Description of a new species of Loxosceles Heineken & Lowe (Araneae, Sicariidae) recluse spiders from Hidalgo, Mexico, under integrative taxonomy: morphological and DNA barcoding data (CO1 + ITS2)

Based on an integrative taxonomic approach, a new species of the genus Loxosceles Heineken & Lowe, 1832, is described from the state of Hidalgo, Mexico. Loxosceles tolantongo sp. nov. is described based on DNA barcoding using cytochrome c oxidase subunit 1 (CO1) and internal transcribed spacer 2 (ITS2), and morphology. For species delimitation, four molecular methods were implemented: 1) corrected p-distances under neighbor joining (NJ); 2) automatic barcode gap discovery (ABGD); 3) general mixed yule coalescent model (GMYC) and 4) Bayesian Poisson tree processes (bPTP). The new species morphologically resembles L. jaca, another species from Hidalgo, but there are morphological differences mainly in the tibiae of the male palp, the seminal receptacles of the females and also the high genetic p-distances. CO1 was more informative than ITS2 for the genetic separation; however, both concatenated genes (CO1 + ITS2) present robust evidence for species delimitation. Loxosceles tolantongo sp. nov. is considered a unique species for four reasons: 1) it can be diagnosed and distinguished by morphological characters (of the male palps mainly, but also of the seminal receptacles of the females); 2) the genetic p-distances with CO1 were high (>10%); 3) the molecular species delimitation methods were congruent under CO1 and CO1 + ITS2; and 4) under CO1 and CO1 + ITS2, the new species is a putative sister group of L. jaca + L. tenango. European Journal of Taxonomy 704: 1–30 (2020) 2

Historically, the taxonomy of spiders has been based mainly on traditional morphology, using genitalic characters, such as male palps and seminal receptacles in females. Sexual characters in spiders are robust and important morphological characters that are still used to separate species and to provide a diagnosis. This means that genitalia evolves, on average, more rapidly than non-genital morphological traits (Huber 2003;Huber & Dimitrov 2014). Furthermore, somatic characters are useful as additional evidence to separate species in some groups of spiders. Thus coloration, color pattern, body proportions and even extreme size differences are useful traits for species separation (Huber et al. 2005;Huber & Dimitrov 2014;Valdez-Mondragón et al. 2019). Traditional morphology alone cannot determine species boundaries in some cases due to the intraspecific variation in sexual structures, and the genus Loxosceles is a good example (Brignoli 1968;Gertsch & Ennik 1983;Valdez-Mondragón et al. 2019). However, the male palps of Loxosceles remain a good character for species identification because of the little morphological variation in comparison with the seminal receptacles of the females, as was demonstrated recently in the description of L. malintzi and L. tenochtitlan by Valdez-Mondragón et al. (2018b, 2019. In spider groups with complicated morphology, the use of molecular markers helps in the discovery of undescribed diversity and species delimitation, as was demonstrated, for example, by Planas & Ribera (2015) with the genus Loxosceles. Although DNA barcodes are being applied in systematics as a useful tool to resolve species delimitation problems, modern taxonomy includes many different sources of evidence, such as traditional morphology, ecology, reproduction and biogeography (integrative taxonomy). Recently described L. tenochtitlan by Valdez-Mondragón et al. (2019) represents the first species of the genus described using an integrative taxonomic approach with multiple lines of evidence, including traditional morphology, geometric and lineal morphology, molecular markers (CO1 and ITS2), and even the biogeographic methods of ecological niche modeling.
In this study, a new species of Loxosceles from the state of Hidalgo is described based on morphological and molecular evidence using an integrative taxonomic approach.

Biological material
Specimens were hand-collected and preserved in ethanol (80%) for morphology and ethanol (96%) for molecular studies. The type specimens and the additional examined material are deposited in the following repositories: Abbreviations AME = anterior median eyes PLE = posterior lateral eyes PME = posterior median eyes

Taxon sampling
The molecular analyses presented here are based on a total of 50 individuals from eleven species, including the new species described here and two outgroups to root the trees: Loxosceles rufescens (Dufour, 1820) and Scytodes thoracica (Latreille, 1802) (Table 1). Three different partitions were used (CO1: 674 bp, ITS2: 435 bp and CO1 + ITS2: 1109 bp).

