Systematics and phylogeography of the Dysdera erythrina species complex ( Araneae , Dysderidae ) in Sardinia

Sardinia is the second largest island in the Mediterranean and, together with Corsica and nearby mainland areas, one of the top biodiversity hotspots in the region. Th e origin of Sardinia goes back to the opening of the western Mediterranean in the late Oligocene. Th is geological event and the subsequent Messinian Salinity Crisis and Pleistocene glacial cycles have had a major impact on local biodiversity. Th e Dysdera woodlouse hunter spiders are one of the most diverse ground-dweller groups in the Mediterranean. Here we describe the fi rst two species of this genus endemic to Sardinia: Dysdera jana sp. n. and Dysdera shardana sp. n. Th e two species show contrasting allopatric distribution: D. jana sp. n. is a narrow endemic while D. shardana sp. n. is distributed throughout most of the island. A multi-gene DNA sequence phylogenetic analysis based on mitochondrial and nuclear genes supports the close relationships of the new species to the type species of the genus Dysdera erythrina. Age estimates reject an Oligocene origin of the new Dysdera species and identify the Messinian Salinity Crises as the most plausible period for the split between Sardinian endemics and their closest relatives. Phylogeographic analysis reveals deep genetic divergences and population structure in Dysdera shardana sp. n., suggesting that restriction to gene fl ow, probably due to environmental factors, could explain local speciation events. ZooKeys 16: 319-345 (2009) doi: 10.3897/zookeys.16.128 www.pensoftonline.net/zookeys Copyright Miquel A. Arnedo et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Launched to accelerate biodiversity research A peer-reviewed open-access journal


Introduction
Th e Mediterranean basin is one of the Earth's biodiversity hotspots for conservation priority (Myers et al. 2000).A dynamic and tumultuous geological history ranks high among the main factors responsible for the outstanding diversity of the region.Th e island of Sardinia constitutes a good example of this ever-changing geological landscape.At the beginning of the Oligocene (~30 million years ago, m.a.), several continental microplates, including present-day Sardinia, Corsica, the Balearic Islands, Calabria, the Kabylies and the Betic-Rift Cordillera, broke off from the north-eastern Iberian Peninsula as a result of geological back-arc extension processes, and started drifting to their present-day location.During the fi rst stage of the back-arc extension, the northern microplate assemblage (i.e., Sardinia, Corsica and Calabria) started drifting counter-clockwise with respect to the Eurasian plate and fi nally collided with the Apulian plate about 20-18 m.a.(Rosenbaum et al. 2002).Th ere is no general consensus on the time of separation of Corsica and Sardinia.Some authors suggest that Sardinia broke off from Corsica 20 to 16 m.a.(Rosenbaum et al. 2002); while others consider that the process would not have been fully completed until about 9 m.a.(Alvarez 1972(Alvarez , 1974)).Th e closing of the Miocene marine gateways between the Mediterranean Sea and the Atlantic Ocean, 5.96 m.a., caused a large drop in the Mediterranean water level (Blondel and Aronson 1999; Krijgsman et al. 1999) and the emergence of land connections between Sardinia and northern Italy and southern France.Th e opening of the Strait of Gibraltar about 5.3 m.a.restored the water exchange between the Atlantic and Mediterranean basins, re-establishing eff ective isolation of the island ecosystems (Loget and Van Den Driessche 2006;Loget et al. 2005).Eustatic sea level changes as a result of Pleistocene glacial cycles could have connected Sardinia with Corsica, and with the mainland, via Elba.Sea level oscillations creating land bridges between Sardinia and Corsica from the Miocene until well into the Pleistocene (5.7 -0.23 m.a.) have been additionally suggested (Arias et al. 1980).
