Litter consumption by macrodetritivores depends more on mechanical than on nutritional constraints
Abstract
Ecosystem functions greatly depend on trophic interactions between consumers and their resources. Resource consumption depends on ingestion, digestion, and allocation processes. Mechanical constraints are expected to influence ingestion, while metabolic and nutritional constraints are expected to influence allocation. Leaf litter are resources presenting a high mechanical and nutritional heterogeneity that depends on plant identity and on physical and microbial processing over the course of decomposition. Litter consumption by detritivores is known to depend on metabolic and nutritional constraints but the importance of mechanical constraints is yet unknown. After accounting for metabolic constraints on consumption rate, we tested the relative importance of mechanical and nutritional constraints in explaining litter consumption rates by detritivores. For this, we exposed 16 leaf treatments (eight leaf species either just leached or leached and microbially conditioned) to four aquatic and five terrestrial detritivore taxa in laboratory no-choice consumption experiments. We investigated two mechanical constraints: grabbing and fragmenting the resource, by measuring suitable couples of mechanical traits for both litter and detritivores. We also investigated four nutritional constraints related to N, P, K and Ca contents in both detritivores and litter. For each constraint, we also tested if trait matching significantly contribute to explain consumption. Our analysis revealed that both mechanical and nutritional constraints are influencing mass-independent consumption rate but that mechanical constraints predominate over nutritional constraints. Litter fragmentation, studied through litter toughness and detritivore biting force, was especially important to explain consumption rate. Nutritional constraints were dominated by P constraints. Trait-matching had very weak importance and was significant only for P constraints. Our findings highlight the importance of mechanical constraints for litter consumption by detritivores.
Introduction
Ecosystem functions greatly depend on trophic interactions between consumers and their resources (Hines et al. 2015, Schmitz et al. 2018, Thompson et al. 2012). A way to better understand feeding interactions involving animals is to describe the trophic network by functional metrics (e.g. functional traits) (Gravel et al. 2016, Schleuning et al. 2023). Wootton et al. (2023) notably built a modular theory of trophic interactions for co-occurring species. According to this framework, consumers have to go through several steps ruling resource location, decision to feed, attack and consumption. Specifically, resource consumption theoretically depends on ingestion, digestion, and allocation processes. The feasibility and the strength of each step of the pairwise consumer‒resource interaction could be explained and predicted by some traits of both the consumer and the resource (Schleuning et al. 2015, Gravel et al. 2016, Wootton et al. 2023). It was notably hypothesized that trophic interactions could be explained not only by traits of consumers and resources alone, but also by a match between those traits (Bartomeus et al. 2016, Laigle et al. 2018). The modularity of Wootton's framework lies in fact that the relative importance of the different steps could change regarding of the type of interaction (e.g. predation, detritivory).
Consumption rates are limited by different constraints. First, ingestion of a resource by a co-occurring consumer partly depends on mechanical constraints. For instance, a match between carabid beetle biting force and prey cuticular toughness successfully predicted trophic interactions (Brousseau et al. 2018). These constraints impact the handling time, i.e. the time needed for a consumer to handle and ingest its resource (Ott et al. 2012). A longer handling time will result in a lower consumption rate. Chemical deterrents such as toxic, repulsive or unpalatable molecules (e.g. tannins, lignin) can also modulate the ingestion rate (David 2014, Wootton et al. 2023). Second, allocation of digested resource can be divided into allocation of energy and nutrients. As a consumer has to fulfil its demands for survival, growth, and reproduction, consumption rate is expected to depend on consumer's metabolic rate which is linked to body mass (Brown et al. 2004, Rall et al. 2012), and on resource's energy content (Neu et al. 2023). Yet, repeated observations of deviations from predictions based on an energy-metabolic approach, led to the conclusion that energetical constraints are far from entirely driving consumption rate (Pawar et al. 2012, Rall et al. 2012, Rota et al. 2018). Lastly, as a consumer also has to meet its demands in nutrients, consumption rate is also expected to depend on both consumer and resource nutrient composition. Ecological stoichiometry theory predicts that consumers will perform better with resources having elements in similar proportions as in their own body (Sterner and Elser 2017). An outcome of this prediction is that a trophic interaction will depend on the match between consumer's need and resources' content of one or more elements that may be in limiting concentrations in their resource (Frost et al. 2006, Danger et al. 2013, Sterner and Elser 2017).
Macrodetritivores are defined as invertebrate animals larger than two mm which mainly contribute to decomposition by feeding on detritus which mainly came from plant (hereafter called litter), but also from animals (necrophagous and coprophagous detritivores) (Brussaard 1998, Marks 2019). They thus contribute to numerous associated ecosystem services (Mancinelli and Mulder 2015). Litter presents a high mechanical and nutritional heterogeneity. This heterogeneity partly depends on the plant identity. It is further enhanced by physical and microbial agents during decomposition, decreasing litter toughness, transforming chemical deterrents, and increasing nutrient concentration over time (Danger et al. 2012, David 2014, Brousseau et al. 2019). While nutritional constraints are well studied, mechanical constraints are largely under considered (Motomori et al. 2001, Clissold 2007, Evans-White and Halvorson 2017). Several experiments suggested that mechanical constraints could predominate over nutritional ones, but such experiments were not always designed to test it formally and if so, they focused on litter traits only (Graça and Cressa 2010, Walpola et al. 2011, Foucreau et al. 2013).
