How soil granulometry, temperature, and water predict genetic differentiation in Namibian spiders (Ariadna: Segestriidae) and explain their behavior

Abstract The Namib Desert is a biodiversity hotspot for many invertebrates, including spiders. Tube‐dwelling spiders belonging to the Ariadna genus are widespread in gravel plains. These sit‐and‐wait predators share a particular behavior, as they spend their life in tunnels in the soil, surrounding the entrance of their burrow with stone rings. We investigated five spider populations taking into account environmental parameters, functional traits, and molecular data. We have chosen the temperature at the soil surface and at the bottom of the burrow, the air humidity, and the soil granulometry to define the environment. The chosen functional traits were the diameter and depth of the burrows, the ratio between weight and length, the thermal properties of their silks, and the number of ring elements. The molecular branch lengths and the evolutionary distance emerging from cytochrome oxidase I gene sequences summarized the molecular analysis. Our study highlights a strong coherence between the resulting evolutionary lineages and the respective geographical distribution. Multivariate analyses of both environmental and molecular data provide the same phylogenetic interpretation. Low intrapopulation sequence divergence and the high values between population sequence divergence (between 4.9% and 26.1%) might even suggest novel taxa which deserve further investigation. We conclude that both the Kimura distance and the branch lengths are strengthening the environmental clustering of these peculiar sites in Namibia.

These factors also produced a very high level of endemism (20% of described species are endemic to Namibia) and stimulated extraordinary adaptive responses to the environment that hosts them. In particular, the Central Namib Desert is a biodiversity hotspot for many vertebrates and invertebrates (Prendini & Esposito, 2010;Simmons, Griffin, Griffin, Marais, & Kolberg, 1998). Moreover, it is estimated that 11% of arachnids in Namibia are endemic and about 90% of the occurring invertebrates might not have been described yet (Ministry of Environment & Tourism, 2014).
Spiders have been also increasingly used as nonconventional terrestrial bioindicators (Conti et al., 2018;Wilczek, 2017). Their ability to survive under extreme conditions has allowed them to colonize several ecosystems. For instance, they can easily mitigate the intense heat by living below the surface (Lawrence, 1962). There are no biogeographical zones where spiders are not present and consequently, their ecological role cannot be underestimated (Nyffeler & Birkhofer, 2017).
About 25 years ago, Costa, Petralia, Conti, Hänel, and Seely (1993) recorded the existence of numerous and large tube-dwelling spider populations on the gravel plains of the Namib Desert. These populations were identified as belonging to the genus Ariadna (Segestriidae).
The genus Ariadna (Audouin, 1826), belonging to the Synspermiata spider family Segestriidae (Michalik & Ramírez, 2014), has an almost worldwide distribution. The taxonomy of Segestriidae seems to be quite chaotic (World Spider Catalog, 2018). This makes an integrated approach based on molecular data as well as ecological features and functional traits the best response to highlight how different populations can face different environmental conditions especially in such extreme habitats. Elsewhere, studies on functional traits like body size have been carried out in order to link phylogeny and ecosystem services as well as to identify phylogenetic relationships between species (Cavender-Bares, Kozak, Fine, & Kembel, 2009;Mulder et al., 2013). As far as we know, functional traits like the depth and diameter of burrows have never been taken into account for taxonomic surveys yet.
A global revision of the Ariadna genus becomes a hard task and to date, only some revisions limited to America have been carried out (Beatty, 1970;Giroti & Brescovit, 2018;Grismado, 2008). In particular, taxonomy and distribution of the Ariadna genus in Namibia (Lawrence, 1928;Purcell, 1904Purcell, , 1908Strand, 1906) are rather confusing, sometimes even without the sampling location of the specimens. Species that are recorded for Namibia are Ariadna insularis (Purcell, 1904(Purcell, ,1908, A. viridis (Strand, 1906), A. masculina (Lawrence, 1928), but taxonomical identification at species level was hard in the Central Namib Desert (Eryn Griffin personal communication). The DNA barcoding method based on cytochrome oxidase I gene appears to be particularly useful for a fine-tuned discrimination when morphological analysis is lacking (Čandek & Kuntner, 2015;Hebert & Gregory, 2005). In particular, DNA barcode reference libraries have been built for spiders both at regional scale (e.g., Astrin et al., 2016;Blagoev et al., 2016;Gaikwad, Warudkar, & Shouche, 2017;Naseem & Tahir, 2018) and at broader geographical scale (e.g., Robinson, Blagoev, Hebert, & Adamowicz, 2009;Barret & Hebert, 2005;Coddington et al., 2016). The purpose of this work is to understand to what extent ecological factors in a hyperarid environment might have led to a specific differentiation within the genus Ariadna with site-specific behavioral features.

