Genetic and phenotypic variation in juvenile development in relation to temperature and developmental pathway in a geometrid moth

Authors


Sami M. Kivelä, Department of Biology, University of Oulu, PO Box 3000, 90014 University of Oulu, Finland.
Tel.: +358 (0)8 553 1213; fax: +358 8 553 1061;
e-mail: sami.kivela@oulu.fi

Abstract

Life histories show genetic population-level variation due to spatial variation in selection pressures. Phenotypic plasticity in life histories is also common, facilitating fine-tuning of the phenotype in relation to the prevailing selection regime. In multivoltine (≥ 2 generations per year) insects, individuals following alternative developmental pathways (diapause/direct development) experience different selection regimes. We studied the genetic and phenotypic components of juvenile development in Cabera exanthemata (Lepidoptera: Geometridae) in a factorial split-brood experiment. F2 offspring of individuals originating from populations in northern and central Finland were divided among manipulations defined by temperature (14 °C/20 °C) and day length (24 h/15 h). Short day length invariably induced diapause, whereas continuous light almost invariably induced direct development in both regions, although northern populations are strictly univoltine in the wild. Individuals from northern Finland had higher growth rates, shorter development times and higher pupal masses than individuals from central Finland across the conditions, indicating genetic differences between regions. Individuals that developed directly into adults tended to have higher growth rates, shorter development times and higher pupal masses than those entering diapause, indicating phenotypic plasticity. Temperature-induced plasticity was substantial; growth rate was much higher, development time much shorter and pupal mass higher at 20 °C than at 14 °C. The degree of plasticity in relation to developmental pathway was pronounced at 20 °C in growth rate and development time and at 14 °C in pupal mass, emphasizing multidimensionality of reaction norms. The observed genetic variation and developmental plasticity seem adaptive in relation to time-stress due to seasonality.

Introduction

Intraspecific genetic variation in life histories is common among populations (Roff, 2002). It arises due to spatial (i.e. geographical) variation in the selection regime (Roff, 2002), as there may be geographical variation in external mortality risk (Reznick et al., 1996), for example. There is also large-scale geographical variation in many environmental factors such as temperature and seasonality, because these factors change gradually with latitude and altitude (Angilletta, 2009). Populations along environmental gradients experience different selection pressures, and adaptation to local conditions results in clinal variation in life history traits under selection. In addition to genetic variation in life histories among populations, there is also within-population life history variation that usually is a consequence of phenotypic plasticity. Phenotypic plasticity is common in life histories (Roff, 2002), and it facilitates the adjustment of the phenotype to the prevailing environment, but plasticity may be nonadaptive as well.

Large-scale environmental gradients affect the evolution of local adaptations in key life history traits. Consequently, latitudinal and altitudinal body size clines are pervasive within the animal kingdom (Stillwell, 2010). Body size clines have been repeatedly reported in insects; body size may either decrease or increase with increasing latitude or altitude (reviewed by Mousseau, 1997; Chown & Gaston, 1999, 2010; Blanckenhorn & Demont, 2004; Dillon et al., 2006). The former represents cogradient and the latter countergradient variation in relation to the length of the growing season (sensu Conover & Schultz, 1995) as growing season length is negatively correlated with latitude and altitude. There is a tendency that body size shows cogradient variation (i.e. converse Bergmann cline) in insects that have long generation times in relation to season length (Blanckenhorn & Demont, 2004), because season length constrains their body size via its effect on development time (Masaki, 1967; Roff, 1980; Iwasa et al., 1994; Chown & Gaston, 1999, 2010; Blanckenhorn & Demont, 2004). Cogradient variation with season length is typically observed in development times of species having long generation times (Masaki, 1967; Telfer & Hassall, 1999; Kiveläet al., 2011), whereas increasing evidence suggests countergradient variation in growth rate in such species (Telfer & Hassall, 1999; Blanckenhorn & Demont, 2004; De Block et al., 2008; Pöykkö & Tammaru, 2010; Kiveläet al., 2011). Countergradient variation in growth rate may evolve as a compensation for the decrease in season length (Blanckenhorn & Demont, 2004), or it may represent a correlated response to selection affecting development time, because there is a strong negative genetic correlation between growth rate and development time (Kiveläet al., 2011). Countergradient variation in body size (i.e. Bergmann cline) is usually found in species with short generation times in relation to season length (Blanckenhorn & Demont, 2004), because environmental factors such as temperature are expected to inflict stronger selection on life histories of these species than season length (Chown & Gaston, 1999). Development times of species having short generation times show cogradient variation with season length (James & Partridge, 1995; Blanckenhorn & Demont, 2004), which, combined with body size variation, suggests countergradient variation in growth rate.

