Metals (e.g., Ni, Cu, Zn) are one of the most common sources of impairment in freshwater ecosystems (U.S. EPA, 2013), yet the chemical reactivity of most metals is a challenge for regulation and remediation. Multi-ecosystem studies have established that there is variability in the concentration at which components of the ecosystem respond to a contaminant (e.g., Di Toro et al., 1992; Iwasaki et al., 2011; Schmidt et al., 2011), which is often related to measurable attributes of the ecosystem (Chapman and Wang, 1998; Clements and Rohr, 2009; Di Toro et al., 1992; Fleeger et al., 2003). Context dependency (i.e., non-random variation in dose response) has been linked to many potential drivers (Clements et al., 2012), including physicochemical characteristics that affect exposure (e.g., light, dissolved organic carbon (DOC), sediment texture) and ecological attributes (e.g., community structure, detoxification mechanisms, predator-prey interactions). A more mechanistic understanding of the ecosystem characteristics that lead to context dependency may promote flexible risk assessment that is more efficient and appropriate for the regulatory processes and remediation.
Due to their strong interaction with other biogeochemical cycles (see review by Chapman and Wang, 1998), the speciation and toxicity of metals is strongly modified by physicochemical conditions. In freshwater ecosystems the majority of metal will be stored in the sediments with only a small fraction partitioned to surface waters (log Kd 3–5 L kg-1) (Allison and Allison, 2005; Burton, 2010). Within sediments, metals are bound or adsorbed to various solid-phase minerals (e.g., organic carbon, metal sulfides, ferric oxides) or dissolved in porewater depending on the pH and redox conditions (Chapman and Wang, 1998; Costello et al., 2011; Di Toro et al., 1992; U.S. EPA, 2005). The speciation of metals within sediments and surface waters has strong control over potential exposure to organisms (i.e., bioavailability) because only freely dissolved metal ions (e.g., Cu2+) interact with biological receptors and cause toxicity (Di Toro et al., 2001; Santore et al., 2001; U.S. EPA, 2005). For risk assessment of a specific ecosystem, a prediction of impairment due to metals requires knowledge of the physicochemical context to establish an accurate estimate of the bioavailable fraction of metal. Currently, site-specific bioavailability models for sediment metals include only a subset of the physicochemical parameters that can modify metal bioavailability (i.e., acid volatile sulfide (AVS) and organic carbon) (Ankley et al., 1996; U.S. EPA, 2005). These bioavailability models have been used in the formulation of chemical criteria that state that no adverse effects are expected when concentrations of AVS exceed concentrations of simultaneously extracted metal (SEM) on a molar basis (Ankley et al., 1996; U.S. EPA, 2005).
While the importance of physicochemical context for metals has been well studied, the role of ecological context has received less attention. The most conspicuous example of ecological context is the differential sensitivity of species to sediment metals (Burton, 1991; Milani et al., 2003). However, not all variation in community response can be explained by species sensitivity, likely due to emergent properties and indirect effects that arise when species interact (Brix et al., 2011; Clements et al., 2012; Fleeger et al., 2003). Further, ecological context may interact with physicochemical attributes because ecological traits (e.g., burrowing, photosynthesis) may modify the physicochemical environment (Chapman and Wang, 1998; Morris and Meyer, 2006). The importance of ecological context in sediment metal response emphasizes the need for in situ studies of entire communities with measurement of multiple endpoints representing both structural and functional attributes of the community (Burton, 1991; Clements et al., 2012). Importantly, attached algae and microbes (i.e., biofilms), which regulate most stream ecosystem functions (Gibbons et al., 2014; Pusch et al., 1998), are relatively understudied when compared to higher consumers such as macroinvertebrates and fishes; further, biofilm (or component species) response to sediment metal has only been marginally considered in the establishment of risk assessment and regulatory criteria for stream ecosystems. The application of ecological context and the inclusion of a wider variety of endpoints in setting chemical criteria would likely lead to more accurate and appropriate risk assessment for protecting freshwater ecosystems.
Our study examined the role of physicochemical and ecological context in toxicological responses in streams near the newly constructed Eagle Mine in the Upper Peninsula of Michigan, United States. This mine is targeting a large Ni and Cu ore body in a relatively undeveloped region that is near the headwaters of many ecologically diverse streams. The biological and physicochemical diversity of the streams in the region provide an opportunity to explore how environmental factors may control the sensitivity of the ecosystem to potential metal contamination. We amended two geochemically different sediments with Ni and Cu to simulate the outcome of potential contamination from this mine’s operation. Further, we studied dose-response relationships from these two sediments (physicochemical context) in two streams to explore how ecological context would influence the sensitivity of these communities to sediment metals. We measured a number of structural and functional endpoints in the epibenthic community to establish if the dose-response relationships were consistent for different components of the community. Our hypotheses included: (1) context dependency (both physicochemical and ecological) in response to sediment metal will be observed for many functional and structural attributes of the epibenthic community, (2) community attributes will not respond equally to sediment metals, and (3) current bioavailability models, which do not account for all context dependency will not be sufficiently protective of all community attributes.
This experiment was designed as a 2×2 factorial experiment to establish if differences in sediment type (physicochemical context) and/or stream location (ecological context) alter toxic response to sediment metals. For each sediment-type × stream-location combination, we deployed sediments with a range of metal concentrations to study dose-response for a variety of functional and structural endpoints. Using two exposure chamber styles, we measured multiple endpoints, including: periphyton function and standing stock, sediment microbial decomposition, grazing rate, and invertebrate abundance and community composition. Experiments and sediment collection were done in the Upper Peninsula of Michigan within the Dead-Kelsey Watershed (Marquette, MI, USA).
