Decomposer communities in human-impacted streams: species dominance rather than richness affects leaf decomposition

Authors


Correspondence author. E-mail: mikko.tolkkinen@ymparisto.fi

Summary

  1. There is compelling evidence that anthropogenic disturbance can decrease biodiversity and impair ecological functioning. A major challenge to biodiversity–ecosystem function research is to disentangle the effects of biodiversity loss on ecosystem functions from the direct effects of human disturbance.
  2. We studied the influence of human disturbance (acidification and eutrophication) and a natural stressor (low pH due to bedrock geology) on leaf-shredding macroinvertebrates, fungal decomposers and leaf decomposition rates in boreal streams. We used pyrosequencing techniques to determine fungal richness and assemblage structure.
  3. Decomposition rates were higher in anthropogenically disturbed than in circumneutral reference or naturally acidic sites, but did not differ between the latter two groups. Macroinvertebrate richness was higher in circumneutral than in human-impacted or naturally acidic sites, and shredder evenness was highest in circumneutral sites. Fungal evenness was also lower in human-disturbed than in reference sites, whereas fungal richness did not vary among site groups.
  4. Decomposition rate in fine-mesh bags was related positively to current velocity and fungal dominance, while in coarse-mesh bags, it was related positively to total phosphorus. In anthropogenically disturbed streams, the effects of low pH were overridden by eutrophication, and increased decomposition rates resulted from disturbance-induced increase in species dominance rather than richness. Furthermore, decomposition rates were positively correlated with abundances of dominant taxa, suggesting that ecosystem processes may be driven by a few key species.
  5. Synthesis and applications. Our results suggest that leaf decomposition rates are insensitive to natural background variation, supporting the use of decomposition assays, preferably accompanied by molecular analysis of fungal assemblages, to assess stream ecosystem health. Instead of focusing solely on diversity, however, more emphasis should be placed at changes in dominance patterns, particularly if management aims are to improve stream ecosystem functioning.

Introduction

The unprecedented loss of biodiversity during the last few decades has prompted a massive amount of research towards the effects of biodiversity on ecosystem functioning. A major challenge to biodiversity–ecosystem function (B-EF) research is to disentangle the indirect effects of biodiversity loss from the direct effects of human disturbance on ecosystem processes (Hillebrand & Matthiessen 2009). For stream ecosystems, there is compelling evidence that species diversity may enhance ecosystem functions. Positive relationships between leaf decomposition rate and leaf-shredding invertebrate (Jonsson & Malmqvist 2000) or aquatic hyphomycete (Dang, Chauvet & Gessner 2005; Duarte et al. 2006) richness have been frequently reported, although neutral relationships (Perkins et al. 2010) are not uncommon either. Most of these studies were microcosm experiments where species richness was manipulated at scales that may not easily translate to scales relevant for biodiversity conservation (Srivastava & Vellend 2005). Field studies have produced less conclusive results: positive (Jonsson, Malmqvist & Hoffsten 2001), neutral (Dangles et al. 2004) and negative (McKie et al. 2009) relationships between shredder richness and leaf decomposition rate have been reported.

In experimental studies, decomposition rates have generally saturated at a low level of diversity (Jonsson & Malmqvist 2003; Dang, Chauvet & Gessner 2005), suggesting that addition of species above a critical level may not further increase decomposition rates. Furthermore, most freshwater B-EF studies have focused on species richness instead of exploring also changes in species' relative abundances (Woodward 2009). However, species evenness may respond more rapidly to environmental change than does richness (Hillebrand, Bennett & Cadotte 2008). Changes to evenness may therefore affect ecosystem functions well before species go extinct, and species dominance rather than richness may be critical in determining the rates of ecosystem functions in real-world ecosystems (Dangles & Malmqvist 2004). Changes in function may strongly depend on species identity, suggesting that the effects of human disturbance on ecosystem processes may be driven by a few particularly efficient taxa (Dangles et al. 2004).

