Once characterized as rare1, the net-casting spiders (family: Deinopidae; C.L. Koch, 1850) are cryptic, challenging to collect, and historically under-sampled and understudied. Living deep in understory habitats throughout tropical and subtropical regions2, deinopids were seldom collected1 and their perceived rarity is likely due to inattention and their unusual habits. Therefore, despite being charismatic spiders admired by most arachnologists, the difficulty in sampling them has resulted in very poor knowledge of the group’s evolutionary history. Deinopids, like the majority of web-spinning spiders, are sit-and-wait predators; however, they have a unique hunting strategy. Instead of sitting in their web, these spiders employ a modified orb web3 that they manipulate using their two anterior leg pairs. Using a silken net spun of a fuzzy, mechanical-capture silk (cribellate silk) rather than gluey silk, deinopids remain motionless until a prospective prey comes into view. They lunge into a forward strike, extending their webs with their legs and then envelope their prey, rendering them immobile1,4,5,6,7. They are also capable of a backward strike detecting vibrations of flying prey, hence are not entirely reliant on their enlarged eyes for prey capture7.

Historically, phylogenetic analyses based on morphology have divided the family into two genera: Deinopis, the ogre-faced spiders, aptly named for their massive posterior median eyes (PMEs), and Menneus, the humpback spiders1. Both genera share the unique net-casting hunting strategy, yet only Deinopis species have distinctively enlarged PMEs. Eye size, function, orientation, and visual field overlap all contribute to how spiders perceive visual signals. While some spiders have visual fields that span 360°8, the spider optical system is highly diversified across taxa and typically aids in prey recognition, hunting, predator avoidance, mating, and courtship9,10,11. There are two eye types, the principal eyes (typically three pairs), which capture light, and a pair of eyes (the PMEs) which are resolution-based10,12. In deinopids, the PMEs are forward-facing with low visual acuity and are primarily responsible for detecting motion10. Remarkably, Deinopis PMEs are the largest simple eyes of any arthropod and are 2000 × more sensitive to light than human photoreceptors13. They are a particularly important feature that enables visual hunting at night in low-light conditions14. Although other characters differentiate the two nominal genera, including abdominal tubercles, genitalic features, and their geographic distributions, deinopids have been notoriously difficult to diagnose based on somatic characters alone, particularly to species level1.

Molecular phylogenetic treatments of Deinopidae have focused on species-level relationships within Deinopis15 and a newly described genus in Asia, Asianopis16, formerly Deinopis. Molecular data revealed that Eastern Hemisphere Deinopis are more closely related to Menneus from South Africa than to Western Hemisphere Deinopis15. Blest et al.17 postulated that Menneus PMEs were a plausible ancestral form of the enlarged Deinopis PMEs. Deinopis paraphyly contradicts a morphology-based phylogeny of deinopids1, and foments questions: did deinopids gain their large PMEs within the clade, or ancestrally followed by subsequent loss, and when did this shift in PME size occur? Furthermore, did this shift in PME size occur once or multiple times across the evolutionary history of deinopids? Eye size reductions and even complete eye loss are typical in troglomorphic organisms, including crustaceans18,19, fish20,21, blind mole rats22, beetles23, and spiders24,25,26,27, where, in eternal darkness, energetically costly eyes may no longer offer fitness benefits28. Troglomorphic arthropods have been useful systems for studying the evolution and genetic underpinnings and regulation of eye size18,29,30; however, in arachnids, the origins, development, and evolution of eye size remains largely unexplored12,31. Alternatively, large PMEs could have evolved multiple times in parallel in deinopids. While energetically costly to maintain, large eye size has been associated with enhanced survival, fitness, and prey capture in vertebrates32,33,34,35 and invertebrates36,37,38.

Trait evolution must be considered across geologic time scales in order to better understand macroevolutionary patterns. Teasing apart the role of vicariance—the geographic separation of populations over time leading to divergent species—from long-distance, often overwater, dispersal (long-distance dispersal—LDD) (e.g.40,41,42,43,44) in generating species distributions is at the center of biogeographic studies. Rigorous methods and model comparison frameworks have allowed researchers to explicitly test dispersal hypotheses45. Disjunct distributions of Southern Hemisphere taxa have largely been attributed to vicariance driven by the breakup of Gondwana, the southern portion of Pangaea, in the Mesozoic46. Deinopids were present on Gondwana prior to the breakup of the supercontinent, with the separation of Africa and South America generating the Eastern and Western Hemisphere deinopid clades15. Both vicariance and LDD have been important in shaping the distribution of deinopids in the Western Hemisphere15. Vicariance and LDD have also shaped the distributions of deinopids across the Eastern Hemisphere; however, the timing of this dispersal has not yet been extensively tested. Eastern Hemisphere deinopids contain both large-eyed Deinopis and Asianopis and small-eyed Menneus, and they are distributed throughout South Africa, Madagascar, Southeastern Asia, Indomalaya, and Australia. Because the Chamberland et al.15 molecular phylogeny of Menneus only contained species from South Africa, we have yet to ascertain when and where Menneus hypothesized eye size reduction, or Deinopis and Asianopis eye size increase, evolved. Did this evolutionary shift in eye size occur on the same continent and then the taxa dispersed (Menneus is monophyletic; trait conservatism) or did these shifts occur multiple times on different continents (Menneus is polyphyletic; trait convergence/parallel evolution) (Fig. 1)? With new Menneus and Deinopis specimens from Australia, we were able to more thoroughly test the historical biogeography of Eastern Hemisphere deinopids, while exploring the evolution of PME size across geographic time scales.

