Morphological and molecular evidence support the taxonomic separation of the medically important Neotropical spiders Phoneutria depilata (Strand, 1909) and P. boliviensis (F.O. Pickard-Cambridge, 1897) (Araneae, Ctenidae)

Abstract The species of the genus Phoneutria (Ctenidae), also called banana spiders, are considered amongst the most venomous spiders in the world. In this study we revalidate P. depilata (Strand, 1909), which had been synonymized with P. boliviensisis (F.O. Pickard-Cambridge, 1897), using morphological and nucleotide sequence data (COI and ITS-2) together with species delimitation methods. We synonymized Ctenus peregrinoides, Strand, 1910 and Phoneutria colombiana Schmidt, 1956 with P. depilata. Furthermore, we designated Ctenus signativenter Strand, 1910 as a nomen dubium because the exact identity of this species cannot be ascertained with immature specimens, but we note that the type locality suggests that the C. signativenter syntypes belong to P. depilata. We also provide species distribution models for both species of Phoneutria and test hypotheses of niche conservatism under an allopatric speciation model. Our phylogenetic analyses support the monophyly of the genus Phoneutria and recover P. boliviensis and P. depilata as sister species, although with low nodal support. In addition, the tree-based species delimitation methods also supported the separate identities of these two species. Phoneutria boliviensis and P. depilata present allopatric distributions separated by the Andean mountain system. Species distribution models indicate lowland tropical rain forest ecosystems as the most suitable habitat for these two Phoneutria species. In addition, we demonstrate the value of citizen science platforms like iNaturalist in improving species distribution knowledge based on occurrence records. Phoneutria depilata and P. boliviensis present niche conservatism following the expected neutral model of allopatric speciation. The compiled occurrence records and distribution maps for these two species, together with the morphological diagnosis of both species, will help to identify risk areas of accidental bites and assist health professionals to determine the identity of the species involved in bites, especially for P. depilata.

species. In addition, we demonstrate the value of citizen science platforms like iNaturalist in improving species distribution knowledge based on occurrence records. Phoneutria depilata and P. boliviensis present niche conservatism following the expected neutral model of allopatric speciation. The compiled occurrence records and distribution maps for these two species, together with the morphological diagnosis of both species, will help to identify risk areas of accidental bites and assist health professionals to determine the identity of the species involved in bites, especially for P. depilata.

Keywords
Andes, Maxent, niche conservatism, Phylogenetics, species delimitation introduction The species of the genus Phoneutria are considered aggressive and amongst the most venomous spiders in the world (Foelix 2010). Currently this genus includes eight large (17-48 mm) nocturnal species that are widely distributed in Central America and South America (Simó and Brescovit 2001;Martins and Bertani 2007). Their venom has a neurotoxic action and many researchers have analyzed its components and the epidemiology of the bites of these species (Gomez et al. 2002;Richardson et al. 2006;Bucaretchi et al. 2008Bucaretchi et al. , 2018Garcia et al. 2008). Most of the clinically relevant bites by this genus are caused by P. nigriventer (Keyserling, 1891) and occur in Brazil (around 4,000 cases per year), with only 0.5% being severe (Bucaretchi et al. 2018).
Phoneutria boliviensis (Pickard-Cambridge, 1897) is a widespread species distributed from Central America (Costa Rica) to central South America (Bolivia), found across many types of ecosystems and geographical barriers that commonly divide other taxa (e.g. the Andes mountain system that separates many cis and trans Andean lowland lineages) (Bartoleti et al. 2018;Salgado-Roa et al. 2018). This species was originally described from the "Madre de Dios" Amazonian region in Bolivia and only the male palp was illustrated because the epigynum of the single female was damaged (Schiapelli et al. 1973). Schiapelli et al. (1973) illustrated the epigynum and the male palp based on other specimens identified by F.O. Pickard-Cambridge as P. boliviensis (a female from Ecuador and a male from Charaplaga, Bolivia). These authors report that male specimen at The Natural History Museum (at their time known as the British Museum of Natural History) in the vial with the syntypes of Ctenus boliviensis was a specimen in good condition of P. nigriventer (Keyserling, 1891). Subsequently, Simó and Brescovit (2001) indicated that they were not able to find the type specimens, and therefore they considered them lost.
At that time, P. boliviensis was known only to occur in the Amazon region, until Valerio (1983) reported this species in Costa Rica. Later, Simó and Brescovit (2001) revised the genus and synonymized several ctenids with P. boliviensis. Simó and Brescovit (2001) acknowledged the large morphological variation of P. boliviensis across its distribution range, but they interpreted this variation as intraspecific and diagnosed it by the truncated apex of the male retrolateral tibial apophysis.
