The rapid cold hardening response of Collembola is influenced by thermal variability of the habitat

Authors

  • Simon Bahrndorff,

    Corresponding author
    1. Ecology and Genetics, Department of Biological Sciences, University of Aarhus, Ny Munkegade, Building 1540, DK-8000, Aarhus C, Denmark;
    2. Mols Laboraotry, Natural History Museum, Strandkærvej 6-8, Femmøller, DK-8400, Ebeltoft, Denmark;
    Search for more papers by this author
  • Volker Loeschcke,

    1. Ecology and Genetics, Department of Biological Sciences, University of Aarhus, Ny Munkegade, Building 1540, DK-8000, Aarhus C, Denmark;
    Search for more papers by this author
  • Cino Pertoldi,

    1. Ecology and Genetics, Department of Biological Sciences, University of Aarhus, Ny Munkegade, Building 1540, DK-8000, Aarhus C, Denmark;
    2. Mammal Research Institute, Polish Academy of Sciences, Waszkiewicza 1c, 17-230 Bialowieza, Poland;
    Search for more papers by this author
  • Claus Beier,

    1. Risø-DTU, Technical University of Denmark, Biosystems Department, Building BIO-330, P.O. Box 49, Frederiksborgvej 399, DK-4000 Roskilde, Denmark; and
    Search for more papers by this author
  • Martin Holmstrup

    1. National Environmental Research Institute, University of Aarhus, Department of Terrestrial Ecology, P.O. Box 314, Vejlsøvej 25, DK-8600, Silkeborg, Denmark
    Search for more papers by this author

*Correspondence author. E-mail: simon.bahrndorff@biology.au.dk

Summary

  • 1It has been argued that species living under unpredictable thermal conditions need to have more flexible physiological capabilities to meet with thermal stress than species living in thermally stable environments. Here we investigate if the ability to rapidly cold-harden in Collembola is influenced by thermal conditions of the habitat.
  • 2Collembola exploit diverse habitats and are therefore exposed to different thermal environments: soil dwelling (euedaphic) species occupy relatively stable environments, whereas surface dwelling (epedaphic) species can be exposed to more fluctuating thermal environments, but a single species can also be found in diverse thermal habitats within its geographic distribution.
  • 3We compared the inherent cold shock tolerance and ability to rapidly cold-harden in three epedaphic, two near surface dwelling (hemiedaphic) and four euedaphic species of Collembola using a similar experimental approach for all species. Additionally we compared three populations of the epedaphic species, Orchesella cincta, sampled along a climatic gradient (Norway, Denmark, Italy).
  • 4Inherent cold shock tolerance was estimated as LT50 by assaying cold shock survival following a 2 h exposure to a range of temperatures from 1 °C to –12 °C. Rapid cold-hardening (RCH) was induced by cooling individuals from 20 °C to a temperature 7 °C above the LT50 during 80 min, followed by 1 h at the specific cold shock temperature, which was close to the LT50 of the particular species.
  • 5There was large variation in cold shock survival among species. The capacity to rapidly cold-harden was found in all three ecotypes.
  • 6Genetic difference in the ability to rapidly cold-harden was seen in O. cincta from different climatic regions, consistent with the predictability of the thermal environment of their habitat. Population differences matched the daily fluctuations in temperature (CV) recorded at the site of collection as well as the day-to-day predictability (autocorrelation). The role of phylogenetic inertia was tested using sequence data from the cytochrome-c oxidase I (COI) gene and no signal of phylogeny was detected that could explain these population differences.
  • 7Our results show that genetic differences in RCH ability exist, consistent with latitudinal gradients in thermal fluctuations and predictability; thus comparative studies can provide important insight when exploring the role of acclimation in the geographical distribution of species.

