Phylogenetic analyses and description of a new species of black widow spider of the genus Latrodectus Walckenaer (Araneae, Theridiidae) from Mexico; one or more species?

A new species of the spider genus Latrodectus Walckenaer, 1805 from Mexico is described based on an integrative taxonomic approach. Latrodectus occidentalis Valdez-Mondragón sp. nov. is described using the molecular markers cytochrome c oxidase subunit 1 (CO1) and internal transcribed spacer 2 (ITS2), morphology of male and female specimens, and Species Distribution Models (SDM). Four molecular methods for species delimitation were implemented. The new species is characterized by having a unique dorsal coloration pattern on the abdomen. Latrodectus occidentalis sp. nov. is considered a distinct and valid species for four reasons: (1) it can be distinguished by morphological characters (genital and somatic); (2) the average interspecific genetic variation is > 2%; (3) 12 haplotypes were recovered within the species, being separated by the next close haplogroup of L. hesperus Chamberlin & Ivie, 1935 (30 mutations); and (4) congruence was observed among the four molecular methods. The number of recorded species of Latrodectus from Mexico increases to four: Latrodectus mactans (Fabricius, 1775), L. hesperus Chamberlin & Ivie, 1935, L. geometricus C.L. Koch, 1841 (introduced), and L. occidentalis sp. nov. The diversity of the genus Latrodectus from Mexico is surely underestimated, and more sampling is needed from the different biogeographical provinces and ecoregions to fill in these gaps.


Introduction
The spider family Theridiidae Sundevall, 1833 comprises 124 genera and 2544 species (World Spider Catalog 2023). Commonly known as 'black-widow spiders', the genus Latrodectus Walckenaer, 1805 contains the biggest spiders in size of the family Theridiidae and are well known by the medical community and general public due to their venom, which can cause neurological symptomatology in humans. Specifi cally, the presynaptic neurotoxin toxic α-latrotoxin affects the vertebrate central nervous system by increasing intracellular [Ca2 + ] in presynaptic neurons, and stimulating uncontrolled exocytosis of neurotransmitters from nerve terminals (Holz & Habener 1998).
In the last two decades, phylogenetic analyses of the genus Latrodectus using morphological characters and the mitochondrial gene CO1, suggest that the species are grouped into two main clades: 1) the geometricus clade which includes Latrodectus geometricus C.L. Koch, 1841, Latrodectus rhodesiensis Mackay, 1972 and Latrodectus umbukwane Wright, Wright, Lyle & Engelbrecht, 2019; and 2) the mactans clade which includes the rest of the species of the genus (Garb et al. 2004;Aguilera et al. 2009;Rueda et al. 2021). Both clades are morphologically supported by the position of the spermathecae in females, being V-shaped in the mactans clade and parallel in the geometricus clade (Garb et al. 2004;Aguilera et al. 2009). Garb et al. (2004) proposed an African origin for the genus, and that the South African species were the fi rst to colonize the Americas with a posterior diversifi cation in North America.
The genus Latrodectus is composed of 34 currently recognized species (World Spider Catalog 2023). Latrodectus geometricus has a cosmopolitan distribution that is considered to be the result of humanmediated introductions in several parts of the world, likely a result of anthropochory from commercial trade, in combination with its ability to adapt to disturbed environments (Chamberlin & Ivie 1935;Garb et al. 2004;Kaslin 2013). Worldwide, the Americas host the greatest diversity of species of Latrodectus, with Argentina having the highest described diversity with nine species (World Spider Catalog 2023).
In Mexico, spiders of the genus Latrodectus are known colloquially in Spanish as "viuda negra", "capulina spiders" or "casampulga" in southeastern Mexico, and in the Nahuatl language as "tzintlatlauhqui" or "cintlatlahua", which translates to: the one with the red butt. Currently, three species are reported for Mexico: L. mactans (Fabricius, 1775), L. geometricus, and L. hesperus Chamberlin & Ivie, 1935(Barreto & Barreto 1994Cortez-Roldán 2018;Valdez-Mondragón et al. 2018;Cabrera-Espinosa 2020;Cabrera-Espinosa & Valdez-Mondragón 2019World Spider Catalog 2023). For a long time, L. mactans has been supposed to be the species with the largest distribution throughout the country, being reported in all 32 Mexican states and commonly associated with anthropized areas (Cabrera-Espinosa & Valdez-Mondragón 2019, 2021; Cabrera-Espinosa 2020). Even some authors such as Chamberlin & Ivie (1935) based on the variation of the coloration pattern of the abdomen, proposed three subspecies for L. mactans: Latrodectus mactans mactans Fabricius, 1775, Latrodectus mactans texanus Chamberlin & Ivie, 1935 and Latrodectus mactans hesperus Chamberlin & Ivie, 1935. Descriptions and observations of specimens were performed using a Zeiss Discovery V8 stereo microscope at LATLAX. Digital photographs of specimens were taken with a Zeiss Axiocam 506 color camera attached to a Zeiss AXIO Zoom V16 stereo microscope. Photographs were edited using Photoshop CS6. Male palps and female epigyna were dissected in ethanol (80%). Female epigyna were cleaned in potassium hydroxide (KOH 10%) for 5 to 10 minutes, following the protocol outlined in Valdez-Mondragón et al. (2018). The habitus, male palps and female epigyna were submerged in 70% alcohol gel (ethanol) and covered with a thin layer of distilled water to minimize diffraction during photography (Valdez-Mondragón & Francke 2015;Valdez-Mondragón et al. 2019). For the photomicrographs, morphological structures (palps, epigyna, prosoma and abdomen) were dissected and cleaned with an ultrasonic cleaner at 20-40 kHz for fi ve minutes; subsequently, they were criticalpoint dried and examined at low vacuum in a Hitachi S-2460N Scanning Electron Microscope (SEM). Descriptions were carried out following Rueda et al. (2021). All morphological measurements are given in millimeters (mm). Scale measurements on photomicrographs are in micrometers (μm). The distribution map was made using Q-QGIS ver. 2.18.

