Complex Patterns of Phenotypic Plasticity: Interactive Effects of Temperature During Rearing and Oviposition

Ecology, 86(4), 2005, pp. 924±934 ᭧ 2005 by the Ecological Society of America

COMPLEX PATTERNS OF PHENOTYPIC PLASTICITY: INTERACTIVE EFFECTS OF TEMPERATURE DURING REARING AND OVIPOSITION

R. CRAIG STILLWELL1 AND CHARLES W. F OX Department of Entomology, University of Kentucky, S-225 Ag Science Center North, Lexington, Kentucky 40546-0091 USA

Abstract. Temperature profoundly affects growth and life history traits in ectothermic animals through selection (i.e., genetic) and through direct effects on the phenotype (i.e., nongenetic/plasticity). We examined the effects of rearing temperature (24Њ,30Њ, and 36ЊC) on adult body size and development time and the interactive effects of temperature experienced during rearing and oviposition on several life history traits (age-at-®rst-reproduction, fecundity, egg size, egg development, and egg hatching) in two populations of the seed beetle, Stator limbatus, collected at different elevations. The higher elevation population was larger and matured sooner than the low-elevation population when raised at the lower temperature, but the reverse was true at the higher temperature suggesting that these populations have adapted to local temperature. There were interactions between the effects of rearing temperature and oviposition temperature for age-at-®rst-reproduction, fecundity, egg development, egg hatching, and two composite measures of ®tness, generating complex reaction norms. The most dramatic example of this was a large maternal effect on egg hatching; females raised at low temperature produced eggs that had substantially reduced hatching when laid at high temperature. Our experimental design also allowed us to explore the adaptive signi®cance of acclimation. Beetles reared at intermediate or low temperature had the highest ®tness at multiple oviposition temperatures. There was little support for the ``bene®cial acclimation hypothesis,'' which predicts that beetles should have higher ®tness at the temperature at which they were reared; acclimation had only a small effect on ®tness. This study shows that temperature-mediated plasticity can be complex, but these complex patterns can yield new insights into the evolution of phenotypic plasticity. Key words: acclimation; adaptation; bene®cial acclimation hypothesis; body size; maternal effects; phenotypic plasticity; Stator limbatus; temperature.