Introduction

Cold hardy invertebrates can be divided into freeze avoiding or freeze tolerant species. All Collembola species are freeze avoiding and die if ice forms in their body fluids (Block 2003). Most studies show that Collembola are typical freeze avoiders with high capacity to supercool during winter (Sømme 1982), but some soil dwelling species also utilize cryoprotective dehydration as a cold tolerance strategy (Holmstrup, Bayley & Ramløv 2002). However, even though Collembola in general have a high capacity for supercooling, chill injuries or even mortality, may occur at temperatures above the supercooling point (SCP) in individuals that are not acclimatized to winter conditions (Coulson et al. 1995). This type of cryo-injury was originally defined by Bale (1993) who termed such organisms as chill-susceptible. This issue has since been addressed by many workers in the field and has been synthesized in the reviews of Sinclair (1999) and Sinclair et al. (2003b). Chill injury in arthropods is probably associated with loss of membrane function as cold shock causes dissipation of trans-membrane gradients of Na+ and K+ and a depolarisation of the cell membrane (Kostal, Vambera & Bastl 2004).

Rapid cold hardening (RCH) is a physiological process that can improve resistance to cold shock substantially in many insects and other invertebrates. Thus, when exposed to moderately low temperatures for minutes to hours a number of species will exhibit a RCH response that will subsequently increase chill tolerance (Chen, Denlinger & Lee 1987; Lee, Chen & Denlinger 1987b; Czajka & Lee 1988; Powell & Bale 2004; Shreve, Kelty & Lee 2004). This hardening response can be induced by ecologically relevant thermoperiods (Kelty & Lee 2001) and can affect both survival to cold stress, but also increase the daily time span where, for example, searching for food and reproductive behaviour is possible (Shreve et al. 2004). RCH therefore seems to be an ecologically important trait that would be under strong natural selection. The physiological mechanism underlying RCH is not well understood but evidence is mounting that rapid reorganization of the membrane properties are involved (Overgaard et al. 2005; Lee et al. 2006; Michaud & Denlinger 2006). In some species it has been shown that slight accumulation of low-molecular weight cryoprotectants occurs during RCH (Chen et al. 1987; Michaud & Denlinger 2006; Overgaard et al. 2007) and cold induced heat shock proteins seem also to be involved in some instances (Joplin, Yocum & Denlinger 1990). RCH has been studied in many insect species but only a few studies of Collembola exist with most of these focusing on rapid changes in SCP (Worland & Convey 2001; Sinclair et al. 2003a; Hawes et al. 2006; Worland, Leinaas & Chown 2006; Slabber et al. 2007).

Recently, there has been much focus on the physiological scope for climatic stress tolerance as an underlying basis for Rapoport's rule, the increase in species latitudinal ranges with increasing latitude (Stevens 1989). It has been convincingly argued that species living at higher latitudes experience a more variable climate and therefore need to have physiological capabilities to meet this variation than species at lower latitudes where climate is less variable. This also applies to acclimation responses such as RCH as a form of phenotypic plasticity (Addo-Bediako, Chown & Gaston 2000; Chown, Addo-Bediako & Gaston 2002; Chown & Terblanche 2007). Stevens (1992) extended Rapoport's rule to elevational gradients in altitudinal range because similar gradients in climatic variability often exist in these cases (see review by Chown & Terblanche 2007).

Collembola, with their stratified distribution along the soil/air boundary, offer an interesting opportunity to study the occurrence of RCH as an adaptive trait in species along an extended gradient of thermal fluctuation similar to an elevational gradient, albeit at a much smaller spatial scale. The subterranean habitat of the soil dwelling species secures a relatively stable thermal environment (Willmer 1982). Surface dwelling or near surface dwelling Collembola, on the other hand, can experience an increased range of thermal fluctuation with much more rapid shifts between extremes. Therefore it may be hypothesised that the capacity for RCH is more developed in those species associated with the soil surface than in soil dwelling species, where RCH could seem redundant. Additionally, many Collembola, such as Orchesella cincta, also occur along wide latitudinal gradients and have adapted to local thermal conditions (Bahrndorff et al. 2006), which could lead to different selection pressures on the capacity for RCH. We therefore tested (i) cold shock tolerance and the effect of RCH on cold shock tolerance in nine species of European Collembola representing three ecotypes, (ii) cold shock tolerance and the effect of RCH on cold shock tolerance in three populations of O. cincta collected along a climatic gradient, and (iii) how microhabitat temperature measurements varied between habitats and correlate with the responses to RCH. Furthermore, in order to distinguish the contributions of the various processes that underlie the pattern, we tested the RCH response for phylogenetic inertia due to phylogenetic relatedness between populations using sequence data from the COI gene.