Depositional sediments were collected from Alder Creek (46° 47.49N, 87° 42.12W) and Salmon-Trout River (46° 47.30N, 87° 52.91W). Sediments were returned to the lab where they were characterized based on acid volatile sulfide (AVS; Allen et al., 1991) and organic matter (OM) content (loss-on-ignition at 550°C for >4 hr) and amended with Ni and Cu. Simultaneously extracted metals (SEM), which are acid-extracted metals liberated during the AVS procedure, indicated that these two sediments contained Ni and Cu at background concentrations (<10 mg kg-1 dw). Sediments from the Salmon-Trout R. had higher concentrations of both AVS and OM (Table 1), and treatments using these sediments will be referred to as high binding capacity (HB). Sediments from Alder Creek, which have lower AVS and OM, are referred to as low binding capacity (LB) sediments. Metal amendments were done in HDPE bottles using the indirect spiking method with NaOH pH buffering (Brumbaugh et al., 2013; Costello et al., 2011; Simpson et al., 2004). Spiked and buffered sediments were equilibrated for 1 week (rolled every 2 days) prior to dilution with reference sediments. For both the LB and HB sediment, the spiked sediment was diluted with reference sediment to establish 3 metal-amended treatments plus a non-amended control sediment (i.e., 8 treatments total) (Table 1). After dilution, the sediments were equilibrated for three weeks by rolling for 1 h weekly and subsequently purging the headspace with N2. Initial Ni and Cu concentrations were determined by microwave assisted acid digestion (3:1 HNO3:HCl) followed by inductively coupled plasma optical emission spectroscopy (ICP-OES), and these treatments spanned a range of non-toxic to potentially toxic concentrations (MacDonald et al., 2000; U.S. EPA, 2005) (Table 1).
(mg kg-1 dw)
(mg kg-1 dw)
Sediments were deployed in two stream reaches within Big Pup Creek and Salmon-Trout River and incubated for 4 weeks. At Big Pup Creek (BP), the experiment was conducted in a third-order reach with an open canopy (46° 42.65N, 87° 42.31W). At Salmon-Trout River (ST), sediments were placed in a second-order reach with a closed canopy (46° 46.96N, 87° 52.63W). Throughout the exposure time, water chemistry was monitored with weekly grab samples (n = 4 per reach) and hourly measurements by a multi-parameter sonde (YSI 6290). Weekly grab samples were analyzed for dissolved organic carbon (DOC), total nitrogen (TN), phosphate (PO43-), and major cations (Ca, Mg, Na, K). Sondes continuously (hourly) measured temperature, dissolved oxygen (DO), pH, conductivity (BP only), and turbidity.
Sediments were placed into two styles of exposure chambers: colonization baskets and chemical exposure substrate (CES) cups (Fig. S1). Colonization baskets were used to assess the macroinvertebrate community and decomposition response to sediment metal and CES cups were used to assess the biofilm (function and structure) and grazing response of snails to sediment metal. Colonization baskets consisted of mesh-lined shallow plastic trays (190 cm2) that were filled with sediment, placed flush with the streambed, and covered with a coarse-mesh (0.5 cm) bag (Burton et al., 2005; Costello et al., 2011). In each colonization basket, we placed a single cotton strip to estimate sediment microbial decomposition (Tiegs et al., 2013, 2007). Cotton strips were placed vertically into the sediment so each strip was exposed to both oxic surface and anoxic deeper sediments. CES cups consisted of a 30-mL HDPE hinge-top cup with a 2.2 cm diameter hole in the cap and a fritted glass crucible cover for biofilm attachment (Costello et al., in review). The cups were filled with sediment, capped to hold the attachment substrate at the surface, and secured to the stream bottom. For a subset of CES cups, we initiated biofilm colonization in a recirculating stream seeded with a mixed biofilm community from the Huron River (Ann Arbor, MI, USA). However, these pre-colonized cups started with a very low algal biomass (<1 g chlorophyll a cm-2) and exhibited similar patterns to cups with clean substrates; therefore, we combined cups with pre-colonized and clean substrates for our analyses. For each sediment treatment, 9 CES cups and 3 colonization baskets were placed in each stream for a total of 72 CES cups and 24 colonization baskets in each stream.
After 4 weeks, exposure chambers were removed from the stream and structural and functional endpoints were assessed (Table 2). Colonization baskets were carefully removed from the stream and 2/3 of each basket was placed into individual 1-L HDPE bottles with 90% ethanol for macroinvertebrate preservation. Preserved macroinvertebrates were separated from the sediment by sieving (45 µm), identified to family level, and enumerated. The invertebrate community composition was summarized with relevant diversity and abundance indices: total abundance, total richness, Simpson diversity, EPT abundance, and relative abundance of dominant taxa (i.e., Chironomidae). The remaining 1/3 (containing the cotton strip) from the three replicate baskets were combined, gently homogenized, and frozen for chemical analysis (total Ni, total Cu, AVS, and SEM). Cotton strips were carefully removed from the sediment, gently agitated in the stream to remove sediment particles, and placed in a 50-mL centrifuge tube filled with site water (no headspace, DO measured) for a respiration assay (Tiegs et al., 2013). Tubes containing cotton strips and site water were placed in an opaque plastic bag within the stream to incubate the chambers at stream temperature. After 2-4 hours of incubation, the centrifuge tubes were removed and DO was measured. At each stream, two tubes containing stream water only were used to correct for background respiration from suspended particles. Cotton strip respiration (Rcot in mg O2 L-1 h-1) was calculated as:
|Biofilm structure||Chlorophyll a (chla), Ni content, Cu content|
|Biofilm function||Net primary productivity (NPP)|
|Invertebrate structure||Richness, abundance, Ephemeroptera Plecoptera Tricoptera (EPT) abundance, relative abundance of Chironomids (% dominant), Simpson diversity index|
|Invertebrate function||Lymnaea stagnalis feeding rate|
|Sediment microbial function||Cotton respiration, decomposition of cotton|
where DOi are DO concentrations (mg L-1) from the stream at the beginning of incubation (DOstream), in the centrifuge tube following cotton incubation (DOcot), and average DO in the tubes incubated with site water only (DOctrl) and ti are times (h) for the incubation of cotton strips (tcot) and site water (tctrl). Following the respiration assay, strips were soaked in 90% ethanol to arrest microbial activity and then air-dried. Decomposition of preserved cotton strips was measured as loss of tensile strength using a tensiometer on a motorized stand (Tiegs et al., 2013). Initial tensile strength was determined from a mean of three pre-incubation strips.
