Spiders feeding on vertebrates is more common and widespread than previously thought, geographically and taxonomically

Abstract. According to a recent global literature survey, a total of 39 out of the 129 known spider families (∼30%) contain species capable of capturing vertebrate prey. The finding that the percentage of spider families engaged in vertebrate predation is so high is novel. Two groups of vertebrate-eating spiders are distinguished: “habitual vertebrate-eaters” vs. “occasional vertebrate-eaters”. The habitual vertebrate-eaters comprise ten spider families (Araneidae, Atracidae, Ctenidae, Lycosidae, Nephilidae, Pisauridae, Theraphosidae, Theridiidae, Trechaleidae, and Sparassidae) to which can be attributed 91% of all reported vertebrate predation incidents. The habitual vertebrate-eaters have evolved prey-capture adaptations such as (1) sufficient physical strength coupled with large body size, (2) the use of potent venoms, and (3) the use of highly efficient prey-catching webs. By contrast, unexpected feeding on vertebrates by the occasional vertebrate-eaters (i.e., Actinopodidae, Agelenidae, Amaurobiidae, Anyphaenidae, Barychelidae, Clubionidae, Corinnidae, Ctenizidae, Cyrtaucheniidae, Deinopidae, Desidae, Dipluridae, Eresidae, Filistatidae, Gnaphosidae, Haplonoproctidae, Linyphiidae, Liocranidae, Miturgidae, Oxyopidae, Pholcidae, Porrhothelidae, Salticidae, Selenopidae, Sicariidae, Sparassidae, Tetragnathidae, and Thomisidae) might be considered as chance events that took place when a tiny vertebrate crossed the path of an opportunistic spider. For a few families (e.g., Idiopidae) their status as habitual or occasional vertebrate predators is still unclear. In conclusion, our survey unveiled a large number of spider taxa previously not anticipated to feed on vertebrate prey. These findings improve our general understanding of spider feeding ecology and provide a first assessment of the significance of vertebrate prey as a food source for spiders.


INTRODUCTION
In biology text books of the past, the narrative was upheld that spiders would feed more or less exclusively on arthropods (Bristowe 1941;Gertsch 1979;Foelix 1982). Anecdotal observations of predation on unusual prey such as fish, frogs, lizards or birds were considered to represent either extremely rare chance events or doubtful reports. So for instance, Foelix (1982) stated in the first edition of his book Biology of Spiders ''. . ..It is rather unusual to find vertebrates among a spider's prey, although tadpoles or small fish may fall victim to certain lycosids or pisaurids. Reports that claim that large 'tarantulas' have been feeding on birds, lizards, or snakes are rarely scientifically verifiable.'' However, during the last 40 years much photographic evidence of vertebrate predation by spiders has been reported in the scientific literature and in the social media, providing irrefutable proof that this type of feeding behavior is not that uncommon in at least some spider groups (e.g., McCormick & Polis 1982;Menin et al. 2005;Toledo 2005;Nyffeler & Knörnschild 2013;Nyffeler & Pusey 2014;Folt & Lapinski 2017;Nyffeler & Vetter 2018;Babangenge et al. 2019;Nyffeler & Altig 2020;Reyes-Olivares et al. 2020;Nyffeler & Gibbons 2021). In the following, we provide an overview of geographically and taxonomically widespread vertebrate predation by different spider families.

METHODS
2.1 Data collection.-A global survey of scientific literature and social media reports, using the search method described in Nyffeler et al. (2017a), resulted in a total of 966 recorded incidents (source: McCormick & Polis 1982;Brooks 2012;Nyffeler & Knörnschild 2013;Nyffeler & Pusey 2014;Nyffeler et al. 2017aNyffeler & Vetter 2018;Weisberger 2019;Nyffeler & Altig 2020;Reyes-Olivares et al. 2020;Fulgence et al. 2021;Nyffeler & Gibbons 2021, 2022; Google Scholar & Google Picture Survey for Sparassidae feeding on vertebrates 2021). From this database, we extracted the following: -How many times each spider family had been engaged in vertebrate predation. By comparing the frequency with which the 39 spider families were engaged in vertebrate predation, it was possible to distinguish two groups: ''habitual vertebrateeaters'' vs. ''occasional vertebrate-eaters.'' -For each spider family how many of nine major geographic regions (Europe, Africa, Asia, Australia, North America, Central America, South America, the Caribbeans, and Antarctica) vertebrate predation had been reported in the scientific literature and social media. We ranked these to get an approximate measure of the geographical extent of this type of feeding behavior for each family of spiders.
