Examined specimens, fossil or extant, are housed in the following collections (curators in parentheses): IBSP – Instituto Butantan, São Paulo, Brazil (A. D. Brescovit); MACN-Ar – División Aracnología, Museo Argentino de Ciencias Naturales “Bernardino Rivadavia”, Buenos Aires, Buenos Aires, Argentina (M. J. Ramírez); MCZ – Museum of Comparative Zoology, Harvard, USA (G. Giribet and M. Srivastava); MNHNCu – Museo Nacional de Historia Natural de Cuba, Havana, Cuba (G. Alayón García); SMF-Be – Amber Collection, Senckenberg Research Institute and Natural History Museum, Frankfurt, Germany (M. M. Solórzano-Kraemer), SMNG – Senckenberg Museum für Naturkunde Görlitz, Görlitz, Germany (A. Christian).

Amber pieces were re-polished at the SMF using a Phonix Beta polishing machine with grinding paper for metallography, wet and dry: Grip 2500 and 4000. After polishing, pieces were embedded in Araldite 2020® Epoxy resin following Sadowski et al.’s (2021) recommendations.

The following specimens have been examined for the new scorings of the morphological matrix, or to discuss the morphology of scytodoids. We scored Ochyrocera diablo Pérez-González, Rubio et Ramírez, 2016 using the data from the original description (Pérez-González et al. 2016).

Althepus maechamensis Li et Li, 2018 (Psilodercidae): THAILAND • 1 ♂ 1 ♀; Chiang Mai, Doi Inthanon National Park; 18.53°N 98.5025°E; 6 Oct 2003; ATOL Expedition 2003 leg.; MACN-Ar 35342.

Drymusa spectata Alayón, 1981 (Drymusidae): CUBA • 1 ♂ 1 immature; Cienfuegos; Comanaqua (Cumanayagua); cueva del Canto; 750 m a.s.l.; 21.8930°N 80.1486°W; 20 August 2002; J. M. Ramos leg.; MNHNCu.

Drymusa armasi Alayón, 1981 (Drymusidae): CUBA • 1 ♂ 1 ♀; Santiago de Cuba, Gran Piedra; under rocks; 1100 m a.s.l.; 20.083333°N 75.623611°W; 21 June 1982; L. F. Armas leg.; MNHNCu.

Drymusa simoni Bryant, 1948 (Drymusidae): HAITI • 1 ♂ (holotype) 1 ♀; Nord-Ouest, LaHotte; 16–17 Oct 1934; P. J. Darlington leg.; MCZ 23101. 2 immatures; same data as the holotype; MCZ 44196.

Loxosceles caribbaea Gertsch, 1958 (Sicariidae): CUBA • 2 ♂ 4 ♀; Santiago de Cuba, Reserva Ecológica Siboney-Jutici, Cueva La Virgen; 19.96083°N 75.7144°W; 04 May 2010; A. Pérez González leg.; MACN-Ar 32734, MACN-Ar 32736.

Loxosceles cubana Gertsch, 1958 (Sicariidae): CUBA • 2 ♂ 2 ♀; Pinar del Río; Viñales; Cueva del Cable; 22.6677°N 83.7094°W; A. Sánchez leg.; 20 April 2012; IBSP 164702.

Loxosceles deserta Gertsch, 1973 (Sicariidae): USA • 1 ♂ 1 ♀; Arizona, Tucson; F. E. Russell leg.; MACN-Ar 21497, MACN-Ar 21498.

Loxosceles hirsuta Mello-Leitão, 1931 (Sicariidae): ARGENTINA • 1 ♂; San Luis, Junín, Establecimiento La Clotilde; 32.19955°S 65.59549°W; 10 January 2017; M. Ramírez et J. Faivovich leg.; MACN-Ar 38707.

Loxosceles laeta (Nicolet, 1849) (Sicariidae): BRAZIL• 3 ♂ 3 ♀; São Paulo, Osasco; 23.5325°S 46.7916°W; 26 Jul. 1982; J. R. Bettinazzi leg.; IBSP 34948.

