Introduction
The persistence of the weed seedbank is critical for the survival and regeneration of weed populations in agroecosystems (Adams et al. Reference Adams, Marsh and Knox2005; Davis Reference Davis2006). Natural weed seed losses, particularly postdispersal seed predation (i.e., the consumption of seeds after they have been released and dispersed away from the parent plant), can significantly reduce the number of viable seeds entering the soil seedbank and thereby suppress future weed infestations (Chauhan et al. Reference Chauhan, Migo, Westerman and Johnson2010; Harrison et al. Reference Harrison, Regnier and Schmoll2003; Maron and Simms Reference Maron and Simms1997). Integrating postdispersal seed predators into weed management strategies offers a sustainable approach to reduce the weed seedbank, but their ecological roles require a deeper mechanistic understanding (Müller-Schärer et al. Reference Müller-Schärer, Scheepens and Greaves2000; van Lenteren et al. Reference van Lenteren, Bolckmans, Köhl, Ravensberg and Urbaneja2018).
Among these predators, earthworms are key soil organisms that contribute to soil structure and nutrient cycling, while also directly and indirectly influencing seedbank dynamics through seed predation and burial (Bertrand et al. Reference Bertrand, Barot, Blouin, Whalen, de Oliveira and Roger-Estrade2015; Eisenhauer et al. Reference Eisenhauer, Schuy, Butenschoen and Scheu2009, Reference Eisenhauer, Butenschoen, Radsick and Scheu2010; Lavelle Reference Lavelle1988; Lavelle et al. Reference Lavelle, Brussaard and Hendrix1999; Regnier et al. Reference Regnier, Hovick, Liu, Harrison and Diekmann2022). Several studies have reported substantial seed losses due to earthworm consumption: 30% reported in Grant (Reference Grant and Satchell1983), 20% to 100% in McRill (Reference McRill1974), and 34% to 83% in Eisenhauer et al. (Reference Eisenhauer, Schuy, Butenschoen and Scheu2009). Notably, recent studies have indicated that earthworms such as Lumbricus terrestris consume seeds selectively, potentially altering weed community composition by disproportionately removing preferred weed seeds (Clause et al. Reference Clause, Margerie, Langlois, Decaëns and Forey2011; Milcu et al. Reference Milcu, Schumacher and Scheu2006; Regnier et al. Reference Regnier, Harrison, Liu, Schmoll, Edwards, Arancon and Holloman2008, Reference Regnier, Hovick, Liu, Harrison and Diekmann2022; Shumway and Koide Reference Shumway and Koide1994; Zaller and Saxler Reference Zaller and Saxler2007). Understanding the drivers of this selectivity is important when considering earthworms as biological weed control agents in agroecosystems.
Studies have reported on earthworm seed preferences, which are influenced by multiple traits, including seed size, shape, texture, and chemical composition (Clause et al. Reference Clause, Margerie, Langlois, Decaëns and Forey2011, Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017; Shumway and Koide Reference Shumway and Koide1994). Typically, earthworms exhibit preferences for small seeds (Shumway and Koide Reference Shumway and Koide1994). In this regard, seeds within the small size class (<2 mm in length), such as redroot pigweed (Amaranthus retroflexus L.), are often preferentially consumed by earthworms, although exceptions exist (Eisenhauer et al. Reference Eisenhauer, Schuy, Butenschoen and Scheu2009; Shumway and Koide Reference Shumway and Koide1994; Willems and Huijsmans Reference Willems and Huijsmans1994). For instance, L. terrestris selectively feeds on nitrogen-rich legume seeds (e.g., white clover [Trifolium repens L.]) in preference to grass seeds (e.g., bird vetch [Vicia cracca L.]) (Eisenhauer et al. Reference Eisenhauer, Butenschoen, Radsick and Scheu2010; Zaller and Saxler Reference Zaller and Saxler2007). Lipids may also drive earthworm preferences, as demonstrated by the attraction of L. terrestris and little tree worms (Satchellius mammalis) to oil-rich seeds such as stinging nettle (Urtica dioica L.) (Clause et al. Reference Clause, Margerie, Langlois, Decaëns and Forey2011, Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). Taken together, these lines of evidence suggest that seed nutrients seem to play a key role in guiding earthworm seed choice, and earthworms are likely to include seeds in their diets for nutritional purposes (Eisenhauer et al. Reference Eisenhauer, Butenschoen, Radsick and Scheu2010; Forey et al. Reference Forey, Barot, Decaëns, Langlois, Laossi, Margerie, Scheu and Eisenhauer2011). However, interactions between seed traits, such as the interplay of nutrient content and physical limitations, remain poorly understood. Furthermore, seed ingestion by earthworms does not always result in destruction, as some seeds survive gut passage with variable effects on their germination (Aira and Piearce Reference Aira and Piearce2009; Milcu et al. Reference Milcu, Schumacher and Scheu2006). For example, Regnier et al. (Reference Regnier, Harrison, Liu, Schmoll, Edwards, Arancon and Holloman2008, Reference Regnier, Hovick, Liu, Harrison and Diekmann2022) have shown that earthworms preferentially move giant ragweed (Ambrosia trifida L.) seeds from the surface deeper into the soil profile, thereby increasing the number of seeds entering the seedbank, protecting them from surface-active seed predators to some extent, and therefore enhancing the chances for successful recruitment of A. trifida in the following seasons. Thus, quantifying the net impact of earthworm seed predation requires evaluating both ingestion preferences and postegestion viability.