Taxonomic identification of aquatic fungi is traditionally based on spore morphology. The number of taxa thus detected is usually quite low, often <20 species per site. Recently, molecular DNA sequence-based methods have allowed the discovery of numerous fungal taxa that have gone undetected via traditional microscopy (Bärlocher 2007). A great majority of these newly detected taxa are very rare, however, and they may thus have a negligible role in ecosystem functions (Woodward 2009) although, under changed environmental conditions, rare taxa may also prove to be functionally important (e.g. Lyons et al. 2005).

Low pH and associated metals are often connected to reduced richness of decomposer assemblages (Niyogi, McKnight & Lewis 2002; Petrin, Laudon & Malmqvist 2008) and depressed decomposition rates (Dangles & Chauvet 2003). Elevated nutrient concentrations, by contrast, may accelerate leaf decomposition mainly because of increased microbial activity (Gulis & Suberkropp 2003; Ferreira, Gulis & Graça 2006). Streams are often affected by both acidification and eutrophication, and taxonomic groups differ strongly in how they respond to the combined effects of these two stressors (Mykrä et al. 2012). Sensitivity to low pH may also vary depending on the source of acidification. Anthropogenic acidification is often linked to atmospheric pollution or discharges from local industry or mining activities (Niyogi, McKnight & Lewis 2002). Stream water pH can also be naturally low because of organic acids from peatland-dominated catchments (Mattsson et al. 2007) or specific geology (Kwong, Whitley & Roach 2009). While humic substances in peatland streams may buffer against harmful effects of acidification, low pH caused by background geology is often associated with toxic metals (Loukola-Ruskeeniemi et al. 1998). Anthropogenic acidification is arguably one of the most detrimental human disturbances to freshwater ecosystems, whereas communities in naturally acidic humic streams may be as functional as those in circumneutral streams (Dangles, Malmqvist & Laudon 2004; Petrin, Laudon & Malmqvist 2007). However, little is known about how naturally low pH caused by bedrock geology affects stream communities and ecosystem functioning.

We examined how land-use-related stress (forestry and agriculture) affects the diversity of decomposer communities (macroinvertebrates and fungi) and leaf decomposition rates in boreal streams. We selected two sets of near-pristine reference sites: (i) circumneutral streams and (ii) streams with naturally low pH due to specific catchment geology. These two sets of nonimpacted streams were compared with human-modified streams. We expected both macroinvertebrates and fungi to have lower taxonomic richness and evenness (i.e. higher dominance) in both anthropogenically impacted and naturally acidic sites compared with circumneutral reference sites. We further expected leaf decomposition rates to differ among the stream types, being fastest in the impacted sites where high nutrients stimulate microbial activity and lowest in naturally acidic sites. Our impacted sites were affected by both low pH and high nutrients, and we therefore expected that the positive effects of nutrients on decomposition might be partially cancelled by low pH; then, decomposition rates in impacted streams would not differ significantly from those in circumneutral streams. Finally, we assessed the environmental determinants of the diversity of fungi and leaf-shredding invertebrates and how variation in decomposer diversity and environmental conditions relates to decomposition rates.

Materials and methods

Study Sites

We conducted our study in 30 s-to-fourth order streams in northern and western Finland between 63° and 66°N and 22° and 27°E. Our data consist of ten anthropogenically disturbed sites and twenty near-pristine reference sites. We selected two sets of ten reference sites: circumneutral sites (near-pristine streams typical of western Finland) and naturally acidic sites with catchments dominated by metamorphosed black shale. Because of black shale, surface waters in these sites have low pH (Loukola-Ruskeeniemi et al. 1998; Table S1 in Supporting Information). All reference streams had little agriculture (<5%) in their catchments, no forestry activities near the sampling sites and no obvious signs of human impact in the riparian zone or in stream channel. The third group (hereafter, anthropogenically disturbed sites) was selected from the coastal areas of western Finland. Due to intensive land use and geological characteristics (acid sulphite soils), these streams frequently face episodic acidification during autumnal runoff. The anthropogenically disturbed sites were selected based on catchment land use and water chemistry (agricultural land use >5%, Tot-> 50 μg L−1, minimum pH <5·5).