Loxosceles rufescens (Dufour, 1820) (Sicariidae): ISRAEL • 1 ♂ 1 ♀; HaZafon, near Elon, Nahal Bezet Nature Reserve; 33.0740°N 35.2379°E; 17 February 2020; I. L. F. Magalhaes, E. Gavish-Regev, Z. Ganem, S. Aharon, N. Givon et M. Arnedo leg.; MACN-Ar 41223.

Loxosceles simillima Lawrence, 1927 (Sicariidae): NAMIBIA • 2 ♂ 2 ♀; Waterberg; 20.352701°S 17.337579°E; M. Stockmann leg.; MACN-Ar 39459.

Loxosceles taino Gertsch et Ennik, 1983 (Sicariidae): DOMINICAN REPUBLIC • 1 ♂ 2 ♀; Pedernales; no collecting date; A. Sánchez leg.; IBSP 164710.

The photographs and Z-stacks images of fossil species were taken under a Nikon SMZ25 microscope, using Nikon SHR Plan Apo 0.5× and SHR Plan Apo 2× objectives with a microscope camera Nikon DS-Ri2 and the NIS-Element software (version 4.51.00; www.microscope.healthcare.nikon.com). In some cases, to prevent diffraction caused by irregular surfaces of the amber piece, we placed a coverslip and a drop of water with sugar on top of the piece to flatten the surface to be photographed.

Morphological observations on extant specimens were made using Leica M165 C and Leica M125 stereomicroscopes. Pictures were taken with Nikon DXM1200 digital camera mounted on a stereoscopic microscope Nikon SMZ1500 and on a microscope Leica DM4000 M, and with a Leica DFC 500 digital camera mounted on a stereoscopic microscope Leica M165 C or Leica M216. Extended focal range images were composed with the Leica Application Suite version 3.6.0. or Helicon Focus 3.10.3–4.62 (https://www.heliconsoft.com, Ukraine). Preparations were carefully cleaned using fine brushes and a thin jet of alcohol from a thinned pipette; some setae were removed to expose structures, especially those on legs, palps, spinnerets, and chelicerae.

For scanning electron microscope (SEM), all preparations were dehydrated in a series of increasing concentrations of ethanol (80%, 90%, 95%, 100%), and critical-point dried. After drying and brushing, they were mounted on adhesive copper tape (Electron Microscopy Sciences, EMS 77802) affixed to a stub and secured with a conductive paint of colloidal graphite on isopropyl alcohol base (EMS 12660). Prior to SEM examination under a high vacuum with a FEI XL30 TMP or a LEO 1450VP, the structures were sputter-coated with Au-Pd.

The imaging of the holotype of L. aculicaput was performed at the Imaging Beamline – IBL P05 - PETRA III at Deutsches Elektronen Synchrotron (DESY) in Hamburg, operated by the Helmholtz-Zentrum Hereon (Greving et al. 2014; Wilde et al. 2016). The specimen was imaged at a photon energy of 18 keV using a commercial CMOS camera system with an effective pixel size of 1.28 µm. The sample to detector distance was set to 3 cm. For each tomographic scan, 3601 projections at equal intervals between 0 and π were recorded. Tomographic reconstruction was done by applying transport of intensity phase retrieval approach and using the filtered back-projection algorithm (FBP) implemented in a custom reconstruction pipeline (Moosmann et al. 2014) using Matlab (Math-Works) and the Astra Toolbox (Palenstijn et al. 2011; van Aarle et al. 2015; van Aarle et al. 2016). For processing, raw projections were binned two times, resulting in an effective pixel size of the reconstructed volume of 2.56. The complete specimen was segmented in three dimensions using region-growing techniques in VGStudioMax (version 3.3.1 www.volumegraphics.com/de, Volume Graphics, Heidelberg, Germany).

