G Model
ARTICLE IN PRESS
AQTOX-4150; No. of Pages 10
Aquatic Toxicology xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Aquatic Toxicology journal homepage: www.elsevier.com/locate/aquatox
Establishing mussel behavior as a biomarker in ecotoxicology Jason T. Hartmann a , Sebastian Beggel a , Karl Auerswald b , Bernhard C. Stoeckle a , Juergen Geist a,∗ a Aquatic Systems Biology Unit, Department of Ecology and Ecosystem Management, Technische Universitaet Muenchen, Muehlenweg 22, D-85350 Freising, Germany b Chair of Grassland Science, Department of Plant Science, Technische Universitaet Muenchen, Alte Akademie 12, D-85350 Freising, Germany
a r t i c l e
i n f o
Article history: Received 3 May 2015 Received in revised form 19 June 2015 Accepted 28 June 2015 Available online xxx Keywords: Bivalves Hall sensors Unionoida Sublethal endpoint Avoidance behavior NaCl
a b s t r a c t Most freshwater mussel species of the Unionoida are endangered, presenting a conservation issue as they are keystone species providing essential services for aquatic ecosystems. As filter feeders with limited mobility, mussels are highly susceptible to water pollution. Despite their exposure risk, mussels are underrepresented in standard ecotoxicological methods. This study aimed to demonstrate that mussel behavioral response to a chemical stressor is a suitable biomarker for the advancement of ecotoxicology methods that aids mussel conservation. Modern software and Hall sensor technology enabled mussel filtration behavior to be monitored real-time at very high resolution. With this technology, we present our method using Anodonta anatina and record their response to de-icing salt pollution. The experiment involved an environmentally relevant ‘pulse-exposure’ design simulating three subsequent inflow events. Three sublethal endpoints were investigated, Filtration Activity, Transition Frequency (number of changes from opened to closed, or vice versa) and Avoidance Behavior. The mussels presented a high variation in filtration behavior, behaving asynchronously. At environmentally relevant de-icing salt exposure scenarios, A. anatina behavior patterns were significantly affected. Treated mussels’ Filtration Activity decreased during periods of very high and long de-icing salt exposure (p < 0.001), however, increased during short de-icing salt exposure. Treated mussels’ Transition Frequency increased during periods of very high and long de-icing salt exposure (p < 0.001), which mirrored the Avoidance Behavior endpoint observed only by mussels under chemical stress. Characteristics of Avoidance Behavior were tighter shell closures with repeated and irregular shell movements which was significantly different to their undisturbed resting behavior (p < 0.001). Additionally, we found that mussels were sensitive to a chemical stressor even when the mussel’s valves were closed. Due to the effects of de-icing salt pollution on freshwater mussel behavior, we suggest better management practices for de-icing salt use be implemented. Our experimental method demonstrated that, with the application of current technologies, mussel behavioral response to a chemical stressor can be measured. The tested sublethal endpoints are suitable for mussel ecotoxicology studies. Avoidance Behavior proved to be a potentially suitable endpoint for calculating mussel behavior effect concentration. Therefore we recommend adult mussel behavior as a suitable biomarker for future ecotoxicological research. This method could be applied to other bivalve species and for physical and environmental stressors, such as particulate matter and temperature. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Pollution is a major driver towards the endangerment of freshwater bivalves (Unionoida) which are the fastest diminishing taxa with all of the European species listed on the International Union for Conservation of Nature’s (IUCN) ‘Red List’ (Bogan, 1993; Geist, 2011; IUCN, 2014). Unionoida conservation is a high priority. As keystone species they contribute to a number of essential functions
∗ Corresponding author. Fax: +49 8161 71 3477. E-mail address: [email protected] (J. Geist).
in aquatic ecosystems (Geist, 2010; Vorosmarty et al., 2010). Due to their high exposure pathway as filter feeders and their limited evasion capabilities, mussels are especially susceptible to organic and inorganic chemicals (Strayer et al., 2004). The number of registered chemical substances is over 100 million (CAS, 2015). This therefore presents a substantial risk to Unionoida (Dudgeon et al., 2006). The consequences of further decline in filter-feeder biodiversity is likely to have far reaching implications towards the failure of lotic and lentic ecosystem services (Cardinale et al., 2012; Chapin et al., 2000; Vaughn, 2010). In recognition of this, Eggen et al. (2004) appealed for the ecotoxicology research community to develop new concepts, tools,
http://dx.doi.org/10.1016/j.aquatox.2015.06.014 0166-445X/© 2015 Elsevier B.V. All rights reserved.
