Arthropods as Vertebrate Predators: A Review of Global Patterns

Arthropod predators preying on vertebrates is generally overlooked in ecological studies, as it is not typically observed in nature and generally considered a rare event. This is likely due to the cryptic nature of these predatory events, the relatively small size of arthropods, and the difficulty in collating published data which is scattered throughout the literature. Although arthropods are known to readily hunt and consume vertebrates, very little is known about these predatory events. In this study, a systematic literature review was conducted to provide a conceptual framework, identify global patterns, and create a searchable database of arthropod preying on vertebrates. This study represents the largest global assessment of arthropod predators and vertebrate prey with over a thousand recorded observations collated from over 80 countries across every continent except Antarctica, where no arthropod predator exists. Arthropod predators were represented by six classes (insects, arachnids, centipedes, and crustaceans: Malacostraca, Ostracoda, Hexanauplia) and over 80 families. Vertebrate prey were represented by five classes (birds, mammals, reptiles, amphibians, fish) and 160 families. The most common prey were frogs consisting of over a third of all observations. The most commonly preyed reptiles were nearly all lizards, half of mammal prey were bats, nearly a third of fish were Cypriniformes, and half of bird prey were passerines. Spiders represented over half of all predatory events found and were the main predator for all vertebrates except birds, which were preyed mostly upon praying mantises. However, prey varied between spider families. For insects, true bugs (Hemiptera) and beetles preyed mostly on amphibians while the aquatic Odonata larvae preyed on amphibians and fish. Decapod predators were observed preying equally between reptiles, birds, and amphibians; with centipedes preying mainly on reptiles and mammals. Predation was mostly recorded from the Americas and Australia, with countries and regions varying between predator and prey groups. Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 10 February 2020 doi:10.20944/preprints202002.0119.v1 © 2020 by the author(s). Distributed under a Creative Commons CC BY license. This study demonstrates that arthropods are indeed an overlooked predator of vertebrates. Recognizing and quantifying these predator-prey interactions is vital for identifying patterns and the potential impact of these relationships on shaping vertebrate populations and communities. Understanding the possible threat of arthropod predators may be especially important to improve the success of conservation efforts by accounting for predators which may currently be overlooked.

Recognizing the predator-prey interactions that exist between arthropods and their vertebrate prey is vital for understanding how arthropods can shape vertebrate populations and communities. This could be especially important for threatened vertebrate populations, with recent evidence demonstrating that arthropod predators have the potential to negatively impact conservation efforts of vertebrates, such as fish (Feher, 2019). While some studies have shown arthropods may have a large impact on vertebrates in experimental settings (Kopp, Wachlevski, & Eterovick, 2006;Nordberg et al., 2018;Pearman, 1995;Wizen & Gasith, 2011), very few studies have examined arthropod predation under natural conditions, resulting in these predation events being considered rare in nature.
Although there are millions of online videos, photos, and newspaper accounts of arthropods preying on vertebrates, these events are only occasionally documented in scientific literature, with most published articles scattered as single individual observations and appearing in the literature as natural history notes. Despite literature searches becoming much easier with the internet, these articles are still difficult to find, resulting in studies incorrectly stating there are only a few cases recorded (Bastos, Oliveira, & Pombal, 1994;Bernarde, Souza, & Kokubum, 1999) or that they are the first published accounts of such predatory events (Hibbitts, 1992). The complications in finding these published accounts will only be further compounded as natural history science continues to decline (Tewksbury et al., 2014). Furthermore, as this information remains scattered throughout the literature it reiterates the presumed rarity of arthropod predation on vertebrates and makes it increasingly difficult to identify patterns and the potential impact of these trophic relationships.
Although these major reviews provide a great overview for a specific subset of arthropod predators, they do not provide a comprehensive overview of these relationships, global patterns, or an easily searchable database. The only comprehensive global review of arthropod predation was conducted four decades ago by McCormick and Polis (1982), where they identified arthropods as an important and overlooked predator of vertebrates. In their major review, they detailed a lack of quantitative data and emphasized the importance of studying arthropod predation in future ecological research. However, their recommendation has since been mostly ignored with very little progress having been made and the importance of these trophic relationships remaining largely unclear.
For this study, a global analysis of arthropods preying on vertebrates was undertaken on the available published scientific literature. This systemic literature review was conducted to provide a conceptual framework, identify global patterns and create a searchable database of arthropod predators and their vertebrate prey.