Despite these advances, critical gaps persist. First, previous studies have focused on European seed species, often in non-agronomic settings, limiting their applicability to North American cropping regions such as the Northern Great Plains (NGP), where weeds such as field pennycress (Thlaspi arvense L.) and green foxtail [Setaria viridis (L.) P. Beauv.] cause significant yield losses (Aira and Piearce Reference Aira and Piearce2009; Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017; Derksen et al. Reference Derksen, Anderson, Blackshaw and Maxwell2002; Eisenhauer et al. Reference Eisenhauer, Butenschoen, Radsick and Scheu2010; Padbury et al. Reference Padbury, Waltman, Caprio, Coen, McGinn, Mortensen, Nielsen and Sinclair2002). Second, the role of lipids, a key energy source for soil fauna, in driving seed preferences has been understudied compared with nitrogen content, as have interactions between seed traits (e.g., lipid content vs. seed morphology) in mediating earthworm selectivity (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017; Eisenhauer et al. Reference Eisenhauer, Butenschoen, Radsick and Scheu2010).
This study aimed to investigate the feeding preferences of L. terrestris for six weed species common to the NGP agroecosystems and to assess how seed traits influence weed seed selection and postegestion viability. We hypothesized that L. terrestris would preferentially consume lipid-rich weed seeds and that undigested weed seeds would suffer reduced germination after passing through the guts of earthworms.
Materials and Methods
Weed Seed Species and Earthworm Species
For this study, seeds of six weeds common to the NGP region were chosen, with three weeds having high lipid content and three having low lipid content. The average lipid content used to categorize weed seeds as high or low in lipid was obtained from the literature and confirmed through our own Goldfisch oil-extraction experiments (Bretagnolle et al. Reference Bretagnolle, Matejicek, Gregoire, Reboud and Gaba2016). The weed seeds with higher lipid levels included shepherd’s purse [Capsella bursa-pastoris (L.) Medik.], T. arvense, and wild mustard (Sinapis arvensis L.), with average lipid levels of 34.28%, 23.90%, and 22.66%, respectively. By contrast, A. retroflexus, S. viridis (caryopsis), and catchweed bedstraw (Galium aparine L.), represented weed seeds with lower average lipid levels of 6.32%, 5.12%, and 4.18%, respectively (Table 1). The weed seeds were collected during the summers of 2020 and 2021 from the Kernen Crop Research Farm outside of Saskatoon, SK, Canada (52.1508°N, 106.5437°W). All seeds were stored at room temperature until use (Gan et al. Reference Gan, McTavish, Bourchier and Smith2023). Loam soil was also collected from unplanted fields at the same location and sieved through a 2-mm metal mesh to remove large plant fragments and background weed seeds.
Means of five seed characteristics for six weed seed species.
a DW, dry weight basis.
Lumbricus terrestris is an anecic earthworm species that inhabits vertical burrows in the soil and forages for litter on the soil surface (Bottinelli et al. Reference Bottinelli, Hedde, Jouquet and Capowiez2020; Bouché Reference Bouché1972).These earthworms were selected as a model species in this study because of their widespread distribution in agricultural fields across the NGP region (Reynolds Reference Reynolds2018), as well as their availability for commercial purchase. Large adults of L. terrestris (3.5 to 6.5 g fresh weight) were purchased from two bait stores in Ontario, Canada: Windsor Wholesale Bait (before November 2022) and Country Bait (after November 2022). These earthworms were hand-collected by the companies from Canadian fields and not reared under controlled conditions. Live earthworms were maintained in Styrofoam™ boxes with perforated lids to allow air circulation. Soil and paper towels were added to each box as food items, and the boxes were stored in a refrigerator at 5 C to reduce metabolic activity (Asshoff et al. Reference Asshoff, Scheu and Eisenhauer2010; Daniel et al. Reference Daniel, Kohli and Bieri1996). No additional food material was provided to avoid any interference with feeding preferences in the multiple-choice feeding experiments.