During field work in Costa Rica, Panama, Colombia, Ecuador and Peru, and careful examination of museum specimens from these countries, we realize that P. boliviensis can be separated into two distinct species. One trans-Andean species, Phoneutria depilata and the true P. boliviensis (cis-Andean) endemic of the Amazon region. Therefore, in this study we revalidated P. depilata that was synonymized with P. bolivienesis by Simo and Brescovit (2001) and we designate a neotype of P. boliviensis collected from the Madre de Dios region of Peru. We follow an integrative taxonomic approach using molecular, morphological, and ecological data to support the separation of these two species. We also provided species distribution models (SDMs) for both species of Phoneutria. Furthermore, we also tested the hypothesis of niche conservatism under an allopatric speciation model (Wiens 2004;Wiens et al. 2011). This hypothesis states that the tendency of lineages to maintain their ancestral ecological niche, and their failure to colonize and adapt to new environments, separate ancestral taxa promoting speciation (Wiens 2004). Therefore, we expect that P. depilata and P. boliviensis, separated by the Andean mountain, present niche conservatism. Phoneutria depilata has been deeply studied in the literature as P. boliviensis, in works regarding its venom composition and toxicity (Estrada-Gomez et al. 2015;Valenzuela-Rojas et al. 2019), natural history (Hazzi 2014;Valenzuela-Rojas et al. 2020), geographic distribution (Valerio 1983;Hazzi et al. 2013), bite accidents to humans (Trejos et al. 1971;Florez et al. 2003) and introductions to Europe through banana shipments (Cathrine and Longhorn 2017;Rozwałka et al. 2017). Unlike P. depilata, except for brief field anecdotal mentions (Torres-Sánchez and Gasnier 2010), there is no such information for P. boliviensis. We have also provided additional information on the natural history of both species.

Museum abbreviations
The material examined and/or collected belongs to the following museums:

Morphological examination and description of species
Specimens were preserved in 95% ethanol. Descriptions and terminology follows Simó and Brescovit (2001) and Martins and Bertani (2007). All measurements were taken in millimeters using the application of LAS in a Leica M205A stereomicroscope. Epigyna were digested with pancreatin solution (Álvarez-Padilla and Hormiga 2007) to enable study of internal structures. Digital images were taken with a Leica DFC425 camera on a Leica M205A stereomicroscope. Extended focal range images were composed using the software package Helicon Focus (version 6.7.1; www.heliconsoft.com) from Helicon Soft Ltd. The SEM images were taken using a LEO 1430VP scanning electron microscope at the Department of Biology of The George Washington University. For scanning electron microscope preparation, structures were cleaned ultrasonically, transferred to 95% and then to 100% ethanol for 10 min in each immersion before being critically-point-dried. The following abbreviations are used: C = conductor, CD = copulatory duct, E = embolus, ELA = epigynal lateral apophysis, ELF = epigynal lateral field, ELG = epigynal lateral guide, EMF = epigynal middle field, FD = fertilization duct, IB = internal bulge of the embolus, LP = lateral projection, MA = median apophysis, RTA = retrolateral tibial apophysis; S = spermatheca, S = subtegulum.

DNA-based analysis
Sampling design. Due to the widespread climatic niche of P. depilata, we sequenced seven specimens from Costa Rica, Ecuador and Panama that were collected from mountain to lowland areas, and from dry to rain forests ecosystems (Table 1). For P. boliviensis, we sequenced six specimens collected in three localities distributed from the north through to the southern part of the Peruvian Amazon, including one specimen from the type locality. In addition, we sequenced a specimen of Phoneutria fera Perty, 1833 collected in the Amazon of Ecuador and added two more sequences of the same species from GenBank (HM575999 and KY017637). As an outgroup, we sequenced one specimen of Spinoctenus escalerete Hazzi et al., 2018, Ctenus datus Strand, 1909, C. aff. amphora Mello-Leitão, 1930and Kiekie curvipes (Keyserling, 1881. In addition, we also added a sequence of Ctenus crulsi Mello-Leitão, 1930, from Gen-Bank (KY017633.1).
Specimens preserved in 95% ethanol were used for DNA extraction using the Qiagen DNEasy kit. Coxae and femora were used for extractions and the rest of the specimen was preserved as a voucher. Two gene fragments frequently used for species recognition and delimitation in spiders (e.g., Montes de Oca et al. 2016;Ballesteros and Hormiga 2018;Salgado-Roa et al. 2018) were amplified for analysis: the mitochondrial cytochrome c oxidase subunit I (~650 bp, COI) and the nuclear internal transcriber subunit 2 (~300 bp, ITS2). The former was amplified using the primers LCO1490 and HCOout (Folmer et al. 1994;Carpenter and Wheeler 1999) and ITS2 was amplified with the primers FITS and RITS (White et al. 1990;Agnarsson 2010) using the conditions previously reported in Ballesteros and Hormiga (2018).