Leaf Decomposition

Six grams of dried alder Alnus incana (L) leaves were enclosed in 15 × 15 cm mesh bags. We used two different mesh sizes to allow (8 mm, coarse) or exclude (0·2 mm, fine) shredding invertebrates. In a preliminary experiment, we detected no effects of mesh size on fungal richness and evenness (coarse vs. fine; t test, both > 0·80), suggesting that fungal communities did not encounter hypoxic conditions during the experiments. Additionally, several studies that have used a larger mesh size for the fine-mesh treatment have reported a potential confounding factor, as these bags are often entered by small-sized shredders (McKie, Petrin & Malmqvist 2006; Flores et al. 2013).

The leaves were collected prior to abscission in September 2009 and air-dried at room temperature (+22 °C) for two weeks. At each site, five bags of both mesh sizes were anchored to the stream bed using house bricks. The experiment started in mid-September, and the bags were removed after 30 days, sealed in zip-lock bags and transferred to a freezer. In the laboratory, litter bags were gently cleaned to remove other material, and invertebrates from the coarse-mesh bags were preserved in ethanol and identified in the laboratory, mostly to species level. From each fine-mesh bag, a subsample (12·5 cm2) was taken for the extraction of fungal DNA and measurement of ergosterol content. The weight of the subsample was taken into account when calculating leaf decomposition rate. The remaining leaf material was dried for 48 h at 60 °C, and subsamples were ashed for 4 h at 550 °C to convert air dry mass to ash-free dry mass (AFDM). Leaf decomposition rates (k) were determined using the negative exponential model (Benfield 1996). Leaf mass loss caused by leaching and handling was measured and accounted for in the analyses. Because differences between stream types in mean water temperature (based on daily measurements in seven streams per group) were negligible (anova: F2,18 = 0·074, = 0·92), decomposition rates were not adjusted for degree days. All fine-mesh leaf bags were used to calculate mean decomposition rates for each site, whereas coarse-mesh bags with <10% of their initial weight remaining were excluded (6% of litterbags).

DNA Isolation, Library Construction and Sequencing

Fungal assemblage structure was examined using pyrosequencing. DNA was extracted from 0·07 g of frozen leaf material using PowerSoil DNA Isolation Kit (MOBIO Laboratories, Carlsbad, CA). rDNA coding regions were amplified using the fungal ITS primers 5′-CTTGGTCATTTAGAGGAAGTAA-′3 and 5′-TCCTCCGCTTATTGATATGC-′3 (White et al. 1990; Gardes & Bruns 1993). The amplicons were sequenced using the GS FLX 454 system (Roche, Basel, Switzerland). Sequences and quality scores were extracted from the original SFF files. ITS sequences were located using ITS extractor (Nilsson et al. 2010), and false positives were removed. Because of varying lengths of sequences, complete ITS1 sequences were extracted from the reads using a custom script. Reads shorter than 250 nucleotides were discarded, and ITS1 sequences were cut off from their flanking regions. On average, 77% (2948–17144) of the sequences remained for downstream analyses. Trimmed and filtered sequences were assembled into contigs with CodonCode Aligner version 3.7.1 (CodonCode Corporation, Dedham, MA) using end-to-end alignments. The minimum percentage identity was 98%, minimum overlap length 20 nt and word length 12 nt. If necessary, the reads were manually edited and reassembled. Singleton sequences were discarded, and the remaining contigs with two or more reads were handled as OTUs (Operational Taxonomic Units). The OTUs were annotated using BLAST (The Basic Local Alignment Search Tool) searches to NCBI (The National Center for Biotechnology Information, USA), GenBank's nonredundant nucleotide database. Naming was based on the best BLAST hits. As there were several OTUs with a nonunique annotation, the contigs were aligned using MUSCLE (Edgar 2004) with default settings. The alignment including all contigs from different samples was used to construct a neighbour-joining tree (Saitou & Nei 1987), which was utilized to cluster the OTUs within and among samples. During this step, OTUs with the same best blast hit or no annotation were also phylogenetically classified.