Multiple-Choice Feeding Experiment
The experimental design was adapted from Eisenhauer et al. (Reference Eisenhauer, Schuy, Butenschoen and Scheu2009) and Clause et al. (Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017) and employed a one-factor randomized complete block design. The independent factor was weed species, with six different seed species included. Each feeding arena (petri dish) was replicated 10 times, and the whole experiment was repeated twice.
Before feeding, all earthworms were starved for 48 h in separate petri dishes (Ø = 15.0 cm, depth = 2.0 cm, with lid, Sarstedt, Nümbrecht, NRW, Germany) to void their guts (i.e., empty the contents of their gut) and to standardize the level of hunger among earthworms (Eisenhauer et al. Reference Eisenhauer, Schuy, Butenschoen and Scheu2009). A piece of moistened filter paper (Ø = 15.0 cm, P5, Medium, Fisherbrand, Pittsburgh, PA, USA) was placed in each petri dish and moistened with 5 ml of deionized water (Ripple FX Water, 4 L, New Westminster, BC, Canada) to maintain an appropriate moisture level. A single earthworm was then placed in the petri dish, and the dishes were kept in a growth chamber (Thermo Scientific, Precision Model 818, Marietta, OH, USA) at 15 C under complete darkness (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017; Eisenhauer et al. Reference Eisenhauer, Butenschoen, Radsick and Scheu2010).
After starvation, each earthworm was rinsed with distilled water and placed individually into a separate experimental feeding arena (one worm per petri dish). The feeding arena comprised a petri dish (Ø = 15.0 cm, depth = 2.0 cm, with lid) lined with a moist filter paper (moistened with 5 ml of deionized water), 1 g of sieved soil, 0.1 g of pure sand (CaribSea, Fort Pierce, FL, USA), and 20 seeds each of six weed species. A total of 120 seeds (i.e., 20 seeds of each weed) were randomly placed in each feeding arena. Soil and sand were added separately and spread evenly in the petri dishes to simulate natural environmental conditions and enhance the grinding of ingested food material in earthworm gizzards (Curry and Schmidt Reference Curry and Schmidt2007; Eisenhauer et al. Reference Eisenhauer, Schuy, Butenschoen and Scheu2009, Reference Eisenhauer, Butenschoen, Radsick and Scheu2010). Sterilized sand was chosen to prevent microbial interference. The petri dishes were then placed in a growth chamber at 15 C under complete darkness for 24 h (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). After the 24-h ingestion period, weed seeds remaining in the feeding arena were counted by species and were treated as non-ingested. The percentage of weed seed ingestion was calculated according to Equation 1:
$$\eqalign{& {\rm{Weed}}\;{\rm{seed}}\;{\rm{ingestion}}\; \cr & {\rm{ = }} \;{{{\rm{No}}.\;{\rm{of}}\;{\rm{seeds}}\;{\rm{offered}}\;{\rm{ - }}\;{\rm{No}}.\;{\rm{of}}\;{\rm{remaining}}\;{\rm{seeds}}} \over {{\rm{No}}.\;{\rm{of}}\;{\rm{seeds}}\;{\rm{offered}}}}\; \times \;{\rm{100}}\% }$$
Weed Seed Egestion and Digestion by Earthworms
After the completion of a 24-h ingestion period, earthworms were gently rinsed and transferred to a new set of petri dishes, which were lined with moist filter paper. These petri dishes were then placed in a growth chamber at 15 C under complete darkness for another 48 h to allow earthworms to egest seeds (Eisenhauer et al. Reference Eisenhauer, Schuy, Butenschoen and Scheu2009). Any seeds fused with worm cast were treated as egested (i.e., excreted from worm gut) and were counted and collected for viability testing. Weed seeds that disappeared from experimental arenas were considered completely digested by earthworms. The percentage of weed seed digestion was calculated according to Equation 2:
$$\scriptsize{\eqalign{& {\rm{Weed}}\;{\rm{seed}}\;{\rm{digestion}}\;\cr& {\rm{ = }}\;{{{\rm{No}}.\;{\rm{of}}\;{\rm{seeds}}\;{\rm{offered}}\;{\rm{ - }}\;{\rm{No}}.\;{\rm{of}}\;{\rm{remaining}}\;{\rm{seeds}}\;{\rm{ - }}\;{\rm{No}}.\;{\rm{of}}\;{\rm{egested}}\;{\rm{seeds}}} \over {{\rm{No}}.\;{\rm{of}}\;{\rm{seeds}}\;{\rm{offered}}\;{\rm{ - }}\;{\rm{No}}.\;{\rm{of}}\;{\rm{remaining}}\;{\rm{seeds}}}} \times {\rm{100}}\% }}$$
Weed Seed Viability after Egestion
Germination of weed seeds recovered from earthworm casts was tested to assess the influence of gut passage on seed viability. Egested seeds were transferred to individual petri dishes each lined with a filter paper moistened with 5 ml of deionized water and incubated in darkness at 15 C for 7 d in a growth chamber. These conditions were chosen to simulate the low-light environment expected for seeds deposited in the soil seedbank by anecic earthworms (Asshoff et al. Reference Asshoff, Scheu and Eisenhauer2010). Control treatments were composed of 20 intact seeds (i.e., weed seeds that had not been used in prior experiments) of each weed species in a petri dish lined with a piece of moist filter paper. The germination experiments were replicated five times for each weed species. Water was added as needed during the germination test. The number of germinated weed seeds was recorded, and the percentage of germination was calculated according to Equation 3 for each weed species.