Phenotypic plasticity may generate within-population life history variation. A well-known example of phenotypic plasticity in ectotherms is the temperature–size rule, the thermal reaction norm of body size where body size decreases with increasing temperature (Atkinson, 1994; Atkinson & Sibly, 1997). The adaptive significance of the temperature–size rule is controversial (Atkinson & Sibly, 1997), and not all insects follow the rule (Walters & Hassall, 2006; Diamond & Kingsolver, 2010; see also Pöykkö & Tammaru, 2010). Another well-documented example of phenotypic plasticity in insect life histories is diapause induction, and the consequent two alternative developmental pathways: diapause and direct development into the adult stage. Diapause is induced by environmental factors, mainly by photoperiod, that indicate the phase of the seasonal cycle (Tauber et al., 1986; Danks, 1987). Because day length on a particular date during the growing season becomes longer with latitude, the genotype-specific critical day length of diapause induction (50% of individuals enter diapause) lengthens with latitude as well (Tauber et al., 1986; Danks, 1987; Gomi, 1997; see also Mousseau & Roff, 1989). Diapause induction determines voltinism (the number of generations emerging per season) of an insect population.

The induction of the developmental pathway is associated with changes in life histories, apparently because individuals following the alternative developmental pathways experience different selection pressures. Generally, offspring of individuals that have developed directly into adults have less time to reach the species-specific overwintering developmental stage before conditions turn adverse than the diapausing offspring of overwintered individuals. This sets time constraint for the direct development particularly when bivoltine phenology is only partial, that is, only a part of the offspring of overwintered individuals develops directly into adults. Owing to this, individuals that develop directly into adults have a shorter juvenile development than diapausing individuals (Wiklund et al., 1991; Nylin, 1992; Blanckenhorn & Fairbairn, 1995; Wiklund & Friberg, 2011; Pöykkö & Hyvärinen, 2012) as it increases the time that is available for their offspring to reach the overwintering stage. Direct development is also associated with a higher growth rate than the diapause pathway (Wiklund et al., 1991; Nylin, 1992; Pöykkö & Hyvärinen, 2012), which may result in larger (Wiklund et al., 1991), smaller (Mousseau & Roff, 1989; Teder et al., 2010; Pöykkö & Hyvärinen, 2012) or equal (Blanckenhorn, 1994; Blanckenhorn & Fairbairn, 1995) body size compared with the diapause pathway, depending on the magnitude of change in development time and growth rate. In addition to variable time constraints, life history differences between the alternative pathways may, at least partially, be generated by different environmental conditions experienced during development. This is evident as predictable variation in environmental conditions is known to affect the optimal and expressed body size (Gotthard et al., 2007; Gotthard & Berger, 2010).

Despite the well-documented life history difference between the alternative developmental pathways, the environmental sensitivity of the difference is poorly understood due to paucity of data (Wiklund et al., 1991; Blanckenhorn & Fairbairn, 1995). Furthermore, almost nothing is known about the possible genetic population-level variation in the reaction norms of growth and development in relation to the developmental pathway and environment [see Blanckenhorn & Fairbairn (1995) for development time and body size]. Finally, the role of temperature in the expression of alternative life histories deserves more attention as temperature-induced plasticity is common, and temperature varies both spatially and temporally (Angilletta, 2009). In this study, we address these issues.

