Constraints, independence, and evolution of thermal plasticity: Probing genetic architecture of long- and short-term thermal acclimation

Significance

Mitigating thermal stress through evolutionary adaptation or physiological plasticity is critical for species’ persistence in changing climates. Sparse knowledge of genetic and physiological architectures of thermal plasticity hampers our ability to predict organismal resilience to climate change. Understanding the independence of short- and long-term plasticity and constraints of basal thermotolerance on plasticity is important for understanding responses to climate change. We show heritable genetic variation for basal cold tolerance and plasticity in a midlatitude Drosophila melanogaster population. High long-term plasticity predicted high short-term plasticity, and basal cold tolerance constrained both plasticity measures. There was no overlap in SNPs associated with either plasticity type. Overlapping molecular function of SNPs suggests shared physiology between long- and short-term plasticity, despite distinct genetic architectures.

Abstract

Seasonal and daily thermal variation can limit species distributions because of physiological tolerances. Low temperatures are particularly challenging for ectotherms, which use both basal thermotolerance and acclimation, an adaptive plastic response, to mitigate thermal stress. Both basal thermotolerance and acclimation are thought to be important for local adaptation and persistence in the face of climate change. However, the evolutionary independence of basal and plastic tolerances remains unclear. Acclimation can occur over longer (seasonal) or shorter (hours to days) time scales, and the degree of mechanistic overlap is unresolved. Using a midlatitude population of Drosophila melanogaster, we show substantial heritable variation in both short- and long-term acclimation. Rapid cold hardening (short-term plasticity) and developmental acclimation (long-term plasticity) are positively correlated, suggesting shared mechanisms. However, there are independent components of these traits, because developmentally acclimated flies respond positively to short-term acclimation. A strong negative correlation between basal cold tolerance and developmental acclimation suggests that basal cold tolerance may constrain developmental acclimation, whereas a weaker negative correlation between basal cold tolerance and short-term acclimation suggests less constraint. Using genome-wide association mapping, we show the genetic architecture of rapid cold hardening and developmental acclimation responses are nonoverlapping at the SNP and corresponding gene level. However, genes associated with each trait share functional similarities, including genes involved in apoptosis and autophagy, cytoskeletal and membrane structural components, and ion binding and transport. These results indicate substantial opportunity for short-term and long-term acclimation responses to evolve separately from each other and for short-term acclimation to evolve separately from basal thermotolerance.

With the predicted climate change, ectotherms must compensate not only for changes in mean temperature but also for an increased frequency of extreme events (1, 2), including rapid shifts from relatively warm temperatures to cold temperatures, and vice versa, as weather fronts dictate (3, 4). The ability to mitigate thermal stress through either evolutionary adaptation or plasticity/acclimation is expected to be important for species’ persistence in the face of climate change (5, 6). Acclimation responses, a form of adaptive phenotypic plasticity, are particularly important for mitigating thermal stress in ectotherms. Although studies of plasticity often consider responses that occur over moderately long time scales—termed “developmental acclimation” (DACC)—acclimation responses can occur over both longer and very short time scales (7–10). Rapid cold hardening (RCH) is one such short-time-scale plastic response wherein ectotherms exposed to mild low temperatures for minutes to hours suffer less from subsequent exposure to lower temperatures (7, 8, 11, 12). Ectotherms can use this rapidly inducible and rapidly reversible form of acclimation to adjust their physiology with very short time lags, tracking daily thermal fluctuations and extending active temperatures, preventing cold-induced loss of reproductive performance, and even preventing cold-induced mortality (7–10, 12, 13).

Much work has focused on the physiological mechanisms and the ecological implications of both DACC and RCH, particularly to cold (14–19). Despite its ecological importance, the physiological and genetic mechanisms underlying RCH have been challenging to unravel using physiological, transcriptomic, or metabolomic techniques, particularly because the physiological responses are both fast and transient (20–25). Thus, there is substantial debate about how much the mechanisms of developmental and rapid thermal acclimation overlap (10, 14, 19, 26).

