Phosphoglucose isomerase genotype affects life-history traits and cold stress resistance in a Copper butterfly


*Correspondence author. E-mail:


  • 1Accumulating evidence suggests that the phosphoglucose isomerase (PGI) locus is under thermal selection. In the Copper butterfly Lycaena tityrus PGI allele frequencies show altitudinal variation, with a single genotype occurring in c. 90% of high-altitude animals. In low-altitude populations variation at this locus is much higher.
  • 2Here, we investigate variation in life-history traits and temperature stress resistance across PGI genotypes in L. tityrus from different lowland populations reared at two temperatures (19 and 24 °C).
  • 3PGI genotype significantly affected larval and pupal development time, growth rate, pupal mass and chill-coma recovery time, but had no effect on heat knock-down resistance. The latter suggests that heat and cold stress resistance are based on differential mechanisms.
  • 4As expected temperature also influenced all traits under investigation, its effect being more pronounced compared to that of PGI genotype (except for pupal mass).
  • 5Patterns found for the PGI genotype dominating in high-altitude populations were consistent with those found for high-altitude animals. Therefore, and because of the direct link between PGI genotype and cold stress resistance, we conclude that PGI is likely to contribute to thermal adaptation in L. tityrus.
  • 6Genotypes promoting rapid development and largest body size were rather rare, suggesting weak selection on both traits and/or rather high associated costs.


For several ectotherms it has been shown that allozyme variation correlates with variation in an array of fitness-related traits including morphological and physiological ones (e.g. Watt 1992; Neargarder, Dahlhoff & Rank 2003; McMillan et al. 2005; Dahlhoff & Rank 2007; Saastamoinen 2007). Usually, those studies supposed the involvement of environmental variables as selective agents causing a given allozyme variation. Where alternative forms of enzymes exist (allelic products of a single locus), selection through temperature is particularly likely because of its effects on enzyme function and metabolic rates, in turn affecting fitness (Borrell et al. 2004; Ward, Jann & Blanckenhorn 2004). In recent years, the importance of environmental effects on the phenotypic expression of individual genes became increasingly clear (DeWitt & Scheiner 2004), and it is furthermore now generally accepted that genetic variation at single loci may affect multiple phenotypic characters (Pigliucci 2004; Pigliucci & Preston 2004).

One possibility to detect interactions between the forces shaping phenotypic expression and those maintaining its underlying genetic variation is to look for variation at a specific gene locus and to identify its effects on the phenotype (Krause & Bricelj 1995). In this context, enzymes that are involved in important metabolic pathways, for example, glycolytic enzymes, are of special interest, as glycolysis is the centre point of all ATP-based energy supplies (Watt 1985). One of these enzymes, phosphoglucose isomerase (PGI), has received much attention since the 1980s, and evidence for a covariance between environmental variables (especially temperature) and PGI is accumulating (Watt 1983, 1992; Rank & Dahlhoff 2002; Neargarder et al. 2003; McMillan et al. 2005). In Colias butterflies, for example, PGI genotypes differ dramatically in their enzyme–kinetic properties, thus affecting glycolytic fluxes and thereby flight performance (Watt 1983, 1992). Likewise, in the Glanville fritillary butterfly, PGI genotype has a significant effect on flight metabolic rate and concomitantly on dispersal rate (Haag et al. 2005). As another example, directional changes in PGI allele frequencies, coinciding with variation in HSP70 expression, temperature stress resistance, running speed, survival and fecundity were found in the leaf beetle Chrysomelia aeneicollis (Dahlhoff & Rank 2000, 2007; Rank & Dahlhoff 2002; Neargarder et al. 2003; McMillan et al. 2005; Rank et al. 2007). However, apart from these studies, our knowledge of the consequences of variation at the PGI locus on life-history and stress resistance traits is still very limited.

