Ethics statements

This study was authorized by the Autonomous Government of Catalonia (Generalitat de Catalunya) and the Natural Parks Department of the Government of Barcelona (Diputació de Barcelona). The University of Barcelona reviewed and approved the project without requirement for ethics approval. Fish were euthanized following the standard protocol recommended by the animal welfare service at the University of Barcelona (anaesthetized using Tricaine methanesulfonate (MS- 222)), and all efforts were made to minimize animal stress and suffering during this study.

Study area

The Vall d’Horta stream (41°40’24’‘N, 2°02’4’‘E; Altitude: 480 m asl) is a first order stream located in the protected area of Sant Llorenç del Munt i l’Obac Natural Park (50 km inland from Barcelona, NE Spain). The main stream course is formed from the confluence of the Pregona and Font del Llor creeks draining to the Ripoll‘s Basin (a tributary of the Besòs River). This hilly area is characterised by a Mediterranean climate and a calcareous geology, with alternating highly permeable and less permeable substrates where springs are located (see [27,38] for a detailed site description). Barbus meridionalis is a common fish within these intermittent streams that find refuge in the remaining permanent pools during periods of hydrological disconnection (usually in summer). In August 2003, a wildfire burned a forested area of 4543 ha, affecting 62% of the Vall d’Horta basin. As a consequence of this wildfire, B. meridionalis became locally extinct in some of the affected streams, even in the pools, potentially due to chemical changes that occurred during the first rainfall events [39]. The fish population has not recovered since the fire, most likely due to natural and human barriers in the lower part of the study site.

We conducted the experiment in a 100 m reach formed by a large pool where riparian vegetation was not burned by the wildfire. This reach was selected because, as observed in the years before the fire, barbels took refuge in these pools to survive periodic drought conditions present in the area when intermittent Mediterranean streams were reduced to isolated pools [40]. Physicochemical water analyses (n = 9) were performed before, during, and at the end of the experiment. The results (presented as the mean ± SE) confirmed that water of this reference stream was hard (conductivity: 520 ± 5 μS cm^-1; pH: 7.9 ± 0.1) and oligotrophic (N-NO₃^-1: 0.29 ± 0.02 mg l^-1; N-NH₃: 0.019 ± 0.003 mg l^-1; P-PO₄^3-<0.01 mg l^-1). The stream discharge averaged 15.7 ± 0.9 l s¹, which, with the very low water velocity in the pool (< 1 cm s^-1), naturally kept the pool water renewed and oxygenated (DO₂: 9.6 mg l^-1, 84.7%) during our study.

Mesocosm design

We performed an enclosure/exclosure mesocosm experiment to manipulate B. meridionalis densities. Removal experiments that simulate the loss of one or more species from a natural community can reveal the consequences of apex consumer extinctions and assess biodiversity-ecosystem function (BD-EF) relationships [41].

We used nine large cages (100 x 100 cm surface, 70 cm height; see Fig. 2) covered with a 10 mm mesh that retained fish but allowed macroinvertebrate emigration/immigration, thereby minimising the impact of our experimental design on the rate of prey exchange with the benthos [42,43]. In each cage, four plastic trays (40 x 40 cm surface, bottom of 1 mm mesh size) were used as replicates (36 trays in total); each tray contained four medium-sized stones for macroinvertebrate colonisation and three glass tiles (2 x 4 cm) for periphyton colonisation (see Fig. 2). Tray substrates within the mesocosms were complex due to the material deposited during the colonisation period; substrate was formed by a mixture of sediment, detritus and leaves, which provided some refuge to invertebrates [44,45] along with the initial added stones. To study the consequences of B. meridionalis extirpation, we tested three treatments with varying barbel density levels in the enclosures: i) no fish; ii) barbels at low density (i.e., 2 individuals m^-2, the known pre-fire density; A. de Sostoa pers. comm.); and iii) barbels at high density (i.e., 4 individuals m^-2, twofold the pre-fire density). Barbels were caught using an electrofishing source downstream from our study site, and individuals selected for the experiment were approximately the same size (total length 101.8 ± 2.6 mm; mean ± SE) and weight (2.3 ± 0.2 g). To ensure similar initial conditions, barbels were kept in observation for 24 h before starting the experiment after electrofishing and transportation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Diagram of the experimental enclosure.

Diagram of the experimental enclosure and one of the four identical trays that contained stones for macroinvertebrate colonisation and glass tiles for periphyton colonisation. Dimensions are indicated.