Methods

maintenance and origin of collembola

Nine species of Collembola which in their natural habitat occur at different depths in the soil profile were used in the experiment (Table 1): three surface dwelling species, Sinella curviseta, O. cincta and Orchesella sp., four soil dwelling species, Protaphorura fimata, Mesaphorura macrochaeta, Folsomia candida and Folsomia fimetaria and two hemiedaphic species (near surface), Hypogastrura assimilis and Proisotoma minuta.

Table 1.  Phylogenetic relationships, ecotype and geographic origin of the Collembola used in the study. LT50 temperatures (mean) for nine species of Collembola based on survival following 2 h of cold exposure
SuperfamilyFamilySpeciesEcotypeOriginLT50 (°C)
EntomobryoideaEntomobryidaeSinella curvisetaEpigeicUSA–1·9
Orchesella cinctaEpigeicDenmark–7·4
Orchesella sp.EpigeicItaly–4·9
OnychiuridaeProtaphorura fimataEuedaphicGermany–9·2
Mesaphorura macrochaetaEuedaphicDenmark–11
IsotomidaeFolsomia candidaEuedaphicGermany–6·5
Folsomia fimetariaEuedaphicDenmark–8·8
PoduroideaHypogastruridaeHypogastrura assimilisHemiedaphicDenmark–11·9
Proisotoma minutaHemiedaphicDenmark–10·8

Three separate populations of O. cincta were collected from each of the following locations: Midtlæger, Norway (59°83 N, 10°79 E), Silkeborg, Denmark, (56°17 N, 09°57 E) and Siena, Italy (43°32 N, 11°32 E). Individuals of O. cincta were synchronized in age (±3 days) and not used before all individuals from all populations had reached maturity. Furthermore, these individuals were reared through to the second generation in the laboratory before LT50 and the RCH response were tested. Orchesella sp. was collected on Sicily, Italy, and held in the laboratory for three generations until used in experiments. The two Orchesella species were held in transparent plastic boxes containing a substrate of water-saturated plaster of Paris : charcoal mix (ratio 9 : 1) and fed green algae growing on small twigs. Twigs were collected in the field continuously and frozen at –20 °C for 5 days prior to use.

The other Collembola species were held in Petri dishes containing water-saturated plaster of Paris : charcoal mix and fed on dried baker's yeast. These species had been held in laboratory cultures for 3–5 years (approximately 30–60 generations) before the experiments were done. All cultures were kept at 20 ± 1 °C in a 12 : 12 light : dark regime.

Temperature data (year 2003) were obtained both at different depths in the soil profile, but also from weather stations close to the site of collection of the three populations of O. cincta. This provided documentation of the temperature regimes and in particular the thermal variability and predictability that the collembolans are exposed to under natural conditions. Temperature recordings every second hour were obtained at different depths in the soil (air, 2 and 10 cm into the soil) from Mols, Denmark (DK) (56°23 N, 10°57 E) from an already existing project, ‘Vulnerability assessment of shrubland ecosystems in Europe under climatic changes’ (VULCAN) (Beier et al. 2004). Air temperature recordings (2 m) obtained from local weather stations were used to predict the temperature exposure that O. cincta would experience at the different geographical sites of collection. These were obtained from Midtlæger, Norway (59°83 N, 10°79 E) (http://eklima.met.no), Mols, Denmark (DK) (56°23 N, 10°57 E) (Beier et al. 2004) and Ciampino, Italy (41°47 N, 12°35 E) (Tang, Waring & Hong 2007). Daily mean temperature recordings from Italy were available from Ciampino and Bologna, and Ciampino was judged to represent Siena best. These recordings were used to calculate temperature fluctuations. Hourly temperature recordings from two sites close to Siena, Italy: Fagna and Montevarchi (http://www.lamma.rete.toscana.it/) were used for autocorrelation analysis.