In the present study, we quantified variation in life-history and temperature stress resistance traits across different PGI genotypes in the butterfly Lycaena tityrus. For this species, we have found altitudinal variation in PGI allele frequencies, causing significant genetic differentiation between high- and low-altitude populations (Karl, Schmitt & Fischer unpublished data). In the high-altitude populations, a single PGI genotype (PGI 2–2) was found in about 90% of all animals, while low-altitude butterflies showed a much larger variation at the PGI locus (Karl et al. unpublished data). Because of the strong covariance between temperature and geographic clines, temperature is generally believed to be one of the most important selective agents causing such clinal variation (Loeschke, Bundgaard & Barker 2000), which is supposedly also the case in L. tityrus. Assuming that the dominance of the PGI 2–2 genotype in L. tityrus high-altitude populations is causally related to thermal selection in these low-temperature environments, we test here whether PGI 2–2 butterflies (from lowland populations) show features typical of high-altitude butterflies, namely whether they show shorter development times, increased growth rates (but see Karl, Janowitz & Fischer 2008), larger body size, increased cold stress resistance, but decreased heat stress resistance as compared to other genotypes.

Material and methods

study organism

Lycaena tityrus (Poda, 1761) is a widespread temperate zone butterfly, ranging from Western Europe to central Asia (Tolman & Lewington 1998). The species is bivoltine with two discrete generations per year in most parts of its range, although populations with one or three generations per year occur (Ebert & Rennwald 1991; Tolman & Lewington 1998). Lycaena tityrus hibernates as L3-larvae. The principal larval host-plant is Rumex acetosa L., but some congeneric plant species such as R. acetosella L. and R. scutatus L. are utilized as well, and may in some regions represent the main hosts (SBN 1987; Ebert & Rennwald 1991; Tolman & Lewington 1998). For this experiment, 18 freshly eclosed, mated females were caught in June 2007 in different bivoltine lowland populations in Germany (western Germany –‘Westerwald’, north-eastern Germany – near the city of Greifswald, southern Germany – near Benediktbeuern) and transferred to Bayreuth University. Females from different populations were included to promote a high genetic diversity at the PGI locus in the offspring generation.

experimental design

Oviposition and butterfly rearing

Field-caught females were kept in a climate chamber at 24 °C and L18 : D6 (24 h light cycle) throughout. For oviposition they were placed individually in translucent plastic pots (1 L) covered with gauze, and were provided with R. acetosa (oviposition substrate), fresh flowers (Crepis sp., Achillea millefolium, Bistorta officinalis, Leucanthemum vulgare) and a highly concentrated sucrose solution (for adult feeding). Eggs were collected daily, pooled across females and transferred to small glass vials. After hatching, larvae were randomly divided among two climate chambers differing in rearing temperature (19 and 24 °C, respectively). Both climate chambers used are located within the same building next to each other. They are identical in terms of construction, lightning and air conditioning. Throughout, photoperiod was set at L18 : D6, and relative humidity at 70%. Larvae were reared in groups (10 individuals each) in translucent plastic boxes (500 mL), containing moistened filter paper and fresh cuttings of R. acetosa in ample supply. Boxes were checked daily and supplied with new food when necessary. Resulting pupae were weighed (to the nearest 0·01 mg; Sartorius microscale MC 210 P) and afterwards kept individually in numbered plastic pots (125 mL). Following adult eclosion, butterflies were kept individually in translucent plastic pots (250 mL) covered with gauze, and were provided with a highly concentrated sucrose solution for adult feeding until the start of stress tolerance assays (see below).