INTRODUCTION in nature. Body size and egg size of ectotherms exhibit clinal variation along latitudinal (e.g., Lonsdale and For many ectothermic animals, the temperature ex- Levinton 1985, Armbruster et al. 2001) and altitudinal perienced during development strongly affects many gradients (e.g., Berven 1982). When individuals are growth and life history traits. For instance, body size reared in a common environment, these differences are (Atkinson 1994), development time (Atkinson 1994), usually found to be partially genetically based. In Dro- and egg/offspring size (Fox and Czesak 2000) generally sophila melanogaster, body size clines are often re- increase with decreasing temperature, whereas fecun- peatable across continents (e.g., James et al. 1995, dity (e.g., Fischer et al. 2003a) generally declines with Van't Land et al. 1995). These clines can evolve rap- decreasing temperature. These phenotypic responses idly, only a couple of decades after introduction to a may be a result of adaptation to different thermal en- new continent (Huey et al. 2000), indicating strong vironments or may be an unavoidable consequence of natural selection on either body size, egg size, or some a temperature effect on the organism's physiology dur- correlated trait. Laboratory natural selection experi- ing development (developmental plasticity). The extent to which temperature-mediated natural selection vs. the ments, in which insects are allowed to adapt to high environmental effect of temperature differentially af- or low temperature, also demonstrate that both body fect traits at different temperatures has generated de- size and egg size evolve to be larger at colder tem- bate between adaptive (genetic) and developmental peratures (Partridge et al. 1994, Azevedo et al. 1996; (nongenetic) hypotheses (Van Voorhies 1996, 1997, but see Santos et al. 2004). Mousseau 1997, Partridge and Coyne 1997). Nongenetic responses to developmental temperature Evidence for genetic adaptation to temperature (phenotypic plasticity) are widespread. For example, comes from studies of patterns of geographic variation growth at lower temperature results in increased adult body size in Ͼ80% of ectothermic animals (Atkinson 1994), and many arthropods typically lay larger eggs Manuscript received 22 March 2004; accepted 26 July 2004; ®nal version received 21 September 2004. Corresponding Editor: at lower temperatures (Fox and Czesak 2000). Non- E. Brodie III. genetic responses to developmental temperature are 1 E-mail: [email protected] thus very similar to those produced by evolution at 924 April 2005 COMPLEX PATTERNS OF PLASTICITY 925 different temperatures making it dif®cult to distinguish and the interactive effects of rearing vs. oviposition between genetic (evolution due to selection) and non- temperature and to assess the adaptive signi®cance of genetic (plasticity) causes for geographic patterns. acclimation. Within central Arizona, USA, S. limbatus Common garden and reciprocal transplant experiments experiences large daily (ranging from 24Њ to 39ЊC; are required to disentangle these factors. mean minimum daily temperature compared to mean Adding to this complexity, patterns of phenotypic maximum daily temperature in Phoenix, Arizona) and plasticity may change if the temperature encountered spatial (25Њ±31ЊC; monthly mean in Oracle, Arizona, in an earlier vs. a later life stage interact to affect a compared to Phoenix, Arizona; Western Regional Cli- phenotype. For instance, immature stages of many in- mate Center, Desert Research Institute) variation in sects develop at different times of the year and in other temperature during August when beetles are most ac- physical localities than adults, subjecting these life tive. S. limbatus is a generalist seed parasite of le- stages to different thermal environments. Also, im- gumes; adults are the only dispersing stage of the life matures are less mobile than adults and are often unable cycle, with larvae being restricted to the seed their to leave unfavorable thermal environments. Neverthe- mother has chosen for them. Since larvae cannot move less, few studies have considered both the separate and between seeds they cannot move between thermal en- interactive effects of temperature experienced during vironments. immature development (rearing) vs. adult (oviposition) In this study, we examine two populations of S. lim- stages of the life cycle (Fox and Czesak 2000). We batus from central Arizona that occur at different el- need to understand how these interactions affect re- evations (Phoenix, ϳ500 m above sea level; Oracle, action norm shape to predict response to selection in ϳ1400 m above sea level) and experience different more complex environments. local thermal environments (mean temperature differ- Studies on the effects of temperature experienced at ence in August of ϳ6ЊC). Our a priori expectations an early and later life stage provide a unique oppor- were that these populations are adapted to a ``warm'' tunity to study the adaptive signi®cance of acclimation and ``cool'' environment and should be genetically dif- (a nongenetic, physiological alteration in a phenotype ferentiated in a number of traits. We ®rst assessed how caused by a change in the environment). It is commonly temperature during rearing affects adult body size and assumed that development at a certain temperature will larval development time in both populations. We then enhance ®tness in the adult stage at that temperature examined the separate and interactive effects of tem- (``bene®cial acclimation hypothesis''; Leroi et al. perature experienced by beetles during rearing vs. ovi- 1994). Alternatively, acclimation may be nonadaptive position periods of the life cycle on age-at-®rst-repro- if rearing in one environment enhances ®tness in al- duction, fecundity, egg size, egg development, and egg ternate environments to which the organism is not ac- hatching. Finally, using the approach described by climated (Huey et al. 1999). For example, development Huey et al. (1999) we tested the bene®cial acclimation, at an intermediate (``optimal developmental tempera- optimal developmental temperature, hotter is better, ture hypothesis''), low (``colder is better hypothesis''), and colder is better hypotheses. or high (``hotter is better hypothesis'') temperature may enhance ®tness in a variety of adult thermal environ- METHODS ments. Recently, factorial designs have been developed Natural history of Stator limbatus to explore the adaptive signi®cance of acclimation (Huey and Berrigan 1996). These designs and recently The seed beetle, Stator limbatus (Horn) (Coleoptera: developed statistical models allow for several com- Chrysomelidae: Bruchinae), is distributed from north- peting hypotheses to be tested simultaneously (Huey ern South America to the southwestern United States et al. 1999). Assessments using this approach have (Johnson and Kingsolver 1976, Johnson et al. 1989, shown that while the bene®cial acclimation hypothesis Nilsson and Johnson 1993). It is a generalist seed her- is supported in some cases, alternative hypotheses often bivore that has been collected from Ͼ70 species of have a greater in¯uence on ®tness (Huey et al. 1999, legumes throughout its wide geographic range, though Gibert et al. 2001). However, these studies have been only one or a few hosts are encountered in most lo- criticized as having examined the adaptive signi®cance cations. S. limbatus commonly uses Acacia greggii (Fa- of developmental plasticity (irreversible as well as re- baceae: Mimosoideae), Parkinsonia ¯orida (Fabaceae: versible changes to the phenotype induced during de- Caesalpinioideae; previously Cercidium ¯oridum), and velopment) instead of traditionally studied acclimation Parkinsonia microphylla (Fabaceae: Caesalpinioideae) (reversible changes to the phenotype induced in later as hosts in central Arizona, USA. Females oviposit life stages) (Wilson and Franklin 2002). Accordingly, directly onto host seeds inside of seed pods that have the adaptive signi®cance of acclimation and how it can either dehisced or been damaged by other organisms be studied is the subject of substantial debate. (e.g., mice, other bruchine beetles such as Mimosestes Here, we use the seed beetle, Stator limbatus (Co- species, etc.). Upon hatching, ®rst-instar larvae burrow leoptera: Chrysomelidae: Bruchinae), to investigate into the seed directly underneath the egg, where de- both genetic and nongenetic responses to temperature velopment and pupation take place. Mating and egg 926 R. CRAIG STILLWELL AND CHARLES W. FOX Ecology, Vol. 86, No. 4 laying start ϳ24±48 h after emerging from the seed light : dark) within 24 h of being laid (N ϭ 85±90 for (in the laboratory). S. limbatus needs only the resources each rearing treatment ϫ population combination; total inside of a single seed to complete development and ϭ 525 families). This was repeated daily until females reproduce; additional food and water are not necessary. laid eggs on ϳ7±8 seeds, after which adults were dis- carded. Study populations Emerging progeny were collected twice daily and We used one S. limbatus