Sardinia is the second largest island in the Mediterranean (Fig. 1), generally mountainous with a few coastal plains.Th e island's mountains are divided into three ranges that largely coincide with areas of high endemism, namely Mount Limbara on the north, the Gennargentu-Supramonte Massif on the central-eastern side, and the Sette Fratelli on the south-east (Grill et al. 2007).Additionally, a separate biogeographic region has been proposed for the Sulcis-Iglesiente territory in the south-western part of Sardinia, based on exclusive plant endemism and a unique blend of geolithologic, geomorphologic, paleogeographic, bioclimatic and vegetation features (Bacchetta et al. 2007).It is worth noting that the Tyrrhenian islands, Sardinia and Corsica, and the Maritime-Ligurian Alps, are among the regions with the highest level of narrow endemism in the Mediterranean basin (Medail and Quezel 1999).Th e woodlouse-hunter spiders of the genus Dysdera Latreille, 1804 include about 250 described species (Platnick 2009) of mostly nocturnal, wandering, ground-dwelling spiders.Although the geographic distribution of the genus includes most of the Western Palaearctic, it is especially diverse in the Mediterranean region, where it constitutes one of the most conspicuous components of the local ground arthropod fauna.Th e erythrina group was originally proposed by Deeleman-Reinhold and Deeleman (1988) to include the type species of the genus, Dysdera erythrina (Walckenaer, 1802), as well as some other species described from the Iberian peninsula, which included D. fuscipes Simon, 1882, D. anonyma Ferrández, 1984and D. veigai Ferrández, 1984.Th is group was proposed to be related to the crocata group, from which it would diff er by the shape and size of the carapace and the chelicera, the spination pattern and the shape of the male bulb (male secondary copulatory structure, located on palpal tarsi).Th e great French arachnologist E. Simon described three subspecies of D. erythrina, namely D. erythrina lantosquensis Simon, 1882 from France, D. erythrina fervida Simon, 1882 from Corsica and the Balearic Islands and D. erythrina provincialis Simon, 1882 from France (Platnick 2009).Subsequently, Řezáč and collaborators (2007) have granted species status to D. lantosquensis, and suggested that the rest of the subspecies should also be considered diff erent species.
Sardinia harbours a rich endemic Dysderidae fauna, which includes four Harpactea Bristowe, 1939 species, one Parachtes Alicata, 1964 and the monotypic cave-dwelling endemic genus Sardostalita Gasparo, 1999(Platnick 2009;Stoch 2003).Endemic Dysdera species have not been reported from Sardinia so far.Th e crocata group is represented by the synanthropic D. crocata C.L. Koch, 1839, D. ancora Grasshoff , 1959and D. ventricosa Grasshoff , 1959, the latter two species also present in Sicily (Stoch 2003).Grasshoff (1959) reported the presence in Sardinia of D. kollari Doblika, 1853, a species previously reported from the eastern Adriatic coast (Chyzer and Kulczyński 1897) and northern Greece, and described the formerly unknown male (Grasshoff 1959) (it turned out to be a misidentifi cation, see below; a drawing of the actual D. kollari male palp can be found in Gasparo 2004).However, the former citation has been suggested to refer to another species (Deeleman-Reinhold and Deeleman 1988).Finally, there is an early record of the presence of D. erythrina from Sardina (Costa 1885), which has not been subsequently confi rmed to date.
In the present paper we confi rm the presence of the erythrina group in Sardinia based on museum collections and newly collected material, and describe two new species, one of them including specimens previously misidentifi ed as D. kollari.Phylogenetic position, time of origin and phylogeographic patterns of the new species are further investigated by means of molecular evidence.

Taxonomy
Morphological methods are described in detail in Arnedo and Ribera (1999) and Arnedo et al. (2000).Taxonomic descriptions follow the format of Arnedo and Ribera (1999).Specimens were examined using a Leica MZ16A stereoscopic microscope equipped with a Nikon DXM1200 digital camera and an ocular measuring graticule.Digital microscope images were edited using the Auto-Montage software package v.4 (Syncroscopy, Frederick, MD, USA)].Male right palps were removed, cleaned by means of ultrasound and examined using either a HITACHI S-2300 Scanning Electron Microscope (SEM) (SCT, Universitat de Barcelona, Barcelona, Spain) operated at 15 kV.Left male palps were used for optic images and right palps were used for SEM images.All morphological measurements are given in millimetres.Eye diameters were taken from the spans of the lens.Carapace and abdomen measurements were taken in dorsal view, abdomen hairs were measured in lateral view, and cheliceral basal segment length was measured in lateral view.Th e prolateral groove of the chelicera was measured in dorsal view, by positioning the chelicera parallel to the background and measuring from the distal end of the margin to the beginning of the cheliceral lamina.Th e fang was measured in ventral view, from the basal segment condyl to the fang distal tip.Th e largest leg article lengths were measured in lateral view without detaching the legs from the specimen, by placing the article being measured in a perpendicular position.Th e female vulva was removed and muscle tissues were digested using a KOH (35%) solution before observation.Leg spination was recorded using the codifi cation method fully described in Arnedo and Ribera (1997).