cold shock tolerance

Cold shock tolerance was tested by exposing adult Collembola to a range of temperatures from 1 °C to –14 °C causing mortality proportions between 0% and 100%. Pilot studies showed that Collembola acclimated to 20 °C were ‘chill susceptible’ (sensu Bale (1993)). Individuals were exposed in 1·5 mL Eppendorf tubes for 2 h in custom made constant temperature cooling cabinets accurate within ±0·2 °C. Five replicates of 10 individuals were used for each temperature tested. Survival was assessed after 24 h recovery at 20 °C on the basis of animals able to walk in a coordinated fashion following gentle tactile stimulation with a fine brush.

rapid cold hardening (RCH)

The ability of collembolans to rapidly cold harden was evaluated by comparing cold shock survival in two experimental groups. The cold shock temperature of the different species was chosen on the basis of the first experiment and aimed at a mortality level of 50%, that is, LT50, or higher. The cold shock temperatures chosen were: S. curviseta (–2 °C), O. cincta (–7 °C), Orchesella sp. (–6 °C), P. fimata (–10 °C), M. macrochaeta (–11 °C), F. candida (–7 °C), F. fimetaria (–9 °C), H. assimilis (–12 °C) and P. minuta (–12 °C). Using Collembola of the same cultures as described above individuals were exposed directly to the 2 h cold shock treatment or with a prior cold hardening treatment that consisted of a linear decrease of temperature over 80 min from 20 °C to a temperature 7 °C above the LT50 for the relevant species. The Collembola were held for an additional 60 min at this temperature before they were exposed to the final cold shock temperature treatment. This protocol is similar to that used in other studies and has proven successful in inducing a hardening response (Kelty & Lee 1999; Overgaard et al. 2005). Controls (no RCH) were held at conditions similar to RCH treated animals at 20 °C until exposure to the final low temperature treatment. The cold hardening treatment was performed using a programmable Binder MK heating/cooling cabinet (Binder, Tuttlingen, Germany) accurate to ±0·2 °C. Five replicates of 10 individuals were tested for each treatment. Survival was assessed 24 h after exposure as described previously.

dna extraction

Genomic DNA was extracted from whole-body homogenates using a proteinase K method. A 710-bp region of the mitochondrial COI gene was amplified by polymerase chain reaction (PCR) from the DNA of two individuals from each population or species using the primers LCO1490 and HCO2198 (Folmer et al. 1994). These primers have proved useful for resolving taxonomic relatedness in Collembola (Hogg & Hebert 2004). The 20-µL PCRs contained 10 µL master mix red (1·5 mm MgCl2; Ampliqon), 1 µL of each primer (10 pmol µL−1) and 1 µL of DNA template. The PCR conditions were 3 min at 94 °C; 35 cycles of 30 s at 94 °C, 40 s at 50 °C, and 1 min at 72 °C, and 7 min at 72 °C. PCR products were cleaned using gel purification kit (E.Z.N.A. gel extraction kit, Omega Bio-Tek) and sequenced in both directions using the primers LCO1490 and HCO2198 on an ABI automated DNA sequencer at Macrogen Inc. The accuracy of sequencing was verified by duplicate sequencing of DNA from selected individuals and sequencing in both directions. The resulting sequences were edited and aligned using the program Geneious Pro, which was also used for UMPGA analysis (version 3·5·6; Biomatters Ltd.). All the sequences obtained have been submitted to GenBank (EU869803–EU869806).