Data acquisition

Larval development time (from hatching to pupation), pupal mass, pupal development time and growth rate (calculated as quotient of pupal mass and larval developmental time) were recorded for all individuals. On day 2 after eclosion, butterflies were randomly assigned to either a cold or heat stress resistance assay. Cold stress resistance was determined as chill-coma recovery time (Hoffmann, Anderson & Hallas 2002; Ayrinhac et al. 2004; Castañeda, Lardies & Bozinovic 2005). Butterflies were placed individually in small translucent plastic cups (125 mL), which were arranged on a tray in a randomized block design. The tray was then exposed for 6 min to –20 °C. This period was selected as preliminary studies showed that longer cold exposure induced significant mortality, and a shorter one very quick recovery. After cold exposure the trays were transferred to an environmental cabinet with a constant temperature of 20 °C. Recovery time was defined as the time elapsed between taking the tray out of the freezer until a butterfly was able to stand on its legs. Only butterflies that had recovered within 1 h were included in further analyses, as mortality is very high for butterflies with longer recovery times. Following recovery, butterflies were frozen at –80 °C for later allozyme analyses.

Heat stress resistance was determined by using a knock-down assay (Sørensen et al. 2005). Butterflies were placed in small, sealed glass vials, which were submerged in a water bath kept at a constant temperature of 47 °C (again in a randomized block design). Butterflies were continuously monitored and heat knock-down time (defined as the time until a butterfly was no longer able to stand upright) was recorded for each individual. To reduce mortality, knocked-down butterflies were immediately taken out of the water bath and, after a short recovery time (to be sure that the butterfly was still alive), they were frozen at –80 °C for further analyses.


PGI genotype was analysed for 1381 individuals. Therefore, half the butterflies’ abdomen was homogenized in Pgm-buffer (Harris & Hopkinson 1987) by ultrasound and centrifuged at 8000 g for 5 min. Cellulose acetate plates were used for electrophoresis applying standard protocols (Hebert & Beaton 1993). Samples were run in TG-buffer (Tris–glycine pH 8·5; see Hebert & Beaton 1993) at 200 V. The alleles were labelled according to their relative mobility, starting with ‘1’ for the slowest.

statistical analyses

Due to the presence of rare genotypes and concomitantly low sample sizes in some groups, life-history and stress resistance data were only analysed for the four most common PGI genotypes (1–1; 2–2; 1–2; 2–3). Despite this restriction, sample sizes were highly unbalanced owing to the differential frequency of genotypes. Nevertheless, all respective individuals were included in the statistical analyses, as analyses using similar group sizes (by drawing random samples from the two most frequent genotypes) yielded qualitatively identical results. Data were analysed using analyses of (co-)variance (an(c)ovas) with PGI genotype, temperature and sex as fixed effects. Pupal mass was added as covariate when analysing stress resistance traits. Pair-wise comparisons were performed employing Tukey's HSD. All statistical tests were performed using jmp (4·0·0) or Statistica (6·1). Unless otherwise stated, least square means ± 1 SE are given throughout the text.


The butterflies analysed (N = 1381) represent seven different PGI genotypes. The two most common genotypes, PGI 1–2 (N = 606) and PGI 2–2 (N = 486), represent 79·1% of all individuals. They are followed by PGI 2–3 (N = 127), PGI 1–1 (N = 86), PGI 1–3 (N = 49), PGI 1–4 (N = 21) and PGI 3–3 (N = 6; see Appendix S1 in Supplementary Material for sample sizes and trait values across all genotypes and treatment groups). All results below are based on N = 1306 representing the four most common genotypes.

effects of pgi genotype on life-history traits

All four life-history traits investigated showed significant variation across genotypes, rearing temperatures, and sexes, the only exception being that there was no sex difference in pupal mass (Table 1). Larval development time varied significantly between genotypes (PGI 1–2: 22·7 ± 0·1 days = PGI 2–2: 22·5 ± 0·1 days > PGI 1–1: 21·9 ± 0·2 days = PGI 2–3: 21·7 ± 0·1 days), rearing temperatures (19 °C: 26·7 ± 0·1 days > 24 °C: 17·7 ± 0·1 days), and sexes (females: 23·3 ± 0·1 days > males: 21·1 ± 0·1 days; Fig. 1a). The significant temperature by sex interaction for larval time indicates that the sex difference was slightly less pronounced at the higher (males 2·0 days faster) than at the lower temperature (males 2·4 days faster).

