Spider external morphology, with a specific focus on male and female genitalia was examined using a ZEISS Stemi 2000-C and a Nikon SMZ1270 stereomicroscope. The taxonomic identification was based on the resources available at Spiders of Europe (Nentwig et al. 2025).

Images of the female epigyne and male pedipalp of T. domestica were taken using a Nikon D5600 DSLR camera mounted on a Nikon SMZ1270 stereomicroscope (Fig. 3B, C). The specimens of P. vagans were photographed using a Kern ODC 825 microscope camera attached to a Kern OZL 464 trinocular stereomicroscope. Images of the female epigyne and male pedipalp were taken using a Kern ODC 825 microscope camera mounted on a Kern OBN 132 trinocular optical microscope (Fig. 3). Multifocal images were compiled using HeliconFocus software, digital images were processed using Photoshop software.

The morphological characterization of the spiders was complemented by DNA barcoding. DNA was isolated from the prosoma or legs of specimens collected in absolute ethanol and kept at 4 °C using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. For Tegenaria spiders, we amplified and sequenced Folmer’s fragment (COI) of the cytochrome c oxidase mitochondrial gene of 26 cave-collected individuals and 1 surface-collected individual (caught near the entrance of nearby Pixaria Cave) using the protocol described in Collard et al. (2025); Nanopore reads were subsequently assembled using amplicon_sorter (Vierstraete and Braeckman 2022). For the other spiders (namely, two specimens of Kryptonesticus, three specimens of Prinerigone, and two specimens of Metellina), a slightly shorter fragment of COI mitochondrial gene was amplified using the LCO1490 (5’-GGTCAACAAATCATAAAGATATTGG-3’; Folmer 1994) and chelicerate_reverse_2 (5’-GATGGCCAAAAAATCAAAATAAATG-3’; Barrett and Hebert 2005) primer pair. Those PCRs were performed in a 40 μL volume containing 1X Green GoTaq^® Flexi Buffer (Promega, Madison, WI, USA), 2.5 mM MgCl₂, 1X BSA, 0.1 mM dNTP, 1 U/μL GoTaq^® Flexi DNA polymerase, 0.1 μM each primer and up to 10 ng/μL DNA. The cycling conditions consisted of an initial denaturation at 95 °C for 2 minutes followed by 5 cycles of denaturation at 94 °C for 40 seconds, annealing at 45 °C for 40 seconds, extension at 72 °C for 1 minute, 35 cycles of denaturation at 94 °C for 40 seconds, annealing at 51 °C for 40 seconds, extension at 72 °C for 1 minute and a final extension step at 72 °C for 5 minutes. Sanger sequencing was performed by Macrogen (Amsterdam, The Netherlands) and the resulting trace files were visually inspected and curated in Chromas v.2.6.6 (Technelysium Ltd., South Brisbane, Australia) and CodonCode Aligner v.3.7.1 (CodonCode Corporation, MA, USA). The resulting sequences were compared to the publicly available databases GenBank (Sayers et al. 2022) and Barcode of Life Data System (Ratnasingham and Hebert 2024). Sequences belonging to the same taxon or related taxa were downloaded and aligned on the MAFFT web server (Katoh et al. 2019; https://mafft.cbrc.jp/align-ment/server/index.html, accessed on the 29^th of March 2025), sequence statistics and haplotype diversity (Hd) were estimated in DnaSP v.6 (Rozas et al. 2017) and haplotype networks were drawn in HaplowebMaker (Spöri and Flot 2020) using the median-joining algorithm (Bandelt et al. 1999).

