1. Introduction
Pholcidae are ubiquitous spiders in tropical and subtropical regions around the world. They occupy a wide range of different microhabitats, with corresponding variation in body size, body shape, leg length, and coloration (Eberle et al. 2018). More than 1,100 new species have been added to the family within the last 20 years, and assignments of new species to higher taxonomic ranks (subfamilies and genera) have usually been unproblematic. This is particularly true at the level of subfamilies, even though subfamily-level relationships and exact compositions of subfamilies have proven difficult to resolve in some cases. For example, relationships of Ninetinae to other subfamilies and the positions of the genera Artema Walckenaer, 1837 and Priscula Simon, 1893 at subfamily level have long been dubious (Huber 2000; Dimitrov et al. 2013; Eberle et al. 2018; Huber et al. 2018) and could only be resolved recently with genomic data (Meng et al. 2025).
Thus, the current subfamily division of Pholcidae is essentially an updated version of Simon’s (1893) classification, with numerous corrections and changes but without any new subfamily-level names (Huber 2011b). For more than a century, no new species have been discovered that would defy assignment to one of the subfamily-level taxa of Simon (1893) or their current emended versions. Here, we report on two newly discovered species for which genomic data suggest they are neither included into nor sister of an existing subfamily; thus, they require the creation of a new subfamily.
In addition, we formalize here the separation of Artema from ‘other Arteminae’, which was first suggested in Eberle et al. (2018) and which is strongly supported by new genomic data (Meng et al. 2025). We thus redefine Arteminae to include only Artema and Priscula and propose a new subfamily name for the ‘other Arteminae’ sensu Eberle et al. (2018) and Huber et al. (2018). Finally, we provide first molecular data for the Chilean genus Aucana Huber, because we initially thought (based on superficial similarity) that this genus might be the closest relative of the new Brazilian species.
4. Discussion
4.1. How reliable is the molecular phylogeny?
In our likelihood mapping analysis, approximately 32% of quartets are positioned at each of the three corners of the triangular graph, indicating a high level of phylogenetic signal in the dataset and the potential to resolve a robust phylogeny (Strimmer and von Haeseler 1997) (Fig. S3). Only 7.2% of the loci violate the assumptions of stationarity, homogeneity, or both (Table S3), and less than 0.6% of the contigs are affected by intra-locus recombination (Table S5). This suggests that the dataset is well-suited for phylogenetic inference. Furthermore, the coalescence-based methods (ASTRAL-IV and wASTRAL) enable us to mitigate the impact of incomplete lineage sorting (ILS) (Zhang and Mirarab 2022). The models GTR+R4+I+FO used in RaxML-NG and CAT-GTR in PhyloBayes-MPI are designed for handling rate and compositional heterogeneity (Lartillot et al. 2013; Kozlov et al. 2019), which alleviates potential influences of model violation.
Our phylogenetic inferences from 30 different analyses (five matrices × six tree construction methods) are largely congruent in terms of subfamily monophylies and inter-subfamily relationships (375 green squares / (30 × 13 panels) = 96%; see Fig. 2a). The monophylies of Pholcinae, Smeringopinae, Modisiminae, Ninetinae, Physocyclinaesubfam. nov., and Arteminae appear beyond doubt. The monophyly of the new subfamily Caipirinae is supported by all analytical methods in our study (Fig. 2a) with 100% bootstrap support (Fig. 1). Non-monophyly of subfamilies was found in only a few analyses: (1) Arteminae, when analyzed with PhyloBayes-MPI using data matrices 3–5 (3 out of 30 analyses), which may be due to an insufficient amount of data (total sites ranging from 9 to 20 Kbp), and (2) Modisiminae, when analyzed with SVDquartets using matrices 1–5. The latter case involves the uncertain placement of three genera (Arenita and two unnamed genera). In most analyses (25 out of 30 analyses), they were sister to all other Modisiminae (Fig. 1), while in a few cases (5 out of 30 analyses) they were sister to Physocyclinaesubfam. nov. (not shown), suggesting that further analysis of this poorly known group is necessary. The inconsistency between SVDquartets and other programs could stem from the fact that finding the species tree with maximum quartet support from a set of quartet trees is an NP-hard problem (Jiang et al. 2001). As SVDquartets uses a heuristic search strategy, it does not guarantee to find the optimal solution (Vachaspati and Warnow 2018).