statistical analysis

Sigmoidal dose–response curves were fitted to survival data using the software Microcal Origin® (Microcal Software, Northampton, MA) and associated LT50 values (the temperature where 50% of the individuals died) were calculated for each species. The effect of RCH on each species was estimated using an unpaired t-test. In two cases, Orchesella sp. and P. minuta a nonparametric test (Mann–Whitney U) was used to test for the effect of RCH, as this test compares the medians which are less affected by deviations from a normal distribution. Survival proportions were arcsine-square-root transformed to improve normality and homogeneity of variances. Levene's test was used to test for equality of variances. Two-way anova was used to test for population and treatment (RCH or no RCH treatment) effects, and interaction of population and treatment on survival for O. cincta. Scheffe's post-hoc test was used to test for population differences in cold shock tolerance and RCH response. Analyses were performed using the statistical package spss (SPSS Inc., Chicago, IL). past version 1·78 (Hammer, Harper & Ryan 2001) was used to carry out autocorrelation analysis on temperature data from Italy, Denmark and Norway, following the recommendations of Deere & Chown (2006) and Chown & Terblanche (2007). Differences in the coefficient of variation of daily mean temperature between sites or soil layers were tested using an F-test following Donnelly & Kramer (1999). A correction for multiple comparisons was performed on all the multiple tests controlling for the false discovery rate (Garcia 2003).

A test for serial independence (TFSI) to determine whether there was a significant positive autocorrelation on RCH was conducted, in order to take into account the bias produced by phylogenetic inertia, phenotypic response (acclimation) and genetic adaptation. To evaluate whether a phylogenetic effect exists among the O. cincta populations, we used the software PI (http://biology.mcgill.ca/faculty/abouheif) (Abouheif 1999). Based on sequence data of the COI gene the topology of the three O. cincta populations was determined. The topology was randomly rotated 2000 times per iteration for all the simulations. To provide the null hypothesis sampling distribution, the observed and randomized mean C-statistics were estimated by rotating the nodes within a given phylogenetic topology, where each individual node had a probability of 50% of being randomly rotated. Furthermore, the original data were shuffled 2000 times. One-tailed probabilities were used to assess the statistical significance of the TFSI as we are only testing for a positive autocorrelation (self-similarity due to phylogenetic descent). TFSI was chosen because it is preferable to other phylogenetic autocorrelation methods when dealing with small data sets such as ours (Martins & Hansen 1997).

Results

Two measures were used to analyze the temperature recordings. The coefficient of variation and the autocorrelation were used to analyze the fluctuations and predictability in temperature, respectively, between sites and at different layers in the soil profile. At the Mols site, Denmark, the fluctuations in temperature (measured as CV) were larger in the top part of the soil (air) as compared with deeper in the soil (soil, –2 cm and soil, –10 cm) (see Table S1 in Supporting Information). Fluctuations in temperature (measured as CV) were largest in Norway, followed by Denmark and then Italy, respectively (Table 2). Day-to-day predictability of temperature changed over season at Mols, Denmark. Daily temperatures were predictable for up to 14 days for the air temperature measurements in July, with temperatures strongly correlated with those 24 h earlier. This is reflected in a significant positive autocorrelations at lags of 24 h and multiples thereof (see Fig. S1 in Supporting Information). In contrast to this air temperatures in November were significantly dissimilar to those experienced 24 h before and this became even more pronounced when looking at soil temperatures (see Fig. S1). However, it is seen that the autocorrelation plots do not change substantially down through the soil layers in March, July and November (see Fig. S1). The predictability or autocorrelation also changed significantly when comparing the Italian, Danish and Norwegian sites (Fig. 1). At the Italian site temperatures showed a positive autocorrelation that persists for up to 14 days although its length and magnitude decreased in December. In contrast to this the autocorrelation plots from Norway show that current temperatures only provide an indication of future temperatures for around a day and a half (Fig. 1). The highest autocorrelation was noted at the Italian site showing that climate is more predictable here than in Norway or Denmark.

