Prey nutrient content is associated with the trophic interactions of spiders and their prey selection under field conditions

Materials and Methods

Fieldwork

Money spiders (Araneae: Linyphiidae) and wolf spiders (Araneae: Lycosidae), the two most abundant spider groups in this study, were visually located along transects in two adjacent barley fields at Burdons Farm, Wenvoe in South Wales (51°26'24.8"N, 3°16'17.9"W) and collected from occupied webs and the ground in daylight hours between April and September 2018. Each belt transect was adjacent to a randomly selected crop tramline and were distributed across the entire field and ran its length. The areas searched were 4 m2 quadrats at least 10 m apart and all observed linyphiids and lycosids were collected. The 300 spiders taken forward for molecular dietary analysis in this study were taken from 64 randomly selected locations along the aforementioned transects. Following collection of spiders, 4 m2 of ground and crop stems was suction sampled in each of these 64 sampling locations for approximately 30 seconds, with the collected material emptied into a bag and any organisms immediately killed with ethyl-acetate. Suction sampling used a ‘G-vac’ modified garden leaf-blower. All material was later frozen at -20 ºC for storage before sorting in the lab. Sticky trap data were also collected, but were not used in this study as suction sampling was found to represent the interactions of spiders more closely (Cuff, Tercel et al., 2024). These invertebrates were collected for background population densities and macronutrient analysis, not for molecular dietary analysis.

All invertebrates were identified to family level using morphological keys: Araneae (Roberts, 1993), Diptera (Ball, 2008), Coleoptera (Duff, 2012), Hymenoptera (Goulet & Huber, 1993), Hemiptera (Unwin, 2001), Collembola (Dallimore & Shaw, 2013) and Chilopoda (Barber, 2008). Further identifications were not carried out due to the inability to identify some of the invertebrate groups further via the associated metabarcoding-derived dietary data (e.g., Sciaridae), and the difficulty associated with finer taxonomic resolution of many damaged or immature specimens. The only taxa not identified to family level were springtails of the superfamily Sminthuroidea (Sminthuridae and Bourletiellidae, which were often indistinguishable following suction sampling and preservation due to the fine features necessary to differentiate them) which were left at super-family, mites (many of which were immature or in poor condition, or lacked appropriate taxonomic keys) which were identified to order level and wasps of the superfamily Ichneumonoidea (which were identified no further due to obscurity of wing venation due to damage); in these cases, these taxonomic assignments were pooled to family-level for later analyses.

Extraction, amplification and sequencing of DNA from the individually collected spiders, and its bioinformatic analysis are described by Cuff, Tercel, et al. (2022) and Drake et al. (2022) and are also detailed in Supplementary Information 1. In short, dietary metabarcoding was carried out using two primer pairs, one excluding predator DNA and the other amplifying it, to overcome the problem of overamplification of predator DNA (Cuff, Kitson, et al., 2023). Amplified DNA was sequenced on an Illumina MiSeq V3 2x300 cartridge, and resultant data screened for false positives following bioinformatic processing via minimum sequence copy thresholds applied according to read counts in controls and control DNA counts present in samples (Drake et al., 2022).

Macronutrient determination

Specimens were taken for macronutrient analysis from the same suction samples collected for invertebrate community identification. Representatives were taken from each family found in the community samples for which specimens were intact, in visually good condition and relatively clean of soil and other contaminants. If specimens were from a relatively uncommon family but unclean, soil and other surface contaminants were physically removed, and the specimen then momentarily dipped in water to remove remaining surface contaminants without greatly dislodging surface lipids. Macronutrient contents were determined following the MEDI protocol (Cuff, Wilder, et al., 2021; Cuff & Wilder, 2021) with minor alterations to account for the small size of most of the invertebrates processed (Cuff, 2021) and with the omission of exoskeletal measurement. During extraction, half volumes (i.e., 500 µl) of solvents were used. For the lipid assays, 15 µl of sulfuric acid was added for a 15 min incubation, followed by only 200 µl of vanillin reagent to increase the concentration and development of analyte for more accurate readings from smaller invertebrates. Lipid and protein standard series were diluted to 50% of the concentration specified in the original protocol (i.e., 0-1 mg ml-1). Carbohydrate assays used 140 µl of reagent with 30 min incubation at 92 °C followed by a further 30 min at room temperature. Carbohydrate standard series were diluted to 1 % of the concentrations specified in the original protocol (i.e., 0-0.02 mg ml-1) to ensure signals overcame the higher limit of detection relative to typical invertebrate carbohydrate content.  Mean macronutrient contents were calculated for each taxon and converted into proportions of the total macronutrient mass detected for each taxon (i.e., macronutrient values are given as % total macronutrient mass). Macronutrient data were allocated to each prey taxon. Where macronutrient data were not available for a family (due to no or very few individuals being present in vacuum samples), average data for that order were used.