We carried out the segmentation and volume rendering of the pedipalps in AMIRA5.4.5 (FEI Visualization Science Group, Burlington, MA, USA). All images were post-processed to adjust contrast and sharpness using Photoshop CS6 (Adobe Inc., San Jose, CA, USA).

We augmented the morphological matrix of Scytodoidea families (Labarque and Ramírez 2012) by including the following taxa: Ochyrocera diablo, Althepus maechamensis, Loxosceles defecta, and L. aculicaput. We added the first two to include all extant Scytodoidea families in the sampling, and the latter two to test their phylogenetic position. Loxosceles deformis was not included because the only known specimen is poorly preserved and most characters needed in the matrix are not observable.

We added the following characters to accommodate the morphological diversity brought by newly incorporated taxa:

(102) Chelicerae promarginal lobe, shape: (0) small and rounded, (1) large and detached, (2) with a distal prolongation (Fig. 7C).

(103) Chelicerae medial lamina, apex shape: (0) simple, (1) bifid (Fig. 1D).

Figure 1. 

Loxosceles defecta Wunderlich, 1988, holotype (SMF-Be 970a; A, B, E, G–I) and paratype (SMF-Be 970b; C, D, F) embedded in the same amber piece, photographs under light microscopy. A habitus, dorsal view. Arrow to incrassate tibia I. B prosoma, dorsal view. C prosoma, subanterior view. Note six eyes in three dyads. D left chelicera, subanterior view. Arrows to bifid apex of cheliceral lamina. E left tibia I, retrolateral view. Arrow to incrassate region with strong macrosetae. F right tibia I, dorsal view. Arrow to strong macrosetae. G left tarsus I, apical view. Note absence of third claw. H right palp, retrolateral. I: left bulb, dorsal. Arrow to prolateral lobe of the cymbium. Scale bars: 1 mm (A, B, E, F), 0.5 mm (C, H), 0.2 mm (I) 0.1 mm (D, G).

(104) Colulus, shape: (0) lobe-like, (1) broad, plate-like (Fig. 7I).

(105) Male, leg I, femur, prolateral macrosetae: (0) absent, (1) present (Fig. 3C).

(106) Labium, apex, shape: (0) straight, (1) pointed, (2) notched (Fig. 7B).

(107) Podotarsite, divisions: (0) single unit, (1) subdivided by additional articulation (see Labarque et al. 2017 for explanations of characters 107–109).

(108) Podotarsite, cuticle on dorsal side: (0) membranous, open podotarsite; (1) sclerotized, closed podotarsite.

(109) Podotarsite, distal dorsal hood: (0) absent, (1) present.

(110) Chelicerae, promarginal lobe, distal prolongation, shape: (0) curved, rounded; (1) straight, pointed (Fig. 7C).

We added legacy genetic data from six target-gene markers from the mitochondrial (12S rRNA, 16S rRNA, COI) and nuclear (histone H3, 18S rRNA, 28S rRNA) genomes, and genomic ultra-conserved elements (Table 1). To maximize morphological and genetic data overlap, we merged data from closely related species into a single terminal in the outgroup genera Stedocys, Ariadna, Althepus, and Ochyrocera (Table 1). The alignment of target-gene markers was made with the online version of MAFFT (Katoh et al. 2019) with the L-INS-I option, which resulted in a total length of 6458 sites (12S 333 sites; 16S 469; 18S 1746; 28S 2348; COI 1235; H3 327).

