Different diversity–functioning relationship in lake and stream bacterial communities

Authors


Correspondence: Irene Ylla, Institute of Aquatic Ecology, University of Girona, Campus Montilivi, E-17071 Girona, Spain. Tel.: +34 972419789; fax: +34 972418150; e-mail: irene.ylla@gmail.com

Abstract

Biodiversity patterns have been successfully linked to many ecosystem functions, and microbial communities have been suspected to harbour a large amount of functionally redundant taxa. We manipulated the diversity of stream and lake water column bacterial communities and investigated how the reduction in diversity affects the activities of extracellular enzymes involved in dissolved organic carbon degradation. Dissimilar communities established in cultures inoculated with stream or lake bacteria and utilized different organic matter compounds as indicated by the different extracellular enzyme activities. Stream bacterial communities preferentially used plant-derived organic material such as cellulose and hemicellulose. Communities obtained from the lake, where the longer residence time might permit the organic matter to age, efficiently degraded lignin-like material and also showed higher peptide degradation capacities. The results highlight a stronger negative effect of decreasing diversity on ecosystem multifunctionality for stream than for lake bacterial communities. We found a relatively higher multifunctional redundancy in the lake as compared to the stream-derived cultures and suggest that community assembly might shape diversity–functioning relationships in freshwater bacterial communities.

Introduction

Biodiversity is important for ecosystem functions such as the production of biomass, the biogeochemical cycling of elements and the ability to resist invasion by exotic species (Hooper et al., 2005; Hillebrand & Matthiessen, 2009; Reiss et al., 2009). Biodiversity–ecosystem functioning relationships are often positive but saturating, suggesting some degree of redundancy (Balvanera et al., 2006; Cardinale et al., 2006). More recently, Hector & Bagchi (2007) and Gamfeldt et al. (2008) highlighted the importance of simultaneous effects of biodiversity on multiple ecosystem functions. Considering that species perform several functions at the same time, redundancy may be lower when multiple functions are considered. However, there might be no single, general relationship between species diversity and the functioning of an ecosystem because the relative contribution of species changes with environmental context and over time (Cardinale et al., 2000). Accordingly, dispersal and colonization are important modifiers of diversity–functioning relationships because these processes allow for compensatory dynamics (Hillebrand & Matthiessen, 2009). In fact, both community assembly (Fukami & Morin, 2003) and dispersal (Venail et al., 2008) have been found to modify diversity–functioning relationship in microbial communities.

In freshwater ecosystems, the activity of heterotrophic bacterial communities drives the degradation of dissolved organic matter (DOM) (Eiler et al., 2003; Kritzberg et al., 2006; Langenheder et al., 2006; Lindström et al., 2010). Stream and lake DOM is constantly supplied from catchment sources or internally produced (e.g. Tranvik, 1992; Battin, 1999; Lutz et al., 2012), and the uptake rates into bacterial biomass are strongly related to the ability of extracellular enzymes to break up large molecules into smaller ones (Chróst, 1990). Therefore, the degradation of the complex pool of organic matter requires the interplay of several extracellular enzymes, and work on biofilm communities showed the importance of bacterial diversity on multifunctionality (Singer et al., 2010; Peter et al., 2011a, b).

Free-living freshwater bacterial communities in streams and lakes are subjected to different environmental conditions (e.g. water residence time, flow velocity, DOM quantity and quality), which affect community composition and functioning (Comte & del Giorgio, 2010; Lindström et al., 2010). Moreover, stream and lake ecosystems likely differ in the way bacterial communities assemble (Besemer et al., 2012; Portillo et al., 2012). In lakes, with their long residence times, one could expect to find communities that are well-adapted to utilize the available C-source and thus perform efficiently in degrading this carbon pool. In contrast, stream communities might rather reflect a stochastic mixture of species assembled from different sources in the catchment (soils, wetlands, sediments, biofilms) (Lindström & Bergström, 2004; Crump et al., 2007; Besemer et al., 2012), and short retention times might not allow for species sorting.

Here, we manipulated the diversity of free-living stream and lake bacterial communities and measured the activity of key extracellular enzymes involved in carbon cycling. We expected the reduction in bacterial diversity to negatively affect ecosystem functioning in both ecosystems but that lake communities sustain higher levels of multifunctional redundancy than stream communities.