We analysed both the genetic and phenotypic components of variation in juvenile growth and development in the moth Cabera exanthemata (Scop. 1763) (Lepidoptera: Geometridae). We documented the induced developmental pathway and measured larval development time, growth rate and attained pupal mass of individuals originating from two geographical regions in relation to temperature and day length in a factorial split-brood experiment. Any region-level variation in the phenotype within a particular manipulation would indicate genetic differentiation between geographically distinct populations, whereas within-brood variation among experimental manipulations would indicate phenotypic plasticity. According to the theory (Roff, 1980; Iwasa et al., 1994), we predict that the genetic component of variation gives rise to season length–related cogradient variation in both development time and body size, and countergradient variation in growth rate. Phenotypic plasticity is predicted to be multidimensional. First, increasing body size associated with decreasing temperature (i.e. temperature–size rule) would be expected in Lepidoptera within a wide range of temperatures (see Davidowitz & Nijhout, 2004). Second, short day length, indicative of oncoming winter, should result in a relatively short development time and a high growth rate (Leimar, 1996; Gotthard, 2008). Third, direct development should be associated with a relatively short development time, high growth rate and small body size, because the direct generation is partial when it emerges in the studied regions, and so directly developing individuals are under more intense time-stress than diapausing individuals (Wiklund et al., 1991; Kiveläet al., 2011; see above). A genetic component in the propensity to enter direct development would also be expected; the propensity to enter direct development should decrease northwards. Possible region-level variation in the genetic architecture underlying the studied traits would result in geographical variation in reaction norms, generating interactions including region. Owing to this, geographical variation in diapause propensity might be associated with developmental pathway by region interactions. For the above reasons, this study was able to capture the expected multidimensional variation in growth and development.

Materials and methods

Cabera exanthemata is a widespread Palearctic moth (Lepidoptera: Geometridae) feeding mainly on willows (Salix spp.) (Mikkola et al., 1989). Adult nectar feeding is rare (Mikkola et al., 1989; own observations), which together with the large number of mature eggs at eclosion (own observations) places the species close to the capital end of the income vs. capital breeder continuum. Diapause is restricted to the pupal stage, but the species has the capability to develop directly into the adult stage within the same season [i.e. to express bivoltine (2 generations per year) phenology] (Mikkola et al., 1989; see also Results).

Adult females were captured from the field with a net from four locations (i.e. populations), two of which were in central (65 °2′–65 °6′N 25 °23′–25 °24′E and 65°0′N 25 °13′E; hereafter referred to as south) and another two in northern (67 °56′N 23 °39′E and 67 °24′N 26 °35′E; hereafter referred to as north) Finland. The phenology of C. exanthemata is strictly univoltine in the north, whereas a small nondiapause generation may emerge in the south, although the species is predominantly univoltine there as well (Mikkola et al., 1989; Hyönteistietokanta, 2011; own observations). Sampling was conducted during the flight period of the diapause generation in June and July.

Wild-caught females (7 and 8 females per population in the north; 5 and 6 females per population in the south) were allowed to lay eggs on Salix phylicifolia L. in a laboratory. Neonate larvae of each female (F1 generation) were placed into a 0.75-L plastic container (10–15  individuals per container) of their own. The containers had moist garden peat at the bottom and were provided with fresh S. phylicifolia leaves daily. To induce diapause, the larvae were reared under a short photoperiod of 8 h : 16 h (light : dark) at 20 ± 1 °C. After pupation, unearthed pupae were covered with moist moss (Sphagnum spp.). After ca. 9 months of hibernation at a dark refrigerator (5 °C), the pupae were placed individually into 0.2-L plastic containers with moist Sphagnum moss and exposed to a temperature of 20 °C and continuous light. Eclosed females were allowed to reproduce in 0.75-L containers (1 female per container) with moist garden peat at the bottom and leaves of S. phylicifolia. To ensure mating, females were accompanied by 1–2 virgin males from the same population. The males were not siblings of the female, but siblings to each other when two males were used.

Sixteen neonate larvae of each F1 generation female [F2 generation; Nnorth = 12 (6 females per population), Nsouth = 10 (5 females per population)] were placed individually into 0.2-L plastic containers with moist garden peat at the bottom and a fresh S. phylicifolia leaf. Offspring of each female were randomly distributed among four manipulations in a factorial design with two temperatures (14 and 20 °C) and two photoperiods [24 h : 0 h and 15 h : 9 h (light : dark)], so that four offspring were subjected to each of the four manipulations. The temperature manipulation simulated conditions during cool and warm summers, and the two photoperiods midsummer and autumn conditions, respectively. The containers were monitored daily, and fresh host plant leaves were provided and water was added to maintain humidity when necessary. Pupation was recorded once a larva burrowed into the peat. Five days later, the pupa was weighed to the nearest 0.01 mg with a precision balance (Mettler Toledo MT 5) and sexed based on the cuticular genital scars (see Scoble, 1992). The pupae were maintained under the same conditions as larvae. Individuals that did not eclose within a month were recorded as diapausing ones.

