Starvation hardiness as preadaptation for life in subterranean habitats

Most subterranean habitats, especially caves, are considered extreme environments, mainly because of the limited and erratic food supply and constant darkness. In temperate regions, many climatic conditions, such as temperature and air humidity, are periodically less adverse or even more favourable in caves than the harsh seasonal weather on the surface. Accordingly, many animal species search for hibernacula in caves. These overwintering, non-specialized subterranean species (non-troglobionts) show various modes of dormancy and ongoing development. Since they do not feed, they all undergo periodic starvation, a preadaptation, which might evolve in permanent starvation hardiness, such as found in most specialized subterranean species (troglobionts). To this end, we performed a comparative analysis of energy-supplying compounds in eleven most common terrestrial non-troglobiont species during winter in central European caves. We found highly heterogeneous responses to starvation, which are rather consistent with the degree of energetic adaptation to the habitat than to overwintering mode. The consumption of energy-supplying compounds was strongly higher taxa-dependant; glycogen is the main energy store in gastropods, lipids in insects, and arachnids rely on both reserve compounds. We assume that permanent starvation hardiness in specialized subterranean species might evolved in many different ways as shown in this study.

Laboratory analyses. Specimens were measured for their size, fresh and dry mass, and lipid, glycogen, and water content at the beginning (November), in the middle (January) and at the end of overwintering (March) after sacrifice by exposure to -20 °C for two hours. Numbers of individuals studied are given in Table 2. The dry mass was determined after 15 days of vacuum desiccation under P 2 O 5 . Lipids were extracted from completely macerated, dry individual samples via diethyl ether extraction 60 by addition of 2 mL diethyl ether, stirring for 5 min, decanting ether and being weighed after 24 h desiccation of air-dried sample under P 2 O 5 . Larger samples with more fats were eluated twice. Glycogen was quantified in the same samples after lipid extraction with the anthrone reaction 61 , by putting a sample in 2 mL of KOH (300 g L -1 ), and boiled for 20 min while shaking. After cooling, the undissolved particles were sedimented by centrifugation (1000 × g, 5 min). The supernatant Statistical analysis. Due to relatively small sample sizes for some species, data for males and females were pooled for statistical analyses. The data distribution for fresh and dry mass, amount of energy-supplying compounds, and water content for all species for the three overwintering periods was first checked for normality using the Shapiro-Wilk test. As some data were non-normally distributed, we summarized the data by means, standard deviations (StD), medians and interquartile ranges (IQR). We examined differences in energy-supplying compounds and water content among species, among higher taxa, and among guilds. For this purpose, we used the one-way ANOVA or Welch's ANOVA (for data with unequal variance) for normally distributed data, and the Kruskal-Wallis test for non-normally distributed data. Where statistically significant differences were found, Tukey HSD post hoc test (for normally distributed data) and the Bonferroni-adjusted Mann-Whitney U test (for non-normally distributed data) were applied. To visualize lipid, glycogen and water content of individual species, higher taxa and guilds, we provided PCA biplots with polygons obtained by the Convex-Hull method. For this purpose, data for individual species, higher taxa and guilds across the three overwintering periods were merged. All tests and PCA biplots were performed in PAST 4.03 program 62 , while dot plots were created with the ggplot2 R package 63 using the ggplot command.
To summarize energy demands during overwintering in species under study, we calculated average per-day consumption and accumulation of the main energy-supplying reserve compounds, respectively. In the calculations, the value 39 kJ g -1 was used for reserve lipids and 17 kJ g -1 for glycogen 61 .

Results
Dynamics of energy-supplying compounds and water content in the studied species during overwintering. Descriptive statistics for fresh and dry mass, and lipid, glycogen, and water content during overwintering for the eleven species is presented in Table 2. In the following text, only statistically significant differences in lipid, glycogen and water content, obtained by various statistical tests, are reported; the results of the post-hoc tests are presented in Supplementary Table S1.
In C. (O.) lefeburiana, we observed statistically significant differences in glycogen (F 9.76 = 4.39, p = 0.044) and water content (F 2 = 10.72, p = 0.001) in three overwintering periods. Both were higher at the end than at the beginning and in the middle of overwintering (Fig. 1a).
In C. illyrica, significant differences were found in lipid (F 2 = 3.78, p = 0.044) and water (χ 2 = 15.65, p < 0.001) content during overwintering. Lipid content was significantly higher in the middle compared to the end of www.nature.com/scientificreports/ www.nature.com/scientificreports/ overwintering. Water content was significantly lower in the middle and at the end, compared to the beginning of overwintering, and also lower at the end, compared to the middle of overwintering (Fig. 1b).