Published UCE and transcriptome sequences were downloaded from the Sequence Read Archives using fastq-dump v. 2.11 (SRA Toolkit Development Team: https://hpc.nih.gov/apps/sratoolkit.html) with the flags –gzip –clip –split-files –qual-filter to remove adapters and trim low-quality base calls. Sequence reads were trimmed using TRIMMOMATIC v. 0.39 (Bolger et al. 2014), with parameters LEADING:20 TRAILING:20 SLIDINGWINDOW:4:20 MINLEN:30 HEADCROP:5 and –phred 33, and assembled using SPAdes genome assembler v3.15.3 (Prjibelski et al. 2020) with parameters single-cell flag and coverage cutoff value auto. Duplicates were removed using CD-HIT version 4.8.1 (Fu et al. 2021) with parameters: sequence identity threshold 0.95, word_length 10. The sequences were processed with the PHYLUCE v. 1.7.1 (Faircloth 2016) pipeline. We matched the assembled sequences to a compilation of UCE probes merging the Arachnid probe set (Starrett et al. 2017), and the Spider probe set (Kulkarni et al. 2020). Sequences were aligned using MAFFT v. 7.455 and trimmed with GBLOCKS v. 0.31b (Castresana 2000), with the default parameters implemented in PHYLUCE. The recovery of UCE markers did not work well with the few taxa represented in our analysis, probably because few markers were shared by several species; this small dataset failed to recover the monophyly of well-established groups, such as the genus Loxosceles. We then used a wider sample of 100 short-read archives from 75 species of Synspermiata and outgroups (Table 1), which recovered 1912 markers with a minimum coverage of 10%, and a total length of 510,125 sites. From this dataset, we selected the target species, which were analyzed together with the target-gene markers and morphological data.

The phylogenetic analysis under maximum likelihood was done with IQ-TREE v. 2.1.3 (Minh et al. 2020), selecting models for each target gene and the concatenated UCE markers by Bayesian information criterion with ModelFinder (Kalyaanamoorthy et al. 2017), and estimating support with 1000 rounds of ultrafast bootstrap (Hoang et al. 2018). The models selected were: UCE GTR+F+I+G4, 12S TIM2+F+G4, 16S TIM2+F+I+G4, 18S TNe+R3, 28S TIM3+F+R3, COI TIM3+F+I+G4, H3 K2P+R2. We removed the invariant characters of the morphological partition and replaced polymorphic scorings with missing entries. Morphological data were analyzed afterwards as two separate partitions, one for 74 unordered characters, treated with the Mk model, and the other for 5 ordered characters, with the Ordered model; both partitions were corrected for the absence of invariant characters with the Asc model. The analysis under maximum parsimony was done with TNT v. 1.5 (Goloboff and Catalano 2016), under equal weights; since 100 out of 100 heuristic searches of 1 RAS followed by TBR branch swapping reached the same tree and length, it is likely that the optimal tree was found. We estimated branch support with 1000 rounds of bootstrap. Synapomorphies were calculated with TNT.

ALS = anterior lateral spinneret, Bu = bulb, Ch = chelicera, Co = colulus, Cy = cymbium, CF = cheliceral fang, e = embolus, Fe = femur, Pa = patella, PLS = posterior lateral spinneret, PMS = posterior median spinneret, SF = stridulatory file, Ti = tibia.

The supplementary figures and data underpinning the analyses reported in this paper are deposited in the Zenodo repository at https://zenodo.org/record/6954956 under www.doi.org/10.5281/zenodo.6954956.

We present new photographs under light microscopy of the holotypes of Loxosceles defecta (Fig. 1), L. deformis (Fig. 2) and L. aculicaput (Fig. 3). This revealed some details that had been overlooked in the original descriptions, such as a third tarsal claw in L. aculicaput (Fig. 3E–F), or the modified legs, bifid cheliceral lamina and stridulatory files in the chelicerae of L. defecta (Fig. 1D–F). The phylogenetic significance of these observations is discussed below (see 4.2.). In the case of L. aculicaput, a large fissure in the amber prevented observation of the ventral aspects of the specimen, and thus only the dorsal side could be imaged (Fig. 3), hampering a detailed examination of some taxonomically important structures, such as the palps (Fig. 3G).

Figure 2. 