Table 1. an(c)ovas for the effects of PGI genotype, temperature and sex on life-history and stress-resistance traits in Lycaena tityrus. Pupal mass (covariate) was added as appropriate. Significant P-values are given in bold
Larval time
 Genotype109·6317·3< 0·0001
 Temperature13355·116321·1< 0·0001
 Sex789·61373·7< 0·0001
 Genotype × Temperature3·330·50·6627
 Genotype × Sex5·530·90·4569
 Temperature × Sex8·213·90·0492
 Genotype ×  Temperature × Sex3·630·60·6358
Growth rate
 Genotype45·9332·7< 0·0001
 Temperature1076·012296·2< 0·0001
 Sex56·21119·9< 0·0001
 Genotype × Temperature7·034·90·0019
 Genotype × Sex0·630·40·7186
 Temperature × Sex7·3115·7< 0·0001
 Genotype ×  Temperature × Sex1·631·10·3444
Pupal time
 Temperature5458·517202·3< 0·0001
 Sex34·7145·8< 0·0001
 Genotype × Temperature3·931·70·1591
 Genotype × Sex0·730·30·8228
 Temperature × Sex3·214·20·0392
 Genotype ×  Temperature × Sex4·832·10·0988
Pupal mass
 Genotype5967·4316·4< 0·0001
 Genotype × Temperature254·030·70·5540
 Genotype × Sex123·730·30·7968
 Temperature × Sex23·210·20·6623
 Genotype ×  Temperature × Sex237·630·70·5819
Chill-coma recovery time
 Genotype × Temperature167288·330·70·5229
 Genotype × Sex41226·430·20·9068
 Temperature × Sex106805·711·40·2313
 Genotype ×  Temperature × Sex158236·830·70·5469
 Pupal mass9563·110·10·7201
Heat knock-down time
 Sex3848598·6183·0< 0·0001
 Genotype × Temperature78752·230·60·6373
 Genotype × Sex121983·130·90·4524
 Temperature × Sex337502·817·30·0071
 Genotype ×  Temperature × Sex99442·530·70·5431
 Pupal mass942221·3120·3< 0·0001
Figure 1.

Means (± 1 SE) for larval time (a), larval growth rate (b), pupal time (c) and pupal mass (d) for Lycaena tityrus males (black symbols) and females (white symbols) across four PGI genotypes and two rearing temperatures (19 and 24 °C). Note the partly different scales on the Y-axis.

Patterns in growth rates were largely opposite to those found in larval time (genotypes – PGI 2–3: 6·40 ± 0·06 mg day−1 > PGI 1–1: 6·09 ± 0·08 mg day−1 = PGI 2–2: 5·91 ± 0·03 mg day−1 > PGI 1–2: 5·76 ± 0·03 mg day−1; temperatures – 24 °C: 7·3 ± 0·03 mg day−1 > 19 °C: 4·8 ± 0·04 mg day−1, sexes – males: 6·3 ± 0·03 mg day−1 > females: 5·7 ± 0·03 mg day−1; Fig. 1b). The sex difference across temperatures was less pronounced at 19 °C (males by 0·37 mg day−1 faster) than at 24 °C (males by 0·80 mg day−1 faster). Further, a significant interaction between genotype and temperature indicates a more pronounced effect of temperature on growth rate in PGI 2–3 butterflies compared to other genotypes.