CES cups were removed from the stream and the biofilm attachment substrate was separated from the sediment. Cups containing sediment were covered with parafilm and frozen for chemical analysis (total Ni, total Cu, AVS, and SEM). Glass disks containing biofilms were gently agitated in the stream to remove any attached sediment particles. Six biofilm disks for each sediment treatment were placed into transparent 160-mL specimen cups with site water (no headspace, DO measured) and incubated in direct sunlight in the stream for 3-5 hours. Following incubation, specimen cups were recovered and dissolved oxygen was measured to estimate net primary production (NPP). At each stream, three specimen cups containing stream water only were used to correct for background NPP. NPP (mg O2 m-2 h-1) was calculated as:
where DOi are DO concentrations (mg L-1) from the stream at the beginning of incubation (DOstream), in the specimen cup following biofilm incubation (DObiofilm), and average DO in the cups incubated with site water only (DOctrl), ti are times (h) for the incubation of biofilm disks (tbiofilm) and site water (tctrl), V is the volume (L) of water in the specimen cups, and A is the surface area of the disk exposed to the stream (m2). For grazing trials, the remaining three biofilm disks for each sediment treatment were placed in 50-mL centrifuge tubes with site water and 4 Lymnaea stagnalis snails (7–14 d old). Snails were allowed to feed for 24 hour before being removed from biofilm disks. Following NPP incubations and grazing trials, biofilm disks were frozen for analysis of biomass, dry mass, and metal (Cu and Ni) content (NPP disks only). Biofilm biomass was estimated by extracting chlorophyll a (chla) with boiling 90% ethanol (Biggs and Kilroy, 2000) and measuring fluorescence. Following chla measurement, ethanol was evaporated, the solid material was weighed and digested with microwave-assisted acid digestion (3:1 HNO3:HCl). Digestates were analyzed for Cu and Ni using ICP-OES. Digested disks were reweighed for measurement of biofilm dry mass.
An analysis of covariance (ANCOVA) approach was used to relate endpoint responses to total sediment metal content, stream location, and sediment type. Total sediment metal content (sum of Ni and Cu molar concentrations) was used as our estimate of metal dose because bioavailability models assume that the applicable metals interact additively (U.S. EPA, 2005). The majority of the total sediment metal was present in the SEM fraction (>90%), which suggested that these metals were potentially bioavailable (Ankley et al., 1996; U.S. EPA, 2005). For simplicity, we discuss changes in response to total metals only; however, identical dose-response patterns were observed in response to total SEM. Significant interactions between sediment metal and stream location and/or sediment type indicate that the functional response between the endpoint and the metal concentration is context dependent. For snail feeding assays, an additional factor (snail presence or absence) was included in our ANCOVA models; significant interactions that included the snail presence/absence factor indicate that the other variables in the interaction significantly altered snail feeding (Peterson and Renaud, 1989). Relatively little variation in metal concentrations among replicate CES cups and baskets allowed us to use total metal concentrations from a single randomly selected CES cup per treatment combination collected at day 28 for all statistical analysis. Further, this allowed us to use lack-of-fit tests to verify that linear ANCOVA models are most appropriate for the given data (Cottingham et al., 2005). Lack-of-fit tests indicated that biofilm response variables (chla, NPP, metal content, snail feeding) responded to the log of the sediment metal concentration, whereas all other endpoints responded linearly. Assumptions of normality and equal variance were verified using Shapiro-Wilks tests and residual plots, respectively. For models that violated assumptions of linear models, we used log (biofilm metal content, cotton respiration) or arcsine square root (% dominant) transformations to meet assumptions. A small number of CES cups were buried or lost and a single invertebrate sample was lost due to poor preservation; we removed these observations from our statistical analysis. Although our experiment was ultimately unbalanced, all of our treatments were represented in the final analysis and our design remained fully-crossed. All statistics were conducted using R 3.0.1 (R Core Team, 2013).