Under the assumption that nutritional constraints drive litter consumption by detritivores, we expect one or a few elements to be limiting consumption rates. Although all organisms are composed of approximately 25 elements, the vast majority of studies focused on C, N and P elements, with a focus on C and N (Sterner and Elser 2017). Yet, other macroelements have rarely been investigated in the context of detritivore consumption. For example, K and Ca are crucial common elements (metabolic cofactors and tegument components) (Peñuelas et al. 2019, Zhang et al. 2022). Alternatively, under the assumption that mechanical constraints limit food ingestion, consumption rate should be explained by traits reflecting mechanical constraints. Based on findings on herbivores, interesting assumptions about how mechanical constraints influence detritivores were tested by field approaches with interesting results (Brousseau et al. 2019, Raymond-Léonard et al. 2019) but they were never experimentally tested before. First, grabbing the limb of the leaf should depend on limb thickness and mandible maximal gape (Brousseau et al. 2019). Second, punching through the limb, cutting and shredding it should depend on the ability of mandibles to break leaf litter mechanical resistance. A detritivore should be able to punch through the limb if its biting force exceeds litter toughness (Ibanez et al. 2013, Brousseau et al. 2019, Raymond-Léonard et al. 2019). However, their effects on consumption have not been extensively tested. Scarce existing studies suspected litter toughness to impair consumption (Motomori et al. 2001, Patoine et al. 2017)
Previous literature based on litter traits solely indicates that mechanical constraints can overcome nutritional ones (Motomori et al. 2001, Graça and Cressa 2005, Patoine et al. 2017). Furthermore, detritivores are considered as being prone to feed on the most conditioned litter (which were subjected to physical and microbial processing agents), i.e. on the resource with highest quality and lowest constraints (Graça and Cressa 2005, Danger et al. 2012, Evans-White and Halvorson 2017). Yet, as conditioning increases litter quality, competition could lead to niche partitioning. For instance, some species would exploit litter at an earlier stage of decomposition (Cummins et al. 1989), or differently exploit litter (grazing from surface versus fragmenting the whole limb) (Danger et al. 2012). Under the assumption that detritivores are rather generalists and opportunistic consumers relying on a large variety of resources, we expect functional trait identity of detritivores and/or litter to separately mainly explain litter consumption by detritivores (e.g. litter toughness impacts all detritivores similarly). On the contrary, under the assumption that detritivores are diet-specialized, we expect that a match between detritivore and litter traits will mainly explain litter consumption (e.g. detritivores with strong mandibles will consume tough leaves) (Ibanez 2012, Brousseau et al. 2018, Neu et al. 2023).
Our study aimed to test for the relative importance of mechanical constraints over nutritional constraints in explaining litter consumption by detritivores. For this, we compared the consumption rates of a large variety of leaf litter by several detritivore taxa representative of aquatic and terrestrial ecosystems in laboratory no-choice consumption tests. Leaf litter was chosen to have a large range of mechanical and nutritional trait values. We investigated two mechanical constraints: grabbing and fragmenting the litter. For this, we measured suitable couples of mechanical traits for both litter and detritivores. We also investigated four nutritional constraints related to N, P, K and Ca contents in both detritivores and litter. Constraints relative to C content were not investigated as they are expected to be linked mainly to metabolic constraints. For each constraint, we also tested if trait matching drives a significant part of litter consumption by detritivores.
Material and methods
Experimental design
We selected eight leaf litter species with a large range of mechanical and nutritional trait values to cover a wide spectrum of potentially different constraints. Each leaf litter species was then either just leached, or leached and microbially conditioned before being offered to detritivores. This ensured enlarging the range of initial mechanical and nutritional litter traits values while limiting the content of chemical deterrents. We performed leaf litter consumption tests by offering one of 16 different leaf litter types to one of nine macrodetritivore taxa (four aquatic and five terrestrial species). Detritivore taxa were chosen to be representative of coarse lifeforms commonly encountered in aquatic and terrestrial ecosystems. As we aimed to assess fundamental rules of pairwise interactions between detritivores and litter, we tested the consumption of a single litter by a single detritivore individual at the time. We also offered litter species that detritivores could not encounter in their natural habitat. We thus sampled litter in a geographical site (Canal du Midi: Toulouse and Ramonville-Sainte-Agne, France) distant from the geographical sites where we sampled detritivores (Montagne Noire, France). We performed a total of 576 consumption tests, corresponding to 144 detritivore–litter pairs: 9 detritivores taxa × 16 litter treatments (8 leaf litter species × 2 litter conditioning treatments (leaching or leaching plus microbial conditioning)) × 4 replicates. Replicates spread out from week to week and were associated with corresponding control tests without detritivores (Supporting information). We performed a total of 128 control tests without detritivores: 16 litter treatments (8 litter species × 2 litter conditioning treatments (leaching or microbial conditioning)) × 2 ecosystem types (aquatic or terrestrial) × 4 replicates.