DNA extraction, amplification and sequencing
Specimens for DNA extraction were preserved in ethanol (96%) and kept at -20°C. DNA was isolated from legs, prosoma, or complete specimens in the case of immatures. DNA extractions were done using a Qiagen DNeasy Tissue Kit following the protocol by Valdez-Mondragón et al. (2019). DNA fragments included approximately 650 bp of the cytochrome c oxidase subunit 1 (CO1) mitochondrial gene and 435 bp of the Internal Transcribed Spacer 2 (ITS2) nuclear gene. The fragments were amplified  (annealing), 1 min at 72°C (elongation); and final elongation 5 min at 72°C. PCR program for ITS2 was as follows: initial step 3 min at 94°C; amplification 40 cycles of 30 s at 94°C (denaturation), 1 min at 53°C (annealing), 1 min at 72°C (elongation); and final elongation 5 min at 72°C. PCR products were checked to analyze length and purity on 1% agarose gels with a marker of 100 bp and purified directly using the QIAquick PCR Purification kit of QIAGEN. DNA extraction and amplification were performed at the Molecular Laboratory at the Laboratorio Regional de Biodiversidad y Cultivo de Tejidos Vegetales (LBCTV), Institute of Biology, Universidad Nacional Autónoma de Mexico (UNAM), Tlaxcala City. Sequencing was performed at the Molecular Laboratory in the Institute of Biology, UNAM, Mexico City. Sequencing of both strands (5′-3′and 3′-5′) of PCR products were performed in a Sequencer Genetic Analyzer RUO Applied Biosystems Hitachi model 3750xL. Sequence data of CO1 and ITS2 are deposited in GenBank with accession numbers: MK936272-MK936303 and MN512427-MN512432 for CO1 and MK957192-MK957225 and MN525280-MN525286 for ITS2 (Table 1).

DNA sequence alignment and editing
Sequences were edited with the programs BioEdit ver. 7.0.5.3 (Hall 1999) and Geneious ver. 10.2.3 (Kearse et al. 2012). Sequences were aligned online using the default gap opening penalty of 1.53 in MAFFT (Multiple Sequence Alignment based on Fast Fourier Transform) ver. 7 (Katoh & Toh 2008) using the following alignment strategy: Auto (FFT-NS-2, FFTNS-i or L-INS-i, depending on data size). These aligned matrices were subsequently used in analyses.

Starting trees under Maximum Likelihood (ML) and Bayesian Inference (BI)
The approaches for DNA barcoding tree-based delimitation explicitly use the phylogenetic species concept. An input starting tree generated with ML using MEGA ver. 7.0 and BI performed by MrBayes ver. 3.1.2 (Ronquist & Huelsenbeck 2003) were implemented, and the analysis recognized monophyletic clusters by searching differential intra-and interspecific branching patterns (Ortiz & Francke 2016). The ML analysis was calculated with the parameters for CO1 and ITS2: Number of replicates = 1000, bootstrap support values = 1000 (significant values ≥ 50%), Models of sequence evolution selected using jModelTest = GTR, Rates among sites = G + I, No. of discrete Gamma Categories = 6, Gaps Data Treatment = Complete deletion, Select Codon Position = 1 st + 2 nd + 3 rd + Noncoding Sites, ML Heuristic Method = Subtree-Pruning-Regrafting -Extensive (SPR level 5), Initial Tree for ML = Make initial tree automatically (Default -NJ/BioNJ). The BI analyses were run with four parallel Markov chains with the following parameters: MCMC (Markov Chain Monte Carlo) generations = 20 000 000, sampling frequency = 1000, print frequency = 1000, number of runs = 2, number of chains = 4, MCMC burnin = 2500, sumt burn-in = 2500, sump burn-in = 2500, Models of sequence evolution selected using jModelTest = GTR, Rates among sites = G + I, Select Codon Position = 1 st , 2 nd and 3 rd . TRACER ver. 1.6 (Rambaut & Drummond 2014) was used to analyze the parameters and the effective sample size (ESS) of the MCMC to ensure the runs converged. FigTree ver. 1.4.3 was used to visualize the topology of the tree with the posterior probability values (PP) at nodes. The models of sequence evolution were selected using the Akaike information criterion (AIC) in jModelTest ver. 2.1.10 (Posada & Buckley 2004). The models selected for CO1 and ITS2 for each partition block were: GTR + G + I (1 st and 2 nd codon positions) and GTR + G (3 rd position). The model selected for ITS2 was GTR + G.

