Plant–animal interactions between carnivorous plants, sheet‐web spiders, and ground‐running spiders as guild predators in a wet meadow community

Abstract Plant–animal interactions are diverse and widespread shaping ecology, evolution, and biodiversity of most ecological communities. Carnivorous plants are unusual in that they can be simultaneously engaged with animals in multiple mutualistic and antagonistic interactions including reversed plant–animal interactions where they are the predator. Competition with animals is a potential antagonistic plant–animal interaction unique to carnivorous plants when they and animal predators consume the same prey. The goal of this field study was to test the hypothesis that under natural conditions, sundews and spiders are predators consuming the same prey thus creating an environment where interkingdom competition can occur. Over 12 months, we collected data on 15 dates in the only protected Highland Rim Wet Meadow Ecosystem in Kentucky where sundews, sheet‐web spiders, and ground‐running spiders co‐exist. One each sampling day, we attempted to locate fifteen sites with: (a) both sheet‐web spiders and sundews; (b) sundews only; and (c) where neither occurred. Sticky traps were set at each of these sites to determine prey (springtails) activity–density. Ground‐running spiders were collected on sampling days. DNA extraction was performed on all spiders to determine which individuals had eaten springtails and comparing this to the density of sundews where the spiders were captured. Sundews and spiders consumed springtails. Springtail activity–densities were lower, the higher the density of sundews. Both sheet‐web and ground‐running spiders were found less often where sundew densities were high. Sheet‐web size was smaller where sundew densities were high. The results of this study suggest that asymmetrical exploitative competition occurs between sundews and spiders. Sundews appear to have a greater negative impact on spiders, where spiders probably have little impact on sundews. In this example of interkingdom competition where the asymmetry should be most extreme, amensalism where one competitor experiences no cost of interaction may be occurring.

Antagonistic relationships are typically a cost to the plant and include herbivory and seed predation.
Relatively little work has focused on antagonistic plant-animal interactions where carnivorous plants and animals compete as predators, despite competition between kingdoms possibly being the most common form of competition (Barnes, 2003;Hochberg & Lawton, 1990;Trienens, Keller, & Rohlfs, 2010;Trienens & Rohlfs, 2011). Jennings, Krupa, Raffel, and Rohr (2010) conducted a laboratory experiment and field study suggesting wolf spiders and sundews compete, while Jennings, Krupa, and Rohr (2016) suggested sundews, wolf spiders, and toads compete. Clearly more extensive field studies are needed to understand the dynamics of plant-animal interactions between carnivorous plants and spiders where they co-exist as predators.
We tested the following hypothesis: Under natural conditions, sundews and spiders consume the same prey creating the potential for interkingdom competition.

| S TUDY SYS TEM
The dwarf sundew (D. brevifolia) has one of the widest distributions of any carnivorous plant in the western hemisphere ranging from North America to South America (United States, Cuba, Mexico, Belize, Brazil, and Uruguay; Correa & dos Santos Silva, 2005;Schnell, 2002).
In North America, the distribution is a coastal band that extends from east Texas to Virginia with disjunct populations in Oklahoma, Arkansas, Alabama, Kansas, Kentucky, and Tennessee. The Kentucky population is the northern most of these and is state-endangered.
This population grows in a 0.81-hectare area in Hazeldell Meadow, Pulaski County, Kentucky. This site is the only remaining, protected Highland Rim Wet Meadow ecosystem left in the state. The associated Robertsville series soil is deep and poorly drained as a result of an underlying fragipan, which creates a shallow water table just beneath the surface. Most sundews grow in a 600 m 2 portion of the meadow where the soil is compressed and depressed. The population fluctuates greatly from year to year and from season to season.
Over a 10-year period, the population has varied from 220,000 to 25,000 plants. This population is comprised of biennial and perennial sundews the proportions of which vary from year to year depending on temperature and precipitation (Krupa, 2019). extreme, amensalism where one competitor experiences no cost of interaction may be occurring.