Molecular phylogeny
Taxa analysed in the present study, along with locality information and the GenBank accession numbers of the sequenced genes are listed in Table 1.Fresh specimens of D. jana sp.n. were not available for DNA analysis.Phylogenetic relationships of D. shardana sp.n. were inferred from a taxonomic sample that included all subspecies of Dysdera erythrina, representatives of the erythrina group (D. cf.inermis), putative related species based on genitalia characteristics (D. lusitanica Kulczyński, 1915, D. valentina Ribera, 2004), as well as two members of the crocata group.D. adriatica Kulczyński, in Chyzer and Kulczyński, 1897 was used as an outgroup based on the results of Macías-Hernández et al. (2008).Th e inclusion of D. cf.inermis specimens from both sides of the Strait of Gibraltar provided a vicariant biogeographic point (opening of the strait 5.3 million years ago) to estimate lineage age, by assuming that the separation of Iberian and Moroccan populations of D. cf.inermis was the result of this geological event.Th e same assumption has been used to calibrate molecular phylogenies in Canarian Dysdera (Macías-Hernández et al. 2008), scorpions (Gantenbein 2004), beetles (Gómez-Zurita 2004) and newts (Carranza and Arnold 2003).A second data matrix was built to investigate phylogeographic patterns in D. shardana sp.n. by sampling all freshly collected specimens amenable to DNA analyses, namely 14 specimens from localities in central Sardinia (Mt. Ferru, Mt. Arci, and Gennargentu).
Protocols for specimen handling, DNA extraction, amplifi cation, and sequencing of the gene fragments followed those of Macías-Hernández et al. (2008).Th e following gene fragments were analyzed: mitochondrial genes cytocrome c oxidase subunit I (cox1, 676 bp), 16S rRNA (16S), the complete tRNA leu UAG (L1) (16S+L1, 387-546 bp), NADH dehydrogenase subunit I (nad1, 362 bp), and the nuclear genes 28S rRNA (28S, 620-768 bp) and histone H3 (H3, 328 bp).Sequences were edited and handled using the computer program GENEIOUS v. 4.5 (Drummond et al. 2009).Alignment of the protein coding genes was trivial, since no length diff erences were observed and no indel events needed to be postulated.Sequences of 16S-L1 and 28S were automatically aligned using the online version of the program MAFFT v. 5.8 (Katoh et al. 2005;Katoh et al. 2002) (manual strategy set to G-INS-i).Gaps were subsequently recoded as separate presence/absence characters with the aid of the program GapCoder (Young and Healy 2003) following Simmons and Ochoterena (2000), to minimise the eff ect of increasing the weight of overlapping multiple non-homologous gaps (Pons and Vogler 2006).Gene partitions were combined in a single data matrix using WINCLADA v.1.00.08 (Nixon 2002), with non-sequenced fragments scored as missing data.