Figure 1
figure 1

Main phylogenetic hypotheses of Deinopidae assessing the monophyly of Menneus. We included all alternative tree topologies, which include non-monophyly of Deinopis and Asianopis in Supplemental Fig. S1. Phylogenies are based on (a) morphological data, (b) incomplete molecular data. The two primary alternative hypotheses we tested with the addition of Menneus from Australia: (c) PME size is reduced once and (d) PME size is reduced twice, independently in Australian and South African Deinopis. Map was created using the base map form Wikimedia Commons ( and country colors were modified using Adobe Illustrator ( (Map: Lisa Chamberland).

In the study, we have integrated a taxon-rich COI dataset with a locus-rich UCE dataset to infer phylogenetic relationships within the family Deinopidae and explicitly test the monophyly of Menneus, Asianopis, and Deinopis. We expanded taxon sampling beyond Chamberland et al.15 to include Menneus specimens from South Africa and Australia and Deinopis from Australia, South Africa, Taiwan, Madagascar, and Mexico (Fig. 2). Specifically, we seek to infer the first genus-level phylogeny of the family and explicitly evaluate the evolution of eye reduction in Menneus. We tested the hypotheses that the ‘regular’ sized PMEs of Menneus either (1) represent an ancestral state, (2) evolved once from a common ancestor shared with Deinopis that then dispersed to other continents, or (3) arose independently from Deinopis in South Africa and Australia. Finally, we examine the ancient Gondwanan biogeographic history of Deinopidae within the context of the newly derived phylogenetic hypothesis.

Figure 2
figure 2

Sampling map of deinopids. Localities colored by current taxonomic genus. Map was created using the base map form Wikimedia Commons ( Map: Lisa Chamberland).



The taxon-rich cytochrome-oxidase I (COI) dataset comprised 258 deinopids and 4 Uloboridae outgroup individuals. Phylogenetic analysis of the COI dataset yielded weak nodal support for a number of important clades, including the placement of Menneus. To resolve the ambiguous phylogenetic relationships, we generated UCEs for 40 deinopid individuals across all major clades. We found strong support for Menneus monophyly and Deinopis paraphyly with the UCE and UCE + COI concatenated datasets (Figs. 3, 4; Table 1). Although the COI-only analyses did not support the monophyly of Menneus, nodal support was low (Supplementary Fig. S2). Furthermore, the tree topology tests on the UCE dataset supported the unconstrained tree, with Menneus monophyletic and Deinopis and Asianopis both paraphyletic, and rejected all four alternative tree topology hypotheses (see “Methods”; Table 1; Supplementary Table S1).

Figure 3
figure 3

Maximum-likelihood phylogenetic tree of Deinopidae inferred using the COI + UCE (75% occupancy) matrix. Clades are colored by geographic location. The concatenated dataset resolves the uncertain placement of Australian Menneus and strongly supports the monophyly of Menneus sister to Eastern Hemisphere Deinopis. Numbers at nodes indicate bootstrap support values. Black stars indicate proposed genera (Deinopis, Menneus, and Asianopis).

Figure 4
figure 4

Dated phylogeny inferred using mcmcTREE. Pie charts are colored by geographic range.

Table 1 Divergence times and average marginal likelihoods and 95% HPD of deinopids based on UCE + COI concatenated MCC phylogeny.

We found strong support for Eastern and a Western Hemisphere deinopid clades across all phylogenetic analyses (Figs. 3, 4). The Western Hemisphere clade contained a Caribbean clade nested within a South and Central American grade, which was consistent with the phylogeny from Chamberland et al.15. There were nine nominal and nine putative species nested within the Eastern Hemisphere clade, including: three Menneus species from South Africa and three from Australia; two Deinopis species from Madagascar, three from South Africa, three from Australia, and at least one from Mexico; and five Asianopis species, including an undescribed Asianopis species from Taiwan (Fig. 3). The relationships of nominal genera differed between the COI only analyses and the analyses that contained UCE data; however, bootstrap support at these nodes were low and were most likely the result of using a single mitochondrial gene47,48,49,50,51,52. The addition of the UCE backbone to the COI dataset resulted in higher bootstrap support at the nodes. Thus, for all subsequent inferences and interpretations, we only considered the UCE only and UCE + COI phylogenies. The Eastern Hemisphere deinopid clade contained Menneus, Asianopis, and paraphyletic Eastern hemisphere Deinopis (Fig. 3). African, Australian, and Malagasy Deinopis were each monophyletic. We found D. cylindrica and an undescribed species, D. sp Hell’s Gates within an African clade. There were also two undescribed species within the Malagasy clade, D. sp Perinet and D. sp Andasibe-Mantadia. Finally, all phylogenetic analyses indicated three Deinopis clades in Australia, including D. subrufa in Eastern Australia, D. unicolor in Western Australia, and an undescribed species in Western Australia (Figs. 3, 4).