Since a large percentage of our dataset is identical to that analyzed in Meng et al. (2025), it is unsurprising that the subfamily relationships recovered herein align well with those reported in Meng et al. (2025). Only a small fraction of the analyses (7 out of 90) did not support the monophyly of certain higher-level groups (Fig. 2a): (1) the clade ((Arteminae, Ninetinae), (Physocyclinae, Modisiminae)); (2) its sister-group relationship with the newly proposed subfamily Caipirinae; and (3) the entire family Pholcidae. Interestingly, all of these discrepancies arose from analyses based on the smaller datasets of matrices 4 and 5 (total sites: 15 Kbp and 9 Kbp, respectively). This suggests that the incongruences are likely due to the limited data in these specific matrices. Finally, although sparse taxon sampling can compromise phylogenetic inference, denser sampling – even when using short marker genes – has been shown to improve accuracy (Nichols et al. 2015; Kulkarni et al. 2023). In this context, Meng et al. (2025) analyzed combined genomic-scale data and six short marker genes for more than 900 pholcid species and recovered subfamily-level relationships congruent with our results.
4.2. Morphological aspects
The sister-group relationship between the two newly described species is strongly supported by molecular data but not immediately obvious in their morphology. The combined similarities listed in the diagnosis of the genus appear suitable for the particular purpose of differentiation from other genera, but rather weak in the context of phylogeny. Individually, most of these characters or traits do not seem to be unique among Pholcidae, i.e. they are unlikely candidates to be non-homoplastic synapomorphies. A possible exception is the transversal ridge on the male genital bulb (Figs 6d, e, 16d, e). For all other items listed in the diagnosis, similar characters and traits occur in other, often distantly related, genera. (1) Pointed male cheliceral apophyses and a corresponding pair of female epigynal pockets also occur in other genera like Belisana Thorell, 1898; Crossopriza Simon, 1893; Quamtana Huber, 2003; and Smeringopus Simon, 1890 (Huber 2003, 2005, 2012, 2022) but are absent in the sister group of Caipiragen. nov., i.e. in Physocyclinaesubfam. nov., Modisiminae, Arteminae, and Ninetinae. (2) A ventral or prolateral process on the male palpal tarsus, at the basis of procursus, exists in many Pholcidae, even though usually less distinct, as for example in Crossopriza, Smeringopina Kraus, 1957 and Wanniyala Huber and Benjamin, 2005 (Huber 2013, 2019, 2022). (3) Dark marks on the carapace are ubiquitous in Pholcidae and only useful as a quick distinguishing trait in comparison to superficially similar Brazilian Ninetinae. (4) A low ratio of male tibia 1 / tibia 2 (~1.05–1.15) is very rare among Pholcidae (Fig. 20), but similar values also occur in Ninetinae. (5) The absence of leg tarsal pseudosegmentation is difficult to evaluate since the pseudosegmentation of some Pholcidae is very indistinct, or the individual pseudosegments are ‘broken’ into many seemingly irregular small platelets (in some Smeringopinae and some Physocyclinaesubfam. nov.; Huber 2000, 2022).
On the other hand, the two species show some remarkable differences that are unusual among closely related and congeneric Pholcidae. The male palpal femur is particularly different: strongly enlarged in Caipira mineira (very similar to most Physocyclinaesubfam. nov., also Artema, some Priscula, some Mesabolivar; e.g. Aharon et al. 2017; Huber 2018; Huber & Villarreal 2020) but much smaller and with strong ventral process in C. baiana (ventral apophyses are common in some Modisiminae genera and usually relatively conserved among close relatives; e.g. Huber 2000). Noteworthy is also the different number of ALS spigots: Caipira mineira has the plesiomorphic set of eight spigots; C. baiana has the reduced set of only two spigots. A reduction of spigots has occurred repeatedly in Pholcidae, even though rarely within genera, e.g., in Belisana; Metagonia Simon, 1893; Pholcus Walckenaer, 1805; and Smeringopus (Huber 2000; 2005, 2011a, 2012). Male clypeus modifications have also evolved repeatedly in Pholcidae, sometimes within genera (Huber 2021), and are present in C. baiana but absent in C. mineira. Finally, the AME have been reduced repeatedly in Pholcidae, sometimes even within genera, as in Leptopholcus Simon, 1893; Mecolaesthus Simon, 1893; Modisimus Simon, 1893; Panjange Deeleman-Reinhold & Deeleman, 1983; Pholcus, Priscula, and Quamtana (Huber 2003, 2011a; Huber & Villarreal 2020); their presence in C. mineira but absence in C. baiana just adds one more reduction to the list. In sum, the discovery of further species may well justify a future split of Caipira into two genera, but currently we see no need and no justification to do so.