Automatic barcode gap discovery (ABGD)
The ABGD species delimitation method uses recursive partitioning with a range of prior intraspecific divergence and relative gap widths, estimating the threshold between intra-and interspecific genetic variation, generating species-level groupings (Ortiz & Francke 2016). ABGD analyses were conducted using both uncorrected and K2P distance matrices with default options: Pmin = 0.001, Pmax = 0.1, Steps = 10, Relative gap width (X) = 1, Nb bins = 20.

General mixed yule coalescent model (GMYC)
The GMYC species delimitation method applies single (Pons et al. 2006) or multiple (Monaghan et al. 2009) time thresholds to delimit species in a Maximum Likelihood context, using ultrametric trees (Ortiz & Francke 2016). Phylogenetic analyses were run in BEAST ver. 2.6.0  using a coalescent (constant population) tree prior. Independent lognormal relaxed clock was applied to each partition, for analyses 20×10 6 generations were run. Convergence was assessed with TRACER ver. 1.6 (Rambaut & Drummond 2014). TREEANNOTATOR ver. 2.6.0 (BEAST package) was used to build maximum clade credibility trees, after discarding the first 25% of generations by burn-in. Following gene tree inference, GMYC was implemented in the web interface for single and multiple threshold GMYC (https://species.h-its.org/gmyc/). The backend of this web server runs the original R implementation of the GMYC model authored by Fujisawa & Barraclough (2013). A single threshold was used for the concatenated matrix.

Bayesian Poisson tree processes (bPTP)
The PTP species delimitation method (Zhang et al. 2013) is similar to GMYC, but uses substitution calibrated (not ultrametric) trees to avoid the potential flaws in constructing time calibrated phylogenies (Zhang et al. 2013;Ortiz & Francke 2016). We employed the Bayesian variant of the method (bPTP) on the online version (https://species.h-its.org/ptp/). It was run on the Bayesian gene trees with default options: rooted tree, MCMC generations = 100 000, Thinning = 100, Burn-in = 0.1, Seed = 123. Haplotypes network for CO1 was constructed to visualize the mutations among haplotypes of species using the TCS algorithm (Clement et al. 2000) in PopArt ver. 1.7 (Leigh & Bryant 2015). The trees generated were edited using Adobe Photoshop CS6.

Etymology
The species name is a noun in apposition and refers to the type locality of the new species: the Tourist Center Grutas de Tolantongo, located in the Mezquital Valley, Hidalgo, Mexico.

Males
Males collected from the Tourist Center Grutas de Tolantongo and from 500 m west of entrance No. 5 are pale brown, with legs the same color as carapace; in some cases the legs are darker than the body. The male holotype is darker brown that all other specimens. The male collected from ʻEl Paraisoʼ trail is dark  The male palps present a little variation. In the case of the holotype the palp has more marked coloration and the embolus is wider at the base (Figs 20-25, 30-33). The seminal receptacles of females are S-shaped, asymmetrical and broadly variable in shape, even in specimens from the same locality (Figs 36-39). The apical lobes are rounded in some specimens, oval in others, or even with a sharp tip (Figs 28,(36)(37)(38)(39). The base of the seminal receptacles is variable, in some specimens it is wider, more rounded and more sclerotized than in others (Figs 28,(36)(37)(38)(39).

Natural history
The specimens of L. tolantongo sp. nov. were collected in a xerophytic forest, the native vegetation type around the Touristic Center Grutas de Tolantongo (Figs 10-15). The microhabitat where the specimens were collected was under big rocks, and from inside rotten and dry agave plants (Figs 12-15). Some specimens were collected close to each other. (Figs 50-51).

