Rapid evolution of larval life history, adult immune function and flight muscles in a poleward-moving damselfly

Authors


  • Data deposited at Dryad: doi:10.5061/dryad.4n704

Abstract

Although a growing number of studies have documented the evolution of adult dispersal-related traits at the range edge of poleward-expanding species, we know little about evolutionary changes in immune function or traits expressed by nondispersing larvae. We investigated differentiation in larval (growth and development) and adult traits (immune function and flight-related traits) between replicated core and edge populations of the poleward-moving damselfly Coenagrion scitulum. These traits were measured on individuals reared in a common garden experiment at two different food levels, as allocation trade-offs may be easier to detect under energy shortage. Edge individuals had a faster larval life history (growth and development rates), a higher adult immune function and a nearly significant higher relative flight muscle mass. Most of the differentiation between core and edge populations remained and edge populations had a higher relative flight muscle mass when corrected for latitude-specific thermal regimes, and hence could likely be attributed to the range expansion process per se. We here for the first time document a higher immune function in individuals at the expansion front of a poleward-expanding species and documented the rarely investigated evolution of faster life histories during range expansion. The rapid multivariate evolution in these ecological relevant traits between edge and core populations is expected to translate into changed ecological interactions and therefore has the potential to generate novel eco-evolutionary dynamics at the expansion front.

Introduction

Global warming is triggering poleward range expansions in many species (Chen et al., 2011). Models predict an increase in dispersal ability in the newly founded populations at the range edge in response to in situ natural selection (Travis & Dytham, 2002) and/or in response to spatial sorting whereby the best dispersers accumulate at the range front (Shine et al., 2011). Linked to these models, a growing number of studies have documented the evolution of adult traits at the range edge that are directly associated with locomotory ability (reviewed in Hill et al., 2011), primarily focusing on locomotory morphology (e.g. Hill et al., 1999; Phillips et al., 2006) in the adult stage.

Less well studied are changes in fitness-related traits in adults that are not directly associated with locomotory ability. An important fitness-related trait is immune function (Sheldon & Verhulst, 1996), and to our knowledge, empirical studies investigating immune function in poleward-expanding species are lacking. An increased investment in immune function in edge populations can be expected for several reasons. More actively dispersing individuals may be more likely to be exposed to a broader range of parasites and pathogens (Møller & Erritzøe, 2001; Barber & Dingemanse, 2010; but see Llewellyn et al., 2012). Empirical and theoretical studies also showed that dispersers are typically these individuals exhibiting the best condition (Bonte & Saastamoinen, 2012) and immune function is often positively correlated with individual condition (Yang et al., 2007). Related to this, the ability to disperse can be reduced in animals with a higher pathogen load (e.g. Nagel et al., 2010), so that an increased immune function may result indirectly from selection for increased dispersal rates.

We also know very little about evolutionary changes in fitness-related traits in the nondispersing larval stage. During range expansion, a faster life history (faster larval growth and development) is predicted to evolve in edge populations as genotypes capable of completing a greater number of generations in a given amount of time will be favoured by selection at the low initial population densities in populations at the expansion front (Phillips, 2009; Phillips et al., 2010). The only study that explored this documented faster growth rates in the tadpoles of the cane toad (Phillips, 2009).

Knowledge on the evolution of adult and larval traits that are not directly related to an individuals’ dispersal ability may give a more complete picture on the evolutionary and ecological dynamics in edge populations. Firstly, these traits may be involved in trade-offs with dispersal-related traits and therefore may affect the evolution of the latter traits (Burton et al., 2010). Such trade-offs are expected to occur, for example, between energetically costly traits such as flight morphology and immunity (Bonte et al., 2012). Secondly, because these traits are known to shape antagonistic interactions (competition, predation and parasitism) (e.g. Stoks et al., 2005; Beckerman et al., 2010), they will be necessary to study in order to understand the ecological consequences of evolutionary trait changes in edge populations.

