Know your foe: synanthropic spiders are deterred by semiochemicals of European fire ants

Many ants prey on spiders, suggesting that web-building spiders may avoid micro-locations near ant colonies or frequented by foraging ants. Here we tested the hypothesis that ant-derived semiochemicals deter synanthropic spiders. To generate stimuli, we exposed filter paper for 12 h to workers of European fire ants, Myrmica rubra, black garden ants, Lasius niger, or western carpenter ants, Camponotus modoc, and then offered select urban spiders in three-chamber olfactometer bioassays a choice between ant-exposed filter paper and unexposed control filter paper. Semiochemical deposits of M. rubra, but not of L. niger or C. modoc, had a significant deterrent effect on subadults of the false black widow, Steatoda grossa, the black widow, Latrodectus hesperus, and the hobo spider, Eratigena agrestis, as well as a moderate (but statistically not significant) deterrent effect on the cross spider, Araneus diadematus. The deterrent effect caused by semiochemical deposits of M. rubra may be attributable to the aggressive nature and efficient foraging of M. rubra in its invaded North American range, exerting selection pressure on community members to recognize M. rubra semiochemicals and to avoid micro-locations occupied by M. rubra.

Many ants prey on spiders, suggesting that web-building spiders may avoid micro-locations near ant colonies or frequented by foraging ants. Here we tested the hypothesis that ant-derived semiochemicals deter synanthropic spiders. To generate stimuli, we exposed filter paper for 12 h to workers of European fire ants, Myrmica rubra, black garden ants, Lasius niger, or western carpenter ants, Camponotus modoc, and then offered select urban spiders in three-chamber olfactometer bioassays a choice between ant-exposed filter paper and unexposed control filter paper. Semiochemical deposits of M. rubra, but not of L. niger or C. modoc, had a significant deterrent effect on subadults of the false black widow, Steatoda grossa, the black widow, Latrodectus hesperus, and the hobo spider, Eratigena agrestis, as well as a moderate (but statistically not significant) deterrent effect on the cross spider, Araneus diadematus. The deterrent effect caused by semiochemical deposits of M. rubra may be attributable to the aggressive nature and efficient foraging of M. rubra in its invaded North American range, exerting selection pressure on community members to recognize M. rubra semiochemicals and to avoid micro-locations occupied by M. rubra.
spiders [3]. Chemical tactics such as insecticide applications [5] are largely ineffective because spiders can avoid insecticides by abandoning their web and rebuilding one elsewhere [3,6]. Natural repellents of spiders, such as chestnuts and lemon oil, are widely advertised in anecdotal accounts but only a few have been experimentally tested [7,8], and none effectively repelled all species of spiders tested [9]. Moreover, there is no immediate ecological reason why these materials are repellent to spiders.
By contrast, there is every reason for spiders to avoid natural predators such as ants that prey on both web-building and cursorial spiders [10][11][12]. At the population level, there is a negative correlation between the density of ant populations and the total biomass of spiders [13,14]. Cobweb spiders, Phylloneta impressa, tend to disperse in response to chemical cues derived from black garden ants, Lasius niger, and the formicine ant Formica clara [15]. Sensing chemical cues of potentially predatory ants is particularly adaptive for subadult web-building spiders which seek suitable micro-locations for settling and building their webs [16]. As web building is a significant time and energy investment [17][18][19], subadult spiders are thought to explore, and ultimately select, primarily those microhabitats that have no or few threats to survival, such as the presence of predatory ants. Flat rock spiders, Morebilus plagusius, e.g. avoid ant-scented rocks when selecting retreat sites [20].
Here we tested the hypothesis that ant-derived semiochemicals deter spiders. As model organisms for our study, we selected three synanthropic ant species [European fire ants, Myrmica rubra; black garden ants, Lasius niger; western carpenter ants, Camponotus modoc (all Formicidae)] and four synanthropic web-building spider species [false black widow, Steatoda grossa; western black widow, Latrodectus hesperus (both Theridiidae); cross spider, Araneus diadematus (Araneidae); hobo spider, Eratigena agrestis (Agelenidae)], all of which are commonly found in and around human dwellings in North America [21].

Ants
Myrmica rubra workers (figure 1) were collected from nests at Inter River Park (49°19'10.9 00 N 123°01'43.7 00 W) in North Vancouver, British Columbia (BC), Canada, whereas workers of L. niger and C. modoc (figure 1) were collected from nests located on the Burnaby campus of Simon Fraser University (SFU, 49°16'33 00 N 122°54'55 00 W), BC. All ants were kept in jars (1-4 l) filled with soil from collection sites and were provisioned with tubes of sugar water retained with a cotton ball. To standardize the presentation of test stimuli according to weight equivalent of ants, 75 workers of each species were weighed in groups of five using a microbalance (TR-204, Denver Instrument Comp., Arvada, CO 80004, USA). Body weights (mean ± s.e.) of individual workers of M. rubra, L. niger and C. modoc amounted to 3.51 ± 5.56, 3.02 ± 4.44 and 43.7 ± 52.7 mg, respectively.