-For each spider family, how many of five different vertebrate categories (i.e., amphibians, reptiles, fish, birds, and mammals; Table 1) were part of the diet. This information provides a rough measure for dietary diversity of a particular spider family.
2.2 Statistical methods.-The two-tailed Mann-Whitney U test (see MacFarland & Yates 2016) was applied to examine whether mean values differed statistically significantly. These analyses were performed online at https://www.socscistatistics. com/tests/mannwhitney/default2.aspx. Prior to that, we examined with the Shapiro-Wilk test software whether the data were normally distributed (Online at http://www. statskingdom.com/320ShapiroWilk.html). As the data gathered in our survey were not normally distributed, we used a nonparametric Mann-Whitney U test. Mean values are followed by Standard Errors (6 SE).
2.3 Taxonomic comments.-Nomenclature of spider taxa was based on the World Spider Catalog 2022 except for the nephilids. Contrary to the World Spider Catalog 2022, we placed the genera Nephila Leach, 1815, Nephilingis Kuntner, 2013, andTrichonephila Dahl, 1911 in the family Nephilidae (sensu Kuntner et al. 2019). Argyroneta aquatica (Clerck, 1757) previously placed in the family Cybaeidae has been transferred to Dictynidae (World Spider Catalog 2022).
We recognized two groups of vertebrate-eating spiders, namely those that are habitually engaged in vertebrate predation (i.e., ''habitual vertebrate-eaters''), and those that feed on vertebrates only occasionally (i.e., ''occasional vertebrate-eaters''). In the following, the two groups will be examined comparatively.
3.3 Habitual vertebrate-eaters vs. occasional vertebrateeaters.-There is compelling evidence that spiders representing ten families are habitually engaged in vertebrate predation. This group includes the Araneidae, Atracidae, Ctenidae, Lycosidae, Nephilidae, Pisauridae, Theraphosidae, Theridiidae, Trechaleidae, and Sparassidae ( Fig. 1). The assignment of these ten families to the group of habitual vertebrate-eaters is based on the fact that together they represent 91% of all recorded incidents of vertebrate predation (Fig. 1). The atracids were included in the group of habitual vertebrateeaters despite the fact that the number of vertebrate prey records available for this family was not as high (n ¼ 20 records; Fig. 1) as in the other families from this group (n .40 records; Fig. 1). Including the atracids is justified by the fact that countless frog bones have been detected in the funnels of the northern tree-dwelling funnel-web spider Hadronyche formidabilis (Rainbow, 1914) in New South Wales, Australia. These were not taken into account in Fig. 1 because it remained unclear how many frog speciments these bones represented. Be that as it may, the information presented by McKeown (1952) strongly suggests that at least this particular atracid species is predominantly a frog eater. With regard to their phylogenetic origins, the ten families can be assigned to three taxonomic groups: the RTA clade (i.e., families Ctenidae, Lycosidae, Pisauridae, Trechaleidae, and Sparassidae), the Araneoidea clade (i.e., families Araneidae, Neph-  Fig. 2. A second group of spider families is also engaged in vertebrate predation, but vertebrate-eating has been reported to occur only occasionally. This group (made up of the 28 families Actinopodidae, Agelenidae, Amaurobiidae, Anyphaenidae, Barychelidae, Clubionidae, Corinnidae, Ctenizidae, Cyrtaucheniidae, Deinopidae, Desidae, Dipluridae, Eresidae, Filistatidae, Gnaphosidae, Haplonoproctidae, Liny-phiidae, Liocranidae, Miturgidae, Oxyopidae, Pholcidae, Porrhothelidae, Salticidae, Selenopidae, Sicariidae, Sparassidae, Tetragnathidae, and Thomisidae) accounts for , 10% of all recorded incidents of vertebrate predation (Fig. 1). Spider species that occasionally consume vertebrate prey tend to be medium-sized (mean ¼ 16.54 6 1.32 mm; median ¼ 15 mm; Tables 1, 2). Representatives of this latter group are depicted in Figs. 4-6. In most of the spider families engaged in habitual vertebrate-eating, this type of feeding behavior has a wide global distribution, whereas in the group of occasional vertebrate-eaters this behavior was found to be less widespread; the difference in the mean number of geographic  regions was statistically significant (Mann-Whitney U test, n 1 ¼ 10, n 2 ¼ 28, Z ¼ -4.59123, P , 0.0001; Table 2). Furthermore, the habitually vertebrate-eating spider families had a significantly more diverse diet (in terms of the number of different vertebrate prey categories included in each spider family's diet) compared with the occasional vertebrate-eaters (Mann-Whitney U test, n 1 ¼ 10, n 2 ¼ 28, Z ¼ -3.97796, P , 0.0001; Table 2).