To obtain a multifaceted picture of trait evolution during range expansion, we investigated a set of ecologically relevant adult and larval traits in replicated edge and core populations of the currently poleward-moving damselfly Coenagrion scitulum. Specifically, we studied three different trait sets during a common garden rearing experiment from the egg stage: larval life history (larval development time, larval growth rate), adult immune function and flight-related traits (relative fat content, relative flight muscle mass, body mass and body size). Based on the above-mentioned theoretical and empirical studies, we expected in the edge populations a faster larval life history, a higher adult immune function and a higher investment in flight-related traits. We reared larvae at low and high food levels as allocation trade-offs may be easier to detect under energy shortage (Reznick et al., 2000). Together with selection forces associated with the process of range expansion, traits may at the same time evolve in response to shorter growth seasons (Conover & Schultz, 1995; De Block et al., 2008) and lower environmental temperatures (Hassall et al., 2008) at higher latitudes. Therefore, we corrected our analyses for possible latitudinal influences by reconstructing latitude-specific thermal regimes experienced by the larval and adult stages in each population.

Materials and methods

Study system, collection and rearing

Coenagrion scitulum (Rambur, 1842) is a Mediterranean damselfly preferring small ponds (Dijkstra, 2006). Up to the 1990s, the northern range limit was situated in northern France, after which a north-eastward range expansion has occurred (Wasscher & Goudsmits, 2010). In 2010, the northernmost limit of the expanding range margin was situated in the southern parts of the Netherlands, and the north-eastern limit in Western Germany (Wasscher & Goudsmits, 2010). We studied two core populations in France (C1: Rosnay and C2: Merlimont) within the historical distribution of the species and two edge populations (E1: Zülpich, Germany, and E2: Cadzand, the Netherlands) situated at the expansion front and founded < 4 years ago when sampled. All populations were located > 150 km from each other. Core population C2 and edge population E1 were chosen because they are situated at approximately the same latitude (Appendix S1, Table S1), which reduced the influence of latitudinal effects when comparing core and edge populations.

Paired females (C1: 10, C2: 20, E1: 11, E2: 10) were caught during June–July 2010 and were placed in oviposition chambers where they were allowed to oviposit on wet filter paper. Despite the high success rate of this method to collect egg clutches of closely related species, the number of females laying eggs was rather low (number of females laying eggs: C1: 2, C2: 11, E1: 5, E2: 2). Eggs were transported to the laboratory in Belgium, where throughout the study, eggs and larvae were kept at 20 °C and a photoperiod of L:D 14:10 h. Larvae were pooled per population and afterwards placed individually in 100-mL plastic cups filled with aged tap water and randomly allocated to the food treatment. Larvae were fed twice a day, 6 days a week, in excess with brine shrimp nauplii.

Fifteen days after hatching, half of the larvae per population were randomly allocated to one of the two food treatments. At the high food treatment, larvae were fed 6 days a week in excess with brine shrimp nauplii; larvae at the low food treatment were fed in excess 3 days a week. In nature, both unconstrained food conditions allowing maximum feeding rates and limited food conditions reducing feeding rates and growth have been documented in damselflies (Corbet, 1999). Some studies even reported serious food limitation in the field with most larvae having empty guts reflecting starvation conditions (e.g. Baker, 1989; Anholt, 1990a). The imposed feeding regimes span large part of this natural variation in food availability. To meet the higher food demands, the daily food ration was doubled in both food treatments when the larvae entered the final instar. The initial numbers of larvae reared per population status (core vs. edge) at each food treatment varied between 299 and 338 (total of 1280 larvae), and the numbers of larvae that survived until emergence per population status at each food treatment varied between 114 and 231 (total of 707 larvae). We studied three different trait sets on the reared animals: larval life history (larval development time, larval growth rate), adult immune function and flight-related traits (relative fat content, relative flight muscle mass and body size). Sample sizes of each measured variable per population at each food treatment are given in Table S1b (Appendix S2).

Life history traits

The total larval development time was quantified as the sum of the larval time needed to reach the final instar (period 1) and the time spend in the final instar (period 2). Larval growth rate was calculated using the dry mass of the adults as ln(dry mass)/total larval development time (Johansson et al., 2001).

