Facilitation costs and benefits function simultaneously on stress gradients for animals

Abstract

Understanding the variation in species interactions along environmental stress gradients is crucial for making robust ecological predictions about community responses to changing environmental conditions. The facilitation–competition framework has provided a strong basis for predictions (e.g. the stress-gradient hypothesis, SGH), yet the mechanisms behind patterns in animal interactions on stress gradients are poorly explored in particular for mobile animals. Here, we proposed a conceptual framework modelling changes in facilitation costs and benefits along stress gradients and experimentally tested this framework by measuring fitness outcomes of benefactor–beneficiary interactions across resource quality levels. Three arthropod consumer models from a broad array of environmental conditions were used including aquatic detritivores, potato moths and rainforest carrion beetles. We detected a shift to more positive interactions at increasing levels of stress thereby supporting the application of the SGH to mobile animals. While most benefactors paid no significant cost of facilitation, an increase in potato moth beneficiary's growth at high resource stress triggered costs for benefactors. This study is the first to experimentally show that both costs and benefits function simultaneously on stress gradients for animals. The proposed conceptual framework could guide future studies examining species interaction outcomes for both animals and plants in an increasingly stressed world.

1. Introduction

Positive interactions among species are an important focus of current ecological research as they are ‘ubiquitous drivers of biodiversity’ [1], provide a potent mechanism for multispecies coexistence [2], and mediate the impact of environmental changes on natural communities [3,4]. The stress-gradient hypothesis (SGH) postulates an increase in the frequency of positive species interactions with increasing environmental stress [5]. If one species ameliorates environmental conditions under stressed conditions, the frequency of facilitative interactions would increase and offset competition thus leading to net positive interactions. While evidence of net facilitative interactions has been largely documented in a wide range of plant communities [3] and for some animal taxa [6–8], we also need to examine the relative importance of the bi-directional nature of interactions, i.e. not only the effects of benefactors on beneficiaries but also the impacts of associated beneficiaries on benefactors [9–11]. Bi-directional interactions are likely important factors for many of the drivers of biodiversity and certainly not exclusive to plants. Consequently, there is a critical opportunity to advance the understanding of interactions, conceptually and empirically, by examining reciprocity on gradients and by testing animals in addition to plants.

If bi-directionality is included in the study of facilitation, structuring interactions into pairs for simplicity and formalizing costs provides a framework for theory and experiments. Three different modes of facilitation can thus be distinguished, based on the costs (−) and benefits (+) for the benefactors/beneficiary species pairs [12]: mutualistic (+/+), antagonistic (−/+) and commensal (0/+). The mutualistic mode has received attention elsewhere [1,13], but a main issue in the mechanistic understanding of facilitation is whether there is a cost for the benefactor species and particularly whether this cost changes on a stress gradient [14]. While facilitative interactions can strongly influence both performance and fitness of a benefactor species, it is the reciprocal nature of these interactions that can be of overriding importance for community dynamics because of the capacity for benefactor species to function as foundation and sometimes keystone in many ecosystems [10].

A system that shows a classical competition–facilitation shift along a stress gradient for a beneficiary species [15,16] is ideal for exploring the relevance of this extension as a merger between interaction theory (i.e. facilitation) and coevolutionary concepts. We hypothesize that fitness costs for benefactors in the presence of beneficiaries (i) depend on whether benefactor species experience net competition at low stress and (ii) can either be relatively constant or increase or at higher stress. We, therefore, predicted four different cases (figure 1a–d). The bi-directional interactions between benefactors and beneficiaries can be a form of commensalism (figure 1a), can shift from commensalism to parasitism at increasing stress (figure 1c) or can lead to parasitism if costs are significant to the benefactor (figure 1b, consistent costs, and figure 1d, increasing costs at higher stress), i.e. the benefactor does not tolerate the beneficiary. Defining the facilitation mode that functions along stress gradients will have important eco-evolutionary implications because parasitic interactions would select for avoidance or disassociation by benefactor species while commensal interactions would likely promote association between benefactors and beneficiaries [11]. This implies that we cannot ignore the cost of facilitation to the benefactor species because it can be the fundamental component that determines evolutionary stability.

Figure 1.

