More complex than expected : Cold hardiness and the concentration of cryoprotectants in overwintering larvae of fi ve Erebia butter fl ies ( Lepidoptera : Nymphalidae )

Understanding the factors restricting the distribution of some insect species to high altitudes is hindered by poor knowledge of temporal changes in their cold hardiness during overwintering. We studied overwintering larvae of fi ve species of Erebia butterfl ies (Lepidoptera: Nymphalidae: Satyrinae) differing in altitudinal distribution: lowland E. medusa, submountain E. aethiops, subalpine E. pronoe, alpine E. cassioides, and subnivean E. pluto. We subjected them to three treatments, AutumnWarm (13/8°C), imitating conditions prior to overwintering; AutumnCold (5/0°C), imitating late autumn conditions; and WinterCold (5/0°C), differing from AutumnCold by a shorter photoperiod and longer exposure to zero temperatures. Supercooling points (SCP) did not differ between species in the AutumnWarm treatment, despite large differences in the concentrations of cryoprotectants (CrPC; lowest in E. medusa and E. aethiops). Lowland E. medusa was freeze-tolerant, the subalpine, alpine and subnivean species were freezeavoidant, whereas submountain E. aethiops displayed a mixed strategy. SCPs diverged in the AutumnCold treatment: it increased in the lowland E. medusa (from –16.5 to –10.8°C) and reached the lowest value in E. cassioides (–21.7°C). In WinterCold, SCP increased in subalpine E. pronoe (from –16.1°C in AutumnWarm and –18.7°C in AutumnCold to –12.6°C). E. medusa decreased and E. aethiops increased their CrPCs between autumn and winter; the highest CrPC was recorded in subnivean E. pluto. CrPC did not correlate with SCP across species and treatments. Cryoprotectant profi les corroborated the difference between lowland and freeze-tolerant E. medusa and the three high altitude freeze-avoidant species, with E. aethiops in an intermediate position. Glycerol was surprisingly rare, trehalose was important in all species, and such rare compounds as monopalmitin and monostearin were abundantly present in E. pronoe, E. cassioides and E. pluto.


INTRODUCTION
Debates on the effects of climate change on biodiversity are much concerned with future fates of cold-adapted range restricted species, such as Lepidoptera in high altitude habitats within temperate zones, under a warming climate (Franco et al., 2006;Konvicka et al., 2014;Scalercio et al., 2014;Konvicka et al., 2016).In this context, it is notable how little is known regarding the mechanisms restricting cold-adapted species to their high altitude habitats.The mechanisms proposed so far include dependency on treeless habitats (and associated plants), which may shrink in extent with increasing timberline (Bila et al., 2016;Roland & Matter, 2016); dependency of overwintering immature stages on stable snow cover for winter insulation and hence threats due to less predictable snow (Vrba et al., 2012;Roland & Matter, 2016); the requirements of adults for thermal heterogeneity (Kleckova & Klecka, 2016;MacLean et al., 2016); or increased pressure from parasitoids or other natural enemies (cf.Pere et al., 2013).A wide variety of species differing in altitudinal distribution and their habitats, but contrary to the earlier study, we simulated the thermal conditions the larvae most likely experience in nature: moderately low temperatures prior to freezing, autumn conditions close to freezing, and mid-winter conditions.We asked how these conditions are mirrored in the SCP, assayed the larval polyol profi les and compared the situations within and across species.We hypothesized that (1) species whose larvae experience unpredictable snow cover will be more cold-hardy than those experiencing more predictable insulating snow cover in their natural habitats; (2) cold hardiness will be highest under simulated winter conditions; (3) polyol content will negatively correlate with the super cooling point; and (4a) polyol profi les will be similar in closely related species or, alternatively (4b) species experiencing similar thermal conditions will be more similar in their polyol profi les.