Regarding pupal time, PGI 2–3 butterflies (11·6 ± 0·07 days) showed a significantly shorter development time compared to PGI 1–1 (12·0 ± 0·08 days) and PGI 1–2 butterflies (11·9 ± 0·04 days), while PGI 2–2 butterflies (11·8 ± 0·04 days) did not differ from any other group. Further, development was significantly shorter at the higher (8·99 ± 0·04 days) than at the lower (14·68 ± 0·04 days) rearing temperature, and was shorter in males (11·63 ± 0·04 days) than in females (12·04 ± 0·04 days; Fig. 1c). Sex differences were slightly smaller at 19 °C (males by 0·3 days faster) than at 24 °C (males by 0·5 days faster; indicated by a significant temperature-by-sex interaction). Pupal mass finally was more than 5% higher in PGI 2–3 than in PGI 1–2 butterflies (PGI 2–3: 130·9 ± 0·9 mg > PGI 1–1: 128·2 ± 1·2 mg ≥ PGI 2–2: 126·2 ± 0·5 mg > PGI 1–2: 123·9 ± 0·4 mg), was higher at 24 °C than at 19 °C (24 °C: 128·4 ± 0·5 mg > 19 °C: 126·3 ± 0·5 mg), but did not differ between the sexes (Fig. 1d).

effects of pgi genotype on stress resistance

Chill-coma recovery time varied significantly between genotypes and temperatures, but was not affected by sex or the covariate pupal mass (Fig. 2a, Table 1). PGI 2–2 animals (406·8 ± 17·7 s) showed by more than 15% reduced recovery times compared to PGI 1–2 animals (484·6 ± 15·7 s), while PGI 1–1 (508·7 ± 39·4 s) and PGI 2–3 butterflies (489·7 ± 34·5 s) did not differ significantly from any other group despite of having the longest mean recovery times (but note the substantial within-group variation, obviously caused by relatively low sample size; see Supplementary Appendix S1). Further, the animals reared at the lower temperature (430·8 ± 17·6 s) showed a reduced recovery time (> 15%) compared to the ones reared at the higher temperature (514·0 ± 18·3 s).

Figure 2.

Means (± 1 SE) for chill-coma recovery time (a) and heat knock-down time (b) for Lycaena tityrus males (black symbols) and females (white symbols) across four PGI genotypes and two rearing temperatures (19 and 24 °C).

In contrast, heat knock-down time was not affected by PGI genotype, but was more than 10% shorter in animals reared at the lower (393·0 ± 14·5 s) vs. the higher temperature (444·0 ± 14·9 s; Fig. 2b), and nearly 40% shorter in males than in females (315·9 ± 14·9 s vs. 521·1 ± 14·5 s). The latter difference was much more pronounced at the higher compared to the lower temperature, as evidenced by a significant temperature by sex interaction (knock-down time was 260·0 s longer in females compared to males at the higher temperature, but only 151·3 s longer at the lower temperature). Further, there was a significant impact of the covariate pupal mass, with larger individuals being on average more resistant to heat stress than smaller ones (R = 0·186, P < 0·0001, N = 655).


A wealth of studies shows correlations between variation in life-history traits and individual fitness (e.g. Munch & Conover 2003; Fischer et al. 2004; Berger, Walters & Gotthard 2006), and at least some of such fitness-related traits vary with PGI genotype (e.g. Neargarder et al. 2003; McMillan et al. 2005; Dahlhoff & Rank 2007; Saastamoinen 2007). If a single locus shows patterns of variation being discordant with other loci, natural selection is likely acting on that locus (Slatkin 1987). Here, we demonstrate a direct link between variation in life-history and stress resistance traits and PGI genotypes in the Copper butterfly L. tityrus. Note in this context that hardly any data have been published to date on the potential association between developmental traits and PGI genotype.