Ergosterol Content of Decomposing Leaves

Fungal biomass was estimated from freeze-dried, pulverized leaf samples using a modified ergosterol assay (Nylund & Wallander 1992). Ergosterol extracts were quantified with high-pressure liquid chromatography (HPLC) using a reverse-phase C18 column equipped with a precartridge and methanol as the eluent (1·0 mL min−1, column temperature 30 °C). Commercial ergosterol (5,7,22-Ergostatrien-3ß-ol, Fluka AG) was used as standard. Results are expressed as ergosterol concentration in the litter (μg g−1 litter DW).

Benthic Samples

In addition to leaf-bag samples, we also collected kick-net samples of benthic invertebrates at each site at the time of leaf-bag removal, except two anthropogenically disturbed sites where flooding and formation of anchor ice prevented sampling. We took four 30-sec kick-net samples (mesh size 0·3 mm) covering most microhabitats present in a riffle and pooled them. The area covered by such a sample is 1·2 m2. Samples were preserved in ethanol and transferred to the laboratory where macroinvertebrates were sorted and identified, mostly to species level. Macroinvertebrates were assigned to functional feeding groups according to Moog (2002), and all statistical analyses were restricted to leaf-shredding invertebrates.

Environmental Variables

Concurrently with invertebrate sampling, we measured a suite of environmental variables known to be important determinants of macroinvertebrate and/or fungal diversity in streams. We included variables describing the structure of the in-stream habitat, as well as water chemistry variables. We measured current velocity at 30 random locations along cross-sectional transects, the number of transects varying from four to eight depending on stream width. Moss cover and substrate size were estimated visually at 10 randomly placed 50 × 50 cm quadrats. Percentage of nine particle size classes and organic sediments were assessed for each quadrat using a modified Wenthworth scale from silt (0) to large boulder and bedrock (9) (see Mykrä et al. 2011). These estimates were then averaged to give a mean particle size for a site. Water samples were analysed for pH, alkalinity, total phosphorus, iron and water colour using national standards (National Board of Waters 1981).

Statistical Analyses

We used analysis of variance (anova) to examine differences between stream types (circumneutral reference, naturally acidic, anthropogenically disturbed) in species richness, evenness, and abundance of shredders (numerical abundance) and fungi (ergosterol content). Evenness was calculated as H'/log(richness), where H' is Shannon diversity. We used two-way anova, with stream type (three levels) and mesh size (two levels) as fixed factors to examine differences in leaf decomposition rates (k). Decomposition rates in coarse-mesh bags were corrected for decomposition in fine-mesh bags (k coarse minus k fine) to obtain the fraction of decomposition attributable to detritivore feeding and physical abrasion (McKie, Petrin & Malmqvist 2006). Homogeneity of variances was checked using Levene's tests, and variables were logarithmic transformed if needed.

Multiple linear regressions were used to examine the relationships between leaf decomposition rate and biological (richness and evenness of shredder taxa/fungal OTUs, ergosterol content, shredder abundance) and environmental variables. For these regressions, macroinvertebrate richness and evenness were calculated using invertebrates from the leaf bags, as these are presumably directly related to leaf decomposition. Because the number of leaf bags varied among streams, we rarefied all samples to the same number of individuals (20). We screened correlations between water chemistry variables and selected the least intercorrelated (< 0·6) ones (total phosphorus and pH). We also included three in-stream habitat variables: current velocity, moss cover and substrate size. Current velocity was selected because physical disturbance can affect leaf decomposition (McKie & Malmqvist 2009), and moss cover and substrate size were selected to represent in-stream habitat characteristics known to be important regulators of boreal stream communities (Mykrä et al. 2011). These three in-stream variables were only weakly intercorrelated (all < 0·30, all > 0·10).