Taxon sampling
Molecular analyses are based on 130 individuals of 22 species of Latrodectus and two outgroups to root the trees: Steatoda borealis (Hentz, 1850) and Latrodectus katipo Powell, 1871. This dataset was used for the analyses of the p genetic distances (uncorrected) with Neighbor Joining (NJ) using the molecular markers CO1 and ITS2 respectively ( Table 1). The species delimitation analyses with the concatenated matrix (CO1 + ITS2) were focused mainly on species from Mexico, using L. bishopi Kaston, 1938 as an outgroup for the topology obtained under Bayesian Inference (BI). Three different partitions were used for BI: 1) CO1 (580 pb), 2) ITS2 (402 pb), and 3) CO1 + ITS2 (983 pb).

DNA extraction, PCR amplifi cation and sequencing
Two legs from immatures were used for the DNA extraction with the Qiagen DNeasy Tissue Kit (Valencia, CA, USA) following the protocol by Valdez-Mondragón et al. (2019). Targeted DNA fragments included approximately 580 bp of the cytochrome c oxidase subunit 1 (CO1) mitochondrial gene and 402 bp of the Internal Transcribed Spacer 2 (ITS2) nuclear gene (Table 1). Amplifi cations were carried out in a Veriti Applied-Biosystems 96 Well Thermal Cycler, in a total volume of 25 μl: 3 μl DNA, 8.7 μl H 2 O, 12.5 μl Multiplex PCR Kit of QIAGEN and 0.4 μl of each molecular marker (forward and reverse). The PCR program for CO1 and ITS2 was: initial step 15 min at 95°C; amplifi cation 35 cycles of 35 s at 94°C (denaturation), 1.5 min at 40°C (annealing), 1.5 min at 72°C (elongation); and fi nal elongation 10 min at 72°C. PCR products were checked to analyze length and purity on 1% agarose gels with a marker of 100 bp and purifi ed directly using the QIAquick PCR Purifi cation kit of QIAGEN according to manufacturer's protocol (Valencia, CA, USA). DNA extraction and amplifi cation 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. Sanger sequencing was done at the Laboratory of Molecular Biology and Health, IB-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 OP652138-OP652221 and OP686984-OP687043 for CO1 and ITS2 respectively (Table 2).

DNA sequence alignment and editing
Sequences were edited using the programs BioEdit ver. 7.0.5.3 (Hall 1999) and Geneious ver. 2021.0.1 (Kearse et al. 2012). Sequences were aligned online with 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 molecular analyses.

p-distances under Neighbor Joining (NJ)
A genetic distances tree was reconstructed using MEGA ver. 10.1.7 software (Kumar et al. 1994), with the following commands: Number of replicates = 1000, Bootstrap support values = 1000 (signifi cant values ≥ 50%), Substitution type = nucleotide, Model = p-distance, Substitution to include = d: Transitions + Transversions, Rates among sites = Gamma distributed with invariant sites (G + I), Missing data treatment = Pairwise deletion.

Automatic Barcode Gap Discovery (ABGD)
The aim of this method is to fi nd gaps in genetic divergence, considering that the intraspecifi c genetic variation is smaller than the interspecifi c divergences. This method fi rst generates an initial data partition into putative species (Initial partitions IP). Then, these initial partitions are recursively partitioned until there is no further partitioning of the data (Recursive partitions RP). ABGD analyses were carried out on the online platform (https://bioinfo.mnhn.fr/abi/public/abgd/) using the following options: K2P distances non-corrected, Pmin = 0.001, Pmax = 0.1, Steps = 10, Relative gap width (X) = 1, Nb bins = 20.

Assemble Species by Automatic Partitioning (ASAP)
This is an ascending hierarchical clustering method in which the sequences are merged into groups that are successively further merged until all sequences form a single group. Partitions are the equivalent to each sequence merge step, then the software analyzes all partitions and scores the most probably groups into a topology or tree (Puillandre et al. 2021). ASAP analyses were run on the online platform (https://bioinfo.mnhn.fr/abi/public/asap/) using Kimura (K80) distance matrices under the following parameters: Substitution model = p-distances, Probability = 0.01, Best scores = 10, Fixed seed value = -.

General Mixed Yule Coalescent (GMYC)
Using ultrameric trees as input, this approach applies single (Pons et al. 2006) or multiple (Monaghan et al. 2009) time thresholds to delimit species in a Maximum Likelihood context (Ortiz & Francke 2016). The phylogenetic analyses were carried out in the software BEAUti and BEAST ver. 1.10.4 ) using a coalescent (constant population) tree prior to generating ultrameric trees. An uncorrelated independent log normal clock was applied to each partition with their respective evolution model and substitution rates (CO1: GTR + I + G; ITS2: K2P; 28S: GTR + I + G). Five independent replicates of 40 million iterations were run for the analyses. Tracer ver. 1.6 (Rambaut et al. 2018) was used to evaluate convergence values, with the ESS (Effective Sample Size) > 200. Tree Annotator ver. 2.6.0 (a BEAST package) was used to construct maximum credibility of clades trees, after discarding the fi rst 25% of each of the fi ve independent run as burn-in. The GMYC method was carried out using the online platform (https://species.h-its.org/gmyc/), which uses the original R implementation of the GMYC model (Fujiwasa & Barraclough 2013).

Bayesian Poisson Tree Processes (bPTP)
This molecular delimitation method is similar to GMYC; however, rather than using an ultrametric tree as input, the models of speciation rate are implemented directly using the numbers of substitutions calculated from the branch lengths. The Bayesian (BI) and Maximum Likelihood (ML) variants were carried out on the online version (https://species.h-its.org/ptp/), with the following parameters: Rooted tree, MCMC = 1 000 000, Thinning = 100, Burn-in = 0.1, Seed = 123. Trees from all analyses were edited with the iTOL online version (https://itol.embl.de/) (Letunic & Bork 2021) and Photoshop CS6.
Following Carstens et al. (2013), we used the congruence integration criteria to delimit different species, which is based on the correspondence among different molecular methods to generate a high support for species hypotheses. When the information and results of different molecular methods are incongruent, we make conservative assumptions regarding the delimitation of putative species, so we selected as different species when all the methods where congruent delimiting the species.
A haplotypes network for CO1 was constructed to visualize the mutations among haplotypes of species using the TCS algorithm (Clement et al. 2002) in PopArt ver. 1.7 (Leigh & Bryant 2015) and edited using Adobe Photoshop CS6.