$${\rm{Weed}}\;{\rm{seed}}\;{\rm{germination}}\;{\rm{ = }}\;{{{\rm{No}}.\;{\rm{of}}\;{\rm{germinated}}\;{\rm{seeds}}} \over {{\rm{No}}.\;{\rm{of}}\;{\rm{egested}}\;{\rm{seeds}}}}\; \times \;{\rm{100}}\% $$
Assessment of Weed Seed Characteristics
The lipid content of weed seeds was determined using the Goldfisch oil-extraction method, with four replicates per seed species (Bhatty Reference Bhatty1985). For each replicate, 2 g of ground seed material (Black & Decker spice grinder, Middleton, WI, USA) was extracted with 40 ml of petroleum ether for 3 h in a continuous circulation system. The recovered lipid was dried in an oven at 105 C for 2 h to remove residual solvent and moisture (Bhatty Reference Bhatty1985). Seed lipid content was calculated as the weight loss of the sample after extraction, expressed relative to the original sample weight.
The length, width, and thickness of weed seeds were measured using a digital caliper (Neiko Tools, Corona, CA, USA). Seed shape index was quantitatively determined following Clause et al. (Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017) and Thompson et al. (Reference Thompson, Band and Hodgson1993), which defined the extent to which a seed deviated from a perfectly spherical shape. To calculate this, seed length, width, and maximum thickness were divided by seed length, which made length consistent across all species. The variance of these three parameters was then calculated to obtain the seed shape index (as shown in Equation 4; Thompson et al. Reference Thompson, Band and Hodgson1993). This index ranges from zero (i.e., perfectly spherical shape) to 0.3 (elongated or needle-like shape). Seed measurements (i.e., length, width, and shape) were cross-validated according to descriptions provided by Bojnanský and Fargašová (Reference Bojnanský and Fargašová2007) (Table 1); readers seeking more detailed description of the seed species can refer to this source.
$${\rm{Seed}}\;{\rm{shape}}\;{\rm{index}}\;{\rm{ = }}\;{[{{\sum {{{({\rm{x}}\;{\rm{ - }}\;{\rm{\bar x}})}^{\rm{2}}}} } \over {\rm{n}}}]^{{\rm{1/2}}}}$$
where x represented seed length, width, and thickness, respectively;
${\it{\bar x}}$
represented the average number of seed length, width, and thickness; n was the sample size.
Data Analysis
Weed seed ingestion, digestion, and germination data were analyzed using mixed-effects models in R statistical software (v. 4.4.2; R Core Team Reference Core Team2023) using the lmer function from the lmerTest package (Kuznetsova et al. Reference Kuznetsova, Brockhoff and Christensen2017). The mixed model was fit by using weed species as the fixed effect and replicate as a random blocking factor, as all six weed seeds were nested within each feeding arena (i.e., offered in the same petri dish). Model assumptions were evaluated by inspecting residual plots. Pairwise contrasts among weed species were performed using estimated marginal means with the emmeans package in R (Lenth et al. Reference Lenth, Singmann, Love, Buerkner and Herve2018).
The variability of weed seed characteristics was analyzed using principal component analysis (PCA) as an exploratory approach to reveal patterns of covariation among seed traits (sensu Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). Seed lipid content, length, and thickness were log-transformed and seed ingestion data were arcsin-transformed to improve normality (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). All variables were then centered and standardized by standard deviation (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). Weed seed ingestion and digestion by L. terrestris were included in the PCA as quantitative illustrative variables, which were projected on the PCA axes to evaluate the associations with weed species and seed characteristics (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). In addition, Pearson’s correlation analysis was conducted to quantify the association between seed ingestion/digestion and individual seed characteristics.