The most parsimonious multiple linear regression (MLR) model was selected using the regression with empirical variable selection (REVS) approach (Goodenough, Hart & Stafford 2012). REVS is based on ‘branch-and-bound’ all-subsets regression approach, and it quantifies the amount of empirical support for each variable in a data set. A series of models is created: the first model contains the variable with most empirical support, the second model contains the first variable and the next most-supported, and so on. The best model is selected based on Akaike Information Criterion (AIC). This approach has been shown to be superior to full, stepwise and all-subsets models for most types of ecological data (Goodenough, Hart & Stafford 2012). To examine the independent influence of each variable significantly related to decomposition in MLR, we constructed partial regression plots by regressing the independent variable of interest (e.g. shredder richness) against another independent variable (e.g. current velocity). Residuals of this regression were then plotted against residuals of the regression of the dependent variable (leaf decomposition) and the other independent variable (e.g. current velocity). This approach partials out the effect of other variables on the relationship and allows examination of the independent effect of an explanatory variable (Moya-Laraño & Corcobado 2008). Residuals were only used for graphical purposes, not for statistical tests, because of the inflated type II error rate. If needed, environmental variables were logarithmic transformed to approximate normality. Analyses were performed using LEAPS (version 2·9 (Lumley 2012)), MASS (version 7.3–8 (Venables & Ripley 2002)), Vegan (Oksanen et al. 2010) and QuantPsyc (Fletcher 2008) packages in R (R Development Core Team 2010).

Results

Biological Response Variables

Shredder richness in benthic samples varied from two to nine species per site. Differences between stream types were significant (F2,26 = 5·60, = 0·009), richness being higher in circumneutral reference sites than in anthropogenically disturbed or naturally acidic sites, whereas the latter two did not differ (Fig. 1a). Shredder evenness also varied significantly between stream types (F2,26 = 4·11, = 0·028), being higher in circumneutral reference sites than in anthropogenically disturbed sites (Fig. 1b). Total shredder abundance varied widely, from 24 to more than 1300 individuals per site, with no significant differences among stream types (F2,26 = 0·53, = 0·529) (Fig. 1c). Shredder richness in the leaf bags varied from one to six species per site. Both richness and evenness were highest in the circumneutral reference sites, but differences among stream types were not significant (F2,25 < 0·561, > 0·50).

Figure 1.

Shredder species richness (a), evenness (b) and abundance (c), and fungal OTU richness (d), evenness (e) and biomass (ergosterol content) (f) in circumneutral reference sites (Neutral), naturally acidic sites (Acidic) and anthropogenically disturbed (Disturbed) sites. Boxplots represent median values with interquartile range. Whiskers represent maximum and minimum values. Sites sharing a letter do not differ significantly (Tukey's test, < 0·05).

The stoneflies Nemoura sp., Leuctra sp. and Taeniopteryx nebulosa L. (34%, 32% and 11% of total shredder abundance, respectively) dominated in benthic samples. Leuctra sp. was the most abundant shredder in circumneutral reference sites (33%) and in naturally acidic sites (55%), while Nemoura sp. dominated in anthropogenically disturbed sites (73%). Abundance of Nemoura sp. correlated negatively with shredder evenness (= −0·48, = 0·01) and positively with total phosphorus (= 0·67, < 0·001). Nemoura sp. (35%), Limnephilidae sp. caddis larvae (19%) and Leuctra sp. (16%) were the dominant shredders in the leaf bags. Nemoura and Leuctra were the two most abundant taxa in both types of reference streams, while Asellus aquaticus L. (19%) was the second-most abundant taxa in anthropogenically disturbed sites. Asellus abundance correlated positively with total phosphorus (= 0·39, = 0·048).

Fungal OTU richness varied from 41 to 163, and leaf ergosterol content from 18·1 to 121 μg g−1 litter DW. Fungal richness did not vary significantly among stream types (F2,27 = 0·58, = 0·565) (Fig. 1d), whereas evenness did (F2,27 = 8·30, = 0·001). Evenness was higher in both types of reference sites than in anthropogenically disturbed sites (Fig. 1e). Ergosterol content also varied significantly among stream types (F2,27 = 8·56, = 0·001), being lower in the circumneutral reference sites than in anthropogenically disturbed or naturally acidic sites, whereas the latter two did not differ significantly (Fig. 1f).