There is a well-supported relationship of a group of Mexican Deinopis sister to South African Deinopis (Fig. 3; Supplementary Figs. S3 and S4), thus supporting at least one Mexican group in the Eastern Hemisphere Deinopis + Asianopis clade. Individuals from this clade share morphological traits that characterize Eastern Hemisphere Deinopis and Asianopis, including bulky (as opposed elongated and narrow) abdomens. A second Mexican clade was sister to the entire Eastern Deinopis + Asianopis clade; however, this result was weakly supported and these individuals were represented by COI data only (Supplementary Fig. S5).

Within Asianopis, we recovered UCE data for A. wangi and A. wuchoai. A. wangi, A. wuchaou, A. zhuanghaoyuni, and A. sp from Taiwan all formed a clade sister to Australian Deinopis. Asianopis was paraphyletic in the UCE + COI phylogenetic inference with A. liukuensis sister to an Australian Deinopis + Asianopis clade. We were unable to recover UCEs for A. liukuensis; however, the COI only and COI + UCE concatenated phylogenetic reconstructions nested Australian Deinopis within the Asianopis clade. The A. liukuensis clade comprised specimens from China and India from Lin et al.16 and one specimen from Taiwan from the current dataset. Still, in all analyses, Asianopis (excluding A. liukuensis) was sister to Australian Deinopis.

Both the UCE only and COI + UCE analyses supported Menneus as monophyletic and sister to Eastern Hemisphere Deinopis + Asianopis clade. Both African and Australian Menneus were reciprocally monophyletic with three putative species in each geographic clade. We recovered UCEs for M. camelus and an undescribed species of Menneus from Africa, but only COI data for M. capensis.

Divergence time estimates and biogeography

Eastern and Western Hemisphere deinopids diverged in the lower Cretaceous and had an ancestral range in Africa + Australia + Asia (Fig. 4, Table 1). Following a vicariant divergence of Eastern and Western deinopids, Menneus diverged from the rest of the Eastern hemisphere deinopids generating a sympatric distribution of South African Menneus and Deinopis (Fig. 4). These vicariant and sympatric divergences generated three distinct clades: Western Hemisphere Deinopis, Eastern Hemisphere Deinopis + Asianopis, and Menneus.

Results from the biogeographic analyses supported LDD and vicariant divergences within Eastern Hemisphere Deinopis + Asianopis (Fig. 4). There are two Mexican deinopid clades that diverge from the South African Deinopis, one in the Upper Cretaceous and a second in the Eocene. In the absence of UCE data for the earlier diverging Mexican clade, the mechanism and direction of the dispersal was uncertain. Following the divergence of Mexican and South African Deinopis, there were two subsequent vicariant divergences: (1) African and Malagasy Deinopis diverged, (2) Malagasy Deinopis diverged from India + Asian Asianopis. Finally, there was a single LDD event from Asia to Australia resulting in the Australian Deinopis clade (Fig. 4). Both the UCE only (A. liukuensis absent) and UCE + COI concatenated biogeographic analyses indicated an ancestral range of Asianopis + Australian Deinopis in Asia. The inferred ancestral range of Australian Menneus was South Africa + Australia and diverged from South Africa around 56 Ma (24.22–94.35, Ma HPD 95%) (Fig. 4, Table 1). Since the date of divergence post-dates when Africa split from Australia53, Menneus likely dispersed via LDD from South Africa to Australia (Fig. 4). The 95% highest posterior density (HPD)54 varied widely across nodes, and is likely a consequence of having only one fossil to date the phylogeny (Table 1, Fig. S5).