Another important question about the evolutionary potential of thermotolerance in the context of local adaptation and anthropogenic change is whether acclimation, either developmental or rapid, is constrained by an organism’s basal thermotolerance (27–31). Basal cold tolerance is defined by an organism’s survival after cold exposure without any precooling period or cold acclimation. Stillman (27) showed that basal heat tolerance and rapid heat acclimation were negatively correlated, or constrained, across populations of several species of Porcelain crabs along a latitudinal cline. However, Calosi et al. (29) found a positive relationship between heat tolerance and magnitude of acclimation among species of European diving beetles. Nyamukondiwa et al. (31) found that basal cold tolerance constrained rapid cold acclimation among Drosophila species in a phylogenetic context, but they found no relationship between basal heat tolerance and rapid heat hardening.

Some models of the evolution of thermal breadth suggest that decreasing basal tolerance to low temperatures can decrease phenotypic plasticity (32). Considering that one projected effect of climate change is increased thermal variability, reduced ability for organisms to predict future temperatures in the short or long term could select against plasticity and for increased basal tolerance, because inappropriate expression of thermal plasticity can be costly (3, 24, 30, 32, 33). Our knowledge about potential constraints to the evolution of thermal plasticity resulting from trade-offs with basal thermotolerance currently is limited to a few animal groups, and all these studies have focused on comparisons among species or among populations within a species. None of these studies considers genetic variation segregating within populations for both basal thermotolerance and acclimation, a substantial knowledge gap for understanding whether thermal plasticity can evolve separately from basal thermotolerance during local adaptation.

Here we use a large set of fully sequenced inbred lines from a single midlatitude population of the fly Drosophila melanogaster (34, 35) to address the genetic basis of both basal cold tolerance and acclimation responses. Specifically, we quantified basal cold tolerance, DACC, and RCH phenotypes and performed a genome-wide association study (GWAS) to identify candidate loci for each of these thermotolerance traits. The Drosophila melanogaster Genetics References Panel (DGRP) is a powerful system for assessing heritable genetic variation in acclimation and other forms of plasticity and patterns of evolutionary constraint on trait evolution and for identifying candidate loci underlying complex traits, because this panel of 192 lines contains substantial genetic variation that should reflect the breadth of thermal phenotypes segregating in this midlatitude population. The power of using a large panel of naturally segregating genotypes allows us to make general statements about evolutionary and ecological patterns at the whole-population level that we expect also will hold more broadly for other populations of temperate ectotherms.

Results and Discussion

Heritability of Basal Cold Hardiness and Acclimation.

We show substantial segregating variation and relatively high heritability for basal thermotolerance of both acute and chronic exposures, as well as both rapid acclimation and DACC capacity in this midlatitude population from Raleigh, NC, reinforcing the notion that there is much potential for both basal low-temperature tolerance and thermal plasticity to evolve in wild populations. Flies reared at 25 °C had substantial variation in basal thermotolerance across lines for both acute exposure to −6 °C for 1 h (heritability 0.15; Fig. 1 and Fig. S1A) and chronic exposure to 0 °C for 1 h (heritability 0.44; Fig. S3A). RCH treatment for 2 h at 4 °C increased survivorship after an acute exposure to cold for 1 h at −6 °C for flies reared at 25 °C across almost all lines (t₁₈₃ = 12.09, P < 0.001; RCH 25; Fig. S2A). To estimate the degree of plasticity induced by RCH treatment, we took the difference between basal levels of survival and survival after RCH as our reaction norm, hereafter termed “RCH capacity” (heritability 0.14). Thirty-four of the 184 DGRP lines showed decreased survival after RCH, yielding a negative RCH score (Fig. 1 and Fig. S1B). For these lines it appears that the initial RCH treatment for 2 h at 4 °C was itself stressful and demonstrates that not all genotypes benefit from RCH under these treatment conditions (32, 36). However, these genotypes are not poor survivors in general, because they survive acute cold stress slightly better than all other genotypes combined (t₃₀ = −2.50, P = 0.018; survivorship difference = −0.14).

Fig. 1.

RCH capacity for flies reared at 25 °C. (Upper) Reaction norms for survivorship proportion for values used to calculate RCH capacity: survivorship after 1 h at −6 °C and following pretreatment for 2 h at 4 °C before 1 h at −6 °C. (Left) Ten negative RCH capacity scores. (Center) Ten average RCH scores. (Right) Ten of the most positive RCH scores, i.e., most extreme improvement in survivorship following pretreatment. (Lower) The histogram shows the distribution of RCH acclimation capacity across the DGRP.