After 28 days of incubation, sediment metal concentrations declined; losses of Cu and Ni in LB sediments (mean loss of 43 and 52%, respectively) were greater than losses of Cu and Ni from HB sediments (mean loss of 4% and 26%, respectively). The decline in sediment metal, in particular Ni, is likely due to diffusive loss, which is common when spiked sediments are equilibrated for short time periods (Simpson et al., 2004). Although substantial amounts of sediment metal were lost, at least 3 of 4 treatments for each sediment remained above non-toxic thresholds for at least one metal (MacDonald et al., 2000; U.S. EPA, 2005). AVS was negatively correlated to sediment Cu concentrations (p < 0.05) due to the insolubility of copper sulfide during the AVS extraction procedure (Simpson et al., 1998). Though there were physicochemical changes to the sediments during the incubation, HB sediments maintained relatively higher concentrations of important metal binding ligands (AVS and OM) than LB sediments. Our two streams did not differ greatly in the water chemistry parameters that modify metal toxicity (Table 3); therefore, any significant stream x metal interactions are likely a result of ecological not physicochemical context. As expected, the metal concentrations we used induced toxic responses in the stream ecosystem; 7 of the 12 measured endpoints exhibited significant responses to increasing concentrations of sediment metal (Table 4). As expected, most endpoints were negatively related to increasing concentrations of sediment metal, with the exception of biofilm metal content, which was positively related to sediment metal concentrations. As we hypothesized, the response to increasing concentrations of sediment metal was not equal across streams, sediment types, or endpoints.
|TOC (mg L-1)a||7.75 (0.63)||9.96 (0.63)|
|Total N (mg L-1)a||0.26 (0.02)||0.20 (0.02)|
|PO43- (µg L-1)a||3.52 (0.46)||6.56 (2.94)|
|Ca (mg L-1)a||11.83 (1.76)||9.48 (3.38)|
|K (mg L-1)a||1.47 (0.57)||1.52 (0.30)|
|Mg (mg L-1)a||2.61 (0.19)||2.48 (0.13)|
|Na (mg L-1)a||1.32 (0.07)||1.32 (0.04)|
|Temperature (°C)b||14.4 (1.4)||15.1 (1.4)|
|pHb||7.69 (0.09)||7.75 (0.07)|
|DO (% saturation)b||98.3 (2.4)||99.1 (0.8)|
|Conductivity (µS cm-1)b||93.2 (5.1)||88.4 (4.0)|
|Turbidity (NTU)b||1.3 (4.0)||3.6 (2.5)|
For the endpoints that responded to sediment metal, we found that six of the seven responses exhibited a significant interaction with sediment type and/or stream location, which indicates that the responses were context dependent (Table 4). Physicochemical context affected the response of the microbial community, ecological context modified the response of the biofilm community (chla and biofilm Ni content), and both factors influenced biofilm primary production and EPT abundance (Table 4). Biofilm NPP was negatively affected by sediment metal only in Salmon-Trout River and, at that location, only when exposed to metals from HB sediments (Fig. 1A). Biofilm chla was negatively affected by sediment metal only in Salmon-Trout River, but biofilm biomass on LB and HB sediments both responded similarly (Fig. 1B). All biofilms in Big Pup Creek showed no response to sediment metals regardless of exposure to LB or HB sediments (Fig. 1A & B). Biofilm Ni content exhibited a weak interaction (p = 0.03) between sediment metal and stream location with less Ni accumulation in biofilms from the Salmon-Trout River (Table 4). However, biofilm Cu content did not exhibit any context dependency (Fig. 1D). In both streams, the dose response of cotton respiration was modified by physicochemical context, with a strong negative response for cotton in LB sediment and no response in HB sediment (Fig. 2A, Table 4). Similarly, cumulative cotton decomposition over the 28-d exposure period (measured as loss of tensile strength) showed similar context dependency (Fig. 2B, Table 4). EPT abundance declined with increasing metal concentration when exposed to LB sediment in Big Pup Creek; we observed no response of EPT to increasing sediment metal for any exposures in Salmon Trout River and any exposures to HB sediments (Fig. 3B). Overall, EPT abundances were much lower in Salmon Trout River (mean = 600 m-2) than Big Pup Creek (mean = 1700 m-2), which was likely responsible for the lack of response to sediment metal observed in Salmon Trout River and the cause for the interaction between physicochemical and ecological context (i.e., significant metal x sediment x stream interaction).
|Metal x sed. x
|NPP||93||< 0.001||< 0.001||0.001||0.002|
|Biofilm Cu||89||< 0.001||0.44||0.91||0.76|
|Cotton resp.||47||0.57||< 0.001||0.11||0.13|
|Cotton decomp.||48||0.052||< 0.001||0.7||0.10|
We observed strong differences in the dose response among our functional and structural endpoints. The biofilm and microbial communities (6 of 6 endpoints) responded much more strongly to sediment metal than the consumer community (1 of 6 endpoints). Although all measured attributes of the biofilm and microbial community responded to sediment metal, the context dependency differed among these communities; the biofilm community response differed between stream (i.e., ecological context) (Fig. 1) and the microbial community response differed between sediment types (i.e., physicochemical context) (Fig. 2). For our six measures of the consumer community, only a single metric (EPT abundance) was significantly affected by sediment metal (Table 4, Fig 3A). Our snail feeding rates were highly variable and exhibited no response to sediment metal (Fig. 3A, Table 4). There was a trend of lower feeding rates for snails exposed to metal-contaminated LB sediments in Big Pup (Fig. 3A), but high variability in controls and low statistical power caused this pattern to be statistically insignificant. The four other invertebrate community metrics (total richness, total abundance, Simpson diversity, % dominant) varied among streams but did not show any response to increasing sediment metal concentrations (Table 4).