Larval growth rate was estimated using the growth model described by Tammaru & Esperk (2007). Within their model, an index of growth rate, c, can be calculated as

image(1)

where mpupa is pupal mass, tlarva is larval development time and b is an allometric exponent relating anabolism to body mass. We used a b value of 0.8, which seems to be a reasonable estimate for a species of this size (Tammaru & Esperk, 2007) and describes the growth rate variation among C. exanthemata individuals well (Kiveläet al., 2011). To test the sensitivity of results, the analysis of growth rate was repeated with b = 0.7 and b = 0.9.

Statistical analyses

Statistical analyses were performed with R 2.10.1 (R Development Core Team, 2009). Variation in pupal mass, larval development time and growth rate was analysed in relation to temperature (14 °C/20 °C), photoperiod (24 h : 0 h/15 h : 9 h), region (north/south), sex and developmental pathway (diapause/direct development) with linear mixed-effects models [function lme (Pinheiro et al., 2009)] fitted with the maximum likelihood method. Random effects were initially defined so that random families were nested within random populations, but variance at the population level was negligible in all cases. Thus, population was removed from the random effects [this was supported by Akaike information criterion (AIC)]. Next, we investigated the need to include family-level random effects that are correlated with temperature and photoperiod manipulations, the best random effect structure being determined on the grounds of AIC. As photoperiod almost completely determined the developmental pathway induced (see Results), both photoperiod and developmental pathway contain almost the same information. To avoid problems in model fitting and inference, we did not include both photoperiod and developmental pathway as fixed effects in the models, but included only developmental pathway as comparison between the developmental pathways was one of our objectives. Fixed effects were therefore initially defined so that they included all main effects and all possible interactions of sex, developmental pathway, temperature and region. After the best structure of random effects was found, fixed effects were reduced according to the principle of hierarchy. As larval growth rates in the temperature manipulations differed by an order of magnitude, growth rate was standardized for the analyses by subtracting the mean and by dividing with the standard deviation.

The need to model heteroscedasticity by adding a variance function to the model was analysed after the structures of both random and fixed effects were determined. Heteroscedasticity was first investigated visually from residual plots, and then the need to model it was analysed by adding an appropriate variance function to the model. In each case, model goodness-of-fit (measured with AIC) improved significantly when a varIdent (Pinheiro et al., 2009) variance function was added to the model. Different variances were allowed for different photoperiods in pupal mass (more variance under continuous light), for different temperatures in larval development time (more variance at 14 °C) and for the four combinations of different photoperiods and temperatures in larval growth rate (variance decreased with decreasing temperature and photoperiod). Model goodness-of-fit was assessed by visual investigation of residual plots.

Results

Diapause induction

Photoperiod significantly affected diapause induction (inline image = 295.3, P < 0.0001). Under the short day length (15 h), all individuals entered diapause in both regions, whereas almost all individuals from both regions developed directly into adults under continuous light (Fig. 1). Thus, there was no region-level variation in diapause induction in relation to the examined day lengths.

Figure 1.

 Diapause induction in relation to day length in the north (left) and in the south (right). Black bars represent individuals that entered diapause, and grey bars individuals that developed directly into adults.

Genetic variation in growth and development

There was significant region-level (genetic) variation in larval development time and larval growth rate, but not in pupal mass (Table 1). Larval development times were slightly shorter and larval growth rates higher in the north than in the south (Table 2; Fig. 2). Even though there was a pattern of higher pupal masses in the north compared with the south (Fig. 2), the difference (see Table 2) did not reach statistical significance in the mixed-model analysis, and region did not enter the final model (Table 1). The difference is, however, significant in a pairwise comparison (Welch’s t-test: t319.917 = 2.56, P = 0.0109).

Table 1.   Analysis of fixed effects in larval development time, larval growth rate and pupal mass in relation to sex, developmental pathway, temperature and region.
TraitFixed effectdf1df2FP
  1. *Random intercepts for families. Heteroscedasticity modelled with varIdent variance function (Pinheiro et al. 2009) allowing different variances for different temperatures.

  2. †Random intercepts for families. Heteroscedasticity modelled with varIdent variance function (Pinheiro et al. 2009) allowing different variances for each of the four combinations of temperature and photoperiod. Standardized values were used in the analysis.

  3. ‡Random intercepts and random temperature and photoperiod effects for families. Heteroscedasticity modelled with varIdent variance function (Pinheiro et al. 2009) allowing different variances for different photoperiods.