Morphological analyses
For the lineal morphometry, 124 adult specimens (36 males and 88 females) of the fi ve putative species of Latrodectus from Mexico observed in this study were dissected and measured. The analyzed features included: 1) angle of the spermathecae (Supp. fi le 1A); 2) number of turns on the embolus; 3) length and 4) width of the genital opening on females (Supp. fi le 1B); 5) length and 6) width of the female carapace (Supp. fi le 1C); 7) length and 8) width of the sternum (Supp. fi le 1D); 9) length and 10) width of the femur of legs I and IV; 11) length and 12) width of the patella + tibia of legs I and IV (Supp. fi le 1E). For legs and carapace measurements, the coeffi cient "T" (length/width of the carapace) and "TT" ([patellar length + tibia]/length of the carapace) of legs I and IV were obtained, which have been used for species description (Melic 2000;Aguilera et al. 2009). The coeffi cient of each measured structure was also obtained (length/width). All measurements are in millimeters (mm) and only taken from adult specimens of both sexes. The RStudio ver. 1.4.1106 program was used to perform normality (Shapiro-Wilk) and equality of variances (Levene) tests to explore the distribution of the data and implement the corresponding analysis. Due to the nature of the data (non-normal), a non-parametric Kruskal-Wallis test was implemented to analyze the measurements of the different structures among the putative species.

Species Distribution Models (SDM)
Species Distribution Models (SDM) were carried out for L. mactans, L. hesperus and L. occidentalis Valdez-Mondragón sp. nov. Species records were obtained from the CNAN and LATLAX databases, Institute of Biology, UNAM; the Global Biodiversity Information Facility (GBIF) (www.gbif.org); and from 19 fi eld trips to 28 states along Mexico between 2017 and 2022 as was previously mentioned. To avoid spatial autocorrelation, spatial fi ltering of the records within a radius of 10 km 2 was performed for L. mactans and L. hesperus using the "Wallace" ver. 1.1.3 package in RStudio (Kass et al. 2022). Since few records (15) exist of L. occidentalis, spatial fi ltering for this species was set to 5 km 2 to reduce the loss of occurrence records. The temporality of the records with climatic variables is another important factor in the creation of the SDM, and so records were separated into two databases: (1) including all records whose dates correspond to the time interval of the climatic variables (Cortez-Roldán 2022), and (2) including the remainder of the records. To increase the number of usable records (> 10) in the SDM, records outside the time interval of the climatic variables were recovered (altitude, average annual temperature, annual precipitation and isothermality) corresponding spatially and ecologically with the records of the fi rst database in RStudio ver. 2022.02.1. For georeferencing and corroborating localities, two programs were used: GeoLocate online version (http://www.geo-locate.org/web/WebGeoref.aspx/) and Google Earth ver. 7.1.5.1557. Geographic coordinates were transformed from NAD83 to WGS84 online on INEGI, and geographical coordinates are given in degrees. The SDM data were generated using the "Kuenm" ver. 1.1.7 package in RStudio (Cobos et al. 2019). For the SDM of L. occidentalis, 15 bioclimatic variables from Mexico proposed by Cuervo-Robayo et al. (2013) were used, with an interval of years spanning from 1910-2009. These layers were downloaded from the web page: http://idrisi.uaemex.mx/distribucion/superfi cies-climaticas-para-mexico. For the SDM of L. mactans and L. hesperus, 15 climate layers from WorldClim ver. 2.1 were downloaded, spanning the interval from 1970-2000, as well as an elevation layer, available on https://www.worldclim.org/data/worldclim21.html (Fick & Hijmans 2017). From the two sets of bioclimatic variables downloaded with a resolution of 30s (~ 1 km 2 ), the variables Bio8, Bio9, Bio18, and Bio19 were excluded, since they show spatial anomalies (Escobar et al. 2014;Marques et al. 2020) (Table 3). The selection of sets of variables for each species were determined based on Analysis of Infl ation Factors (AIF) < 10 (Set 1), Contribution Percentage > 5% (Set 2), and Jacknife of AUC ( (Olson et al. 2001). The calibration area and the cut of the climatic variables were carried out in RStudio. For the creation of the candidate models, two runs were made with different values of the regularization multiplier as suggested by Cobos et al. (2019b) using the "Kuenm" ver. 1.1.7 package for RStudio. In order to the fi rst analysis, all possible combinations of linear (l), quadratic (q), product (p), threshold (t), and hinge (h) features were tested with different regularization multipliers (1, 2, 3, 4, 5, 6, 7, 8, 9, and 10). The second analysis was performed with the same combinations, but with fi ner regularization multipliers (intervals of 0.1) within the threshold of the values selected in the fi rst run. The extrapolation analyses were carried out with the "Kuenm_mmop" function available in the package used to elaborate the SDMs (Cobos et al. 2019a).