| Field sampling
This study focused on the largest of the subpopulations of D. brevifolia growing in the meadow. Ten 400 cm 2 plots were established in July 2011. Each plot was counted periodically over the duration of this study. From August 2012 to August 2013, the study site was visited on 15 days over the seasons weather permitting; during significant snow cover, heavy rain, and standing water, data collection was not possible. We systematically walked along a transect on the eastern edge of the study site (where the densest patches of sundews occur), from south to north, identifying all sheet-webs. After this transect was surveyed, we moved one meter to the west and again walked the length of the sundew population. The goal was to locate 15 sheet-webs with spiders on each sample date. On some collection days, due to weather, we were unable to find 15 of these webs. Spiders occupying each web were collected with an aspirator, and the location marked and identified with a numbered flag.
Each spider was preserved in a separate 1.5 ml microcentrifuge tube filled with 100% EtOH and maintained on ice. Spiders were transferred to a −20°C freezer upon return to the laboratory. The area of a web was calculated by measuring the longest horizontal facial dimension and the dimension perpendicular to it, then calculating an ellipse with these two measures as the radii (Hesselberg, 2010;Welch, 2013). The shape of a web was calculated by dividing length by width.
A 40 cm 2 metal frame subdivided into a string grid of 100 units was placed on the ground at each collection site with the flag at center. The percent of grid units with at least one sundew was used as a measure of percent cover. The distance and diameter of the three nearest sundews from the site of a sheet-web were recorded.
In addition to sites where sheet-web spiders were collected (henceforth referred to as spider sites), 15 sites with sundews that lacked sheet-webs (sundew sites) were randomly selected and flagged. Percent sundew cover was also recorded for each sundew site. Additionally, 15 sites that lacked sheet-webs and sundews (control sites) were randomly selected and flagged. After which, circular sticky traps were placed at each site using 60 mm dia. pieces of clear transparency sprayed with Tangle-Trap Insect Trap Coating Spray (The Tanglefoot Company). Each of these was placed on top of 60-mm-dia. Petri dishes painted with dark brown water paint and depressed into the ground to be flushed with the soil surface (modified from sampling protocol described by Harwood, Sunderland, & Symondson, 2001, 2003. These traps were collected after 24 hr, immediately put on ice for transport to the laboratory and placed F I G U R E 1 (a) Fluctuation in number of sundews (Drosera brevifolia) for each of 10, 400 cm 2 plots at Hazeldell Meadow from August July 2013 to August 2014; (b) Mean diameter (±1 SE) of sundews measured over the study period in a laboratory freezer until all captured arthropods were identified and counted.
On each sampling day, 8-20 ground-running spiders were collected. Percent sundew cover and both distance and diameter of the three nearest sundews from the point where a ground-running spider was captured were recorded. Sticky traps were not set out at capture sites of ground-running spiders, because they were highly mobile ranging over a large area. The 150 leaves were kept frozen until captured arthropods were identified.

| Direct spider-sundew interactions
A 70-mm-dia plastic ring was pressed into the soil surrounding 17 sundews that covered 346 mm 2 which was 9% of the area within the ring. Individual Neoantistea agilis were dropped into the arena one at a time and observed for at least 60 s or until they stopped moving for 60 s. The following behaviors were recorded: (a) runover-a spider ran over a sundew without having sundew mucilage attach to the spider; (b) avoid-a spider approached a sundew with front legs barely touching, stopped, and then moved away; and (c) pull away-a spider ran over a sundew, had sundew mucilage attach to its body then pulled free.

| Molecular analysis of predation
Spiders were identified to species, when possible, and whole-body DNA extractions were performed using QIAGEN DNeasy Tissue Kits (QIAGEN Inc.) following the manufacturer's animal tissue protocol.
The DNA extracted from spiders was then screened for the presence of prey DNA using a general Collembola (hereafter referred to as springtails) primer (Chapman, Schmidt, Welch, & Harwood, 2013). PCR procedures, as described by Chapman et al. (2013), were followed which optimized the primers and screened for cross-reactivity against 155 nonspringtail species. Positive tests for springtails in the diet of spiders were determined by electrophoresis of 10 μl of PCR product in 2% SeaKem agarose (Lonza) stained with 0.1 mg/μl GelRed ™ (Biotium, Inc.). Even though flies and springtails are most commonly captured by sundews (Ellison & Gotelli, 2009), grounddwelling spiders primarily consume springtails (Chapman et al., 2013;Harwood, Sunderland, & Symondson, 2001, 2003. Because of this, springtails are the most likely common prey for sundews and spiders in this study; thus, molecular analysis for flies in spider gut content was not performed.

| Statistical analysis
Data were analyzed using R statistical software version 3.6.1 (R Core Team, 2015). We tested for specific predictions of competition, including negative relationships between (i) sundew abundance and shared prey abundance, (ii) spider presence and shared prey abundance, (iii) spider web size and shared prey abundance, (iv) spider presence and sundew abundance, and (v) sundew abundance and spider web size. We used linear mixed-effects models (lmer function in R) to model predictor variables as fixed effects and sampling date as a random effect to control for the effects of time. ANOVA tables were calculated to compare means between groups. General linear mixed models (glmer function is R) were used when non-Gaussian data called for a Poisson link function. To ensure estimated coefficients would be on the same scale and facilitate comparisons of effect sizes, explanatory variables were standardized by centering means and scaling standard deviations prior to regression analysis.
Chi-square tests compared the likelihood of spider gut contents containing springtail DNA between the two spider types as well as percent sundew coverage between spiders that had and had not recently consumed springtail prey.