As T. domestica is the dominant species in the communal web, our preliminary microbiome analysis has focused on this species. It included six individuals caught at the cave entrance, 13 individuals caught in the first section of the spider wall encountered when walking from the entrance into the cave, and 13 individuals caught in the second section of the spider wall (both on the left bank of the stream); as well as one surface individual caught near the entrance of Pixaria Cave, a few kilometers away, as a comparison. Whereas the first section of the Sulfur Cave spider wall is characterized by numerous egg sacs visible on the web, the second section is almost devoid of egg sacs, suggesting that reproduction chiefly occurs on the first section. Whole opisthosomas were used for DNA extraction, because the gut of spiders is not easy to separate from surrounding tissues. Although there are several organs located in the abdomen that contain microbial communities, the midgut is the largest one and it has been suggested in the literature that microbes associated with other organs do not influence the results significantly (see Kennedy et al. 2020 and references therein). DNA was extracted using the DNeasy PowerSoil Pro kit (Qiagen, UK), then 16S amplification and Nanopore sequencing was performed using the protocol described in Collard et al. (2025). The obtained reads were then analyzed in Emu (Curry et al. 2022) using its default database (Stoddard et al. 2015; O’Leary et al. 2016; Schoch et al. 2020).

The density of the T. domestica population was estimated by counting individual funnel-shaped webs and then extrapolating these counts to the surface area occupied by the colony. A 15 × 15 cm quadrant was placed near the cave wall where the colony was present, at 30 randomly selected locations. At each location two high-resolution pictures were taken of each quadrat. The number of individual webs per quadrat, identified by the central funnel structure, including those overlapping the center of the web, was counted from the pictures. Note that this methodology is likely to slightly overestimate the total individual count in the colony, as some funnel webs may be abandoned/unoccupied. This survey was carried out in October 2023, April 2024 and March 2025. To calculate the surface area occupied by the colony, the length (L) and width (W) of the web-covered section of the wall were measured, and the area was computed using the formula: Area = L × W, assuming the colony forms an approximately rectangular patch. The P. vagans population density was estimated in March 2025 by counting individual spiders visible within 15 × 10 cm quadrants photographed at 30 random locations. The total estimated population size was extrapolated to half of the surface area determined for T. domestica, reflecting the observed spatial distribution pattern of P. vagans. The density of flies associated with the colony was estimated using a similar approach: individuals were counted within quadrats based on photographs, and the mean density was extrapolated to the total area occupied by the colony.

A total of 86 egg sacs of T. domestica were collected from the surface of the colonial web during three sampling periods: June 2024 (n = 15), October 2024 (n = 40), and March 2025 (n = 31). Each egg sac was carefully opened using fine-tipped tweezers, and the number of eggs per clutch was recorded. Photographs were taken for further documentation. To assess egg size, 20 egg clutches were randomly selected, and three eggs from each clutch were measured using a Stage Micrometer Microscope calibrated to a precision of 0.01 mm.

For each species of invertebrate, a single individual was transferred into a tin capsule. Larger species were ground in a ball mill (Retsch Mixer Mill MM200, Haan, Germany), and 1–2 mg were weighed into the tin capsules. All samples were dried at 60 °C for 24 hours. Animal and biofilm samples were analyzed using a coupled system of an elemental analyzer (NA 1500, Carlo Erba, Milan, Italy) and a mass spectrometer (MAT 251, Finnigan, Bremen, Germany) adapted for the analysis of small sample sizes (Langel and Dyckmans 2014). Ratios of the heavy isotope to the light isotope (¹³C/¹²C, ¹⁵N/¹⁴N, denoted as R) were expressed in parts per thousand, relative to the standard using the delta notation with δ¹³C or δ¹⁵N = (R_sample/R_standard -1) × 1000 (‰); R_sample and R_standard being the respective target isotope ratio (¹⁵N/¹⁴N or ¹³C/¹²C, respectively). Vienna PD Belemnite limestone (V-PDB) and atmospheric nitrogen were used as standard for ¹³C and ¹⁵N, respectively. Acetanilide (C₈H₉NO, Merck, Darmstadt) was used for internal calibration. Details on the analysis are given in Reineking et al. (1993).

