In total, 69 specimens of Speolepta leptogaster from 37 different caves, bunkers, wells, tunnels and cellars in Hesse (maximum of three per site) and one from a bunker in Poland were analyzed. The samples (Table 1) were collected and identified by members of the Hesse Federation for Cave and Karst Research (Germany) (Reiss et al. 2009) and stored in ethanol. In order to ensure good DNA preservation, all molecularly-processed samples were dated no later than six years old. Prior to DNA isolation, images of all specimens were taken with the camera “Moticam Model Moticam 5“. Specimens were measured using the measuring function of the image-capturing program “Motic Images Plus 2.0”. The bodies of the specimens were examined to determine potential differences in phenotype, which could correspond to different haplotypes.

A small piece of the posterior end of the body (approx. one sixth of the total animal) was used for larval samples. In the case of pupae and imagines, a larger portion of the abdomen (approx. ¼) was macerated. DNA isolation was performed according to the instructions of the DNEasy Blood & Tissue Kit (Qiagen Sample and Assay Technologies, Hilden, Deutschland) for the column purification of animal tissue.

A 662 bp fragment of the Cytochrome C Oxidase subunit 1 (COI) gene was amplified using the primers C1-J-2195 5’-TTGATTTTTTGGTCACCCTGAAGT-3’ and TL2-N-3014 5’-TCCAATGCACTAATCTGCCATATTA-3’ established by Simon et al. (1994). The PCR reactions were executed in a Peqlab Primus Advanced 96 thermo cycler. Each PCR-mix (25 µL) contained 2.5 µL PCR-buffer (10×, without MgCl₂), 2.0 µL MgCl₂ (100 mM), 0.3 µL dNTPs (20 mM), 1.0 µL of each primer, 0.3 µL Taq-DNA-Polymerase, 1.5 µL BSA (bovine serum albumin, 10 mg/mL), 11.4 µL ddH₂O and 5.0 µL template DNA. PCR conditions were as follows: initial denaturation step 1 min 94 °C, 30 cycles of denaturation (30 sec, 94 °C), annealing (30 sec, 43.7 °C) and elongation (30 sec, 72 °C) and a final elongation step with 7 min at 72 °C. PCR products were bi-directionally sequenced using the Sanger chain termination method (Sanger et al. 1977). Sequencing service was provided by the Laboratory Centre of the Biodiversity and Climate Research Centre (Frankfurt am Main, Germany) with their own sequencing protocols. Editing and assembly of the forward and reverse sequences were done using Geneious v5.4.6. The sequences were manually trimmed to a length of 659 bp to fit the shortest sequence retained. Sequence alignment was performed using the MAFFT v6.814b plug-in for Geneious under default settings and the automatic algorithm option.

The software DnaSP v5 (Librado and Rozas 2009) was used to calculate the number of haplotypes, the total haplotype and nucleotide diversity and to perform neutrality tests. Haplotype diversity (Hd) can result in values between 0 and 1. A value of Hd = 1 implies that two randomly picked samples will always demonstrate two different haplotypes (= 100% diversity). The nucleotide diversity (π) illustrates the average genetic distance between two sequences estimated as an average for all sequences. The value can range from 0 (no changes in all sequences) to a theoretical 1 (every base is replaced). The calculation for Fu’s Fs (Fu 1997) and Tajima’s D (Tajima 1989) tests can be positive, negative or have a zero value. A significant positive value for Tajima’s D may point to over dominant selection or a population bottleneck event. Significant negative values point to a population expansion event or purifying selection. A value of zero or non-significant values cannot reject the neutral model of molecular evolution. Fu’s Fs shares the same characteristics but adds the support for genetic hitchhiking with a negative value. Standard settings were used and the sequences were trimmed to the same length. The program TCS 1.21 (Clement et al. 2000) was used to reconstruct a haplotype network by Statistical Parsimony (Templeton et al. 1992). Networks were created by a 95% connection probability.