Amplified products were sent to Macrogen USA (Rockville, Maryland) for sequencing. Contigs were formed using GENEIOUS 6.0.6 (http://www.geneious.com; Kearse et al. 2012) and COI sequences were checked for stop codon position, then queried against NCBI BLAST nucleotide database to check for contamination. Multiple sequence alignments were completed using the Q-INS-I search strategy using MAFFT. Gaps were treated as missing data for the phylogenetic analysis.
The best partitioning scheme and substitution models were explored using Parti-tionFinder 2.1.1 using the ''greedy" search strategy and the correction of the Akaike information criterion (AICc). Four partition schemes were used as input data: first, second and third codon position for COI, and ITS-2 as a whole. Phylogenetic analyses were performed using parsimony (MP), maximum likelihood (ML) and Bayesian inferences (BI). The parsimony analyses were carried out in TNT v. 1.5 (Goloboff et al. 2008;Goloboff and Catalano 2016) using 100 random addition sequences followed by TBR branch swapping algorithm and retaining 10 trees per replicate. Branch support was assessed using 1000 replicates of jackknife resampling (Farris et al. 1996). The Bayesian analyses were performed in MrBayes 3.2.6 (Ronquist and Huelsenbeck 2003) running 20 million generations from four Markov Chain Monte Carlo chains (MCMC). Trees and parameters were sampled every 1000 generations, 25% of the trees were discarded as burn-in and the remainder were used to calculate posterior probabilities. To check that the run was long enough for the chains to converge, the probabilities of the marginal parameters were observed in Tracer v. 1.5 (Rambaut et al. 2014b). The maximum likelihood analyses were performed with the package IQ-TREE 1.4.2 (Nguyen et al. 2015) and ultrafast bootstrap (UFBoot) were used as support measure (Minh et al. 2013).
To measure relationships between haplotypes, we constructed haplotype medianjoining networks for each marker using PopArt v1.7 (Leigh and Bryant 2015). Due to the small genetic variation found in the allele network of the nDNA, we only calculated genetic distances for the mDNA. Uncorrected genetic distances (uncorrected p-distance) were calculated within and among Phoneutria species pairs using MEGA v.10 (Kumar et al. 2018). We performed both genetic distance and tree-based species delimitation methods in order to distinguish species of Phoneutria. The Automatic Barcoding Gap Discovery (ABGD) method (Puillandre et al. 2012) was used to identify breaks between the intraspecific and interspecific diversity (this is known as the barcode gap). This method relies on just pairwise genetic distances and therefore does not used phylogenetic information. Because ABGD was designed for single locus analysis, we only used this method with the COI sequences data. The analysis was performed through the web-server (https://bioinfo.mnhn.fr/abi/public/abgd/abgdweb.html) using default settings and the uncorrected p-distances option.
The three remaining methods used are tree-based. First, we applied the general mixed Yule coalescent model (GMYC, Pons et al. 2006;Fujisawa and Barraclough 2013) using GMYC web server (https://species.h-its.org/gmyc/). This method models the Yule and coalescent processes on an ultrametric tree to determine the transition between intra and interspecific divergences. The ultrametric tree was estimated in BEAST 2.6.0 (Bouckaert et al. 2014) using a coalescent constant population as a tree prior. An uncorrelated relaxed clock with log normal distribution and GTR+Gamma substitution model for each codon was applied. We ran the analysis with 20 million generations of MCMC. Trees and parameters were sampled every 1000 generations, 25% of the trees were discarded as burn-in and the remainder were used to calculate posterior probabilities. To check that the run was long enough for the chains to converge, the probabilities of the marginal parameters were observed in Tracer v.1.5 (Rambaut et al. 2014b). TreeAnnotator version 2.6.0 (BEAST package) was used to build maximum clade credibility trees. For the second method, we applied a Bayesian framework of the multi-rate Poisson tree process (mPTP, Kapli et al. 2017). This approach differs from GMYC in modelling coalescent and speciation events as relative to numbers of substitutions rather than time (Kapli et al. 2017). The minimum branch length was calculated and used as an input together with a likelihood tree (estimated as above). We ran the alignment with 2 independent replicates of MCMC of 5,000,000 generations, sampling every 1000 with a burn-in of 10% of the total length of the chain. GMYC and mPTP were designed to model single locus data, and because ITS-2 market lumped the three morphologically diagnosable species in one, we only show the results with COI.