Parsimony analyses of the data matrices were conducted with the program TNT v. 1.1 (Goloboff et al. 2003), using the following search strategy: 1,000 iterations of Wagner trees constructed with random addition of taxa and subsequent TBR branch swapping, holding fi ve trees per iteration up to a total maximum of 10,000.Clade support was assessed via jackknife resampling using 1,000 replicates with individual heuristic searches consisting of 15 iterations of Wagner tree construction using random  addition of taxa, holding fi ve trees per iteration and an overall maximum of 10,000.Th e program jMODELTEST v. 0.1.1 (Posada 2008) was used to select the substitution model that best fi ts the data with the fewest parameters (including branch lengths as parameters), as indicated by the Akaike information criterion (AIC) (Akaike 1973).Bayesian inference analyses were conducted with MRBAYES v.3.1.2(Ronquist and Huelsenbeck 2003) and run remotely at the University of Oslo's Bioportal computer resources (http://www.bioportal.uio.no).Th e substitution models suggested by jMODELTEST were specifi ed for each gene fragment and a standard discrete model was implemented for the gaps scored as absence/presence data (Nylander et al. 2004).Th e substitution estimates were allowed to vary independently between each partition.Two independent runs with four simultaneous MCMC (Markov Chain Monte Carlo) chains (one cold and three heated), each starting with random starting trees, were carried out simultaneously, sampling 1,000 generations until the standard deviation of the split frequencies of these two runs dropped below 0.01 (10 million generations).Th e program TRACER v. 1.4 (Rambaut and Drummond 2003) was used to ensure that the Markov chains had reached stationary by examining the eff ective sample size (ESS) values and also to determine the correct number of generations to discard as a burn-in for the analysis.

Species
Maximum likelihood searches were conducted with the stand-alone version of the computer program RAxML v. 7.0.3(Stamatakis 2006).Individual models of nucleotide substitution (GTR+Γ+I, with default number of Γ-categories) were specifi ed for each gene partition.Th e best trees were selected from 100 multiple inferences, and clade support was assessed by means of 100 non-parametric bootstrap resampled replicates of the original matrix.
Uncorrected genetic distances were estimated with MEGA v.4.0 (Tamura et al. 2007).A haplotype network was estimated from the cox1 sequences using statistical parsimony, with probability connection limit set to 95%, as implemented in the program TCS v. 1.21 (Clement et al. 2000).
Lineage age estimates were obtained with the aid of the computer program R8S v. 1.71 (Sanderson 2003).Branch lengths were re-estimated with the computer program RAxML forcing the topology obtained in the Bayesian analysis and specifying independent GTR+I+G substitution models for each gene fragment.Th e same program was used to generate 100 trees with the same topology by bootstrapped branch lengths, to generate confi dence intervals of the times of divergence.Trees were rooted and very short or zero branch length branches were pruned with the help of the software TreeEdit v. 1.0 (Rambaut and Charleston 2001).A preliminary cross-validation analysis (Sanderson 2002) conducted with r8s selected the penalized likelihood method with a smoothing parameter value of 3200 and additive penalty function as the best options for reconstructing divergence times in our data set.Subsequent analyses with the preferred method were conducted to obtain optimal ages and corresponding confi dence intervals.Th e resulting chronograms were visualised and manipulated with the assistance of the computer program FigTree v. 1.1.2(http://tree.bio.ed.ac.uk/software/fi gtree/).