We performed consumption tests in microcosms made of clean plastic containers with 50 g of clean sand in the dark at 10°C. For aquatic detritivores, we added 200 ml of water collected from their stream. For terrestrial ones, we sprayed 5 ml of tap water on the sand. We starved detritivores for three days prior to the consumption test. In each microcosm we placed one individual (assigned randomly) and five discs of one litter treatment (litter species × conditioning treatment) that were previously freeze-dried, weighted, and rehydrated in tap water for 1 h. Tests were stopped when consumption visually reached 75% of initial discs surface or after three days. At the end of the consumption tests, remaining discs fragments larger than 1 mm were collected, freeze-dried and weighted. When one individual died during the first 24 h of the test, we immediately replaced it with a new individual. When the individual died later during the test, we repeated the consumption test the week after, with corresponding control treatments. At the end of consumption tests, detritivores were starved for 24 h, and were weighted (aquatic animals were gently blotted with paper towel). We converted fresh body mass into dry mass using a linear relationship established for each taxon. For each detritivore taxon, we conserved half of the individuals in 70% ethanol for dissection and we froze the other half for chemical analyses. All weight measurements were determined at the nearest 0.1 mg.
Leaf litter
We collected dead leaves at abscission from October to November along the Canal du Midi (Toulouse and Ramonville-Sainte-Agne, France) from a limited number of individuals (≤ 5) for each species. We used eight tree species belonging to eight different families, namely Ailanthus altissima (Simaroubacea), Robinia pseudoacacia (Fabacea), Juglans regia (Juglandacea), Carpinus betulus (Betulacea), Acer platanoides (Aceracea), Prunus avium (Rosacea), Quercus petrea (Fagacea), Platanus × hispanica (Platanacea). Leaves were air-dried and stored in the dark until being leached with tap water during 24 h. We then cut 1 cm diameter leaf discs with a cork-borer, avoiding main veins. Some leached discs were microbially conditioned. They were incubated in a mix of decomposing dead leaves until they were visually microbially colonized (soft and discolored). For aquatic conditioning specifically, we collected 50 litres of stream water and dead leaves from the same stream where we collected aquatic individuals. We added Fertiligène Naturen fertilizer (NPK: 3 – 2 – 5) at 0.5 ml l−1. We left it three days in a greenhouse with constant oxygenation for microorganisms' development. We then filled tanks with filtered (63 µm) water and placed one fine-mesh bag of monospecific litter discs of each litter species per tank with constant oxygenation. For terrestrial conditioning, we collected dead leaves from Montagne Noire forest soil (beech, chestnut, hazelnut) and grinded it with a garden shredder. We left it three days in a green house after humidification with the same fertilizer at 0.5 ml l−1. We then placed monospecific leaf discs in a fine-mesh bag of each litter species between two layers of fragmented litter in each tank. Tanks were regularly humidified with the same fertilizer.
To test for mechanical constraints of grabbing and fragmenting litter, we measured thickness and toughness on eight discs from controls, respectively. We measured limb thickness to the nearest 0.001 mm with a Helios-Preisser digital micrometer, avoiding main veins. We measured litter toughness as the penetration pressure (kPa) needed for a 2.2 mm diameter steel rod to penetrate through a leaf disc. We used a custom-made penetrometer, such as described in Graça et al. (2005), fitted to a digital force tester (CS225 Series, Chatillon) measuring force to the nearest 0.1 N. To test for nutritional constraints, we quantified nitrogen (N) content using a total nitrogen and organic carbon analyzer (TOC L, Shimadzu) on three replicates of 20 mg. We quantified phosphorus (P), potassium (K) and calcium (Ca) content with an induced coupled plasma – mass spectrometer (ICP-MS) on three replicates of 5 mg.
Macrodetritivores
We collected macrodetritivores in la Montagne Noire, a metamorphic forested massif east of Toulouse (France) (Supporting information). We hand-captured aquatic detritivores from February to March in a stream mainly boarded by beech (Fagus sp.), hazelnut (Corylus sp.), and chestnut (Castanea sp.) trees. We hand-captured terrestrial detritivores in April in a site dominated by ash (Fraxinus sp.). We identified all taxa to the lowest taxonomic level, mostly species. Only Plecoptera and Tipula larvae were identified to the genus, Nemoura and Tipula, respectively (Tachet et al. 2010). Case-bearing Trichoptera larvae were identified as Potamophylax cingulatus (Waringer and Graf 2011), and Amphipods as Gammarus fossarum (Tachet et al. 2010). Terrestrial detritivores were either Isopods (Philoscia affinis and Porcellio monticola) (Vandel 1962, Oliver and Meechan 1993) or Diplopods (Cylindroiulus londinensis, Polydesmus inconstans and Glomeris marginata ) (Blower 1985, David 1995). At the laboratory we sorted them by taxon, and left them 1–3 days in the dark at 10°C for acclimatation before being starved.
To assess metabolic constraints, we converted detritivore fresh body mass into dry mass using a linear relationship established for each taxon with a subset of individuals both fresh- and freeze-dry-weighted (n ≥ 30, p < 0.001, R2 ≥ 0.57). Other individuals were only fresh-weighted before being conserved in 70%-ethanol for dissection. Their dry mass was then inferred with the estimated linear relationship. To assess nutritional constraints, four pools of at least four individuals for each detritivore taxon (resulting in a total of four analytical replicates per taxon) were grinded into powder and analyzed following the same procedure as for litter samples. For Nemoura alone, only two pools were analyzed due to the very small body weight of this taxon.