Please cite this article in press as: Hartmann, J.T., et al., Establishing mussel behavior as a biomarker in ecotoxicology. Aquat. Toxicol. (2015), http://dx.doi.org/10.1016/j.aquatox.2015.06.014
G Model AQTOX-4150; No. of Pages 10 2
ARTICLE IN PRESS J.T. Hartmann et al. / Aquatic Toxicology xxx (2015) xxx–xxx
and approaches to assess the effects of complex chemical mixtures on all levels of the ecosystem, i.e., organisms, communities and ecosystem functions. The use of standard ecotoxicology methods, exposing standard organisms (i.e., algae, daphnia, and fish) to measure standard endpoints (i.e., mortality, growth and fertility) is known to often underestimate the toxic effects on the diverse range of aquatic organisms, particularly for filter feeders (Connon et al., 2012). The sublethal effects of chemical exposure, such as behavior response, can be 10–100 times more sensitive than standard lethal exposure parameters (Hasenbein et al., 2015; Robinson, 2009). The behavior response of mussels is the result of physiological and environmental factors. Behavioral measurements reflect the combination of conditions representing a series of acute cumulative effects (Bae and Park, 2014). Thus, the development of sublethal ecotoxicological methods, by monitoring mussel filtration behavior, may provide the opportunity to measure the consequence of chemical stress on bivalves. This could offer an important addition to the suite of currently used ecotoxicological tools and explain an element causing freshwater mussel decline. The measurement of mussel valve movements (valvometry) using remote sensing technologies were first implemented by Kramer et al. (1989) as a tool for early warning alarms for aquatic pollution. Recent technology developments (e.g., Halls sensor technology and software) provide the ability to collect valvometric measurements with very high resolution in real-time (Robson et al., 2009; Sow et al., 2011). A Hall sensor is a transistor that, with the application of a controlled current, creates an output voltage proportional to the strength of a magnetic field (Nagai et al., 2006). The attachment of a Hall sensor and a magnet to the valves of an adult mussel allows the recording of their valve gap movements as a change in voltage output. The recording of gap changes provides evidence of a mussel’s response to disturbances as the closing of a mussel’s shells is an indicative evasion behavior (Lorenz et al., 2013). To our knowledge, Hall sensor technologies have not been used to improve the understanding of bivalves’ filtration behavior or evasive behavioral response to environmentally relevant chemical exposure scenarios. Mussel behavior could be established as a biomarker in ecotoxicology studies and provide the tool to assess mussel filtration behavior as a sublethal endpoint. To discuss and demonstrate the use of mussel behavior as a biomarker, we use the example of de-icing salt pollution. Sodium chloride (NaCl) is the most common component of de-icing salt and a substantial contaminant to freshwater ecosystems, particulary during winter and spring (Kaushal et al., 2005). Across the northern hemisphere millions of tons of NaCl are dispersed on hard surfaces to maintain community safety and economic activity during periods of freezing temperatures (Environment Canada, 2004; Thunqvist, 2004; US EPA, 2010). Chloride mass balance studies by Howard and Haynes (1993) and Perera et al. (2013) of the Highland Creek watershed east of Toronto (Canada) revealed that up to 60% of the de-icing salt, applied to hard surfaces, runs into surface water bodies almost immediately after application. Consequently, Cl− concentrations in urban freshwater streams have been reported to peak as high as 5000 mg/L in North America from de-icing salt contaminated snow melt (Corsi et al., 2010; Kaushal et al., 2005). This Cl− concentration is in the order of brackish conditions (i.e., Cl− concentrations between 500 mg/L to 5000 mg/L). The effects of de-icing salt on mussel larvae (glochidia) mortality has been investigated in a few studies, but not for adult mussel behavior (Beggel and Geist, 2015; Blakeslee et al., 2013; Gillis, 2011). NaCl is used in toxicology as a reference toxicant (ASTM, 2006), making it a suitable chemical to demonstrate our methodology. In our study, we use the European Unionoida species Anodonta anatina (Linnaeus 1758) as the model species. The purpose of this study was to establish mussel behavior patterns as a biomarker for ecotoxicology research for Bivalvia. The