Species studied
All the species of Erebia studied are univoltine, i.e. they have a single generation per year.They all feed on grasses and overwinter as larvae, low in grass tussocks, attached by their legs to the bases of grass stems (Sonderegger, 2005).In the high altitude species, prolonged two-year (semivoltine) development may occur, varying among sites and years (Kleckova et al., 2015;Zakharova & Tatarinov, 2016).The fi ve species studied form an altitudinal cline, as far as their maxima and optima are concerned.E. medusa (Denis & Schiffermüller, 1775) is a lowland species.It mainly occurs at low altitudes, but can ascend to ca 2000 m a.s.l. in the Alps.Its distribution ranges from Central France to Northern China.It inhabits late-successional grasslands, e.g.abandoned meadows and pastures, forest-steppes, grassy woodland clearings and margins (Schmitt, 2002;Stuhldreher & Fartmann, 2015).Larvae overwinter in the third instar (Sonderegger, 2005).
E. aethiops (Esper, 1777) is a submountain species.Its Eurosiberian distribution ranges from France to Western Siberia.It inhabits sparse woodlands, grassy forest clearings and margins (e.g., Slamova et al., 2013) from sea level to about 1500 m a.s.l. in the Alps.It overwinters as a second-instar larva (Sonderegger, 2005).
E. pronoe (Esper, 1780) is a subalpine species.This European species inhabits most of the continent's high mountains, except in Scandinavia and mountains surrounding the Mediterranean basin.It prefers mountain grasslands (Kleckova et al., 2014) from about 900-2800 m a.s.l.It overwinters as small, fi rst instar larvae (Sonderegger, 2005).
E. cassioides (Reiner & Hochenwarth, 1792) is an alpine species.Distributed in the Eastern Alps and Apennines at altitudes of about 1600-2600 m.It is a member of the taxonomically diffi cult E. tyndarus complex, in which populations from the Western Alps and Balkan peninsula were recently considered to be separate species (Schmitt et al., 2016).The habitats are sparsely vegetated rocky substrates (Kleckova et al., 2014).Larvae overwinter in the fi rst or second instar (Sonderegger, 2005).
E. pluto (De Prunner, 1798) is a subnivean species.European populations, occur in the Alps and central Apennines.It uses the most extreme habitats, sparsely vegetated unstable screes in alpine to subnivean zones (1800-3000 m a.s.l.).Larval development lasts two years, the larvae overwinter either in the second or the last (fi fth) instar (Sonderegger, 2005).
Survival during winter is important in determining an insects' abundance and requires major restructuring of physiological functions: an individual interrupts its development and shifts to maintenance processes and to tissue and cell protection (Lee, 1991) In numerous temperate zone species, this involves diapause (Denlinger, 1991), a process initiated long before the onset of cold temperatures, in late summer or early autumn, and entering this phase depends on appropriate reading of external, typically photoperiodic, clues (Kostal et al., 2014).Some species, however, just increase their cryoprotectant content while not undergoing true diapause (Pullin & Bale, 1989c).
There are two major strategies for surviving winter frosts (Sinclair et al., 2015), both of which are reported for species of Erebia (cf.Vrba et al., 2012).Freeze-tolerant insects allow for the freezing of extracellular body liquids, whereas freeze-avoidant species decrease the freezing temperature of body fl uids so that they do not freeze; some insects may display mixed strategies, depending on presence or absence of external ice (Sinclair et al., 2015).Survival at low temperatures is facilitated by synthesis of polyols, low-molecular sugars and alcohols that decrease the temperature of crystallization of body fl uids (= supercooling point, SCP), and contribute to the protection of cell membranes and proteins (Ramlov, 2000).Examples of cryoprotective polyols in Lepidoptera include glycerol, sorbitol, glucose, fructose and trehalose (Pullin & Bale, 1989b, c;Vrba et al., 2014a;Williams et al., 2014).There is only fragmentary information on how individual polyols contribute to cold hardiness, how they are connected with the two strategies, how their content varies during winter and how it varies among related species.It seems that usually, one or two compounds found in high concentrations colligatively modify the supercooling point, whereas several rarer compounds contribute non-colligatively to stabilizing macromolecules and biomembranes (Storey & Storey, 1991;Kostal et al., 2001).