Molecular analyses of p genetic distances under Neighbor Joining (NJ)
The analyzed matrix of CO1 includes 213 terminals of 25 putative species of the genus Latrodectus ( Fig. 1). Specimens used in this study, including GenBank accession numbers and localities of the sequences used, are listed in Table 1. In this phenetic analysis under p genetic distances, two large groups are recovered: the geometricus "clade" (Fig. 1, red box) and the mactans "clade" (Fig. 1, blue box), with signifi cant bootstrap statistical support values of 60% and 80%, respectively. Based on the barcoding criterion of 2%, 25 groups corresponding to 25 putative species were recovered. Most terminals from other regions of the world form groups with bootstrap values ≥ 90%, with interspecifi c p genetic distances > 2%, recovering several species already described (e.g., L. geometricus, L. rhodesiensis, L. umbukwane, Precipitation of the wettest quarter Bio17 Precipitation of the driest quarter Within the mactans "clade", L. tredecimguttatus is divided into two groups, group D with a high bootstrap value (100%) and an intraspecifi c genetic distance of 0.2% (Fig. 1), and group G composed of two terminals containing L. tredecimguttatus and L. renivulvatus, and an intraspecifi c genetic distance of 0.1%. The genetic distance between groups D and G within L. tredecimguttatus is 7%, being samples from Iran and Spain (Appendix 1). Similar to L. tredecimguttatus, not all terminals of L. pallidus were found grouped together. Three of the four terminals from Israel are grouped with high bootstrap value (100%) and an intraspecifi c genetic distance of 1% (group H), while one terminal from Iraq groups with another L. thoracicus also from Iraq with an interspecifi c genetic distance of 2.5% between them (Fig. 1). The intraspecifi c genetic distance between L. pallidus from Iraq and group H from Israel is 9% (Fig. 1, Appendix 1). The terminals of L. thoracicus from Chile are grouped with terminals of L. mirabilis and L. variegatus from Argentina (group I) with high bootstrap support (100%), showing intra-and interspecifi c distances of 0.5% and 5%, respectively ( Fig. 1). Group K is composed of L. corallinus and L. diaguita from Argentina, with high 100% bootstrap support and an intraspecifi c genetic distance of 0.1% and interspecifi c distance > 5% ( Fig. 1, Appendix 1).
Regarding the terminals from North America in the CO1 tree ( Fig. 1, gray box), L. mactans from the United States groups with terminals from Mexico, the states of Tamaulipas and San Luis Potosi ( Fig. 1, group M). These terminals showed an intraspecifi c genetic distance of 0.4% and high bootstrap support of 100% (group M, red branch Fig. 1). Terminals of L. occidentalis sp. nov. (group N, yellow lines Fig. 1) have an intraspecifi c variation of 0.5% with a high bootstrap value of 100% and an interspecifi c genetic distance > 6% ( Fig. 1; Appendix 1). Latrodectus hesperus had the highest number of terminals used in the analysis (Fig. 1), yet the species was found to be composed of three different groups (P, Q, S), with average genetic distances between groups > 6%. Terminals of L. hesperus from Canada form two distinct groups, the fi rst with three terminals (group P) and an intraspecifi c genetic distance of 0.2% and high bootstrap support (100%, Fig. 1). Group Q (green branch) includes sequences from Canada and four sequences from Querétaro, Mexico, with an average intraspecifi c genetic distance of 0.3% and an interspecifi c distance of 6.7% with respect to group P (Fig. 1). Group Q of L. hesperus groups with group R (blue branch) with a high a bootstrap value of 100%. Group R includes several populations from an undetermined species (Latrodectus sp. 2) from the central region of Mexico, including the states of: Durango, Guanajuato, Hidalgo, Oaxaca, Querétaro, San Luis Potosí, Sinaloa, and Tlaxcala. Group R shows an intraspecifi c genetic variation of 1%, and 2.9% genetic distance from group Q (Fig. 1). Comprising group S (purple branch) are terminals of L. hesperus from the United States and terminals from Mexico (Baja California, Baja California Sur, Coahuila, Durango, Guanajuato, Querétaro, San Luis Potosi, Sinaloa and Sonora), with an average intraspecifi c genetic distance of 0.8%, and interspecifi c variation > 6% with respect to groups P, Q, and R ( Fig. 1, Appendix 1).
The analyzed matrix of ITS2 includes 58 terminals of specimens only from Mexico (Fig. 2). Since no GenBank sequences have been published for North American species at the ITS2 marker, all sequences in these analyses were generated for this study. Three species are recovered: L. occidentalis sp. nov.
(group A, yellow box), L. hesperus (group E, blue box), and Latrodectus sp. 1 + Latrodectus sp. 2 (groups C and D) (Fig. 2). Overall, the ITS2 p distance analysis with NJ shows less resolution than the CO1 analysis and does not recover the same topology and groups of putative species for the Mexico samples (Figs 1-2). However, L. occidentalis (Fig. 2) was recovered with a high bootstrap support in both analyses (75% in ITS2 and 100% in CO1). Terminals in this clade come from the Mexican states of Guerrero, Jalisco, Michoacán, Colima, and Guanajuato (Fig. 2). Group B is composed of eight terminals of Latrodectus sp. 2 from the states of Durango and Guanajuato (Fig. 2). Groups C and D are composed of terminals of the species Latrodectus sp. 2 as well as on individual of L. hesperus. In addition, group C contains one terminal of Latrodectus sp. 1 (ARA-0954) from Querétaro, with a bootstrap value of 80% (Fig. 2). Group E, with a bootstrap value of 65%, is composed entirely of terminals pertaining to L. hesperus from Baja California, Baja California Sur, Sinaloa, Sonora, and Guanajuato (Fig. 2).

Species delimitation analyses under COI + ITS2
In addition to the single locus species delimitation analyses, concatenated analyses (COI + ITS2) were run using only populations/species from Mexico (Fig. 3). Using a Bayesian Inference (BI) tree, the four different molecular methods for species delimitation, including the p genetic distances (NJ), were congruent among most methods (Fig. 3). The analyses recovered fi ve putative species with high Posterior Probabilities (PP) support values of 100% under BI: L. mactans, L. occidentalis sp. nov., Latrodectus sp. 1, Latrodectus sp. 2, and L. hesperus (Fig. 3).
The barcoding methods ASAP and ABGD (IP and RP) were congruent in recovering four species: L. mactans, L. occidentalis sp. nov., [Latrodectus sp. 1 + Latrodectus sp. 2], and L. hesperus (Fig. 3). The GMYC and PTP analyses were incongruent in the number of species recovered, with GMYC recovering seven species, and PTP recovering 11 and 13 species with the ML and BI analyses, respectively (Fig. 3).
Latrodectus mactans is recovered as a single species in all methods except PTP (IB), where the terminals from the Unites States and Mexico are separated into two different species. Only L. occidentalis sp. nov. was recovered as a distinct species in all four molecular methods, including the p genetic distances under NJ (Fig. 3).