Four graphical models of predicted costs for benefactor species interacting with a beneficiary species along environmental stress. For simplicity, we consider a system that shows a classical competition–facilitation shift along a stress gradient for a beneficiary species. The bi-directional interactions between benefactors and beneficiaries are either a form of commensalism (a), shift from commensalism to parasitism at increasing stress (c) or parasitism (b and d). Note that because we had neither mechanistic hypothesis nor empirical evidence of benefits for the benefactor in presence of the beneficiary, we did not consider this case in the conceptual models.

Here, we proximately tested this predictive framework using three invertebrate models living in a broad array of environmental conditions; potato tuber moths, stream detritivores and rainforest carrion beetles. The net facilitation was previously established as a quantifiable and emergent outcome of behavioural mechanisms by which consumers facilitated resource access to other consumers (figure 2 and legend for details, [6,7,19]). Because positive interactions among animals can be deferred in space or time [4], and thus might be more difficult to detect than among plants, our study models dealt with animals whose interactions were constrained to patch resources (potato heaps, leaf packs and carrion pads). We modelled a stress gradient using food quality stress thereby assuming that consumers cannot realize maximum growth on a low-quality diet (i.e. sensu Grime's definition [20]). Using these models, we experimentally tested the following predictions: (i) sensu a strict SGH interpretation, positive interactions increase with environmental stress, (ii) facilitation will influence the fitness (survival, growth) of not only the beneficiary species but also the benefactors, and (iii) resource limitation is an effective form of stress gradient for animals that influences the net outcomes of interspecific interactions. This predictive framework assesses whether the SGH can be effectively used to describe animal interactions and establish the relevance of bi-directionality to the theory of facilitation as a whole.

Figure 2.

Facilitation mechanisms involved in interspecific interactions for the three studied consumer models. (a) The trichopteran shredder larvae Anomalocosmoecus feed by cutting large leaves through the middle, thereby increasing the number of leaf pieces and the total leaf perimeter available to amphipod Hyallela that feed preferentially on edges [17]. While competition is the rule between these two species fed with high-quality litter, more positive interactions occur for low-quality litter [6]. (b) The larvae of the specialist potato tuber moth feeder Tecia solanivora are highly efficient burrowers and facilitate tuber access to larvae of the generalist potato moth, Symmetrischema tangolias. Symmetrischema tangolias benefits from the presence of T. solanivora when feeding on tubers with a thick epidermis but not when feeding on tubers with a thin epidermis [7]. (c) Both Canthon aequinoctialis and Onthophagus haematopus are abundant diurnal carrion-eater beetles in the western Amazon [18]. While large C. aequinoctialis individuals detach pieces of carrion material from the surface of the pile, the smaller O. haematopus individuals enter the pile, preferably at a site previously disturbed by another beetle, and then dig tunnels below the carrion.

2. Material and methods

(a). Model study systems

Beyond their relevance for testing the hypotheses (figure 2), our three study pairs of species allow testing the SGH in a wide range of ecological conditions in the relatively poorly studied tropical realm [21]. The aquatic detritivores (Hyalella sp., Crustacea, Hyalleliae and Anomalocosmoecus sp., Trichoptera, Limnephilidae) are widespread in lotic environments of the tropical Andes and belong to two taxa (Amphipods and Limnephilids) that have a well-recognized role in litter decomposition, a key process in aquatic ecosystems [22]. Potato moths (Tecia solanivora and Symetrischemma tangolias, Lepidoptera, Gelechiidae) are widespread in potato storages of high elevation (2800–3400 m) agricultural landscapes in the Northern Andes, where they represent one of the most damaging crop pest complexes [23]. The beetles (Canthon aequinoctialis and Onthophagus haematopus, Coleoptera, Scarabaeidae) are among the most abundant species of the dung and carrion beetle community in the highly diverse Amazonian rainforest of Ecuador, accounting for about 40% of total abundance (see appendix 1 in [18]).

(b). Resource quality×species mixture experiments

To test our predictions, we established protocols in which we manipulated both resource quality (from low to high food palatability) and species mixtures (from species in isolation to pairwise mixture). Those protocols are detailed for stream invertebrates in [6] and potato moth in [7]. The protocol for carrion beetles is detailed in the electronic supplementary material, appendix S1.