A Mantel-test (Mantel 1967) was used to test for correlation between geographic distance vs. genetic distance. The geographic distance matrix was calculated with the Geographic Distance Matrix Generator 1.2.3 (Ersts 2014) using a WGS84 spheroid and geodesic distances in km. Genetic distances were calculated as p-distances using the software MEGA 5.2 (Tamura et al. 2011) and the pairwise deletion option. The Mantel-test was performed in XLSTAT 2013 (Addinsoft) using the linear Pearson product-moment correlation coefficient (r) and 10, 000 permutations. A value of r = 0 implies no linear correlation, whereas a minimal / maximal value of -1 / +1 indicates total negative / positive correlation.

No gaps were present in the COI-alignment. In total, 14 haplotypes (H1-H14) were revealed. Haplotype diversity was high with a value of Hd = 0.7017, whereas nucleotide diversity was low with π = 0.00165 (Table 2). The haplotype network (Figure 1) indicated a highly frequent haplotype, H1, which consisted of 39 samples (57%) and a second relatively frequent haplotype, H2, which comprised 11 samples (16%). Eight haplotypes were singletons (H3, H4, H9-H14), two consisted of two (H6 and H7) and two of three samples (H5 and H8). There were only four haplotypes (H4, H8, H9 and H11) which were not directly connected to the most frequent haplotype H1.

When the different haplotypes are placed in a geographical context (Figure 2), H1 can be revealed throughout the study region of Hesse. Haplotype 2 demonstrates a similar distribution except in the central underground sites and the most northern parts of Hesse. Haplotype 8 occurs once in southern Hesse and in the sampling site in Poland. All other haplotypes have been found within only a single locality. At a few localities several haplotypes co-occurred: In north-western Hesse H14+H1 and H2+H1; in north-eastern Hesse H9+H1, H2+H1 and H13+H1 and in south-eastern Hesse, H10+H6+H1.

The results of both neutrality tests were significantly negative (p < 0.01) and point to (a) past population expansion event(s) (Table 2). In our dataset, 28.5% of all mutations (4/14) were non-synonymous leading to a change of the respective amino acid.

The Mantel-test (r = 0.17, p < 0.0001) reveals a significant positive low correlation between genetic distance and geographic distance (Table 2).

The region of the Central German Uplands (including our target area of Hesse) was covered with permafrost during the Last Glacial Maximum (18, 000 – 24, 500 years ago) (Clark et al. 2009). It is widely accepted, that ecologically diverse species were able to (re-)colonize this area when the climatic conditions became more suitable after the ice shields had retracted northward again (Taberlet et al. 1998, Hewitt 1999, Petit et al. 2002). Our genetic analyses are in agreement with a similar postglacial scenario for Speolepta leptogaster. The recent demographic structure of this species in Hesse is best explained by (a) past population expansion event(s). Most probably and after population bottlenecks, the low-frequency and most often cave endemic haplotypes have evolved from the high-frequency central haplotypes already present in the ancestral gene pool of the species (i.e. the founder effect). The significant but weak pattern of isolation-by-distance can be interpreted in terms of a certain level of connectivity between populations, which may be explained by dispersal of adult specimens. An interesting finding is the identity of haplotypes found in Poland and southern Hesse. Yet, a pan-European sampling is needed to address any further hypotheses (e.g. passive anthropogenic transportation vs. active surface-dispersal).

In general, our results imply good dispersal ability for Speolepta leptogaster. This is further supported by surface records of adult specimens compiled from the literature (Table 3). Although all developmental stages can be found in caves throughout the year, there are more findings of surface imagines in summer than in winter. Whether this is due to the biology of Speolepta leptogaster or to the smaller number of traps being laid out in winter cannot be determined.

Animals associated with underground habitats can be classified into different categories. Those classifications (i.e. ecological, behavioral, morphological) have been under constant change and revision. Multiple authors created new categories as well as split up old ones (Shiner 1854, Racovitza 1907, Vandel 1965, Hamilton-Smith 1971, Sket 2008). However, the main ecological categories to which we refer here have remained more or less unchanged: eutrogloxenes (normally not living in caves), subtroglophiles (partially living in caves, but without permanent populations), eutroglophiles (able to complete several underground generations) and eutroglobionts (solely living in caves). As a result of this ecological continuum of cave-association, i.e. from “only found by chance” to “being obligate subterranean”, probable cave-associated (e.g. morphological and behavioral) adaptations are multifarious.