Structural equation modeling (SEM) was then employed as a confirmatory approach to test hypothesized causal relationships between seed characteristics and ingestion and to determine which traits most strongly influenced earthworm seed selection (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). By simultaneously accounting for correlated predictors, SEM provides a statistically robust framework to quantify the relative contributions of each trait and to distinguish between direct and indirect effects on earthworm ingestion (Krishnakumar and Nagar Reference Krishnakumar and Nagar2008; Ullman and Bentler Reference Ullman, Bentler and Weiner2012). Variables included were selected based on the results of PCA analysis and correlation analysis, and those variables were log- or arcsin-transformed (as described earlier) before the SEM analysis. The model fit was assessed using a chi-square test and the root-mean-square error of approximation (RMSEA), with acceptable fit indicated by a nonsignificant chi-square and low RMSEA (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). Results were presented as standardized path coefficients with associated P-values. This analysis was conducted in R using the lavaan package (Rosseel Reference Rosseel2012) and visualized with lavaanPlot (Lishinski Reference Lishinski2024).
Results and Discussion
Weed Seed Ingestion and Selection: Lipid-Driven Preferences
Seed ingestion of L. terrestris differed among weeds offered in the multiple-choice feeding arenas. Overall, the highest ingestion was observed for S. arvensis and C. bursa-pastoris (97% and 96%, respectively). Ingestion of T. arvense seeds was 82%. In contrast, only 71% and 70% of G. aparine and A. retroflexus seeds were ingested, respectively. Setaria viridis seeds had the lowest ingestion, averaging 44% (Figure 1).
Lumbricus terrestris seed ingestion percentages in a six-choice seed-feeding experiment. Mean values of seed ingestion are shown with bars indicating standard error of the mean. The letters above the bars indicate significant differences among weed species (derived from emmeans in R and adjusted by Tukey’s method, α < 0.05). Seed species were arranged based on their average lipid content levels from the highest to the lowest: 34.28%, 23.90%, 22.66%, 6.32%, 5.12%, and 4.18%, respectively, from the left to right. The purple bars represent weed species with high lipid content, whereas the grey bars represent species with low lipid content.
The PCA analysis revealed that the first two principal components captured a large proportion (88.79%) of the total variation in seed characteristics (Table 2; Figure 2). The first principal component, which was primarily influenced by seed size attributes such as width, thickness, and length, accounted for 63.78% of the total variability. The second principal component, which was primarily influenced by seed shape and lipid content, explained an additional 25.01% of the total variability. Along this component, weed species with irregular shapes and low lipid content (e.g., A. retroflexus) were separated from species with a round shape and relatively high lipid content (e.g., S. arvensis). Variation in weed seed ingestion was predominantly associated with the second component (69.9%), indicating that L. terrestris preferentially ingested seeds that were round and lipid-rich.
Eigenvector scores of seed characteristics in three main principal component analysis (PCA) components (PC1–PC3) a .
a Values in parentheses indicate variance accounted for by each component.
b Seed ingestion and digestion responses were used as illustrative variables to test their association with different seed traits.
Principal component analysis biplot of six weeds (points) and five seed characteristics (solid arrows). Lumbricus terrestris seed ingestion and digestion percentages were projected onto the biplot as illustrative variables (dashed arrows). Weed seeds include Capsella bursa-pastoris, Thlaspi arvense, Sinapis arvensis, Amaranthus retroflexus, Setaria viridis, and Galium aparine.
Correlation analysis revealed weed seed ingestion by L. terrestris was positively correlated with lipid content (Figure 3). By contrast, seed length, width, and shape were negatively correlated with weed seed ingestion, indicating that L. terrestris tended to avoid long, wide weed seeds and seeds of elongated shape. SEM confirmed that lipid content was the primary determinant of weed seed ingestion by L. terrestris (Figure 4). The overall SEM model explained 59.8% of the variance for weed seed ingestion by L. terrestris, and the lipid content of weed seed emerged as the strongest predictor of ingestion by these earthworms (based on a standardized path coefficient of 0.57).
Correlation matrix for weed seed ingestion and digestion and seed characteristics, including seed lipid content, length, width, thickness, and shape. Pearson’s correlation coefficients are provided for each variable combination, with positive correlations shown in green and negative correlations shown in blue. Significance levels are indicated as follows: ***P < 0.001; **P < 0.01; *P < 0.05; n.s. = nonsignificant.