We followed a common protocol for DACC to cold by rearing flies at 18 °C throughout their lifecycle before testing them for basal cold tolerance and RCH capacity (26, 37). DACC increased survival after acute exposure to cold for 1 h at −6 °C compared with flies reared at 25 °C across all lines (t₁₈₂ = −2.92, P = 0.0039; Figs. S1C and S2A). To estimate the degree of plasticity, we used the difference between basal levels of survival after our acute cold stress for 1 h at −6 °C for flies reared at 25 °C and survival after the same acute cold stress after DACC at 18 °C as our reaction norm, hereafter termed “DACC capacity” (heritability 0.12, Figs. S1D and S2B). Physiological studies of both DACC and RCH suggest some shared mechanistic components, such as the use of cryoprotectants, membrane lipid composition, and some heat-shock proteins (9, 16, 21, 24). If mechanisms of short- and long-term acclimation overlap, we predict that DACC will reduce the degree of plasticity induced by RCH treatment.

Because DACC induced large increases in cold hardiness across almost all lines, we tested each developmentally acclimated line for RCH capacity at a lower acute exposure temperature of −8 °C for 1 h rather than −6 °C (heritability 0.22; Fig. S1E). Exposing lines that had undergone DACC to RCH-inducing conditions of 2-h exposure at 4 °C substantially increased survivorship to an acute cold exposure for 1 h at −8 °C (t₁₈₄ = −34.99, P < 0.001). RCH ability had heritability of 0.17 across lines that already were developmentally acclimated, very similar to the estimates of heritability for RCH at 25 °C and for DACC (Figs. S1F and S2 A and C). When considering scores for RCH capacity, only two lines that were acclimated throughout development showed slight negative effects of RCH on acute cold survival, whereas 34 lines showed negative effects of RCH treatment when reared at 25 °C throughout the lifecycle. Thus, DACC was synergistic with RCH in improving survival of acute cold, further suggesting that there is some separation in the genetic and physiological basis of these two routes to acclimation (19). In addition, the magnitude of the RCH response when flies were not developmentally acclimated (0.20 ± 0.46, mean ± SD) was substantially smaller than the magnitude of either the DACC response (0.44 ± 0.40, P < 0.001) or the RCH response when flies already were developmentally acclimated (0.45 ± 0.44, P < 0.001). The greater magnitudes of response in developmentally acclimated flies suggest that, as time available for acclimation increases (i.e., during DACC), environmental cues activate additional suites of mechanisms, significantly increasing the magnitude of protection provided by thermal plasticity.

Basal Hardiness Constrains Plasticity.

Whether acclimation, developmental or rapid, is constrained by an organism’s basal thermotolerance is a matter of debate with implications for understanding the evolution of acclimation responses (27–31). If basal hardiness constrains either short- or long-term plasticity, we expect a negative correlation between basal hardiness and scores for acclimation capacity across genotypes. DACC was very strongly negatively correlated with basal cold hardiness for both chronic and acute cold exposures (Table 1 and Figs. S3B and S4A). Similarly, for flies developmentally acclimated by rearing at 18 °C, RCH capacity also was negatively correlated with survival of acute cold exposure, although the correlation was not as strong as the correlation between DACC and survival of acute cold exposure (Table 1 and Fig. S4B). There was only a slight negative correlation between the RCH capacity of developmentally acclimated flies reared at 18 °C and basal hardiness to chronic cold exposure for flies reared at 25 °C (Table 1 and Fig. S3C). The RCH capacity of lines reared at 25 °C was not correlated with chronic basal cold hardiness (Table 1 and Fig. S3D) but was negatively correlated with acute basal tolerance (Fig. S4C). The negative correlation between basal cold tolerance and DACC capacity was substantially stronger than the negative correlation between basal cold tolerance and RCH, suggesting that basal cold tolerance places greater constraints on long-term plasticity, because DACC provides the opportunity for individuals to alter their physiology more dramatically than does a short-term (RCH) treatment (Table 1). All correlations also were tested without genotypes that reached 100% survival following acclimation to ensure that these genotypes were not artificially driving observed patterns, and the sign and magnitude of correlations remained for all trait interactions (Table S1).

Table 1.