Results
Results of the phylogenetic analyses conducted on all gene fragments concatenated in a single matrix are summarised in Fig. 2. Heuristic searches in TNT under equal weights yielded 3 trees of 2,015 steps.Th e best fi t-models selected by the AIC for each data set were: TrN+I+G, TIM2+G, TIM2+G, GTR+G and TrNef+G for the cox1, 16S-L1, nad1, 28S and H3, respectively.All analyses support the sister group relationship of D. shardana sp.n. with the present and former subspecies of D. erythrina.All analyses support the hypothesis that D. erythrina sensu stricto and D. erythrina provincialis are sister species, while the sister-species relationship of D. erythrina fervida and D. lantosquensis is only contradicted by the maximum likelihood analysis, although with low support.D. valentina is resolved as the sister group to the clade formed by D. shardana sp.n. and D. erythrina sensu lato, although only the Bayesian analysis provides high support for this relationship.Th e phylogenetic analyses do not support a close relationship between either D. cf.inermis or D. lusitanica with the erythrina group.Parsimony partial analyses were run to assess the contribution of each data partition.Th e levels of support across the partial analyses were higher in the mitochondrial partition, followed by the 28S gene fragment, and the histone 3 was the less resolved.Instances of incongruence among the partial analyses were, in general, poorly supported.Th e mitochondrial genes   yielded a single tree of 1800 steps, which mirrored results of the combined data set except for the position of D. valentina, which was sister to the D. lancerotensis+D.crocata clade (JS=60).Th e analysis of the 28S gene fragment yielded 5 trees of 136 steps, the strict consensus of which supports the sister group relationship of D. valentina and the erythrina group, but does not resolve the position of D. shardana sp.n.In addition, D. lantosquensis formed a clade with D. erythrina and D. erythrina provincialis, although with low support (JS=62).Th e histone 3 fragment yielded a single tree of 63 steps.Th e erythrina group, D. shardana sp.n. excluded, was recovered as a clade (JS=60), but the haplotype k105 of D. erythrina provincialis grouped with D. fervida (JS=61).Lineage age estimates are also summarized in Fig. 2. Th e split of the ancestor of D. shardana sp.n. from the remaining species of the erythrina group is estimated at 6.4 m.a.although confi dence intervals recovered are very wide (from 1.9 to 10 m.a.).Th e fourteen D. shardana sp.n. individuals sequenced for the cox1 fragment resulted in 9 unique haplotypes.Statistical parsimony analysis of these resulted in two separate networks (Fig. 3) with strong geographic structure.Th e fi rst network included the haplotypes sampled from Mt. Arci and the second network connected the haplotypes from Mt. Ferru, on the one hand, and those from Gennargentu, on the other.One specimen from Mt. Ferru remains isolated from any of the networks.Uncorrected genetic distances for the cox1 gene between D. shardana haplotypes and between and within haplotypes of the species related to D. shardana sp.n. are summarized in Table 2. Mean genetic distance between D. shardana sp.n. haplotypes (3.8%) was very similar to those observed in closely related species (4.6% and 3.4%.for D. lantosquensis and D. erythrina provincialis, respectively).Diagnosis.Dysdera jana sp.n. can be distinguished from continental species of the erythrina group by the shape of the posterior-distal part of the tegulum, which is expanded in the continental species.It can also be distinguished from D. erythrina and D. erythrina provincialis by the lateral concave shape of the chelicera and a heavily wrinkled (foveate) carapace (Figs 4,6).In addition, continental females of the erythrina group exhibit a characteristic hourglass-shaped spermatheca, which is very attenuated in D. jana sp.n. (Fig. 11).It diff ers from the other Sardinian species of the erythrina group by the spineless dorsal tibia 4, the position and length of the fi nger-like process on the tip of the male bulb (compare Figs 15 and 21), and by the fusion of the anterior sclerotised stripes of the vulva VA (compare Figs 11 and 13).
Etymology.Th e name in apposition refers to the "Domus de janas", characteristic Sardinian tombs dug in the rock often according to a labyrinth architecture.Domus de Jana translates in Sardinian as "House of the fairies or of the witches".It also happens to be the name of the fi rst author's goddaughter.
Abdomen 3.73 mm long.Abdominal dorsal hairs 0.07 mm long; thin, curved, not compressed, pointed; uniformly, thickly distributed.Vulva DA (Fig. 10) clearly distinguishable from VA; DA slightly wider than long; DF wide in dorsal view.MF margins not fused, poorly developed, membranous.VA rectangle-like; frontal region completely sclerotized (Fig. 11); posterior region internal margin sclerotized, forming two stripes fused at the anterior part, diverging backwards; AVD absent.S attachment not projected under VA; arms as long as DA, slightly curved; tips not projected; neck as wide as arms.TB usual shape.
Distribution.Known from several localities around the Supramonte region, a mountain and highland range in Sardinia, lying in the north-western part of the Gulf of Orosei.Diagnosis.Dysdera shardana sp.n. diff ers from continental species of the erythrina group in the shape of the posterior-distal part of the tegulum, which is expanded in the continental species.In addition, continental females of the erythrina group exhibit a characteristic hourglass-shaped spermatheca, which is very attenuated in D. shardana sp.n.It can also be distinguished from D. erythrina and D. erythrina provincialis by the lateral concave shape of the chelicera and a heavily wrinkled (foveate) carapace (Figs 5,7).It diff ers from D. jana sp.n., also from Sardinia, in the presence of dorsal spines on tibia 4, the position and length of the fi nger-like process on the tip of the male bulb (compare Figs 15 and 21), and by the fusion of the anterior sclerotised stripes of the vulva VA (compare Figs 9 and 11).