Molecular analyses and species delimitation
The analyzed matrices include 49 individuals of ten species of Loxosceles, 38 individuals for the CO1 data set and 42 individuals for ITS2 (Table 1,  . Specimens used in this study, GenBank accession numbers and localities of the specimens are listed in Table 1. The average genetic p-distance among analyzed species was 15.9% for CO1 and 8.4% for ITS2 (Figs 52-53). Corrected p-distances from the CO1 and ITS2 data recovered nine species of Loxosceles, both with high bootstrap support values (Figs 52-53). Based on the genetic analyses, L. tolantongo sp. nov. is closely related to L. jaca + L. tenango with CO1 (Fig. 52), and with L. jaca with ITS2 (Fig. 53); with an average p-distance between L. tenango and L. tolantongo sp. nov. of 9.6%, and between L. jaca and L. tolantongo sp. nov. of 10.5% for CO1 and 1.8% for ITS2 (Tables 3-4). Molecular analyses with CO1 and the concatenated matrix (CO1 + ITS2) (Figs 54, 56) indicate that four different species delimitation methods, including the morphology, were congruent to delimit L. tolantongo sp. nov. as a valid and different species, and recovered a total of nine species. However, using the ABGD species delimitation method under recursive partitions (RP), 14, 11 and 10 species were recovered (Fig. 54). The molecular analyses are consistent with the morphology; the nine species of Loxosceles used in this study were previously described only with morphological characters, except for L. tenochtitlan which was described with different lines of evidence (morphological and molecular) by Valdez-Mondragón et al. (2019). Molecular analyses with ITS2 indicate that of the four different molecular species delimitation methods, only NJ and GMYC (Yule) were consistent with the morphology for delimitation of L. tolantongo sp. nov. (Figs 53, 55). ABGD and bPTP species delimitation methods do not recover L. totaltongo as a different species, but they recover other species, such as L. zapoteca Gertsch, 1958, L. colima Gertsch, 1958 (except with initial partitions (IP) (ABGD)) and L. malintzi (Fig. 55). The haplotype network analysis with CO1 data is concordant with the results of the different species delimitation analyses (Fig. 57). There were > 10 mutations between haplotypes under CO1 for all the species (Fig. 57). Regarding L. tolantongo sp. nov. and L. jaca + L. tenango, the haplotype network was concordant with the species delimitation, and showed a total of 39 and 36 mutations respectively between haplotypes under CO1 (Fig. 57).

Discussion
There are currently two important tasks to which DNA barcodes markers are being applied in modern systematics. The first is distinguishing between species (equivalent to species identification or species diagnosis) and the second is the use of DNA data to discover new species (equivalent to species delimitation and species description) (DeSalle et al. 2005). For some groups of organisms, such as spiders, traditional morphology in some cases fails to determine species boundaries, and identifying morphologically inseparable cryptic or sibling species requires a new set of taxonomic tools, including the analysis of molecular data (Jarman & Elliott 2000;Witt & Hebert 2000;Proudlove & Wood 2003;Hebert et al. 2003Hebert et al. , 2004Bickford et al. 2007;Hamilton et al. 2011Hamilton et al. , 2014Hamilton et al. , 2016  In recent studies, molecular evidence has suggested that the known diversity within the genus Loxosceles could be highly underestimated (Binford et al. 2008;Duncan et al. 2010;Planas & Ribera 2014, 2015Tahami et al. 2017;Valdez-Mondragón et al. 2019). One important factor leading to the underestimation is widespread intraspecific variation in sexual structures, mainly in the seminal receptacles of the females (Figs 36-43), as was noted in this work and previously by Brignoli (1968), Gertsch & Ennik (1983), and recently by Valdez-Mondragón et al. (2018b, 2019 in the case of the species from Mexico. However, sexual structures such as the male palps remain a good character for species identification because of their little morphological variation in comparison with the seminal receptacles of the females (Valdez-Mondragón et al. 2018a, 2018b, 2019. It is important to consider that male palps may be remarkably similar, as indicated for L. tolantongo sp. nov and L. jaca (which showed slight differences), which can also make species identifications difficult, thus contributing to such underestimations. Although L. tolantongo sp. nov. morphologically resembles L. jaca, the new species could be diagnosed based on the sexual characters of the male palps  and the seminal receptacles of the females (Figs 28,(36)(37)(38)(39). Furthermore, molecular markers provided additional evidence to delimit the new species within an integrative taxonomic context. Identifying morphologically inseparable cryptic or sibling species  Table 3. Genetic p-distance matrix from the CO1 data among Loxosceles tolantongo sp. nov., L. tenango Gertsch, 1973 andL. jaca Gertsch &Ennik, 1983. Average p-distance = 6.7%.  Table 4. Genetic p-distance matrix from the ITS2 data between Loxosceles tolantongo sp. nov. and L. jaca Gertsch & Ennik, 1983. Average p-distance = 1.8%.