Comparison of cold hardiness in four Erebia species whose overwintering larvae were acclimated in constant conditions (Vrba et al., 2012) revealed freeze-tolerance in a lowland species, E. medusa (Denis & Schiffermüller, 1775), and freeze-avoidance in three species from mountain and alpine environments: E. sudetica (Staudinger, 1861), E. epiphron (Knoch, 1783) and E. tyndarus (Esper, 1781).These four species respond counter-intuitively to low temperatures: the lowland species is more cold hardy (surviving -21°C) than its mountain or alpine congeners (Vrba et al., 2012).It is not known, however, how cold hardiness in Erebia spp. is connected to polyol profi les, or how the profi les change depending on external conditions.
Here, we study the overwintering of Erebia larvae in more detail.As in Vrba et al. (2012)  The females were kept in plastic boxes, in groups of up to 5 individuals, containing a selection of grasses from the original localities.The eggs were allowed to hatch, the newly hatched larvae were placed into cooled incubators (Sanyo MIR-153), where the outdoor natural photoperiod was maintained by adjusting manually on a weekly basis.Newly hatched larvae of all the species studied were reared under a summer temperature regime (20°C day / 10°C night) until they reached their respective overwintering instars in September.Duration of this summer regime thus depended on species' phenology (larval hatching -E.medusa: mid-June, E. pluto: early August, E. aethiops, E. pronoe, E. cassioides: mid-August).
In mid-September, larvae of each species were divided into two groups, which were subsequently reared under two different thermoperiodic regimes: warm treatment (13°C day / 8°C night) and cold treatment (5°C day / 0°C night), with the photoperiod following outdoor conditions.For SCP identifi cation and cryoprotectant assays, we sampled 16 and 10 larvae, respectively, at two different phases of overwintering: autumn (end of October, i.e. after 45 days exposure: AutumnCold and AutumnWarm treatments) and winter (mid-January, i.e. 4 months: WinterCold treatment only).Samples for the SCP and cryoprotectant assays were always taken on the same day for a single species.

Cold hardiness measurements
The SCP was measured individually using a PICO recorder with hand-made type K thermocouples (Hanson & Venette, 2013), which were attached to the body of the experimental caterpillars in a syringe (Brunnhofer et al., 1991) using zincoxide thermoconductive paste.Starting at 5°C, the larvae were gradually cooled in a Calex desktop freezer with a manually controlled cooling rate of 1°C per minute [considered fast enough for this experiment and still biologically realistic (Nedved et al., 1995)] by gradual submerging the syringe with a caterpillar into a box containing aluminium marbles.After the exotherm was detected, the larva was kept in the cooling device until its body temperature decreased again to the value of the crystallization temperature and then removed and warmed up at room temperature.

Cryoprotectant analysis
The cryoprotectant concentration (CrPC) assays followed procedures described in (Kostal & Simek, 1995;Kostal et al., 2007).The larvae were weighed, stored frozen at -80°C, then thawed, homogenized in 70% ethanol and the samples centrifuged.The supernatant was purifi ed using a hexane treatment and after drying in a hydrogen stream derivatized with O-methylhydroxylamine (oximation, 80°C for 30 min) and trimethylsilylimidazol (silylation, 80°C for 30 min).After derivatization and re-extraction into isooctane, 1 μl aliquots were injected using an AOC-20s autosampler into a gas chromatograph Shimadzu GC-2014 equipped with a fl ame ionization detector and controlled by GC Solution software (all from Shimadzu, Japan).The concentrations of lowmolecular mass sugars and polyols (putative cryoprotectants) were determined after separation on a DB-1MS capillary column (29 m × 0.25 mm, 0.25 μm fi lm thickness from Agilent Technologies, USA).Hydrogen carrier gas fl ow rate was 1.25 mL/ min.The injector and detector temperatures were 270 and 320°C, respectively.The GC oven temperature program started at 140°C, held for 0 min, programmed at 10°C/min ramp to 200°C, held 2 min, programmed at 7°C/min to 235°C, held 0 min, 50°C/min to 280°C and held 6 min.Gas chromatography coupled to mass spectrometry (Trace Ultra gas chromatograph with programmed temperature vapourizing injector and GC/MS DSQ quadrupole instrument with electron and chemical ionization, both from Thermo Fisher Scientifi c, USA) and a capillary column DB 1-MS (30 m × 0.25 mm, 0.25 μm fi lm thickness from Agilent Technologies, USA) were used for polyol identifi cation by comparing the spectra to those of authentic standards.Helium carrier gas fl ow rate was 1.1 mL/min, 1 μl injected, injector temperature 180°C, MS ion source and transfer line temperatures were set to 220 and 280°C, respectively.The oven temperature program was the following: initial temperature 110°C, held for 0 min, 10°C/min ramp to 182°C, held 0 min, 2°C/min ramp to 196°C, held 0 min, 35°C/ min ramp to 300°C and held 2.5 min.The chemicals used were purchased from Sigma-Aldrich Co. Carbohydrate System Check Mix from RESTEK, USA, was used as a QC control for checking polyols quantifi cation.Concentrations were expressed as μg/mg fresh body mass.