effects of pgi genotype on life-history traits

PGI genotype considerably affected larval and pupal development time, growth rate, and pupal mass in L. tityrus. As PGI 2–2 is the dominant genotype found in high-altitude butterflies (Karl et al. unpublished data), a reduced development time with concomitantly higher growth rates was expected for this compared to other genotypes. Such results were for instance obtained for Drosophila buzzatii populations from the coolest highland localities (Bubliy & Loeschcke 2005). Our results on L. tityrus, however, show that the high-altitude genotype had intermediate to long development times. This can be explained by the fact that indeed high-altitude populations do exhibit longer development times in L. tityrus, presumably caused by a change in voltinism (low-altitude populations are bivoltine, while high-altitude ones are monovoltine; Karl et al. 2008). Of course, developmental traits may be affected by a large number of different genes (Chippindale, Ngo & Rose 2003), but the trend towards relatively long development times in addition to the pronounced overall variation across genotypes clearly suggests that PGI does affect developmental pathways (including growth rates, see below). Similarly, variation in development time among PGI genotypes was shown in the leaf beetle C. aeneicollis, though such differences were thought to be related to differences in the expression of heat shock proteins rather than PGI genotype per se (McMillan et al. 2005).

It is a common belief that ‘faster is better’ in ecology and evolutionary biology, and accumulating evidence suggests that growth rate in itself is a target of natural selection (Arendt 1997; Nylin & Gotthard 1998; Munch & Conover 2003). Although it is intuitively appealing that consequently growth rates should be maximized, recent studies convincingly implicate that growth rates are optimized rather than maximized (Arendt 1997; Nylin & Gotthard 1998). This in turn implies that fast growth carries costs such as a lower viability (Chippindale et al. 1997), a higher weight loss during metamorphosis (Fischer et al. 2004), or a higher predation risk (Gotthard 2000; Munch & Conover 2003). Against this background it is interestingly to note that in L. tityrus growth rates in PGI 1–1 and PGI 2–3 butterflies, having low frequencies of 6% and 9%, were clearly higher than in PGI 1–2 (43%) and PGI 2–2 butterflies (35%). This may suggest, in line with the above considerations, that selection is not favouring genotypes promoting fast development.

A large number of studies, many of which were concerned with clinal variation, suggest that ectotherms tend to be larger in colder environments (James & Partridge 1995; James, Azevedo & Partridge 1995). However, recent work indicates that this may not necessarily be the case, and that many different outcomes are possible (Blanckenhorn 1997; Blanckenhorn & Demont 2004). In Copper butterflies, different studies could not find an association between the temperature conditions at the place of population origin and body size (Karl et al. 2008). Accordingly, high-altitude genotypes (PGI 2–2) were not of particularly large size in this study.

Another interesting pattern is that there was no trade-off between fast growth and body size, that is, slow-growing individuals did not become large, and fast-growing ones not small (cf. Blanckenhorn 1999; Davidowitz, D’Amico & Nijhout 2004). In contrast, PGI 2–3 individuals with the highest growth rates also showed the highest pupal masses, while PGI 1–2 butterflies with the lowest growth rates were smallest. It remains unclear though why the (under laboratory conditions) obviously highly efficient genotype PGI 2–3 is relatively rare (9%) in nature. Furthermore, the two most common genotypes were smallest in size, suggesting that the costs associated with achieving and/or maintaining large body size largely outweigh any potential benefit such as increased fecundity or mating success (Roff 1992; Blanckenhorn 2000).

effects of pgi genotype on temperature stress resistance

In addition to life-history traits, PGI genotype also affected cold but not heat stress resistance. The genotype dominating in high-altitude populations (PGI 2–2) exhibited the shortest chill-coma recovery times. This is consistent with an increased cold stress resistance in high-altitude L. tityrus populations (Karl et al. 2008), and suggests that the PGI locus is under thermal selection (see also Watt 1994; Dahlhoff & Rank 2000; McMillan et al. 2005). The mechanisms underlying the association between PGI variation and cold resistance are not yet known for L. tityrus. In the willow leaf beetle, however, allelic variation at the PGI locus across a latitudinal gradient is linked to variation in the expression of HSP70 (Dahlhoff & Rank 2000). This may also apply to L. tityrus, as significant differences in HSP70 expression between high- and low-altitude individuals were recently found (Karl et al. unpublished data).