Etymology.Th e name in apposition refers to the Shardana people, one of several groups of "Sea Peoples" that appear in fragmentary historical records (Egyptian inscriptions) for the Mediterranean region in the second millennium B.C.According to some scholars they settled in Sardinia after the period of Ramses III (around 1180 B.C.).
Description.Holotype male num.CRBA 1246.Figs 5, 7, 9, 19-22.Carapace (Fig. 5) 3.01 mm long; maximum width 2.35 mm; minimum width 1.68 mm.Dark red, darkened at borders; foveate at borders, slightly wrinkled in the middle, covered with small black grains.Frontal border roughly round, from 1/2 to 3/5 carapace length; anterior lateral borders parallel; rounded at maximum dorsal width, back lateral borders straight; back margin wide, straight.AME diameter 0.18 mm; PLE 0.18 mm; PME 0.16 mm; AME slightly backwards from frontal border, separated from one another by about 2/3 diameter, close to PLE; PME close to each other, less than 1/4 PME diameter from PLE. Labium trapezoid-shaped, base wider than distal part; longer than wide at base; semicircular groove at tip.Sternum brownish orange, darkened on borders; wrinkled; uniformly covered in slender black hairs.
Male copulatory bulb (Fig. 9 curved,compressed,pointed;uniformly,) DA clearly distinguishable from VA; DA slightly wider than long; DF wide in dorsal view.MF margins not fused, poorly developed, membranous (Fig. 12).VA rectangle-like, pointed expansion at middle frontal part (Fig. 13); frontal region completely sclerotized; posterior region internal margin sclerotized, forming two divergent stripes; AVD absent.S attachment not projected under VA; arms as long as DA, slightly curved; tips not projected; neck hardly visible.TB usual shape.
Variation.Carapace ranges in length from 2.50 mm to 3.42 mm, males (n=7) average 2.79 (2.50-3.06),female (n=8) average 3. 07 (2.68-3.42).In general male legs  bear fewer spines that female legs, one male form Mt. S. Antonio representing an extreme case of reduced spination.Spination variability summarised in Table 4. Distribution.Known from several localities around the western half of Sardinia, and also in the Gennargentu region on the central-eastern side of the island, spanning from 360 to above 1100 m elevation.

Discussion
Molecular phylogenetic analyses unambiguously support the close relationship of Dysdera shardana sp.n. with the species of the erythrina group.Morphological synapomorphies for this clade include a large, semicircular expansion (AR) at the DD tip, running parallel to the bulb proximal-distal axis and a bilobular, granulated posterior diverticulum in the female vulva.Both D. shardana sp.n. and D. jana sp.n. exhibit a regular-shaped tegulum.Th is suggests that the particular shape of the tegulum with a mid-part constraint and an expansion around the P region is most likely a synapomorphy of the clade formed by D. lantosquensis and D. erythrina.Th e erythrina group was not defi ned in cladistic terms and hence it is not possible to assess its limits.Th e taxonomic sampling of the phylogenetic analysis conducted in this study included two species closely resembling those originally proposed to be part of the erythrina group, namely D. cf.inermis (closely similar to D. fuscipes) and D. lusitanica (resembling D. veigai).Th e phylogenetic analyses revealed that they are not closely related to D. erythrina.Deeleman-Reinhold and Deeleman (1988) did not justify the inclusion of D. fuscipes and D. veigai species into the erythrina group and thus it is diffi cult to evaluate convergence in morphological traits.All these species, however, share a particular shape of the P, which is in most cases short and with the proximal margin parallel, instead of perpendicular, to the proximal-distal axis of the bulb.Close examination of this structure, however reveals, obvious diff erences among the species (e.g., the particular shape observed in the D. lantosquensis / D. erythrina clade is probably synapomorphic for this group).Th e eastern Iberian species D. valentina was included in the analyses based on the similarities in the distal part of the male bulb DD to D. erythrina.All analyses support (although only marginally under parsimony and maximum likelihood) the sister group relationship of D. valentina with the D. shardana sp.n. / D. erythrina clade, suggesting that the presence of a long, arch-like expansion on the anterior-distal side of the DD, running parallel to a short L (see Ribera 2004, Figs 1A-B) could constitute a synapomorphy of a more inclusive clade that would also embrace the erythrina group.More morphological comparative data and new phylogenetic analyses based on more exhaustive taxonomic sampling would be required to confi rm the latter suggestion.