The current bioavailability models predict that our spiked sediments will be non-toxic when metal concentrations in LB and HB are below 0.8 and 2.5 µmol g-1, respectively. The non-toxic threshold accurately predicted non-toxic conditions for the microbial and invertebrate community, but in the biofilm community toxicity was observed at concentrations below the non-toxic thresholds. Our dose-response function predict a 10% reduction in NPP for biofilm in the Salmon-Trout River at just 0.31 µmol metal g-1 if growing on the HB sediment (Fig. 1A). The dose-response models predict a 10% reduction in chla at 0.21 and 0.47 µmol metal g-1 for biofilms growing on LB and HB sediment, respectively. The modeled 10% reductions in biofilm function and structure, which occur below the non-toxic thresholds, are observed only in Salmon-Trout River as toxicity was not observed in Big Pup Creek. For the microbial community in the LB sediment regardless of stream location, our dose-response function predicts a 10% reduction in Rcot and tensile strength at 1.2 and 1.7 µmol metal g-1 dw, respectively, which is at a concentration that exceeds the non-toxic threshold. We observed a 10% reduction in EPT abundance in Big Pup Creek when sediment total metal concentrations were 1.1 µmol g-1, which is in excess of the AVS concentration (0.8 µmol g-1). For both the microbial community and EPT abundance, there was no response to metal in the HB sediment even though the metal concentrations exceeded the non-toxic threshold.
Our study has demonstrated that understanding whole-ecosystem responses to sediment metal benefits from examining many components of the community and considering the importance of physicochemical and ecological context. The frequency at which we observed context dependency adds to the growing evidence (Brix et al., 2011; Roman et al., 2007) that risk assessments that do not incorporate context-dependency (e.g., MacDonald et al., 2000) are inappropriate except for preliminary screening of ecosystems. Further, not all components of the community responded equally to metal, nor were their responses modified by the same contextual factors. For restoration or regulatory goals intended to protect entire ecosystems, careful study of the functional and structural attributes of communities is required for comprehensive protection. Broadly, a greater mechanistic understanding of the biogeochemical and ecological interactions that manifest as context dependency (e.g., Clements et al., 2012) would improve our ability to predict impairment, restore ecosystems, and understand how anthropogenic activity is modifying the structure and functioning of freshwater ecosystems.
The divergent dose-response relationships we observed on our two sediment types emphasizes the strong role sediment geochemistry plays in metal cycling and exposure (Burton, 2010; Chapman and Wang, 1998; U.S. EPA, 2005). For our measures of the macroinvertebrate community (i.e., EPT abundance) and sediment microbial function, we observed negative effects of metals only on the LB sediments and only when metal concentrations exceeded AVS concentration, which supports the context-dependent bioavailability models (Ankley et al., 1996; U.S. EPA, 2005). The absence of a toxicological response for macroinvertebrates and sediment microbes exposed to the HB sediment is somewhat surprising considering the amended metal concentrations greatly exceeded AVS binding capacity. However, metals in excess of AVS and OM can still be unavailable to organisms due to complexation with other ligands (e.g., Fe oxides), microhabitats with physicochemical conditions different from bulk sediment, or physical separation of metals from organisms (Ankley et al., 1996; Costello et al., 2011; Di Toro et al., 1992; U.S. EPA, 2005). The sediment context-dependency of the biofilm response differed substantially from the sediment microbial and invertebrate response with either no evidence of sediment context dependency (chla) or the HB sediment eliciting an effect at lower concentrations than the LB sediment (NPP). This deviation in context dependency between the biofilm response and the microbial and invertebrate responses suggests that models of metal bioavailability may not be equally applicable to all members of a community.
We observed evidence of ecological context dependency for invertebrates (i.e., EPT abundance) and many of the measures of the biofilm community. The context dependency observed for EPT abundance likely results from community differences between the two streams; the invertebrate community Big Pup Creek had a greater abundance of EPT than Salmon-Trout River and a higher relative abundance of mayflies, specifically from the families Baetidae and Ephemerellidae. The sediments placed in Salmon-Trout River were colonized by some mayflies (i.e., Ephemerellidae), but stoneflies (i.e., Leuctridae) and caddisflies (i.e., Molannidae) were a greater proportion of the EPT abundance. The greater sensitivity to sediment metals of EPT in Big Pup Creek matches expectations of interspecies variability in metal tolerance, which suggests that taxa in the order Ephemeroptera are highly sensitive to metals (Carlisle and Clements, 2009; Clements et al., 2000; U.S. EPA, 1996). Additionally, the stronger response of EPT abundance in Big Pup Creek may have resulted from elevated dietary metal exposure; in Big Pup Creek, a greater proportion of sediment metal (particularly Ni) was found in the biofilm matrix when compared to Salmon-Trout River. However, due to the experimental design (i.e., separate biofilm and invertebrate exposure chambers) and the minimal biofilm colonization on the colonization trays, we do not believe that this is the primary mechanism for the differences in EPT response between Big Pup Creek and Salmon-Trout River. Nonetheless, dietary metal exposure has been implicated in chronic invertebrate responses at very low metal concentrations (Brix et al., 2011), and any invertebrates feeding on biofilms growing on metal-contaminated sediments do have the potential to be exposed to metals.