population from Oracle, development time was recorded. Beetles were weighed Arizona (Oracle population), and one from Phoenix, on an electronic balance (AT261 Delta Range; Mettler Arizona (Phoenix population). The Oracle population Toledo, Columbus, Ohio, USA) to the nearest 0.1 mg. was collected along Highway 77 and adjacent roads in Virgin beetles were placed in individual 35-mm petri Oracle, Pinal County, Arizona, USA (32.62Њ N, 110.78Њ dishes at their assigned oviposition treatment for 24± W; ϳ1400 m above sea level) in August 2000 from 36 h before we allowed them to mate (N ϭ 24±33 per mature seed pods of Ͼ20 A. greggii trees. Mean annual rearing temperature ϫ oviposition temperature ϫ pop- temperature at this site is ϳ16.7ЊC and the August ulation combination; i.e., there were three rearing treat- monthly mean is ϳ25.2ЊC (Western Regional Climate ments ϫ three oviposition treatments ϫ two popula- Center, Desert Research Institute, Reno, Nevada, tions ϭ 18 groups of 24±33 families). This assured that USA). The Phoenix population was collected just north eggs laid later in the experiment had been matured at of Phoenix, Maricopa County, Arizona, USA (33.79Њ the correct oviposition temperature. Ten A. greggii N, 112.12Њ W; ϳ500 m above sea level) in August 2000 seeds were present in the female dishes because the from seed pods of Ͼ20 P. ¯orida trees. Mean annual presence of hosts facilitates egg maturation (Savalli and temperature at this site is ϳ20.9ЊC, and the August Fox 2002). After 24±36 h a randomly selected virgin monthly mean is ϳ31.3ЊC. Seed pods were shipped male originating from the same population and rearing back to the laboratory where colonies were established regime was placed in the female's dish. Each dish was from Ͼ200 emerging adult beetles of each population. checked for eggs twice daily. If eggs were present, Colonies were maintained on A. greggii seeds at ϳ100 adults were transferred to a second 35-mm petri dish families each generation (at 30ЊC, 15:9 light : dark) pri- containing 10 A. greggii seeds. Females were allowed or to the experiment, which started in February 2002. to oviposit for 24 h after which they were transferred Beetles were reared on A. greggii in the laboratory to a third dish (60 mm) containing 30 A. greggii seeds because they have high survivorship (Ͼ90%) at 30ЊC upon which they were allowed to oviposit until death on this host (Fox et al. 1994). (to estimate lifetime fecundity). Egg size and egg mortality were assessed by ex- Experimental design amining only eggs laid in the female's second dish We used a completely randomized design with a full because egg size is known to change over a female's 3 ϫ 3 factorial treatment arrangement to examine the life and may in¯uence survivorship of larvae (Fox and effects of rearing temperature on growth traits and the Mousseau 1996, Fox et al. 1997). This also ensured effects of rearing vs. oviposition temperature on life that the eggs measured were matured at the correct history traits. In brief, full-sib families (family sizes oviposition temperature. Egg size of each female was were too small to allow the use of a split brood design) determined by measuring the length of 1±3 haphazardly of beetles from both the Oracle and Phoenix popula- chosen eggs using an ocular micrometer on a 100ϫ tions were assigned randomly to one of three rearing stereo microscope (0.005-mm precision). Measuring treatments (24Њ,30Њ,or36ЊC). These temperatures are egg mass is not practical because removing eggs from within the range of temperatures normally encountered the seed is destructive. However, egg length is highly in the ®eld (see Introduction). One-third of the families correlated with egg mass (Fox and Mousseau 1996) within each rearing treatment were randomly assigned and, therefore, a good estimate of overall egg size. Egg to each of three oviposition treatments (24Њ,30Њ,or mortality was determined for each individual egg and 36ЊC). All progeny of a family experienced only one was divided into three categories: did not develop (de- rearing ϫ oviposition treatment combination; siblings noted by clear eggs), developed but did not hatch (in- were not split among the treatments. Offspring were dicated by a visible embryo), and hatched (denoted by reared in temperature-controlled growth chambers. Pe- the presence of frass in the eggs). In total 3160 off- tri dishes were rotated daily to minimize the effects of spring were reared to adult and weighed. Reproductive spatial variation in temperature. traits were recorded for 1372 females. To create families, virgin males and females were paired and placed with 20 A. greggii seeds in a 35-mm Data analyses petri dish and allowed to mate and oviposit at 30ЊC, We analyzed all data using SAS 8.2 (SAS Institute, 15:9 light : dark. They were checked daily for eggs. If Cary, North Carolina, USA). Normal probability plots eggs were present, seeds were scraped to one egg/seed revealed that the data (except egg development and egg (to eliminate larval competition) and placed in their hatching) were approximately normally distributed, so assigned rearing treatments (24Њ,30Њ,or36ЊC, 15:9 no transformations were performed. We used mean trait April 2005 COMPLEX PATTERNS OF PLASTICITY 927 values of each family (approximately three offspring per family on average) in our ANOVAs (SAS PROC GLM) to remove the non-independence of siblings within treatments. For egg mortality data, we used logistic regression analysis (SAS PROC GENMOD) on individual eggs (11 756 total eggs) with family included as a covariate nested within treatment interactions to control for the variability among families on egg mortality (the family effect was highly signi®cant for ␹2 ϭ Ͻ both analyses; egg development, 18 51.44, P ␹2 ϭ Ͻ 0.0001; egg hatching, 18 74.96, P 0.0001). For simplicity, results are presented by main effects and interactions although analyses for each trait were performed using a full model including main effects and all possible interactions. We created two composite measures of ®tness to assess the acclimation hypotheses. The ®rst composite measure of ®tness (rC1) was the number of offspring produced by a female that survived to a ®rst-instar larva (proportion of offspring hatching ϫ fecundity). The second (rC2) was adjusted by the age at which each female reproduced (rC1/age-at-®rst-reproduction). Nor- mal probability plots indicated that both rC1 and rC2 were approximately normally distributed and thus no transformation was performed. Since body size and development time are affected by rearing temperature only, they are not included in our estimates of ®tness. Egg size is included in ®tness estimates through its effects on larval survivorship and female fecundity. To evaluate the acclimation hypotheses, we used AN- OVA with orthogonal polynomial contrasts (Huey et al. 1999). We subsampled the original data set to create the balanced design necessary for orthogonal polynomial contrasts (Sokal and Rohlf 1995). The number sampled from each treatment combination was based on the treatment combination with the lowest sample size. Separate analyses were run for each population. We used a statistical model in which the linear (rearing ) and quadratic (rearing ) effects of rearing tem- (l) (q) FIG. 1. The effect of rearing temperature on (A) female perature, and their interactions with oviposition tem- body mass, (B) male body mass, and (C) female development perature, allowed us to assess the relative impact of time of beetles from Phoenix (low-elevation) and Oracle each acclimation hypothesis (Huey et al. 1999). The (high-elevation) populations in Arizona, USA (means Ϯ 1 colder is better hypothesis predicts that rearing at a SE). lower temperature will enhance ®tness at all oviposition temperatures (rearing is signi®cant with a neg- (l) support for several hypotheses simultaneously, the relative coef®cient) whereas the hotter is better hypothesis ative importance of each can be determined by com- predicts that rearing at a higher temperature will en- paring F ratios. Huey et al. (1999) provides details of hance ®tness at all oviposition temperatures (rearing (l) the analysis and an extensive discussion of the hy- is signi®cant with a positive coef®cient). The optimal potheses. developmental temperature hypothesis predicts that rearing at an intermediate temperature will enhance RESULTS ®tness at all oviposition temperatures (rearing(q) is signi®cant with a negative coef®cient). The bene®cial ac- Main effects of rearing temperature climation hypothesis predicts that an individual's ®t- Body mass decreased with increasing rearing tem- ness will be greatest when oviposition temperature and perature for both male and female beetles (Fig. 1A, B; ϫ Ͻ rearing temperature are the same (rearing(l) ovipo- Table 1; P 0.0001), consistent with the typical pattern ϫ sition and rearing(q) oviposition interactions signif- in ectothermic animals (Atkinson 1994). At all rearing icant). Although the analysis can indicate signi®cant temperatures males were larger than females (highly 928 R. CRAIG STILLWELL AND CHARLES W. FOX Ecology, Vol. 86, No. 4