Loxosceles deformis Wunderlich, 1988, holotype (SMF-Be 968a), photographs under light microscopy. A, B habitus, dorsal view. C left tibia I, prolateral. D palps, subventral view. E right palp, retrolateral view. F right bulb, retrolateral view. Scale bars: 2 mm (A), 1 mm (B, C), 0.5 mm (D, E), 0.1 mm (F).

Figure 3. 

Drymusa aculicaput (Wunderlich, 2004) comb. nov., holotype (SMNG 07/36287-422), photographs under light microscopy. A, B habitus, dorsolateral. C right femur I. Arrows to anterior macrosetae. D spinnerets, dorsolateral view. E, F, H right tarsi (E: III–IV, F: II, H: I), prolateral. Arrows to third claw. Note highly articulated podotarsite. G Chelicerae and palp, subdorsal view. Arrow to tip of right embolus. Scale bars: 1 mm (A), 0.5 mm (B), 0.2 mm (C), 0.1 mm (D, G), 0.05 mm (E, F, H).

After scanning the amber piece containing the holotype of L. aculicaput, we rendered a 3D reconstruction of the specimen (Figs 4, 5) that revealed additional details of key ventral structures of the spider, such as a notched labium (Fig. 4C, inset), a broad colulus (Fig. 5A, B) and the shape of both pedipalps (Fig. 5D–J). It also allowed better visualization of the spinnerets, which are disposed of in a compact group (Fig. 5B); the ALS are about the same length as the PLS (Figs 3D, 5C). The resolution did not resolve some of the smaller structures, such as the number of tarsal claws (although these can be seen in Fig. 3E, F), the presence and number of cheliceral teeth, or the presence of a double row of teeth in the first and second prolateral claws of the legs. The 3D rendering of the genitalia shows some aberrant structures (marked with an asterisk in Fig. 5) that we interpret as artifacts (perhaps thin layers of air surrounding the specimen) since each is only present in a single palp and similar structures are lacking in related species (see Fig. 7).

Figure 4. 

Drymusa aculicaput (Wunderlich, 2004) comb. nov., holotype (SMNG 07/36287-422), rendered volume after image stack obtained with synchrotron radiation micro-computed tomography. A, B habitus, dorsal. C habitus, ventral. Inset showing notched labium. D clypeus, subanterior. Scale bars: 1 mm (A), 0.5 mm (B–D).

Figure 5. 

Drymusa aculicaput (Wunderlich, 2004) comb. nov., holotype (SMNG 07/36287-422), rendered volume after image stack obtained with synchrotron radiation micro-computed tomography. Spinnerets (A–C), left palp (D–H), and right palp (I, J). A ventral. B apical. C dorsal. D subprolateral. E retrolateral. F prolateral. G bulb and cymbium, apical. H bulb and cymbium, dorsal. I prolateral. J retrolateral. Asterisks are artifacts in the rendering. Scale bars: 0.1 mm (only A–C, F, I to scale).

The target-gene and UCE datasets produced similar trees (Figs S1, S2), with higher support in the last case, as expected. The combination of all sequence data (Fig. S3) produced a tree with all the groups in agreement with the UCE analysis. The morphological dataset produced trees with the same resolution for Sicariidae, Scytodidae, Periegopidae, and Drymusidae, but discordant with the molecular data in the more basal splits, also differing between maximum likelihood (ML) and maximum parsimony (MP) analyses (Fig. S4). The ML and MP analyses of the complete dataset produced very similar results, only differing in three polytomies of the MP tree, that were resolved in the ML tree (Fig. 6; Fig. S5). Because these resolved branches in the ML tree had bootstrap below 75% and no molecular or morphological synapomorphies, we base our subsequent analyses on the phylogeny with those branches collapsed (Fig. 6). The morphological matrix, DNA alignments, and the total evidence tree found under ML are available as Supplementary material 2, 3, and 4, respectively.

Figure 6. 