objective was to test the suitability of three sublethal endpoints; Filtration Activity, Transition Frequency (change in activity status) and, Avoidance Behavior, in an innovative ecotoxicological experiment using environmentally relevant potential exposure scenarios simulating winter storms. With a ‘pulse-exposure’ experimental design and Hall sensor technology, we tested the hypothesis that exposure of adult A. anatina to de-icing salt (NaCl) alters their filtration behavior patterns. 2. Materials and methods 2.1. Study animals and experimental setup Anodonta anatina mussels were collected from ponds in Bavaria (Germany) and were kept outdoors in an 8000 L tank supplied with groundwater until experimentation. Following DNA extraction as described by Geist and Kuehn (2005), the mussels were genetically validated as A. anatina following the molecular identification method in Zieritz et al. (2012). Thirty mussels were used in the experiment with the mean length, height, and thickness of 9.02 ± 0.73 cm (mean ± SD), 5.26 ± 0.47 cm, and 3.2 ± 0.42 cm, respectively. Five mussels were placed randomly in one of six 25 L aquaria (25 cm × 40 cm × 25 cm with 4 cm of gravel substrate). Three aquaria were used for the treatment replicates and three aquaria for control replicates. All aquaria were kept in a water bath 12.5 ◦ C ± 0.5 ◦ C and filled with local groundwater (pH: 8.03, electrical conductivity at 20 ◦ C: 1072 S/cm, and 460 mg/L CaCO3 ). The mussels were fed 5 mL of Shellfish Diet 1800TM (Reed Mariculture Inc., U.S.A.) per aquarium once a week. A semi-closed system was designed to control the Cl− concentration within each aquarium without directly disturbing the mussels. A peristaltic pump (Ismatec MCP Standard, IDEX Health & Science GmbH, Germany) with Tygon® tubing (ID 3.2 mm, SaintGobain Performance Plastics Corporation, France) was used to create a flow-through exposure system. A stock solution of 10 g/L NaCl (99.5% purity, Merck KGaA, Germany) or fresh groundwater was pumped into the aquaria to achieve the desired Cl− concentration. Control aquaria received fresh groundwater during treatment phases. Two aeration stones were placed in each aquarium to ensure even mixing of the saline solution and to maintain dissolved oxygen levels. 2.2. NaCl exposure scenarios The mussels in the treatment aquaria were exposed to three dynamic NaCl exposure events designed to simulate possible deicing salt runoff events, each a week apart (Fig. 1). The three scenarios included: I A simulated single heavy runoff event after extensive application of road salts. Exposure duration of 30 h and a peak Cl− concentration of 3125 mg/L. II Three simulated runoff events within a short time frame. Total exposure duration of 48 h, with three peak Cl− concentrations up to 1750 mg/L and with low concentrations of 1000 mg/L between peaks. III A simulation of a period of cold conditions with salt application but without runoff of de-icing chemicals followed by a heavy runoff event with numerous de-icing salt applications. Exposure duration of 72 h and a peak Cl− concentration of 3840 mg/L. 2.3. Hall sensors, data measurement and data collection Hall sensor technologies were used in this experiment similar to the methods described by Wilson et al. (2005) and Robson et al.
Please cite this article in press as: Hartmann, J.T., et al., Establishing mussel behavior as a biomarker in ecotoxicology. Aquat. Toxicol. (2015), http://dx.doi.org/10.1016/j.aquatox.2015.06.014
G Model AQTOX-4150; No. of Pages 10
ARTICLE IN PRESS J.T. Hartmann et al. / Aquatic Toxicology xxx (2015) xxx–xxx
(2009). A Honeywell SS495A linear positions Hall sensor (Honeywell, U.S.A.) and a 0.2 g N42 neodymium magnet were glued to the opposite valves of each mussel using a cyanoacrylate adhesive, Dupla Plant Fix (Dohse Aquaristik GmbH, Germany). The Hall sensors were water proofed using synthetic rubber protective coating (Performix PLASTI DIP® , U.S.A.). Mussel movements were recorded as voltage readings (1 V) every minute for each mussel for the 48 days of the experiment using National Instruments hardware (module NI-9923 and cRIO-9074 controller) and a program developed in LabVIEW 2013TM (National Instruments, Austin U.S.A.). 2.4. Behavior data analysis Mussel behavior was analyzed for significant differences between the treatment group and control group (each group consisting of the three replicate aquaria containing 5 mussels each), between exposure and non-exposure phases within each group, and the individual behavior response to de-icing salt exposure. Three behavior endpoints were tested, Filtration Activity, Transition Frequency (per day) and Avoidance Behavior. The Filtration Activity was measured as the fraction of time a mussel’s shells were open and considered to be filtering over a specific period of time. The Transition Frequency was the number of observations where a mussel’s status changed from open to closed and vice versa. The Avoidance Behavior was a mussel’s evasive response to the chemical stressor observed when the mussel’s valves were closed. This behavior is distinctly different to the mussel’s undisturbed resting behavior. The status of each individual was determined as open (active filtering) or closed (not filtering) by using the mean voltage of the experimental period. Voltage measurements greater than the mean were considered open and values less were considered closed. To reduce noise a 10 min moving average was calculated for each data point. Eleven periods were extracted from the data for statistical analysis of significant effects for Filtration Activity and Transition Frequency endpoints (Fig. 1). The periods were selected before, during, and after each NaCl exposure scenario plus 7 and 14 d after the final exposure. Each period covered a duration of 48 h (i.e., 2880 data points) except for exposure scenarios I and III which lasted 30 h and 78 h, respectively. Each period was analyzed for normal distribution using Shapiro–Wilks Normality Test and Levene’s Test for heterogeneity of the variance. The values for Transition Frequency were normalized with a square root transformation.