Statistical analysis
We originally intended to subject all fi ve species to all three treatments and thus to attain a balanced experimental design.However, we did not obtain suffi cient numbers of larvae of E. cassioides and E. pluto.Therefore, only three species (E.aethiops, E. medusa, E. pronoe) were subjected to all three treatments; E. cassioides was subjected to AutumnCold and WinterCold treatments, and E. pluto to AutumnCold treatment only (cf.Fig. 1).This unbalanced design required re-calculating some of the below analyses separately within species and treatments.SCPs and CrPC s were compared separately for AutumnCold (all fi ve species), WinterCold (four species) and AutumnWarm (three species) using one-way ANOVAs.Factorial ANOVAs were used to compare SCPs from the two cold treatments, Autumn-Cold and WinterCold (four species, less E. pluto) and all three treatments (three species: E. aethiops, E. medusa and E. pronoe).We also correlated average SCP and CrPC values across all species and treatments, expecting that low SCP would be associated with high CrPC.
For analyses of cryoprotectant profi les, we used a multivariate ordination technique, the redundancy analysis (RDA) computed in CANOCO 5 (Ter Braak & Smilauer, 2013).RDA relates the composition of samples, here, the specifi c cryoprotectant concentrations measured per individual, to predictors characterizing the samples (i.e., species identity, treatment) and tests the significance of the ordination using Monte Carlo tests.We used nontransformed data and 999 permutations for each Monte Carlo test.
We ran the RDAs fi rst separately within treatments, i.e. with fi ve species for AutumnCold, four species for WinterCold and three species for AutumnWarm.These analyses visualized and statistically tested the differences among species subjected to these three treatments.Subsequently, for the four species for which at least two experimental treatments were available, we ran separate single-species analyses to visualize the differences in cryoprotectant profi les between treatments and to directly test for treatment effects.
In the AutumnWarm treatment, which simulates conditions prior to fi rst freezing, E. aethiops, E. medusa and E. pronoe did not differ in SCP, which were ≈ -16°C, despite signifi cant differences in CrPC, which were much lower in E. medusa and E. aethiops than in E. pronoe (Fig. 1).In the AutumnCold treatment, which simulates conditions before the onset of very low temperatures, SCPs were highly signifi cantly different.It was highest in the lowland E. medusa; followed by similar values in the submountain E. aethiops, subalpine E. pronoe and subnivean E. pluto, and lowest in the alpine E. cassioides.The pattern was only partly mirrored by the CrPCs, which were low for the two low altitude species (E.medusa, E. aethiops), followed by the subalpine E. pronoe and alpine E. cassioides, and highest in the subnivean E. pluto.In the WinterCold treatment, which simulates winter conditions, SCP was lower in the alpine E. cassioides than in lowland E. medusa, submountain E. aethiops and the subalpine E. pronoe.CrPC again showed a different pattern, with the lowest value recorded for the lowland E. medusa, followed by the submountain E. aethiops, and then by E. pronoe and E. cassioides.
The average values of SCP and CrPC were negatively correlated across species and treatments (N = 12), revealing that high CrPC is associated with a low temperature at which the body fl uids freeze, but this was not statistically signifi cant (Pearson's r = -0.428,P = 0.165).