Since the biological diversity of subterranean species and ecology of subterranean habitats is high (Moseley 2008), every categorical and thus limited ecological classification system will elicit problems in classifying all cave organisms. This is particularly challenged by a “species from both worlds” living in a transitional environment or ecotone, as which caves must be considered (Moseley 2008). The situation is even more complicated by the fact that Speolepta leptogaster belongs to the holometabolic insects with different developmental stages demonstrating different ecological requirements. This species is thus a prime example that an ecological classification system for cave-dwelling species, based on four categories may be too ambiguous particularly in cases where different life stages may behave ecologically distinct. Here we discuss the species three effective life stages, including the larva, pupa and imago in respect to the ecological classification system. We characterize the single developmental stages and potentially the whole species.

The larvae hatch and live solely within subterranean habitats reaching from microcavities to macrocaverns (or caves). They are incapable of surviving on the surface, which is related to their preference for a highly water-saturated atmosphere and adaptation to oxygen respiration. They lack a trachea system but are able to respire oxygen through a very thin cuticula spanning the entire body surface (Schmitz 1912). Author’s observations (SZ and AW) point to a behavior of avoiding strong air currents. Since the pupal stage is immobile and directly succeeds the larval developmental stage, it can only occur in the same habitat. The imago however, possesses elongated legs (sometimes regarded as a troglomorphy) (Schönborn 2003) and demonstrates a sluggish flight. At the same time, imagines do not feed and probably only survive a few days to weeks, thus complicating the inference of their ability to survive outside the subterranean habitat. However, several surface records of imagines (Table 3) and sightings of larvae in the transition and entrance zones of underground sites (Zaenker 2008, Weber 2012) are known. Furthermore, it seems that adult specimens travel between caves during the night.

Conclusively, not all developmental stages are capable of surviving outside the subterranean zone. Thus, the completion of a life-cycle aboveground will be incomplete – eliminating the eutrogloxenes as a potential ecological category for Speolepta leptogaster. Due to the short life-span of adults rarely leaving the subterranean zone, the complete life-cycle (or a very large proportion) is within the subterranean environment – leaving the eutroglophiles and eutroglobionts. From an evolutionary point of view and for long-term survival, Speolepta leptogaster should be classified as a eutroglophile, i.e. being able to complete several underground generations but having the ability of surface dispersal. But pinpointing Speolepta leptogaster to a single ecological category clearly underestimates the ecological versatility of the species.

Even after 150 years, some facts about the biology of Speolepta leptogaster are still unknown. Specifically, these include the number of larval stages and the durations of the different life stages. Although Schmitz (1912) claimed that he had successfully bred Speolepta leptogaster in captivity, he provided no information about the duration of any of the life stages. Still, a comparison with related Mycetophilidae (including the ecologically-similar keroplatid Arachnocampa luminosa (Skuse, 1890)) (Baker 2010, Li et al. 2011) may enable a vague picture of the potential life-cycle and the life-spans of the different developmental stages: a) egg: approx. 2 weeks; b) larval stages: 6 – 12 months; c) pupa: approx. 2 weeks and d) imago: 4 – 20 days (Figure 3).

Two types of larvae were present in our material. They can be distinguished by a small deviance of the ocelli position and by body measurements (Figure 4). In type A, the ocellus is situated adjacent to the antenna whereas in type B, the ocellus is approximately 15 to 20 µm behind the antenna. The head capsule of larvae type B was thinner (A: about 210 µm, B: about 190 µm), slightly shorter (A: about 300 µm, B: about 280 µm) and generally more pointed than in type A. The labrum of larvae type B displays a deeper indentation in the middle of the frontal end. At the proximal end of the head area where the vermiform body begins, the larvae type A displays an edged transition, whereas with larvae type B, it is more semicircular. In the sample set, three larvae were found that appeared as depicted in type B. Two of which had only half the total length of the average length of a normal larva (4.7 mm and 5.0 mm compared to 10 mm). A third specimen was of normal size. Since different larval stages between hatching and pupating are reported in other Mycetophilidae (Madwar 1937), this might explain the morphological discrepancy of larval type A and B in Speolepta leptogaster.