Larval time*Sex130053.1< 0.0001
Developmental pathway130055.0< 0.0001
Temperature13005959< 0.0001
Region1206.100.023
Developmental pathway × temperature13004.740.030
Growth rate†Developmental pathway130062.6< 0.0001
Temperature13002585< 0.0001
Region12012.40.0022
Temperature × region130015.50.0001
Developmental pathway × temperature130065.8< 0.0001
Pupal mass‡Sex129879.3< 0.0001
Developmental pathway129825.0< 0.0001
Temperature1298160< 0.0001
Developmental pathway × temperature129820.3< 0.0001
Sex × temperature129813.50.0003
Sex × developmental pathway129821.1< 0.0001
Table 2.   Means and 95% confidence intervals of larval development time, growth rate and pupal mass in different subsets of the data.
Trait (units)Mean (95% CI)
FemaleMaleNorthSouth14 °C20 °CDiapause pathwayDirect pathway
Development time (days)23.9 (22.9, 24.8)22.8 (22.0, 23.6)23.0 (22.2, 23.8)23.6 (22.6, 24.5)28.8 (28.5, 29.1)18.1 (17.9, 18.3)23.6 (22.8, 24.4)22.8 (21.9, 23.7)
Growth rate (SD units)−0.054 (−0.220, 0.112)0.042 (−0.103, 0.188)0.073 (−0.083, 0.230)−0.087 (−0.237, 0.063)−0.937 (−0.946, −0.928)0.870 (0.781, 0.960)−0.188 (−0.311, −0.066)0.203 (0.023, 0.382)
Pupal mass (mg)71.3 (69.4, 73.2)63.4 (62.1, 64.7)68.3 (66.6, 69.9)65.2 (63.5, 66.9)62.1 (60.5, 63.8)71.3 (69.9, 72.7)62.3 (60.9, 63.8)71.8 (70.2, 73.4)
Figure 2.

 Larval development time (a), larval growth rate (b) and pupal mass (c) in females (closed symbols and continuous lines) and males (open symbols and dashed lines) in relation to region. Whiskers indicate 95% confidence intervals.

Phenotypic plasticity in growth and development

There was significant phenotypic plasticity in larval development time, growth rate and pupal mass in relation to the developmental pathway induced (significant main effect of developmental pathway, Table 1). Individuals that developed directly into adults tended to have shorter development times and higher growth rates, and they attained higher pupal masses than individuals entering diapause (Table 2; Fig. 3). The degree of plasticity depended on temperature (significant developmental pathway × temperature interaction, Table 1). There was more plasticity at 20 °C than at 14 °C in both larval development time [developmental pathway (direct development) × temperature (20 °C): estimate =  −0.602 days, t300 = −2.18, P = 0.030] and growth rate [developmental pathway (direct development) × temperature (20 °C): estimate = 0.594 SD units, t300 = 8.11, P < 0.0001] (Fig. 3). On the contrary, there was more plasticity at 14 °C than at 20 °C in pupal mass [developmental pathway (direct development.) × temperature (20 °C): estimate = −6.02 mg, t298 = −4.22, P < 0.0001] (Fig. 3).

Figure 3.

 Larval development time (top row), larval growth rate (middle row) and pupal mass (bottom row) in females (closed symbols, continuous lines) and males (open symbols, dashed lines) in relation to temperature (left column: 14 °C, right column: 20 °C) and developmental pathway. Whiskers indicate 95% confidence intervals. Note the remarkable scale difference between the y-axes of the growth rate plots.

There was also much phenotypic plasticity in relation to temperature as such (significant main effect of temperature, Table 1). Individuals reared at 14 °C had much longer development times and much lower growth rates, and they attained lower pupal masses than individuals reared at 20 °C (Table 2; Fig. 4). In larval growth rate, there was also a significant temperature × region interaction (Table 1), indicating that the difference between the regions was pronounced at 20 °C [temperature (20 °C) × region (south): estimate = −0.257 SD units, t300 = −3.86, P = 0.0001] (Fig. 4). In pupal mass, there was a significant sex × temperature interaction (Table 1), because sexual size dimorphism was pronounced at 20 °C compared with 14 °C [sex (male) × temperature (20 °C): estimate = −4.84 mg, t298 = −3.76, P = 0.0002] (Fig. 3). In pupal mass, there was also a significant sex × developmental pathway interaction (Table 1), indicating that sexual size dimorphism was pronounced in the directly developing generation [sex (male) × developmental pathway (direct development): estimate =  −6.77 mg, t298 = −4.59, P < 0.0001] (Fig. 3). Sexual dimorphism was found in larval development time and pupal mass, females having longer development times and higher pupal masses than males (Table 2; Figs 2 and 3).