Relationships among Scytodoidea inferred under maximum likelihood using a combined dataset (morphology + target genes + ultraconserved elements). Taxa with “ch” in their names are chimeras formed by combining morphological data of a species with sequence data of a congener. Bootstrap percentages are above branches, apomorphies below branches (state 0 in regular font, state 1 in bold font, states 2–3 indicated below character number).

The total evidence analysis (Fig. 6) recovered a monophyletic Scytodoidea including Ochyroceratidae, Psilodercidae, and Scytodidae as three successive splits in the base of the superfamily. Sicariidae is recovered including Sicariinae (Hexophthalma + Sicarius) and two extant species of Loxosceles grouped with the fossil L. defecta. This family is recovered as sister to Periegopidae + Drymusidae. Within Drymusidae, the South African Izithunzi Labarque, Pérez-González and Griswold, 2018 is recovered as the sister group to a clade containing American Drymusa Simon, 1892 and Loxosceles aculicaput. This latter species is recovered in a clade of Antillean species (D. armasi Alayón, 1981 and D. spectata Alayón, 1981 from Cuba).

Sicariidae is supported by five unambiguous morphological apomorphies, only one of which is observable in the fossils (third tarsal claw absent; Fig. 1G; Labarque and Ramírez 2012: fig. 8). Loxosceles is supported in our tree by a single unambiguous morphological apomorphy (cheliceral lamina with bifid apex; Fig. 1D; Magalhaes et al. 2017a: fig. 14A); two additional potential synapomorphies from spigots, chars. 73-1 and 82-1, were not visible in the fossil species. Drymusidae is supported by the promarginal lobe of the chelicerae with a distal prolongation (Labarque et al. 2018: fig. 3D), and the presence of two major ampullate gland spigots (Labarque et al. 2018: fig. 4C). Drymusa is supported by short, conical setae in the promarginal lobe of the chelicerae (see Labarque and Ramírez 2012) and the presence of a single aciniform gland spigot on the posterior lateral spinnerets (Fig. 7H). Finally, we recover a clade containing the extant D. armasi and D. spectata (both from the Antilles) and L. aculicaput, supported by three synapomorphies: the broad, plate-like colulus (Figs 5A, 7I), femoral macrosetae (Figs 3C, 7A) and the notched labium (Figs 4C, 7B).

Figure 7. 

Extant drymusids from the Greater Antilles: Drymusa spectata Alayón (A, D) Drymusa armasi Alayón (B, C, F), and Drymusa simoni Bryant (E, G–I). A: male habitus, ventral (MNHNCu 51). Arrow to femoral macrosetae. B: female, mouthparts, ventral (MNHNCu 28). Arrow to notch in labium. C: female (MNHNCu 28), cheliceral lobe, arrow to distal prolongation. D: left palp, prolateral (MNHNCu 51). E: holotype male, left palp, prolateral (MCZ 23101). F: right palp, retrolateral (MNHNCu 28). G: immature paratype, left tarsal claws II, retrolateral (MCZ 44196). Notice double row of teeth in the proclaw. H: immature paratype, spinnerets, apical (MCZ 44196). Inset showing single spigot in the right posterior lateral spinneret. I: immature paratype, colulus, ventral (MCZ 44196). Scale bars: 1 mm (A), 0.5 mm (B, D–F), 0.05 mm (C, G–I).

Figure 8. 

Loxosceles cubana Gertsch, male (IBSP 164702), left leg I. Inset in A marking area with retrolateral macrosetae on the tibia. A, B Retrolateral view. C Prolateral view. Scale bars: 1 mm.

All three species are redescribed and re-diagnosed below. Based on the phylogenetic results, we propose the transfer of Loxosceles aculicaput from Sicariidae to Drymusa in Drymusidae, resulting in Drymusa aculicaput (Wunderlich, 2004) comb. nov. and new family assignment.