Ancestral state reconstruction

There was a single shift in PME size, from large to small, in Deinopidae at the point of Menneus diversification (Fig. 5a–c, Supplementary Fig. S7). PME size was highly conserved and had a strong phylogenetic signal estimated D (− 2.11). The average PME diameter (scaled to carapace width at the PLEs) and total ocular distance of adult females were significantly larger for Deinopis and Asianopis compared to Menneus (Fig. 5; Supplementary Fig. S8). The tests of PME/carapace ratio and raw averages for the Mann–Whitney non-parametric test of mean difference between large and typical size PMEs were both significant p < 0.0001. In the phylogenetic principal component analysis (pPCA), the first principal component (PC) (PC1, 90.2% of the variance) was strongly affected by carapace measurements: length (− 0.983), width at widest point (− 0.963), width at PLE (− 0.934). The second PC (PC2, 4.3% of the variance), with a negative loading on PME diameter (− 0.469) and PME row width (− 0.410) and positive loadings on anterior median eye (AME row width (+ 0.214) and carapace length (+ 0.163) (Supplementary Table S2). The directionality of the vectors in the biplots are close together, indicating these traits are highly positively correlated (Fig. 5d). PC1 reflected body size whereas PC2 reflected PME and AME size and eye row width; however, while the three genera generally clustered in the pPCA, the combination of morphological characters used did not strongly differentiate the three genera. There was no significant difference between PME size and total ocular distance between Deinopis and Asianopis; however, Menneus PMEs and total ocular distances were significantly smaller (Fig. 5e,f). Average AME size was highest in Menneus, although this was not significant (Fig. 5g).

Figure 5
figure 5

Summary of eye size data. (a) MCC phylogeny with branch width representing the log of PME to carapace ratio. The size of the dots represents raw eye diameter (mm) and are colored by eye type: PME (yellow), AME (blueish green), PLE (blue), ALE (yellowish green). (b) A. subrufa and (c) M. sp Gurragawee with eyes colored by type- also represented in dot phylogeny. (d) pPCA biplot of deinopid genus clusters (n = 29) based on carapace and eye dimensions. Clades, boxplots, and pPCA are colored by genus: Asianopis (magenta), Deinopis (orange); Menneus (purple). Boxplots represent (e) PME diameter scaled to carapace width, (f) total ocular distance, and (g) AME scaled to carapace width.


Monophyly of Menneus and evidence for secondary loss of enlarged PME

Phylogenetic data (UCE and UCE + COI datasets) strongly support the monophyly of Menneus, whereas Deinopis and Asianopsis are rendered paraphyletic (Figs. 3, 4). These findings contradict the prevailing taxonomic hypothesis of the genus Deinopis (MacLeay 1839), as well as the more recently proposed new genus Asianopis16. Instead, all analyses support a deep divergence between Western Hemisphere Deinopis and Eastern Hemisphere deinopids (Deinopis + Asianopis), and the UCE and UCE + COI phylogenies support divergence between Menneus and Eastern Hemisphere Deinopis. Recent taxonomic studies have transferred a number of the Asian Deinopis species into Asianopis16,55,56; however, Australian and Malagasy Deinopis have yet to be moved into Asianopis. Based on the geographic and taxonomic coverage of the family in this study, including the deep divergences between the Eastern Hemisphere Deinopis + Asianopis clade, the Menneus clade, and the paraphyly of Deinopis with respect to Asianopis, we propose the transfer (see Taxonomic Section below) of the Eastern Hemisphere Deinopis into Asianopis Lin et al.56, thus retaining the three genera within Deinopidae: Deinopis in the Western Hemisphere, Asianopis in the Eastern Hemisphere, and Menneus (Simon 1876). Based on the phylogenetic hypothesis, we also propose the transfer of a number of Central American and Mexican Deinopis to Asianopis. Although these results highlight the need for further taxonomic treatment of the family and the formulation of a new diagnosis of Asianopis, we see no other working alternative that would adequately reconcile these phylogenetic results with the current taxonomy. That is, considering Asianopis and Menneus as junior synonyms of Deinopsis would fail to appropriately acknowledge the family’s phylogenetic structure, complex and worldwide biogeography, remarkable morphological divergence, and lineage divergence over major evolutionary time scales. The error bars around all major divergences are large, which is likely due to limited data and the use of a single fossil and broad priors to date the phylogeny. Still, major divergences are consistent with the previously published dated phylogeny of deinopids15.

We found support for a single reduction of PME size in deinopids (Fig. 5a, Supplementary Figs. S7, S9), indicating enlarged PMEs were an ancestral deinopid trait with subsequent size reduction in Menneus. Although it has been postulated that Deinopis species evolved large PMEs through adaptation to hunt exclusively at night, leaving Menneus to hunt at dusk and twilight, empirical testing of this hypothesis has been limited to two Australian species: D. subrufus and M. unifasciatus57. Extensive field observations indicate that Deinopis and Menneus hunt at the same time in low-light conditions (pers. obsv.). The observation that both genera are hunting at the same time and in the same locality, using the same hunting strategy, raises the question—what are the potential ecological shifts and evolutionary tradeoffs driving morphological divergence? The reduction in Menneus eyes size could be interpreted as regressive evolution. Regressive evolution, the loss or reduction of non-functional, formally adaptive characters, has been repeatedly observed in the eyes of subterranean23,58 and cave dwelling organisms59, animals of the deep ocean, and among species that make the switch from diurnal to nocturnal habitats60,61,62. Alternatively, although PME size are reduced in Menneus, they are not non-functional. Analyses of the anatomical structures have indicated Deinopis PMEs are more sensitive than Menneus17; however the extent of the visual sensitivity between the two genera remain largely unknown and understudied.