https://doi.org/10.1371/journal.pone.0117630.g002

Sampling and laboratory protocols

The field experiment was conducted in late spring of 2010 before pool disconnection (flow averaged 15.7 ± 0.9 l s¹), over the course of five weeks. Three weeks were allowed for periphyton and macroinvertebrate colonisation, a time previously described as adequate for equilibrating the mesocosm and background macroinvertebrate densities [46]. Two weeks were allowed for barbel interaction. During the colonisation period, the cage tops were opened to facilitate aerial colonisation and the entrance of organic material. Before the addition of barbels to the experimental enclosures, one tray per cage (n = 9) was removed and sampled to test if there were differences in colonisation among cages. Barbel density levels were randomly assigned to enclosures, and the cage tops were closed following barbel introductions to avoid bird or mammal predation. After two weeks of interaction, mesocosms were destructively sampled with the same effort for each tray (n = 27; 9 trays per treatment). Tray contents (with stones) were carefully washed in a 250 μm mesh sieve and preserved in 4% formalin until being processed in the laboratory. All samples were sorted, counted and identified. Taxonomic resolution was primarily to the genus level, including Chironomidae. Some Diptera were identified to the family level, and Oligochaeta, Ostracoda, Cladocera, Copepoda, Hydracarina and terrestrial invertebrates identified to higher levels. Each taxon was categorised as either secondary or primary consumer according to Merritt and Cummins and Tachet et al. [47,48]. Periphyton net primary production was measured as the net accumulation of chlorophyll-a on artificial substrata [49]. Chlorophyll-a was measured after extraction in acetone (90%) for 24 h in the dark at 4°C, sonication for 5 min at 40 kHz, and filtration (GF/F Whatman 0.7 μm-pore size). Following Jeffrey and Humphrey [50], chlorophyll-a concentration was determined spectrophotometrically (Perkin-Elmer, Lambda UV/VIS).

In order to test if B. meridionalis also feeds on predatory invertebrates (not only on primary consumers), and therefore, if intraguild predation occurs, we analysed barbels’ gut contents. Barbels were euthanised using an overdose of anaesthetic (MS-222). Gut contents were preserved in 4% formalin, sorted, counted, and identified at the same taxonomic level as the benthic samples.

Data analysis

To test differences among the three barbel density treatments, we used the non-parametric Kruskal-Wallis test (K-W test). Then, pairwise Mann-Whitney U-tests were used to detect significant differences between treatments. We compared total macroinvertebrate abundance (total number of individuals m^-2), taxa richness (number of different taxa), rarefied taxa richness (taxa richness corrected by macroinvetebrate abundance in the sample), Simpson’s diversity index (D, calculated as , where n_i is the number of individuals of taxon i and N is the total number of macroinvertebrates [51]), abundance of common taxa (number of individuals of each abundant taxon, i.e., > 50 ind m^-2 in the treatment lacking barbels), and periphyton net primary production (net accumulation of chlorophyll-a) among the three treatments.

We used permutational multivariate analysis of variance (PERMANOVA, ‘Adonis’ function in R) on the Bray-Curtis distance matrix, after the log-transformation of the macroinvertebrate abundance data, to test differences in macroinvertebrate community composition among treatments. Afterwards, we used indicator species analysis, using ‘IndVal’ test in R, to identify which taxa of the macroinvertebrate communities could serve as indicator for each barbel density treatment. The ‘IndVal’ test calculated the indicator value for each taxon, combining measurements of taxon specificity to each established barbel density treatment with taxon fidelity within each treatment [52]. The significance of ‘IndVal’ measures was tested using the Monte Carlo test with 9999 permutations.

We also calculated predator:prey ratios for abundance and richness, dividing the abundance (or richness) of secondary consumers by that of primary consumers for each sample. To test for intraguild predation, we also categorised each taxon found in the gut contents as either primary or secondary consumer, and calculated the proportion (%) of each category in the contents. All statistical analyses were performed in R 2.15.2., we used ‘vegan’ and ‘labdsv’ packages [53].

Ecosystem structure: ‘mesopredator release’ and ‘prey release’

Mesopredators were more abundant in mesocosms lacking barbels, supporting the ‘mesopredator release’ hypothesis (see Fig. 1B), which confirms that the loss of small-bodied top predators may have this main common effect with large-bodied predator extirpations [8–10,54]. Several predatory invertebrates that characterised the enclosures lacking barbels (e.g. Zavrelimyia sp., Parasigara sp. and Stictonectes sp.; see Table 1) dominated barbel gut contents, indicating that fish predation contributed to density reduction for these taxa in the presence of barbels. Other taxa, such as Chaoborus sp., were not found in barbel gut contents, suggesting that the density decline for some taxa was likely the result of induced emigration. Mesopredator abundance thus appears to be controlled by the top predator through the combination of predation and possible non-consumption impacts such as competition or induced emigration. Moreover, mesopredator richness also increased in top predator absence. Consequently, a basic element of trophic webs was altered [64]: predator:prey ratios differed among the barbel density treatments (see Fig. 4E-F). Even though predator:prey richness ratio has been previously considered invariant due to underlying community assembly rules [65–67], our results support other studies that did not find conservative predator:prey ratios [68,69] and suggest that secondary and primary consumers respond unequally to the presence of a top predator.