All statistical analyses were performed in R v.4.4.2 (R Core Team, 2024). To compare the number of eggs among sampling months, we conducted a one-way analysis of variance (ANOVA), followed by Tukey’s Honest Significant Difference (HSD) post hoc test for pairwise comparisons. Graphical representations of the data were generated using the ggplot2 package v.2.3.5.2 (Wickham 2016). Tabular data files containing the sequences and associated metadata downloaded from the BOLD database were transformed using the dplyr package v.1.1.4 (Wickham et al. 2023) and the sequences were exported to fasta files using the dat2fasta function of the phylotools package v.0.2.2 (Zhang 2024). Finally, we calculated Shannon’s diversity index for the microbiome of cave- and surface-dwelling Tegenaria domestica individuals using a custom code.

The large spider colony in Sulfur Cave was mainly observed on the left bank of the sulfidic stream of Sulfur Cave, in a permanently dark zone, starting at approximately 50 m from the cave entrance. In this section of the cave, the passage is relatively narrow, and the ceiling is mostly low (Fig. 5). The location of the colonial spiderweb coincides with an area where an unusually dense swarm of small chironomid flies identified as Tanytarsus albisutus Santos Abreu, 1918 (a member of the triangularis group), hovering above the sulfidic stream that flows along the left cave wall.

The estimated total surface area occupied by the spider colony was ~ 106 m². Based on the extrapolated quadrat counts, the total population size of T. domestica was estimated at 69,113 spiders with the mean density of 652 ± 149 individuals per m² (range: 178-2756). For P. vagans the estimated total number of individuals was 42,400, with a mean density of 823 ± 556 per m² (range: 133-2200). The density of the chironomid flies resting on the cave wall was estimated at 45,500 individuals per m². Based on this density and the observed spatial distribution pattern, where flies cover approximately half of surface area occupied by the spider colony, the total number of flies was extrapolated to be 2,414,440 individuals.

The mean number of eggs per T. domestica clutch was 16.15 ± 0.62 SD (range: 6–33 eggs) with significant seasonal variations (ANOVA: F_[2,83] = 69.42, p < 0.001). Tukey’s post hoc test showed that clutches collected in June (mean ± SD: 26 ± 3.12) contained significantly more eggs compared to those collected in October (14.5 ± 3.6) (mean difference: 11.550, 95% Confidence Interval (CI) = [8.966, 14.133], p < 0.001) and March (13 ± 3.73) (mean difference: 12.419, 95% CI = [9.735, 15.103], p < 0.001), while no significant difference was found between October and March (mean difference: 0.869, 95% CI = [-1.172–2.911], p = 0.569) (Fig. 6). The mean egg size was 0.53 ± 0.05 mm, with values ranging from 0.4 to 0.6 mm.

Figure 6. 

Monthly variation in the number of eggs per egg sac in Tegenaria domestica.

The δ¹⁵N and δ¹³C values in adult Tanytarsus albisutus (Diptera, Chironomidae), as well as in both agelenid and linyphiid spiders inhabiting the colony, ranged from −2‰ to −10‰, in contrast to conspecifics or ecologically similar surface-dwelling taxa, which showed values between 1‰ and 4‰ (Fig. 7).

Figure 7. 

Mean (+ standard deviation) of δ¹³C and δ¹⁵N values of four spider species (Tegenaria domestica, Metellina merianae, Prinerigone vagans, Lepthyphantes magnesiae, Araneae), their potential food resource (the non-biting midge Tanytarsus albisutus) and the basal food resource (biofilm) of the cave food web in Sulfur Cave located at the border between Albania and Greece (animal drawings by Svenja Meyer). ‘Surface spiders’ include representatives of the families Lycosidae, Salticidae, and Pholcidae.