Finally, we applied the Bayesian Phylogenetics and Phylogeography software (BPP, Yang 2015), a is species delimitation approach based on the Multi Species Coalescent Model. This method uses a Bayesian modelling framework to estimate posterior probabilities of species assignment's multilocus gene trees, considering uncertainties in the coalescent process. We carried out joint species delimitation and species tree estimation (A11 analysis), assigning individuals a priori to a species based on the phylogeny and morphology. For the root age of the tree (τ) and the ancestral population size (θ), four combinations of priors were used. Combinations were among deep divergence times (τ = G(1, 10)) and shallow divergence times (τ = G(2, 2000)), and large populations sizes (θ = G(1, 10)) and small populations sizes (θ = G(2, 2000)). We performed 100,000 iterations, sampling every 2, using the 10% of the chain as burn-in. Because mDNA has a different mutation rate and effective population size than nDNA, we did analysis with mDNA+nDNA and mDNA alone. As mDNA obtained similar results, we only provide the results of the multilocus dataset. Currently, all species delimitation methods differentiate simplifying assumptions on the potential real parameter space relevant to species delimitation. Therefore, any of these assumptions could be violated easily in a particular empirical system, consequently only congruently delimited lineages across the different methods were considered as species (Carstens et al. 2013).

Species model distributions and niche comparisons
We estimated the distribution of P. boliviensis and P. depilata using the Maxent algorithm (Phillips et al. 2006;Elith et al. 2011). We used occurrence records from the literature, fieldwork and museum specimens examined by us (herein after LIFIMU database). In addition, we used iNaturalist (https://www.inaturalist.org/) as a novel procedure in spiders to obtain more distribution records for these species. iNaturalist is a citizen science platform that provides unprecedented access to documenting species diversity and distribution across the world (Hochmair et al. 2020). Users upload media (mostly images) of biological findings to the iNaturalist data portal that are later identified to some taxonomic level by the iNaturalist community. Because in most cases, spiders can only reliably be identified by examining their genitalia under the stereoscope, these new apps that rely on images for species identification have not been used on spiders, to our knowledge. However, for these medically relevant spiders, it is possible to identify them using only images ( Fig. 1A-D). In the case of P. depilata, after extensive fieldwork and the study of museum specimens, we have been able to conclude that this is the only Phoneutria species distributed in the Trans Andean region reaching Central America (Nicaragua). Thus, we can assign with high certainty Phoneutria images from these regions to P. depilata. Phoneutria boliviensis is endemic to the Amazonian region and it co-occurs with two more species of Phoneutria: P. fera and P. reidyi (F. O. Pickard-Cambridge, 1897). However, P. boliviensis is the only species that has two conspicuous lateral white bands in the anterior area of the carapace (Fig. 1C, D, 4A). In addition, males of P. boliviensis have dark black grooves in the carapace (Fig. 1D, 4A). Therefore, Phoneutria images from the Amazon region with these coloration features were identified as P. boliviensis.
To mitigate the impact of uneven sampling in our occurrence data, we applied a distance correction by taking only one point within a radius of 10 km. We obtained 19 bioclimatic predictor layers summarizing annual trends, seasonality and extremes in precipitation and temperature at a spatial resolution of 30 arc-seconds (i.e. 1 km 2 ) from the WorldClim database (Fick and Hijmans 2017). In order to reduce collinearity of the predictor variables, we selected the following variables (Pearson <0.7): annual mean temperature (Bio1), mean diurnal range (Bio2), temperature seasonality (Bio4), annual precipitation (Bio12), precipitation seasonality (Bio15) and precipitation of warmest quarter (Bio18). The modelling area selected for P. depilata was the trans-Andean region until Nicaragua and for P. boliviensis the Amazon and Orinoquia basins (cis-Andean region). We selected these regions considering species accessible area M (diagram by Barve et al. 2011) based on the geographical extension of gathered records of both species and the distribution of terrestrial ecoregions (Olson et al. 2001) and biogeographic regions of endemism (Morrone 2014) in the Neotropical Region.
We ran the models selecting a logistic output and random seed, and the maximum number of background points maintained at 10,000. To assess model performance, we applied k-fold cross validation procedure splitting the occurrences into training and testing records (70% and 30%, respectively), and replicating this process 15 times. Models were evaluated using the Area Under the Curve Metric (AUC) that compares model results with null expectations using a threshold-independent measure. We average the AUC values obtained in the replicates and created confidence intervals values to assess model significance from random model expectations (AUC > 0.5). In order to make the binary distribution maps, habitat suitability values were converted in presence and absence using the 5 th percentile as the threshold value (Liu et al. 2005(Liu et al. , 2013. In addition, areas with high probability of presence, but disjunct from areas where specimens have been recorded, were excluded from the prediction (Helgen et al. 2013).
