If so, flies evolved in constant environments may have diverged in membrane physiology in ways that our assay did not detect.

THE EVOLUTIONARY GENETICS OF ACCLIMATION

The ability to acclimate the architecture of cell membranes in response to developmental temperature has not been lost in populations that evolved in constant environments (i.e., the slopes of the reaction norms are not equal to zero). The shared baseline capacity for membrane acclimation across selective environments indicates that plasticity was present in the ancestral population and remains an important response for D. melanogaster (Overgaard et al. 2008), perhaps genetically correlated with the capacity to acclimate the membrane in response to cellular stressors other than temperature (e.g., Montooth et al. 2006). However, our data demonstrate that selection favors a greater capacity for acclimation in thermal environments that vary temporally across generations. Plasticity in the PE/PC ratio enables membrane fluidity to remain relatively constant across thermal gradients, which presumably maintains fitness across developmental temperatures by preventing the disruption of core cellular processes that underlie physiological function (Hazel 1995; Hochochka and Somero 2002). Because cell membrane physiology also changes in response to more rapid thermal shifts (Hochochka and Somero 2002), future studies of reversible responses will help determine whether genetic correlations exist between developmental and reversible plasticity at the level of the cell.

Developmental plasticity emerges from the action of natural selection acting on allelic variation independently in multiple environments, when character states are not perfectly genetically correlated across environments (Via and Lande 1985; Via 1993). Increases in cellular plasticity could arise through selection favoring alleles that regulate the pathways underlying membrane acclimation. Selection in variable environments should increase the frequency of alleles that confer higher fitness cellular phenotypes in each environment. The alleles may be environmentally specific and modify traits independently in different environments, or selection may favor alleles with greater environmental sensitivity that produce phenotypes that more closely match each environment (Via 1993; Via et al. 1995). In the absence of environmental heterogeneity, this second class of alleles may experience relaxed selection, and if they confer an associated cost of plasticity (Van Tienderen 1991) these alleles should decrease in frequency in populations experiencing a constant thermal environment.

Although several components of the membrane acclimated in response to developmental temperature, only the PE/PC ratio showed greater plasticity in populations that were evolved in the variable thermal environment. The enhanced capacity for acclimation results from a greater ability of these populations to reduce the PE/PC ratio at higher temperatures. If selection had favored alleles with environmental specificity that decrease the PE/PC ratio at 25°C, then we would have expected to see these alleles favored in the constant 25°C selective environment. Yet, we saw no evidence of this type of adaptive membrane response in the constant H populations. Instead our results indicate that the greater plasticity maintained by the thermally variable environment is conferred by alleles with greater thermal sensitivity to adjust lipid composition at 25°C, but with an associated cost that results in a decreased frequency of these alleles in populations that evolved in constant thermal environments (Van Tienderen 1991).

Although the capacity to acclimate the ratio of PE to PC consistently evolved across replicate populations, this does not indicate that the genetic response was necessarily the same. There may be many loci segregating alleles that potentially modify this ratio and these may follow different selection-drift trajectories in different replicate populations with a shared phenotypic outcome. An understanding of the genetic basis of these traits requires determining what underlying biochemical and transcriptional changes are leading to the phenotypic evolution of the PE/PC ratio, in addition to analyzing sequence divergence among selective environments. Homeoviscous adaptation has been well studied as a physiological response to the environment, but much remains unknown at the level of genetic regulation. The sterol regulatory element binding protein SREBP and the lipid-signaling enzyme Phospholipase D are good candidates for modifying the PE/PC ratio in response to temperature (Miller et al. 1993; Dobrosotskaya et al. 2002; Montooth et al. 2006). These candidates provide a good foundation to elucidate the genetic basis of enhanced cellular plasticity in variable environments.

Associate Editor: J. Kelly

ACKNOWLEDGMENTS

We would especially like to thank S. Yeaman and M. Whitlock for making their selection lines available. M. Angilletta Jr. assisted in generating the isofemale lines. C. Meiklejohn, M. Wade, M. Whitlock, and two anonymous reviewers provided comments that greatly improved our manuscript. D. Daleke, V. Kostal, J. Overgaard, and A. Postle provided advice on the extraction, measurement, and quantification of lipids. This research was funded by grants from the Society for Integrative and Comparative Biology and the Indiana Academy of Science to BSC, Indiana University to KLM, the Hutton Honors College to NPF, and the METACyt Initiative to LAH and JAK.

Ancillary