In A. aurantiacus, lipid content decreased gradually and differed significantly among the periods (F 2 = 6.95, p = 0.003). At the beginning, it was higher than in the middle and at the end of overwintering. Significant differences were also found in glycogen content by period (χ 2 = 20.56, p < 0.001). More glycogen was present at the beginning and end than in the middle of overwintering. Water content increased over the course of overwintering and differed significantly among the periods (F 21.5 = 351.2, p < 0.001); amounts were significantly higher in the middle and at the end as compared to the beginning of overwintering (Fig. 2a).
Glycogen content in G. annulatus gradually diminished during overwintering. We found significant differences among the periods (F 2 = 41.99, p < 0.001). A post-hoc test revealed lower glycogen content in the middle and at the end than at the beginning of overwintering, and lower at the end than in the middle of overwintering (Fig. 2b).
In G. titanus, we found significant differences in glycogen (F 2 = 16.24, p < 0.001) and water (F 2 = 9.42, p = 0.001) content during overwintering. Glycogen content was significantly lower at the end compared with the beginning and the middle of overwintering, while water content was significantly higher at the end than at the beginning of overwintering (Fig. 2c).
In M. menardi, we found significant differences in lipid content among the periods (χ 2 = 11.91, p = 0.003). Lipid content was lower in the middle than at the beginning of overwintering (Fig. 2d).
In L. schreibersii, we observed significant differences in glycogen content during overwintering (F 18.3 = 15.40, p < 0.001). Glycogen content was higher at the end than at the beginning and in the middle of overwintering (Fig. 3a).
In T. dubitata, there were significant differences in glycogen content among overwintering periods (χ 2 = 18.49, p < 0.001). Glycogen content was higher at the beginning and in the middle than at the end of overwintering (Fig. 3b).
In S. libatrix, we found statistically significant differences in lipid content (χ 2 = 18.03, p < 0.001) among the time periods. Lipid content was significantly higher at the beginning than in the middle and at the end of overwintering (Fig. 3c).
In T. cavicola, there were statistically significant differences in lipid content among the periods (F 2 = 5.02, p = 0.014). Lipid content was significantly higher at the beginning and in the middle than at the end of overwintering. We also found significant differences in water content among the periods (F 2 = 71.84, p < 0.001). The water content was higher at the end than at the beginning and in the middle of overwintering, and higher in the middle than at the beginning of overwintering (Fig. 3d).
In T. neglectus, lipid content differed significantly between periods (F 2 = 12.41, p < 0.001). Lipid content was higher at the beginning than in the middle and at the end of overwintering (Fig. 3e).   www.nature.com/scientificreports/ Dynamics of energy-supplying compounds and water content in higher taxa and guilds during overwintering. Figures 5 and 6 show the general trends in lipid, glycogen, and water content during overwintering for three higher taxa and three guilds, respectively. The results of the post-hoc tests are presented in Suplementary Table S2.
In molluscs (Fig. 5a), we found significant differences in water content among the periods (χ 2 = 8.73, p = 0.013); there was more water at the beginning than in the middle and at the end of overwintering.
In arachnids (Fig. 5b), lipid and glycogen content gradually decreased, while the water content gradually increased during overwintering, and values significantly differed by period (lipids: χ 2 = 15.78, p < 0.001; glycogen: χ 2 = 6.68, p = 0.035; water: χ 2 = 18.06, p < 0.001). Lipid content was higher at the beginning than in the middle and at the end of overwintering, and glycogen content was higher at the beginning that at the end. Water content was higher in the middle and at the end than at the beginning of overwintering.  www.nature.com/scientificreports/ In insects (Fig. 5c), there were significant differences in lipid (χ 2 = 14.70, p = 0.001) and water (χ 2 = 7.73, p = 0.021) content during overwintering. Lipid content was higher at the beginning than in the middle and at the end of overwintering, while water content was higher at the end than in the middle of overwintering.  www.nature.com/scientificreports/ In Guild 1 (Fig. 6a), lipid (χ 2 = 10.03, p = 0.007) and glycogen (χ 2 = 7.14, p = 0.028) content differed significantly among overwintering periods. Lipid content was higher at the beginning of overwintering than in both following periods, and glycogen content at the beginning than at the end of overwintering.
In Guild 2 (Fig. 6b), there were significant differences in lipid (χ 2 = 12.78, p = 0.002) and water (χ 2 = 15.67, p < 0.001) content during overwintering. Lipid content was higher at the beginning and in the middle than at the end, and water content was higher in the middle and at the end than at the beginning of overwintering.