Materials and methods

Experimental setup

Gradients in bacterial diversity of stream and lake communities were established by dilution-to-extinction (Szabo et al., 2007; Peter et al., 2011a, b). The dilution approach results in the nonstochastic removal of organisms – depending on their initial relative abundance (Franklin et al., 2001). With each dilution step, the least abundant species are removed from the community. This means that the low diversity treatment represents a limited set of species that are present also at the highest diversity.

The bacterial communities were sampled from a stream (Fibyån, Sweden, N 59° 53.386′, E 17° 20.681′) and a lake (Ekoln, Sweden, N 59° 46.984′, E 17° 37.593′) in August 2009. Fibyån is a small, humic-rich stream that drains a protected forest area near Uppsala with a current velocity of 0.3 m s−1 during low flow in summer. Dissolved organic carbon concentration is high with 32.9 mg C L−1 and pH is circumneutral with 6.8. Conductivity is also relatively high with 56.4 μS cm−1. Ekoln is a sub-basin of Lake Mälaren, the third-largest lake (surface area: 1096 km2) in Sweden. Ekoln (surface area: 94.1 km2) is a eutrophic lake with 10–40 μg P L−1 and 1500–2200 μg N L−1, with highest concentrations after snow melting in April–May. DOC concentrations range between 12 and 15 mg C L−1, pH between 7.5 and 7.9. Ekoln has a theoretic water retention time of 1.2 years, and besides several other streams and rivers, Fibyån drains into Ekoln. During sampling, no algal bloom event was noticed. Approximately 2 L of water was sampled and filtered (GF/F, Whatman, Maidstone, Kent, UK) to eliminate eukaryote predators and particle-associated bacterial communities. Subsequently, most of the collected water was filtered (0.2 μm, Supor-200. Pall, Sweden) and autoclaved twice at 121 °C for 20 min, with a 24-h interval. This procedure prevents contamination of cultures by spores that might survive the first autoclaving step.

To prepare the dilution series, an aliquot of the same GF/F-filtered water (not autoclaved) was kept in the dark at 4 °C for app. 48 h. Bacterial abundance was determined by staining with DAPI (4,6-diamidino-2-phenylindole) and counting under an epifluorescence microscope. Afterwards, cell numbers in the inocula were adjusted to 107 cells mL−1 and stepwise diluted (1 : 10) with sterile lake or stream water to 101 cells mL−1. The different dilution steps and sterile water (control) were used to inoculate triplicate batch cultures (labelled as A, B and C) in 120-mL glass vials, which resulted in a total of 48 cultures. We include 107, 106, 105, 104, 103, 102 and 101 cells as nominal inoculum sizes and a sterile control (100). The cultures were allowed to grow at 20 °C (in situ temperature) in the dark for 7 days, and bacterial abundance was measured daily. After 7 days of incubation, extracellular enzymes and bacterial diversity were measured.

Bacterial abundance

Bacterial abundance was monitored by flow cytometry of Syto13 (Molecular Probes, Invitrogen, Carlsbad, CA) stained cells (del Giorgio et al., 1996). Every 24 h, 1 mL of each culture was sampled, fixed with 3.7% final conc. formaldehyde and stored at 4 °C. Bacterial abundance was analysed with a CyFlow Space (Partec, Görlitz, Germany) equipped with a Robbywell 96-well plate autosampler. 200 μL of sample was loaded into the 96-well plate and stained with a 1.25 μM final concentration Syto13 solution. The detector gain settings were optimized for the samples to 465 FS1 (fluorescence at 508 nm) and 245 FFC (forward scatter).

Molecular analyses (DNA extraction, PCR amplification, T-RFLP)

At the end of the experiment, 50 mL of each culture was filtered onto 0.2-μm membrane filters (Supor-200, Pall), which were stored at −80 °C. DNA was extracted from the filters using the Ultraclean Soil DNA extraction kit (MoBio Laboratories, Carlsbad, CA). Frozen filters were cut into pieces, directly added to the bead tubes and treated according to the manufacturer's instructions for maximum yield. DNA extracts were used as templates for PCR amplification of the 16S rRNA genes with the universal primers 27-forward, labelled with hexachlorofluorescein (HEX) and unlabelled 519-reverse. Thermocycling was carried out with a MyGene MG 96 Thermocycler (Longene Scientific Instruments, Hangzhou, China) using an initial 30-s denaturation at 98 °C, 28 cycles of 98 °C for 10 s, 50 °C for 30 s and 72 °C for 30 s followed by a final 7-min extension step at 72 °C. PCR products were purified using MultiScreen PCRμ96 plates (Millipore, Billerica, MA). Reactions of the PCR product with the restriction enzymes HaeIII and HinfI and corresponding buffer were incubated at 37 °C for 18 h (Liu et al., 1997). Terminal fragments were sized by electrophoretic separation and detection on a capillary sequencer (ABI 96, Applied Biosystems, Carlsbad, CA). The size and quantity of terminal restriction fragments were analysed using GeneMarker (ver. 1.7) software.