To test niche conservatism among these two species, we used the niche similarity test (Peterson et al. 1999) and niche equivalency test (Graham et al. 2004) in the R package Ecospat (Di Cola et al. 2017). First, we performed an environmental principal component analysis (PCA-env) (Broennimann et al. 2012), calibrated with the accessible areas of the two species. We then created a grid of 100 × 100 cells over the ordination space, and a kernel density function was applied on the occurrence data in order to estimate Schoener's D index (Schoener 1968) with the first to principal components. This metric estimates niche overlap and D values ranging from zero, when niches do not overlap, and one, when niches completely overlap. Finally, the niche equivalence test and the niche similarity test were performed using 1000 simulated replicates in the R package Ecospat (Di Cola et al. 2017). Both metrics assess the statistical significance of a measured niche similarity against null model niches taken randomly from the modelling area. However, while niche equivalency test is estimated comparing the empirical D value with random relocation of the occurrence records on different distribution ranges (species lineages), the similarity test is estimated through random shifts of the niches within the available conditions of the study area (Warren et al. 2008;Broennimann et al. 2012).

Phylogenetic and species delimitation analyses
The tree topologies of the parsimony, Bayesian and maximum likelihood analyses were congruent in recovering with high support metrics the monophyly of the genus Phoneutria and the three morphologically recognized species: Phoneutria depilata, P. fera and P. boliviensis. Therefore, only the likelihood tree is shown (Fig. 2), and the main discrepancies amongst analyses relate to the relationships of Phoneutria species mentioned below. While the likelihood and parsimony analyses indicated that P. boliviensis is the sister species of P. depilata, the Bayesian analysis suggests that P. boliviensis is the sister species of P. fera. The incongruent nodes receive very low support values of jackknife, posterior probabilities, and ultrafast bootstraps. mtDNA haplotype networks (Fig. 3) revealed three major haplogroups that were congruent with the three species clades found in the phylogenetic analyses. Phoneutria fera haplotypes were separated from P. boliviensis by 29 mutations, and P. boliviensis was separated from P. depilata by 39 mutations. However, nDNA network ( Fig. 3) shows that P. depilata and P. fera share alleles, and alleles of P. boliviensis are separated from this group just by one mutation (Fig. 3). Average genetic mDNA distance for P. depilata-P. fera was 8.2%, P. depilata-P. boliviensis 7.4%, and P. fera-P.boliviensis 6.1%. For intraspecific variation comparisons, the mean p-distance for P. depilata was 2%, P. fera 1% and P. boliviensis 1%.
The ABGD method indicates four species, separating two specimens of P. boliviensis (GH2782 and GH2783) as a separate species. Instead, the mPTP species delimitation analysis indicated with high support (ASV = 0.99) the delimitation of three morphologically recognizable species. In addition, The GMYC analysis produces the same result. The posterior probabilities for the three species in each model tested in BPP were always many times higher than the alternatives scenarios: 0.97 for deep divergence and large population size, 0.33 for deep divergence and small population size (the next most likely scenario was one species with 0.10), 0.71 for shallow divergence and small population size (the next most likely scenario was two species with 0.14) and 1.00 for shallow divergence and large population size. Thus, three parameter combinations suggest the same number of species. Occurrence records, potential distribution and niche conservatism The Fig. 11A, B show the compilation of occurrence records for the two species of Phoneutria obtained from LIFIMU database and iNaturalist. Qualitatively, iNaturalist records match relatively well with the known distribution range of both species of Phoneutria, and the localities where the records came from are the same localities or the same regions of the localities of LIFIMU database. However, it is important to highlight that iNaturalist provided more occurrence records for both species than LIFIMU database. For instance, in P. depilata iNaturalist extends its distribution range to Honduras and there is more density of records in the inter-Andean Valley of Magdalena in Colombia, and the Choco region in Ecuador. In the case of P. boliviensis, iNaturalist provides more distribution records in the Amazon of Ecuador, where LIFIMU database has only one record.
Distribution models of P. boliviensis and P. depilata presented high performance compared to random expectations (AUC = 0.84 ± 0.10 SD for P. boliviensis and AUC = 0.84 ± 0.06 SD for P. depilata). The distribution model of P. depilata highlighted areas with different levels of suitability across Central and South America (Fig. 12A), with highest suitability values located in lowland and premontane areas, and from dry to tropical rain forest ecosystems. This species is well distributed in the inter-Andean Valleys of Magdalena and Cauca in Colombia. In addition, P. depilata is distributed in many areas of the Choco region of Ecuador and Colombia, and the Caribbean region reaching to Honduras. For Phoneutria boliviensis, the distribution model (Fig. 12B) indicated suitable values in lowland ecosystems of the West Amazon including Brazil, Bolivia, Colombia, Ecuador, Peru and small portion of Venezuela (although without a confirmed occurrence record). The Fig. 13A, B depicts the binary maps of the predicted distribution range of both species of Phoneutria.