| Fieldwork
Adult specimens of Ariadna spiders were collected from March 25 to April 21, 2012. Namibia can be easily classified into three F I G U R E 1 Location of the sampled areas. The enlargement shows the research stations inside the Namib Naukluft Park terrestrial biomes Desert, Karoo, and Savanna (Mendelsohn, Jarvis, Roberts, & Robertson, 2002). Of our research stations, four (G, M, R, W) are within the Desert biome and one (K) is in the Savanna biome ( Figure 1). The sites investigated can be described as follows: The G site (23°19.0′38.4″S, 15°2.0′23.3″E) lies in the Central Namib Desert, 56 km from the Atlantic coast and 25 km from the Gobabeb Research and Training Station. Fog is less frequent, and humidity is lower than in the areas closer to the coast but wind can blow strongly (Kaseke, Wang, & Seely, 2017;Viles, 2005;Wentworth, 1922 Like G, local precipitation is pulsed and unpredictable (Agnew, 1997;Jürgens, Burke, Seely, & Jacobsen, 1997) and fog is sparse if compared to the coastal areas. The mean annual humidity is lower, whilst the mean annual temperature is higher than else (Seely, 1987).
The R site (23°0.0′32.7″S, 14°43.0′38.0″E) is a part of the famous lichen area in the Namib Desert, 22 km from the Atlantic coast and 10 km from the Rooikop airport. A thick fog daily and strong winds are typical of this site (Costa & Conti, 2013;Seely, 1987;Viles, 2005) where a gravel plain consisting of a gravelly sandy sediment with small quartz pebbles, rich in lichens.
The W site (23°36.0′32.9″S, 15°10.0′2.7″E), characterized by the presence of some specimens of the dwarf gymnosperm Welwitschia mirabilis, is located 72 km from the Atlantic coast and 14 km from the Gobabeb Research and Training Station. The W site includes a 3-20 m wide river dry bed that is a dry tributary of the Kuiseb River (Henschel & Seely, 2000).
We measured the temperature at the soil surface and collected the spiders as described in the next subsection.

| Spider sampling
A total of 88 adult Ariadna specimens (about 20 specimens per site) were collected from their own burrows during our 2012 survey. For each specimen, we estimated weight using a Sartorius balance (model CPA225D) and total body length using a Borletti caliper (measurement error of 0.02 mm). We also measured in field the depth below surface and diameter of entrance burrows of each spider. After measurements, spiders were individually placed alive in Falcon tubes (50 ml) filled by half with sand collected from the burrow. They were kept at temperature of 20-22°C, approximately 67%-69% relative humidity, and photoperiod corresponding to that of the sampling area (i.e., 13 hr 15'L:10 hr 45'D) until their shipping to Italy for molecular analyses. Figure 2 shows the morphological difference between the burrows of Ariadna spiders. The individual burrow of these spiders is conspicuously different, due to its vertically oriented tube, internally covered with silk, and with a circular entrance surrounded by a stone ring, with sometimes lichen bits (Costa, Petralia, Conti, & Hänel, 1995), and the features of the burrow rings vary according to population and habitat (Costa et al., 2000).