In the one surface individual of T. domestica analyzed (collected near Pixaria Cave), we detected 30 different bacterial genera. By contrast, the microbiome of T. domestica from Sulfur Cave showed a much lower bacterial diversity (Fig. 8). The mean Shannon diversity index was 0.82 for Sulfur Cave individuals (standard deviation: 0.25), in contrast to 2.99 for the surface-dwelling individual. A large proportion of the bacterial reads in Sulfur Cave individuals originated from intracellular symbionts, primarily Mycoplasmopsis, Mycoplasma, and Wolbachia.

Figure 8. 

Comparison of the Shannon diversity index of the microbiome of the opisthosoma of one surface individual of Tegenaria domestica caught near Pixaria Cave vs. 29 individuals of the same species from Sulfur Cave (three Sulfur Cave individuals with insufficient number of reads were excluded from the analysis). S0 individuals were caught at the entrance of Sulfur Cave, S1 on the first section of the spider wall and S2 and the second section.

We report the discovery and detailed analysis of an extraordinary colonial spider assemblage in Sulfur Cave. Morphological examinations and molecular analyses identified Tegenaria domestica as being responsible for weaving the enormous colonial web. Prinerigone vagans (Andouin, 1826), a small linyphiid spider, was also observed co-habiting the web of the much larger agelenids. The colony covers a surface area of over 100 m² and represents the first documented case of colonial web formation in these species, comprising an estimated 69,000 individuals of T. domestica and more than 42,000 of P. vagans. The methodology used to estimate spider density may lead to an overestimation due to the presence of abandoned funnel webs that are difficult to distinguish from those in use.

Tegenaria domestica is a cosmopolitan species with a global distribution (Nentwig et al. 2025; World Spider Catalog 2025). It is a nocturnal and synanthropic spider usually found close to human habitations, such as in buildings, under stones, in dark recesses, i.e. closets and basements, with adults present year-around (Nentwig et al. 2025). While frequently found in caves across Europe (Mammola et al. 2018, 2022) and considered a characteristic species for Balkan caves (Bristowe 1958; Deltshev et al. 2011), prior literature did not describe T. domestica exhibiting gregarious behavior and forming large colonial webs. The species typically builds a web resembling an open-cut funnel, with a tubular retreat open on both ends and its edge widening on one side to form a sheet-like capture surface (Nentwig et al. 2024). On a close examination, the large colonial web appears to be the result of joining numerous individual funnel-shaped webs, each strategically positioned in a spot of abundant trophic resource availability. Sections of the web may detach from the wall under their own weight. The colonization of Sulfur Cave by T. domestica was most likely driven by the abundant food resources represented by the dense swarm of chironomids thriving in the cave.

Prinerigone vagans, the co-habiting spider species within the colony, apparently did not elicit a predatory response from T. domestica. We hypothesize that the absence of light impairs the visual detection capabilities of T. domestica, rendering the small P. vagans inconspicuous, particularly given P. vagans stationary ambush predatory strategy with brief movements when prey items are in close proximity within the web. Although we did not conduct a specific study to quantify seasonal variations, our observations indicate no discernible changes in the species composition and abundance of the two spider species that co-habit the large colony in Sulfur Cave. Comparable densities of these spiders were observed throughout the year during our field trips in May, June, and November 2023, in April, July, and October 2024, and in March 2025. Sustaining a predator colony of this size inside a cave requires a substantial and consistent supply of food and this is unlikely to be met solely by allochthonous resources from the surface, where resources become rather scarce during the cold season.

Metellina merianae (Scopoli, 1763) (Tetragnathidae) (Fig. 9) has been previously reported from Sulfur Cave (Audy et al. 2022). The juveniles prey upon collembola, while the adults feed on the two species of chironomids: the Tanytarsus albisutus and the larger Chironomus riparius (Meigen, 1804), mostly present in the deeper sections of the cave. M. merianae builds solitary orb webs and was not observed within the colonial web, yet it is quite abundant in close proximity to the colony where the dense chironomid swarm occurs. Specimens observed close to the colonial web appear to exhibit larger body size, presumably due to the readily available and abundant food resources. However, this species was observed throughout the cave, including the areas covered by gypsum crusts that appear rather dry. As a troglophile species (Mammola et al. 2018, 2022, Nentwig et al. 2025), M. merianae appears to be versatile and able to adapt to both wet and dry cave habitats.