Briefly, we first established a resource quality gradient by using (i) leaves from five different plant species with different palatability, (ii) tubers from five potato varieties differing in thickness of tuber epidermis and (iii) tuna fish pads showing a gradient of four different freshness conditions (see electronic supplementary material, table S1). We assumed that while both benefactor and beneficiary species would easily consume high-quality resources, the feeding activity of the benefactor species might be necessary to permit beneficiaries feeding on recalcitrant resources (figure 2). The selected gradient of food resources naturally occurs in the environmental conditions in which the species live, and thus the gradient is representative of the extent of limitations that these species experience.

After collection in the field, we placed both benefactor and beneficiary species in experimental containers and measured whether their performance varied with resource stress. Three levels of the species mixture factor were set up including each species in isolation and one pairwise mixture. We used a replacement density approach [24] to control for changes in abundance by fixing total density in each chamber to correspond to ranges detected in natural systems. Thus, the chambers with species mixtures were comprised of half the densities per species used in the monoculture/single-species chambers. For each resource stress gradient, replicates were physically arranged in 5–10 blocks including each of the three treatments to further examine the outcome of species interactions (see Data analysis). The sample sizes for each experimental design varied from 180 individual consumers per species to 1500 depending on the specific system (see electronic supplementary material, table S1).

At the end of the experiments (long enough to detect changes in consumer fitness), we measured the survival and individuals' growth for each of the ‘resource × species’ treatments. For aquatic detritivores and carrion beetles, growth was measured as the difference in lipid biomass between individuals collected at the beginning and end of the experiment (see electronic supplementary material, appendix S1). Moth growth was measured as pupa biomass at the end of the experiment because potatoes were inoculated with moth eggs.

(c). Data analysis

To quantify the outcome of the interactions between benefactors and beneficiaries along the food resource gradient, we used the relative interaction index or RII [25] for survival and growth data at all treatment levels. RII ranges from pure competition (−1) to pure facilitation (+1) and was calculated as follows:

2.1

where P_+N and P_−N represent the performance of the beneficiary (or benefactor) in the presence and absence of benefactor (or beneficiary), respectively. We assumed that the interaction had a cost for the benefactor when RII was significantly negative. We generated stress–species interaction curves in which we calculated values of relative resource quality stress by dividing the value of each resource quality treatment (e.g. leaf palatability, epidermis thickness and tuna pad dryness) by the value of the most palatable food treatment (see [6,7] for further details). For example, the five potato periderm thickness values of 0.02, 0.06, 0.12, 0.16 and 0.2 mm corresponded to relative resource stress of 10, 29, 55, 78 and 100%, respectively.

To assess the effect of the different resource stress treatments on single and interacting species, mixed-effect linear models were applied to survival rates, growth and RII, using the lme4 package [26] in R v. 3.4.3 [27]. Species identity (benefactor/beneficiary), species treatment (single or two-species) and resource stress level were treated as factors while block was modelled as a random effect. We used the lmerTest package with the Satterthwaite approximation for degrees of freedom to generate F and p-values. We also applied post hoc contrasts using the emmeans package to differentiate between the different treatments and identify potential interaction type shifts along the stress gradients (e.g. a shift from commensalism to parasitism, figure 1c).

3. Results

In the three study models, the experimental resource quality gradient had a significant effect on both survival and growth of the beneficiary consumers alone (except for O. haematopus survival), but it did it not significantly influence the benefactor species in single-species treatments (electronic supplementary material, table S2 and S3). Beneficiaries had 16–44% lower survival and 19–59% lower growth in the highest stress, i.e. low resource quality, treatments. The presence of benefactors significantly increased the performance of tuber moths (both survival, +31% and growth, +18%), aquatic invertebrates and carrion beetles beneficiaries (only growth, +16% and 17%, respectively) at high resource stresses (electronic supplementary material, tables S1 and S2). In all treatment combinations, the random effect of blocks did not contribute to improving model fit (AIC ≤ 2 between random effect and fixed effect models).

Consistent with stress gradient models, beneficiary consumers showed a significant switch from negative/neutral to neutral/positive interactions with increasing resource stress (RII, figure 3 and table 1). The strongest effects were measured for potato moth larvae (in terms of both survival and growth), while net positive effects were more limited for aquatic invertebrates (significant switch from negative to neutral interactions for both survival and growth) and carrion beetles (significant switch from neutral to positive interactions only for growth). Although benefactors generally improved the fitness of beneficiaries, these species never provided a reciprocal benefit for benefactors. We found no significant costs at relatively lower stress for any benefactor species but for tuber moths there was evidence of costs, i.e. significantly negative RII values at the two highest levels of resource stress (figure 3). Consequently, we found support for the commensalism hypothesis (figure 1a) for aquatic detritivore and carrion beetle benefactor species and the commensalism–parasitism hypothesis (figure 1c) for the tuber moth benefactor species.