TABLE 1. ANOVAs (Type III sums of squares) for the ef- signi®cant sex effect; Table 1; P Ͻ 0.0001) as observed fects of rearing temperature and population on adult body mass (including the effect of sex) and female development in other studies of Stator limbatus (Savalli and Fox time of Stator limbatus reared from populations found in 1998). Body size (of both sexes) decreased ϳ6±9% Arizona, USA. between 24Њ and 30ЊC and decreased ϳ4±10% between 30Њ and 36ЊC. Beetles took longer to develop at lower Source df FPtemperatures (Fig. 1C; Table 1; P Ͻ 0.0001; data col- Mass (mg) lected for females only). Rearing temperature (R) 2 178.02 Ͻ0.0001 Age-at-®rst-reproduction (measured as time between Population (P) 1 0.92 0.34 Њ Sex (S) 1 65.10 Ͻ0.0001 emergence and ®rst oviposition) was lowest at 30 C R ϫ P 2 8.56 0.0002 (Fig. 2A, B; Table 2; P Ͻ 0.0001). Female lifetime R ϫ S 2 1.21 0.30 fecundity decreased monotonically with increasing P ϫ S 1 0.95 0.33 Ͻ R ϫ P ϫ S 2 0.62 0.54 rearing temperature (Fig. 2C, D; Table 2; P 0.0001). Error 997 Interestingly, females reared at 30ЊC (the intermediate Female development time (d) temperature) laid the smallest eggs (vs. those raised at Њ Њ ϭ R 2 10 279.40 Ͻ0.0001 24 C and 30 C; Fig. 2E, F; Table 2; P 0.01), which P 1 0.96 0.33 is not the monotonic relationship (larger eggs at lower R ϫ P 2 3.37 0.04 temperatures) found for most arthropods (Fox and Cze- Error 496 sak 2000).