Statistical analysis

We have assessed nutritional dynamics through a combination of multivariate models and network-based null modelling. All analyses were conducted in R v.4.0.3 (R Core Team, 2020).

To compare the nutritional balance of prey consumed by different spider groups, the mean nutrient contents of all prey consumed by each spider were calculated and compared using a multivariate linear model (MLM) via the ‘manylm’ command in mvabund (Wang et al., 2012). Differences were visualised using ternary plots via ‘ggtern’ (Hamilton & Ferry, 2018) and ‘ggplot2’ (Wickham, 2016). How spider diets differ between spider groups (genera, sexes and life stages) and how this is related to the nutrient contents of those prey was assessed using a fourth corner analysis (FCA). Fourth corner analyses assess how the relationship between the presence of species (or consumed resources in a dietary context) and environmental (or consumer) traits relates to species traits (or prey traits; (Brown et al., 2014). First, overall relationships between dietary composition and spider traits were assessed using a multivariate generalized linear model (MGLM) via the ‘manyglm’ command in the ‘mvabund’ package (Wang et al., 2012) with a binomial error family. These relationships were identified via likelihood ratio test using the ‘anova.manyglm’ command. A fourth corner analysis was performed using the ‘trait.glm’ command in mvabund with the ‘R’, ‘Q’ and ‘L’ matrices representing dietary detections of prey families in each spider, spider trait data (genus (a proxy for many unmeasured traits such as morphology), sex and life stage) and prey proportional macronutrient contents, respectively, with a binomial error family. Log-likelihood ratio tests were carried out using the ‘anova.traitglm’ command with 999 bootstrap iterations and Monte-Carlo resampling. The model was repeated with the least absolute shrinkage and selection operator (LASSO) applied, which is a method of penalised likelihood that reduces model terms to zero if they lack predictive power (i.e., do not reduce the Bayesian information criterion), thereby selecting models with greater predictive accuracy (Brown et al., 2014).

To assess whether the proportions of mean prey nutrient contents deviated from those expected based on random foraging, null diets were simulated using network-based null models in ‘econullnetr’ (Vaughan et al., 2018) with the ‘generate_null_net’ command. The ‘generate_null_net_indiv’ function (Cuff, Windsor, et al., 2023) was used to generate null diets for each individual spider based on local prey communities determined via suction sampling. The mean prey macronutrient contents of spider diets were compared between expected and observed diets using a MLM in mvabund, and significant differences visually represented through a ternary plot using ggtern. To ascertain how differences between spider groups factor into any deviations from random nutrient intake, the difference in macronutrient proportions between expected and observed spider diets was also compared between spider genera, life stages and sexes in a MLM.

To relate prey preferences of different spider groups to different prey and their macronutrient contents, observed interactions were compared against null models based on prey abundances using the ‘generate_null_net’ command in econullnetr (as above) for each of the spider groups and, separately, for individual spiders. Ternary plots representing preference effect sizes for prey of varying macronutrient contents were generated using the group-specific data via ‘ggtern’. The observed interactions of individual spiders were divided by the interactions expected in the null model; infinite values (i.e., zero interactions expected and more than zero observed) and NAs (e.g., no interactions expected nor observed) were converted to zero. These observed/expected values were compared between spider groups via permutational multivariate analysis of variance (PerMANOVA). These results were visualised by plotting mean standardised effect sizes for each spider genus, sex and life stage from the prey choice null models via ggplot2.