Figure 3.

Relationship between resource quality stress and benefactor–beneficiary interactions (measured by the relative interaction index (RII); see the Material and methods section) for the three study models (potato moth, aquatic shredders and carrion beetles). The RII was calculated using two performance measures: survival and growth. The black dotted line indicates net neutral interactions. Error bars are standard deviations of replicate experimental units. * denotes significant differences (p < 0.05) in performance at a given stress level compared to the performance at the lowest stress level.

Table 1.

Linear mixed model effect of resource stress on RII values for the six different consumer species (benefactors and beneficiaries). Relative interaction index (RII) measured either with survival or biomass values. SE is standard error. Pos., positive effect; Neu., neutral effect; Neg., negative effect (figure 3). All random effects were non-significant after comparison.

	estimate	s.e.	t-value	p-value	direction
(a) RII survival
aquatic detritivores (random effect variance: 0.0003)
benefactor	0.0001	0.0003	0.359	0.721	Neu → Neu
beneficiary	0.0013	0.0004	3.266	0.0022***	Neg → Neu
tuber moths (random effect variance: 0.0003)
benefactor	−0.0011	0.0003	−3.714	0.0006***	Neu → Neg
beneficiary	0.0035	0.0005	6.947	8.8 × 10−9***	Neu → Pos
carrion beetles (random effect variance: 0.0042)
benefactor	−0.0003	0.0006	−0.401	0.695	Neu → Neu
beneficiary	0.0000	0.0000	0.736	0.471	Neu → Neu
(b) RII growth
aquatic detritivores (random effect variance: 0.0000)
benefactor	0.0000	0.0000	0.148	0.883	Neu → Neu
beneficiary	0.0022	0.0004	5.794	1.00 × 10−6***	Neg → Pos
tuber moths (random effect variance: 0.0000)
benefactor	−0.0016	0.0003	−5.268	3.22 × 10−6***	Neu → Neg
beneficiary	0.0024	0.0005	5.319	2.7 × 10−6***	Neu → Pos
carrion beetles (random effect variance: 0.0017)
benefactor	−0.0007	0.0007	−0.977	0.345	Neu → Neu
beneficiary	0.0024	0.0008	3.045	0.0087*	Neu → Pos

4. Discussion

The SGH has been shown to effectively describe the net outcome of plant interactions in communities for many ecosystems globally with a relatively consistent shift towards more positive interactions at increasing stress levels [28]. Our study supports the premise that this hypothesis can be extended to previously poorly explored taxa such as arthropods and their associated systems such as terrestrial and aquatic tropical environments. We also found that most benefactors paid no significant cost of facilitation along stress levels (full commensalism hypothesis), but that the commensalism–parasitism hypothesis was supported for the potato moth model. To our knowledge, this study is the first to experimentally show that both costs and benefits function simultaneously on environmental stress gradients for animals.

In this experiment, all beneficiary taxa performed poorly under low food quality environment, but, once their access to the resources was facilitated, they benefit from benefactors without receiving competitive effects or at least the net outcome was positive [16,29]. This supports the application of the SGH to animal systems, which are key, yet poorly studied models to further examine stress through resources. Hines & Gessner [30] argued that species neighbouring effects that reduce stress might be less important for mobile animals than for plants and sessile animals. In alpine systems for example, Barrio et al. [4] found no clear trend in the strength and direction of interactions among herbivores across productivity gradients. However, there is also evidence of stress alleviation in co-occurring animal species [31]. For example, a diversity of desert and alpine animals cope with extreme temperatures by digging burrows that can benefit to other animals [4], regardless their mobility (mobile animals may even be more facilitated). This situation is likely to apply to other types of facilitation processes involving mobile animals (e.g. shrub–animal interactions [32]).