Figure 4.

 Larval development time (a), larval growth rate (b) and pupal mass (c) in individuals of northern (closed symbols and continuous lines) and southern (open symbols and dashed lines) origin in relation to temperature. Whiskers indicate 95% confidence intervals.

The analysis of growth rate variation was relatively robust for the value of b used. When b = 0.7 was used, the same final model was found as with b = 0.8, and the qualitative results remained unchanged. When b = 0.9 was used, main effect of sex and sex × region interaction entered the final model in addition to the terms presented in Table 1. With b = 0.9, males had significantly higher growth rates than females [females, −0.102 (95% CI, −0.253 to 0.050) SD units; males, 0.079 (95% CI, −0.075 to 0.234) SD units] and the difference between the sexes was pronounced in the north [sex (male) × region (south): estimate = −0.00336 SD units, t298 = −2.07, P = 0.040].

Discussion

There was region-level variation in larval development times and growth rates, indicating genetic differentiation between the geographical regions. Contrary to predictions, there was no region-level variation in propensity to enter direct development. Within-brood variation in the measured life history traits was remarkable in relation to temperature, photoperiod and developmental pathway, indicating phenotypic plasticity, the degree of plasticity in relation to developmental pathway being sensitive to temperature. Growth rates of individuals of northern and southern origin responded differently to temperature, which indicates a genotype by environment interaction in this trait. However, there was no region (i.e. genotype sensu lato) by developmental pathway interactions, which would have been predicted.

Induction of the developmental pathway

There was no region-level variation in diapause induction in relation to the photoperiodic manipulations applied, and almost all northern individuals deriving from strictly univoltine populations entered direct development under continuous light. A low propensity to enter direct development in the north would have been expected as local adaptations in the critical day length of diapause induction are necessary for an appropriate phenological response. Therefore, the critical day length is found to be positively correlated with latitude (Tauber et al., 1986; Danks, 1987; Gomi, 1997; see also Mousseau & Roff, 1989). A likely reason for our result is that the critical day length of diapause induction falls between the 2 day lengths used.

Continuous light indicates midsummer, especially in the northern region that is north of the Arctic Circle. As the oviposition period of overwintered females coincides with midsummer, the larval stage that is sensitive to the induction of developmental pathway will be reached in late summer in the wild. This is because the induction of developmental pathway takes place late in the larval stage in the lepidopteran species studied in that respect (Friberg et al., 2011 and references therein). Hence, an individual that would be at the sensitive stage of larval development at midsummer would probably have enough time to produce surviving offspring during the same summer. Owing to this, there is apparently no selection against the direct development pathway in the north, because past selection has moved the critical day length to such an extreme value that the direct development pathway is never expressed under natural conditions. On the other hand, a day length of 15 h occurs in late August in the south and in early September in the north. Autumn frosts before these dates are not exceptional, and thus the day length of 15 h indicates absolutely too late date for direct development in both regions.

Region-level genetic differentiation

The observed cogradient variation in development time and countergradient variation in growth rate in relation to season length are in accordance with earlier data on insects (Masaki, 1967; Telfer & Hassall, 1999; Blanckenhorn & Demont, 2004; De Block et al., 2008; Pöykkö & Tammaru, 2010), including the species studied here (Kiveläet al., 2011). These findings strongly suggest that the time constraint due to seasonality affects the evolution of these traits. As season length becomes shorter northwards, the time available for growth is shorter in the north than in the south. This, combined with fecundity selection for large body size favouring long development time (see Honěk, 1993; Roff, 2002), is predicted to translate into northward decrease in development time (Roff, 1980; Iwasa et al., 1994), that is, into cogradient variation. Countergradient variation in growth rate may arise as a correlated response to the evolution of development time (see Kiveläet al., 2011), or as a compensation for decreasing development time to prevent decrease in body size and fecundity (Blanckenhorn & Demont, 2004).