FIG. 2. The effect of both rearing temperature and oviposition temperature on (A, B) age-at-®rst-reproduction, (C, D) lifetime fecundity, and (E, F) egg length of females from the (A, C, E) Oracle and (B, D, F) Phoenix populations (means Ϯ 1 SE). April 2005 COMPLEX PATTERNS OF PLASTICITY 929

TABLE 2. ANOVAs (Type III sums of squares) for the ef- Overall, females raised at 36ЊC laid eggs that were fects of rearing temperature, oviposition temperature, and population on life history traits of Stator limbatus. less likely to develop than were eggs laid by females Њ Њ ␹2 ϭ ϭ raised at 24 and 30 C (Fig. 3A, B; 2 5.73, P Source df FP0.06). There was no effect of rearing temperature on ␹2 ϭ ϭ Age at ®rst reproduction (d) egg hatching (Fig. 3C, D; 2 1.90, P 0.39; but see Rearing temperature (R) 2 16.58 Ͻ0.0001 the interaction results discussed below). In general, rC1 Oviposition temperature (O) 2 307.46 Ͻ0.0001 (the ®rst composite measure of ®tness) decreased with Population (P) 1 2.8 0.09 increased rearing temperature (Fig. 4A, B; Table 3; P R ϫ O 4 2.38 0.05 Ͻ R ϫ P 2 0.89 0.41 0.001) while rC2 (the second composite measure of O ϫ P 2 1.93 0.15 ®tness that considers age-at-®rst-reproduction) was R ϫ O ϫ P 4 1.29 0.27 greatest at 30ЊC (Fig. 4C, D; Table 3; P Ͻ 0.001). Error 478 Lifetime fecundity Main effects of oviposition temperature R 2 64.38 Ͻ0.0001 Oviposition temperature had a larger effect than did O 2 114.84 Ͻ0.0001 P 1 0.55 0.46 rearing temperature on age-at-®rst-reproduction, fe- R ϫ O 4 6.51 Ͻ0.0001 cundity, and egg size (comparison of F ratios in Table R ϫ P 2 0.24 0.79 2). Females took longer to initiate reproduction at lower O ϫ P 2 1.11 0.33 Ͻ R ϫ O ϫ P 4 0.72 0.58 oviposition temperatures (Fig. 2A, B; Table 2; P Error 470 0.0001). Fecundity decreased sharply from 30Њ to 36ЊC, Њ Egg length (mm) but there were only small differences between 24 and Њ Ͻ R 2 4.7 0.01 30 C (Fig. 2C, D; Table 2; P 0.0001). Females also O 2 21.26 Ͻ0.0001 laid the smallest eggs when they oviposited at 30ЊC P 1 68.26 Ͻ0.0001 (Fig. 2E, F; Table 2; P Ͻ 0.0001), which was similar R ϫ O 4 1.23 0.30 R ϫ P 2 0.94 0.39 to the effect of rearing temperature on egg size. There O ϫ P 2 1.81 0.17 was no overall effect of oviposition temperature on R ϫ O ϫ P 4 0.43 0.79 ␹2 ϭ ϭ either egg development (Fig. 3A, B; 2 0.50, P Error 463 ␹2 ϭ ϭ 0.78) or egg hatching (Fig. 3C, D; 2 0.42, P 0.81)

(but see the interaction results discussed below). rC1 decreased with increasing oviposition temperature

FIG. 3. The effect of both rearing temperature and oviposition temperature on (A, B) proportion of eggs that developed into an embryo and (C, D) proportion of those that developed that hatched, for eggs laid by female beetles from (A, C) Oracle and (B, D) Phoenix populations (means Ϯ 1 SE). 930 R. CRAIG STILLWELL AND CHARLES W. FOX Ecology, Vol. 86, No. 4

FIG. 4. The effect of both rearing temperature and oviposition temperature on the composite measures of ®tness, (A, B) Ϯ rC1 and (C, D) rC2, for beetles from (A, C) Oracle and (B, D) Phoenix populations (means 1 SE).