While all tested beneficiary taxa benefit from benefactors, it is not certain whether these benefits measured under controlled experiments may have population and community-level consequences in natural settings. The complexity of natural systems includes confounding factors such as community diversity and evenness, dispersal, or trophic interactions that are likely to complicate predictions derived from experimental, species interaction studies [19]. The possibility of scaling-up our results on pairwise interactions to community levels will likely depend on the study model. For example, a heap of potatoes in traditional storage structures in the Andes is not strikingly different from our experimental units [23]. Contrastingly, a dozen of species of dung and carrion beetles can interact within a single resource pad [19] and many features of resource colonization such as attraction, phenological co-occurrence and avoidance of predators may play a role in determining the form and intensity of interspecific interactions among consumers. Consequently, the models used here may not encompass the full extent of variation that natural populations experience but are nonetheless representative models.

The feedback effect on benefactors from beneficiaries along a resource stress gradient is a critical research gap examined here. While neutral feedback seems to be the rule in these specific systems, we also detected some negative feedback for the growth of T. solanivora, the benefactor potato moth. These results are in line with previous plant studies where negative and neutral feedback effects have most often been reported ([14,33] but see [1,11] for cases of mutualistic interactions). Importantly our manipulations, together with the conceptual framework (figure 1), further revealed how environmental stress can modulate feedback effects of beneficiaries on benefactors. In most cases, feedback was neutral at low-stress levels and did not change at higher-stress levels, in line with the commensalism hypothesis. These findings support previous studies on cost for cushion plant along altitudinal gradients [11]. In only one case, T. solanivora paid a cost for facilitation at higher-stress levels, supporting the commensalism–parasitism hypothesis (figure 1c). In herbivores, reduced survival and growth in benefactors is likely the result of competition for resources with the beneficiary [34]. Once their access to the resources has been facilitated, S. tangolias can be a strong competitor and interfere with other tuber exploiters [23]. Another possibility is that benefactors have reduced abilities to cope with harsh environments and thus are less tolerant towards interspecific competition even though the strength of competition may be unchanged [8,35]. Regardless of the mechanism, the net outcome for the benefactors generally suggests that costs are not a significant factor for arthropods and this strongly suggests positive coevolution for these sets of species.

Even though we found one case of negative feedback on benefactors, we do not expect facilitation costs to be common in animal communities for several reasons (see [15], for an in-depth discussion on facilitation costs and benefits). First, animal benefactors can avoid parasitic beneficiaries by reducing their interaction with them. Avoidance responses, either spatial or temporal, are likely the most common mechanism allowing benefactors to minimize their costs with potential seasonal trade-offs between facilitation and competition. Second, animal benefactors can tolerate the beneficiaries (e.g. through niche differentiation [36]) to reduce their negative impact, thereby shifting the interaction from parasitism towards commensalism. It has been shown that negative feedback effects of plants hosted by benefactor cushions depend on the species richness and composition forming the cushion-associated species assemblage, with monocultures having stronger negative impacts than diverse species assemblages [14]. Furthermore, it is also likely that facilitation costs can be limited in diverse animal communities. Third, there could be regulatory feedback loops to minimize the negative effect of beneficiaries on benefactors. Hart & Marshall [8] showed that, in sessile marine invertebrates, the reciprocal competitive effect of a beneficiary species on its benefactor could limit facilitation under stressful conditions by limiting the benefactor's population growth.

Overall, understanding the long-term and large-scale costs and benefits of facilitation in animal communities remains a challenging research area. Further refinements of theoretical facilitation models (e.g. [2]) and tests of hypotheses presented in the conceptual framework will improve understanding of the ecological and evolutionary importance of species interaction outcomes in an increasingly stressed world. Experiments that address inclusive measures of fitness will also inform estimates of the sensitivity of these ecological interactions to changing selection processes through climate and species loss.

Ethics

The work complied with the Association for the Study of Animal Behaviour guidelines for the use of animals in research.

Data accessibility

Data are available in the electronic supplementary material.

Authors' contributions

O.D. laid out the conceptual framework. O.D., C.C. and M.H. performed the experiments and analysed the data. O.D. wrote the first draft of the manuscript with significant contribution of C.J.L. All authors read and approved the final manuscript.

Competing interests

We declare we have no competing interests.

Funding

O.D. warmly thanks the financial support by the ECOBIO Department of the French National Research Institute for Sustainable Development (IRD) for his stay at Cornell University.