The phylogenetic results we obtained are consistent with those of the most recent phylogeny with a broad sampling of Synspermiata families (Ramírez et al. 2021; ultra-conserved elements), although that study lacked Ochyroceratidae in its sampling. Our findings partially contradict the results found with morphology by Labarque and Ramírez (2012), who recovered Scytodidae (rather than Sicariidae) as closer to Periegopidae + Drymusidae; because our dataset includes a morphological matrix almost identical to Labarque and Ramírez’s, it seems that the signal of the massive number of molecular markers (1912 markers) overrides that of morphological data (79 variant characters). Our results also contrast with those of Labarque et al. (2018; Sanger sequences) and Li et al. (2020; transcriptomes), who obtained Ochyroceratidae as sister to Scytodidae. It seems the phylogenetic position of ochyroceratids and psilodercids is far from settled (e.g., see the low bootstrap below Althepus maechamensis in the parsimony analysis, Fig. S5), although the evidence that they are indeed members of Scytodoidea is mounting. Discussing synapomorphies for the extended Scytodoidea (including Ochyroceratidae and Psilodercidae) is difficult since the morphology of both families needs further study. We recovered a single morphological synapomorphy for Scytodoidea (including Ochyroceratidae and Psilodercidae), the fused 3^rd abdominal entapophyses. This might be an artifact of our taxon sampling, since some members of the Pholcoidea (the sister group of Scytodoidea) present this character state as well; also, it is open to interpretation, since the entapophyses are vestigial in Ochyroceratidae, Psilodercidae and Pholcoidea. The following node (Psilodercidae + the remaining Scytodoidea families) has a single synapomorphy: the simple (as opposed to branched) lateral tracheae in the posterior respiratory system.

We tested the position of Loxosceles defecta in the phylogeny of Scytodoidea and found it to be a true member of Loxosceles, supported in our analysis by the bifid cheliceral lamina (Fig. 1D). Other characters are also consistent with this genus, such as the absence of a third tarsal claw (Fig. 1G). Loxosceles is divided into several species groups (see Gertsch 1967, Gertsch and Ennik 1983) and we strived to place the fossil into one of those groups by comparing its morphology to that of extant species (L. simillima, L. hirsuta, L. laeta, L. rufescens, L. deserta, L. cubana, and L. caribbaea, each belonging to a different clade; see Binford et al. 2008, Magalhaes et al. 2019). The cheliceral stridulatory files of L. defecta consist of relatively deep and well-delimited ridges (Fig. 1D) when compared to the shallow scales that had been documented for L. rufescens (see Labarque and Ramírez 2012: fig. 16E), and thus this was a potentially useful character. However, examination of several other Loxosceles revealed that deep ridges are the most common state in this genus, and only L. rufescens (among the species examined) presents shallow scales. Another potential synapomorphic state is the prolaterally expanded cymbium (Fig. 1I, arrow). Gertsch and Ennik (1983: 280) noted the “tarsus broadly lobed on the prolateral side” as diagnostic of the L. reclusa species group. Confirming this, we only observed this character state in L. reclusa, L. cubana, L. taino, L. caribbaea and L. defecta, suggesting it might be a synapomorphy of the reclusa group and indicating the fossil species belongs here (the prolateral shape of the cymbium is not observable in L. deformis, see Fig. 2D–F). Finally, L. defecta shows a few unusual features: the first tibia is sinuous, distally incrassate, and bears macrosetae (Fig. 1E–F). All Loxosceles species examined by us have straight tibiae; only Loxosceles laeta has a sinuous first metatarsus with strong setae. The only known species with a sinuous tibia is L. carinhanha Bertani, von Schimonsky et Gallão, 2018, from Brazil (see Bertani et al. 2018: fig. 34), but it apparently belongs to the distantly related amazonica group, although its phylogenetic position remains untested. Thus, it might be that this character state evolved independently in L. carinhanha and L. defecta. Regarding the strong macrosetae, the only species examined by us whose tibia bears strong macrosetae is L. cubana (Fig. 8). Loxosceles caribbaea and L. taino (examined by us) and other members of the reclusa group from the mainland (A. Valdez-Mondragón, pers. comm.) lack such tibial macrosetae, although a few species present strong macrosetae in the first femur. Thus, the strong macrosetae in the tibia may be a synapomorphy uniting L. defecta with L. cubana, and indicate the fossil could be more closely related to the Antillean species of the reclusa species group.