Species
requires a new set of taxonomic tools, including DNA and additional sources of evidence (integrative taxonomy) (Jarman & Elliott 2000;Witt & Hebert 2000;DeSalle et al. 2005;Hebert et al. 2003Hebert et al. , 2004Bickford et al. 2007;Hamilton et al. 2011Hamilton et al. , 2014Hamilton et al. , 2016Ortiz & Francke 2016, Valdez-Mondragón et al. 2018a, 2018b, 2019. As was mentioned by Hamilton et al. (2011), we can rarely delimit species based only on molecular data, and that is why, in addition to the molecular evidence, the geographical data provides information to delimit these species. Loxosceles jaca and L. tolantongo sp. nov. were found in different localities, Cardonal and Jacala, Hidalgo, respectively. The type locality of the new species is located in the Tolantongo Canyon, 500 m deep, where the Tolantongo river is located, which can serve as a geographic barrier for Loxosceles spiders that usually have limited dispersions (Sandidge & Hopwood 2005;Binford et al. 2008;Vetter 2008Vetter , 2015Foelix 2011 Planas & Ribera (2014, 2015 found genetic distances between species of Loxosceles from the Canary Islands to be > 12% using CO1, whereas Tahami et al. (2017) found genetic distances between species from the Middle East ranging from 17.5 to 20.6% for CO1. In this work, the genetic distances were over 15% between the different species. They also reported average low genetical distances inside the same species (0.26% and less than 1%) and in this work the intraspecific genetic distances are low (< 1%) (Tables 3-4; Fig. 52). The corrected genetic p-distances showed that CO1 performed better for species delimitation than ITS2 (Figs. 52-55). Agnarsson (2010) concluded that ITS2 was insufficient to solve the relationships between closely related species, but also indicated that ITS2 is an easily amplified and sequenced marker for use in spider phylogeny despite its limits at lower taxonomic levels. The uses of the ABGD species delimitation method are recursive partitioning, with a range of prior intraspecific divergence and relative gap width to estimate the threshold between intra-and interspecific variation which intends to produce species-level groupings (Ortiz & Francke 2016). However, this species delimitation method is sensitive to sampling effect, as was demonstrated in the spider genera Aphonopelma Pocock, 1901 and Bonnetina Vol, 2000 from North America by Hamilton et al. (2014) and Ortiz & Francke (2016), respectively, tending to give moderately over-splitting solutions as was the case of our work (Fig. 55). PTP is similar to GMYC, but uses substitution calibrated (not ultrametric) Numbers above bars represent the delimitation methods: 1 = morphology (M); 2 = neighbor joining (NJ); 3 = ABGD with initial partitions (IP); 4-5 = ABGD with recursive partitions (RP); 6 = GMYC yule analysis; 7 = GMYC coalescent analysis; 8 bPTP with ML; 9 = bPTP with IB. Numbers below bars represent species recovered for each delimitation method. Red numbers correspond to Bayesian posterior probabilities, black numbers are bootstrap support values from the ML analysis.
trees to avoid the potential flaws in constructing time-calibrated phylogenies, as was proposed by Zhang et al. (2013).
The work of Cao et al. (2016) with the spider genus Pseudopoda Jäger, 2000 (Sparassidae Bertkau,1872) suggested using concatenated fragments CO1 + ITS2, which identifies and diagnoses species in a more appropriate way than the separated mitochondrial barcodes. In this work, the concatenated matrix showed that all species delimitation methods were congruent with the morphology, recovering L. tolantongo sp. nov. as a separated species (Figs 54-56). Carstens et al. (2013) suggested that researchers should apply a wide range of species delimitation analyses to their data and place their trust in delimitations that are congruent across methods. Using several species delimitation methods, incongruence across the different results is evidence of a difference in the power to detect cryptic lineages across one or more of the approaches used to delimit species and could indicate that assumptions of one or more of the methods have been violated. In these cases the assumptions for species delimitations should be conservative (Carstens et al. 2013). In this work, the different molecular species delimitation methods and their variants were congruent and consistent to separate L. tolantongo sp. nov. of L. jaca + L. tenango. Loxosceles tolantongo sp. nov. is considered a unique species for four reasons: 1) it can be diagnosed and distinguished by morphological characters (male palps mainly); 2) the high genetic p-distances with CO1 (> 10%); 3) the four different molecular species delimitation methods were congruent under CO1 and CO1 + ITS2 (concatenated matrix); 4) under CO1 and CO1 + ITS2, L. tolantongo sp. nov. is a putative sister group of L. jaca + L. tenango.
Although the new species described herein has a small range of occurrence, it is considered as not threatened. The Tourist Center Grutas de Tolantongo, which belongs to the Socieded Cooperativa Ejidal Grutas Tolantongo (Cooperative Ejido Society of the Tolantongo Grottos), has managed to preserve much of the resort and the ejido land around it in its natural state, preserving many native species distributed in the region.