| DNA barcode sequencing
Total genomic DNA was extracted from 38 entire spider specimens using the DNeasy Blood & Tissue Kit (Qiagen, Milan, Italy) according to the manufacturer's instructions. COI sequences were obtained using the primers HCO2189 (5-TAA ACT TCA GGG TGA CCA AAAAAT CA-3) and LCO1490 (5-GGT CAA CAA ATC ATA AAG ATA TTGG-3) (Folmer, Black, Hoeh, Lutz, & Vrijenhoek, 1994). All PCR amplifications were carried out in 25 µl total volume using approximately 50 ng of the isolated DNA as a template. In addition, each PCR contained 1X Taq DNA polymerase buffer (supplied by the respective Taq DNA polymerase manufacturer), 1.5-2 mM of MgCl 2 , 200 mM of each dNTP, 10 pmol of each primer and 0.5 U of Taq DNA F I G U R E 2 Ariadna burrow features from the investigated sites (clockwards: the G, K, M, W, and R sites). The burrow rings of the G and M sites include most commonly 6-7 quartz stones, similar in size, shape, and color and arranged in only one layer, but M rings are less regular than G ones. In the K site, the burrow rings include numerous small stones, placed in a single layer and differing in size, shape, and structure. In the R site, the burrow rings include 6-15 quartz stones mixed with pieces of lichens and arranged in one to four strata to shape a typical turret. Finally, in the W site, the burrows are dug on the slopes of the tributary and the ring stones are numerous, irregular and arranged in up to four strata polymerase (Platinum Taq DNA polymerase, Invitrogen). An initial denaturation at 94°C for 15 min was followed by 35 cycles (denaturation at 94°C for 30 s, annealing at 51°C for 1 min and extension at 72°C for 1 min) and a final extension at 72°C for 10 min. Negative controls were included in all PCR runs to ascertain that no crosscontamination occurred. Double-stranded products were checked with agarose gel electrophoresis, purified with the QIAquick PCR purification kit (Qiagen) and subsequently sequenced in the forward and reverse direction by Genechron (http://www.genechron.it/ index.php/sangersequencing) using an ABI Prism 3100 automated sequencer (Applied Biosystems). Sequences were carefully checked and deposited in GenBank (http://www.ncbi.nlm.nih.gov/genbank).
TA B L E 1 Specimens from spider populations used for the COI molecular analysis.

| Genetic analysis
The chromatograms obtained were edited using BioEdit ( to confirm the identity of the obtained fragments (the 5' COI region). The sequence divergences within and between Operational Taxonomic Units (OTUs) were calculated using the distance model Kimura-2-Parameters (Kimdist) and the bootstrapping proportion (1,000 iterations) was computed according to Hillis and Bull (1993).
The Kimdist dendrogram using the Neighbor-Joining algorithm, as clustering method for analysis of barcoding data (Hajibabaei, Singer, Hebert, & Hickey, 2007), was generated using as outgroups all the

| Statistical analysis
For statistical analyses, we considered six environmental variables: temperature at soil surface (Tsurf), temperature at the bottom of the burrow (Tdepth), humidity, and granulometry (i.e., percentage of gravel, silt, and sand). In addition, we examined five functional traits: diameter of the burrow entrance (DIA), burrow depth (DEPTH), bodymass index (BMI, calculated as ratio between weight and length of each spider), thermal properties of the silk considered as total normalized enthalpy of melting (DSC-Hm as in Conti et al., 2015), and total number of ring elements at the higher stratum (RINGS). Finally, we used two molecular variables: the value of branch lengths (Tree-BL, i.e., the sum of units of substitutions per site of the sequence alignment) and the evolutionary distance emerging from the Kimdist distinguishing between transitions from purine to purine or from pyrimidine to pyrimidine and transversions from purine to pyrimidine or from pyrimidine to purine (Kimura, 1980).
The normality of the data was tested by a Shapiro-Wilk test (Shapiro & Wilk, 1965) and the homogeneity of variance by a Levene's test (Levene, 1960). To test for differences between populations, we used a one-way ANOVA (using ΔT calculated as difference between Tsurf and Tdepth as temperature variable). Post hoc comparisons were conducted using the Dunnett's test (Dunnett, 1964)