Ascomycota was the most common fungal group comprising 70·6% of all recorded OTUs, followed by Basidiomycota and Chytridiomycota that comprised 20·4% and 5·7% of the OTUs, respectively. Based on BLAST analysis, OTUs were largely similar to aquatic hyphomycete taxa previously found on decomposing leaves in streams (Nikolcheva, Bourque & Bärlocher 2005) (see Table S2). The most common fungal identity was the uncultured fungus clone AA45-20 (33% of all OTUs), which has been reported in only a single study in a woodland stream in North America (Kelly et al. 2010) (these strains were 100% similar by the sequenced region). The proportional abundance of the uncultured fungus clone AA45-20 correlated strongly with fungal evenness (= −0. 56, = 0·002) and total phosphorus (= 0·53, = 0·003), suggesting that it may have been the most influential fungal OTU in creating the observed patterns of fungal evenness and leaf decomposition rates.

Leaf Decomposition Rates

Leaf decomposition rates (coarse minus fine-mesh bags) in coarse-mesh bags varied from 0·001 to 0·026 day−1 in the circumneutral reference sites, from 0·003 to 0·032 day−1 in naturally acidic sites and from 0·009 to 0·058 day−1 in anthropogenically disturbed sites. Corresponding figures for the fine-mesh bags were 0·007–0·012 day−1, 0·005–0·011 day−1 and 0·004–0·016 day−1, respectively (Fig. 2). Decomposition rates were higher in coarse than in fine-mesh bags (F1,54 = 8·07, = 0·006), and they also differed among stream groups (F2,52 = 8·39, = 0·001), being significantly higher in the disturbed sites than in circumneutral or naturally acidic reference sites (Tukey's test, < 0·011). Decomposition was faster in anthropogenically disturbed sites for both mesh sizes, but more strongly so for coarse than fine-mesh bags (Fig. 2), as also indicated by the marginally significant interaction between mesh size and stream type (F2,52 = 3·00, = 0·06).

Figure 2.

Leaf decomposition rates in circumneutral reference sites (Neutral), naturally acidic sites (Acidic) and anthropogenically disturbed (Disturbed) sites. Decomposition rate in coarse minus fine-mesh bags represents the fraction of decomposition attributable to detritivore feeding and/or physical abrasion.

Relationships Between Biological and Environmental Variables

Shredder richness in benthic samples was positively related to moss cover and pH, although moss cover was the only significant predictor of richness (Table 1). Together these variables explained 34% of variation in shredder richness. Shredder evenness was correlated to total phosphorus (negative relationship), current velocity (positive) and particle size (nonsignificantly negative) (32% of variation explained) (Table 1). Shredder abundance was related to both total phosphorus (positive) and current velocity (negative), together explaining 31% of variation in abundance.

Table 1. Summary of the regression with empirical variable selection (REVS) approach to select the best multiple linear regression models [based on Akaike Information Criterion (AIC)] between the biological response variables and environmental variables
Dependent r 2 P AICIndependent Coefficient P
Shredder richness0·34<0·0132·1Moss cover0·410·02
pH0·280·11
Shredder evenness0·320·02−86·9Tot-P −0·450·01
Current velocity0·360·04
Particle size−0·280·10
Shredder abundance0·31<0·01338·9Tot-P 0·48<0·01
Current velocity −0·440·02
Fungal OTU richness0·340·01180·4Particle size0·420·01
Tot-P −0·370·05
Fungal OTU evenness0·42<0·01−169·4Tot-P −0·65<0·01
Ergosterol0·60<0·01181·4pH−0·76<0·01
Tot-P 0·260·05
Moss cover0·210·12

Fungal richness was related positively to particle size and negatively to total phosphorus, the latter coefficient only bordering at significance. Fungal evenness was negatively related to total phosphorus that explained 42% of variation in evenness. Leaf ergosterol content was related negatively to pH and positively to total phosphorus (Table 1).

Leaf Decomposition Rates in Relation to Biological and Environmental Variables

Leaf decomposition rate in fine-mesh bags was best explained by current velocity (positive relationship) and fungal evenness (negative), which together explained 48% of variation (Table 2, Fig. 3). Total phosphorus was the only significant predictor of decomposition rate in coarse-mesh bags (coarse minus fine), explaining 22% of variation (Table 2). Asellus aquaticus was the only shredder species whose abundance was positively related to decomposition rates (= 0·53, = 0·006). The proportional abundance of two fungal OTUs correlated with decomposition rate in fine-mesh bags: the uncultured fungus clone AA45-20 (= 0·54, = 0·002) and Anguillospora filiformis Greath. (= 0·42, = 0·02). Partial regression plots showed that both current velocity and fungal evenness were related to decomposition rates in fine-mesh bags (Fig. 3a–d), while in coarse-mesh bags, decomposition was only related to total phosphorus (Fig. 3e).