The factorial ANOVA of SCP for the four species exposed to the two winter treatments (Table 1) pointed to a marginally signifi cant increase in SCP in the subalpine E. pronoe and alpine E. cassioides subject to the WinterCold treatment (cf.Fig. 1).For CrPC, the low CrPC recorded in submountain E. aethiops, and the already high autumn CrPC recorded in subalpine E. pronoe, substantially increased in winter (Fig. 1).The comparison for the three species subjected to all three treatments (Table 1), revealed the following changes among treatments: SCPs, which did not differ in the AutumnWarm treatment, diverged in the AutumnCold treatment (it increased in E. medusa, remained unchanged in E. aethiops, and decreased in subalpine E. pronoe) and further diverged in the WinterCold treatment (increased in E. pronoe).CrPC of E. aethiops and E. pronoe, but not E. medusa, signifi cantly increased from the AutumnCold to WinterCold treatment.

Cryoprotectant profi les
We detected 16 compounds with a putative cryoprotectant function (Table 2).Three of these (myo-inositol, erythritol and ribose) never reached a concentration of > 0.1 μg/ mg fresh mass.The three compounds with highest average concentrations, across all treatments, were: trehalose, glucose and saccharose in E. medusa; and trehalose, monostearin and monopalmitin in the remaining four species.
All three RDA analyses comparing cryoprotectant profi les among species within individual treatments revealed signifi cant patterns.In all three treatments, the fi rst ordination axes differentiated low altitude species (with E. medusa always the most extreme) from alpine/subnivean species (Fig. 2).The second ordination axes pointed to a difference between E. aethiops and the remaining species in AutumnWarm and WinterCold treatments, and between E. aethiops and E. pronoe relative to the remaining three species in the AutumnCold treatment.
The AutumnWarm treatment (3 species) revealed that lowland E. medusa contained high concentrations of glucose, but also trehalose and ribitol.The concentrations of fructose, ribose and saccharose were relatively high in E. aethiops.Arabinitol, maltose, monopalmitin, monostearin and sorbitol were characteristic of the subalpine E. pronoe (Fig. 2).In the AutumnCold treatment (5 species), the lowland E. medusa again had high concentrations of glucose, which was now accompanied by erythritol, ribitol, saccharose and threitol.E. aethiops had high concentrations of fructose and ribose, while E. pronoe, the alpine grasslands dweller, contained relatively high concentrations of maltose, plus monopalmitin and monostearin.The two species presumably adapted to the harshest conditions, E. pluto and E. cassioides, had the highest concentrations of compounds believed to be responsible for cold hardiness: glycerol and trehalose, plus other common cryoprotectants: arabinitol, mannitol and sorbitol (Fig. 2).The situation was similar in the WinterCold treatment (Fig. 2), except for the high concentration of glycerol (still accompanied by monopalmitin, monostearin, trehalose and threitol) in E. pronoe and E. cassioides.Sorbitol and trehalose were high in E. aethiops, while the lowland E. medusa contained high concentrations of arabinitol, erythritol, glucose, ribitol and sucrose.
Comparing the effects of treatments for each species separately pointed to differences in the polyol profi les recorded in the three treatments for E. medusa, E. aethiops and E. pronoe, but not in the two cold treatments for E. cassioides.The distinctions between WinterCold and the two autumn treatments were always greater (i.e., fi tted by the fi rst canonical axes) than those separating the two autumn treatments (Fig. 3).
In lowland E. medusa (Fig. 3A), a decrease in temperature (AutumnCold) was associated with an increase in the concentration of monopalmitin, monostearin, saccharose, threitol and trehalose.With continuing winter, there was not only a remarkable increase in erythritol, but also an increase in arabinitol, glucose, mannitol and sorbitol.In E. aethiops (Fig. 3B), maltose prevailed in the Autumn-Warm, followed by fructose in the AutumnCold treatment.Then, in the WinterCold treatment, there was big increases in multiple cryoprotective compounds, mainly arabinitol, erythritol, glycerol, ribitol and trehalose, which were similar to the increase in total CrPC concentration (see Table 1, Fig. 1).Similar changes occurred in E. pronoe (Fig. 3).Here, the main AutumnWarm compound was maltose, accompanied by fructose and sorbitol.The increase in CrPC in WinterCold was due to arabinitol, glucose, glycerol, mannitol, ribitol and trehalose.Finally, the analysis for E. cassioides indicated a non-signifi cant increase in all cryoprotectants between AutumnCold and WinterCold (Fig. 3D).