Correlations for traits of physiological interest

Trait	Correlation	df	P value
Acute basal tolerance
vs. DACC	−0.98	181	<0.001
vs. RCH at 18 °C	−0.85	180	<0.001
vs. RCH at 25 °C	−0.56	182	<0.001
Chronic basal tolerance
vs. DACC	−0.46	172	<0.001
vs. RCH at 18 °C	−0.15	172	0.042
vs. RCH at 25 °C	0.011	173	0.87
DACC vs. RCH at 25 °C	0.55	181	<0.001
DACC vs. RCH at 18 °C	0.012	178	0.86
RCH at 25 °C vs. RCH at 18 °C	0.045	178	0.54

RCH and DACC are the capacity values calculated from survivorship following acclimation and survivorship without acclimation. The acute survivorship for RCH at 25 °C was calculated at −6 °C for 1 h and for RCH after DACC at 18 °C was calculated at −8 °C for 1 h.

If DACC and RCH induce the same underlying mechanisms to produce greater cold tolerance, we expect that lines with high capacity for DACC will also have high capacity for RCH (17, 19, 26). We do find that DACC and RCH capacity are positively correlated with each other, suggesting functional similarities between DACC and RCH (Table 1 and Fig. S5A). However, there still is variation among lines in the relationship between DACC capacity and RCH capacity, suggesting that mechanisms are not completely overlapping. Furthermore, if there were complete mechanistic overlap between DACC and RCH, we would expect a negative correlation between DACC capacity and RCH capacity in flies that already were developmentally acclimated (RCH 18 °C). However, most lines were capable of expressing a RCH response when developmentally acclimated (Fig. S2 A and C), and we found no relationship between DACC capacity and RCH capacity in developmentally acclimated individuals reared at 18 °C (Fig. S5 B and C).

Our findings are consistent with the previous suggestion that there is some mechanistic overlap between DACC and RCH (19). The longer-term DACC response may induce a larger degree of change in any mechanisms that overlap between DACC and RCH and also may induce additional mechanisms unique to DACC (10, 19, 26). There is substantial variation in the timing and number of extreme swings in temperature, such as those associated with fast-moving cold fronts, from year to year. Thus, the ability to induce a RCH response even during the process of seasonal DACC probably is beneficial at midlatitude sites, such as the site of origin of this population (3, 4). Selection for the ability to maintain a rapid plastic response even when developmentally acclimated may contribute to maintaining the nonoverlapping physiological architecture of rapid and long-term thermal plasticity responses.

Association Mapping and Functional Mutant Testing.

Given that previous physiological research and our own correlations described above suggest that only some mechanisms are shared between long-term and short-term acclimation (10, 19), we expect that the degree of genetic overlap between RCH and DACC responses will be small or undetectable. We found no SNPs overlapping among RCH at 25 °C (61 SNPs), RCH after DACC at 18 °C (103 SNPs), and DACC (66 SNPs, P < 10⁻⁵) (Dataset S1). In addition, there was no overlap in genes associated with significant SNPs for any of the acclimation treatments. Linkage disequilibrium (LD) among SNPs of each acclimation phenotype may not be independent, and we can estimate LD among all significant SNPs to find any degree of dependency of the genetic architecture. There was high LD within each individual plasticity phenotype, but SNPs with high LD across plasticity phenotypes always were in close proximity to one another (Dataset S2). The lack of overlap and the contrasting patterns of LD in plasticity phenotypes are consistent with distinct genetic architectures for DACC and RCH and are in accordance with our hypothesis of limited mechanistic overlap for long- and short-term thermal plasticity.

In contrast, there was some overlap in SNPs in basal tolerance to acute stress and acclimation, mirroring our assessment of physiological constraint in these phenotypes. Overlap was highest between DACC and tolerance to acute cold (33 of 360 SNPs within 27 unique genes) and was lowest between RCH for flies reared at 25 °C and tolerance of acute cold (3 of 360 SNPs with one gene overlap) (Dataset S1). RCH for flies developmentally acclimated by rearing at 18 °C and acute cold stress shared 11 of 360 total associated SNPs. The degree of overlapping SNPs is in agreement with the degree of phenotypic associations we observed, confirming that some of the constraint in these phenotypes is genetically founded, i.e., 9.16% of SNP overlap and a phenotypic correlation of −0.98 between DACC and tolerance to acute cold compared with 0.83% of SNP overlap and a phenotypic correlation of −0.56 between RCH at 25 °C and basal tolerance to acute cold.