Natural history. Phoneutria boliviensis is the smallest species of the genus and it inhabits in sympatry with P. fera and P. reidyi. Torres-Sánchez and Gasnier (2010) indicated that P. boliviensis seems to be restricted to periodically indudated forests because they have never been detected in "terra firme" forests. In Peru, this species was also very common in swamp forests (aguajales) dominated by the large, dioecious palm Mauritia flexuosa. However, we also found that boliviensis is not exclusive to inundated forests but also can be found in "terra firme" forests and even in the Amazonian foothills in Caqueta, Colombia. In these non-inundated ecosystems, P. boliviensis is found in secondary forests and forest edges. This species lives in the leaf litter and low vegetation. It is interesting to highlight that in the Amazon of Colombia, Ecuador and Peru, we always found P. boliviensis in sympatry with P. fera but never with P. reidyi. (Strand, 1909) fig. 2 (female). Ctenus peregrinoides : Strand, 1910: 318 (syntypes: two females from Guatemala, in ZMB 30717, not examined). New synonymy. Phoneutria depilata : Schmidt, 1954: 417-418. Phoneutria colombiana Schmidt, 19561956: 28 (female holotype from Colombia, in SMF, not examined). New synonymy.  Simó and Brescovit (2001) distinguished the Amazonian specimens of Phoneutria boliviensis from the specimens from Colombia and Central America (which we identify now as P. depilata) based on the epigynal morphology: "In specimens from Central America to Colombia it is triangular, with a wide base and a narrow apex, but in specimens from Ecuador to Bolivia the apex is more rounded". Based on the fact that Simó & Brescovit were able to distinguish these epigynal morphological differences among these two Phoneutria species and that the only species of Phoneutria in the trans-Andean region is P. depilata, we suggest that Ctenus peregrinoides (from Guatemala) is a junior synonym of P. depilata. Strand described Ctenus signativenter in 1909 based on immature syntypes from Paramba, Ecuador (one male and two female syntypes, all immatures, 3500 ft, 28 April 1898, Rosenberg leg., in ZMB 306, not examined). We have designated Ctenus signativenter as a nomen dubium because the exact identity of this species cannot be ascertained with immature specimens, but we note that the type locality suggests that the C. signativenter syntypes belong to P. depilata. Based on the epigynal morphology (Schmidt 1956, fig.3), we synonymize Phoneutria colombiana with P. depilata. Both peregrinoides and colombiana had been synonymized with P. boliviensis by Simó and Brescovit (2001).

Comments. In their revision of Phoneutria
Other material examined. Nicaragua: Región Autónoma de la Costa Caribe Sur: one female, Escondido River (      Diagnosis. Males of P. depilata resemble those of P. boliviensis by the truncated apex of the RTA (Fig. 9C, D), but differ from this and the remaining Phoneutria species by the lateral pronounced projection of the locking lobes visible in ventral view (Fig. 9B). In addition, males present an embolus with an internal bulge which is absent in P. boliviensis; and a much larger tegulum (Figs 7B, 9B). Females of P. boliviensis also resemble those of P. depilata by the general configuration of the epigynum but differ by the narrow area of the EMF (Fig. 8A), copulatory ducts slightly sclerotized (Fig. 8A), and large spermatheca heads (Figs 8B, 10D). In addition, both males and females of P. depilata can be distinguished from P. boliviensis and the remaining Amazonian species (P. perty and P. fera) by the four conspicuous series of yellow dots in the ventral side of the abdomen (this character is also present in Phoneutria eickstedtae Martins & Bertani, 2007).
Natural history. This species is found in disturbed habitats associated with both dry and humid tropical forests (0-1700 m), usually on the ground with sparse litter and low vegetation (Hazzi 2014). The range of eggs per egg sac is 430-1300, and spiderlings emerge 28-34 days after the egg sacs are produced. Sexual maturity occurs after 14-17 molts, and spiders mature 300-465 days after emerging from the egg sac (Hazzi 2014). Valenzuela-Rojas et al. (2020) reported that P. depilata is an euryphagous predator with a broad diet made up predominantly of arthropods and to a lesser extent of small vertebrates (Gekkonidae, Hylidae, and Sphaerodactylidae). There are human bite records of this species reported in Costa Rica and in banana plantations in Colombia (Flórez et al. 2003). All the cases reported have occurred with adults, and most of them have presented mild to moderate envenomation symptoms, with only one patient presenting severe symptoms such as renal failure (Flórez et al. 2003). Estrada-Gómez et al. (2015) partially characterized the venom of this species, detecting Ctenitoxin-Pb48 and Ctenitoxin-Pb53, which showed a high homology with other Ctenitoxins (family Tx3) from P. nigriventer, P. keyserlingi and P. reidyi affecting voltage-gated calcium receptors (Cav 1, 2.1, 2.2 and 2.3) and NMDA-glutamate receptors. Valenzuela-Rojas et al. (2019) found that the venom of P. depilata was significantly more effective on vertebrate (geckos) than invertebrate (spiders) prey in both LD50 and immobilization time. In addition, males had slightly more toxic venom (LD50) to geckos and much more toxic venom to spiders when compared to females (Valenzuela-Rojas et al. 2019). For two periods, March to May and October to November, adult males and females with egg sacs are always found in homes in the Inter-Andean Cauca Valley of Colombia. This likely indicates two reproductive peaks that coincided with the two rainy seasons during those same periods (N. Hazzi, unpub. data).