In Guild 3 (Fig. 6c), lipid (χ 2 = 7.32, p = 0.026) and glycogen (χ 2 = 6.28, p = 0.043) content differed significantly between periods. Lipid content was higher at the beginning than in the middle of overwintering, and glycogen content was higher in the middle and at the end than at the beginning of overwintering.
Total amount of energy-supplying compounds and water content in higher taxa and guilds. Ordination diagrams in Fig. 7, based on PCA, provide general information on lipid, glycogen and water content for species, higher taxa and guilds during overwintering. The results of the post-hoc tests are presented in Tables 3 and 4.
In the two mollusc species, there was considerable overlap in lipid, glycogen, and water content levels during overwintering, and consequently, there were no statistically significant differences (Fig. 7a).
In insects, we also observed significant differences in lipid (χ 2 = 25.68, p < 0.001), glycogen (χ 2 = 12.38, p = 0.015) and water (χ 2 = 121.40, p < 0.001) content (Fig. 7c). In Troglophilus cavicola, T. neglectus, L. schreibersii and S. libatrix lipid contents were significantly higher than in T. dubitata. Glycogen content in S. libatrix was significantly higher than in L. schreibersii. Water content in T. cavicola was significantly higher than in L. schreibersii, T. dubitata and S. libatrix. Moreover, water content in T. neglectus was significantly higher than in L. schreibersii, T. dubitata and S. libatrix. It was also higher in L. schreibersii than in T. dubitata and S. libatrix, and higher in T. dubitata than in S. libatrix. In testing for differences between troglophilic L. schreibersii and the other insect species under study, it was found that lipid content was significantly higher (U = 2184.00, p = 0.049), while glycogen content was significantly lower in L. schreibersii (U = 2009.00, p = 0.010) than in the other insects.
In Guild 1, there were significant differences in lipid (U = 121.00, p < 0.001), glycogen (U = 114.40, p < 0.001) and water (U = 148.20, p < 0.001) content among species (Fig. 7d). Lipid and glycogen contents were higher in G. annulatus, G. titanus, T. neglectus and S. libatrix than in C. (O.) lefeburiana and C. illyrica. In addition, lipid content was also higher in G. titanus, T. neglectus and S. libatrix than in G. annulatus. We found higher glycogen content in G. titanus than in G. annulatus, and in T. neglectus and S. libatrix than in G. titanus. Water content in C. (O.) lefeburiana and C. illyrica was higher than in the other species of this guild. It was also higher in G. annulatus than in G. titanus, T. neglectus and S. libatrix, and higher in G. titanus and T. neglectus than in S. libatrix. www.nature.com/scientificreports/ In Guild 2, lipid (χ 2 = 10.67, p = 0.005), glycogen (χ 2 = 33.57, p < 0.001) and water (χ 2 = 55.72, p < 0.001) contents differed significantly among species (Fig. 7e). Lipid and water content were higher in A. aurantiacus and T. dubitata than in T. cavicola, and glycogen content was higher in A. aurantiacus than in the other two species.
Among Mollusca, Arachnida and Insecta, we also found significant differences in lipid (χ 2 = 123.40, p < 0.001), glycogen (χ 2 = 95.53, p < 0.001) and water (χ 2 = 134.50, p < 0.001) content (Fig. 7g). In insects, lipid content was significantly higher than in mollusks and arachnids, and in arachnids, lipid content was higher than in molluscs. On the other hand, in molluscs, glycogen content was higher than in arachnids and insects. Moreover, in insects, glycogen content was higher than in arachnids. In molluscs, water content was higher than in arachnids and insects. In contrast to glycogen content, in arachnids, water content was higher than in insects.
Among the three guilds, significant differences were observed in glycogen (χ 2 = 68.81, p < 0.001) and water (χ 2 = 31.94, p < 0.001) content (Fig. 7e). In species assigned to Guild 1 and Guild 2, glycogen content was higher than those of Guild 3. In addition, in species of Guild 1, water content was higher than in the species of the other two guilds.  www.nature.com/scientificreports/ Transmission electron microscopy. Figure 8 illustrates intracellular conditions in selected arachnids and insects. Presentation of gastropods is missing because of damaged slices, useless for the TEM analysis (see explanation in Methods). In most species, the amount of reserve compounds diminished during winter. However, TEM reveals variation in the pattern of diminishing. In the harvestman A. aurantiacus (Fig. 8a), at the beginning of overwintering, the cytoplasm of adipocytes in the epithelial cells of the midgut diverticula was crowded with lipid droplets and glycogen granules. Consequently, the lipid droplets were in close contact, and many neighbouring droplets fused. Between the droplets, the cytoplasm was locally electron-dense because of numerous glycogen granules and protein granules. In the middle and at the end of overwintering, the diameter of lipid droplets decreased slightly.