Table 2.  Coefficient of variation (CV; mean percentage) in daily temperature in Norway, Denmark and Italy and F-test values. Asterisks indicate significant differences in CV between sites (*P < 0·05; **P < 0·01). Data are log(x + 10) transformed. See text for locations
MonthsSiteDK v. IN v. IN v. DK
Norway (N)Denmark (DK)Italy (I)FPFPFP
January691·72454·36181·182·51*3·82**1·52 
February449·80168·43131·251·28 3·43*2·67*
March297·12229·13111·532·05 2·66*1·30 
April394·18330·38220·741·50 1·79 1·19 
May227·99213·22115·861·84 1·97 1·07 
June199·57163·59127·881·28 1·56 1·22 
July216·36174·55 94·771·84 2·28 1·24 
August260·10226·76114·551·98 2·27 1·15 
September228·55224·08119·931·87 1·91 1·02 
October338·50274·67196·291·40 1·72 1·23 
November278·54150·47127·771·18 2·18 1·85 
December539·82292·79268·931·09 2·01 1·84 
Figure 1.

Autocorrelation plots for temperature (air) recorded in Norway (Midtlæger), Denmark (Mols) and Italy (Fagna and Montevarchi (June)) at three different times in a year (June, September and December). The grey lines represent the 95% confidence intervals. The values on the y-axis are the autocorrelation coefficients and hourly time lag on the x-axis.

species differences in response to rch

We first assessed the inherent cold shock tolerance of each species (see Fig. S2) to obtain estimates of LT50 of un-acclimated Collembola. LT50 differed substantially between species where H. assimilis showed the lowest LT50 (–12 °C) and S. curviseta the highest LT50 (–1·9 °C) (Table 1).

The two surface dwelling species S. curviseta and O. cincta increased their survival rate significantly from 16% and 20% to 80% and 90%, respectively, when pre-exposed to a RCH treatment, whereas Orchesella sp. significantly increased its survival rate from 0% to 24% when pre-exposed to a RCH treatment. The two hemiedaphic species, H. assimilis and P. minuta, also increased their survival rate from 46% to 81% and 1% to 71%, respectively, although this was not significant for H. assimilis. The soil dwelling species, F. fimetaria and F. candida significantly increased their survival rate from 23% and 8% to 64% and 50% when pre-exposed to a RCH treatment, whereas M. macrochaeta did not significantly increase the survival rate. Protaphorura fimata was the only species that did not increase survival when pre-exposed to a RCH treatment (Fig. 2). Comparing the response to RCH between surface (including hemiedaphic) and soil dwelling species, the largest response was seen in surface dwelling or hemiedaphic species.

Figure 2.

The effect of rapid cold hardening on cold shock survival in nine species of Collembola. Survival proportions (mean ± SE, n = 5) after 2 h of cold exposure are given. Individuals were either rapidly cold hardened before exposure (RCH) or exposed directly (no RCH) to low temperature treatment. Species names and specific temperature exposure are indicated on each graph. Asterisks indicate significant differences in survival between treatments (*P < 0·05; **P < 0·01).

population differences in response to rch

When comparing cold shock tolerance between populations, the Norwegian and Danish populations of O. cincta had a survival rate of around 25% whereas the Italian population had a survival rate of 14% (Fig. 3). When the populations were pre-exposed to a RCH treatment the pattern changed substantially. Survival rate increased in all three populations: the Norwegian population to 90% and the Danish and the Italian populations to 65% and 45%, respectively. Two-way anova indicated significant effects of population (F = 11·381; P < 0·001), treatment (F = 86·726; P < 0·001) and interaction between population and treatment (F = 4·584; P < 0·05) suggesting local adaptation with respect to RCH (Fig. 3). A post-hoc test showed that survival in the Norwegian population was significantly higher than in the Danish and Italian populations (P < 0·05 and P < 0·01, respectively).