Adult immune function

We immune-challenged a subset of the adults after hardening of the exoskeleton for 12 h by inserting a nylon filament and subsequently measured its degree of encapsulation (for details, see Appendix S3). This is a standard technique for measuring responses to an immune challenge in insects (e.g. Rantala et al., 2000). Only adults that survived the immune challenge were used to assess immune function. The immune challenge treatment followed with slight modifications the methodology of Rantala et al. (2000).

We additionally quantified the activity of phenoloxidase (PO). PO is a key component of insect immune function (Gonzalez-Santoyo & Cordoba-Aguilar, 2012). We measured the activity of PO using the protocol of Stoks et al. (2006) (for details, see Appendix S3). Enzyme activity was measured spectrophotometrically as the slope of the reaction that converts l-DOPA to dopachrome. We adjusted PO activity with the protein concentration.

Flight-related traits

For each adult that emerged successfully (i.e. with fully expanded wings), we quantified four flight-related traits based on the protocols described in Swillen et al. (2009): adult body dry mass, flight muscle mass, fat content and exoskeleton mass (as a measure of size) (for details, see Appendix S3).

Statistical analyses

Data were analysed using mixed models (sas v. 9.3) with the replicated populations nested in the population status as random effects. We used between- and within-subject degrees of freedom (Proc mixed, ddfm = bw) to determine the denominator degrees of freedom (Schluchter & Elashoff, 1990). Due to an error when separating the larvae within each population, we could only trace back the female identity of the offspring of a subset of females per population (number of remaining females per population when larvae with unknown female identity were removed: C1:2, C2: 3, E1: 3, E2: 1). All other offspring could only be unambiguously attributed to their population of origin. For this subset of larvae with known family identity, analyses with family nested in population as random factor could not detect any significant effects on the response variables.

We analysed the effects of population status, food treatment and sex on larval development time with a repeated-measures anova with the durations of period 1 and period 2 as repeats. The effects of population status, food treatment and sex on larval growth rate, encapsulation response and exoskeleton mass were analysed using an(c)ovas. Dry body mass and fat mass were added in the analysis of the encapsulation response to correct for the potential condition dependence of the immune function (Yang et al., 2007). For the response variables that could be affected by the insert treatment, dry body mass, flight muscle mass, fat content and PO activity, we added the insert treatment to the an(c)ovas as a covariate. Exoskeleton mass was added as covariate when analysing flight muscle mass and fat content, to correct for adult size differences. Dry body mass, fat mass and protein concentration were added as covariates when analysing PO activity. Means of all measured traits for the two core and the two edge populations as a function of sex and food level are given in Table S1 (Appendix S2).

To specifically correct for the effects of latitude-specific thermal regimes, we performed supplementary analyses for traits that were found to differ between core and edge populations (for details, see Appendix S1).

Results

Life history traits

Larval survival until emergence was higher at high food (F1,1275 = 58.78, P < 0.0001, Appendix S4, Fig. S1b) and did not differ between core and edge populations (population status: F1,2 = 2.82, P = 0.24; Population status × Food treatment: F1,1274 = 0.27, P = 0.60).

Differences in larval development between core and edge populations were dependent on the combination of period and food treatment (Population status × Food treatment × Period, Table 1). Separate analyses per period showed a faster development in edge individuals in period 1 across food levels (Table 1, Fig. 1a); in period 2, edge individuals developed significantly faster but only at the high food level (Population status × Food treatment, Table 1, Fig. 1b; population status at low food: F1,268 = 0.32, P = 0.574, at high food: F1,387 = 6.65, P = 0.010). Larvae developed faster at high food (Table 1, Fig. 1a–b), especially in period 2 (Food treatment × Period, Table 1). Males developed faster than females (sex, Table 1).