DISCUSSION
The fi ndings regarding the mean super cooling point did not support our initial expectation, derived from Vrba et al. (2012) that species experiencing less predictable snow cover (i.e., the lowland E. medusa and submountain E. aethiops) should be more cold hardy than the alpine and subnivean species (E.pronoe, E. cassioides and E. pluto).The only (weak) support for such a pattern was the lower SCP recorded for E. cassioides, a species associated with stony substrates, than in the alpine grasslands dwelling E. pronoe (cf.Kleckova et al., 2014).Our second prediction that the highest levels of cold hardiness would be recorded during winter was not the case for the two alpine species, E. cassioides and E. pronoe.Their high autumn super cooling ability could be due to a less predictable snow cover in autumn than later in winter.In the lowland E. medusa and submountain E. aethiops, no changes in cold hardiness were recorded between the AutumnCold and WinterCold treatments.Our third prediction, i.e. that a higher cryoprotectant content would be associated with a lower SCP, was also only partly supported; the two values were correlated negatively, but not statistically signifi cantly.Arguably, the patterns in cold hardiness were complicated by the two different cold hardiness strategies of the species studied and hence differences in polyol profi les among individual species.Individual cryoprotectants also tend to replace their precursors during the season, complicating their long term comparison.

Variation in cold hardiness and cryoprotectant content
Consistently with Vrba et al. (2012) we recorded the freeze tolerant cold hardiness strategy in lowland E. medusa and freeze-avoidant strategies in the three high-altitude species (E.pronoe, E. cassioides, E. pluto).A mixed situation was recorded in submountain E. aethiops, in which some of the larvae (those with high SCP) survived ice formation in their body fl uids.
Freeze-avoidance seems to be an evolutionarily older and more common strategy (Vernon & Vannier, 2002).The apparently less common freeze tolerance occurs either in habitats with prolonged periods of extremely low temperatures (Turnock & Fields, 2005) or in unpredictable climates with repeated freeze-thaw events (Sinclair et al., 2003).We speculated earlier (Vrba et al., 2012) that lowland E. medusa is more likely to experience freezing temperatures without snow cover and freeze-thaw cycles than the alpine species.The results recorded here document that freeze tolerance in E. medusa occurs independently of prewinter acclimation conditions.
A fi fth of the individuals of the submountain E. aethiops assayed for SCP survived the freezing of their body fl uids, Table 2. Putative cryoprotectants: mean concentrations (and their standard deviations) expressed in μg/mg fresh body mass of all the fi ve species tested and three treatments (AW -AutumnWarm, AC -AutumnCold, WC -WinterCold).Only compounds reaching a higher concentration than 0.1 μg/mg fresh body mass are listed..90Monostearin 0.37±0.16 0.51±0.14 0.43±0.19 2.87±0.59 3.43±1.233.75±1.60 12.31±3.82 13.81±2.33 14.55±3.14 9.33±4.15 12.09±3.72 12.60±4.63Monopalmitin 0.20±0.080.26±0.060.23±0.101.63±0.31 2.05±0.79 2.36±1.00 7.16±2.088.13±1.64 8.88±1.68 5.34±2.646.70±2.18 7.74±2.which indicates the existence of a mixed strategy.So far, evidence for alternating strategies within one species, or co-existence of strategies within a single population, are extremely scarce.A lepidopteran example of a mixed strategy might be Papilio zelicaon Lucas, 1858 (Lepidoptera: Papilionidae), in which winter-acclimated pupae partly survive and are partly killed by internal ice formation (Williams et al., 2014).Two beetle species, Dendroides canadensis Latreille, 1810 (Pyrochroidae) (Horwath & Duman, 1984) and Cucujus clavipes Fabricius, 1781 (Cucujidae) (Kukal & Duman, 1989) may adapt their cold hardiness strategies along a latitudinal gradient, or accord-ing to acclimation conditions (Sformo et al., 2010).In the case of E. aethiops, this species' range includes warm nonalpine habitats, such as piedmont forest steppes and lowland wooded savannas (Franco et al., 2006;Slamova et al., 2011), as well as mountain forests openings and grasslands (Sonderegger, 2005).In diverse areas such as the Alps, such contrasting habitats may well occur within an individual female's dispersal range (cf.Slamova et al., 2013) and the species hence could have evolved either mechanisms to adapt strategies according to external conditions or polymorphism within populations.The freeze tolerant strategy is less energy demanding and does not require emptying the gut, which combines the possibility of food intake and ability to survive unpredictable cold snaps at the same time.On the contrary, the predictable, cold subnivean conditions in montane habitats may favour an alternative freeze-avoiding strategy via investment in super cooling (cf.Sinclair et al., 2003).More detailed, family controlled studies of this phenomenon are needed.