The context-dependency of the biofilm response differed greatly from expectations of models of metal bioavailability (Ankley et al., 1996; U.S. EPA, 2005). For biofilm function (i.e., NPP) we observed differences between our sediment types but the relationship was opposite of expectations (i.e., stronger dose-response from sediments with greater content of binding ligands). In addition to the observed physicochemical context dependency, we also observed strong ecological context dependency for the biofilm dose-response. Of greater significance, we observed negative effects of sediment metals (i.e., 10% reduction in NPP) at concentrations below the well-established non-toxic thresholds (U.S. EPA, 2005). We did not identify the taxa comprising the biofilm in our study, thus we cannot make mechanistic inferences about the role of biofilm community composition on sediment metal sensitivity. However, it is probable that our streams had dissimilar biofilm communities because the two stream reaches differed greatly in canopy cover, which can alter biofilm structure (Johnson et al., 2009). In addition to biofilm responses deviating from existing bioavailability models, we observed differences between functional and structural endpoints (i.e., in the Salmon-Trout River, biofilms on HB and LB showed reduced chla whereas NPP was reduced only for biofilms on HB). It has been demonstrated that biofilm structural and functional responses are not equally sensitive to ecosystem stress; functional endpoints only respond at the highest levels of stress whereas structure responds at low levels of stress (Crossey and La Point, 1988; Niyogi et al., 2002). However, for the treatment that did show a reduction in biofilm function (i.e., HB in Salmon-Trout), we observed a dose response at concentrations that were comparable to concentrations eliciting a structural response. In all, we observed impairment of biofilm function and structure at concentrations below non-toxic thresholds and the context dependency for biofilms does not match context dependency for invertebrates and fishes. This suggests that, although bioavailability models do account for certain aspects of physicochemical context (i.e., AVS), these models may not be protective for biofilms.
The disconnect between our observed biofilm responses and existing bioavailability models may be a result of non-equilibrium metal dynamics in sediments, primarily oxidation of surface sediments by biofilms. Biofilms are not passive organisms solely responding to sediment metals; rather their metabolic processes actively alter biogeochemical cycles, which can modify metal speciation and exposure. Photosynthetic activity by biofilms increases pH and O2 concentrations in overlying waters and surface sediments, which has been implicated in diel cycles of metals (i.e., water column metals decline during the day) observed in some contaminated systems (Beck et al., 2009; Morris et al., 2005). Oxidation of sediments can cleave metal-sulfide bonds and mobilize metals (De Jonge et al., 2012a; Simpson et al., 2012). These mobilized metals can either sorb to other metal-binding ligands (e.g., Fe oxides; De Jonge et al., 2012a) or in the absence of alternative binding ligands remain dissolved and elicit a toxic response (De Jonge et al., 2012b; Simpson et al., 2012). For our experiment, we observed greater photosynthetic activity (i.e., higher background NPP) and greater diel O2 fluctuations (i.e., DO standard deviation was 3x greater) in the open-canopy Big Pup Creek. The increased O2 production in Big Pup Creek during the day supports greater sediment oxidizing potential, which could lead to mobilization of sulfide-bound metals. The biofilm response to low concentrations of sediment metal in Big Pup Creek may be a result of greater exposure to metals released from oxidized metal sulfides. The instability of metal sulfides under oxidizing conditions is likely responsible for the inability of sulfide-based bioavailability models to predict non-toxic conditions in Big Pup Creek. Sediment bioavailability models assume that metal speciation is stable and dissolved and particulate metals are in equilibrium (Ankley et al., 1996; U.S. EPA, 2005); however, our data and other studies of processes that oxidize sediments (e.g., bioturbation and resuspension; Peterson et al., 1996; Simpson et al., 1998) reveal that non-equilibrium dynamics can be common and need to be considered in sediment risk assessment. Integration of sediment oxidation processes into sediment risk assessment is further complicated by the spatial scale at which these mechanisms operate, which is often within the top few millimeters of sediment. Additional research on spatial and temporal variability in sulfide and metal speciation could greatly improve our ability to predict metal bioavailability in dynamic and heterogeneous environments.
The discrepancy in responses between invertebrates, sediment microbes, and biofilms suggests that even if we include context dependency in bioavailability models, a universal model applicable to all organisms may not be feasible. For sediment microbes, AVS-based bioavailability models offer good predictions of non-toxic conditions. Similarly, for invertebrates AVS-based models predict non-toxic responses well, but models could be improved by including community composition and tolerance as an additional contextual consideration. For biofilms, AVS-based measurements of bulk sediment are not sufficiently protective and new models incorporating non-equilibrium processes or more targeted sampling protocols (e.g., microsensors in surface sediment) would need to be developed. Ultimately, the difference in efficacy of these models may require prioritization of regulatory and restoration goals and selective application of an appropriate lens of context-dependency. For example, if a regulatory goal is protecting aquatic invertebrate biodiversity across a broad geographic range, then these data and other studies (Burton et al., 2005; Costello et al., 2011; Di Toro et al., 1992) suggest that current no-effects thresholds that include sediment context (i.e., AVS) offer an appropriate screening-level approach. Alternatively, if a single ecosystem were being restored to improve biodiversity, it would be constructive to incorporate community context by determining the sensitivity of the anticipated community under reference conditions. If protection or restoration of ecosystem functions is a goal (Millennium Ecosystem Assessment, 2005), then the current bioavailability models may not be adequately protective of biofilms, which control most stream ecosystem functions (Gibbons et al., 2014; Pusch et al., 1998). Our study of two ecosystems is not sufficient for developing widely-applicable sediment metal criteria for biofilms, yet these results should stimulate research on the interactions between biofilm and metal cycles. In all, these data suggest that context dependency can be used to improve metal sediment risk assessment, but the application of contextual filters must be judicious as models may not be equally applied to all constituents of a community.
The following datasets were generated and will be archived at the Dryad Digital Repository: http://doi.org/10.5061/dryad.7sd19
© 2014 Costello and Burton. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Contributed to conception and design: DMC, GAB
Contributed to acquisition of data: DMC
Contributed to analysis and interpretation of data: DMC
Drafted and/or revised article: DMC, GAB
Approved the submitted version for publication: DMC, GAB
The authors do not have any competing interests that may influence the interpretation or presentation of this manuscript.