Table 1. Repeated-measures anova and separate anovas per period testing for the effects of population status (edge vs. core), food treatment (low vs. high) and sex on larval development time in Coenagrion scitulum. The successive durations of periods 1 and 2 were considered as repeats.
SourceBoth periodsPeriod 1Period 2
d.f. F P d.f. F P d.f. F P
Population status1,6775.590.0181,6805.710.0171,6560.510.477
Food treatment1,677174.34< 0.0011,68098.74< 0.0011,656191.57< 0.001
Sex1,6777.210.0071,6803.370.0671,6568.410.004
Period1,6557034.16< 0.001      
Population status × Food treatment1,6770.050.8221,6781.480.2241,6567.500.006
Population status × Sex1,6770.350.5561,6770.260.6111,6540.070.790
Food treatment × Sex1,6772.460.1171,6793.390.0661,6550.110.739
Population status × Food treatment × Sex1,6760.300.5861,6760.980.3221,6531.370.242
Population status × Period1,65544.45< 0.001      
Food treatment × Period1,65520.64< 0.001      
Sex × Period1,6551.080.299      
Population status × Food treatment × Period1,6554.360.037      
Population status × Sex × Period1,6540.230.633      
Food treatment × Sex × Period1,6553.230.073      
Population status × Food treatment × Sex × Period1,6531.490.223      
Figure 1.

Mean (+ SE) larval development time during period 1 (egg hatching until final instar) (a) and period 2 (time spent in final instar) (b), and larval growth rate (c) in core and edge populations of Coenagrion scitulum at low and high food levels. Left plots represent females, and right plots represent males. Shown are least-square means.

Larval growth rate was higher at high food (Table 2, Fig. 1c). The effect of population status on larval growth rate depended upon food level (Population status × Food treatment, Table 2). Separate analyses per food level showed edge individuals to have a higher growth at high food (F1,250 = 5.48, P = 0.020), but not at low food (F1,120 = 0.13, P = 0.722). Males and females did not differ in larval growth rate (sex, Table 2).

Table 2. anovas testing for the effects of population status (edge vs. core), food treatment (low vs. high) and sex on larval growth rate, adult encapsulation response and adult exoskeleton mass in Coenagrion scitulum.
SourceGrowth rateEncapsulation responseExoskeleton mass
d.f. F P d.f. F P d.f. F P
Population status1,3701.780.1831,808.130.0061,4810.010.904
Food treatment1,370138.89< 0.0011,780.380.5411,481127.49< 0.001
Sex1,3690.350.5531,791.270.2641,4801.630.203
Population status × Food treatment1,3706.210.0131,730.110.7441,4814.040.045
Population status × Sex1,3670.030.8691,761.100.2981,4780.410.521
Food treatment × Sex1,3680.060.8141,773.860.0531,4790.510.477
Population status × Food treatment × Sex1,3661.090.2971,720.010.9391,4770.390.530
Body mass   1,804.890.030   
Fat content   1,740.940.335   

Immune function

The adult encapsulation response did not depend upon the larval food treatment (Table 2, Fig. 2b). Edge adults had a higher encapsulation response than core adults (population status, Table 2, Fig. 2b). The encapsulation response did not differ between females and males (sex, Table 2).

Figure 2.

Mean (+ SE) adult exoskeleton mass (a), adult encapsulation response (b) and adult PO activity (c) in core and edge populations of Coenagrion scitulum at low and high food levels. Left plots represent females, and right plots represent males. Shown are least-square means. Encapsulation response is corrected for adult body mass and fat content; PO activity is corrected for adult body mass, fat content and protein concentration of the haemolymph.

Adult PO activity did not depend on the larval food treatment (Table 3, Fig. 2c). Differences in PO activities between core and edge populations were opposite in females and males (Population status × Sex, Table 3). Separate analyses for females and males showed a higher PO activity in edge females compared with core females (population status: F1,164 = 5.28, P = 0.023), whereas edge males had a lower PO activity compared with core males (population status: F1,168 = 4.43, P = 0.037). Overall, PO activity was lower in adults that were immune-challenged with an insert (Table 3, Fig. 2c).