Based on morphological evidence, there is little doubt that the two new Sardinian Dysdera species are closely related.Th ey are extremely similar in somatic morphology (some minor diff erences in leg spination) and share at least one exclusive character: the presence of a fi nger-like projection at the tip of the male bulb (although it is not completely clear whether the process actually grows from the same position, see diagnosis).
Unfortunately, the lack of genetic data for D. jana sp.n. precludes discerning whether the two endemics are sister species or whether one of them is actually more closely related to the D. erythrina /D.lantosquensis clade.Examples of local diversifi cation in Sardinia have been reported in other ground-dwelling animals.Th e cave beetle genera Ovobathysciola and Patriziella form a monophyletic group that includes six endemic species in Sardinia.Molecular clock analysis suggests that separation among the Ovobathysciola / Patriziella species started about 16 to 10 m.a., assuming that the clade originated as a result of the split of the Corsico-Sardinian plate from the Iberian peninsula (~30 m.a.) (Caccone and Sbordoni 2001).Th e Hydromantes cave salamanders, on the other hand, include fi ve endemic lineages in Sardinia, which have been recently shown to form a paraphyletic group with regard to closest relatives in south-eastern France and northern-central Italy (Carranza et al. 2008).In this case, molecular time estimates suggest that the Messinian Salinity Crisis facilitated the two independent colonisation events that gave rise to present day Sardinian endemics (although a single colonisation of the island followed by a back colonisation of the mainland is an equally parsimonious explanation).Th e unresolved phylogenetic position of D. jana sp.n. hinders distinguishing between a scenario of a single colonisation of Sardinia followed by in situ speciation of the two new Dysdera species from a scenario with a colonisation of Sardinia by the common ancestor of the D. shardana sp.n. / D. erythrina clade followed by back colonisation of the mainland.Regardless of the scenario, time estimates clearly suggest that the split of D. shardana sp.n. from mainland relatives post-dates the opening of the western Mediterranean basin.Time estimates of the split of D. shardana sp.n. and of the diversifi cation of the D. erythrina / D. lantosquensis clade are close enough to the Messinian Salinity Crisis time period to suggest that this geological event may have played a role in the origin of these lineages, as has also been proposed for the harvestmen genus Trogulus (Schönhofer and Martens 2008).Confi dence intervals of the time estimates, however, are too large to rule out other alternative scenarios.Moreover, it has been shown based on mitochondrial substitution rates that trans-Gibraltar lineage divergences in the scorpion Buthus occitanus predate the opening of the strait of Gibraltar (Gantenbein and Largiader 2003), which opens the door to the possibility that our calibration point is too young and hence the time of colonization of Sardinia is an underestimation.