Spiders tested
All specimens of S. grossa (figure 1) were F 1 subadult offspring of mated females captured on SFU's Burnaby campus [22], whereas specimens of L. hesperus and E. agrestis (figure 1) were F 1 subadult offspring of mated females collected on Centennial Beach Boundary Bay Regional Park, Delta, BC (49°01'10.9 00 N 123°02'32.1 00 W). Spiderlings were housed singly in a Petri dish (100 × 20 mm) containing a moist cotton wick and-based on body size-were provisioned with Drosophila vinegar flies or Phormia regina blow flies once a week.
All A. diadematus were subadults, collected on the day of bioassays on SFU's Burnaby campus. Following bioassays, they were released into a designated non-collection zone on campus.

General experiments design
The effects of ant-derived deposits on aversion responses by spiders were tested in still-air, dual-choice olfactometers [8,23] kept at room temperature and a 12 L : 12 D photoperiod. Olfactometers (see fig. 1 in [8] for a photographic illustration) consisted of three circular Pyrex glass chambers (3.5 × 10 cm inner diameter (ID)) with removable glass lids linearly interconnected by glass tubes (each 2.5 × 1 cm ID). The bottoms of lateral chambers were lined with circular filter paper (Whatman, Maidstone, England). Treatment and control stimuli were assigned to lateral chambers such that the treatment stimulus was equally often presented in the left and right lateral chamber of an olfactometer to minimize any royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 210279 potential effect of side bias. To prepare a treatment stimulus, ants were placed in one lateral chamber and prevented from leaving by a wet cotton ball inserted in the glass tube interconnecting the lateral and central chamber. The wet cotton ball not only blocked the chamber exit, but also provided a source of moisture for the ants. To ensure symmetry of the experimental design, a wet cotton ball was also inserted in the glass tube interconnecting the central chamber and the second lateral chamber. As the quantity of semiochemicals deposited by ants was probably correlated with their body size or weight, equal weight equivalents of ants were used to standardize the preparation of treatment stimuli; hence, 37 M. rubra, 43 L. niger and 3 C. modoc were confined in the treatment chamber. After 12 h of (overnight) confinement, the ants and the cotton balls were removed. Then, a bioassay spider was introduced into the central chamber and kept in darkness for 24 h, following which its final position was scored under red light. Spiders positioned in lateral chambers were classed as responders to treatment or control stimuli, whereas those in the central chamber were recorded as non-responders. Spiders located in an interconnecting glass tube were scored as non-responders if they were closer to the central chamber than to the respective lateral chamber. All spiders were tested only once, and olfactometers were washed in detergent water (Sparkleen, Fischerbrand, Toronto, Canada) and oven-dried between replicates.  As only semiochemical deposits of M. rubra, but not of L. niger or C. modoc, deterred S. grossa (see Results), follow-up experiments 5-8 (table 1) focused on M. rubra semiochemicals, and tested whether they deter only S. grossa (Exp. 5), or also deter L. hesperus (Exp. 6), E. agrestis (Exp. 7) and A. diadematus (Exp. 8).
With evidence that M. rubra semiochemicals deter at least three spider heterogeners (see Results), experiments 9 and 10 (table 1) then tested dose-dependent effects of deterrent semiochemicals by offering S. grossa a choice between filter paper left untreated (control) or soiled with semiochemicals from either 37 M. rubra workers (Exp. 9; the same dose as in Exps. 3, 5-8) or 111 M. rubra workers (Exp. 10; a threefold higher dose).

Statistical analysis
R [24] was used to perform one-sided binominal tests to analyse data for the hypothesized repellent effect of ants on spiders in two choice experiments 1-10 [25]. Subsequently, the p-values of experiments were adjusted using the Benjamini-Hochberg method to account for multiple comparisons [26].

Results
When subadult S. grossa were offered a choice between two lateral olfactometer chambers, each containing a control stimulus (untreated filter paper), they chose the right and left chamber 11 and 10  (7) 10 111 M. rubra no deposits S. grossa 30 (10) a Equal weight equivalents of ants (37 M. rubra, 43 L. niger and 3 C. modoc) were used to standardize the preparation of test stimuli (chemicals deposited by ants on filter paper during 12 h).   Figure 2. Responses of subadult Steatoda grossa that were given a choice in three-chamber olfactometers [23] between two test stimuli, both being untreated filter paper (Exp. 1), or one, being untreated filter paper, and the other being filter paper previously exposed to worker ants of Myrmica rubra (  , and subadult Araneus diadematus (Exp. 8) that were given a choice in three-chamber olfactometers [23] between two test stimuli, one being untreated filter paper and the other being filter paper previously exposed to 37 worker ants of Myrmica rubra. Shown within bars and square inserts are the number of spiders responding to treatment or control stimuli, and not responding to stimuli, respectively. For each experiment, an asterisk ( Ã ) denotes a statistically significant treatment effect (onesided binomial tests; p < 0.05).