Table 3. anovas testing for the effects of population status (edge vs. core), food treatment (low vs. high), sex and the insert treatment on PO activity, body dry mass, relative fat content and relative flight muscle mass in adult Coenagrion scitulum.
SourcePOAdult body massFat contentFlight muscle mass
d.f. F P d.f. F P d.f. F P d.f. F P
Population status1,3390.010.9121,3750.200.6571,3650.840.3601,3653.500.062
Food treatment1,3370.800.3701,379118.78< 0.0011,3620.450.5021,36628.19< 0.001
Sex1,3392.200.1391,3781.250.2641,3640.710.4011,3620.060.807
Insert1,33910.090.0021,3770.050.8291,3610.270.6051,3641.270.260
Population status ×  Food treatment1,3340.060.8051,3740.510.4751,3560.000.9791,3631.050.305
Population status ×  Sex1,3398.780.0031,3720.090.7591,3630.570.4501,3570.220.640
Population status ×  Insert1,3360.710.3991,3730.330.5651,3600.120.7291,3550.000.973
Food treatment ×  Sex1,3270.000.9591,3710.030.8691,3550.000.9861,3590.370.544
Food treatment ×  Insert1,3350.230.6341,3760.680.4111,3570.620.4331,3600.840.361
Sex × Insert1,3382.900.0901,3700.000.9691,3590.030.8551,3611.010.315
Population status ×  Food treatment ×  Sex1,3240.040.8401,3690.480.4891,3520.030.8661,3560.020.888
Population status ×  Food treatment ×  Insert1,3331.700.1931,3670.030.8691,3530.150.7031,3530.000.980
Population status ×  Sex × Insert1,3250.050.8221,3680.420.5161,3583.270.0711,3540.020.887
Food treatment ×  Sex × Insert1,3260.320.5721,3660.020.8771,3541.240.2651,3580.150.698
Population status ×  Food treatment ×  Sex × Insert1,3230.240.6251,3650.000.9761,3510.160.6861,3520.090.764
Body mass1,339182.53< 0.001         
Fat content1,3280.140.710         
Adult exoskeleton mass      1,366238.78< 0.0011,366212.28< 0.001
Protein concentration1,33925.29< 0.001         

Flight-related traits

Adult exoskeleton mass was higher at the high food treatment (Table 2, Fig. 2a). Exoskeleton mass did not differ between edge and core populations (population status, Table 2), but edge individuals showed a stronger reduction in exoskeleton mass when reared at low food (Population status × Food treatment, Table 2). Males and females did not differ in exoskeleton mass (sex, Table 2). Adult body mass was higher at high food (food treatment, Table 3, Fig. 3a), and females were heavier than males (sex, Table 3). Adult body mass did not differ between core and edge animals and was not affected by the insert treatment or their interactions (all P > 0.40, Table 3).

Figure 3.

Mean (+ SE) adult dry mass (a), relative fat content (b) and relative flight muscle mass (c) in core and edge populations of Coenagrion scitulum at low and high food levels and in the absence or presence of an insert. Left plots represent females, and right plots represent males. Shown are least-square means. Fat content and flight muscle mass are corrected for exoskeleton mass.

Size-corrected fat content was not dependent upon any of the treatments nor upon their interactions (Table 3, Fig. 3b). Size-corrected flight muscle mass was higher at high food (Table 3; Fig. 3c) and was nearly significantly higher in edge adults than in core adults (population status, Table 3, P = 0.062). Sex nor the insert treatment or any interactions had an effect on flight muscle mass (all P > 0.20, Table 3).

Contrasting the effects of the latitude-specific thermal regimes and range expansion

Individuals originating from ponds with a lower number of larval degree-days had a lower relative flight muscle mass, a lower PO activity in the subset of females, a nearly significant lower PO activity in the subset of males (P = 0.073) and a nearly significant faster development in period 2 at high food (P = 0.071) (Table 4). The number of larval degree-days did not affect larval development time in period 1, growth rate and encapsulation response (Table 4).

Table 4. Summary of anovas testing for the effects of population status (edge vs. core) and latitude-specific thermal regimes (larval degree-days and mean air temperature during the flight period) on traits that showed (nearly) significant differences between core and edge populations based on previous analyses (Tables 1-3). Effects on development time in period 1, encapsulation response and flight muscle mass were investigated with the full data set; effects on development time in period 2 and larval growth rate were investigated at high food; and effects on PO activity were investigated in females and males separately. The full results are given in Appendix S5 (Table S1).
TraitData setLarval degree-daysAir temperaturePopulation statusLeast-squares mean
P Slope ± SE P Slope ± SE P CoreEdge
Development time 1Full0.623   0.017209.9 ± 7.7183.9 ± 7.7
Development time 2High food0.071(84 ± 46) × 10−4  0.0234.75 ± 0.9431.47 ± 0.98
Growth rateHigh food0.617   0.02(470 ± 32) × 10−5(573 ± 31) × 10−5
EncapsulationFull0.223 0.209 0.00639.1 ± 2.649.5 ± 2.5
PO activityFemales0.038(131 ± 63) × 10−40.097−2.07 ± 1.240.0117.7 ± 1.113.3 ± 1.2
PO activityMales0.073(65 ± 36) × 10−40.008−1.38 ± 0.510.2189.3 ± 1.012.0 ± 1.3
Flight muscle massFull0.016(35 ± 14) × 10−50.003−0.83 ± 0.03< 0.0010.931 ± 0.0241.114 ± 0.028