Table 2. Summary of the regression with empirical variable selection (REVS) approach to select the best multiple linear regression models [based on Akaike Information Criterion (AIC)] between decomposition rates (k) and biological and environmental explanatory variables
Dependent r 2 Overall PAICIndependent Coefficient P
Decomposition, coarse - fine0·230·01−165·1Tot-P0·480·01
Decomposition, fine0·48<0·01−359·4Current velocity0·54<0·01
Fungal evenness−0·380·01
Figure 3.

Regression plots (a, c, e) and corresponding partial regression plots (b, d) between decomposition rates and selected [(based on Akaike Information Criterion (AIC)] explanatory variables. Partial regression plots show the effects of a given variable on decomposition rate when the effects of other variables are controlled for. Open circles = circumneutral reference sites, filled circles = naturally acidic sites, triangles = anthropogenically disturbed sites.

Discussion

Human activities have reduced the biodiversity of stream communities, with consequences on ecosystem functioning. However, while microcosm experiments have rather consistently demonstrated a positive relationship between species richness and ecosystem functions, field experiments and surveys have produced more variable results (Gessner et al. 2010). In our study, anthropogenic disturbance had a strong effect on macroinvertebrate and fungal assemblages, as well as on leaf decomposition rate. Species richness of both taxonomic groups was unrelated to decomposition, whereas fungal evenness was negatively correlated to decomposition, suggesting that human-induced changes to dominance rather than richness influence ecosystem functions. Despite differences in shredder diversity between naturally acidic and circumneutral streams, decomposition rates did not differ among these two stream types, indicating redundancy among species and possible adaptation to naturally low pH (Petrin, Laudon & Malmqvist 2007).

Nutrient enrichment often increases dominance by homogenizing resources and thus favouring competitively superior species (Hillebrand et al. 2007). Nutrient addition also increases leaf decomposition rates (Gulis & Suberkropp 2003), mainly by stimulating fungal activity (Dang, Chauvet & Gessner 2005). In our study, leaf decomposition was indeed faster in impacted streams, despite these streams also having relatively low pH. Furthermore, fungal evenness was significantly lower in the nutrient-stressed than in reference sites. Fungal evenness was the only significant biological predictor of decomposition rates, suggesting that change in fungal dominance structure was the main biological mechanism driving enhanced decomposition in impacted streams. Current velocity was also related to decomposition but only in fine-mesh bags. Velocities were generally higher in the impacted streams and could have directly affected decomposition, but it is equally possible that physical fragmentation was enhanced by high microbial activity.

The degree of dominance in the shredder guild was high in the impacted streams, but shredder evenness did not show any relationship to decomposition. This is contrary to Dangles & Malmqvist (2004) who reported faster leaf decomposition when shredder communities exhibited high dominance, regardless of the identity of the dominant species. In fact, the dominant shredder in their streams was the same nemourid stonefly that was the most abundant shredder in our study also. The isopod Asellus aquaticus was also abundant and positively related to decomposition in coarse-mesh bags. Indeed, Asellus may often have a disproportionate effect on stream ecosystem functions, for several reasons (see also Bergfur et al. 2007). First, it has a purely aquatic life cycle and is therefore present all year around (Murphy & Learner 1982). It is also highly tolerant to eutrophication (e.g. Woodward et al. 2012), and the combination of high nutrients and presence of Asellus could partly explain the enhanced breakdown rates in our impacted streams. That shredder species identity may matter more to leaf decomposition than does species richness has been previously documented by several authors (e.g. Woodward 2009; Dangles et al. 2011).