| Spiders captured
A total of 172 sheet-web spiders were collected during the study belonging to three genera (Grammonota, Neoantistea, and Tennessellum) in two families (Hahniidae and Linyphiidae). Neoantistea was the most common genus (71.5%) of sheet-web spiders collected. A total of 188 ground-running spiders, all belonging to the family Lycosidae, were collected and represented five genera (Allocosa, Pirata, Pardosa, Rabidosa, and Schizocosa). Pardosa was the most common (71%).

| Prey captured by spiders
Molecular analysis of the gut contents of the 360 spiders we collected revealed that 54.9% of sheet-web spiders and 52.1% of ground-running spiders tested positive for the presence of springtail DNA, indicating frequent consumption during the period of this study ( Figure 2). Sheet-web spiders testing positive for springtail DNA in their guts were found in areas with significantly lower sundew cover than spiders lacking springtails (F 1,151 = 2.848, p = .047, one-tailed; Figure 3). In contrast, the presence of springtail DNA in the guts of ground-running spiders did not differ with respect to sundew cover (F 1,169 = 0.083, p = .387; Figure 3).

| Prey availability to sundews and spiders
Springtails were by far the most common arthropod captured by sticky traps (91% of prey caught). Springtails were also the most common prey trapped on the sundew leaves (40.6%). Springtails were present in the majority of both sheet-web spider guts (54.9%) and ground-running spider guts (52.1%).
The number of springtails caught on sticky traps placed at sundew, spider, and control sites differed significantly (F 2,734 = 296.94, p < .001; Figure 4). Pairwise comparisons showed that sticky traps placed in sundew sites captured significantly fewer springtails than traps set in spider (p < .001) and control sites (p < .001), but there was no difference between spider and control sites (p > .05). This overall effect emerged despite a marked reversal for two winter samples when sundew num- (iv) spider web presence and ground-running spider presence were highly significant predictors of sundew abundance. Sundews were significantly more abundant where spiders were absent (F 2,211 = 493.47, p < .001; Figure 8). Sundew sites averaged 40.5% sundew cover, while F I G U R E 2 Percentage of captured sheet-web spiders and ground-running spiders that contained springtail DNA in their guts for each sampling date F I G U R E 3 Both sheet-web spiders and ground-running spiders with springtail DNA in their gut contents and those without springtail DNA in their guts compared to the percentage of sundew coverage that the site of capture spider sites averaged 6.9% sundew cover, and ground-running spider sites averaged 8.7% sundew cover. Pairwise comparisons found no statistically significant difference in sundew cover between sites with the two spider types and highly significant differences between sundew sites and sheet-web sites (p < .001) and ground-running spider sites (p < .001); (v) there was a highly significant negative correlation between web area and percent sundew (Figure 9), helping to distinguish the effects of sundew abundance from the separate effect of grass.
Sheet-webs located in the open with sundews were significantly larger than webs located in the adjacent (within 5 m) grass areas with no sundews (F 1,45 = 14.092, p < .001; Figure 10). Thus, sheet-webs were smaller in the open area than in the grass, but within the open area, wherever sundew abundance was greater, they were smaller still.
The remaining 16 arthropods included five spiders, two mites, four aphids, two leafhoppers, two crickets (Order: Orthoptera), and one beetle (Order: Coleoptera). Sixty-two (41.3%) of the individual sundew leaves examined had springtails trapped on their trichomes.

| Direct spider-sundew interactions
Seventeen N. agilis were observed in the arena for a total of 62 min during which they made contact with sundews 221 times. On 16 occasions, spiders stopped when their front legs made contact with a sundew then turned away avoiding further contact. Spiders ran over sundews 205 times with mucilage attaching to their legs only 24 times. Of these, it took spiders 9.3 s on average to pull away from sundews. The 17 spiders averaged 2.9 mm in length and were smaller than the average sundew (4.7 mm dia) in the arena. where sundews were sparse or lacking. Both sheet-web and groundrunning spiders were found in areas where sundews were sparse and springtail activity-density high. Molecular analysis of the gut content of 360 spiders (both sheet-web and ground-running) revealed that 50.3% tested positive for springtail DNA, signifying recent consumption of this prey.