Ͻ (Fig. 4A, B; Table 3; P 0.001) but rC2 was highest diction, females from the Phoenix population laid larg- at the intermediate oviposition temperature (Fig. 4C, er eggs than females from the Oracle population (Table D; Table 3; P Ͻ 0.01). 2; P Ͻ 0.0001; least-square means Ϯ 1 SE; Oracle, 0.55778 Ϯ 0.00095 mm; Phoenix, 0.56896 Ϯ 0.00096 ϫ Main effects of population and population mm). temperature interactions ϫ When raised at 24ЊC, Oracle beetles were larger than Rearing temperature oviposition temperature Phoenix beetles, whereas Phoenix beetles were larger interactions than Oracle beetles when raised at 36ЊC (signi®cant There were interactions between rearing temperature rearing temperature ϫ population interaction; Fig. 1A, and oviposition temperature for age-at-®rst-reproduc- B; Table 1; P ϭ 0.0002). Likewise, Oracle females tion (Fig. 2A, B; Table 2; P ϭ 0.05) and lifetime fe- matured sooner than Phoenix beetles when raised at cundity (Fig. 2C, D; Table 2; P Ͻ 0.0001). For example, 24ЊC, whereas Phoenix beetles matured sooner at 36ЊC, Phoenix females that were raised at 36ЊC reached re- though the effect was very small (signi®cant rearing productive maturity later than Phoenix females raised temperature ϫ population interaction; Fig. 1C; Table at 30ЊC except when they laid eggs at 24ЊC (Fig. 2B). 1; P ϭ 0.04). Both of these results are consistent with Likewise, females raised at 24ЊC laid more eggs than our expectation that Oracle beetles are adapted to lower females raised at 30Њ and 36ЊC but Phoenix females temperatures and Phoenix beetles are adapted to higher raised at 30ЊC actually laid slightly more eggs at 36ЊC temperatures. than Phoenix females raised at 24Њ or 36ЊC (Fig. 2D). We also predicted that these two populations would We also detected signi®cant interactions between rear- be genetically differentiated in several life history ing temperature and oviposition temperature for egg ␹2 ϭ ϭ traits. In particular, because most ectothermic animals development (Fig. 3A, B; 4 10.84, P 0.03) and ␹2 ϭ ϭ evolve to lay larger eggs in colder environments (e.g., egg hatching (Fig. 3C, D; 4 16.01, P 0.003) Berven 1982, Azevedo et al. 1996), we expected the indicating large maternal effects. The most dramatic Phoenix population to lay more and smaller eggs (due maternal effect was on egg hatching; when females to a trade-off between these two traits; Smith and Fret- were raised at 24ЊC they produced eggs that hatched well 1974) than the Oracle population. There was no very poorly at 36ЊC suggesting that maternal accli- signi®cant difference in mean fecundity between pop- mation to low temperature comes with a substantial ulations (Table 2; P ϭ 0.46) but contrary to our pre- ®tness cost if they need to lay eggs at high temperature. April 2005 COMPLEX PATTERNS OF PLASTICITY 931

TABLE 3. ANOVAs (Type III sums of squares) with orthogonal polynomial contrasts for the composite measures of ®tness, rC1 and rC2.

Oracle Phoenix Source df F df F

rC1 Rearing temperature (R) 2 69.65*** 2 36.89*** Oviposition temperature (O) 2 22.31*** 2 14.11*** R ϫ O 4 6.66*** 4 5.07*** R(l) 1 37.30*** 1 26.60*** R(q) 1 7.33** 1 1.62 ϫ O R(l) 2 13.21*** 2 8.99*** ϫ O R(q) 2 0.11 2 1.15 Error 144 144

rC2 R 2 23.90*** 2 15.21*** O 2 17.53*** 2 5.39** R ϫ O 4 2.73* 4 1.74 R(l) 1 4.91* 1 2.78² R(q) 1 30.15*** 1 8.00** ϫ O R(l) 2 1.67 2 3.45* ϫ O R(q) 2 3.79* 2 0.03 Error 144 144 Note: Abbreviations: l, linear; q, quadratic. * P Ͻ 0.05; ** P Ͻ 0.01; *** P Ͻ 0.001, ² P Ͻ 0.10.

The maternal effect on egg hatching is not mediated hypothesis was the most strongly supported of the ac- through egg size (interaction effect in logistic regres- climation hypotheses (negative rearing(q), Table 3; Or- ␹2 ϭ ϭ Ͻ ϭ sion when egg size is included as a covariate; 4 acle, F1, 144 30.15, P 0.001; Phoenix, F1, 144 8.00, 15.92, P ϭ 0.0031) indicating that females are in¯u- P Ͻ 0.01); females that developed at 30ЊC have higher encing egg hatching through egg composition. Also, ®tness at all oviposition temperatures over females de- Њ Њ Њ females raised at 24 C had higher ®tness (rC1) when veloping at 24 and 36 C (Fig. 4C, D) due to the large ovipositing at 24Њ and 30ЊC, but lower ®tness when effect of temperature on age-at-®rst-reproduction. ovipositing at 36ЊC compared to females raised at 30Њ However, there was a large negative linear effect of Њ ϫ ϭ and 36 C (highly signi®cant rearing temperature ovi- rearing temperature (Table 3; Oracle, F1, 144 4.91, P Ͻ ϭ Ͻ position temperature interaction; Fig. 4A, B; Table 3; 0.05; Phoenix, F1, 144 2.78, P 0.10) and a sig- P Ͻ 0.001). There was also a suggestion of an inter- ni®cant interaction with oviposition temperature (Table Ͻ ϫ ϭ Ͻ action for rC2 (Table 3; Oracle only, P 0.05); females 3; Oracle, oviposition rearing(q), F2, 144 3.79, P Њ Њ ϫ ϭ that were raised at both 24 and 30 C had higher ®tness 0.05; Phoenix, oviposition rearing(l), F2, 144 3.45, than females raised at 36ЊC, but differences between P Ͻ 0.05), indicating that, although intermediate tem- females raised at all three rearing temperatures was perature was best, ®tness dropped off fastest as tem- small when they oviposited at 36ЊC (Fig. 4C). perature increased (i.e., colder is better hypothesis) and there was some bene®cial effect of acclimation (ben- Acclimation e®cial acclimation hypothesis).