Structural equation modeling of the effects of seed characteristics of six weeds on seed ingestion by Lumbricus terrestris earthworms. Standard path coefficients are shown next to each arrow. The model demonstrated good fit to the data (χ2 = 0.006, df = 1, P = 0.939) and explained 59.8% of the variance in seed ingestion (R2 = 0.598).
Our results support the hypothesis that earthworm preferences differ between weed species and that weed seed ingestion by L. terrestris was likely driven by seed lipid content. Weed seeds with high lipid levels (averaging 22.66% to 34.28%) were preferentially ingested by L. terrestris, which was consistent with another study suggesting seed lipid content is a critical factor in determining seed palatability to L. terrestris (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). The preference of L. terrestris for lipid-rich seeds could be related to earthworm biology and feeding ecology (Curry and Schmidt Reference Curry and Schmidt2007). Earthworms possess sensory cells in the buccal cavity epithelium and chemoreceptors in the prostomium that allow them to detect chemical compounds, including sugars and volatile organic compounds (Edwards and Arancon Reference Edwards and Arancon2022; Laverack Reference Laverack1960; McManus and Wyers Reference McManus and Wyers1979; Zirbes et al. Reference Zirbes, Mescher, Vrancken, Wathelet, Verheggen, Thonart and Haubruge2011). The detection of seed-derived chemical signals is thus the likely sensory mechanism that enables earthworms to assess the chemical composition of seed species to identify and select seeds that have high lipids. Lipids are essential nutrients for the biology of earthworms, as they offer a very efficient energy source that is more readily absorbed than proteins (Levey and Grajal Reference Levey and Grajal1991). Earthworms like Aporrectodea longa accumulate and store long-term energy in the form of lipids before they enter diapause. Consequently, L. terrestris may meet its lipid requirements and in so doing reduce the seedbank size of certain weeds by preferentially consuming lipid-rich weed seeds.
The preference of L. terrestris for seed physical traits, however, can be overridden by lipid preference, which aligns well with previous literature (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017). Lumbricus terrestris consumed more large weed seeds with higher lipid content (e.g., S. arvensis) than small seeds with lower lipid content (e.g., A. retroflexus), despite size constraints (Willems and Huijsmans Reference Willems and Huijsmans1994). This suggests that the morphological characteristics of seeds may pose physical limitations that can affect the ability of earthworms to ingest weed seeds. In cases where these physical limitations are not highly impactful (i.e., do not significantly affect earthworm ingestion), the nutritional value of weed seeds, especially the lipid content, is likely to be a key driver of weed seed choice by L. terrestris earthworms. Furthermore, as the lipid content of the tested weed seeds was negatively correlated with seed length in our study, longer seeds may have less lipid content. Thus, there may be an optimal seed length/lipid concentration ratio that determines the attractiveness of weed seeds to earthworms as proposed by Clause et al. (Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017).
Our results indicated that weed seed ingestion is negatively influenced by irregular seed shapes and long seed length. This contradicts previous findings that seed length and shape did not affect seed selection by L. terrestris (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017) and further highlights that variations between experiments and seed species may be due to the complex interactions between species-specific traits of weed seeds. Therefore, generalized conclusions remain untenable, and large studies with numerous weed seeds of various species are needed to develop a more comprehensive understanding of earthworm weed seed preferences.
Weed Seed Digestion by Lumbricus terrestris
The digestion of weed seeds also varied among species. Capsella bursa-pastoris seeds had the highest digestion (32%), followed by G. aparine seeds (26%) (Figure 5). Seeds of S. viridis, T. arvense, S. arvensis, and A. retroflexus were digested less often by L. terrestris, and no differences were observed among these four weeds. Variation of seed digestion among weeds has been shown by Eisenhauer et al. (Reference Eisenhauer, Schuy, Butenschoen and Scheu2009), who also reported that seed digestion by L. terrestris is determined by seed species identity. The low digestion of seeds (i.e., high-quality food) by L. terrestris is consistent with some research suggesting that earthworms typically consume poor-quality food and feature low food assimilation rates (Curry and Schmidt Reference Curry and Schmidt2007; Eisenhauer et al. Reference Eisenhauer, Schuy, Butenschoen and Scheu2009). The relatively low digestion shown by L. terrestris also suggests digestion varies among earthworm species. For example, green worms (Allolobophora chlorotica) digested almost all seeds that were ingested (e.g., spiny restharrow [Ononis spinosa L.] and bird’s-foot trefoil [Lotus corniculatus L.]) (Clause et al. Reference Clause, Forey, Eisenhauer, Seal, Soudey, Colville and Barot2017).