Body size (pupal mass) was slightly larger in the north than in the south, which is opposite to the predicted response of smaller body size in the north (see Roff, 1980; Iwasa et al., 1994). Although this difference did not reach statistical significance in the mixed-model analysis, confidence intervals and a pairwise comparison suggest a difference between the regions (see Results). These findings suggest that the northward increase in growth rate overcompensates for the northward decrease in development time (sensu Blanckenhorn & Demont, 2004). However, it would be premature to conclude that body size of C. exanthemata shows countergradient variation in relation to season length (i.e. follows a Bergmann cline) in general. This is because the latitudinal body size cline in C. exanthemata is complex (Kiveläet al., 2011), and only two distinct regions were compared in this study. If the same regions (and populations) that we refer here as to north and south are compared pairwise in the data of Kiveläet al. (2011), the result is the same, body size being larger in the north than in the south [Welch’s t-test: t225.736 = 3.74, P = 0.00024; north: mean = 72.5 (95% CI, 70.9–74.0) mg; south: mean = 68.4 (95% CI, 66.9–69.9) mg]. Hence, the present data should not be used in inferring clinal variation in any of the measured traits, but only the pattern and degree of co- vs. countergradient variation between the studied regions. Furthermore, this study rigorously indicates the presence of countergradient variation in growth rate as growth rate was higher in the north than in the south at both temperatures (cf. Conover & Schultz, 1995).

Phenotypic plasticity

All the measured traits were phenotypically plastic in relation to developmental pathway and temperature. The induction of developmental pathway does not only affect the diapause phenotype (i.e. whether diapause is expressed or not) but also affect other key life history traits. In accordance with many earlier observations, the direct development pathway was associated with a short development time (Wiklund et al., 1991; Nylin, 1992; Blanckenhorn & Fairbairn, 1995; Wiklund & Friberg, 2011; Pöykkö & Hyvärinen, 2012), high growth rate (Wiklund et al., 1991; Nylin, 1992; Pöykkö & Hyvärinen, 2012) and large body size (Wiklund et al., 1991) compared with the diapause pathway. These responses seem adaptive in relation to the time constraint due to season length in populations with partially bivoltine phenology. The early maturation of individuals following the direct development pathway prolongs the time that their (diapausing) offspring can invest in growth and development before the onset of adverse conditions. Hence, it is understandable that a relatively short development time and a high growth rate in the direct development pathway seem to be general responses. We are not aware of any counterexamples in respect of this, although the directly developing generation would on theoretical grounds be expected to have a longer development time than the diapause generation under certain conditions when egg is the overwintering developmental stage (Iwasa et al., 1994; Roff, 2002). An alternative explanation for the difference in development times between diapausing and directly developing individuals would be temporal variation in predation risk (Remmel et al., 2009; see also Teder et al., 2010). If directly developing individuals experience higher predation risk than diapausing ones, a shorter development time would be favoured in the direct development pathway as it decreases the time of exposure to predators. This mechanism would require a completely bivoltine phenology, where the larval periods of directly developing and diapausing individuals do not overlap. Thus, it is an unlikely explanation in this case. Even so, directly developing individuals may experience a higher mortality risk than diapausing ones as a cost of high growth rate (Gotthard, 2000), which favours short development time in the direct development pathway.

Body size was larger in the direct development pathway than in the diapause pathway, although the opposite pattern was predicted. Existing data are mixed in respect of the body size difference between the pathways (Mousseau & Roff, 1989; Wiklund et al., 1991; Blanckenhorn, 1994; Blanckenhorn & Fairbairn, 1995; Teder et al., 2010; Pöykkö & Hyvärinen, 2012). Also, the fecundity difference between the pathways may be in any direction (Spence, 1989; Blanckenhorn, 1994; Karlsson & Johansson, 2008; Karlsson et al., 2008; Larsdotter Mellström et al., 2010), which is understandable assuming that fecundity is strongly correlated with body size (Honěk, 1993; Roff, 2002). It is realistic to assume that high fecundity is selected for (e.g. Roff, 2002). Thus, the lack of consistency in the body size difference between the pathways suggests that the magnitude and direction of variation in body size is a nonadaptive consequence of the consistent and adaptive responses in development time and growth rate. Accordingly, the magnitude of change in development time and growth rate in relation to each other would determine the change in body size and explain the mixed observations. The nonadaptive nature of body size variation between the pathways may be illustrated with the observed plastic effects of temperature. At both temperatures, direct development was associated with a relatively short development time compared with the diapause pathway, but the reduction in development time was only ca. 1% at 14 °C, whereas it was ca. 5% at 20 °C. However, growth rate showed considerable plasticity in relation to developmental pathway at both temperatures, direct development increasing growth rate ca. 30% at 14 °C and ca. 39% at 20 °C. As a consequence, direct development was associated with ca. 22% increase in body size compared with the diapause pathway at 14 °C, whereas the increase was only ca. 9% at 20 °C. According to these results, phenotypic differences between the alternative pathways are very sensitive to the environment (see also Wiklund et al., 1991), body size being an extremely sensitive trait where even qualitative changes may be possible.