For rC1, the colder is better hypothesis was more strongly supported than the other hypotheses (negative DISCUSSION ϭ Ͻ rearing(l), Table 3; Oracle, F1, 144 37.30, P 0.001; Both rearing and oviposition temperature strongly ϭ Ͻ Phoenix, F1, 144 26.60, P 0.001); females raised at affected growth and life history traits in the seed beetle, 24ЊC had higher ®tness at all oviposition temperatures Stator limbatus. The rearing and oviposition temper- (except 36ЊC) than those raised at 30Њ and 36ЊC (Fig. ature effects were not always additive, nor in the same 4A, B). Even though females developing at lower tem- direction, and thus reaction norms were highly com- perature performed better overall, the relationship be- plex. These complex patterns are important because tween ®tness and temperature was nonlinear for Oracle immature insects are usually subject to different ther- ϭ females (negative rearing(q), Table 3; F1, 144 7.33, P mal environments than adults. Rearing at a cold (rC1; Ͻ 0.01), suggesting that development at an interme- colder is better hypothesis) or intermediate (rC2; opti- diate temperature (optimal developmental temperature mal developmental temperature hypothesis) tempera- hypothesis) also enhances ®tness. There was some sup- ture enhanced ®tness at several adult oviposition tem- port for the bene®cial acclimation hypothesis (ovipo- peratures; acclimation had only a small effect on ®tness ϫ ϭ sition rearing(l), Table 3; Oracle, F2, 144 13.21, P (bene®cial acclimation hypothesis). Furthermore, Ͻ ϭ Ͻ 0.001; Phoenix, F2, 144 8.99, P 0.001). In con- growth traits of the Oracle (high-elevation) and Phoe- trast, for rC2 the optimal developmental temperature nix (low-elevation) populations responded differently 932 R. CRAIG STILLWELL AND CHARLES W. FOX Ecology, Vol. 86, No. 4 to rearing temperature providing evidence that popu- how reaction norm shape is affected by experiences at lations of S. limbatus are differentially adapted to tem- different stages of the life cycle. In this study, we perature. showed that temperature-mediated plasticity of age-at- ®rst-reproduction, fecundity, egg development, egg Adaptation to temperature hatching, and two composite measures of ®tness were Studies of fruit ¯ies have shown that populations affected by interactions between rearing temperature sampled from warm and cool climates exhibit popu- and oviposition temperature. This is particularly im- lation ϫ temperature interactions for survivorship and portant since larvae experience very different thermal development time (Norry et al. 2001, Bochdanovits and environments than adults; S. limbatus larvae are re- de Jong 2003), suggesting that these populations have stricted to individual seeds while adults can move from adapted to their native climates. We found that the high- host to host. Such nonadditive effects of temperature er elevation population (Oracle) of S. limbatus was on traits closely associated with ®tness will complicate larger and matured sooner than the lower elevation pop- efforts to predict evolutionary response to selection ulation (Phoenix) when raised in the colder environ- (Gibert et al. 2001). ment (24ЊC) while the reverse was true in the hotter The interaction between rearing temperature and ovi- environment (36ЊC), consistent with the hypothesis that position temperature affected not only a female's phe- these populations are differentially adapted to temper- notype (age-at-®rst-reproduction and fecundity) but ature. Van der Have and de Jong (1996) argued that also the survivorship of her offspring in early stages genotype ϫ temperature (and by extension population of development (egg development and egg hatching). ϫ temperature) interactions can be due to differential Females raised at low temperature produced eggs that thermal sensitivity of enzymes that control cellular had substantially reduced hatching when those females growth and differentiation. Because the Oracle and oviposited at high temperature (Fig. 3C, D). This large Phoenix populations experience different thermal en- maternal effect on egg hatching was not due to an effect vironments, it is likely that enzymes that control growth of egg size, implying that females reared at colder tem- and differentiation have evolved different thermal sen- peratures are altering egg composition in such a way sitivities, generating the variation in body size and de- that offspring survival is compromised at hotter tem- velopment time we found among populations. How- peratures. Alternatively, cold-acclimated males may ever, further work is needed to reveal the mechanisms produce sperm that are of lower quality at hotter tem- that are responsible for producing these types of pat- peratures. Our study cannot distinguish a paternal ef- terns. fect from a maternal effect. Temperature-mediated pa- Females from the Phoenix population laid larger rental effects in¯uence offspring ®tness in Drosophila. eggs than females from the Oracle population. This For instance, parents raised at higher temperatures pro- contradicted our prediction that the higher elevation duced offspring that had higher ®tness than offspring population should produce larger eggs and is opposite from parents raised at a low or intermediate tempera- to patterns observed in other species, but is most likely ture (Gilchrist and Huey 2001). Also, male ¯ies whose due to host plant adaptation. Natural selection favors parents had developed at a higher temperature had im- larger eggs when S. limbatus females lay on Parkin- proved territorial success over males whose parents had sonia ¯orida, and populations collected from P. ¯orida been raised at a lower temperature (Zamudio et al. have evolved larger eggs than populations collected 1995). Though they are dif®cult to interpret due to their from Acacia greggii (Fox and Mousseau 1996, Fox et complex nature, temperature-mediated cross-genera- al. 2001). That these populations were collected from tional effects are likely to be important in the evolution different host plants (Oracle, A. greggii; Phoenix, P. of life histories (Mousseau and Fox 1998). Therefore, ¯orida) possibly confounds interpretation of popula- future studies should concentrate on the dynamic and tion differences but cannot account for the population complex nature of cross-generation effects of temper- ϫ temperature interaction we observed for body size ature. and development time. If population differences are due to host plant adaptation, then we should have ob- The adaptive signi®cance of acclimation served differences between the populations across all Acclimation is assumed to be bene®cial and ubiq- temperatures as was demonstrated with egg size. uitous in organisms, but most empirical tests of the bene®cial acclimation hypothesis have challenged this Complex patterns of phenotypic plasticity generality (Leroi et al. 1994, Huey et al. 1999, Gibert Interactions between temperatures experienced in et al. 2001). Several studies have supported alternative separate stages of the life cycle can dramatically affect hypotheses such as the optimal developmental tem- reaction norm shape (Gibert et al. 2001). Because nat- perature hypothesis (Huey et al. 1999, Gibert et al. ural selection will act on the entire reaction norm pro- 2001). These studies have been criticized as having duced by the total effect of temperature experienced evaluated the adaptive signi®cance of developmental throughout an individual's life, it is important to know plasticity (irreversible as well as reversible changes to whether these effects are additive or nonadditive and the phenotype induced during development) instead of April 2005 COMPLEX PATTERNS OF PLASTICITY 933 traditionally studied acclimation (reversible changes to hypothesis. For example, development at stressful tem- the phenotype induced in later stages of the life history) peratures can severely impact ®tness, overwhelming and thus are not direct tests of the bene®cial accli- any bene®cial effects of acclimation (Wilson and mation hypothesis (Wilson and Franklin 2002). Al- Franklin 2002). Likewise, development at constant though the distinction between developmental plastic- temperatures (as well as chronic exposure to any en- ity and traditional acclimation is important, focusing vironmental factor) may be stressful for organisms and too narrowly on the life stage in which plastic responses could lead to incorrect evaluations of the bene®cial are induced may obscure the ultimate goal of under- acclimation hypothesis (Woods and Harrison 2002). standing whether environmentally induced changes to The temperatures used in this study were selected be- the phenotype are adaptive (Woods and Harrison 2002, cause they are within the normal range of temperatures Fischer et al. 2003b). encountered in the ®eld but inclusion of other temper- Our results showed that larvae that had developed atures and variation in temperature within treatments at a lower (colder is better hypothesis) or intermediate could produce different results. (optimal developmental hypothesis) temperature had the greatest ®tness as adults. There was some evidence Conclusions to support the bene®cial acclimation hypothesis but in Responses of growth traits of low- and high-eleva- no case was the bene®cial acclimation hypothesis tion populations of the seed beetle, Stator limbatus,to ranked highest among the hypotheses. Therefore, this rearing temperature indicate that these populations study is in agreement with other recent empirical anal- have adapted to their local thermal environments. In yses suggesting that acclimation has only a small effect addition, beetles developing at a low (colder is better on ®tness (Leroi et al. 1994, Huey et al. 1999, Gibert hypothesis) or intermediate temperature (optimal de- et al. 2001). Our rejection of the bene®cial acclimation velopmental temperature hypothesis) had large ®tness hypothesis is not unexpected because the temperatures advantages as adults over beetles that developed at a that larvae experience within seeds are unlikely to be reliable cues for thermal conditions they will experi- hotter temperature; there was little evidence that ac- ence as adults, which may prevent the evolution of climation is adaptive (bene®cial acclimation hypothe- adaptive acclimation. Alternatively, selection may fa- sis). Most importantly, we detected strong interactions vor physiologically robust individuals that have the between the effects of rearing and oviposition temper- ability to perform well in a variety of adult thermal ature on multiple traits, producing complex reaction environments (Gibert et al. 2001). Our analyses were norms and indicating that reaction norm shape is af- based on indirect estimates of ®tness and should thus fected by nonadditive effects of temperature experi- be interpreted with caution. However, a study that mea- enced across life stages. Complexity in plasticity will sured ®tness directly showed that bacteria acclimated make it dif®cult to predict evolutionary responses to to 41.5ЊC had reduced ®tness when tested at 41.5ЊC natural selection, but an understanding of this com- compared to bacteria that had been acclimated to 32ЊC, plexity should yield new insights into the adaptive sig- demonstrating that acclimation is not bene®cial (Leroi ni®cance of plasticity itself. et al. 1994). ACKNOWLEDGMENTS We based our test of the bene®cial acclimation hypothesis on estimates of ®tness but Woods and Harrison We thank K. Carrico, N. Correll, C. Elkins, J. Smith, B. Wallin, and S. Wyatt for help weighing and mating beetles (2002) argue that acclimation can be better understood during the experiment. We also thank C. Rauter and J. Moya- by exploring responses of individual traits. When the LaranÄo for their comments on experimental design and help life history traits were evaluated individually, the hy- on statistical analysis, and A. Amarillo-Saurez, K. Allen, J. pothesis that was best supported by our data varied Moya-LaranÄo, C. Rauter, B. Wallin, and the Evolutionary Ecology Reading Group at the University of Kentucky for among traits (colder is better [fecundity, egg devel- comments on the manuscript. Two anonymous reviewers pro- opment], hotter is better [egg hatching], and optimal vided helpful comments on an earlier version of the manu- developmental temperature [age-at-®rst-reproduction]; script. R. C. Stillwell wishes to dedicate this paper in memory no hypothesis was supported by egg size) but there was of my mother, Linda L. Stillwell, who was killed in a car no support for the bene®cial acclimation hypothesis. accident shortly before I came to graduate school and started this project. This one is for you mom. Financial support was Therefore, an evaluation of acclimation based solely provided by a National Science Foundation grant (NSF DEB- on individual traits is largely uninterpretable because 01-10754) to C. W. Fox. developmental temperature can simultaneously alter a large number of traits in opposite directions. The best LITERATURE CITED solution may be to simultaneously explore the respons- Armbruster, P., W. E. Bradshaw, K. Rueff, and C. H. Hol- es of multiple traits, the relationships between those zapfel. 2001. Geographic variation and the evolution of traits, and their consequences for ®tness. reproductive allocation in the pitcher-plant mosquito, Wyeomyia smithii. Evolution 55:439±444. Temperatures used in experimental designs of accli- Atkinson, D. 1994. Temperature and organism sizeÐa bio- mation studies can have substantial implications for logical law for ectotherms? Advances in Ecological Re- conclusions when assessing the bene®cial acclimation search 25:1±58. 934 R. CRAIG STILLWELL AND CHARLES W. FOX Ecology, Vol. 86, No. 4