Figure 3.

The effect of rapid cold hardening on cold shock survival in three populations of Orchesella cincta. Survival rates (mean ± SE, n = 5) after 2 h of cold exposure are given. Individuals were either rapidly cold hardened before exposure (RCH) or exposed directly (no RCH) to low temperature treatment (–7 °C). Populations tested were from Norway (Bergen), Denmark (Silkeborg) and Italy (Siena).

genetic distance and phylogeny

A 687-bp fragment of the COI gene was sequenced from individuals of O. cincta and O. bifasciata (used as outgroup) and used to establish the phylogeny between populations. Nucleotide composition in O. cincta was A = 25·8%, T = 34·7%, C = 19·5% and G = 19·0%, showing an A–T bias. The UMPGA analysis showed that individuals of O. cincta from Norway and Denmark grouped together with very low differentiation, and individuals from Italy grouped together. As expected O. bifasciata branched out at the base node (see Fig. S3). We found no significant phylogenetic autocorrelation for RCH (P > 0·05) and the incorporation of phylogeny in further statistical analyses was therefore not necessary (Gittleman & Kot 1990; Abouheif 1999).

Discussion

RCH can rapidly lower the temperature threshold of survival, increasing survival of cold shock and it may also have a positive effect on fitness related traits such as mating and feeding activity at low temperatures (Kelty & Lee 1999). RCH can therefore enable insects to respond quickly to a changing thermal environment and maintain critical functions such as locomotory activity and fecundity when exposed to otherwise critical temperatures (Chen et al. 1987; Lee, Chen & Denlinger 1987a; Czajka & Lee 1990; Powell & Bale 2004). These clear fitness effects suggest that the ability to RCH may be under strong natural selection.

We used cold shock survival as a proxy for the beneficial effects of RCH in an attempt to elucidate the evolutionary and ecological significance of RCH at the inter- and intraspecific level in Collembola. Collembola show diverse physiological adaptations to the environmental stress that they experience depending on the habitat, geographically or locally, in which they live (Joosse & Verhoef 1987; Holmstrup et al. 2002; Kærsgaard et al. 2004). The temperature recordings we obtained from different soil layers and at different geographical locations (Table 2 and Table S1) show that fluctuations are largest in the air as compared with the surface and soil, but also increasing from southern towards northern Europe. The predictability on the other hand (see Fig. S1) changes over seasons, but does not change much between air, surface and soil. Also predictability is much lower at the Norwegian site than at the Danish and Italian site (Fig. 1). Thus RCH should be under strongest selection in the Northern population of O. cincta due to lower predictability and higher fluctuations. On the other hand, this pattern is not as pronounced locally, at least when comparing the microclimate at the surface and in the soil. Thus, the climatic patterns both locally and on a regional scale are in good agreement with the differences in RCH response between populations of O. cincta and between species with different habitat choice in the soil profile.