Among organisms where regressive evolution has been observed, there are three primary, not mutually exclusive, hypotheses that seek to explain character reduction: (1) relaxed selection or neutral mutations—the relaxation of selection leading to reduction (e.g. loss of eyes, reduction of eye size)63,64,65; (2) natural selection, in which reduced visual structures or the absence of eyes are advantageous66,67,68; (3) indirect selection, pleiotropy, or ‘genetic hitchhiking’, in which the gene related to trait reduction is linked to other adaptive trait(s) (e.g. olfaction or taste)29,69,70,71. D. spinosa has been shown to use auditory cues to capture their prey7. Future studies could, for example, explore whether Menneus has evolved morphological and/or behavioral characters that are advantageous for capturing aerial prey (e.g. additional sensitive vibratory sensors such as trichobothria). Furthermore, in D. spinosa, individuals use their large eyes to hunt cursorial prey, which ultimately increases their dietary breadth14; however, enlarged PMEs are metabolically costly. Stafstrom et al.72 found that there was a potential tradeoff between brain size and PME size. Deinopis PMEs are more complex13,73 and twice as sensitive compared to Menneus PMEs17. We postulate that relaxed selection in environments where such eyes do not offer major fitness benefits may explain this hypothesized secondary loss. Explicit tests of differential time and mode of hunting and evaluation of the underlying genetics of phenotypic differences may help tease apart potential evolutionary mechanisms. Finally, we propose an alternative hypothesis in which Menneus have far more sensitive PMEs than spiders with similarly sized eyes. If there was strong selection for increased eye size, neural connections, and light detection mechanisms in Deinopis, an energy cost could be spared by reducing eye size but retaining the high sensitivity. In that case there is a secondary reduction in eye size, but not, a reversal to the ancestral spider condition. Future detailed studies on receptor function, light sensitive cells, and visual/neural system communication could find that taking that advanced eye and making it smaller in Menneus would be advantageous even with certain anatomical simplifications from Deinopis. Deinopids have tremendous potential as a unique system for testing these alternative hypotheses and for studying visual evolution.

Early Gondwanan vicariance followed by long-distance dispersals

The disjunct geographic distributions of deinopids can be attributed to Upper Cretaceous vicariance during the breakup of West Gondwana, followed by subsequent transoceanic dispersal events. Initially formed around 500 Ma, Gondwana, the southern portion of the supercontinent Pangaea, began breaking away in the Middle Jurassic around 170 Ma74,75,76,77. However, the continents did not all break apart in succession. Instead, continents were simultaneously separating, with phases of these overlapping divergent events lasting tens of millions of years. After the initial separation, land bridges persisted between continental masses, providing passageways for biotic exchange78. Even amongst taxa with low vagility, there is mounting evidence for early Gondwanan vicariance followed by subsequent LDD events in taxa, including plants79,80,81,82, cave shrimp83, freshwater fish84 and arachnids85,86. Of course, in the epochs immediately following continental breakup, the distances between major landmasses were shorter and thus dispersal among these would be more likely than among the continents in their current positions. Following these early divergences driven by the breakup of West Gondwana, we concluded that subsequent deinopid diversification occurred via long-distance, trans-oceanic dispersal events.

Dispersal into Asia and Australia

With a largely tropical and sub-tropical range, deinopids do not follow the typical anti-equatorial distributions across the Southern Hemisphere found in many Gondwanan lineages87,88,89,90,91,92,93; nonetheless, vestiges of the ancient Gondwanan connections are still reflected in the biogeographic history of deinopids. Following the divergence from Menneus, Eastern Hemisphere Deinopis/Asianopis dispersed from South Africa to Madagascar where they subsequently spread into Asia. The Deinopis/Asianopis lineage dispersed and diversified throughout Asia, with later dispersal to Australia via Indonesia. The timing of the divergences and biogeographic analyses suggest that Deinopis may have arrived in Asia from India after India collided with Southeast Asia in the Eocene (Out of India hypothesis). The duration of the hypothesized IndoMadagascar landmass is debated. These may have separated as early as 130–125 Ma, but remained connected to India until around 84 Ma94,95 after which it is thought that fracture zones, plateaus, periodically emerged and facilitated stepping stone dispersal between Madagascar and India96. Such hypothesized stepping stone dispersal from Africa into Asia via India and Madagascar has been found in taxa with low vagility97,98,99. Alternatively, Deinopis could have dispersed through Northern Africa and Eurasia followed by subsequent extinction similar to sand scorpions100; however, our molecular phylogeny shows that the Madagascar clade is more closely related to Asia and India than Africa, consistent with the IndoMadagascar plus stepping stone hypothesis. Because this molecular dataset only included COI data for A. luikenisis from China and Taiwan, further sampling throughout India and Southeast Asia will be required to more rigorously test this dispersal hypothesis. Alternatively, these data support the hypothesis that Menneus dispersed to Australia via a single, trans-oceanic dispersal event, although divergence times of Australian and African Menneus post-date an Out of India hypothesis. With 37% of the total 46 described deinopid species in the Eastern Hemisphere (putative and genetic transfers), much of Africa, Asia, and Indomalaya remains under-sampled. More extensive sampling is required, particularly of Madagascar and Indomalaya, to more completely understand Menneus biogeographic history.