| M E TH O DS
An extensive literature search of vertebrate predation by arthropods was undertaken between November 2019 and June 2020.
Scientific articles, reviews, bulletins, newsletters, books, theses, dissertations, government reports and conference proceedings were searched using Google Scholar, BioOne and Web of Science. The literature search was conducted using key search terms (e.g., arthropod, spider, insect, predation, prey, vertebrate, fish, bird, amphibian, etc.), synonyms (e.g., lizard, squamata, reptile), and Boolean language (AND, OR). Literature was found indirectly by reviewing relevant articles cited within the original article. Google Scholar was also used to examine literature that cited the original as well as those recommended by the database using the "cited by" and "related articles" links, respectively. Many secondary citations were from Herpetological Review, for which although indexed by Web of Science, abstracts and full text are not archived or searchable. Therefore, their archive from 1967 to 2019 was downloaded and a comprehensive search of their pdf archives was undertaken. Twitter and ResearchGate were also used to enquire about and obtain further citations.
Predatory events were included only if they met the following conditions: (a) an arthropod predator was directly observed attacking and then consuming or attempting to consume a vertebrate; (b) if the prey was not alive it was assumed by the authors to have been caused by the predation event; (c) the event occurred in the field or a natural (not laboratory) experiment (laboratory experiments were thus removed from previous reviews); and (d) the prey was post-hatching (larvae, juveniles, adults). The class, order and family of the prey and predator, as well as the country where the event occurred, were recorded, if available.
A review of all journals referenced was also undertaken on Web of Science to determine what proportion of scientific journals were indexed in a scientific database. An additional literature search was conducted using BioOne to determine the respective bias in the number of published articles between predator and prey groups.
All articles between 1965 and 2020 were searched using key search terms, synonyms, and Boolean language; for example, (lizard OR gecko OR snake OR Serpentes OR Squamata OR squamata OR testudines OR turtle) was used to search for the number of scientific articles on Squamata. All search terms also included "AND (conservation OR biodiversity OR ecology)" to filter for ecological journals.
The data set is openly available from the Dryad Digital Repository (https://doi.org/10.5061/dryad.9p8cz 8wd6). A list of the data sources is also found in Supporting Information Appendix S1.

| Statistical analysis
A chi-square test was used to compare whether the number of predation events within predator and prey groups was independent of their total number of scientific publications. A post-hoc analysis was conducted using a Bonferroni test to determine any significant differences within the groups (Beasley & Schumacker, 1995).
Since sampling efforts are inherently biased with certain taxa better represented in particular countries than others, stabilized inverse probability weights were calculated to compensate for the imbalance between groups and regions. Inverse probability weights reduce selection bias by adding larger weights to underrepresented observations and lower weights to those overrepresented (Austin & Stuart, 2015;Hernán & Robins, 2020). Stabilized inverse probability weights are calculated by taking the conditional probability of being selected given a set of an observed confounding variables and dividing by the marginal probability of the confounding variable. Thus, the stabilized inverse probability weight for each country within a predator or prey class (W i ) can be calculated with the formula: where P(C = c i |X = x i ) is the probability of a predation event in the ith country within the ith predator or prey class, and P(C = c i ) is the observed probability of the ith country within the study irrespective of predator or prey class. Weights equal to or close to 1 indicate that the conditional probability is equal to its marginal probability and is thus unbiased, weights lower than 1 indicate the observation is overrepresented based on its marginal probability, while weights greater than 1 indicate the observation is underrepresented.

| R E S U LTS
A total of 1,309 arthropod predation events were found from 737 references (data available from Dryad at https://doi.org/10.5061/ dryad.9p8cz 8wd6). References included books, theses and conference proceedings, with over 90% comprised of articles from peerreviewed journals, bulletins, and newsletters (Table 1). Published articles came from a total of 235 different journals with 161 of (1) them (68.51%) indexed in the Web of Science. Nearly a third of all reported events were obtained from Herpetological Review, followed by Herpetological Notes with just over 5% (Table 1). However, only ten publication titles recorded over five predation events, with most containing just one or two documented events. Over 7% of articles were obtained from non-English publications (Table 1) South American countries (Figure 4e).