Discussion
Our data support the conclusion that semiochemical deposits of M. rubra worker ants have a significant deterrent effect on three spider species (S. grossa, L. hesperus, E. agrestis) and that they express a moderate deterrent effect on a fourth spider species tested in our study, the cross spider A. diadematus. Conversely, semiochemical deposits of L. niger and C. modoc worker ants failed to induce a discernible behaviourmodifying effect on the spiders tested. Our findings that semiochemical deposits of M. rubra worker ants, but not of L. niger or C. modoc worker ants, prompted aversion responses by S. grossa have multiple potential explanations, such as the specifics of the experimental design, contrasting life-history traits of ants, and niche overlap, or not, between ants and spiders.
As part of the experimental design to prepare ant semiochemical deposits as test stimuli for spiders, we selected diverse taxonomic species of ants that greatly varied in body size and weight. Assuming that larger ants deposit greater amounts of semiochemicals, we standardized the amount of deposits between experiments by testing equal weight equivalents of ants, using 37, 43 and 3 worker ants of M. rubra, L. niger and C. modoc, respectively, to generate a test stimulus. However, contrary to our assumption, the body weight of ants and the amount of semiochemical deposits may not be positively correlated, and equal numbers, rather than equal weights, of M. rubra, L. niger and C. modoc worker ants may have been required to generate standardized test stimuli. Alternatively, the semiochemicals deposited by M. rubra may have significantly greater potency as spider deterrents than those of L. niger and C. modoc. Worker ants of M. rubra are omnivorous scavengers and prey on many invertebrates [27]. In their invaded North American range, populations of M. rubra occur in extremely high densities and appear more aggressive than their counterparts in Europe. These characteristics, coupled with efficient foraging and aggressive nest defence, have enabled M. rubra to outcompete native ants and lower the arthropod biodiversity in invaded communities [28]. It is conceivable then, that over evolutionary time arthropod community members, including spiders which may fall prey to M. rubra, have learned to respond to semiochemical cues of M. rubra and to settle in (micro) habitats void of M. rubra. If so, this would provide ecological rationale for our data showing that semiochemical deposits of M. rubra have deterrent effects on S. grossa, L. hesperus and E. agrestis.
Insufficient niche overlap between M. rubra and A. diadematus, and thus a lack of opportunity to learn each other's semiochemical signals or cues, may explain why semiochemical deposits of M. rubra had only a weak (and statistically not significant) deterrent effect on A. diadematus. As orb-weavers, A. diadematus females build their webs above ground [29], physically well separated from the subterranean colonies of M. rubra. Females of S. grossa, L. hesperus and E. agrestis, in contrast, build their three-dimensional cobwebs near ground level [29] with greater likelihood of frequent encounters with foraging M. rubra workers.
The identity of the deterrent semiochemical(s) deposited by M. rubra workers remains unknown. Communication signals such as trail or alarm pheromones [30][31][32][33][34] Figure 4. Responses of subadult Steatoda grossa that were given a choice in three-chamber olfactometers [23] between two test stimuli, one being untreated filter paper and the other being filter paper previously exposed to 37 or 111 worker ants of Myrmica rubra (Exps. 9 and 10, respectively). Shown within bars and square inserts are the number of spiders responding to treatment or control stimuli, and not responding to stimuli, respectively. For each experiment, an asterisk ( Ã ) denotes a statistically significant treatment effect (one-sided binomial tests; p < 0.05).
royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 8: 210279 olfactometers had no immediately obvious incentive to release pheromone and coordinate activities. Yet, signalling in ants is complex and we are just beginning to grasp that complexity. While the functional role of most exocrine glands in M. rubra [35] is still unknown, any gland may have released the semiochemical(s) that prompted the deterrent effect on spiders. Alternatively, the semiochemicals are not released from glands but originate from the ants' body surface.
Irrespectively, the rather remarkable deterrence of M. rubra semiochemical deposits against S. grossa, L. hesperus and E. agrestis warrant the identification of these deterrents through proven-effective techniques in arthropod chemical ecology [36]. Once identified, the origin of these deterrents could be traced to a specific exocrine gland and/or the body surface of ants. Moreover, synthetic replica of these deterrents could be developed, together with concurrently known spider deterrents [9], for earth-friendly manipulation of synanthropic spiders.
Data accessibility. All data are presented in the manuscript and in the electronic supplementary material [37]. Authors' contributions.