Data analyses
We calculated individual rate of litter consumption (C in mg day−1) as follows: , where Mf and Mi are the final and initial mass (mg) of litter discs offered to detritivores, respectively. Mfc and Mic are the final and initial mass (mg) of corresponding litter discs in control conditions, respectively. ∆t is the test duration. To take into account the effect of detritivore body mass on individual consumption rate, we built a log–log linear model with all consumption rate values above 0.001 mg day−1. This threshold was based on visual analysis of the data and removed 91 quasi-null observations. The exponential coefficient allowed to compute mass-independent rate of litter consumption (Ci in mg mg−1−c day−1) as follows: Ci = C × B−c, where B is the individual body mass, and c is the estimate for the effect of body mass in the log–log linear model.
To asses which part of the variability in mass-independent consumption rate was explained by the origin ecosystem type (aquatic or terrestrial), detritivore taxon, litter species, and microbial conditioning treatment (aquatic, terrestrial, or no conditioning), we fitted a linear mixed model (lmer function from ‘lme4' package) (ln(Ci) ~ 1|Ecosystem/Detritivore taxon/Conditioning/Leaf species) following the procedure by Messier et al. (2010). As advised by Messier et al. we used bootstrap (n = 500) to precisely estimate the variability effects. To assess how much of data variability could be explained by trait-matching, we performed a linear model explaining mass-independent consumption rate by the interaction of detritivore taxa and litter treatment (ln(Ci) ~ Detritivore taxon × Litter treatment). Due to the nature of the experimental design, we wanted to to test if replicates had an influence on consumption rate. We then performed a mixed model explaining ln(Ci) with ‘Replicate week number' as a fixed factor (Supporting information), and the identity of each detritivore taxon – litter treatment pair as a random factor (ln(Ci) ~ Replicate week number + (1|Detritivore taxon – Litter treatment pair)).
In order to illustrate mechanical and nutritional trait profiles variations of litter and detritivores used in the experiment, we performed two PCA on leaf litter traits and detritivores traits separately. We selected the best model explaining mass-independent consumption rate through a two-steps process. First, we identified the best model accounting for mechanical constraints, and the best model accounting for elemental constraints. Second, we fused these two models and applied a step-wise selection to remove non-significant variables. For the first step, we tested 1) which trait-based constraints best explain mass-independent consumption rate and 2) for each constraint, whether trait-matching can help better explain mass-independent consumption rate. Rohr et al. (2016) suggested the use of matching-centrality models to investigate the influence of single traits and of a match between consumer and resources traits on trophic interactions. As we expected non-linear relationships between mass-independent consumption rate and traits, we only used generalized additive models (GAM) with the gam function (Rohr et al. 2016, Brousseau et al. 2018). We used the lowest smoothing parameter (k = 3) to avoid over-fitting (Crawley 2012, Brousseau et al. 2018). We computed matching terms of the form (X − Y)2, where X and Y are standardized (scaled) detritivore and litter traits, respectively (Rohr et al. 2016, Brousseau et al. 2018). We built six different pairs of models (Table 2): two of them 1) are related to mechanical constraints (grabbing and fragmenting the litter by detritivores) and four of them 2) are related to nutritional constraints (N, P, K and Ca contents in both detritivores and litter). Within each pair of models, a model with corresponding detritivore trait and litter trait, and a ‘matching term' reflecting the match between those two traits [gam(ln(Ci) ~ s(X, k = 3) + s(Y, k = 3) + s((X − Y)2, k = 3))], was compared to a reference model with both traits but without the matching term [gam(ln(Ci) ~ s(X, k = 3) + s(Y, k = 3))]. Log transformations were applied when necessary. We considered a matching term to have a significant influence if it was significant in the first model and if the first model was significantly better (assessed with an ANOVA and by comparing the Akaike information criterion (AIC) values) than the reference model. For the second step, we aimed at identifying the traits and matching terms of all of the five constraints which best explain consumption rate. For this, we built a final model (III) by combining the best model (I) reflecting mechanical constraints with the best model (II) reflecting nutritional constraints. We performed a step-wise selection to remove non-significant variables. To assess the relative importance of each variable in a model, we compared the deviance explained by the model with the deviance explained by the same model from without the variable (Crawley 2012). Residuals distribution of all models was visually checked. All analyses were performed with R ver. 4.0.3 (www.r-project.org).