At this stage, it is not clear what factors may have prompted the speciation of the two new Sardinian endemics.Th e two species show allopatric distribution, generally associated to middle (500 m) and high elevations (1000 m).D. jana sp.n. seems to have a narrow distribution, restricted to the Supramonte region, while D. shardana sp.n. is widespread across most of the islands, except for the northern and southernmost areas (Mt.Limbara and Sette Fratelli, respectively).Interestingly, species of the cave-dwelling beetle lineage Ovobathysciola / Patriziella show a very similar geographic distribution pattern to D. jana sp.n. (O.graffi ti is endemic to the Supramonte region) and D. shardana sp.n. (the rest of the Ovobathysciola / Patriziella species are found in Mt.Arci, Mt.Ferru and Gennargentu), although with contrasting levels of species diversity and phylogenetic relationships (compare fi g. 2 in Caccone and Sbordoni 2001 with our Fig.3).Genetic analysis of several D. shardana sp.n. populations, however, revealed strong phylogeographic patterns in this species.Populations from Mt. Arci form an independent network (i.e., highly divergent haplotypes) from those sampled from Mt. Ferru and Gennargentu, which are also separated (the southernmost known populations on Sulcis-Iglesiente could not be analysed genetically).Th is observation does not necessarily imply that the former haplotype groups should be considered independent evolutionary lineages, but it does suggest that there are environmental factors (e.g., unsuitable lowland habitat due to warmer climatic conditions) restricting gene fl ow across geographic areas separated by just 50 km.Environmental restrictions to gene fl ow could eventually lead to speciation and hence explain the presence of two sibling species of the erythrina group on Sardinia.

Conclusion
Two new species of the woodlouse-hunter spider genus Dysdera are described from Sardinia.Phylogenetic evidence and morphological features suggest that they are closely related to the type species of the genus D. erythrina.Time estimates reject an Oligocene origin of the new Dysdera species and instead suggest that the Messinian Salinity Crisis shaped the diversifi cation of Sardinian endemics and their closest relatives.Deep genetic divergences and population structure in Dysdera shardana sp.n. identifi es restriction to gene fl ow due to environmental factors as the most plausible explanation for local speciation events.

Figure 2 .
Figure 2. Chronogram of the species sampled in the present study, based on the preferred Bayesian tree obtained by simultaneous analyses of the cox1, 16S, nad1, 28S and H3 partitions.Bars on branches denote support as follows: anterior bar refers to posterior probability, middle bar to maximum likelihood bootstrap support, posterior bar to parsimony jackknife support.Black bar: posterior probability >0.95 or ML bootstrap, Parismony jackknife > 70, grey bar: posterior probability <0.95 or ML bootstrap, Parsimony jackknife <70, white bar: this particular clade was not recovered in the analyses.Numbers on branches are the age of the clade in millions of years before the present, estimated using penalized likelihood (set penalty=add, smoothing=3200), number in brackets refer to bootstrap confi dence intervals.

Figure 3 .
Figure 3. Haplotype network constructed using statistical parsimony (95%).Dots on branches refer to missing haplotypes.Node size proportional to number of haplotypes (1 or 2), nodes colour code and numbers inside refer to localities in Fig. 1.

Table 1 .
Species included in the phylogenetic analysis and GenBank accession numbers for the gene sequences.
) T slightly shorter than DD; external, internal distal border sloped backwards.DD bent about 45 o in lateral view; internal distal border not expanded.IS, ES equally developed; IS truncated at DD middle part.DD tip (Figs 19-21) straight in lateral view; AR present, parallel to DD; fi nger-like distal projection at AR retrolateral side, shorter than AR length.C absent.AC absent.LF absent.L reduced to distal part.LA absent.F absent.AL absent.P (Fig. 22) fused to T; sloped forming an angle of about 135° to T in lateral view; lateral length about 1/4 of T width; ridge present, parallel to T; not expanded, upper margin smooth; not distally projected; back margin slightly folded towards internal side.Paratype female num.CRBA 1246.Figs 12-13.All characters as in male except: Carapace 3.19 mm long; maximum width 2.55 mm; minimum width 1.68 mm.AME diameter 0.18 mm; PLE 0.19 mm; PME 0.16 mm; AME slightly back from frontal border, separated from one another by about 2/3 diameter, touching PLE; PME very close to each other, less than 1/4 PME diameter from PLE. Sternum very slightly wrinkled, mainly between legs and frontal border.Chelicerae 1.38 mm long; fang 1.48 mm; basal segment proximal dorsal side scantly covered with piligerous granulations.Chelicera armed with three teeth and lamina at base; M>B=D.Legs orange.Lengths of female described above: fe1 2.