Although the genetic architectures of the three acclimation phenotypes differ, overrepresentation of functional processes in short-term and long-term acclimation suggests that these two forms of plasticity share some physiological pathways. SNPs associated with either DACC or RCH after DACC were enriched for several functional groups in DAVID analyses (Table 2). SNPs associated with RCH at 25 °C treatment, although not significantly enriched via DAVID, revealed functions such as iron binding, apoptotic cell clearance, cell adhesion, calcium ion binding, cytoskeleton, cuticle components, and oxidation–reduction after manual annotation (38, 39). Cell membrane-associated genes were enriched both for DACC and RCH after flies were developmentally acclimated at 18 °C, consistent with a deep literature recognizing the role of cell membrane lipids in thermotolerance and acclimation responses (21, 40, 41). Cytoskeleton and cuticle components (Cpr62Bb, rhea, trio, Cpr66D) also were overrepresented, highlighting their importance for mitigating cold stress events (22, 42–44). Consistent with a substantial body of literature showing that RCH reduces apoptosis after cold stress (39, 45, 46), enrichments in both RCH and DACC highlight genes associated with autophagy and apoptosis [e.g., Ecdysone-induced protein 74EF (Eip74EF), sima, betalnt-nu, CG30116, Fish-lips (Fili, a leucine-rich repeat), and lola].

Table 2.

Functional enrichments (DAVID) for DACC and RCH at 18 °C

DAVID enrichment	No. of genes	% enrichment
DACC
Nucleus	8	4.7
Membrane	8	4.7
Phosphoprotein	7	4.1
Glycoprotein	6	3.5
Transcription regulation	6	3.5
Alternative splicing	6	3.5
Transcription	6	3.5
Calcium	4	2.3
Apoptosis	2	1.2
Phosphoric monoester hydrolase	2	1.2
RCH at 18 °C
Alternative splicing	8	2.9
Developmental protein	7	2.6
Nucleus	7	2.6
Cell membrane	4	1.5
Lipoprotein	3	1.1

There were no significant functional enrichments (% Enrich) for RCH at 25 °C, but manual inspection of significant genes on FlyBase.org provided biological and functional characteristics (see main text). Genes submitted to DAVID were significant for each phenotype for each sex independently and for sexes combined.

Of particular interest, alternative splicing was enriched for both short- and long-term plasticity, and transcription was enriched as a class in long-term plasticity (Table 2). Several studies have associated changes in transcriptional regulation with long-term (47–49) but not with short-term acclimation responses (25). Long-term acclimation may provide the additional time needed to enhance both the physiological and transcriptional milieu further for a larger-magnitude acclimation response than seen with short-term acclimation. Splicing has been shown to be up-regulated in response to artificial selection on cold-tolerance traits (50), and alternatively spliced isoforms of a thermally responsive gene (stv) have been associated with a protective response after multiple heat exposures (51).

To assess the relative importance of several candidate loci and physiological processes for both basal cold tolerance and either DACC or RCH, we used a series of available mutants, all in the w1118 Drosophila melanogaster genetic background (Table 3). The w1118 genotype has basal cold tolerance and plasticity responses consistent with the naturally segregating variation we observed among lines of the DGRP. Mutants showing reduced plasticity in this background suggest that these genes/processes may be important for acclimation responses in general. GWAS studies often generate many candidate SNPs. By testing whether mutants in specific genes/processes affect plasticity, we can narrow our lists of candidates to those that are most likely causal, eliminating many false positives and setting the stage for future work that will test whether specific allelic variants in these candidate loci/physiological processes are truly causal (52).

Table 3.