| Constraints | Traits used in the model | p-value of the matching term | Deviance explained with matching term | Deviance explained without matching term | Difference between the two models | |
|---|---|---|---|---|---|---|
| Mechanical | punching through and shredding the limb | litter toughness – biting force | F1 = 1.5 | 31.0% | 30.8% | F−2.3 = 1.4 |
| p = 0.22 | p = 0.23 | |||||
| ∆AIC = 0.5 | ||||||
| grabbing the limb | litter thickness – mandible gape | F1 < 0.1 | 20.9% | 20.9% | F−1 = 0 | |
| p = 0.98 | p = 1 | |||||
| ∆AIC = 2.0 | ||||||
| Nutritional | P limitation | litter P – detritivore P | F1 = 10.6 | 24.9% | 23.5% | F−1.8 = 5.6 |
| p = 0.001 | p = 0.005 | |||||
| ∆AIC = − 7.2 | ||||||
| N limitation | litter N – detritivore N | F1 = 1.2 | 15.1% | 14.9% | F−1 = 1.4 | |
| p = 0.28 | p = 0.24 | |||||
| ∆AIC = 0.6 | ||||||
| K limitation | litter K – detritivore K | F1 = 2.4 | 13.4% | 13.4% | F−0.2 = 0 | |
| p = 0.18 | p = 1 | |||||
| ∆AIC = 0.9 | ||||||
| Ca limitation | litter Ca – detritivore Ca | F1 = 0.2 | 7.0% | 7.0% | F−1 = 0.25 | |
| p = 0.62 | p = 0.62 | |||||
| ∆AIC = 1.8 | ||||||
Results
Leaf litter
Litter toughness varied over a 17-magnitude range across the whole dataset (from 38 kPa for Ailanthus with aquatic conditioning, to 656 kPa for Quercus without conditioning, Supporting information). Conditioning decreased litter toughness by 42 ± 4% (mean ± SE value). It only slightly affected litter thickness, decreasing it by 9 ± 4%. Conditioning increased N, P, K and Ca content by 37 ± 7%, 29 ± 8%, 24 ± 18% and 14 ± 5%, respectively. Litter mean consumption rate varied from 0.07 (Platanus without conditioning) to 2.96 mg day−1 (Ailanthus with terrestrial conditioning). Conditioning considerably increased consumption rate (by 250 ± 85%). The first two axis of PCA on litter traits (Fig. 1a) represented 76.4% of variability. Litter treatments are quite well scattered along the first axis (49.7% of variability) which can be interpreted as an axis of ‘litter quality' with more recalcitrant litter (typically Platanus and Quercus) on the left, and more palatable litter Ailanthus on the right. The conditioned litter is always more on the right-hand side than the unconditioned litter from the same species (Juglans stands as an exception, being more influence by the second axis than by the first one). A-posteriori mass-independent consumption rate vector is closed to the first axis, pointing to the right, consistently with the above-mentioned interpretation of ‘litter quality'.
Principal component analysis of (a) litter traits, and (b) detritivore traits. On the left (A) panel, litter treatments are indicated by the first letters of the litter genus and species, with ‘(A)' if the litter was conditioned in aquatic conditions, ‘(T)' if the litter was conditioned in terrestrial conditions or without additional letter if the litter is just leached. On each panel, mass-independent consumption (Ci) was added a-posteriori as a supplementary variable (dotted blue arrow).
Macrodetritivores
Mortality was low over the entire experiment (< 3%). Mean dry body mass ranged from 1.0 Nemoura to 156.7 mg Cylindoiulus (Supporting information). Biting force (F index) varied from 0.50 for Nemoura, to 2.64 for Glomeris. N, P and K content varied from 4.32, 0.86 and 0.28% (for Cylindroiulus, Nemoura and Cylindoiulus) to 10.80, 2.10 and 2.14% (for Nemoura, Polydesmus and Potamophylax), respectively. Insect larvae exhibited very low Ca content (from 0.21% for Nemoura, to 0.32% for Tipula) compared to crustaceans and millipedes (from 8.74% for Philoscia, to 17.78% for Polydesmus). Considering ratios for N, K and P elements of all 144 detritivore taxon – litter treatment couples, P was the element with the highest ratio for 138 detritivore – litter couples. K and N only displayed the highest ratio on five (for Tipula and Potamophylax on Ailanthus and Ailanthus with aquatic conditioning, and for Tipula on Robinia without conditioning) and one (Nemoura on Juglans without conditioning) occasions, respectively. Mean P ratio was 28.3 ± 0.6 over the entire dataset, meaning that P content was on average 28 times higher in detritivores than in the litter. The highest ratio value (80.7) was for the P element for Polydesmus with Platanus without conditioning. The first two axis of PCA on detritivore traits (Fig. 1b) represented 83.0% of variability. Detritivore taxa are roughly grouped by habitat, body size, sclerotization and phylogenetic proximity: insect larvae and aquatic taxon are on the left part of the panel, with high N and K content. Crustaceans are close to the second axis. Two millipedes, Glomeris and Cylindroiulus, are on the right part of the panel (with high body mass and biting force), while Polydesmus shows different characteristics.
Sources of variation of litter consumption rate
Mean consumption rate of detritivore taxa varied from 0.15 Nemoura to 1.44 mg day−1 Potamophylax. Mean consumption rate divided by detritivore dry body mass varied from 0.004 Glomeris to 0.291 mg mg−1 day−1 Potamophylax, meaning that Potamophylax daily consumed a mean quantity of litter representing 29% of its body mass (Supporting information). Maximum consumption rate per detritivore body mass reached 1.37 mg mg−1 day−1, for a Potamophylax individual on Ailanthus with aquatic conditioning.
A log–log linear model showed that detritivore body mass weakly but significantly influenced individual consumption rate (n = 485, R2 = 0.02, F1,482 = 7.8, p = 0.005). We then used the estimated coefficient (c = 0.11) to correct for the body mass effect and compute mass-independent consumption rate (Ci).
In a model explaining ln(Ci) with Ecosystem/Detritivore taxon/Conditioning treatment/Leaf species as random nested effects, the ecosystem (aquatic or terrestrial), detritivore taxon, conditioning treatment (aquatic, terrestrial, or just leached), and leaf species explained < 0.0 ± 0.0, 11.6 ± 3.4, 22.3 ± 4.0 and 34.3 ± 3.0% of variability, respectively (Table 1A).
| (A) lmer( ln(Ci) ~ 1 | Ecosystem / Detritivore taxon / Conditioning / Leaf species ) | |||||
|---|---|---|---|---|---|
| Ecosystem | Detritivore taxon | Conditioning treatment | Leaf species | Residuals | |
| Variability explained (%, mean ± SE) | < 0.0 ± 0.0 | 11.6 ± 3.4 | 22.3 ± 4.0 | 34.3 ± 3.0 | 31.8 ± 2.7 |
| (B) lm(ln(Ci) ~ Detritivore taxon × Litter treatment ) | |||||
| Detritivore taxon | Litter treatment | Detritivore taxon × Litter treatment | Residuals | ||
| Variability explained (%) | 23.5 | 32.9 | 11.1 | 32.5 | |
| F and p values | F8 = 39.0, p < 0.001 | F23 = 18.9, p < 0.001 | F112 = 1.3, p = 0.028 | ||
In a model explaining ln(Ci) with Detritivore taxon × Litter treatment, both parameters and the interaction term were significant (Table 1B). Detritivore taxon, litter treatment, and the interaction term accounted for 23.5, 32.9 and 11.1% of the sum of squares, respectively.