The status of several spider families as vertebrate-eaters is still undecided. For instance, Australian idiopids and the Floridian jumping spider Phidippus regius C. L. Koch, 1846 may eventually be determined to be habitual vertebrate-eaters as both have repeatedly been reported to feed on small vertebrates (see McKeown 1952;Butler & Main 1959;Main 1996;Nyffeler et al. 2017a. Nevertheless, in this paper the idiopids were not assigned to either group, and salticids were assigned to the occasional vertebrate-eaters because of a lack of sufficient quantitative data on these two spider families' natural diets (see Appendix 1). The same applies to some anyphaenid species in the genera Katissa Brescovit, 1997 and Patrera Simon, 1903. Neotropical anyphaenids have repeatedly been observed to devour eggs/embryos of glass frogs (Centrolenidae) suggesting that these spiders might be habitual predators of frog eggs (see Nyffeler & Gibbons 2022;Fig. 5). In the present study, the anyphaenids were assigned to the occasional vertebrate-eaters because of insufficient evidence. Many more extensive quantitative investigations on the feeding habits of spiders from these last mentioned families are needed before their status as habitual or occasional vertebrateeaters can unambiguously be determined.

DISCUSSION
4.1 Prey-capture adaptations of the habitual vertebrateeaters.-The ten spider families considered to be habitual vertebrate-eaters ( Fig. 1) appear to be adapted to the capture of vertebrate prey in a number of different ways (see McCormick & Polis 1982). The most important prey capture adaptations are (1) sufficient physical strength coupled with large body size, (2) the use of potent venoms, and (3) the use of prey-catching webs (Escoubas & Rash 2004;Lüddecke et al. 2022). Taking into account that vertebrates usually are large, bulky prey compared to the size of their spider attackers, a certain physical strength coupled with large enough body size is a prerequisite for spiders capable of subduing vertebrates (Escoubas & Rash 2004). Indeed the habitual vertebrate-eaters are, on average, significantly larger than the occasional vertebrate-eaters (Table 2; Mann-Whitney U test, n 1 ¼ 10, n 2 ¼ 28, Z ¼ -2.86745, P , 0.01). This is illustrated by the examples of the goliath birdeater Theraphosa blondi (Latreille, 1804) (Theraphosidae), the giant huntsman spider Heteropoda maxima Jäger, 2001 (Sparassidae), and the giant fishing spider Ancylometes rufus (Walckenaer, 1837) (Ctenidae) -three habitual vertebrate-eaters belonging to the world's largest spiders. An exception to the rule are the widow spiders (Latrodectus spp.; Theridiidae) as well as other members of the theridiid family, which habitually kill and devour vertebrate prey despite the fact that they are of only small to moderate size (usually~9-13 mm in the case of the widow spiders; Fig.  2E; Table 1). The smallest theridiid witnessed killing and eating a tiny hatchling lizard had a body length of only 3 mm (Welter & Fauth 1996). The widow spiders compensate for their smaller size by constructing webs of extreme stickiness and extraordinary strength and by their use of a highly potent vertebrate-specific toxin (see below). Another adaptation is the use of potent venom to paralyze and kill small vertebrates. All 39 spider families engaged in vertebrate-eating ( Fig. 1; Table 1) are equipped with venom glands and they all use venom to immobilize their prey (Foelix 2011;Lüddecke et al. 2022). Some unique properties specifically targetting vertebrate nervous systems have been detected in the venoms of at least seven prominent spider families engaged in vertebrate-eating (atracids, ctenids, lycosids, pisaurids, theraphosids, theridiids, and trechaleids (see Herzig & Hodgson 2009;Wang et al. 2013;Valenzuela-Rojas et al. 2019, among others.)). The best known of the vertebratespecific neurotoxins is the highly potent a-latrotoxin of the widow spiders (Latrodectus spp.; Nyffeler & Vetter 2018).
Still another adaptation is the use of prey-catching webs. Enders (1975) theorized that the use of webs increases the size of prey that spiders can capture. According to our survey ( Fig.1), half of the 966 recorded incidents of vertebrate predation were by web-weaving spiders and half by webless species. Only three families -namely the Araneidae, Nephilidae, and Theridiidae -construct webs of high tensile strength (i.e., ''high energy absorbing webs'' sensu Craig 1987) particularly suitable for catching large, bulky prey including small vertebrates and large arthropods. Spiders from these three families construct webs of silks considered to be the toughest known biological materials (Agnarsson et al. 2010;Babb et al. 2017). In addition, these webs contain sticky capture threads whose function is to retain prey long enough to be located and attacked by the spiders (Agnarsson & Blackledge 2009). The exceptionally strong webs of the araneids/nephilids permit the capture of large beetles and cicadas as well as small bats and birds ( Figs  Knörnschild 2013), whereas widow spiders (Theridiidae) are capable of trapping large beetles, Jerusalem crickets, scorpions, small snakes, lizards, and even mice and rats ( Fig. 2E; Gertsch 1979;Nyffeler & Vetter 2018;Nyffeler & Gibbons 2021).