| D ISCUSS I ON
There was no evidence of intraguild predation of sundews on spiders during this study. Only 5 very small spiders were found attached to sundew leaves. These accounted for 2.2% of all prey captured by sundews. Furthermore, sheet-web spiders (N. agilis) that were only 38% the size of the average sundew ran over sundews 93% of 221 encounters during the arena experiment indicating sheet-web spiders were not being directly impacted by these plants. The much larger wolf spiders often were observed to run over sundews unimpeded during our field study.
The results of this study suggest that the plant-animal interaction between sundews and spiders is most likely exploitative competition. Sheet-web and ground-running spiders were common and ubiquitous in the meadow, yet least common where sundews were most dense. Furthermore, where sundews were dense, sticky traps showed lower activity-density of springtails. When sundews were dormant, springtail activity-density was high around these dense stands of sundews. Our results indicate sundews were drawing down springtail numbers. Thus, springtails may have been a limiting resource in the presence of sundews. Spiders responded by avoiding these areas.
Those sheet-web spiders in the highest densities of sundews were more likely to lack springtails in their guts than those caught in other locations. Sheet-web spiders produce semipermanent webs that are initially built small and gradually expand over time if  (Janetos, 1982). Spiders continually monitor the quality of their microhabitat and adjust silk output to match foraging success. This is referred to as the probe web hypothesis (Welch, Haynes, & Harwood, 2012). Thus, where sundews are dense and prey less abundant, smaller, newer sheet-webs should occur.
These spiders are more likely to move away once the foraging patch has been assessed to be of lower quality. Although not quantified during this study, we frequently found the smallest sheet-webs near dense patches of sundews lacked spiders suggesting the webs were abandoned.
Spiders are generally considered to be food limited (Anderson, 1974;Wise, 1993). Our data suggest this because fewer springtails were captured by larger sheet-webs where sundews were absent.
By virtue of being near dense patches of sundews, fewer springtails occurred, which may have limited their availability to spiders. Spiders are mobile predators capable of assessing prey levels and selecting patches where prey is abundant (Harwood, Sunderland, & Symondson, 2001, 2003Uetz et al., 1999). Spiders in our study were sit-and-wait predators (sheet-web species) and active foragers (ground-running species) both having the option to relocate although inherent risks are associated with website abandonment (Scharf, Lubin, & Ovadia, 2011 Furthermore, Barnes (2003) argued the greater the taxonomic distance between competitors, the more likely one will displace the other. Asymmetrical competition should be most extreme between species of different kingdoms to the point that amensalism (species A has a competitive effect on species B, but species B has no effect on species A) occurs (Hochberg & Lawton, 1990). Amensalism may describe the interaction between sundews and spiders under natural conditions. The potential for this asymmetry exists because sundews only extract nutrients from prey for growth and reproduction and will not die without prey (Dore Swamy & Ran, 1971;Ellison & Gotelli, 2009;Millett, Jones, & Walron, 2003, while spiders will die without prey as they acquire both nutrients and energy from prey (Toft, 2013;Wilder, 2011).
Amensalism was not observed in two previous laboratory studies. In one study (Jennings et al., 2010), wolf spiders (Rabidosa rabida) reduced seed production of pink sundews (Drosera capillaris) when prey (small crickets) availability was low. In a second study (Jennings et al., 2016), spiders (Sosippus floridanus) and oak toads (Anaxyrus quericicus) confined in terraria competed with pink sundews causing changes in sundew growth and trichome density depending on density of prey (crickets). However, none of these animal predators could relocate. Thus, the question remains whether under natural conditions spiders have a negative impact on sundews or whether the interaction is as asymmetrical as our current field study suggests.
The forms of competition to most likely negatively impact sundews are with other plants. This is especially true for small species like dwarf sundews. Sundews, like most carnivorous plants, depend on disturbance, especially fire, to compete with other angiosperms.
As larger, faster growing angiosperms outcompete sundews for space and sunlight, disturbance reduces the asymmetry of competition that is detrimental to sundews. Furthermore, intraspecific competition may also have a greater impact on sundews than from any plant-animal interaction. Those that grow in low densities may face less competition for prey such as flies and springtails, than those growing in dense patches.

| CON CLUS IONS
Doing extended field observations on sundew-spider interactions when spiders can move unimpeded is essential to understanding the dynamics between these two wet meadow predators.
Springtails were abundant prey at our study site, and they were consumed by sundews, sheet-web spiders, and ground-running spiders. Thus, the potential existed for competition between sundews and spiders albeit asymmetrical competition. Exploitative competition best describes the interaction between sundews and spiders since spiders avoid areas with high densities of sundews where they can draw down prey. However, it is uncertain whether under natural conditions, spiders can negatively impact sundews.
Thus, the interkingdom competition observed during this study is not only asymmetrical but probably an example of amensalism with spiders having no effect on sundews. Whether spiders able to move freely can have a negative impact on sundews in nature will require further investigation and field experiments.

ACK N OWLED G M ENTS
We wish to thank the Pulaski County Conservation Commission and The Nature Conservancy and the Office of the Kentucky Nature Preserves for maintaining and protecting Hazeldell Meadow. We thank Kelton Welch for being so generous with his time and expertise. We also wish to thank Mike Draney for spider identification and both Rebecca Wente and Monica Nguyen for their help in the field.
This project was funded by the Kentucky Science and Engineering Foundation (KSEF-2679-RDE-015).

CO N FLI C T O F I NTE R E S T
None declared.