In a mixed model with ‘Replicate week number' as a fixed factor, and the identity of each detritivore taxon – litter treatment pair as a random factor, replicate weeks were significantly influencing ln(Ci) (Anova: chisq10 = 24.1, p = 0.007). Yet, they accounted for only 1.8% of the total sum of squares.
Trait-based models explaining litter consumption rate
Our analysis revealed that both mechanical and nutritional constraints are influencing consumption rate (Table 2). Concerning mechanical constraints, the best model explaining consumption rate was the model based on litter toughness and detritivore biting force (Deviance explained = 30.8%). For this model, the matching term did not significantly increase the performance of the model. Consumption rate first decreased with litter toughness (∆ Deviance explained = 20.6%), the decrease being stronger at low toughness values (below 400 kPa), than at high toughness values (above 400 kPa) where consumption reaches a minimal plateau (Fig. 2a, Supporting information). Biting force index F influenced consumption rate with a humped-back relationship, consumption being maximal for F values around 1 (∆ Deviance explained = 10.2%).
Visual representation of the best generalized additive models (GAMs) accounting for (a) mechanical constraints, and (b) elemental constraints. Squares represent the mean mass-independent consumption rate (Ci) value (n = 4) at each point. On the left panel (a), consumption rate decreases strongly with litter toughness until 400 kPa. Above this toughness value, consumption rate is very limited and depends almost exclusively on biting force. Biting force influences consumption rate according to a humped-back shape relationship with consumption rate being maximal for biting force values around 1 (ln(1) = 0). The model explained 20.9% of deviance. On the right panel (b), consumption rate mostly depends on litter P according to a positive relationship. Detritivore P content influences consumption rate according to a humped-back shape relationship with consumption rate being maximal for P content values around 1.3% (ln(1.3) = 0.26). P-matching term effect is not represented but is visible by the non-symmetrical effects of litter and detritivore P contents. The model explained 24.9% of deviance. See also the Supporting information.
Concerning nutritional constraints, the best model was based on elemental P constraints (Table 2, Deviance explained = 24.9%). For this model, the matching term between P content in detritivores and in litter significantly increased the performance of the model (∆ Deviance explained = 1.4%). Consumption rate first increased, almost linearly, with litter P content (∆ Deviance explained = 19.6%) (Fig. 2b, Supporting information). Detritivore P content influenced consumption rate according to a humped-back relationship (∆ Deviance explained = 2.0%), consumption being maximal for detritivore P content values around 1.3%. Lastly, the matching term negatively influenced consumption rate, meaning that consumption rate decreases when the gap between detritivore and litter P content increases (∆ Deviance explained = 1.4%). To assess the mechanisms behind the influence of the P match on the model, we extracted predicted values of ln (Ci) for detritivore taxa with highest (2.10%) and lowest P (0.82%) contents (Supporting information). We then fitted one linear model to each prediction. Predicted consumption rate of detritivores with the highest P content was positively influenced by litter P with a slope coefficient of 2.22 ± 0.05 (mean ± SE) (F1,18 = 2038, p < 0.001, R2 = 0.99). On the other hand, predicted consumption rate of detritivores with the lowest P content was positively influenced by litter P with a slope coefficient of 1.42 ± 0.06 (F1,18 = 548, p < 0.001, R2 = 0.97). Only the model accounting for Ca constraints presented potential concerning residuals distribution due to bimodal distribution of detritivore Ca content. Although not presented we also performed models accounting for putative C constraints which have proven to be very uninformative (Deviance explained < 4%, data not shown).
For each constraint type, either mechanical or nutritional, the best above-mentioned model outperformed the others by a substantial margin (Table 2, gap in deviance explained of 10.1% and 9.8%, for mechanical and nutritional constraints respectively).
The final synthetic model explained mass-independent consumption rate with litter toughness, biting force, litter P parameters, and the P matching term (Fig. 3, Supporting information). Detritivore P parameter was removed because it did not significantly improve the model. The model explained 36.6% of deviance. Consumption rate was first influenced by biting force (∆ Deviance explained = 8.4%) according to a similar humped-back shape as in the model accounting for mechanical constraints. Litter toughness was the second explaining parameter (∆ Deviance explained = 5.6%) with the same negative relationship as in the best model accounting for mechanical constraints. Consumption rate then linearly increased with litter P content (∆ Deviance explained = 4.2%). Lastly, the P matching term decreased consumption rate (∆ Deviance explained = 1.4%), as in the best model accounting for nutritional constraints.