Crossing Wallace’s line

Wallace’s line is one of the most striking separations of biomes in the world. Wallace101 posited that the fauna of South America and Africa, despite being separated by the Atlantic Ocean, were more similar than the faunas of Asia and Australia. Based on our results, deinopids dispersed from Africa to Asia and then from Indomalaya to Australia 91–52 Ma (Fig. 4), which may reveal intriguing implications of dispersal across Wallace’s line. Furthermore, the Australian Deinopis diverged from its sister Indomalayan clade around 72 Ma (Fig. 4). This divergence pre-dates the Indo-Australian archipelago (IAA) and a deep sea and a large trench separated the two regions until around 40 Ma when islands began forming102. Recently the IAA has been discovered to be a more common route of dispersal than previously predicted in terrestrial animals102,103,104. We postulate that upon broader sampling in this region, we may find range expansions across the IAA in the Late Eocene to Early Miocene when landmasses were forming102.

South Africa, Madagascar, and the Neotropics

We found evidence for at least two trans-oceanic dispersals of South African Deinopis. First, Malagasy Deinopis diverged from South African Deinopis in the Paleocene, post-dating Lower Cretaceous continental drift and vicariance. Madagascar separated from Africa and moved southeastward in the Middle to Early Late Jurassic (160–155 Ma)105,106 and was one of the first of the Gondwanan landmasses to break away from the supercontinent. Madagascar was in its final position relative to Africa approximately 120 Ma (118–130 Ma)107,108. The relative contributions of Gondwanan vicariance and long-distance dispersal among Malagasy lineages remain a source of debate96,109,110,111. Madagascar’s biodiversity primarily comprises endemic flora and fauna that have either been present since Gondwana or were a product of long-distance dispersal events in the Cenozoic, mostly from Africa (400 km away), but including all major biogeographic regions96.

Second, we found evidence for at least one trans-oceanic long-distance dispersal between South Africa and Mexico in the Eocene. Discoveries of such extreme trans-oceanic dispersals are becoming increasingly common41,99,112 and can be attributed to salt-water tolerance113, long-distance flight114,115,116, and wind-dispersal in plants99 and ballooning spiders117,118. Exchanges between Africa, the Western Indian Ocean, and the Neotropics have been especially well documented in tropical flora119,120,121,122,123,124,125. Dispersal events from the Neotropics to Africa are facilitated by westerly winds125,126, whereas dispersals from Africa to the Neotropics, the rarer of the two dispersal scenarios125, typically follow the trade winds126. While there was at least one dispersal event between Africa to the Mexico, further sampling of Eastern Hemisphere deinopids in the Neotropics is necessary to rigorously test dispersal hypotheses.


Deinopids have a rich biogeographic history characterized by complex and ancient patterns of vicariance coupled with long distance dispersal since their emergence in the Cretaceous. While LDD has played an important role in shaping the distributions of deinopids, these events are not so prevalent as to completely obfuscate ancient Gondwanan signatures. The evolutionary history of Deinopidae provides fascinating examples of eye size reduction and unexpected trans-oceanic dispersals. Whereas most studies in eye size evolution are troglomorphic adaptations in the form of reductions/loss of eyes, here, we document an example of a single reversal to ‘normal’ sized eyes from the uniquely derived condition of grossly enlarged PMEs in ogre-faced spiders. The PME size reductions in Menneus provides a framework for considering and testing hypotheses of the potential selective forces that might be driving these divergent morphologies.