Lumbricus terrestris seed digestion percentages in a six-choice seed-feeding experiment. Mean values of seed digestion are shown with bars indicating standard error of the mean. The letters above the bars indicate significant differences among weed species (derived from emmeans in R and adjusted by Tukey’s method, α < 0.05). The purple bars represent weed species with high lipid content, whereas the grey bars represent species with low lipid content.
No correlation existed between weed seed ingestion and digestion, which agrees with previous studies (Eisenhauer et al. Reference Eisenhauer, Schuy, Butenschoen and Scheu2009). In our experiment, the most frequently digested weed seeds were C. bursa-pastoris and G. aparine, whereas the most frequently ingested weed seeds were S. arvensis, C. bursa-pastoris, and T. arvense. This might be because weed seed digestion by L. terrestris is highly impacted by the morphological characteristics of seeds (i.e., seed length, width, thickness, and shape), as seed digestion had a negative correlation with seed length and width. As such, small weed seeds, especially those shorter in length, are more easily digested by L. terrestris and are less likely to survive the grinding of sand particles and other intestinal material (Eisenhauer et al. Reference Eisenhauer, Butenschoen, Radsick and Scheu2010). However, this can only explain the high digestion for C. bursa-pastoris but not G. aparine, which has a much stronger seed coat and a rough surface texture. One explanation is that seeds with thick (or hard) seed coats may require a longer time to be digested, as seeds are protected by lignified seed coats, but lignin generally is not digested by earthworms without the participation of microorganisms (Curry and Schmidt Reference Curry and Schmidt2007; Vidal et al. Reference Vidal, Watteau, Remusat, Mueller, Nguyen Tu, Buegger, Derenne and Quenea2019). Digestion of seeds thus requires damage to the seed coat before seeds can be processed. This is supported by research showing that seeds that are present longer in the worm gut are more likely to be destroyed (Gan et al. Reference Gan, McTavish, Bourchier and Smith2023). Combining this evidence, and given that G. aparine had the largest size among weed seeds tested and has a rough, hard seed coat with trichomes, these seeds are likely to remain longer in the worm gut, resulting in higher damage levels on seed coats due to the gizzard contraction along with sand particles (Gan et al. Reference Gan, McTavish, Bourchier and Smith2023). Ultimately, this could result in a higher digestion. Such trait- or species-specific impacts could reshape weed communities by disproportionately reducing the seedbank size of certain weeds.
Seed Viability after Passage through the Gut of Lumbricus terrestris
Earthworm gut passage altered germination outcomes in a species-specific manner (Laossi et al. Reference Laossi, Noguera and Barot2010). Sinapis arvensis weed seeds had increased germination after egestion (by 9.1%) (Figure 6). In contrast, C. bursa-pastoris, T. arvense, and A. retroflexus weed seeds had notable declines in germination after egestion, decreasing from 67.7%, 38.9%, and 56.9% (i.e., controlled seed germination tests from fresh seeds) to 12.3%, 26.9%, and 16.2%, respectively. Germination of G. aparine and S. viridis remained unchanged. These results are in line with prior research showing that the influence of earthworm ingestion on weed seed germination can be positive (Asshoff et al. Reference Asshoff, Scheu and Eisenhauer2010; Eisenhauer et al. Reference Eisenhauer, Butenschoen, Radsick and Scheu2010), negative (Clause et al. Reference Clause, Margerie, Langlois, Decaëns and Forey2011, Reference Clause, Barot and Forey2015; Decaëns et al. Reference Decaëns, Mariani, Betancourt and Jiménez2003; Drouin et al. Reference Drouin, Bradley, Lapointe and Whalen2014; Li et al. Reference Li, Fan, Qian, Yuan, Meng, Guo and Lv2020; Milcu et al. Reference Milcu, Schumacher and Scheu2006; Regnier et al. Reference Regnier, Harrison, Liu, Schmoll, Edwards, Arancon and Holloman2008), or neutral (Grant Reference Grant and Satchell1983). The reasons behind these variations in germination are likely multifaceted and include weed seed size, seed coat hardness, and other factors, but more research is needed to confirm the relationship between these factors.
Germination percentages of seeds recovered from earthworm casts.