Adults from populations with lower mean air temperatures had a higher relative flight muscle mass and a higher PO activity (only a trend in females, P = 0.097), whereas mean air temperatures did not affect encapsulation response (Table 4).

The latitude-specific temperature regimes could only fully explain the higher PO activity in core males compared with edge males. In contrast, the differentiation between core and edge populations remained for all other traits when the analyses, investigating trait differentiation between core and edge populations, were corrected for the effect of latitude-specific temperature regimes. Furthermore, edge populations had higher relative flight muscle masses when the analysis was corrected for temperature regimes (Table 4).

Discussion

We found strong evidence for phenotypic differentiation between core and edge populations in both the larval and adult stage. Given that animals were reared from the egg stage in a common garden experiment, these phenotypic differences likely reflect genetic differences (Hill et al., 2011) and hence rapid evolution at the expanding range front. Although we cannot fully exclude the possibility of a contribution of maternal effects, these are unlikely to fully explain the here observed multivariate phenotypic differentiation. Two other studies on coenagrionid damselflies have shown that maternal effects only played a minor, nonsignificant role in shaping life history traits (Strobbe & Stoks, 2004; Shama et al., 2011). In general, maternal effects decay through ontogeny (Lindholm et al., 2006), and this may especially be expected in animals with a complex life cycle.

Besides the effects of the range expansion process per se and temperature regimes, traits in each population may be shaped by local conditions such as predator presence, parasite pressure and food abundance (Stoks & Cordoba-Aguilar, 2012). Because we studied only two edge and two core populations with a limited number of families per population, we cannot fully exclude the possibility that some of the documented differentiation between core and edge populations is influenced by random sampling effects related to these local conditions. Yet, we consider it unlikely this would fully explain the here observed directional differences between edge and core population as these are in accordance with a priori predictions. Moreover, in another study focusing on larval behaviour using two core and four edge populations and a much larger number of families, we found consistently higher larval activity levels in edge populations (L. Therry, E. Lefevre, D. Bonte & R. Stoks, unpublished data), which is in line with the here documented faster life history in edge populations. This suggests that our results truly reflect differentiation between core and edge populations rather than being the result of random sampling effects due to a low number of populations or families.

Life history traits

Edge larvae showed a faster life history: shorter development times and higher growth rates. Theory predicts selection for a faster life history at the expansion front to achieve a greater number of generations per year, which is favoured at the low population densities in newly colonized habitats (Burton et al., 2010). Higher larval growth rates were documented in populations at the expansion front of the cane toad Bufo marinus (Phillips, 2009). We hypothesize that a similar process acts at the expanding range edge of C. scitulum: low population densities at the range margin enhance the selection for genotypes that can maintain a univoltine life cycle also at higher latitudes. One may argue that the faster life history in edge populations can be shaped by countergradient variation to shorter growth seasons at higher latitudes (Conover & Schultz, 1995). However, after correcting for the influence of larval degree-days, larvae from edge populations still displayed a faster development and higher growth rates than larvae from core populations. This suggests that edge populations experience strong selection for short life cycles driven not only by the thermal regimes but also by the expansion process per se.