Extracellular enzyme activities

We measured the activity of five extracellular enzymes: cellobiohydrolase (EC 3.2.1.91) and β-glucosidase (EC 3.2.1.21; involved in the degradation of cellulose), β-xylosidase (EC 3.2.1.37; involved in degradation of hemicelluloses), leucine aminopeptidase (EC 3.4.11.1; drives peptides decomposition) and phenol oxidase (EC 1.14.18.1), which transform a broad spectrum of phenolic molecules like lignin (Romaní et al., 2012). β-Glucosidase, β-xylosidase and cellobiohydrolase were determined spectrofluorometrically using methylumbelliferone (MUF)-linked artificial substrates (Sigma-Aldrich). Leucine aminopeptidase was analysed using the fluorescent-linked artificial substrate L-leucine-7-amido-4-methylcoumarin hydrochloride (Leu-AMC, Sigma-Aldrich). Phenol oxidase activity was measured using L-3,4 dihydroxyphenylalanine (Sigma-Aldrich). All enzymatic activities were measured under saturating conditions; we used 0.3 mmol L−1 final substrate concentration for β-glucosidase, β-xylosidase and leucine aminopeptidase, 0.8 mmol L−1 final substrate concentration for cellobiohydrolase and 5 mmol L−1 final substrate concentration for the measurement of phenol oxidase activity. These substrates were added to three replicated samples of 4 mL for each culture. β-Glucosidase, β-xylosidase and leucine aminopeptidase were incubated for 1.5 h in the dark at 20 °C, and under continuous shaking; cellobiohydrolase and phenol oxidase, however, were incubated for 2 h. Blanks and standards of MUF and AMC were included. At the end of the incubation period, glycine buffer (pH 10.4) was added (1 : 1 v : v), and fluorescence was measured at 360/465 nm excitation/emission for MUF and AMC using a plate reader (Ultra 384, Tecan, Switzerland). Blanks were subtracted from the samples to correct for abiotic hydrolysis of the substrate or fluorescent substances in the medium. Phenol oxidase activity was measured following the method outlined by Sinsabaugh et al. (1994). At the end of incubations, absorbance was measured at 460 nm (Lambda 40, Perkin Elmer).

Mean and standard error of each enzyme activity were estimated using all measurements of cultures exhibiting growth. We further calculated enzyme activity ratios, which are indicative of the origin or the complexity of the organic matter. The ratio β-xyl/β-gluc points to the origin of the polysaccharides; a higher value of the ratio indicates that the organic matter being used is derived from hemicellulose instead of cellulose (specifically cellobiose). The ratio cbh/phox shows the origin of the complex material (cellulose or phenolic compounds), thus higher values of the ratio mean that the organic matter origin is more cellulolytic rather than ligninolytic. On the other hand, the ratio (β-gluc + β-xyl)/cbh indicates the complexity of the available polysaccharides. Higher values of the ratio (β-gluc + β-xyl)/cbh indicate higher capacity for decomposition of simple polysaccharides (like cellobiose and hemicellulose) rather than the use of complex polysaccharides (like cellulose).

Multifunctionality

To address multifunctional effects of the diversity treatment, we followed the logics outlined in Gamfeldt et al. (2008) and defined a specific level of enzyme activity to be sufficient to sustain community functioning (see also: Peter et al., 2011a, b). If any of the individual activities would drop below the threshold, we would consider this specific function lost, and consequently, the likelihood to sustain multifunctionality would be reduced. We first defined thresholds of 50%, 75% and 90% for the maximum activity for each enzyme. Subsequently, we tested for each replicate to be present above 1 or below 0 the threshold and summed the relative likelihoods. Hence, a value of 1 indicates that the activities of all five extracellular enzymes are above the respective threshold in all replicated cultures, while a value of 0 indicates that none of the enzyme activities reached above the threshold level.