Discussion
Despite the medical importance of Phoneutria, its taxonomy and systematics have been always debated and there is still disagreement about the exact number of species in the genus. For instance, the last two taxonomic revisions of the genus contradict the boundaries of some species. Simó and Brescovit (2001) lumped several species into the medically relevant species P. nigriventer and only recognized five valid species. Martins and Bertani (2007) split Phoneutria nigriventer into three species, some of which had been recognized by other previous authors as valid. In the case of Phoneutria depilata, this species has been found co-occurring with P. boliviensis for several decades and many works on P. depilata have been published with the species misidentified as P. boliviensis (Valerio 1983;Hazzi et al. 2013;Hazzi 2014;Estrada-Gomez et al. 2015;Valenzuela-Rojas et al. 2019, 2020. The combination of detailed morphological (coloration and genitalia morphology) and molecular data has allowed us to distinguish P. depilata from P. boliviensis, and therefore reconsider the status P. depilata as a valid species.
Previous works of DNA barcoding in Lycosoidea have shown a range of genetic distances among congeneric species of 4-6.9% (Correa-Ramírez et al. 2010;Planas et al. 2013). Our analyses of the three species of Phoneutria resulted in interspecific distances between 6.1 to 8.2%, indicating similar genetic divergence to other Lycosoidea congeneric species. In addition, these divergences are also congruent with p-distances reported in other congeneric species of spiders (Barrett and Hebert 2005;Bidegaray-Batista et al. 2014;McHugh et al. 2014;Hormiga 2017;Agnarsson et al. 2018;Montes de Oca et al. 2016;Ballesteros and Hormiga 2018;Valdez-Mondragón 2020). Moreover, interspecific distances among haplotypes were, by far, higher than intraspecific variation between species haplotypes. For instance, the higher number of mutations was 6 between intraspecific haplotypes, compared with the lower number of mutations of interspecific species haplotypes (P. boliviensis-P. fera = 29). Interestingly, ITS-2 presented few segregating sites, and it was only able to differentiate haplotypes of P. boliviensis from the remaining two species just by one mutation step.
The distance-based method (ABGD) split P. boliviensis into two species, one which was not monophyletic. Several species delimitation studies in spiders have also shown that the ABGD method is sensitive to sampling and tends to over-split species when compared with other methods (Hamilton et al. 2014;Ortiz and Francke 2016;Valdez-Mondragón 2020). Instead, the phylogeny-based species delimitation methods employed in this study were congruent in identifying the three species of Phoneutria, corresponding completely with the morphological data. However, GMYC and mPTP methods grouped the three Phoneutria species into one, when only ITS-2 was used (an expected result due to the low genetic variation of this marker, as mentioned above). Because these two species delimitation methods were designed for single locus data (Pons et al. 2006;Fujisawa and Barraclough 2013;Kapli et al. 2017), we also implemented the BPP which explicitly models the evolution of multilocus data (Yang 2015;Luo et al. 2018). The results of this analysis also supported the existence of three species of Phoneutria. Although the ITS-2 has rarely been used in studies of spiders compared to other nDNA markers (e.g. 28S and histone H3), several studies have started to use it for DNA barcode and species delimitation recently. In Anelosimus species (Agnarsson 2010) and Gasteracantha cancriformis (Chamberland et al. 2020), this marker has insufficient variation to resolve relationships within species and among closely related species. However, for species of the genus Theridion, ITS-2 has shown a perfect match with the morphology-based species delimitation (Domènech et al. 2020). In addition, this marker has also shown to be informative with species of Loxosceles (Valdez-Mondragón et al. 2019). Therefore, ITS-2 sometimes can be useful for species identification and delimitation, and it should be used together with COI.