Generally, seeds ingested by earthworms are subjected to both physical and chemical damage during their transit through the worm gut (Clause et al. Reference Clause, Barot and Forey2015). The species-specific differences in germination potential may depend on seed size and seed coat hardness. During gizzard contractions in the earthworm gut, seeds can suffer physical damage to their seed coats, which can alter germination potential. Seed germination could thus decrease or seeds can become unviable after being destroyed by the grinding of gut contents, especially small weed seeds (Aira and Piearce Reference Aira and Piearce2009; Milcu et al. Reference Milcu, Schumacher and Scheu2006; Regnier et al. Reference Regnier, Harrison, Liu, Schmoll, Edwards, Arancon and Holloman2008). This could explain the observed declines in germination for C. bursa-pastoris, T. arvense, and A. retroflexus. On the other hand, seeds may germinate due to the weakening of seed coats and breaking of physical dormancy, which is likely to lead to higher seed germination rates, as was observed for S. arvensis in our experiment (Ayanlaja et al. Reference Ayanlaja, Owa, Adigun, Senjobi and Olaleye2001; Koptur Reference Koptur1998). However, variation in germination also depends on the damage to seed coats (Koptur Reference Koptur1998; Stanway Reference Stanway1974), and is ultimately influenced by the intestinal contents of earthworms such as grit levels (i.e., sand particles). According to Gan et al. (Reference Gan, McTavish, Bourchier and Smith2023), mixed feeding with grit and seeds embeds grit onto seed coats, and an increase in grit levels can lead to higher damage during transit through the worm gut. Thus, the addition of sand may facilitate the breaking of seed coat or seed dormancy, resulting in increased seed germination.
Because the influence of earthworms on weed seed germination is species specific, their selective feeding behavior could affect weed seed dynamics and weed communities (Clause et al. Reference Clause, Margerie, Langlois, Decaëns and Forey2011). Such impacts could be substantial, especially in fields where earthworms occur in high density or where earthworms are introduced as exotic species (Cassin and Kotanen Reference Cassin and Kotanen2016; Cray et al. Reference Cray, Gaudon and Murphy2022). Therefore, as earthworm abundance increases, the weed seedbank is more likely to be affected (Decaëns et al. Reference Decaëns, Mariani, Betancourt and Jiménez2003). Given that L. terrestris is an exotic species that is widely distributed across the NGP region (Reynolds Reference Reynolds2022), their specific preference for seeds high in lipids may alter the seedbank dynamics by vertical translocation of the preferred seeds or by direct digestion (Regnier et al. Reference Regnier, Harrison, Liu, Schmoll, Edwards, Arancon and Holloman2008). As a result, these seed species may exhibit lower recruitment in the field, thereby impacting weed community composition. A recent study in China supports the idea that earthworms influence the weed community (Li et al. Reference Li, Fan, Qian, Yuan, Meng, Guo and Lv2021). The same study showed that the introduction of earthworms (Pheretima guillelmi) into field microcosms (four 6 by 50 m plots under rotation farming with corn [Zea mays L.] and broccoli [Brassica oleracea (L.) var. botrytis] before the experiments) significantly suppressed total weed density and biomass by 59.2% and 66.6%, respectively, compared with plots earthworms were absent. To our knowledge, however, only a few studies conducted in the NGP region have considered the direct impact of earthworm introduction on weed seedbank dynamics and weed recruitment in agricultural fields (Regnier et al. Reference Regnier, Harrison, Liu, Schmoll, Edwards, Arancon and Holloman2008, Reference Regnier, Hovick, Liu, Harrison and Diekmann2022; Smith et al. Reference Smith, Gross and Januchowski2005). Large-scale field research is needed on earthworm–weed seed interactions, seed palatability to and selection by earthworms, and the consequences of earthworm feeding on weed seed dynamics.
In summary, our study demonstrates that L. terrestris earthworms selectively ingest weed seeds based primarily on seed lipid content and morphology, with secondary effects on seed egestion and seed viability. The impacts of earthworms on seed germination were rather complex, and therefore weed seed–earthworm dynamics could be species specific. Nonetheless, this experiment supports our initial hypothesis that earthworms discriminate between the weed seeds common to the NGP region and select certain seed species for ingestion/consumption. Our data suggest that the lipid content of seed species is likely to play an important role in seed preference by L. terrestris earthworms when seed physical characteristics do not impose major constraints on ingestion. Where lipid-rich weed seeds are rare or challenging to consume, small and spherical seeds are likely to be preferred. Given this, earthworms would be expected to act as weed seed predators for many species, and their feeding preferences may lead to shifts in weed communities when present at high enough densities. Future research should consider field validation and explore interactions with soil microbiota, grit content, and other weed seed predators.
Acknowledgments
We would like to thank Natural Sciences and Engineering Research Council of Canada (NSERC) for providing funding support.
Funding statement
Funding was provided by a Discovery Grant awarded to CJW by the Natural Sciences and Engineering Research Council of Canada.
Open access