Funding for invertebrate taxonomy was provided by Rio Tinto. Kent State University Libraries provided funding for publication costs associated with this article.
We thank Michelle Sawyers, Stephanie Tubbs Aselage, Molly Harkness, Tom Aepelbacher, Kyle Fetters, Shehara Waas, Olivia Rath, Cori Cramer, Maggie Grundler, Meghan Myers, and Anna Harrison for field and lab assistance. Scott Tiegs provided equipment and expertise for the cotton strip decomposition assay. Stephen Hamilton and David Weed provided assistance with stream water chemistry. Jen Daley and anonymous reviewers provided useful comments on previous versions of this manuscript.
Allen HE, Fu G, Boothman W, Di Toro DM, Mahony JD. 1991. Determination of acid volatile sulfide and selected simultaneously extractable metals in sediment . Washington, DC: US Environmental Protection Agency.
Allison JD, Allison TL. 2005. Partitioning coefficients for metals in surface water, soil, and waste . Washington, DC: US Environmental Protection Agency.
Ankley GT, Di Toro DM, Hansen DJ, Berry WJ. 1996. Technical basis and proposal for deriving sediment quality criteria for metals. Environ Toxicol Chem 15: 2056–2066.
Beck AJ, Janssen F, Polerecky L, Herlory O, de Beer D. 2009. Phototrophic biofilm activity and dynamics of diurnal Cd cycling in a freshwater stream. Environ Sci Technol 43: 7245–7251.doi: 10.1021/es900247y
Biggs B, Kilroy C. 2000. Stream periphyton monitoring manual . Christchurch, NZ: NIWA for the New Zealand Ministry for the Environment.
Brix KV, DeForest DK, Adams WJ. 2011. The sensitivity of aquatic insects to divalent metals: a comparative analysis of laboratory and field data. Sci Total Environ 409: 4187–4197. doi: 10.1016/j.scitotenv.2011.06.061.
Brumbaugh WG , Besser JM , Ingersoll CG , May TW , Ivey CD , et al. 2013. Preparation and characterization of nickel-spiked freshwater sediments for toxicity tests: Toward more environmentally realistic nickel partitioning. Environ Toxicol Chem 32: 2482–2494.
Burton GA. 1991. Assessing the toxicity of freshwater sediments. Environ Toxicol Chem 10: 1585–1627.
Burton GA. 2010. Metal bioavailability and toxicity in sediments. Crit Rev Environ Sci Technol 40: 852–907. doi: 10.1080/10643380802501567
Burton GA , Nguyen LTH , Janssen C , Baudo R , McWilliam RA , et al. 2005. Field validation of sediment zinc toxicity. Environ Toxicol Chem 24: 541–553.
Carlisle DM, Clements WH. 2009. Sensitivity and variability of metrics used in biological assessments of running waters. Environ Toxicol Chem 18: 285–291.
Chapman P, Wang F. 1998. Ecotoxicology of metals in aquatic sediments: Binding and release, bioavailability, risk assessment, and remediation. Can J Fish Aquat Sci 55: 2221–2243. doi: 10.1139/f98-145
Clements WH, Carlisle DM, Lazorchak JM, Johnson PC. 2000. Heavy metals structure benthic communities in Colorado mountain streams. Ecol Appl 10: 626–638.
Clements WH, Hickey CW, Kidd KA. 2012. How do aquatic communities respond to contaminants? It depends on the ecological context. Environ Toxicol Chem 31: 1932–1940. doi: 10.1002/etc.1937
Clements WH, Rohr JR. 2009. Community responses to contaminants: Using basic ecological principles to predict ecotoxicological effects. Environ Toxicol Chem 28: 1789–1800.
Costello DM, Rosi-Marhsall EJ, Shaw LE, Grace MR, Kelly JJ. nd. A novel method for measuring the effects of chemical stressors on biofilm structure and function. In review.
Costello DM, Burton GA, Hammerschmidt CR, Rogevich EC, Schlekat CE. 2011. Nickel phase partitioning and toxicity in field-deployed sediments. Environ Sci Technol 45: 5798–5805. doi: 10.1021/es104373h
Cottingham KL, Lennon JT, Brown BL. 2005. Knowing when to draw the line: Designing more informative ecological experiments. Front Ecol Environ 3: 145–152.
Crossey M, La Point T. 1988. A comparison of periphyton community structural and functional responses to heavy metals. Hydrobiologia 162: 109–121.
De Jonge M, Teuchies J, Meire P, Blust R, Bervoets L. 2012a. The impact of increased oxygen conditions on metal-contaminated sediments part I: Effects on redox status, sediment geochemistry and metal bioavailability. Water Res 46: 2205-2214. doi: 10.1016/j.watres.2012.01.052
De Jonge M, Teuchies J, Meire P, Blust R, Bervoets L. 2012b. The impact of increased oxygen conditions on metal-contaminated sediments part II: Effects on metal accumulation and toxicity in aquatic invertebrates. Water Res 46: 3387–3397. doi: 10.1016/j.watres.2012.03.035
Di Toro DM , Allen HE , Bergman HL , Meyer JS , Paquin PR , et al. 2001. Biotic ligand model of the acute toxicity of metals. 1. Technical basis. Environ Toxicol Chem 20: 2383–2396.
Di Toro DM , Mahony JD , Hansen DJ , Scott KJ , Carlson AR , et al. 1992. Acid volatile sulfide predicts the acute toxicity of cadmium and nickel in sediments. Environ Sci Technol 26: 96–101.