The recorded SCP values were nearly identical for the three Erebia species subjected to the treatment mimicking pre-freezing in early autumn (AutumnWarm), but diverged in the treatment mimicking onset of autumn freezing (Au-tumnCold).Moreover, for two species earlier assayed (Vrba et al., 2012), E. cassioides and E. medusa, our present AutumnCold values differed dramatically from the previously reported values, being much lower for the alpine E. cassioides (-19.0°C vs. -8.4 ± 2.8°C in the earlier study) and much higher for the lowland E. medusa (-17.0 ± 2.3°C vs. -11.5°C in present study).
This difference is clearly due to the acclimation conditions and nicely mirrors the contrasting cold hardiness strategies.In Vrba et al. (2012) the larvae were simply acclimated to constant 5°C.Assaying the cold hardiness of animals acclimated to a constant 5°C is frequently used in cold hardiness studies, as it facilitates rapid inter-species comparisons (Denlinger et al., 1992).In natural settings, however, external conditions change during winter, causing the changes in SCP values recorded here and elsewhere (Vrba et al., 2017).
The AutumnCold treatment temperatures, to which the alpine and freeze-avoidant E. cassioides larvae were exposed, likely crossed a threshold for a decrease in SCP (Vesala et al., 2012).The larvae dramatically increased their cryoprotectant concentration and decreased their SCP.Due to limited material, we were not able to assay the SCP of E. cassioides in the AutumnWarm treatment, which might be particularly informative.
In E. medusa, it is likely that the exposure to freezing temperatures initiated mechanisms that allow controlled freezing at relatively high subzero temperatures (Turnock & Fields, 2005).Although repeated freezing and thawing increases risks of tissue damage (cf.Marshall & Sinclair, 2011), freezing at higher temperatures decreases energy costs due to repeated synthesis of cryoprotectants (Voituron et al., 2002).Increase in SCP occurred also in the subalpine E. pronoe, in this case between AutumnCold and WinterCold treatments, which differed in the length of acclimation.
All these differences in SCP recorded under slightly different conditions imply that standardized cold hardiness measuring procedures can convey some information about the physiological capacities of the species studied, but are of limited value for understanding an insects' ability to survive extremely cold conditions.
Cryoprotectant concentrations mirrored SCP only to a limited extent.Alpine species always had a higher CrPC than lowland species.The two high altitude species, E. pronoe and E. cassioides, had high CrPCs already during the warm autumn treatment.In this respect, they resembled the boreomontane Colias palaeno (Linnaeus, 1761) (Lepidoptera: Pieridae), a species that has a high CrPC in autumn, but can decrease its cryoprotectant content in mild winters (Vrba et al., 2017).In contrast, the submountain E. aethiops doubled its cryoprotectant content between AutumnCold and WinterCold treatments, probably in association with the plastic reactions to external cues in this mixed-strategy species.Correlating individual cryprotectant content with SCPs provided some indication of why total cryoprotectant content did not mirror cold hardiness.Compounds such as glucose or sucrose, used as precursors in the synthesis of other polyols (Kostal et al., 2004), reached high concentrations under conditions resulting in a high SCP (i.e., in Au-tumWarm treatment).Also, the presence of two alternative cold hardiness strategies in the species studied precluded a straightforward association between total cryoprotectant content and supercooling point, because freeze tolerant and freeze avoidant species use cryoprotectants in different ways (Turnock & Fields, 2005).On the other hand, we showed earlier that individual populations of the boreomontane C. palaeno may dramatically differ in cryoprotectant profi les but not in their cold hardiness (Vrba et al., 2014a).The role of specifi c cryoprotectants on the level of cold hardiness may differ across species and climatic conditions, as can be demonstrated, e.g., by the different role of glycerol content on cold hardiness in several species of Lepidoptera (Andreadis et al., 2008;Hou et al., 2009;Trudeau et al., 2010).In addition, the synthesis of cryoprotectants is only a part of a more complex mechanism of increasing cold hardiness.Other important components are the synthesis of antifreeze proteins and active removal of ice-nucleating agents from the body (Somme, 1982;Duman, 2001).Clearly, surveying changes in cryoprotectant profi les may also provide more precise insights into the overwintering of alpine Lepidoptera.