In M. menardi (Fig. 8b), at the beginning of overwintering, the cytoplasm of adipocytes in the epithelium of the midgut diverticula was filled with large to huge lipid droplets and numerous glycogen granules. In the middle and at the end of overwintering, the size and abundance of both energy-storing compounds was reduced, and autophagic structures appeared.
In S. libatrix (Fig. 8c), at the beginning of overwintering, the adipocytes in the fat body were characterized by numerous huge lipid droplets, protein granules and glycogen granules. In the middle and at the end of overwintering, the size of the lipid droplets and protein granules and the abundance of the glycogen granules diminished.  www.nature.com/scientificreports/

Discussion
In temperate zone in winter, many climatic conditions in caves, such as temperature and air humidity, are more favourable as compared to conditions on the surface. Consequently, many epigean animals take advantage of such conditions to find hibernacula in caves, all showing relatively high level of starvation hardiness in dormancy.
As originally predicted, the responses of nine inactive and two active terrestrial species included in this study showed large differences in stores of energy-supplying compound and energy consumption during overwintering www.nature.com/scientificreports/ starvation. In addition, overwintering strategies were even more heterogeneous than predicted. Here we discuss the main findings.
Dynamics of energy-supplying compounds and water content in the studied species during overwintering. Considering the fact that the eleven species studied belong to different higher taxa, we confirmed conspicuous differences in the response to starvation and in the consumption of reserve energy-supplying compounds during overwintering. The consumption of these compounds is strongly higher taxa-dependant; in the gastropods C. (O.) lefeburiana and C. illyrica, the energy-supplying stores consist almost exclusively of glycogen 64,65 , in insects lipids strongly prevail, and in arachnids the use of both types of reserve compounds seems to be balanced. Enhanced glycogen metabolism in both invertebrates and in vertebrates allows faster and more vigorous reaction to adverse stimuli than lipid metabolism does 49,66 . Gyas annulatus, e.g., which overwinter closely above a snow-water stream, accumulates more glycogen that the congeneric G. titanus, which overwinter in the same cave in unthreatened microhabitats; glycogen metabolism enables individuals of G. annulatus to flee quickly away from sudden water level rise 49 . Cave-dwelling individuals of the Mexican cavefish Astyanax mexicanus with enhanced glycogen metabolism show the same strategy in energy metabolism: they swim a few-times slower, but reach maximum swimming speed a few times faster than their predominantly lipid-fuelled riverine Table 3. Pairwise post-hoc comparisons for differences in lipid, glycogen and water content among species within higher taxa, and among higher taxa; statistically significant differences are in bold. www.nature.com/scientificreports/ counterparts 66 . Thus, enhanced glycogen metabolism might have an adaptive role for quick reaction to threating events in subterranean habitats. Arachnids and insects can store a great amount of lipid reserves in the form of cytoplasmic lipid droplets in adipocytes 16,24 . Lipids are the most economical supplies for long-term starvation, and are therefore primarily consumed in overwintering and exclusively subterranean residents 17,24 . The highest consumption in S. libatrix and T. neglectus indicates that metabolic depression 67 is limited.
The apparently surprising accumulation of energy-supplying compounds in L. schreibersii during winter perfectly illustrates the opportunistic feeding habit of species adapted to conditions in caves, generally considered extreme habitat 11,34 , which mostly refers to a limited and erratic food supply. Laemostenus schreibersii feed continuously during first half of winter, e.g., on guanobionts, until the adverse conditions dictate cessation. Thus, this beetle optimally exploit food when available. Then, they probably move from the cave into the adjacent fissure network and remain active 68 , or, alternatively, they enter torpor-a postponed dormancy-as they do in laboratory below 6-7 °C (own unpublished data), which might correspond to quiescence.
Because of glycolysis, water content decreases during winter in glycogen-fuelled dormant organisms, which is well evident in our two gastropod species from the beginning to the end of overwintering. The increase in water content in C. (O.) lefeburiana must be considered an artefact, since these snails apparently consumed water before being sampled at the end of overwintering. In contrast, lipolysis produces metabolic water, which contributes to a higher water content 55 as found in our dormant arachnids and insects. A net loss of body water in glycogen-fuelled organisms, and its net gain in lipid-fuelled organisms well comply with the findings on the consuption of energy-supplying reserves and vice versa. Thus, trends in water content are of supplementary importance for the interpretation of the main energy-fuelling processes in these organisms. All these findings must be taken into account when considering species of different higher taxa.