4.3. Notes on the biogeography of <tp:taxon-name><tp:taxon-name-part taxon-name-part-type="subfamily" reg="Caipirinae">Caipirinae</tp:taxon-name-part></tp:taxon-name>
The distribution of Caipirinaesubfam. nov. is restricted to limestone outcrop terrains in northeastern Brazil (Fig. 4). Each described species is recorded from a single biogeographic district of the Caatinga domain, separated by over 500 km. While C. mineirasp. nov. was collected in the Peruaçu district, C. baianasp. nov. was collected in the Irecê district (sensu Moro et al. 2024). These karstic districts are separated by the crystalline terrains of the Southern Depressão Sertaneja district (SDS in Fig. 4) and the Chapada Diamantina subprovince (CPD in Fig. 4). Pholcidae sampling in this region is sparse (see Fig. S4) and it is thus likely that other species or additional occurrences of the described species will be found in future focused campaigns. Likewise, the existence of a third limestone outcrop district in the Caatinga domain (i.e., the Potiguar district; see Moro et al. 2024) raises the possibility of additional Caipirinae taxa further north. It is worth noting that targeted pholcid sampling is fundamental to reduce this sampling bias, as previously demonstrated for Ninetinae taxa (see Huber et al. 2023b, 2024a).
As sister-group to a large lineage of Pholcidae hypothesized to have originated during the Cretaceous period (Meng et al. 2025), Caipirinaesubfam. nov. appears to be a relict taxon, as previously reported for other organisms in the Peruaçu district. A striking example is Relictopiolus galadriel Pérez-González, Monte and Bichuette, 2017, a troglobitic Kimulidae harvestman from another cave of the Parque Nacional Cavernas do Peruaçu (Pérez-González et al. 2017), the type-locality of C. baiana. Relictopiolus galadriel diverged from its epigean sister-taxon Tegipiolus pachypus Roewer, 1949 about 40 Mya, during the Eocene, corroborating the hypothesis of an ancient divergence between relictual taxa from this region and their sister-taxa (Pérez-González et al. 2017). These two species are the only currently known kimulids from eastern Brazil, with the remaining species recorded from northern South America and Central America (Pérez-González et al. 2017). The Peruaçu district also harbors several other troglobites, including other harvestmen, whip-spiders, pseudoscorpions, isopods, amphipods, crickets, heteropterans, centipedes, and fish (Pérez-González et al. 2017; Polhemus and Ferreira 2018; Monte and Bichuette 2020; Ázara et al. 2020; Cardoso et al. 2021). Furthermore, the richest hotspot of subterranean biodiversity in South America (i.e., the Água Clara cave system; see Ferreira et al. 2023), is in the Peruaçu district.
The floristic composition of the Peruaçu district is well-studied and distinct from the Caatinga sensu stricto and sandy Caatinga of other regions in the Caatinga domains, thus forming a particular sub-type of Caatinga, called ‘arboreal Caatinga’ (Moro et al. 2024). Conversely, the flora of the Irecê district is the least studied karstic region in the Caatinga domain (Moro et al. 2024). Both districts share the high abundance of caves, with several troglobites described for the Irecê district too, including other pholcids (Machado et al. 2011), as well as scorpions (Esposito et al. 2017), isopods (Cardoso et al. 2022), amphipods (Bueno et al. 2022), beetles (Gnaspini et al. 1997) and catfishes (Bockmann and Castro 2010). The Pholcidae species Metagonia diamantina Machado, Ferreira & Brescovit, 2011, described from the Irecê district (Ferreira et al. 2011), also represents a relict taxon. This species belongs to the potiguar species-group whose representatives are typically found in low primary productivity environments such as caves and high-altitude regions (Huber et al. 2022). This distribution suggests that species of the potiguar group of Metagonia may have been replaced by modern taxa in more favorable environments (Huber et al. 2022).
These abundant records of endemic and cave-associated taxa within both limestone outcrop districts underscore the importance of allopatric speciation in shaping species composition in this region. This pattern is corroborated by studies on other taxa, such as spiny orb-web spiders (Magalhaes et al. 2024) and frogs (Teixeira et al. 2012).