Functional mutants tested

Gene	RCH at 25 °C		RCH at 18 °C		DACC
	Score	P value	Score	P value	Score	P value
Background	0.48		0.90		0.74
Atg7	0.25	0.017	0.87	0.27	NA	NA
CG8398	0.31	0.32	0.49	<0.001	NA	NA
CG32111	0.38	0.023	0.68	0.0077	0.78	0.62
CG33275	0.56	0.12	0.82	0.21	0.70	0.56
CG33275	0.20	0.22	0.85	0.30	0.86	0.058
dpr5	−0.14	0.16	0.56	<0.001	0.52	0.016
Eip74EF	0.54	0.49	0.71	0.0056	0.00	0.32
Fili	0.54	0.51	0.48	<0.001	0.59	0.10
Glu-RI	0.20	0.17	0.81	0.16	0.78	0.63
Prosap	0.48	0.98	NA	NA	NA	NA
Px	0.34	0.42	0.65	0.0065	0.64	0.16
Syt12	0.37	0.23	NA	NA	0.55	0.016
VGlut	0.43	0.55	0.67	<0.001	0.40	<0.001

We were unable to test some mutants because of poor growth and viability when developmentally acclimated at 18 °C. A significant interaction term between genotype and environment (cold stress with and without acclimation) indicated that the mutant differed in acclimation capacity compared with the genetic background control (w1118), represented here under the P value column. Boldface type indicates significance at P < 0.05. The score represents the difference between survivorship following acute stress (at −6 °C for 1 h) for RCH at 25 °C and DACC and at −8 °C for 1 h for RCH after DACC at 18 °C and survivorship after acclimation for 2 h at 4 °C followed by the acute stress. For survivorship proportions used to calculate score, see Dataset S2.

Consistent with observations that RCH reduces cold shock-induced apoptosis (39, 53), mutants in genes associated with regulation of autophagy had reduced capacity for RCH both when flies were developmentally acclimated at 18 °C and at 25 °C when flies were not developmentally acclimated (Atg7 and Eip74EF; Table 3). Similarly a mutant for Fili, a gene associated with regulation of apoptotic processes, showed a significantly reduced RCH response at 18 °C (Table 3) (41). This observation further reinforces the importance of apoptosis in clearing and controlling cellular death in response to cold stress, especially RCH (45, 53, 54). Autophagy is a broadly conserved mechanism for clearing cellular components damaged by stress, such as mitochondria or protein aggregates, and overexpression of autophagy has been associated with greater stress tolerance (55–59). Our working hypothesis is that RCH increases the cellular autophagy response, allowing cells to clear cold-damaged components more efficiently before signals of damage activate apoptotic pathways, enhancing cellular stress tolerance and cell survival after acute cold exposure.

A mutant for px, a gene associated with imaginal disk-derived wing development and autophagy during wing morphogenesis, also had reduced RCH response when developmentally acclimated at 18 °C (Table 3) (60–64). A mutant in Syt12, a gene associated with calcium and lipid signaling, two processes previously suggested to be important in short-term acclimation responses (64) and maintenance of membranes for maintaining osmotic balance and cell viability (65), showed a significantly reduced DACC response (Table 3) (66). A mutant in VGlut, implicated in energy transmission, synaptic activity, and transmembrane transport (67, 68), also showed a reduced RCH response, which is consistent with the importance of membrane fluidity, cellular exchange, and energetics (40, 41, 44, 69). In addition, a mutant in a transcription factor-binding gene involved in the significantly enriched function of transcription (CG32111) (70) significantly increased the degree of short-term plasticity after RCH treatment in both developmentally acclimated individuals reared at 18 °C and individuals reared at 25 °C.

Both long- and short-term acclimation can increase survivorship and performance, including expanding activity periods and protecting fecundity, in the face of cold stress (7, 11). Here we show that a midlatitude population of D. melanogaster contains substantial segregating genetic variation for both long- and short-term acclimation responses, each response having relatively high heritabilities, suggesting the potential for evolution to shape thermal plasticity. Studies of phenotypic convergence across traits and taxa have shown selection can shape the genetic architecture of populations in diverse ways (71, 72). Understanding the potential for selection to shape cold-tolerance phenotypes, including both long- and short-term acclimation, will require a more detailed understanding of both whether the segregating variants we have identified here are causal for cold tolerance and how the complex interactions between these variants may facilitate or constrain selection for basal tolerances or plasticity.

Methods

All 192 DGRP lines were obtained from the Bloomington Stock Center. Stocks were maintained at 25 °C on a 12-h light, 12-h dark cycle. Five males and five females from each line were anesthetized with CO₂ and placed in a standard narrow Drosophila vial for egg-laying at 18 °C or 25 °C. After 3 d of egg-laying, parents were removed for density control of experimental flies. Offspring were sorted by sex and were allowed to age for 5–7 d (11, 18). Three replicates of 10–20 flies were moved to empty vials to avoid the influence of the food on the freezing temperature of flies.