Final generalized additive model accounting for both mechanical and nutritional constraints. The influence of the two first parameters (biting force and litter toughness) on mass-independent consumption rate (Ci) are presented. Squares represent the mean consumption rate value (n = 4) at each point. Consumption rate decreases strongly with litter toughness. Biting force influences consumption rate according to a humped-back shape relationship with consumption rate being maximal for biting force values around 1. The differences in predictions with Fig. 1A are due to nutritional constraints. Non-represented variables are litter P content and the P matching term. The model explained 36.6% of deviance. See also the Supporting information.
Discussion
In this study, after taking into account the influence of body mass, consumption rate was best explained by mechanical constraints, especially by litter toughness and detritivore biting force. Trait-matching had a very low explanatory power. Only the P–P match significantly improved the model based on P nutritional constraints. Overall, the final model explained up to one third of the variability in the data.
In the final model accounting for both mechanical and nutritional constraints, mechanical traits explained a higher portion of deviance compared to nutritional traits. This indicates that mechanical constraints dominated over nutritional constraints in our experiment. This is in line with the consideration that consumers first have to be able to process and ingest the resource before they can assimilate it and feed again in order to match their nutritional requirements (Wootton et al. 2023). If mechanical constraints dominate over nutritional constraints, we can expect consumption rates of detritivores to be limited by their ability or by the time they need to handle and process their resource. On the contrary, if nutritional constraints dominate over mechanical constraints, we can expect detritivores to apply a strategy of compensatory feeding in increasing consumption rate to compensate for a low amount of nutrients in the resource (Ott et al. 2012, Danger et al. 2013). We did not observe compensatory feeding, thus reinforcing the interpretation of a predominance of mechanical constraints. Those results then support the idea that mechanical constraints drive litter consumption, as hypothesized by other studies (Motomori et al. 2001, Danger et al. 2012, Foucreau et al. 2013, Patoine et al. 2017). Among investigated nutrients, P content was the first parameter explaining consumption rate before N, K and Ca contents. The match between detritivore and litter P content was the only matching term significantly improving the explanation of consumption rate. The weak influence of matching terms for both mechanical and nutritional constraints makes sense as the interaction between detritivore taxa and litter treatments explained a low amount of consumption rate variability. Accordingly, as litter treatment explained a higher amount of variability than detritivore taxa, litter traits were expected to have a stronger influence than detritivore traits in the models which is consistent with our results.
As in other consumption experiments using both aquatic and terrestrial detritivores, the ecosystem parameter (aquatic versus terrestrial) explained a very low amount of data variability (Rota et al. 2022). Observed patterns can then be considered as being relatively generic across detritivores coming from these two ecosystems. Detritivore body mass had a surprisingly weak effect. The power coefficient of 0.11 that fitted our data strongly deviates from the value of 0.75 that is expected under the metabolic theory of ecology (Brown et al. 2004). A lower coefficient than 0.75 is expected if other traits than body mass drive consumption, thus decreasing the importance of metabolism and body mass effect (Pawar et al. 2012, Rall et al. 2012). As detritivores were starved prior to our consumption tests, the choice of not consuming litter would result in a high energetic imbalance. However, our experiment exposed detritivore taxa to resources with diverse mechanical and nutritional constraints that could limit consumption. This would explain in part the low influence of body mass. Finally, as the distribution of replicates over time had a very weak effect on mass-independent consumption rate (1.8% of total sum of squares), we are confident that it did not change our interpretations.
Litter consumption rate decreased with litter toughness until it reached a horizontal asymptote for toughness values greater than 400 kPa (Fig. 2a, Supporting information). This pattern, identified in the first model accounting for mechanical constraints, was found to be very similar in the final model summarizing mechanical and nutritional constraints (Supporting information). A non-linear influence of resource toughness on consumption rate is expected under the hypothesis that consumers need to break through the resource item to process it, increasing the time they need to handle the resource (Clissold 2007, Ott et al. 2012). When resource toughness is higher than the critical biting force of consumers, we can expect them to be unable to process food. Consumption should then be minimal. Our data then suggest a critical biting pressure around 400 kPa for the detritivore taxa we tested. This value is of the same order of magnitude as what is found in the few previous studies that investigated ability of detritivores to overcome resource toughness (Danger et al. 2012). The influence of biting force on litter consumption followed an unexpected humped-back shape (Fig. 2a, Supporting information). We rather expected a monotonically increasing relationship as consumers with stronger mandibles should spend less time and efforts handling and processing the food, which should allow them to consume larger quantities in the same amount of time (Ott et al. 2012). This humped-back relationship between consumption rate and the index for biting force could be explained by a tradeoff between biting force and other traits limiting consumption. We also expected that a match between litter toughness and biting force would drive consumption rate. It was not the case, which highlights the need to know more about how detritivores mechanically process they food. For example, Potamophylax was the taxon with the highest consumption rates, even for the toughest and most recalcitrant litter treatments (Supporting information). Because of their habit of cutting discs from leaves to make their cast (Waringer and Graf 2011), they can be expected to have very strong mandible capacities. Yet, they displayed average biting force index values. Other traits such as mandibles sharpness should then be considered as it may also contribute to explain why detritivores cut more or less efficiently through tough leaves (Clissold 2007).