Statistical analyses

Nonmetric multidimensional scaling (nMDS) and multivariate analysis of similarity (anosim) were calculated using Bray–Curtis similarities on the peak height (= relative abundance) data derived from the molecular fingerprinting of the 16S rRNA genes using the software past ver. 2.01 (Hammer et al., 2001). Multivariate dispersion, that is, the variability in community composition among stream and lake communities was calculated based on Bray–Curtis dissimilarities using PERMDISP2 (Anderson et al., 2006). Variability in the enzymatic activity (β-glucosidase, β-gluc; β-xylosidase, β-xyl; cellobiohydrolase, cbh; peptidase; and phenol oxidase, phox) and enzyme activity ratios (cbh/phox, β-xyl/β-gluc and β-gluc + β-xyl/cbh) were analysed by multivariate analysis of variance (manova). This analysis was used to test for the single-source effects of the dilution gradient and the sample source (stream or lake) and their interactions. Enzymatic activity variables were log (x + 1)-transformed to attain homogeneity of variances and normal distribution. These statistical analyses were performed using the spss software package for Windows (ver. 14.0.1, SPSS Inc, 1989–2005). To test for differences in the likelihoods to sustain multifunctionality between lake and stream samples, we applied bootstrap t-tests using Rundom Pro version 3.14 and 10 000 bootstraps of t.

Results

Abundance

All cultures with a nominal inoculum size of 103 cells and above and three cultures with 102 cells (102C stream, 102A and 102B lake) exhibited growth. None of the sterile controls and cultures with 101 cells exhibited bacterial growth. Cell numbers in the batch cultures increased steeply during 5 days until the stationary phase was reached. After 7 days of incubation, cell numbers were stable at app. 3 × 106 and 5 × 106 cells mL−1 in stream and lake communities, respectively.

Bacterial community composition (T-RFLP)

In total, 72 different operational taxonomic units (OTUs) were detected by terminal restriction fragment length polymorphism (T-RFLP). Forty different OTUs were detected by the restriction enzyme HaeIII and 32 OTUs by the enzyme HinfI. Only 21 of the 72 OTUs were detected in both lake and stream communities. Twenty-two OTUs occurred exclusively in lake samples, whereas 29 OTUs were only detected in samples from the stream. The number of OTUs was weakly correlated with the dilution gradient in lake samples and in stream samples (R = 0.582, P = 0.011; R = 0.468; P = 0.078, respectively). On average, 18 OTUs were detected in both lake and stream cultures inoculated with 107 cells, while on average, 14 OTUs were detected in the lake cultures inoculated with 102 cells and 10 OTUs in the most diluted stream culture.

We used nMDS to visualize community composition (Fig. 1). When all the samples from both sites were plotted together, the stream and lake bacterial community samples clustered separately, suggesting clearly different community composition. This was supported by anosim, which reported significant differences in community similarities in lake and stream cultures (R = 0.79, P < 0.01). Multivariate dispersion, that is, the distances from the centroid and median of the multivariate data sets (e.g. Fig. 1), was not significantly different in communities derived from the lake and from the stream (P = 0.79).

Figure 1.

nMDS plot using Bray–Curtis similarities of the 16S rRNA gene community profiles of the lake (○) and stream (△) cultures. Numbers indicate nominal inoculum size.

Extracellular enzyme activities

Cellobiohydrolase and β-xylosidase activities were higher in stream than in lake samples, while peptidase and phenol oxidase activities were higher in lake samples (Table 1, Fig. 2). Similar values were found for the β-glucosidase in both ecosystems (Table 1, Fig. 2). The activities of all five extracellular enzymes decreased along the dilution gradient with the highest values measured in cultures with highest diversity, that is, lowest dilution (107) and lower values in more diluted communities (102; Tukey's test, P < 0.001; dilution effect, Table 1, Fig. 2). However, the dilution effect was different depending on the sampling source (stream or lake). In lake communities, there was a progressive reduction in β-glucosidase and peptidase activities along the dilution gradient. In contrast, in stream communities, there was a strong and pronounced decrease in these same activities between dilution step 105 and 104. For the phenol oxidase activity, the decrease was small for lake communities and even the activity recovered at the 103 and 102 dilutions. In contrast, in stream communities, there was a decrease especially from 105 up to 102, enlarging the differences between lake and stream communities (Fig. 2). In the stream samples, the enzyme activity ratios cbh/phox and β-xyl/β-gluc were higher than in lake samples. However, in lake samples, the ratio (β-gluc + β-xyl)/cbh was enhanced (Table 1, source effect, Fig. 3). The enzymatic activity ratio (β-gluc + β-xyl)/cbh decreased with increasing dilution in both the stream and lake communities (Table 1, dilution effect and dilution × source effect, Fig. 3).