Cryoprotectant profi les
In all three treatments, the analyses of differences in cryoprotectant profi les among species revealed a lowland-alpine gradient, partly supporting our prediction that species inhabiting similar conditions should employ similar cryprotective compounds.The lowland and freeze-tolerant E. medusa differed from the high altitude and freeze-avoiding E. cassioides, E. pluto and E. pronoe.
All the species studied had high contents of trehalose, a compound commonly found in insects, while the glycerol content, frequently cited as major cryoprotectant in Lepidoptera (Cha & Lee, 2016) , including the boreomontane butterfl y Colias palaeno (Vrba et al., 2017), was surprisingly low.The four high altitude species contained high volumes of monopalmitin and monostrearin, two compounds rarely associated with cryoprotective functions (Vrba et al., 2017), while the lowland E. medusa contained metabolically active precursors (glucose, sucrose) and rarer compounds, such as ribitol, erythritol and threitol.A different situation occurred in E. aethiops, which contained precursors or constitutive cryoprotectants different from both E. medusa and the alpine species, such as a relatively high concentration of fructose in both autumn treatments.
Autumn and winter samples were differentiated along the main ordination axis, whereas the vertical axis, indicating secondary gradients in cryoprotectant variation, corresponded to differences within autumn.For the three species with signifi cant results, the patterns corroborated the differences between lowland and mountain/alpine species outlined above.Thus, E. medusa synthesized trehalose already in autumn, similar to species in which cold hardiness is achieved before the onset of freezing [e.g. C. palaeno in Vrba et al. (2017)].In contrast, E. aethiops and E. pronoe increased their cryoprotectants only after exposure to freezing.
The contents of cryoprotectants in overwintering Erebia larvae thus vary considerably both among species and during the overwintering period.Putatively, this diversity seems to be linked both with the diversity of cold hardiness strategies (tolerant, mixed, avoidant), which in turn probably allows Erebia butterfl ies to exist in a high diversity of climatic conditions, from lowlands to subnivean zones and from wetlands to semiarid conditions.Notably, Shimada (1988) and Williams et al. (2014) also demonstrate that a high diversity of cryoprotective strategies is linked with diversity of cryoprotectant profi les in another widely distributed butterfl y genus, Papilio Linnaeus, 1758.

CONCLUSION
Our study of fi ve ecologically different species of Erebia revealed a high level of diversity in the physiology of cold hardiness within this butterfl y genus, including different profi les of cryoprotectants, which may or may not provide similar levels of cold hardiness in the different species.It is tempting to speculate that switching among biosynthetic pathways, leading to synthesis of species-specifi c cryoprotectant mixtures, allowed Erebia to adapt to a diversity of extreme conditions.Such an inference, however, requires a better understanding of inter-specifi c cryoprotectants and diversity of cold hardiness in other insect taxa.Such knowledge is currently fragmentary, and the modest sampling of taxa used in this study is still much wider than that used in studies on other lepidopteran genera [cf.Pullin & Bale (1989a): two Nymphalidae species; Kukal et al. (1991): four species of Papilio; Vrba et al. (2017): two species of Colias)].It is now necessary to undertake a wider and more systematic sampling of taxa in order to distinguish whether the diversity of strategies and cryoprotectant profi les found in Erebia is particular to this genus, or perhaps Satyrinae butterfl ies in general, or a general characteristic of groups adapted to climatically extreme environments.