Taxonomic section

Based on phylogenetic structure and biogeographical distributions we propose the following new combinations (* indicates those taxa included in this study): Asianopis anchietae (Brito Capello127) NEW COMBINATION; Asianopis aspectans Pocock128 NEW COMBINATION; Asianopis aurita Pickard-Cambridge129 NEW COMBINATION; Asianopis camela Thorell130 NEW COMBINATION; Asianopis cornigera Gerstaecker131 NEW COMBINATION; Asianopis cylindrica* Pocock132 NEW COMBINATION; Asianopis fasciata Koch133 NEW COMBINATION; Asianopis fasciculigera Simon134 NEW COMBINATION; Asianopis giltayi Lessert135 NEW COMBINATION; Asianopis guineensis Berland & Millot136 NEW COMBINATION; Asianopis kollari Doleschall137 NEW COMBINATION; Asianopis labangan Barrion-Dupo & Barrion138 NEW COMBINATION; Asianopis longipalpula Strand139 NEW COMBINATION; Asianopis luzonensis Barrion-Dupo & Barrion138 NEW COMBINATION; Asianopis madagascariensis Lenz140 NEW COMBINATION; Asianopis mediocris Kulczyński141 NEW COMBINATION; Asianopis ornata Pocock142 NEW COMBINATION; Asianopis ravida Koch143 NEW COMBINATION; Asianopis reticulata Rainbow144 NEW COMBINATION; Asianopis schomburgki Karsch145 NEW COMBINATION; Asianopis schoutedeni Giltay146 NEW COMBINATION; Asianopis subrufa* Koch143 NEW COMBINATION; Asianopis tabidus Koch133 NEW COMBINATION; Asianopis unicolor* Koch143 NEW COMBINATION.


Taxon sampling

We collected a total of 42 deinopid individuals (29 Deinopis and 13 Menneus) from the east coast of Australia in January 2019 using standard aerial search and vegetation beating methods described in Coddington et al.147. Most of the specimens were collected at night as deinopids are highly cryptic and are most easily spotted when their nets reflect a blue hue under headlamp lights. The specimens were fixed in 95% ethanol in the field and kept at − 20 to − 80 °C in the lab. Collaborators also provided additional Deinopis specimens from Western Australia, Taiwan, Mexico, Madagascar, South Africa, and Colombia.

DNA extractions, sequence generation, and data processing

Detailed methods for molecular methods, sequence editing, and read processing are available in the Supplemental Methods S1. We extracted DNA from leg tissue for 65 individuals using the OMEGA BIO-TEK E.Z.N.A. DNA extraction kit. We amplified COI: cytochrome c oxidase subunit I, a mitochondrial locus. We obtained COI sequence data for seven species from GenBank for outgroups, following148. We included existing deinopid COI data available on DRYAD149 and Genbank from Lin et al.16 for additional global Deinopis samples and Asianopis samples, respectively. See Supplementary Table S3 for full taxon sample lists, GPS locality information, and Genbank accession numbers. Sequences were aligned in MAFFT150 and then edited by eye and checked for stop codons in Mesquite151.

For UCE data, we sampled at least two individuals from each major clade and their geographic representation, including from African Menneus, Australian Menneus, African Deinopis, Australian Deinopis, Malagasy Deinopis, Asianopis, North and Central American Deinopis, and Caribbean Deinopis. UCE libraries were hybridized to the Spider probeset148 150 bp paired-end reads were sequenced on the HiSeq4K at the University of California Davis DNA Technologies (See Supplemental Methods S1 for more detail). We obtained UCE data for four outgroup species from Kulkarni et al.148 (NCBI Sequence Read Archive, BioProject PRJNA575576). UCE and COI matrices were concatenated in AMAS152 to generate the UCE + COI combined dataset.


We inferred phylogenetic trees for the four molecular datasets: COI-only (258 deinopid individuals, four outgroup individuals; 1279 base pairs), UCE-only (40 deinopids, four outgroup individuals; 75% occupancy matrix; 1018 loci), COI + UCE concatenated, and COI with a UCE backbone phylogeny using Maximum Likelihood153 and coalescence-based methods154. In IQ-TREE 2153, we implemented the built-in model finder to consider invariant site and Gamma rate heterogeneity under the greedy strategy with the AICc criterion to determine molecular substitution models for each codon position. The COI dataset was partitioned by codon using the GTR + I + G and the UCE dataset was partitioned by locus using the GTR model. We ran four separate analyses on each dataset with 1000 pseudoreplicates of ultrafast bootstrapping155,156 on the codon-partitioned COI and COI-backbone datasets, and the loci-partitioned UCE, COI + UCE concatenated datasets. We generated trees from individual UCE loci using IQ-TREE. These trees were used to infer a coalescence-based species tree in IQ-TREE and ASTRAL-II v5.7.1154.

Tree topology tests

We implemented tree topology tests in IQ-TREE 2153 to evaluate the relationships of Asianopis, Deinopis, and Menneus, and to further assess the monophyly of these three groups. We tested the unconstrained tree and four alternative tree topologies (Supplementary Fig. S1) in IQ-TREE using the RELL approximation157, which uses bootstrap proportions, p-values with the Shimodaira-Hasegawa test158 and AU test159, and expected likelihood weights160. Using the UCE dataset, we inferred species trees in IQ-TREE for the datasets constrained to four alternative hypotheses (1) Menneus is paraphyletic (clades = New World Deinopis, Australian Deinopis + Menneus; African Deinopis + Menneus; Malagasy Deinopis; Asianopis); (2) Menneus is monophyletic and Deinopis and Asianopis form a clade (clades = Menneus, all Deinopis + Asianopis); (3) Menneus, New World Deinopis, Old World Deinopis, and Asianopis are each monophyletic; (4) Menneus, all Deinopis, and Asianopis are monophyletic (clades = Menneus, Deinopis, Asianopis) (Supplementary Fig. S1).