Although the degree of phenotypic plasticity in relation to developmental pathway was dependent on temperature, there was plasticity in relation to temperature as such in all measured traits. Development times were much longer and growth rates much lower at 14 °C than at 20 °C, apparently because of temperature sensitivity of ectotherm metabolic and development rates (Angilletta, 2009). Temperature sensitivity of metabolic and development rates may also explain the reduced plasticity in relation to developmental pathway at 14 °C compared with 20 °C in both development time and growth rate. On the other hand, body size of many, if not most, ectotherms decreases with increasing temperature in accordance with the temperature–size rule (Atkinson, 1994; Atkinson & Sibly, 1997). Reversal of the rule, which we observed, has been found in at least three other insects as well (Walters & Hassall, 2006; Diamond & Kingsolver, 2010; see also Pöykkö & Tammaru, 2010). Our findings may be explained with a physiological model, predicting a humped thermal reaction norm of body size in another lepidopteran, Manduca sexta (Davidowitz & Nijhout, 2004). The model predicted that body size increases with increasing temperature at low temperatures, but there is a point where the pattern reverses, giving rise to the temperature–size rule at high temperatures. The relatively small body size in C. exanthemata at 14 °C suggests a thermal regime within the range of temperatures where the reversal of the temperature–size rule would be expected. This is emphasized by the very low growth rate, which implies that 14 °C was close to the lower thermal limit for growth. With the present data, it is impossible to assess on which side of the hump of the reaction norm 20 °C was, or whether the thermal reaction norm of body size is humped in C. exanthemata. An alternative explanation is the environment dependency of the expression of the temperature–size rule or its reversal (Diamond & Kingsolver, 2010).

Genetic variation in phenotypic plasticity, appearing as a region by developmental pathway interaction, would have been expected. This is because only the diapause pathway is expressed under natural conditions in the north, but it is occasionally expressed in the south. Consequently, direct development pathway is not under selection in the north, whereas it may, at least occasionally, be under selection in the south. Relaxed selection on the direct development pathway would be expected to result in fluctuation in the expression of traits under direct development, but the results do not indicate this. Although there were no region by developmental pathway interactions, there was a region by environment interaction in growth rate. This interaction arose, because the between-region difference in growth rate was pronounced at 20 °C, emphasizing countergradient variation in growth rate.

Conclusions

It is intriguing that individuals from populations where the direct development pathway is never expressed had a high propensity to enter direct development under the simulated midsummer conditions. This indicates that part of the reaction norm may be concealed from natural selection, because certain range of environmental conditions is never experienced in the wild. The nonexpressed parts of the reaction norms could be affected by genetic drift, but we did not find evidence for this as there were no region by developmental pathway interactions. The region-level consistency of reaction norms in relation to developmental pathway suggests that some developmental mechanism or constraint maintains the reaction norms. The nonexpressed parts of the reaction norms in the northern study populations may be beneficial, if changing conditions (e.g. global warming) facilitate a bivoltine phenology at some point.

Phenotypic differences between the alternative developmental pathways are sensitive to environment, which emphasizes that the multivariate nature of the reaction norms should be taken into account when generalizing the results. The slope of the reaction norm in relation to the developmental pathway axis seems to be sensitive to the environment, which may explain the mixed empirical results concerning body size.

Acknowledgments

We are grateful to Heikki Pöykkö and Marko Mutanen for commenting on an earlier version of the manuscript and David Carrasco for his help in maintenance of the P generation in the laboratory. This study was financed by the Ella and Georg Ehrnrooth foundation (grant to S.M.K & P.V.), the Jenny and Antti Wihuri foundation (S.M.K.) and Emil Aaltonen foundation (P.V.). All Finland’s guidelines and legal requirements for the use of animals in research were followed.

Data deposited at Dryad: doi: 10.5061/dryad.3ck64027

Ancillary