F I G U R E 4
Documented arthropod predation events on vertebrates around the world by the inverse probability weights of prey class by country. Size represents total proportion of the taxa and colour represents weights. Weights equal to 1 indicate the number of observations is unbiased, a weight of 0.1 indicates the number of observations is biased towards the country by a multiple of 10 and a weight of 5 indicates the observations is underrepresented by a multiple of 5. Data visualization created with Tableau Public 2020.1 (Tableau Software, Seattle, WA) This is unsurprising since the total standing biomass of the global spider community is nearly 25 million metric tons, preying on a similar order of magnitude of prey as all the whales in the world's oceans (Nyffeler & Birkhofer, 2017). However, these results may also be simply due to spiders being the subject of many review articles (Babangenge et al., 2019;Blondheim & Werner, 1989;Butler & Main, 1959;Maffei, Ubaid, & Jim, 2010;Menin et al., 2005;Morris, 1963;Neill, 1948;Nyffeler, Edwards, et al., 2017;Nyffeler & Knörnschild, 2013;Nyffeler & Pusey, 2014;Nyffeler & Vetter, 2018;Steehouder, 1992). Nevertheless, spiders possess strong fangs capable of piercing vertebrate skin and injecting them with neurotoxins, many specific to the nervous system of vertebrates (Garb & Hayashi, 2013;Gregio, Heleno, Von Eicksted, & Fontana, 1999).