Energy consumption in the studied species during overwintering. Species in diapause (Guild 1), i.e., C. (O.) lefeburiana, C. illyrica, G. annulatus, G. titanus, S. libatrix and T. neglectus, that do not move and Table 4. Pairwise post-hoc comparisons for differences in lipid, glycogen and water content among species within guilds, and among guilds; statistically significant differences are in bold italics. www.nature.com/scientificreports/ develop during overwintering, were consequently expected to have low consumption of energy. However, this was not the case for. T. neglectus 25 and S. libatrix 20 . The high energy consumption is most likely due to poorly suppressed basal metabolic rate during overwintering, which could indicate poor energetic adaptation to the habitat. All other species in the guild show very cautious spending of energy for the basal metabolic rate.
On the other hand, species with ongoing ontogenesis under fasting conditions (Guild 2), i.e., A. aurantiacus 44,54 , T. dubitata 53 and T. cavicola 55,56 , were accordingly expected to have high energy consumption. Amilenus aurantiacus and T. dubitata show low energy consumption, although maturating during overwintering. This might indicate a better energetic adaptation of these species to the habitat. However, high energy consumption in T. cavicola was expected because of maturation process.
Moderate consumption was expected for the two, year-round active, opportunistic feeding species (Guild 3), i.e., M. menardi and L. schreibersii. Meta menardi showed a relatively low energy consumption, because they occasionally feed in winter 45,50,51 and in this way replenish the energy reserves that were consumed while food was not available. Thus, M. menardi undergo a compulsory overwintering starvation at times of scarce opportunity to feed 56 , which might partly correspond to hypometabolism as a mode of dormancy 67 . In accordance with extending feeding in winter, L. schreibersii accumulated about 1 J/g energy per day.
Lipids were found to be the major energy-supplying compounds 24 , whereas glycogen metabolism should be studied in more detail, as the utilization of triacylglycerides may be associated with glycogen resynthesis 40 for either glucose release or moulding processes 18 . Our results suggest that energetic strategies evolve independently; each overwintering species deserves separate consideration with this respect. Accordingly, we did not confirm our hypothesis on co-evolution of functional and energetic traits. www.nature.com/scientificreports/ Cytological findings. A cytological approach allowed visual comparison of the size and relative abundance of energy-supplying compounds in selected tissues or organs. Similar changes were observed in all three species assessed by TEM, which we attribute to the consumption of energy-supplying compounds during overwintering. Lipid droplets in the cells decreased in size and abundance, and glycogen and protein granules decreased in abundance; any variation in the glycogen granules cannot be assessed by TEM micrographs. In addition, autophagic structures appeared in M. menardi in the middle and at the end of overwintering, providing an additional source of energy in this and some other overwintering species, e.g. 44,45 . The cytological evidence of reserve compounds was in complete agreement with the biochemical findings, and was, thus, of supporting importance.
Artefact outcomes. The sampling approach used in this study may lead to some artifactual results, which could be particularly evident when considering individual species on a fine scale. Inherent variability within the species may be a consequence of the specific life history of the individuals collected, such as differences in feeding opportunities, stage and timing of reproduction, food availability and microclimatic conditions in overwintering sites. Some other data suggest inappropriate timing of sampling, as in the case of C. (O.) lefeburiana, which obviously began to feed before sampling, given the high water content at the end of overwintering. On the other hand, in many species, e.g., in T. cavicola, the increase in glycogen rate alone at the end of overwintering indicates changes in metabolic activity before entering anew epigean, active ecophase in the life cycle. Despite such cases, the general trends considered here remain well established, as confirmed in experimentally designed studies 21,50 . Differences in the measuring methods of lipid and glycogen amounts, e.g. 25,[69][70][71] , negligibly affect the large-scale comparative results.

Concluding remarks
With the exception of L. schreibersii, which abundantly feed in caves in winter, nine other studied non-troglobiont species undergo a programmed periodical starvation, and M. menardi a compulsory starvation. However, starvation hardiness is the only collective trait in these ten species, an obligate, preadaptive trait for life in the subterranean environments. This preadaptation might evolve into a permanent adaptation, as experienced in troglobionts, by economizing the use of energy-supplying compounds and energy. Another two important steps in adaptation to permanent life in the habitat are the start of feeding and completion of the reproduction in the habitat. The first applies to M. menardi, and the second might apply to L. schreibersii. While studies on M. menardi well compile, very little is known about the biology and ecology of L. schreibersii and many other species studied. It is a great challenge to fulfill these knowledge gaps in the future in these and in other non-troglobiont species, e.g. [72][73][74][75] .

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/