Table 1. Results of manova for the enzymatic activities and the enzymatic activity ratios considering single-source effects and interactions of two factors: dilution and source (stream or lake)
 DilutionSource (Stream–Lake)Dilution×source
Enzymatic activity
β-Glucosidase (nmol MUF mL−1 h−1)P < 0.0010.7150.042
Peptidase (nmol AMC mL−1 h−1)P < 0.0010.0010.069
 β-Xylosidase (nmol MUF mL−1 h−1)0.0010.0640.933
 Cellobiohydrolase (nmol MUF mL−1 h−1)0.0050.0330.302
 Phenol oxidase (μmol DIQC mL−1 h−1)0.001P < 0.0010.031
Enzymatic activity ratios
β-xyl/β-gluc0.1110.0570.147
cbh/phox0.008P < 0.0010.106
(β-gluc + β-xyl)/cbhP < 0.0010.0040.806
Figure 2.

Enzyme activities degradation values (mean±SE) vs. dilution gradient (diversity) in stream and lake communities, n = 3.

Figure 3.

Extracellular enzymatic activity ratios between β-xylosidase and β-glucosidase (a), between cellobiohydrolase and phenol oxidase (b) and between the sum of β-glucosidase and β-xylosidase divided by cellobiohydrolase (c). All ratios are estimated for stream and lake communities along the dilution series (nominal inoculum size of 102 to 107 cells), n = 3.

Multifunctionality

The effect of diversity loss on multifunctionality, measured as the probability to find all five enzymes active above a certain threshold, differed between cultures derived from the stream and the lake (Fig. 4). The probabilities to sustain multifunctionality along the diversity gradient were generally higher in lake samples; however, significant differences between lake and stream samples were only found at a threshold level of 50% (bootstrap t test, P = 0.047). At this threshold level, all but the most diluted lake-derived cultures along the diversity gradient sustained functioning, while already the second dilution of the river inoculum sufficed to reduced multifunctionality. As expected, the consequences of species loss on joint ecosystem functioning were much more dramatic when the threshold level increased from to 75% and 90% of maximum enzyme activity. With increasing thresholds, the reduction in the microbial diversity led to a greater loss of multifunctionality, especially in stream communities, although differences between lake and stream cultures were not significant (bootstrap t test, P = 0.3 and P = 0.8, respectively). Only the most diverse cultures of both lake and stream communities were able to sustain multifunctionality close to maximum activity (90% threshold level).

Figure 4.

The likelihood to sustain multifunctionality along the dilution gradient in lake and stream bacteria water column communities. Multifunctional probability refers to the probability for all five enzymes to be represented above a threshold level of 0.5, 0.75 and 0.9 of the highest recorded enzyme activity; n = 3.

Discussion

A number of studies have shown that broad-scale functional abilities of microbial communities might not be controlled by taxonomic diversity, indicating functional redundancy in these communities (Degens, 1998; Fernandez et al., 1999). However, a negative effect of bacterial diversity loss on multifunctionality has been previously demonstrated for bacterial biofilm communities (Peter et al., 2011a, b), suggesting that the degree of redundancy is limited when several functions are performed at the same time. Here, we show that the concerted activities of multiple extracellular enzymes are lost when bacterial diversity is reduced in communities of free-living bacterial communities of a stream and a lake. The experiment revealed a stronger negative effect of decreasing diversity on ecosystem multifunctionality for the stream than for the lake bacterial community (Fig. 4), especially at the lowest threshold level. This indicates a relatively higher multifunctional redundancy in lake cultures as compared to the stream cultures. Multifunctional redundancy implies that for each taxa eliminated, at least one of the remaining taxa was able to provide the same function at the same level (Lawton, 1994). Hence, ecosystem multifunctionality may be more sensitive to reduced diversity than individual functions. For example, although multiple populations may be capable of performing a function, they may not all perform it with the same efficiency or they may not generate the same metabolic byproducts, which may inhibit or stimulate the expression of other extracellular enzymes (e.g. in lignin decomposition pathways) (Chapin et al., 1997). Similarly, the distribution of well-performing species in communities might affect the way multifunctionality erodes with decreasing diversity. If well-performing species dominate a community, a reduction in diversity due to dilution will affect the functioning of these communities only at high dilutions.