| D I S CUSS I O N
They also have a diversity of tactics such as active hunting, sitand-wait ambush, and for many, the production of webs that can entangle many small animals. These webs are so strong that birds are often found entangled in them, with many reviews published regarding this phenomenon (Abbott, 1931;Brooks, 2012;Kasambe, Thosar, Rathore, Shivkar, & Sasi, 2010). Another major difference between spiders and other arthropods is that spiders can grow much larger than most arthropods, including very large species such as the goliath bird eater (Theraphosa blondi) in the Theraphosidae family. Their size and the ease of discovering prey caught on their spider webs also make it much easier to observe predation events compared to other arthropods.
Spiders were found to differentiate among their target prey based on the different predatory strategies of their families.
Wandering spiders (Ctenidae) are nocturnal, venomous, ambush hunters that preyed mostly on frogs, which are easy targets for these spiders due to their similar nocturnal behaviours and sedentary lifestyle (Valenzuela-Rojas et al., 2019). Tangle-web spiders (Theridiidae) are known for their web building and venomous bite, including those from black widows (Latrodectus) (Nyffeler & Vetter, 2018). This group of spiders are the most common arthropod found in human dwellings, which may explain why the most common prey were lizards and rodents, other common house inhabitants (Leong et al., 2017). Nursery web spiders (Pisauridae), which include fishing spiders, unsurprisingly preyed most on the aquatic animal groups of amphibians and fish, in accordance with previous studies (Baba et al., 2019;Bleckmann & Lotz, 1987;Nyffeler & Pusey, 2014). Lastly, orb-weaver spiders (Araneidae) often feed on birds and bats, which they would be expected to commonly encounter due to their occurrence in gardens, fields and forests. Orb-weavers are known for their strong and very large webs (including the largest web and longest bridge line ever recorded), which makes them well suited to capture these large flying prey (Kuntner & Agnarsson, 2010). Nonetheless, these results are also likely due to the differences between the groups and their chance of being observed. Spiders that build webs are generally more conspicuous and visible to humans, who will notice and recognize the captured prey, especially since prey items typically remain stuck in the web for a period of time. Observers will remember where they saw webs and may therefore consciously or subconsciously look for other prey items within these webs. This contrasts with non-web weaving spiders, in which an observer must be in the right place at the right time to be lucky enough to see a predation event.
Other common predators were insects, with different orders also preying on different vertebrate groups. Insects were mainly from North and South America, south-eastern Asia, central Europe and southern Africa. They were biased towards North and South America and underrepresented in Central America and Australia.
The largest number of events was from the Hemiptera order, mostly giant water bugs in the Belastomidae family and recorded more than would be expected based on their number of scientific articles. These large, aquatic predators are known to be able to consume everything from turtles to fish due to their large mandibles (Ohba, 2019). In this study, they mostly preyed on aquatic vertebrates such as amphibians and fish. Odonata larvae are another group of large aquatic predators that were found to mostly consume amphibians and fish. Praying mantises as vertebrate predators were reported more than would be expected and were mostly observed preying on birds. Mantises not only eat fledglings but are also well known to capture small birds, especially hummingbirds as they hover in mid-air (Fisher, 1994;Hildebrand, 1949;Lorenz, 2007;Murray, 1958), and are the only known insect to impale their prey with its legs (Rivera & Callohuari, 2019). Malacostraca were also documented more than expected and were documented mainly from Brazil and the United States, where they were underrepresented, | 9 VALDEZ with a bias towards Central America and the Caribbean. Decapod crustaceans, which mostly consisted of crabs, did not exhibit any preference and were found to prey equally on reptiles, amphibians and birds, which may have to do with their wide range of terrestrial, arboreal and aquatic lifestyles (Andrade, Júnior, Júnior, & Leite, 2012;Wehrtmann, Hernández-Díaz, & Cumberlidge, 2019).
Scorpions were another common predator, being arachnids with a large venomous stinger on their tails and an aggressive hunting strategy, and mainly consumed reptiles and mammals, which are also found in forest environments (Jestrzemski & Schütz, 2016).
The largest group of vertebrate prey were amphibians, representing over 40% of all prey and consisting nearly exclusively of frogs. They were mainly documented throughout the Americas, are well known to be preyed upon by a wide range of predators, including carnivorous plants, and it has been stated that "practically anything will eat an amphibian" (Duellman & Trueb, 1994). This may be because while many frogs have excellent escape mechanisms (e.g., quick and powerful jumping abilities, production of skin toxins, gliding with large webbed toes, etc.), they are especially vulnerable due to their soft, easily penetrable skin. Compounding this vulnerability, many frogs typically require both aquatic larval and terrestrial adult life stages, with many species also arboreal. This requirement of multi-life stages increases their exposure to the large variety of predators that are present across these habitats. These various predators include diving beetles that are known to feast on tadpoles (Gould, Valdez, Clulow, & Clulow, 2019), water bugs that feast on metamorphs and juveniles (Fadel et al., 2019), and spiders, which may be the most important arthropod predator of adult frogs (Toledo, 2005). The Hylidae group of tree frogs have all these qualities, which may explain why they were the largest order preyed upon. This tendency for frogs to get eaten explains their typical r-selected strategy, with very large clutch sizes compared to other vertebrate groups, which can sometimes number in the tens of thousands (Lever, 2001).
Another large prey group was reptiles, which, compared to their number of scientific articles, were recorded more than would be expected. Reptiles were found throughout the United States, Australia, Brazil, as well as the regions of southern Asia, South America and southern Africa. Observations were biased towards most regions, except for Central and South America. The largest families were geckos, skinks and snakes, which were preyed upon by a diverse group of predators. Although they have various defence mechanisms, groups such as geckos are easy targets due to their relatively small size, compared to larger lizards with stronger teeth and claws. Although snakes are better equipped at protecting themselves given their speed, venoms and fangs, they may still be subdued by the neurotoxins of many of these arthropods, especially scorpions where up to 10% of their diet is snakes (Greene, 1997 (Safarek et al., 2010) and spiders (Poo, Erickson, Mason, & Nissen, 2017) consume amphibian eggs, with other studies demonstrating the severe threat of ant predation on turtle nests (Buhlmann & Coffman, 2001;Erickson & Baccaro, 2016;Holbrook, Mahas, Ondich, & Andrews, 2019;Parris, Lamont, & Carthy, 2002) and bird clutches (Menezes & Marini, 2017). This study also did not include situations where animals were not consumed but were caught in spider webs and subsequently died simply due to being entrapped (Brooks, 2012;Duca & Modesto, 2007;Graham, 1997;Kasambe et al., 2010;Martin & Platt, 2011;Walther, 2016). Both situations would likely remove a considerable number of individuals from a population.
This study demonstrates and further repeats the original assumption of McCormick and Polis (1982) nearly half a century ago, that arthropod predation remains underestimated and quantitative data assessing its impact on vertebrate communities are a necessary link between field and theoretical ecology. Recognizing and understanding arthropod predation and its impact on vertebrate communities have become even more vital in recent years, especially for vulnerable groups with small populations. Evidence from conservation management projects has already demonstrated that arthropod predators can have an impact on the success of conservation efforts (Feher, 2019;Valdez, 2019). However, studies on food webs have been in significant decline during the last four decades (McCallen et al., 2019), with very few ecological studies on predation or food webs (Carmel et al., 2013). It has, therefore, become even more imperative to investigate and quantify the effects of arthropod predators on vertebrate communities within habitats and ecosystems. Although a small number of studies have quantified possible effects, they have occurred under small laboratory settings (Cabrera-Guzmán, Crossland, & Shine, 2012;Pearman, 1995) or involved clay models (Nordberg et al., 2018). However, new methodologies and technological advances can help quantify arthropod predation, such as camera traps (Hobbs & Brehme, 2017), sentinel prey (Lövei & Ferrante, 2017), prey baits and gut analyses . Recognizing and quantifying these interactions will fill the gap of knowledge remaining in our ecological understanding.
This will help to not only understand the role arthropod predators play in shaping vertebrate communities, but also improve the success of conservation efforts by accounting for predators that may currently be overlooked in threat abatement plans.

AC K N OW LE D G M E NT S
I wish to thank John Gould for reviewing and proofreading the manuscript. I also wish to acknowledge the contribution from the