‘Mesopredator release’ did not lead to a negative or a null effect on primary consumers (see Fig. 4A), which conflicts with the original IGP theory [13–15,70]. In contrast, top predator absence led to increased primary consumer abundance (i.e., ‘prey release’), which indicates that the top predator was more effective than mesopredators at suppressing prey. A growing body of literature has posited that top predator presence does not necessarily lead to higher prey abundance if the mesopredator exclusively uses alternate prey [71] or is cannibalistic [72]. However, these new perspectives on IGP are difficult to apply in empirical studies because models continue to oversimplify real food webs (e.g. by modelling food webs with just one intermediate predator). The IGP meta-analysis of Vance-Chalcraft et al. [73] concluded that top predator presence usually leads to ‘prey release’, as predicted by trophic cascade theory, however, it suggested that this is unclear in lotic ecosystems. In this sense, our results showed that the role of the apex consumer was not functionally replaced by the remaining species [74,75], suggesting that the predator assemblage is more important than diversity per se [6,76], with species identity being the critical factor.

Our study confirmed top predator extirpation modified the whole community composition. This finding was previously reported for intermittent streams exclusively by Williams et al. [31], who found fish have a top-down effect on macroinvertebrate assemblages in isolated pools. But to our knowledge, our study is the first in demonstrating top predator extirpation can change community composition in a running intermittent stream. The treatment lacking barbels was the only that contained a large number of associated indicator taxa (see Table 1). Therefore, the presence of B. meridionalis prompted a macroinvertebrate community that was a subset of the macroinvertebrate community without the top predator. The responses of invertebrate populations to barbel presence were highly taxon-dependent, which supports evidence elsewhere that taxa within a trophic level are not functionally equivalent [75,77]. No taxon was however positively affected by barbel presence. We found a statistically significant response even from highly mobile taxa that could rapidly recolonise the enclosures by drift [56,78], indicating a strong top-predator impact. These results indicate that some invertebrates have difficulty co-occurring with this apex consumer. Thus, the local extinction of B. meridionalis offered a competitive advantage for these vulnerable species to predation, and did not lead to an extinction cascade, which conflicts with the predator-mediated coexistence theory [16]. Likewise, it contrasts with several studies that relate top predator extinctions to a decline in biodiversity [9,12]; we did not find a relationship between top predator loss and total taxa richness or Simpson’s diversity, only for mesopredator richness that increased in top predator absence.

Several studies have emphasised that top predators may be functionally extinct from an ecosystem before being extirpated [18,54,79]. Management efforts to maintain threatened top predators at persistent levels can be ecologically irrelevant if the top predator population does not reach a functionally effective abundance. In our study, the top predator at low density (i.e., pre-fire density) led to an effective suppression of mesopredators, modified the whole macroinvertebrate community composition, and increased indirectly periphyton primary production, compared to the treatment without barbels. However, part of the top predator functional role was only revealed at higher fish density, since the suppression of mesopredator richness and primary consumers’ abundance did not occur at low top predator density. These results place apex consumer density as a continuum factor that modulates top predator effects in the ecosystem, confirming that studies about functional extinction thresholds that research top-down effects of apex consumers’ extinctions at different densities are particularly relevant for ecosystem restoration and conservation purposes.

Ecosystem function: primary production response

Periphyton net primary production was significantly lower in the absence of B. meridionalis (see Fig. 5), confirming a strong trophic cascade effect that modified ecosystem function. This effect could occur through several different mechanisms, which are not necessarily mutually exclusive. Changes in primary consumer density could not fully explain the decline in primary production in top predator absence (see Fig. 4A). However, primary production could be top-down controlled by one or more taxa due to differences in the strength of this interaction, with herbivore identity being the key in the herbivore-producer interface. In this case, B. meridionalis extirpation could have increased the abundance of taxa that placed strong pressures on periphyton, triggering a trophic cascade without increasing the total abundance of primary consumers. Another explanation could be that predatory invertebrates were actually omnivorous, and ‘mesopredator release’ (see Fig. 4C) led to the increased consumption of periphyton. In addition to density-dependent causes, top predator presence could have led to higher primary production through a trait-mediated effect, reducing foraging activity by herbivores [77]. Although positive interactions have been studied less frequently by benthologists [2], B. meridionalis presence could have had a direct positive effect on periphyton production via nutrient release and/or by increasing light availability as a result of reduced sediment deposition through feeding foraging movements [35]. These results demonstrate that trophic cascades can be strengthened at the herbivore-producer interface, and contrast with those of Shurin et al. [80], which established that predators more strongly affected primary consumers compared to producers.

Our primary production results have implications for the management of natural and human-altered ecosystems. For instance, our results could modify the general view of how predatory fish abundance is linked to primary production in freshwater ecosystems, given that our results conflicted with traditional trophic cascade theory (which holds that each trophic level is related to the level above and below it in a direct and negative way [11]). In agroecosystems, biological-control practitioners often consider IGP, a very common interaction among aphidophagous predators and parasitoids [14,81]. In this context, Finke and Denno [15] advised against promoting diverse predator assemblages in which IGP was common because it would weaken the suppression of herbivore pests and reduce productivity. These kinds of generalisations can lead to ineffective management practices, particularly given that our results showed that IGP did not dampen the trophic cascade and that neither IGP nor diversity were linked to cascade strength. Instead, in agreement with Borer et al. [82], cascade strength depended on the identity of predators and herbivores. Therefore, we recommend that managers place more importance on species identity in decision-making processes to better predict management outcomes.