Fleeger JW, Carman KR, Nisbet RM. 2003. Indirect effects of contaminants in aquatic ecosystems. Sci Total Environ 317: 207–233. doi: 10.1016/S0048-9697(03)00141-4
Gibbons SM , Jones E , Bearquiver A , Blackwolf F , Roundstone W , et al. 2014. Human and environmental impacts on river sediment microbial communities. PLoS One 9: e97435. doi: 10.1371/journal.pone.0097435
Iwasaki Y, Kagaya T, Miyamoto KI, Matsuda H, Sakakibara M. 2011. Effect of zinc on diversity of riverine benthic macroinvertebrates: estimation of safe concentrations from field data. Environ Toxicol Chem 30: 2237–2243. doi: 10.1002/etc.612
Johnson LT, Tank JL, Dodds WK. 2009. The influence of land use on stream biofilm nutrient limitation across eight North American ecoregions. Can J Fish Aquat Sci 66: 1081–1094. doi: 10.1139/F09-065
MacDonald DD, Ingersoll CG, Berger TA. 2000. Development and evaluation of consensus-based sediment quality guidelines for freshwater ecosystems. Arch Environ Contam Toxicol 39: 20–31. doi: 10.1007/s002440010075
Milani D, Reynoldson TB, Borgmann U, Kolasa J, 2003. The relative sensitivity of four benthic invertebrates to metals in spiked-sediment exposures and application to contaminated field sediment. Environ Toxicol Chem 22, 845–854.
Millennium Ecosystem Assessment. 2005. Ecosystems and Human Well-being: Synthesis . Washington, DC: Island Press.
Morris JM, Meyer JS. 2006. Photosynthetically Mediated Zn Removal from the Water Column in High Ore Creek, Montana. Water Air Soil Pollut 179: 391–395. doi: 10.1007/s11270-006-9232-9
Morris JM, Nimick DA, Farag AM, Meyer JS. 2005. Does biofilm contribute to diel cycling of Zn in High Ore Creek, Montana? Biogeochemistry 76: 233–259.
Niyogi DK, Lewis WM, McKnight DM. 2002. Effects of stress from mine drainage on diversity, biomass, and function of primary producers in mountain streams. Ecosystems 5: 554–567. doi: 10.1007/s10021-002-0182-9
Peterson CH, Renaud PE. 1989. Analysis of feeding preference experiments. Oecologia 80: 82–86.
Peterson GS, Ankley GT, Leonard EN. 1996. Effect of bioturbation on metal-sulfide oxidation in surficial freshwater sediments. Environ Toxicol Chem 15: 2147–2155.
Pusch M , Fiebig D , Brettar I , Eisenmann H , Ellis BK , et al. 1998. The role of micro-organisms in the ecological connectivity of running waters. Freshw Biol 40: 453–495.
R Core Team. 2013. R: A language and environment for statistical computing . R Development Foundation for Statistical Computing.
Roman YE, De Schamphelaere KAC, Nguyen LTH, Janssen CR. 2007. Chronic toxicity of copper to five benthic invertebrates in laboratory-formulated sediment: Sensitivity comparison and preliminary risk assessment. Sci Total Environ 387: 128–140.
Santore RC, Di Toro DM, Paquin PR, Allen HE, Meyer JS. 2001. Biotic ligand model of the acute toxicity of metals. 2. Application to acute copper toxicity in freshwater fish and Daphnia. Environ Toxicol Chem 20: 2397–2402.
Schmidt TS , Clements WH , Zuellig RE , Mitchell KA , Church SE , et al. 2011. Critical tissue residue approach linking accumulated metals in aquatic insects to population and community-level effects. Environ Sci Technol 45: 7004–7010. doi: 10.1021/es200215s
Simpson SL, Angel BM, Jolley DF. 2004. Metal equilibration in laboratory-contaminated (spiked) sediments used for the development of whole-sediment toxicity tests. Chemosphere 54: 597–609. doi: 10.1016/j.chemosphere.2003.08.007
Simpson SL, Apte SC, Batley GE, 1998. Effect of short-term resuspension events on trace metal speciation in polluted anoxic sediments. Environ Sci Technol 32: 620–625. doi: 10.1021/es970568g
Simpson SL, Ward D, Strom D, Jolley DF. 2012. Oxidation of acid-volatile sulfide in surface sediments increases the release and toxicity of copper to the benthic amphipod Melita plumulosa. Chemosphere 88: 953–961. doi: 10.1016/j.chemosphere.2012.03.026
Tiegs SD, Clapcott JE, Griffiths NA, Boulton AJ, 2013. A standardized cotton-strip assay for measuring organic-matter decomposition in streams. Ecol Indic 32, 131–139. doi: 10.1016/j.ecolind.2013.03.013
Tiegs SD, Langhans SD, Tockner K, Gessner MO 2007. Cotton strips as a leaf surrogate to measure decomposition in river floodplain habitats. J North Am Benthol Soc 26: 70–77.
U.S. EPA. 1996. Water Quality Criteria Documents for the Protection of Aquatic Life in Ambient Water: 1995 Updates . Washington DC: United States Environmental Protection Agency.
U.S. EPA. 2005. Procedures for the derivation of equilibrium partitioning sediment benchmarks (ESBs) for the protection of benthic organisms: Metal mixtures (cadmium, copper, lead, nickel, silver and zinc) . Washington DC: United States Environmental Protection Agency.
U.S. EPA. 2013. The Watershed Assessment, Tracking & Environmental Results System (WATERS) . Washington DC: United States Environmental Protection Agency. Available at www.epa.gov/waters/ir. Accessed August 16, 2013.