By means of these types of prey-capture adaptations, habitual vertebrate predators among the spiders are able to not only kill vertebrate prey many times larger and heavier than themselves, they also are able to subdue highly toxic predaceous vertebrates such as Australian brown snakes (Pseudonaja spp.), rattlesnakes (Crotalus spp.), Neotropical lanceheads (Bothrops spp.), and New World coral snakes (Micrurus and Micruroides spp.; e.g., Fig. 2E; Nyffeler & Gibbons 2021).

Prey capture behavior of the occasional vertebrate-
eaters.-In this paper, spiders from 28 families are considered to be occasional vertebrate-eaters . The spiders from these families had previously not been observed (or very rarely observed) subduing vertebrates. In the case of some of these spiders, involvement in vertebrate predation appears to be unexpected and such species most likely are not adapted to vertebrate predation. The significantly smaller average body size of these spiders suggests that they have reduced physical strength compared with the larger-sized habitual vertebrateeaters (see above). Also some spider species from this group (e.g., tetragnathids and linyphiids) produce small, feeble webs (i.e., ''low energy absorbing webs'' sensu Craig 1987) not suitable for the interception of vertebrates. The unexpected occurrence of vertebrate predation in spiders of this type might be considered as chance encounters that took place when a tiny vertebrate (eventually weakened by malnutrition or disease) crossed the path of an opportunistic spider. An incident of this type has been documented from South Africa where a crab spider (Thomisus blandus Karsch, 1880) overpowered and devoured a tiny Cape dwarf gecko of onlỹ 4.5 cm total length ( Fig. 6A-C; Wood 2017). 4.3 How does vertebrate-eating affect the food supply of spiders?-So far only a few quantitative assessments of vertebrates as spider prey exist. In several studies from very different regions of the globe, vertebrates were documented to make up , 1% of the spiders' total prey (Appendix 1 - Zimmermann & Spence 1989;Henschel 1990;Hodar & Sanchez-Pinero 2002;Szymkowiak et al. 2005;Bryan et al. 2015), whereas in many other published dietary studies vertebrates were not even present in spider diets (see Nentwig 1987;Nyffeler 1999). A few studies have documented that vertebrates made up a considerable percentage (8-30%) of the total diet (Appendix 1 -McKeown 1952; Lapinski & Tschapka 2013;Baba et al. 2019;Valenzuela-Rojas et al. 2019;Ramírez et al. 2021;pooled data Edwards 1980Wilder & Simpson 2022). These latter studies were based on low prey numbers (n ¼ 21-77) and must therefore be interpreted with caution.
Large theraphosids in the genera Avicularia Lamarck, 1818, Grammostola Simon, 1892, Lasiodora C. L. Koch, 1850, Theraphosa Walckenaer, 1805, etc. were assumed to feed heavily on frogs and other vertebrates (see Emerton 1926;McCormick & Polis 1982;Hillyard 1994;Nyffeler & Altig 2020;Nyffeler & Gibbons 2021;Rick West, pers. comm.). From a point of view of optimal foraging theory it would make perfect sense if such large spiders (weighing 10-100 g) would feed to a large extent on vertebrate prey since those are usually more profitable food items than most arthropod prey due to their larger body mass (see Nyffeler & Knörnschild 2013;Nyffeler & Pusey 2014). In line with this, the Brazilian tarantula Grammostola actaeon (Pocock, 1903) was labeled as a specialized vertebrate-eater in the literature (see Berland 1932;Millot 1949;Gertsch 1979;Hillyard 1994) due to its acceptance of small snakes, frogs, and lizards as prey when offered to it in captivity (see Emerton 1926).