We could not include Loxosceles deformis in the phylogeny, since the only known individual is poorly preserved and details of the legs, chelicerae, and spinnerets cannot be observed. However, the genitalia is similar to that of L. defecta, and it has the flattened, ribbon-shaped embolus that is common in the Loxosceles reclusa species group (see Gertsch and Ennik 1983). It also lacks the third claw and has an elongate palpal femur that is synapomorphic of Loxosceles (see Magalhaes et al. 2017a). Thus, we conclude that it is a true member of the genus, most likely in the Loxosceles reclusa species group.

We here transfer Loxosceles aculicaput to Drymusa. This species presents three tarsal claws (Fig. 3F), a well-developed podotarsite (Fig. 3F), relatively short anterior lateral spinnerets (Fig. 3D), and the spinnerets are in a compact group (Fig. 4B). All of these characters dispute a placement in Loxosceles, which presents only two tarsal claws in a poorly developed podotarsite (Labarque and Ramírez 2012: fig. 5A, B), relatively long anterior spinnerets, and a diastema between the anterior and posterior spinnerets (Magalhaes et al. 2017b: fig. 1). Our phylogenetic results (Fig. 6) suggest a placement in Drymusa instead, particularly close to extant Antillean species, based on three characters: the broad colulus (Figs 5A, 7I), the femoral macrosetae (Figs 3C, 7A) and the notched labium (Figs 4C, 7B). Interestingly, two of these characters were revealed by imaging the piece with SRµCT, highlighting the usefulness of this technique in the study of fossils. The genitalia of Drymusa aculicaput comb. nov. are here studied in detail for the first time (Fig. 5D–J). They are similar to other species of Drymusa, including the Antillean representatives, in the short, incrassate tibia and the cymbium with a small apical extension (see Fig. 7D–F), but are clearly different from the extant Hispaniolan species Drymusa simoni (see 5.2.1).

Penney (1999) observed that extant, endemic species of Drymusa can be found today in Hispaniola and that fossils would shed light on the timing of arrival of this group to the islands. The discovery that Drymusa aculicaput comb. nov. belongs in this genus indicates the genus arrived at Hispaniola during (or before) the Miocene. It is also a crown Drymusa, showing the genus had diversified by this epoch, and thus that it is an ancient group. It is not unlikely that the separation between the African Izithunzi and the American Drymusa pre-dates the separation of these continents, as is the case of Sicarius and Loxosceles (Binford et al. 2008; Magalhaes et al. 2019). A more complete phylogeny of drymusids, building on previous efforts by Labarque et al. (2018), will be useful to put this hypothesis to test, and will certainly benefit from the new morphological data on D. aculicaput comb. nov. gathered here.

Our phylogenetic and taxonomic revision of spider taxa preserved in Dominican amber and previously assigned to Loxosceles allows us to conclude:

(1) Loxosceles defecta and L. deformis are bona fide members of this genus, and L. defecta can be further placed in the L. reclusa species group based on the prolaterally expanded palpal cymbium; this species also presents an incrassate first tibia bearing macrosetae. Tibial macrosetae are also present in the extant species Loxosceles cubana, suggesting that L. defecta may be closely related to extant taxa from the Antilles.

(2) Loxosceles aculicaput presents three claws and shares a broad colulus, femoral macrosetae, and a notched labium with extant members of Drymusa inhabiting the Antilles, and thus was misplaced in Loxosceles; we here propose the new combination Drymusa aculicaput, representing the first known fossil of Drymusidae. This extends the stratigraphic range of crown Drymusa to the Miocene and indicates these spiders had already reached the islands by then.