Azevedo, R. B. R., V. French, and L. Partridge. 1996. Ther- mental test of the bene®cial acclimation hypothesis. Pro- mal evolution of egg size in Drosophila melanogaster. Evo- ceedings of the National Academy of Sciences (USA) 91: lution 50:2338±2345. 1917±1921. Berven, K. A. 1982. The genetic basis of altitudinal variation Lonsdale, D. J., and J. S. Levinton. 1985. Latitudinal dif- in the wood frog Rana sylvatica. I. An experimental anal- ferentiation in copepod growthÐan adaptation to temper- ysis of life history traits. Evolution 36:962±983. ature. Ecology 66:1397±1497. Bochdanovits, Z., and G. de Jong. 2003. Temperature de- Mousseau, T. A. 1997. Ectotherms follow the converse to pendence of ®tness components in geographical popula- Bergmann's rule. Evolution 51:630±632. tions of Drosophila melanogaster: changing the association Mousseau, T. A., and C. W. Fox. 1998. Maternal effects as between size and ®tness. Biological Journal of the Linnean adaptations. Oxford University Press, New York, New Society 80:717±725. York, USA. Fischer, K., P. M. Brake®eld, and B. J. Zwaan. 2003a. Plas- Nilsson, J. A., and C. D. Johnson. 1993. Laboratory hybrid- ticity in butter¯y egg size: why larger offspring at lower ization of Stator beali and S. limbatus, with new host re- temperatures? Ecology 84:3138±3147. cords for S. limbatus and Mimosestes amicus (Coleoptera: Fischer, K., E. Eenhoorn, A. N. M. Bot, P. M. Brake®eld, and Bruchidae). Southwest Naturalist 38:385±387. B. J. Zwaan. 2003b. Cooler butter¯ies lay larger eggs: Norry, F. M., O. A. Bubliy, and V. Loeschcke. 2001. De- developmental plasticity versus acclimation. Proceedings velopmental time, body size, and wing loading in Dro- of the Royal Society of London B 270:2051±2056. sophila buzzatii from lowland and highland populations in Fox, C. W., and M. E. Czesak. 2000. Evolutionary ecology Argentina. Hereditas 135:35±40. of progeny size in arthropods. Annual Reviews of Ento- Partridge, L., B. Barrie, K. Fowler, and V. French. 1994. mology 45:341±369. Evolution and development of body size and cell size in Fox, C. W., M. E. Czesak, and R. W. Fox. 2001. Conse- Drosophila melanogaster in response to temperature. Evo- quences of plant resistance for herbivore survivorship, lution 48:1269±1276. growth, and selection on egg size. Ecology 82:2790±2804. Partridge, L., and J. A. Coyne. 1997. Bergmann's rule in Fox, C. W., and T. A. Mousseau. 1996. Larval host plant ectotherms: is it adaptive? Evolution 51:632±635. affects the ®tness consequences of egg size in the seed Santos, M., P. F. Iriarte, W. CeÂspedes, J. BalanyaÁ, A. Font- beetle Stator limbatus. Oecologia 107:541±548. devila, and L. Serra. 2004. Swift laboratory thermal evo- Fox, C. W., M. S. Thakar, and T. A. Mousseau. 1997. Egg lution of wing shape (but not size) in Drosophila subob- size plasticity in a seed beetle: an adaptive maternal effect. scura and its relationship with chromosomal inversion American Naturalist 149:149±163. polymorphism. Journal of Evolutionary Biology 17:841± Fox, C. W., K. J. Waddell, and T. A. Mousseau. 1994. Host- 855. associated ®tness variation in a seed beetle (Coleoptera: Savalli, U. M., and C. W Fox. 1998. Sexual selection and Bruchidae): evidence for local adaptation to a poor quality the ®tness consequences of male body size in the seed host. Oecologia 99:329±336. beetle, Stator limbatus. Animal Behaviour 55:473±483. Gibert, P., R. B. Huey, and G. W. Gilchrist. 2001. Locomotor Savalli, U. M., and C. W. Fox. 2002. Proximate mechanisms performance of Drosophila melanogaster: interactions in¯uencing egg size plasticity in the seed beetle, Stator among developmental and adult temperatures, age, and ge- limbatus. Annals of the Entomological Society of America ography. Evolution 55:205±209. 95:724±734. Gilchrist, G. W., and R. B. Huey. 2001. Parental and devel- Smith, C. C., and S. D. Fretwell. 1974. The optimal balance opmental temperature effects on the thermal dependence between size and number of offspring. American Naturalist of ®tness in Drosophila melanogaster. Evolution 55:209± 108:499±506. Sokal, R. R., and J. Rohlf. 1995. Biometry. Third edition. 214. W. H. Freeman, New York, New York, USA. Huey, R. B., and D. Berrigan. 1996. Testing evolutionary Van der Have, T. M., and G. de Jong. 1996. Adult size in hypotheses of acclimation. Pages 205±237 in I. A. Johnston ectotherms: temperature effects on growth and differenti- and A. F. Bennett, editors. Animals and temperature: phe- ation. Journal of Theoretical Biology 183:329±340. notypic and evolutionary adaptation. Cambridge University Van't Land, J., P. Van Putten, H. Villarroel, A. Kamping, and Press, Cambridge, UK. W. Van Delden. 1995. Latitudinal variation in wing length Huey, R. B., D. Berrigan, G. W. Gilchrist, and J. C. Herron. and allele frequencies for Adh and a-Gpdh in populations 1999. Testing the adaptive signi®cance of acclimation: a of Drosophila melanogaster from Ecuador and Chile. Dro- strong inference approach. American Zoologist 39:323± sophila Information Service 76:156. 336. Van Voorhies, W. A. 1996. Bergmann size clines: a simple Huey, R. B., G. W. Gilchrist, M. L. Carlson, D. Berrigan, and explanation for their occurrence in ectotherms. Evolution L. Serra. 2000. Rapid evolution of a geographic cline in 50:1259±1264. size in an introduced ¯y. Science 287:308±309. Van Voorhies, W. A. 1997. On the adaptive nature of Berg- James, A. C., R. B. R. Azevedo, and L. Partridge. 1995. mann size clines: a reply to Mousseau, Partridge and Coy- Cellular basis and developmental timing in a size cline of ne. Evolution 51:635±640. Drosophila melanogaster in laboratory and ®eld popula- Wilson, R. S., and C. E. Franklin. 2002. Testing the bene®cial tions. Genetics 140:659±666. acclimation hypothesis. Trends in Ecology and Evolution Johnson, C. D., and J. M. Kingsolver. 1976. Systematics of 17:66±70. Stator of North and Central America (Coleoptera: Bruchi- Woods, H. A., and J. F.Harrison. 2002. Interpreting rejections dae). USDA Technical Bulletin 1537:1±101. of the bene®cial acclimation hypothesis: when is physio- Johnson, C. D., J. M. Kingsolver, and A. L. Teran. 1989. logical plasticity adaptive? Evolution 56:1863±1866. Sistematica del genero Stator (Insecta: Coleoptera: Bru- Zamudio, K. R., R. B. Huey, and W. D. Crill. 1995. Bigger chidae) en Sudamerica. Opera Lilloana 37:1±105. isn't always betterÐbody size, developmental and parental Leroi, A. M., A. F. Bennett, and R. E. Lenski. 1994. Tem- temperature and male territorial success in Drosophila me- perature acclimation and competitive ®tness; an experi- lanogaster. Animal Behaviour 49:671±677.