Morphological analyses and sexual dimorphism
For the linear morphometry, 20 measurements were obtained from males (Table 4) and 24 from females (Table 5) of the next structures: 1) carapace, 2) sternum, 3) legs I and IV, and 4) female epigyna. Due to the reduced number of males and females of Latrodectus sp. 1 and males of L. mactans, neither species was included in the analyses. Of the fi ve species, Latrodectus sp. 1 is the species with the largest males, with a carapace of up 3 mm, whereas the smallest males belong to L. hesperus (Table 4). Latrodectus sp. 2 is the species with the largest females, with an average carapace length of 4.18 mm, and L. hesperus once again is the species with the smallest specimens, with females having a carapace length of 3.43 mm ( Table 5). Females of L. occidentalis sp. nov. have the longest leg I compared to the other putative species (Table 5).
Sexual dimorphism is marked in all fi ve species, with the females of L. mactans, L. occidentalis sp. nov., and Latrodectus sp. 1 being between 1.2 and 1.8 times as large as their male counterparts. Females of Latrodectus sp. 2 and L. hesperus have carapaces more than twice (2.07 and 2.24 respectively) the size of their male counterparts (Tables 4-5). However, males of all fi ve species have proportionally longer legs I than females, with the tibia-patella twice as long as the carapace (TT1) in the males of L. mactans, Latrodectus sp. 1, Latrodectus sp. 2, and L. hesperus, and more than three times as long in males of L. occidentalis (Table 4). Compared to males, the proportions of tibia-patella I and carapaces in females is no longer than 2.0 in all fi ve putative species (Table 5). Regarding the females of the four putative species analyzed, L. mactans and L. occidentalis sp. nov.
show the greatest statistical differences, with signifi cant differences in 15 of the 24 measurements analyzed ( Table 6). Females of L. mactans had statistical differences with Latrodectus sp. 2 in the femur IV coeffi cient and length of the genital opening of the epigyna (Figs 5-6), and with L. hesperus in the tibia-patella I and IV coeffi cients, femur IV length, and the TT4 coeffi cient (Figs 5-6). Between females of L. occidentalis and Latrodectus sp. 2, statistical differences were found in all measurements except in the genital structures (Figs 5, 6). Differences between females of L. occidentalis and L. hesperus were seen in the tibia-patella I coeffi cient, the three femur I measurements, the femur IV coeffi cient, the TT1 coeffi cient, and the length and width of the genital opening of the epigyna (Figs 5-6). Females of L. hesperus and Latrodectus sp. 2 showed differences in the coeffi cients of the sternum, tibia-patella I and IV, femur I and IV, TT1 and TT4, and the femur I and IV lengths (Figs 5-6).
Fifteen measurements presented signifi cant differences (p < 0.05) in both males and females for all species ( coeffi cients (Fig. 7). Males of L. hesperus showed signifi cant differences in sternum length and tibiapatella IV width in comparison with the other two species (L. occidentalis sp. nov. and Latrodectus sp. 2) (Figs 7-8).

Species Distribution Models (SDM)
The SDM of L. mactans predicts the presence of the species across large parts of the states of Tamaulipas, Coahuila, Nuevo Leon, and San Luis Potosi, as well as the northern parts of Hidalgo, Puebla, and Veracruz (Fig. 9). This region corresponds to the Tamaulipeca and Sierra Madre Oriental biogeographic provinces below 1000 m (Morrone 2004(Morrone , 2005(Morrone , 2017 (Figs 9, 12). The SDM of L. occidentalis sp. nov. shows a potential distribution throughout several states from the western region of Mexico (Fig. 10). This region corresponds to the Pacifi c coast and the Sierra Madre del Sur, encompassing the biogeographic provinces of the Pacifi c lowlands, the southern region of the Mexican Altiplano, part of the Trans-Mexican Volcanic Belt, and the Balsas Depression, at altitudes below 2000 m (Figs 10, 12). Finally, the SDM analyses predicted L. hesperus to have a widespread distribution from northwest to central Mexico, with records in nine states, all between 0-2000 m (Figs 11-12).

Differential diagnosis
Females of Latrodectus occidentalis sp. nov. can be identifi ed from those of other species of Latrodectus from Mexico by the combination of several features. Dorsal coloration: distinct red stripes and dots on dorsal surface of the abdomen: 1) one V-shaped stripe on anterior part (Fig. 25, red arrow), 2) long and sigmoid stripe dorsally along the abdomen (Fig. 25, blue arrow), and 4) two pairs of stripes towards lateral part of abdomen (Fig. 25, green arrows). Ventral coloration: abdomen with hourglass pattern always complete, wide, and never separated as in L. hesperus and L. mactans (Fig. 26). Body measurements (Figs 5-6): sternum coeffi cient (length/width) longer than other species analyzed herein, tibia-patella I longer than other species, tibia-patella coeffi cient longer than other analyzed species, femur I longer than other species, femur I and IV coeffi cient longer than other species, TT1 and TT4 coeffi cients slightly longer than other species. Angle between the spermathecae position in dorsal view longer than other species. Genital opening of epigynum wider than other species (Figs 29, 31-42 (ventral views)). Copulatory ducts forming four or fi ve loops around spermathecae (Figs 30, 33-43 (ventral views), 72). Males: similar to females, but abdomen oval (Figs 17, 27-28). Coloration: ventrally, abdomen with hourglass pattern always complete, thinner than females (Fig. 28). Body measurements (Figs 7-8): sternum coeffi cient (length/width) longer than other species analyzed herein, tibia-patella 1 length and     tibia-patella coeffi cient longer than other species, tibia-patella 1 coeffi cient longer than other species, tibia-patella 1 coeffi cient longer than other species, length and coeffi cient of femur I longer than other species, TT1 coeffi cient longer than other species, tibia-patella IV longer than other species, femur IV longer than on other species. Embolus with 3 coils, located distally on cymbium (Figs 44-48, red arrows), last coil extending medially, curving downwards along retrolateral part of palp, continuing ventrally, ending in a long thin tip (Figs 44-60). In L. mactans and L. hesperus (native species from North America), the embolus has 2 coils (Cabrera-Espinosa & Valdez-Mondragón 2021: fi gs 30-37).