Figure 9. 

Metellina merianae. Female (left) and male (right) in individual webs on the cave wall.

Lepthyphantes magnesiae Brignoli, 1979 (Araneae: Linyphiidae) was observed on the cave wall opposite the colonial web. It builds individual webs and it shares the habitat with a pseudoscorpion species belonging to the genus Neobisium. Both these predator species have been noticed in the humid areas at the base of walls, where they appear to feed on collembola and chironomid flies. Lepthyphantes magnesiae is considered a rare species endemic to the Balkan Peninsula.

In the deep recesses of the cave, far from the cave entrance and the colonial web, two additional web-building spider species were also observed: Kryptonesticus eremita (Simon, 1880) (Nesticidae), previously reported from Sulfur Cave by Audy et al. (2022), and a very small, blind, and depigmented species of Cataleptoneta (Leptonetidae). Both K. eremita and Cataleptoneta sp. were present in large numbers but were spatially limited to areas where the cave walls were moist and free of gypsum crusts.

Haplotype networks revealed both similarities and contrasts among the spider species found in Sulfur Cave. For T. domestica, K. eremita, and P. vagans, the cave individuals shared distinct haplotypes not found in broader regional datasets, though all were closely related, i.e. within three mutations, to widespread haplotypes from Europe, the Middle East, or Asia. In contrast, M. merianae showed a more complex and reticulated haplotype network, with two individuals from Sulfur Cave belonging to different haplotypes, one shared with German samples. Despite the geographical proximity of some samples to Sulfur Cave (e.g. from Slovenia, Bulgaria, North Macedonia), they were quite distant in the network. This suggests that M. merianae is both more geographically mobile and more genetically diverse than the other spider species found in Sulfur Cave.

The reproductive effort, measured as the number of eggs per sac, differed from values reported in the literature for surface-dwelling conspecifics. Females of T. domestica are known to lay six to eight egg sacs at an interval of 20–25 days, with the first cocoon containing up to 100 eggs, followed by smaller successive clutches (Trabalon et al. 1992). Numerous egg-sacks were observed in the large spider web, but their number was not estimated as the multi-layered web made it impossible to count the egg clutches. The underlying factors contributing to a potential reduction in clutch size in the Sulfur Cave population of this species are yet to be determined, although environmental factors such as perpetual darkness and sulfidic conditions have been shown to be associated with reduced fecundity and increased offspring size (Riesch et al. 2010).

At the microbiome level, we found that, by contrast to one surface-dwelling individual of T. domestica, collected in the vicinity of the cave, in which approximately 30 different bacterial genera were detected, the cave-dwelling individuals investigated exhibited markedly lower bacterial diversity. In these individuals a large proportion of the reads originated from intracellular bacterial symbionts, such as Mycoplasmopsis, Mycoplasma and Wolbachia. Although the results are quite preliminary and do not allow to distinguish the relative contributions of the various parts of the opisthosoma to the measured diversity, it suggests that the population of T. domestica from Sulfur Cave presents a much-reduced microbial diversity than surface ones.

Preliminary observations indicate that in the deep cave sections, the food web is based on chemosynthetic carbon fixation in terrestrial microbial biofilms that cover the moist cave walls. These biofilms consist of sulfur-oxidizing microorganisms (unpublished). Numerous specimens of Graeconiscus sp. (Isopoda, Trichoniscidae) and dense populations of collembola appear to feed on this cave-wall microbiome, forming the base of a food web that supports two species of spiders and other small terrestrial predators such as centipedes, pseudoscorpions, mites, and beetles (Sarbu et al. 2024).