The predominance of P content over other elements is expected because P is often the most limiting nutrient for a large variety of organisms (Sterner and Elser 2017). A few studies highlighted the importance of P for detritivores (Danger et al. 2013). In almost all detritivore – litter pairs of our study, P was the element for which content in detritivores was the highest, relatively to litter content. This would indicate that P is most likely the most frequently limiting element. Another argument comes from the study by Frost et al (2006) who computed the C:P threshold elemental ratios (TER) by atoms for many aquatic organisms (i.e. the resource C:P ratio at which consumer growth limitation switches from one element to another). They found that detritivore organisms typically have TERC:P around 1000. As our litter had C:P ratios of 2424 ± 188 ([min − max] = [1014–4791]), we are confident that P was a limiting element for detritivores. Consumption is expected to depend on detritivores and litter P proper contents but also possibly on their match. The P match only slightly contributed to consumption in our data. Yet, compared to the model without P matching term, adding the P match led to model predictions that are in line with ecological stoichiometry theory predictions (Sterner and Elser 2017). This theory predicts that consumers with high elemental requirements will be more impacted by resources of low element content, compared to other consumers with modest requirements (Ohta et al. 2016). Accordingly, in the best model accounting for nutritional constraints, the predicted consumption rate of detritivores with high P content (i.e. probably having highest P requirements) is more negatively impacted by low litter P content, than for detritivores with low P content (i.e. probably having lowest P requirements). Even though we demonstrated the existence of a P match explaining trophic interactions, the underlying mechanisms remain unclear. It seems unlikely that detritivores can taste the amount of P in the resource and adapt their consumption in response. Detritus P content is known to increase with microbial decomposition and with microorganism's biomass (Danger et al. 2012). Detritivores are able to detect the presence of microorganisms (Motyka et al. 1985, Graça and Cressa 2005) and consumption rate of detritivores is known to increase in presence of microorganisms (Zimmer et al. 2003, Swan and Palmer 2006). It would make sense to postulate that the direct influence of microorganism's presence on consumption rate would explain an indirect influence of resource P content. The weakest performance of models based on N content is surprising considering the major importance attributed to litter N content in the literature (García-Palacios et al. 2013, Frouz et al. 2015, Zhang et al. 2015). Yet, as N is often correlated to P (Li et al. 2021) and as N is more systematically investigated than P, this could blur the importance of P over N in past studies. Results about nutritional constraints and their relative importance with mechanical constraints should nonetheless be interpreted with caution. Elemental content we measured in our experiment did not reflect biological needs for detritivores or bioavailability for litter. Acute measures of elemental needs (e.g. N content in proteins vs N content in nucleic acids) or of litter elements' bioavailability (e.g. N content in tannins vs soluble N) may change some of the observed patterns and may improve our understanding of nutritional constraints.
As a conclusion, we contribute to assess fundamental pairwise trophic rules of interactions by using single detritivore – litter pairs. To be completed, other steps of the interaction have to be investigated (e.g. chemical deterrents acting on selection or digestion) to be fully understood and predicted. Also, our no-choice experimental design might underestimate the importance of trait-matching that might arise in realised trophic niches when consumers have the choice to select, or complement their food. Furthermore, the low effect of trait-matching on consumption is in line with the assumption that detritivores are generalists and opportunistic feeders, relying on various resources (Crenier et al. 2017, Rubio-Ríos et al. 2023). Contrarily to consumption of dead resources, other trophic interactions (herbivores – green plants, pollinators – flowers, predators – preys) are expected to drive co-evolution between the two actors of the interaction who have to be adapted to their interacting organism (Ibanez 2012, Ibanez et al. 2013, Brousseau et al. 2018, Neu et al. 2023). From an ecosystem perspective, these fundamental pairwise rules are expected to be modulated by indirect interactions with other biotic trophic network components (e.g. predation, competition, facilitation) or by abiotic parameters fluctuations.
Acknowledgements
– We thank Clément Castille, Florian Chapeau, Mathilde Joffre, Loubna El Madouri, Estelle Ribaut and Valentina Soto for their help on the field. We also thank Corinne Pautot for her help in the laboratory, and Christophe Laplanche for his advice on statistical analyses. We thank the Fast Bio-analyses of Trace Elements platform (FBil, Univ. Toulouse UPS, INPT, CNRS), David Baqué, Frédéric Candaudap, Mathis Flamant, Sophia Hansson, Laetitia Leroy, and Gaël Le Roux for their help and advice on ICP analyses. We thank the LEFE Plateform for physical and chemical analyses (PAPC, Univ. Toulouse UPS, INPT, CNRS) and Frédéric Julien. Finally, we warmly thank Rebecca Oester for her advice, corrections and comments on the manuscript.
Funding
– This experiment was funded by the ELEMENTARY projectFrench national program EC2CO, Ecosphère Continentale et Côtière)
Author contributions
Théo Marchand: Conceptualization (equal); Data curation (lead), Formal analysis (lead); Investigation (equal); Methodology (equal); Project administration (lead); Supervision (supporting); Validation (equal); Visualization (lead); Writing – original draft (lead); Writing – review and editing (equal). Lola Estabes: Data curation (supporting); Investigation (equal); Validation (equal); Visualization (supporting); Writing – review and editing (supporting). Benjamin Pey: Conceptualization (equal); Formal analysis (supporting); Funding acquisition (lead); Methodology (equal); Project administration (supporting); Resources (lead); Supervision (lead); Validation (equal); Writing – review and editing (equal).
Open Research
Data availability statement
Data are available from the Dryad Digital Repository: https://doi.org/doi:10.5061/dryad.6m905qg6z (Marchand et al. 2024).