Citizen science platforms have provided unprecedented access to documenting species diversity and distribution across the world (Amano et al. 2016). In the case of iNaturalist, this platform presents more than 46,765,000 observations of more than 291,200 putative species of animals and plants (Horn et al. 2018). Recently, various studies have used this platform to detect disease in red mangroves (Rossi 2017), document biodiversity and distribution of echinoderms and termites (Michonneau and Paulay 2015;Hochmair et al. 2020), the rediscovery of threatened rare species (Wilson et al. 2020), and the discovery and description of new species (Winterton 2020). To our knowledge, this is the first study that has used iNaturalist to gather occurrence records on venomous species to estimate distribution models. For the two species of Phoneutria studied here, iNaturalist presented higher and more widely distributed records than our database, compiled using literature, examination of specimens from different museums, and years of personal fieldwork. Thus, our study demonstrated iNaturalist's ability to gather occurrence records and improve distribution knowledge of conspicuous and large, venomous spiders that inhabit synanthropic environments, like species of Phoneutria. Unfortunately, for the two remaining Amazonian species of Phoneuria (P. reidyi and P. fera), based on our limited knowledge, it is only possible to distinguish these two species with genitalia images and not with photographs of the habitus at this time. Therefore, we were not able to include the information of iNaturalist to model their potential distribution.
Phoneutria boliviensis and P. depilata live in lowland areas, and sometimes premontane ecosystems as well (Valerio 1983;Simó and Brescovit 2001;Hazzi et al. 2013). The distribution models corroborate that suitable areas for both species are lowland rainforest ecosystems. However, the model also indicated dry and premontane tropical ecosystems reaching elevations of 1600 m as suitable areas for P. depilata, which is congruent with the occurrence records and previous observations about the wide niche plasticity of this species (Hazzi 2014). It is also important to highlight that the species distribution maps (SDM) indicated that a large area of the Pacific of Colombia is unsuitable for this species. However, we think that the species may be present along this area but there are no records as this is one of the less explored regions of this country (Arbeláez-Cortés 2013; Arbeláez-Cortés et al. 2017). The compiled occurrence records and SDMs obtained for these two species, together with the morphological diagnosis, could have significant use in identifying risk areas of accidental bites and help health care personnel determine the species involved, especially for P. depilata which has been involved in bite accidents (Trejos et al. 1971;Florez et al. 2003).
Phylogenetic niche conservatism has been suggested as one of the potential forces in speciation and species richness patterns in the tropics (e.g., Wiens and Graham 2005;Wiens et al. 2011;Pyron et al. 2015). Under the allopatric speciation model, especially when allopatric lowland taxa are separated by a geographic barriers, one may expect that the tendency of species to maintain their ancestral climatic niche prevents them from adapting to new environments (such as mountains), isolating, and promoting speciation. (Wiens 2004;Pyron et al. 2015;Posso-Terranova and Andrés 2016). Phoneutria depilata has an allopatric distribution with respect to the three Amazonian species of Phoneutria (P. fera, P. boliviensis and P. reidyi). The Andes works as the geographic barrier that separates P. depilata (trans-Andean species) from the Amazonian species (cis-Andean species), a biogeographic pattern commonly see with many other Neotropical taxa (Albert et al. 2006;Weir and Price 2011;Richardson et al. 2015;Bartoleti et al. 2018;Salgado-Roa et al. 2018). The niche comparison analysis of these two species, using equivalency and similarity tests, indicated that both species presented niche conservatism. However, the phylogenetic analyses using different optimality criteria were not able to support, with high confidence, that P. depilata and P. boliviensis are sister species. In addition, we did not have samples of P. reidyi. Nevertheless, we think that it is still possible to conclude that Amazonian and the trans-Andean species of Phoneutria have conserved their climatic niches because the three Amazonian species are sympatric, occupying the same kind of ecosystems (climatic areas). Furthermore, the allopatric species P. depilata has a climatic niche similar to the Amazonian species P. boliviensis. These results are also congruent with other allopatric lowland cis and trans-Andean taxa that have conserved their climatic niches (Albert 2010;Richardson et al. 2015).
In conclusion, using morphological and molecular data, together with species delimitation methods our study revalidates Phoneutria depilata as a valid species separate from P. boliviensis. Both species have allopatric distributions separated by the Andean mountains, and species distribution models indicated lowland tropical rain forest ecosystems as the most suitable environments for these species. In addition, this work demonstrated the value of citizen science platforms like iNaturalist for occurrence records and improving species distribution knowledge. Phoneutria depilata and the three Amazonian species presented niche conservatism following the expected neutral model of allopatric speciation. Finally, the morphological diagnosis of these two species and the distribution maps provided in this work will be useful for future studies in venom, epidemiology of bites, and systematics of this venomous groups of spiders.