Predation on vertebrates appears to contribute overall rather insignificantly to the food supply of most spiders. But from the perspective of a single individual, opportunistically feeding on a vertebrate prey item may be quite rewarding. The occasional capture of vertebrate prey may be sufficient to fulfill the energetic needs for reproduction, as suggested for large orb-weavers in the ''rare, large prey hypothesis'' (Nyffeler & Knörnschild 2013). The exceptionally long ''handling time'' in terms of processing vertebrate prey suggests that spiders can extract an exceptionally large amount of liquified prey biomass from that kind of over-sized prey items (vertebrate prey is often several times larger than the spider itself). While it typically takes 0.5-2 hours to consume an insect prey (M. Nyffeler, pers. obs.), spiders were observed feeding on vertebrate prey for many hours up to several days (see Nyffeler & Knörnschild 2013;Nyffeler & Pusey 2014;Nyffeler & Altig 2020;Nyffeler & Gibbons 2021). Excess food -which is not instantly used for the metabolism -is stored in the form of glycogen and lipids in resorptive cells of the interstitial tissue as an energy reserve (Foelix 2011). In times of low food availability, the spiders can then draw on these energy reserves, which allows them to bridge longer periods of hunger. However, oftentimes a spider will eat only a small portion of a dead vertebrate. What is left is later consumed by scavengers (ants, wasps, flies, molds) and returned to the natural cycle of nutrients.

CONCLUDING REMARKS
Our survey unveiled a larger number of vertebrate-eating spider species from families previously not anticipated to prey on vertebrates. Interestingly, 16 (¼ 41%) of the total of 39 spider families documented to be engaged in vertebrate predation are known to also occasionally feed on earthworms and/or gastropods, implying that many spider taxa have a more diverse diet than previously thought (see Nyffeler & Symondson 2001;Nyffeler et al. 2017b). The capability of spiders to supplement their usual arthropod diets with different types of alternative food might be of nutritional significance, most notably during periods of food scarcity (also see Nyffeler et al. 2016Nyffeler et al. , 2017b. The findings presented here improve our general understanding of spider feeding ecology and provide a first assessment of the contribution of vertebrate prey as an alternative food source.  [4][5][6]. We thank Richard Vari † (Smithsonian Institution, Washington DC) for the identification of a characiform fish from Ecuador (Fig. 2). We also wish to thank Graham Alexander (University of the Witwatersrand), Aaron Bauer (Villanova University), and Werner Conradi (Port Elizabeth Museum) for the identification of a South African gecko, and Rick Shine (University of Sidney and Macquarie University) for the identification of an Australian skink (see Figs. 4,6). Vincent Albouy (OPIE, Carrières-sous-Poissy), Florentino Braña (Universidad de Oviedo), Pierre-André Crochet (Centre d'Ecologie Fonctionnelle et Evolutive, Montpellier), and Ivan Ineich (Muséum National d'Histoire Naturelle, Paris) identified a lacertid from France (Fig. 4). A group of Australian bat scientists -including Kyle Armstrong, Mike Craig, Damian Milne, Chris Pavey, Terry Reardon, Martin Rhodes, Monika Rhodes, and Christopher Tidemann -identified a bat from Queensland, Australia based on photographs, whereas Andrew Amey (Queensland Museum) identified an Australian frog (see Fig. 2). We also wish to thank Rick West (Sooke, British Columbia) who provided the senior author with valuable information on theraphosid biology for many years. Furthermore, we thank Barbara Thaler-Knoflach (University of Innsbruck) for reporting to us an account of a lizard being eaten by a theridiid spider in Austria and Rene Jacob (near Hannover, Germany) for informing us on his observation of a dead fire salamander found in an agelenid web. Volker Framenau (University of Western Australia) assisted in getting access to literature. Astri Leroy (The Spider Club of South Africa) helped us to get access to thomisid pictures from South Africa. Valuable comments of the subject editor Yael Lubin (Ben-Gurion University) and an anonymous reviewer helped to improve the manuscript. We also express our gratitude to the editor-in-chief Deborah Smith (University of Kansas) for her help with the layout of the photos. Last but not least, the following photographers are greatly acknowledged for granting permission to use their photos: Gordon Ackermann (Switzerland), Vincent Albouy (Annepont and Carrières-sous-Poissy, France), Jesse Delia (AMNH New York), Carmen Fabro (Cape Tribulation, Australia), Donell S. Frank (College Station, TX), Ed Germain (Sydney, Australia), Lawrence L.C. Jones (Tucson, USA), Barbara Maslen (Daintree, Australia), William Shipway (Sydney), and Mathew Wood (formerly South Africa, now in Bahrein). LITERATURE CITED Ackermann G. 2012. Lepidodactylus lugubris (Squamata: Gekkoni- Table 2.-Comparison of habitual vertebrate-eaters vs. occasional vertebrate-eaters regarding mean spider body length, extent of global distribution, and dietary diversity. The statistical significance of the difference of the mean values for each of the three comparisons was tested using a two-tailed Mann-Whitney U test. A 39 th family (family Idiopidae) was not included in this comparison because its assignment to either of these two groups of families is still unclear.  Figure 3.