Molecular dating

We implemented MCMCTree in the PAML v4.8161 package on the UCE only dataset to estimate divergence times within Deinopidae. Prior to the analyses, we removed redundant subspecies so that only one individual per species (with the least amount of missing data) remained. To set up the model and model parameters, we assigned a birth–death model for the tree and an UCLN clock model (uncorrelated relaxed clock). A GTR substitution model was used with flat Dirichlet prior distribution. Previous phylogenetic dating analyses have used the fossil Palaeomicromenneus lebanensis to calibrate the stem of Deinopidae15,148,163. However, the taxonomy of this fossil is dubious164, and it has been transferred into the extinct family Salticoididae165; therefore we did not use this fossil as a calibration point in the dating inferences. Instead, we employed a molecular calibration point from recent phylogenetic reconstruction on Araneomorphs166 to conservatively use maximum age of 150 Ma as a soft constraint for the Deinopidae crown. We used a single fossil, Seppo koponeni to calibrate a soft constraint with the minimum root age of Deinopidae + outgroups of 132.9 and a maximum of 250 Ma. We ran two independent Markov chain Monte Carlo (MCMC) runs for 20,000 samples, resampling every 10,000 iterations for a total of 200,000,000 iterations with a burn-in set to discard the first 20,000,000 iterations. Outputs were assessed for stationarity in Tracer v1.7.1167 and visualized using FigTree v1.4.4168.

Ancestral range estimates

We estimated the ancestral ranges of deinopids and performed biogeographic analyses in BioGeoBEARS169 using the UCE-only MCC tree inferred in MCMCTree. We used eight isolated biogeographic areas including Africa (F), Asia (A), Australia (U), Neotropics (S), Nearctic (N), Madagascar (M), India (I), and the Caribbean (C), five of which were based on updated zoological regions of Holt et al.170 and on the most recent global biogeographic studies on spiders171. These regions were also selected due to their high endemism, geologic histories, and land availability. We included a time slice model based on the breakup of the continents following descriptions in Seton et al.’s172 of the historical geology and modeled after recent invertebrate dispersal models43,44 (Supplementary Table S4; See Supplemental Methods S1). We tested these models under the Dispersal‐Extinction‐Cladogenesis (DEC) and with the (+ j) parameter, which considers for jump dispersal or founder event speciation169,174. We compared Akaike information criterion (AIC) and relative likelihood scores across all models (a natural log of 2 was considered significant)175. We used the MCMCTree-generated UCE and UCE + COI dated phylogeny with outgroups pruned to only include Deinopidae specimens.

Ancestral state reconstruction and eye measurements

Measurements of adult female specimens were taken with a Leica M205 C scope using the micrometer scale tools in the Leica Application Suite (LAS Version 4.13.0) (Supplementary Table S5; See Supplemental Methods S1). We measured the total ocular distance and individual diameters of PME, anterior median eye (AME), anterior lateral eye (ALE), posterior lateral eye (PLE) diameter for all adult females. We also measured carapace length, width, and carapace width at the PLEs. A Mann–Whitney non-parametric test of mean difference was used to determine whether there was a statistically significant difference between large and small PME. We used a phylogenetic principal components analysis (pPCA) in R using the function phyl.pca in the R package phytools176 to determine if eye and carapace size, ten and three characters respectively, were partitioned among the three genera: Deinopis, Asianopis, and Menneus. We used a pruned MCC tree, which included 29 species throughout the tree that were absent of missing data. Data points included carapace length and width, eye diameter of all four types of eyes (PME, AME, PLE, and ALE) and the eye row width across each of these. We also calculated total ocular distance, the sum diameter of all eight eyes. We generated biplots in R of the scores of the 23 total traits obtained from the pPCA.

We tested the phylogenetic signal of the binary trait for eye size, “large” for Deinopis and “small/typical” for Menneus, and for the outgroups. We calculated D, which is a permutation-based model that measures the phylogenetic signal strength in a binary trait assuming the underlying continuous trait is evolving under Brownian Motion177 and was appropriate for eye size, which is a variable and potentially continuous trait within each genus.

We mapped the states for discrete characters, large and small PME, along the species level MCC phylogeny using the ace function in ape178 and in phytools176. To estimate ancestral states for discrete PME characters, we employed a continuous time Markov chain (Mk) model to give character probability distributions of ancestral states at internal nodes in the tree. We also used stochastic character mapping, an MCMC approach to sample character histories from their probability distributions.