Etymology
This species is a name in apposition and refers to the distribution of the species in the Spanish language: 'Occidente de México' ('western Mexico'), 'occidentalis', that includes the states of Guerrero, Michoacán, Colima, and Jalisco (type locality).
LEGS. Uniformly moderately setose throughout, tapering in metatarsi and tarsi (Figs 73-75). Femora I wider basally, femora I and IV slightly curved in lateral view. Tibiae I slightly curved distally. Three tarsal claws present, lateral claws with one row of teeth, which become larger distally. Tarsi IV with distinct comb macrosetae.
LEGS. Uniformly moderately setose throughout, tapering in metatarsi and tarsi. Femora I slightly wider basally, femora I and IV less curved in lateral view than female. Tibiae I and IV slightly curved distally. Three tarsal claws present, lateral claws with one row of teeth, which become larger distally. Tarsi IV with distinct comb macrosetae.

Variation
Females have black carapace and abdomen coloration, black or dark brown legs, never light shades (Figs 25-26). Globular abdomen always black with distinctive red lines in dorsal region (Figs 61-69, red coloration is lost under ethanol), lines may be outlined by a thin white line in juveniles. The frontal region of the dorsal sigmoidal line may or may not be separated at an isolated point between the rest of the line and the V-shape. The two pairs of lines laterally may be attenuated or reduced, but always present 25,(61)(62)(63)(64)(65)(66)(67)(68)(69). Males always present a black body coloration (Fig. 17), legs black or dark brown with light brown on the mid tibia, metatarsus, and tarsus (Figs 17-18). Abdomen oval with distinctive red lines on the dorsal region, which may be outlined by a thin white line 27). As in females, the frontal region of the transverse dorsal line may or may not be separated at an isolated point between the rest of the line and the V-shape (Fig. 27). Males can vary greatly in body size   (Table 4), however sexual dimorphism is always marked, with females being more than twice the size of males, but males with proportionately longer legs (Tables 4-5). The genital opening of females varies in shape, but is always wider than long (Figs 29,(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)), the spermathecae are arranged in a  V-shape surrounded by the copulatory ducts which may have four or fi ve turns around them ). The angle of the spermathecae may vary, but they are never parallel as in L. geometricus (Table 5). Palps of males show little variation, sclerites with little variation, embolus always with three complete turns (Fig. 46); however, there is variation in the apical part (Figs 49-60).

Natural history
The specimens were collected in their cobwebs in natural and anthropized localities (Figs 19-24). The species is distributed mainly in tropical deciduous forests in the western states of Mexico: Jalisco, Guerrero, Michoacán, Colima, Guanajuato, Morelos, and Oaxaca (Fig. 21). The specimens from the type locality were collected inside the same house, under chairs, on window frames, under stowed dishes, and furniture (Figs 23-24, red arrows). Specimens collected in nature were found under and between big boulders on the ground (Figs 19,22,red arrow), under mounds of rocks, inside cavities on walls along road-cuts, at the bases of alive cacti such as "nopales" (Opuntia spp.) and columnar cacti, and even under rotten cacti . Some females were collected with egg sacs, which are usually big, oval and yellow in coloration (Fig. 15), differing from other Mexican species (e.g. Latrodectus sp. 1 and sp. 2) that have smaller egg sacs whitish or pale-yellow in coloration.