Leaf-shredding invertebrates responded to human impact, as expected, by having lowest richness in the impacted streams. By contrast, fungal richness did not differ among the stream types, and one reason for the lack of a relationship between fungal richness and ecosystem function may have been the very high fungal diversity. In experimental studies, the number of taxa is typically low and decomposition rate saturates at low levels of richness, suggesting that even large reductions in fungal richness may not affect decomposition (Bell et al. 2009). Experiments, however, oversimplify fungal communities by having even inoculum mixtures, while natural assemblages usually comprise a few dominants and a large number of rare species. Using more realistic fungal assemblages, Dang, Chauvet & Gessner (2005) found no effect of fungal richness on leaf decomposition rate, arguing that species identity and trait structure were more important determinants of decomposition than was diversity. Diversity effects are also readily outweighed by excess nutrients (Bärlocher & Corkum 2003), further suggesting the primacy of microbial community composition over diversity on decomposition. Studies using fungal spore morphology and older molecular techniques have undoubtedly underestimated fungal richness. However, although pyrosequencing produces substantially higher estimates of richness, it does not tell which fungal OTUs contribute to decomposition and by how much. In fact, our results suggest that the effect of fungi on decomposition is driven by a few key taxa whose presence was significantly related to decomposition rate. Given that ergosterol content and fungal dominance were strongly correlated and responded consistently to nutrient stress (higher ergosterol content and higher dominance in nutrient-stressed streams), we are confident that fungal OTU evenness reflected functionally important changes in fungal assemblages. Species dominance, and the identity of the dominant taxa, may therefore be a more appropriate measure of human disturbance than is species richness per se, particularly in such hyperdiverse communities as aquatic fungi.

High abundances of leaf-shredding stoneflies in acidic streams have been related to high leaf litter standing stocks (Woodward 2009). Increased shredder abundance, in turn, may compensate the functional effects of reduced species diversity in these streams (Petrin, Laudon & Malmqvist 2007). In our study also, shredder abundances were higher in naturally acidic than circumneutral streams, albeit not significantly so. Fungal biomass was positively related to nutrients, further indicating that the effects of low pH were largely overdriven by eutrophication. Fungal biomass was also related to pH, but negatively so, indicating that some fungal species may benefit from competitive release at low pH and that the effects of acidification and eutrophication on fungal assemblages may be synergistic (Dangles & Chauvet 2003).

Our results suggest that leaf decomposition rates are not particularly sensitive to natural background variation, supporting the use of leaf breakdown assays for stream bioassessment (Gessner & Chauvet 2002; Woodward et al. 2012). Nevertheless, using both functional and taxonomic measures will in most cases allow the most comprehensive assessment of biological responses to environmental stressors. On a broader perspective, our impacted sites are only moderately enriched, belonging to the ‘zone of maximum uncertainty’ where functional indicators of ecosystem health (e.g. leaf decomposition) react more sensitively to eutrophication than do traditional structural measures (Woodward et al. 2012). Fungal assemblages did, however, respond to this level of human impact, and leaf breakdown experiments accompanied by molecular analysis of fungal assemblages are therefore a promising candidate for stream bioassessment, particularly as the ‘next-generation sequencing’ techniques are becoming readily available. Nevertheless, instead of focusing solely on diversity, more emphasis should be placed at changes in dominance patterns and identification of species with a disproportionate effect on ecosystem processes. As the direct control of lotic biodiversity may be impossible in most situations, stream managers need to carefully balance which of several interacting stressors should be given priority in any given situation. Wagenhoff, Townsend & Matthaei (2012) showed recently that managing fine sediments attains primacy over eutrophication, while our study suggests that controlling excessive nutrient inputs is critical, particularly if the key objective of management is to improve ecological functioning of stream ecosystems.

Acknowledgements

We thank Taina Romppanen and Tarja Törmänen for laboratory assistance, and Brendan McKie and two anonymous reviewers for constructive comments on previous drafts of the manuscript. Our study was funded by Maj and Tor Nessling foundation (to MT), Academy of Finland (to HM and TM) and University of Oulu (Thule Institute). This is a contribution from the National Doctoral Program in Integrated Catchment and Water Resources Management (VALUE).

Ancillary