Keeping insects in laboratory cultures for extended periods may lead to changes in stress-related traits because the conditions (especially a regular supply of rich food resources) are much different from natural conditions (Harshman & Hoffmann 2000). In the present study, constant culturing temperature (20 °C) could have reduced the ability to rapidly cold-harden. Nevertheless, eight out of nine species tested in the present study benefited from a RCH pre-treatment in terms of cold shock survival, whereas one soil dwelling species, P. fimata, was negatively influenced. Three surface- or near surface dwelling species (O. cincta, S. curviseta and P. minuta) showed the most dramatic effect of RCH. Orchesella sp. from Sicily also increased its cold shock survival due to RCH even though this species rarely would be exposed to very low temperatures in its natural habitat. This observation is consistent with studies of Chen, Lee & Denlinger (1990) who found that tropical lowland flesh fly species did not benefit much from a prior RCH treatment as compared with those from more temperate areas. There was a weak trend that the surface dwelling species benefited most from a RCH treatment on cold shock tolerance (Fig. 2). There are different reasons that could explain why we did not detect any clear differences in response to a RCH treatment between ecotypes. First, this could be explained by the fact that the CV of temperature was only significantly different between air and surface/soil (not between soil surface and deeper soil layers) and that the species tested in this study might be able behaviourally to avoid these temperature fluctuations by migrating deeper into the surface–soil interface. Also the predictability does not seem to differ much between air, surface and soil. Second, in trying to standardize the RCH protocol for the whole assembly of species (see Methods), we may not have chosen the acclimating treatment giving the strongest RCH response in every species. Terblanche et al. (2007) have shown for the tsetse fly that depending on the acclimation procedure, critical thermal limits may differ considerably. However, such a multifactorial test design was outside the scope of our study, and comparing RCH effects obtained with different protocols between species would also have been fraught with difficulties. Gaston, Chown & Evans (2008) discuss the problems associated with interspecies comparisons of physiological traits across geographic and climatic gradients. Our results represent a similar comparison although our case is not spanning a geographic gradient, but a vertical one (along soil depth) which we assumed would also represent a gradient in temperature variability over a small distance. As pointed out by Gaston et al. (2008) interspecific patterns in a trait (here: RCH in surface vs. soil dwelling species) may be difficult to show if the populations tested originate from a limited portion of the particular species’ range. In other words, the ideal situation would have been to have populations of all nine species from the very same site, and preferably have repeated sets of comparisons from other parts of the geographic range of the species.

Whereas our data do not support the hypothesis that surface dwelling species are more ‘plastic’ than soil dwelling species, it is evident that climate (and predictability) strongly varies with latitude and that variation in traits related to fitness along latitude can be the result of evolutionary adaptation to different climatic conditions. Evaluating traits in natural populations can provide indirect evidence of environmental adaptation to the local thermal environment (Hoffmann, Sørensen & Loeschcke 2003). The present study clearly shows genetic differentiation in the significance of RCH between populations of O. cincta. As would be expected, the northern population experiencing the coldest and most fluctuating (but least predictable) environment also benefited most from the RCH treatment. The present results are in accordance with the findings of Chen & Walker (1993) suggesting the presence of genetic variation in cold shock selected lines of Drosophila melanogaster that had been exposed to a prior RCH treatment. Furthermore, they are in agreement with other studies on O. cincta which showed evidence of thermal adaptation along environmental gradients (Bahrndorff et al. 2006). Also the soil dwelling species, M. arctica, shows evidence of thermal adaptation to cold and drought along environmental gradients (Bahrndorff et al. 2007), although in this study the environmental and genetic components could not be separated. It would therefore be expected that population's adapted to colder conditions (highly fluctuating and unpredictable) would show an increased response to a RCH treatment that is genetically inherited. This is in concordance with our present study. The hypothesis that RCH is under strong natural selection is also supported by the non-significant results of the TFSI, which allows us to exclude the possibility that observed population differences were the result of phylogenetic constraints. Our conclusions do not suffer from weaknesses associated with the assumptions of an inappropriate model of evolutionary change since TFSI does not require branch lengths and does not assume any model of evolutionary change (Abouheif 1999; Reeve & Abouheif 1999).

Acknowledgements

This study was made possible with support from the international graduate school ISOBIS and the Natural History Museum of Aarhus (SB) and by grants from the Villum Kann Rasmussen Foundation (MH, CB) and the Danish Natural Sciences Research Council (VL, CP). This study has been partly supported by a Marie Curie Transfer of Knowledge Fellowship BIORESC of European Community's Sixth Framework Program (contract number MTKD-CT-2005-029957) and the Con Gen program (funded by the European Science Foundation (grant number: 21-01-0526)) (CP). The authors thank F. Frati and G. Spinsanti for sending individuals of O. cincta from Siena. The authors also thank T.C. Hawes, S.L. Chown and J.S.F. Barker for valuable and constructive comments that greatly improved an earlier version of the manuscript.

Ancillary