Discussion
The use of separate lines of evidence in systematic studies, including morphological, molecular, ecological, and biogeographical data, provides more robust hypotheses when identifying and describing species that cannot be delimited by traditional morphology (DeSalle et al. 2005;Zhang et al. 2013;Valdez-Mondragón et al. 2019;Navarro-Rodríguez & Valdez-Mondragón 2020;Hazzi & Hormiga 2021). In recent years, molecular methods for species delimitation have provided a new way to resolve problems with species complexes or underestimated diversity. This has been especially useful in morphologically conservative groups by providing infra-specifi c genealogical information from DNA markers, which allows for objective implementation of modern species concepts (e.g., biological, phylogenetic, genotypic cluster, cladistics, etc.). Analyzing the data with a wide variety of species delimitation methods and delimiting lineages that are consistent across results follows the integrative taxonomy approach of species delimitation (DeSalle et al. 2005;Carstens et al. 2013;Luo et al. 2018;Valdez-Mondragón 2020).
Mitochondrial markers have been shown to be robust in the delimitation, revalidation, and description of several species within the genus Latrodectus, and often recover the main species groups or clades "mactans" and "geometricus" (Garb et al. 2004;Aguilera et al. 2009;Wright et al. 2019;Rueda et al. 2021). In this study, both of these clades were recovered in the phenetic and phylogenetic analyses; however, some discrepancies within the groups were found, as also recorded by Garb et al. (2004). For the terminals analyzed from Mexico, two of the fi ve putative species belong to species already described (L. mactans and L. hesperus). However, some terminals from the state of Querétaro are grouped with terminals of L. hesperus from Canada, which refl ects a possible introduction by anthropochory from Canada to Mexico or, more likely, from Mexico to Canada, although more research is needed to be conclusive. This would not be the fi rst case of species of Latrodectus being introduced by anthropochory, as L. geometricus has been recorded in several countries around the world and is considered cosmopolitan and invasive (Garb et al. 2004;Cabrera-Espinosa & Valdez-Mondragón 2021;Rueda et al. 2021). Recently, using the CO1 barcoding marker, Choi et al. (2019) recorded L. hesperus from South Korea, representing another human-mediated introduction of the species. Although Chamberlin & Ivie (1935) mentioned that L. hesperus is distributed as far south as Mexico, there were no records of this species in Mexico from that time. Recent faunistic inventories, however, have since reported the presence of L. hesperus in the Mexican states of Baja California Sur, Chihuahua, Coahuila and Hidalgo, in addition to additional records of L. mactans and L. geometricus from several other states (Castañeda-Gómez et al. 2012;Jiménez et al. 2015;Salceda-Sánchez et al. 2017;Desales-Lara et al. 2018; Cabrera-Espinosa 2020; Cabrera-Espinosa & Valdez-Mondragón 2021).
Species delimitation is the process by which individuals from different populations are assigned to one or more already described or new species, while also the intraspecifi c variation is assessed (Rannala & Yang 2020). Aguilera et al. (2009) noted that molecular species delimitation methods are good tools for recognizing supposedly cryptic species or potential species complexes, as well as identifying species that had been synonymized through the use of solely morphological characters (e.g., Latrodectus thoracicus). However, a discrepancy between different species delimitation methods is often found, mainly due to the different statistical capacities of each model to recognize species or lineages (Carstens et al. 2013; Navarro-Rodríguez 2019; Valdez-Mondragón 2020; Nolasco & Valdez-Mondragón 2022). Therefore, the use of a concatenated matrix such as CO1 + ITS2 in this study, has shown to provide robust evidence for species delimitation analyses, mainly in tree-based analyses. These tend to separate over-sampled lineages (Pérez-Delgado et al. 2021), such as Latrodectus sp. 2 and L. hesperus in this work. To avoid possible overestimates of putative species of Latrodectus from Mexico, conservative estimates of the total number of species were calculated when discrepancies were observed between different molecular delimitation methods (Cartens et al. 2013).
Based on the criteria of genetic distances (p) and species delimitation methods, we consider that the putative species Latrodectus sp. 2 might correspond to L. mactans mexicanus, a subspecies not recognized by Levi (1959). However, further study is required to determine if specimens from the central region of Mexico constitute a separate species or possibly another species complex. As found in this work, previous taxonomic studies of Latrodectus using molecular markers have recovered interspecifi c genetic distances greater than 2%, except for some species pairs such as L. thoracicus-L. variegatus, and L. corallinus-L. diaguita (Garb et al. 2004;Aguilera et al. 2009;Rueda et al. 2021). The polyphyletic allocation of specimens of L. hesperus from USA-Mexico (group S, Fig. 1) and from Canada-Querétaro (Mexico) (group Q, Fig. 1) has been previously reported by Barrett & Hebert (2005) and Rueda et al. (2021), with a corrected genetic distance (K2P) greater than 7%, suggesting the possibility that the Canadian specimens correspond to a new and undescribed species.
Phylogenies are a useful approach to estimating relationships between groups of organisms; however, establishing genealogical relationships at the population level through the use of molecular characters (genes) is very complicated with traditional phylogenetic methods (e.g., Neighbor-Joining, Parsimony, or Maximum Likelihood) (Clement et al. 2002). Haplotype networks, in particular the TCS method (Fig. 4), offer an alternative way to estimate and visualize relationships between sampled organisms and can help to elucidate phylogeographic histories (Clement et al. 2002). Genetic analyses of populations combined with phylogenetic and morphological analyses have been used to recognize cryptic species population or the existence of gene fl ow between them. This was observed for haplotypes m1 in L. mactans, oc4 in L. occidentalis sp. nov., hap5 in Latrodectus sp. 2, and h6 and h21 in L. hesperus (Fig. 4). Between CO1 haplotype groups of the putative species recognized in this work, we found a maximum of 16 mutations, whereas Rueda et al. (2021) report a maximum of seven mutations between haplotype groups using the Median Joining Networks method. Despite this discrepancy, the haplotype networks from both results recover and corroborate the different estimated putative species (Planas & Ribera 2015;Valdez-Mondragón et al. 2019;Navarro-Rodríguez & Valdez-Mondragón 2020;Rueda et al. 2021).
The use of sexual characters to recognize species is widely used in spiders and other arthropods since sexual structures both evolve faster than other (i.e., somatic) structures and generally present little intraspecifi c variation in order to maintain compatibility between the sexes (Huber et al. 2005). It has been documented in spiders that such sexual characters are not always reliable for distinguishing between species (Huber et al. 2005;Aguilera et al. 2009). This has led to a complicated taxonomic history of the genus Latrodectus that has to date been based essentially on the use of morphological characters, mainly primary sexual structures (Chamberlin & Ivie 1935;Gonzales 1954;Levi 1959;Kaston 1970;Lotz 1994;Aguilera et al. 2009); such as male palps and female epigyna, which, as was demonstrated herein, present interspecifi c and intraspecifi c variation. The use of linear measurements as complementary characters in the description of species in most cases is not informative due to large variations in size of adult specimens, oftentimes within the same population (Aguilera, pers. com.). Traditionally, specimen measurements have focused on the carapace, tibia, or tibia-patella of leg I (Chamberlin & Ivie 1935;Lotz 1994). Although the use of solely linear somatic measurements is not appropriate for species delimitation as was demonstrated in this work, they provide useful evidence to help understand both interspecifi c and intraspecifi c variation, in addition to being useful for the descriptive purposes in species accounts.