Immune function

Edge adults had a stronger immune response as measured by a higher encapsulation response than core adults. This was only partially reflected in the patterns of PO activity as edge females had higher PO activities than core females, whereas the opposite pattern was observed in males. Different patterns in immune function between the sexes likely reflect differences in life history between females and males (Rolff, 2002). The discrepancy between the two measurements of immune function suggests that besides PO, other components involved in the encapsulation response, such as melanin precursors (Slominski et al., 1988) or dopa decarboxylase (Kim et al., 2000), were increased in edge individuals. While it may seem counterintuitive that PO activity decreased in response to the immune challenge, a similar decrease in PO activity was documented in immune-challenged honey bees where it was explained by a failure to replenish PO after its utilization during the immune response (Laughton et al., 2011).

The higher encapsulation response in the edge populations could not be explained by the investigated latitude-specific thermal regimes and likely is associated with the range expansion process per se. A higher immune function could evolve during range expansion through direct selection imposed by higher encounter rates with parasites and pathogens in the newly founded edge populations (Møller & Erritzøe, 2001; Barber & Dingemanse, 2010). The here studied encapsulation response towards the inserted nylon filament specifically mimics an immune challenge caused by water mites whereby melanization of the stylome can result in the death of the water mite (Yourth et al., 2001). This may translate in interpopulation differentiation where populations with higher mite loads have a higher encapsulation response as demonstrated in another Coenagrion species (Kaunisto & Suhonen, 2013). However, in our study, water mite densities were lower in edge compared to core populations (L. Therry, unpublished data), and hence, higher encapsulation responses in edge populations cannot be explained by higher water mite pressures in the edge populations. However, the encapsulation response is a general defence mechanism against a range of other pathogens and parasites such as viruses, bacteria, fungi, nematodes and parasitoids (Honkavaara et al., 2009). While we cannot exclude that the densities of these general pathogens and parasites systematically differ between core and edge populations, we hypothesize that the higher encapsulation response in the edge populations may result from the selection for increased dispersal ability at a moving range front (Burton et al., 2010). In support of this idea, it has been shown that ectoparasitism by water mites reduces the flight ability of damselflies (Reinhardt, 1996; Nagel et al., 2010). Thus, selection for a higher dispersal ability may indirectly select for a higher encapsulation response as individuals with a higher encapsulation response likely carry less parasites and hence have a higher dispersal potential. This may result in a higher encapsulation at the range front because of a genetic correlation with dispersal ability but may also occur under spatial sorting in the presence of a pure (i.e. nongenetic) phenotypic positive correlation between flight ability and encapsulation. Under this scenario, animals with a higher flight ability (exactly because of their higher flight ability) and with a higher encapsulation (because of the positive phenotypic correlation) end up at the range front; if the encapsulation response is heritable, this will lead to a higher encapsulation response also in the following generations at the range front.

Flight-related traits

The relative investment in flight muscles was higher in edge populations; however, this was only significant when the analysis was corrected for the latitude-specific thermal regimes. The thoracic muscle mass is a strong predictor of flight muscle power output in Odonata (Schilder & Marden, 2004). Increased investment in flight-related traits has been observed in edge populations of several poleward-moving insect species (reviewed in Hill et al., 2011). Such pattern is predicted by models as a result of natural selection whereby individuals that disperse most rapidly at the range front benefit because of reduced competition due to lower densities of conspecifics at the edge (Travis & Dytham, 2002). Additionally, only the best dispersers, and hence animals with the highest relative flight muscle mass, may end up in the edge populations (so-called spatial sorting, Shine et al., 2011). Note that spatial sorting shapes phenotypes at the expansion front irrespective of how the underlying genes affect organism's survival or reproductive process, making this process fundamentally different from natural selection (Shine et al., 2011).

The lower relative flight muscle mass in populations with shorter larval growth seasons likely reflects a decreased investment in flight muscles under stressful conditions in the larval stage (Plaistow & Siva-Jothy, 1999). In contrast, relative flight muscle masses were higher in populations with lower temperatures during the adult flight period, which is congruent with a scenario of countergradient variation where individuals compensate the decreased muscle efficiency at lower temperatures by a higher investment in flight-related traits (Hassall et al., 2008). Note that we could exclude that the higher investment in flight muscles in edge populations was a response to latitude-specific temperature regimes and the correction for the latitude-specific temperature regimes even enhanced the pattern of differentiation in relative flight muscle mass between core and edge populations.