| Soil and temperature
As expected, the soil granulometry of the investigated biomes is different, with the sites G, M, R, and W close to each other and rather far away from K, where the silt component is about 5-times higher than G (Figure 3)

| Analysis of variance
The statistical differences between populations according to ANOVA (Figure 4) , 1975;Wefer, Berger, Siedler, & Webb, 1996). Moreover, the sites G and M are mainly composed of sand, and therefore, their soils are highly permeable. Local humidity and granulometry are probably why these spider burrows are relatively deeper compared to K and W ones. Low depths imply spiders make little silk, and therefore, they do not need many elements to stabilize their home (Figures 2 and 4).
R depicts rather mild climatic conditions (moderate temperature and high humidity as in Seely, 1987)  as suggested by the largest number of items of the rings that are also different in size, shape, and structure (shown in Figure 2 and analyzed in Figure 4, upper panel).

| Multivariate analysis
As mentioned in the Materials and Methods, we performed a Varimax rotation of the principal components (PC) to maximize the independence between them (Figures 5 and 6 Therefore, multivariate analysis clearly shows how the sites can be easily distinguished based on environmental data alone ( Figure 5).
In particular, after the Varimax rotation, the first dimension (D1) is explained by microclimate (79.3% as we considered Tsurf and Tdepth separately avoiding ΔT to minimize redundancy;  Table 3).
Looking to the PCA of Figure 5 it is evident that on average the sites are much more scattered due to heterogeneity of the considered environmental variables. In particular, G, M, and R are stretched horizontally along the first dimension due to a strongly different microclimate. On the contrary, K and W stretched vertically along the second dimension due to highly different granulometry (Figure 3).
Surprisingly, using all variables together, the multivariate ordination of our Ariadna specimens is much less scattered than using   Figure 6 with Figure 5).
Comparing Figure 5 with

| DNA Barcoding analysis
The fragment of 617 bp of COI sequences investigated corresponds to the barcode region proposed by Hebert, Cywinska, Ball, and de-Waard (2003) and Hebert, Ratnasingham, and deWaard (2003)   and G-M-R although we strongly confirm that these Namibian taxa must be all species belonging to the worldwide distributed Ariadna genus as described by Beatty (1970). Summarizing, this remarkable coherence between microclimate, behavioral traits and evolutionary lineages for our five Ariadna populations makes clear how easily behavioral ecology provides the right perspective to recognize different taxa of spiders and possibly other invertebrates.

ACK N OWLED G EM ENTS
The authors thank Dr Alessandro Marletta for his help to collect some specimens. Financial support was provided by the Ministry F I G U R E 7 Neighbor-joining (NJ) tree generated by COI sequences of our Ariadna spiders and all the Segestriidae deposited in GenBank when accessed December 13, 2018. The bootstrap support (>70%) for each clade is indicated above the branches. The bar indicates the distance scale

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
EC and GC designed the study; EC and GC sampled the material; EC measured the spider traits in situ; AMP and VF generated the COI data, which were analyzed by AMP; EC and CM performed the statistical analysis; EC, AMP, and CM led the data quality assessment and interpreted the results. EC and CM led manuscript writing.

DATA ACCE SS I B I LIT Y
COI sequences were obtained using the primers HCO2189 (5-TAA ACT TCA GGG TGA CCA AAAAAT CA-3) and LCO1490 (5-GGT CAA CAA ATC ATA AAG ATA TTGG-3). Final DNA sequence assembly deposited in GenBank with the accession numbers as in Table 1.