Recent taxonomic works on Latrodectus have used the "T" (length/width of carapace) and "TT" (length tibia-patella of leg I or IV) coeffi cients to refl ect body proportions and determine intra and interspecifi c variation (Melic 2000;Aguilera et al. 2009;Rueda et al. 2021). In our study, the T coeffi cient was not found to be useful for explaining the variation of the putative species from Mexico; however, as seen in Rueda et al. (2021), the TT coeffi cient for legs I and IV was useful for recognizing morphological differences between species in both females and males. Except for the genital opening coeffi cient in females, the rest of the coeffi cients analyzed herein (i.e., sternum, tibia-patella I and IV, femur I and IV), showed signifi cant differences in both males and females and thus are useful for the identifi cation of specimens and as complementary taxonomic characters. As in previous works (Rueda et al. 2021), no qualitative characters were found in the sexual structures of the putative species analyzed in this study; however, the length, width, and angle in which the female spermathecae are arranged with respect to each other showed differences as quantitative characters.
The dorsal pattern of the abdomen, mainly on females, has been recognized as an important and complementary taxonomic character in the diagnosis of some species of Latrodectus (e.g., L. geometricus, L. rhodesiensis, L. pallidus, L. lilinae, L. tredecimguttatus, L. curacaviensis, L. garbae) (Melic 2000;Rueda et al. 2021) since it allows the species recognition without observing the genitalia. The fi rst taxonomic revision of Latrodectus was published by Pickard-Cambridge (1902) using color patterns, setae, or spines on the abdomen to distinguish all species known at that time. Chamberlin & Ivie (1935) carried out a study on the "black widow spiders" from northern Mexico, recognizing three subspecies of L. mactans: Latrodectus mactans mactans Fabricius, 1775, Latrodectus mactans texanus Chamberlin & Ivie, 1935 and Latrodectus mactans hesperus Chamberlin & Ivie, 1935. Levi (1959 grouped these subspecies as one single species (L. mactans) based on structural similarities of the palps in adult males. In this same work, only three species were recorded for the Americas: L. geometricus, with a tropical distribution; L. mactans, limited to tropical regions; and L. curacaviensis, an American endemic distributed from Canada to Argentina. Levi (1958) also suggested the possibility that Latrodectus hasselti Thorell, 1870, Latrodectus indistinctus O. Pickard-Cambridge, 1904and Latrodectus tredecimguttatus (Rossi, 1790 were synonyms of L. mactans since he found no obvious morphological differences. Eight years later, Levi (1966) admitted that two or more species might present morphological similarities in the male palps, rendering this character unreliable at the species level. Recently, Cabrera-Espinosa (2020) and Cabrera-Espinosa & Valdez-Mondragón (2021) updated the adult female abdomen dorsal pattern colorations of Latrodectus from Mexico, reporting eight different dorsal patterns of the eleven types of L. mactans previously reported by Levi (1959). The geographic regionalization of some dorsal coloration patterns of the abdomen of Mexican L. mactans suggests that a complex of unidentifi ed species may be present.
As for the putative species analyzed in this study, only L. occidentalis sp. nov. is easily identifi able and diagnosable by the dorsal pattern of the abdomen of adult females. While primary sexual structures have traditionally been used for decades in spider taxonomy, body color patterns on the carapace and abdomen have been useful to diagnose groups that lack differentiation in sexual structures (e.g., Psilochorus Simon, 1893, Pholcus Walckenaer, 1805 (Pholcidae), and Maratus Karsch, 1878 (Salticidae)) (Huber et al. 2005;Waldock 2013Waldock , 2014Huber & Dimitrov 2014).
Although Species Distribution Models (SDMs) are not a species delimitation method, they allow for the estimations of species' geographic ranges based on environmental conditions associated with each record (Kaslin 2013). SDMs have been an additional criterion in recent years for taxonomic decisions, providing important information that models the distributions of lineages or species . The SDMs of the three species found their distributions to be well defi ned by altitude. This variable is likely driven by associated temperature and precipitation, as previous works have found the importance of these variables as limiting factors in the distribution of other species of Latrodectus (Vink et al. 2011;Kaslin 2013). The genus Latrodectus, unlike other groups of spiders in Mexico, displays a broad distribution across different climates, vegetation types, and altitudes, ranging from sea level in tropical deciduous forests to cold temperate forests above 2300 m (Cabrera-Espinosa & Valdez-Mondragón 2019, 2021; Cabrera-Espinosa 2020). Similar to Taucare-Ríos et al. (2016), our results found that temperature is the most important factor in the distribution of species of Latrodectus from Mexico, with this variable (maximum temperature of the warmest month and mean diurnal range) being present in the three SDM data sets.
Contrary to what is reported by Levi (1959) and Cabrera-Espinosa & Valdez-Mondragón (2019, 2021), we report that L. mactans does not have a wide distribution in Mexico. Rather, the distribution is limited by the Sierra Madre Oriental in the west and possibly by the Trans-Mexican Volcanic Belt in the south, with records of L. mactans found in the Great Plains ecoregion on the coast of the Gulf of Mexico (Figs 9, 12). This suggests that L. mactans is only distributed in the Tamaulipeca, Veracruzana, and lower parts of the Sierra Madre Oriental biogeographical provinces. Of the three species, L. occidentalis sp. nov. is the only one with a distribution in the Neotropical region of Mexico, mainly associated with the lowland forests of the Pacifi c coast (Figs 10, 12). The highlands (> 2000 m) of the Trans-Mexican Volcanic Belt and the Sierra Madre del Sur present a barrier for the distribution of this species to reach the central region of Mexico. The distribution of L. hesperus is reported by Chamberlin & Ivie (1935) to extend from southern Canada to Mexico, down the western coast of the USA. This distribution is recovered in our SDM analysis, while also extending to the central region of Mexico on both sides of the Sierra Madre Occidental. However, molecular analyses in this study and previous works (e.g., Barrett & Hebert 2005;Rueda et al. 2021) fi nd the populations of L. hesperus from Canada to likely represent an undescribed species. Unfortunately, the material from these populations deposited in GenBank lacks locality information, so it was not possible to verify that these records lie within the area estimated by our SDM.
In conclusion, the diversity of the genus Latrodectus is underestimated in Mexico, and more sampling is needed throughout the country, mainly from distinct biogeographical provinces. As previously reported, we also fi nd that traditional morphology alone does not provide robust characters for species-level identifi cation for Mexican and North American lineages. However, the combination of sexual features such as male palps and female epigyna plus the use of somatic characters such as the dorsal pattern of the abdomen in females, as well as some somatic linear measurements, proved to be informative characters for the identifi cation of some species in the genus. Altogether, our integrative taxonomic approach using morphological and molecular data (CO1 and ITS2) adds to the knowledge of this group and increases the species sampling from North America.