Plasticity under food shortage

The imposed low food treatment was stressful as, in line with many other empirical studies (reviewed in Nylin & Gotthard, 1998) including those on damselflies (Stoks & Cordoba-Aguilar, 2012), it lowered larval survival, growth and development rates and resulted in lighter adults. Noteworthily, larval food stress resulted in adults with a smaller relative flight muscle mass. This may suggest that food stress in the larval aquatic stage may negatively affect dispersal in the adult stage (Benard & McCauley, 2008). This is congruent with the study by Anholt (1990b) where larvae of the damselfly Enallagma boreale reared under low food and high crowding conditions showed lower dispersal rates.

We observed no trade-offs at the population level between costly traits such as flight muscle mass and the encapsulation response as core populations consistently showed the highest values of both traits (Reznick et al., 2000). Similarly, at the individual level, no trade-offs between costly traits were observed (Appendix S6). The absence of clear bivariate trade-offs may reflect that the trade-offs are multivariate and that other unmeasured traits were negatively affected (Edwards & Stachowicz, 2010) or that trade-offs may be expressed later in life (Dmitriew & Rowe, 2007). For example, a trade-off between flight ability and reproduction is documented in many species (reviewed by Zera & Denno, 1997), and a higher investment in flight morphology was associated with a lower fecundity in a northward-expanding butterfly (Hughes et al., 2003). Furthermore, the faster life history and higher investment in costly traits is likely obtained by a higher food acquisition, and this is in line with the results from another study whereby larval activity in edge populations was consistently higher compared with core populations (L. Therry, E. Lefevre, D. Bonte & R. Stoks, unpublished data). Possibly, the faster life history of edge individuals is disadvantageous in particular conditions, such as at low food conditions and at high predation pressures. The higher growth rate of edge populations disappeared at low food conditions and edge populations showed a higher reduction in size when larvae were reared at low food conditions, providing some evidence for Population type × Environment interactions whereby the fitness of edge individuals may be expected to decrease faster than that of core populations at further decreasing food levels.

Conclusions & Perspectives

The here documented multivariate evolution in ecologically relevant larval (life history) and adult (flight muscle mass and encapsulation response) traits between edge and core populations of C. scitulum could largely be attributed to processes directly linked to the recent range expansion. Natural selection for a higher flight performance at the range front, potentially in association with spatial sorting, likely shaped the higher investment in relative flight muscle mass in edge populations. Moreover, we hypothesize that this process also underlay the here for the first time documented higher immune function in animals at the expansion front. Furthermore, several aspects of the faster larval life history in edge populations could be attributed to the expansion process per se, and probably evolved in response to selection for a higher population growth rate (here selection for short generation times) to exploit the initial lower densities in the edge populations.

Our results highlight that, irrespective of the underlying process, evolution of traits in poleward-moving populations under global warming is not limited to locomotory traits in adults, but also includes immunological traits and traits in the larval nondispersive stage. Given that the studied traits are known to modify interactions with competitors, predators and parasites in damselflies (Stoks & Cordoba-Aguilar, 2012), the here observed trait differentiation between edge and core populations is expected to translate into changed ecological interactions and therefore has the potential to generate novel eco-evolutionary dynamics (Pelletier et al., 2009; Hanski, 2012) at the expanding range edge. A future challenge will therefore be to study whether (see Palkovacs et al., 2012) and how ecological interactions changed in the edge populations and whether this affects further range expansion and the local communities (Bolnick et al., 2011) in the edge habitats.

Acknowledgments

We thank Geert De Knijf, Cédric Vanappelghem, Klaus-Jürgen Conze, Dietmar Glitz, Frantz Veillé and Jochen Rodenkirchen for their help in guiding us to populations, Sharon Schillewaert and Silke Sterck for assistance in the field and the ‘Office National des Forêts’ for access at the ‘RBD de la côte d'Opale’. Two anonymous reviewers provided useful comments. All experiments conformed with the permit by the Flemish Agency for Nature and Forest (ANB-Flanders). This work was supported by research grants from the Fund for Scientific Research (FWO) Flanders (G.0610.11) to DB and RS and the KU Leuven Research Fund (GOA/2008/06 and Excellence Center Financing PF/2010/07) to RS.

Ancillary