Flies in vials without food were placed at 0 °C for 16 h to simulate chronic cold exposure. After 16 h, flies were removed from the cold, placed in vials with food medium, and allowed to recover for 24 h at 25 °C before survivorship was assessed (21, 26, 73). Flies were considered alive if they showed coordinated movement 24 h after exposure (12). Discriminating temperatures for acute cold stress were determined previously for a similar population of D. melanogaster from a scale of −5° to −8 °C for flies reared at 18 °C and 25 °C (9, 12, 73, 74). To calculate RCH for flies reared at 25 °C (RCH at 25 °C), flies were pretreated for 2 h at 4 °C followed by cold exposure at −6 °C for 1 h (i.e., RCH) or were moved directly from room temperature to −6 °C for 1 h (i.e., acute survivorship) (73, 75). Survivorship after acute cold exposure when flies were reared at 25 °C was compared with survivorship after acute cold exposure when flies developmentally acclimated by rearing at 18 °C were placed in a cold incubator at −6 °C to calculate a DACC capacity score. Because survivorship increased to such high proportions at −6 °C for flies reared at 18 °C, a discriminating temperature of −8 °C was chosen to test flies exposed to RCH-inducing conditions after DACC because survivorship after exposure to −8 °C was similar to that of flies reared at 25 °C and exposed to −6 °C for 1 h. RCH for flies developmentally acclimated by rearing at 18 °C (RCH at 18 °C) was assessed by pretreating flies by exposure to 4 °C for 2 h followed by exposure to −8 °C for 1 h.

Survivorship was calculated for each sex independently and for both sexes combined, because survivorship of both sexes was correlated (Table S2). Averages across replicates were used as an overall survivorship proportion for each genotype of the DGRP. Acclimation scores were calculated using the mean survivorship of each line to determine the magnitude and direction of the RCH at 25 °C, DACC, or RCH at 18 °C response for each line using the formula: Acclimation Capacity = % Survival_Acclimated − % Survival_{Nonacclimated}. Heritability was calculated using the lmer package in R. Phenotypic correlations for each acclimation score were assessed between RCH at 25° and 18 °C and DACC. Basal tolerance and plasticity trade-offs compared both acute survivorship and chronic survivorship proportions with the acclimation scores for each line. Average magnitudes of acclimation responses were compared via paired t tests.

Genome-wide association mapping for RCH at 25° and 18 °C and DACC were done using all available DGRP lines, for each sex separately and both sexes combined, using the publicly available SNP data from Mackay et al. (dgrp2.gnets.ncsu.edu; DGRP Freeze 2.0) (34, 35). Top SNPs/genes were taken at a standard GWAS significance level of P < 10⁻⁵. LD was calculated using SNPs at the P < 10⁻⁵ significance level. SNPs for each line were taken from the DGRP Freeze 2.0, and LD was assessed using PLINK, version 1.9 (76), accounting for nonhuman chromosome and only within chromosome associations. We used DAVID (77) to assess enrichment of functional ontology using genes associated with each acclimation phenotype for each sex and for the pooled sexes.

Candidate genes for mutant analysis were selected based on the availability of mutant stocks and biological and molecular function as described on FlyBase, highlighting mutants of physiological interest to cold tolerance. Mutants were selected based on one of two criteria. Some were available mutants in genes with top-candidate SNPs from our genome-wide associations, and some mutants were selected because they represented candidate processes that were indicated by a combination of our enrichment analyses (Table 2) and standing mechanistic hypotheses from the literature (Dataset S2). In addition, although these genes were not in our top SNP list (P < 10⁻⁵), several SNPs within each gene had a GWAS P value < 0.05, suggesting some differentiation in acclimation response. All mutants selected shared the same genetic background (w1118), a background that shows substantial short- and long-term cold acclimation responses (Table 3). The 12 mutant lines and the background genotype (w1118) were obtained from the Bloomington Stock Center (Table 3 and Dataset S2). ANOVA was calculated for each mutant compared with the background independently and for each acclimation treatment. A significant interaction effect of genotype and treatment indicates that the mutant has altered thermal plasticity compared with the genetic background control.