In contrast, despite variation in heat stress resistance across high- and low-altitude L. tityrus populations (Karl et al. 2008), no significant variation across PGI genotypes in heat resistance were detected in this study. Thus far, however, there is no evidence for a causal link between knockdown resistance and chill-coma recovery, and further our findings corroborate the notion that the mechanisms underlying increased cold tolerance are at least partly uncoupled from the mechanisms increasing heat tolerance (see also Chown 2001; Klok & Chown 2003; Sørensen et al. 2005). Certainly, PGI is not the only locus under thermal selection, but other yet unknown loci may also contribute to thermal adaptation, some of which may cause the reduced heat resistance in high-altitude populations. Finally, an association between PGI and heat resistance cannot be excluded, as different methods to assess thermo-tolerance may involve different genetic pathways (Folk et al. 2006).

effects of rearing temperature and sex on life-history traits and temperature stress resistance

The differences across sexes and rearing temperatures were largely consistent with previous results (Fischer & Fiedler 2000; Karl & Fischer 2008; Karl et al. 2008). As expected for an ectothermic organism, development times decreased with increasing temperature, accompanied by increased growth rates (see also Blanckenhorn 1997; Fischer, Brakefield & Zwaan 2003; Burke et al. 2005; Van Doorslaer & Stoks 2005). In contrast to earlier findings (Fischer & Fiedler 2000; Karl & Fischer 2008; Karl et al. 2008), individuals reared at the lower temperature did not achieve a higher pupal mass, though this effect can be restricted to the adult phase due to differential weight losses during metamorphosis (Fischer et al. 2004). As predicted from protandry theory (Fagerström & Wiklund 1982), males showed generally shorter development times than females. Further, lower temperatures caused increased cold, but reduced heat stress tolerance. Such plastic responses to the prevailing temperature conditions were also found in a number of other studies (e.g. Ayrinhac et al. 2004; Hoffmann, Shirriffs & Scott 2005; Zeilstra & Fischer 2005; Karl & Fischer 2008). While there was no sex difference in chill-coma recovery time, heat tolerance was reduced in males as compared to females. Similar patterns, again indicating different underlying mechanisms, were also obtained in other studies (e.g. Chen & Walker 1994; Gilchrist, Huey & Partridge 1997; Sørensen, Dahlgaard & Loeschke 2001; Folk et al. 2006; Jensen, Overgaard & Sørensen 2007).


Our study adds to the accumulating evidence that PGI is a locus under thermal selection (Watt 1983, 1994; Dahlhoff & Rank 2000; McMillan et al. 2005). In extension to previous studies, we show that PGI not only affects cold tolerance, but also all life-history traits under investigation here (viz. development time, growth rate, pupal mass). Given the large variation in such traits associated with variation at the PGI locus, PGI can be considered a pleiotropic gene of large effect (although genes linked to the PGI locus may also contribute to the variation found). Most interestingly, the patterns caused by variation in PGI genotype are in broad agreement with those across high- and low-altitude populations, that is, the PGI genotype dominating in high-altitude populations showed increased cold tolerance and rather long development times associated with rather low growth rates (cf. Karl et al. 2008). This genotype is also present in low-altitude populations (c. 35%). Its increase to c. 90% in the high-altitude populations is, based on the current results, likely to be caused by thermal selection on the PGI (and possibly associated) locus (see also Rank & Dahlhoff 2002). This strongly supports the notion that the PGI locus is heavily involved in thermal adaptation in arthropods (Neargarder et al. 2003; McMillan et al. 2005), although it is not related to heat stress tolerance in L. tityrus. Future studies will focus on the mechanisms underlying the association between PGI genotypes and cold tolerance, involving the expression of stress-inducible heat shock proteins.


Financial support was provided by the German Research Foundation (DFG grants Fi 846/1–3 and 1–4 to KF). Insightful comments provided by two referees considerably improved the quality of this manuscript.