Differences in community assembly or in environmental conditions between the lake and stream might result in such species-abundance distributions. In streams, which are characterized by very short water retention times, bacterial communities might assemble as a stochastic mix of cells collected and transported from different sources in the catchment like soils, wetlands, sediments and upstream water bodies (Crump et al., 2007; Besemer et al., 2012). Hence, stream microbial community might not have sufficient time to establish interspecific interactions, which support multifunctionality (Hector & Bagchi, 2007; Gamfeldt et al., 2008). In contrast, the much longer residence time of the lake might allow taxa that are best suited to utilize the available C-source to be selected by competitive exclusion, and hence, a well-adapted community may form (Venail et al., 2008).

The differential response of stream and lake bacterial communities to diversity loss might also be related to contrasting environmental conditions and subjected to seasonal changes. Both Ekoln and Fibyån show pronounced changes throughout the season, with highest loads of DOC and nutrients after snow and ice-melt. Such seasonal dynamics will likely influence the importance of species diversity on multifunctionality; nevertheless, in this study, samples were collected in basal moment characterized by stable and low discharge in the stream and in the lake. However, the results are a snapshot of the multifunctionality response to diversity at these two different ecosystems for the specific collected communities, and the relative importance of environmental changes on this relationship remains to be tested.

The consequences of species loss on ecosystem functioning were much more dramatic when the threshold level (i.e. the probability to perform each specific enzyme activity at the maximum level) increased from 0.5 to 0.75 and 0.9. With more stringent thresholds, virtually no multifunctional redundancy was observed in any of the communities. However, which threshold is required in natural ecosystems to sustain ecosystem functioning will depend on functions as well as ecosystems (Gamfeldt et al., 2008; Peter et al., 2011a, b). In aquatic ecosystems, degradation of most organic compounds requires the activity and interplay of a diverse set of enzymes. Therefore, the loss of multifunctionality with biodiversity could affect not only the capacity to use specific compounds but also complex degradation pathways and the capacity to use distinct organic matter sources.

Although the stream and lake bacterial communities were located in close spatial proximity, community composition in the cultures differed significantly (Fig. 1). Also, the dissimilar stream and lake bacterial communities were able to use different organic matter compounds as indicated by the different extracellular enzyme capabilities. The much higher rates of cellobiohydrolase and the cbh/phox and β-xyl/β-gluc ratios in stream communities indicate a preferential use of large polymeric carbon compounds (cellulose and hemicellulose decomposed respectively by the enzymes cellobiohydrolase and β-xylosidase) probably derived from plant material (Artigas et al., 2009; Sinsabaugh & Follstad Shah, 2011). In contrast, the rates of peptide degradation (leu-aminopeptidase) were higher in lakes, suggesting a greater availability of high-quality organic matter (providing C and N) probably derived from phytoplankton exudates or decaying cells (Debroas, 1998; Naito et al., 2012). Lake bacteria in the water column communities showed an elevated phenol oxidase activity indicating the use of lignin compounds, such as those derived from decaying woody and plant material (Sinsabaugh, 2010). The low cbh/phox ratio in lake communities further suggests a major degradation of recalcitrant and complex carbon substrates like lignin than polymeric polysaccharides (Sinsabaugh & Follstad Shah, 2011).

Although multifunctionality was less affected by diversity loss in the lake bacterial communities than in stream communities, in both cases diversity loss caused a similar reduction in the (β-gluc + β-xyl)/cbh enzyme ratio (Fig. 3). The cellobiohydrolase activity was always lower than that of β-glucosidase and β-xylosidase, but its absolute reduction through the diversity gradient was lower. This point to a reduction in the capacity to decompose simple polysaccharides relative to the degradation of large cellulose molecules along the diversity gradient. The maintenance of the cellobiohydrolase activity might be related to its key role in the degradation of cellulose, which is among the most abundant biopolymers (Atlas & Bartha, 1987).

In conclusion, we demonstrate that a reduction in the bacterial diversity in freshwater ecosystems leads to a loss of extracellular enzyme activity, especially when multiple ecosystem functions are considered at the same time. Our findings indicate that diversity–functioning relationships differ between lakes and streams and that difference in community assembly might account for this.

Acknowledgements

This study was funded by grants from the Malmèns Stiftelse to HP, the Swedish Research Council to LJT and from the Generalitat de Catalunya (beques per estades de recerca fora de Catalunya) to IY. We thank X. Feng for assistance with the molecular analysis.

Ancillary