3
To determine whether NaCl exposure had a significant effect on A. anatina’s Filtration Activity and Transition Frequency, we analyzed the data with linear mixed-effects models from the lme4 package in R version 2.12.1 (Bates et al., 2014; R Core Team, 2014). We tested four models for each dependent variable (Filtration Activity and Transition Frequency). Model 1 included the individual as random factor and period as fixed factor (period + (1/individual)). Model 2 contained aquaria as additional fixed factor (aquaria). In Model 3, group (treatment and control) as a fixed effect was added (group differences + period). In Model 4, the interaction between group differences and period (group differences × period) were included. We compared the models using the ANOVA function in the lme4 package in R. The Akaike Information Criterion (AIC) was then used to evaluate the most suitable model. Analysis of variance (ANOVA) was used to test for significant effects for the Filtration Activity and Transition Frequency endpoints with the aov() function in R version 2.12.1 (R Core Team, 2014). To verify the experimental method, the control was tested for significant differences between the three aquaria and between the eleven periods. The three periods before each of the exposure scenarios I, II, and III (periods 1, 4, and 7) were tested for their suitability to be used as an ‘Internal Standard’ for the treatment group. The internal standard was then then validated by comparison with the control mussels. To determine if de-icing salt exposure had an effect on A. anatina filtration behavior (i.e., Filtration Activity and Transition Frequency endpoints), the treatment group was compared against both the control and the internal standard. The exposure scenarios I, II and III (periods 2, 5 and 8) were compared together to test if mussel behavior changed during chemical exposure. The periods immediately after each exposure scenarios (periods 3, 6, and 9) were compared together to test for delayed or ongoing behavior changes. To detect long term effects, the three periods after the final exposure (time period 9, 10 and 11) were tested together. If ANOVAs identified significant effects, post-hoc tests (Tukey’s) were computed. Additionally, each group was also tested for difference with a multivariate ANOVA (MANOVA) approach combining the dependent variables, i.e., Filtration Activity and Transition Frequency using the lm() function implemented in the R package CAR (Fox and Weisberg, 2011). These statistical analyses were conducted using R version 2.12.1 (R Core Team, 2014). Significance was accepted at p < 0.05.
Fig. 1. An example of an Anodonta anatina’s individual behavioral responses to NaCl exposure. The roman numerals I–III represent environmentally relevant NaCl exposure scenarios, which simulate winter exposure to de-icing salt pollution over 48 days. For statistical analysis of measurable effects from NaCl exposure, eleven periods were selected from the Hall sensor readings (grey boxes). The letters A–C are three examples of closed valve behavior, disturbed, avoidance, and undisturbed, respectively. Photographs on the left illustrate mussels with Hall sensors attached to their valves, with the upper picture presenting an open (i.e., filtering) and the lower picture representing a closed (i.e., non-filtering) specimen.
Please cite this article in press as: Hartmann, J.T., et al., Establishing mussel behavior as a biomarker in ecotoxicology. Aquat. Toxicol. (2015), http://dx.doi.org/10.1016/j.aquatox.2015.06.014
G Model
ARTICLE IN PRESS
AQTOX-4150; No. of Pages 10
J.T. Hartmann et al. / Aquatic Toxicology xxx (2015) xxx–xxx
4
The Avoidance Behavior endpoint was analyzed first graphically by inspecting the filtration behavior pattern of each individual. Then six examples of the closed behavior, selected at regular intervals along the timeline, were investigated for each living mussel. The control group yielded a total of 78 observations. The observations for the treatment group were separated into periods of non-exposure to de-icing salt (71 observations) and the periods during exposure (16 observations). The closed behavior observations during the exposure scenarios occurred mostly during exposure scenario III (13 observations). Exposure scenarios I and II were too short for more entire closure examples to occur during these periods. A log-linear regression coefficient (R2 value) for each closed example and the time required for each movement from open to closed phase and reverse were compared. Homogeneity in the frequency of different closed behaviors between the control group and the treatment group were analyzed using Chi-Square (2 ) test.
for control and treatment mussels, respectively. However, considerable variations for the duration of each mussel’s open and closed phases were found within individuals and between individuals’ behavior patterns. Over the experimental period, the mean of mussels’ open behavior duration was 13.7 ± 8.1 h, while the closed behavior duration was 17.4 ± 11.5 h. The maximum active filtration duration of a mussel was 139 h (5.8 d) while the maximum resting duration was 159 h (6.6 d) with a coefficient of variation across all mussels of 43% and 59%, for open and closed, respectively. Based on the duration of filtering and resting phases and the large variation in filtration behavior patterns, no circadian rhythm or synchronism of the mussel’s filtration behavior patterns were observed. Mortality and Hall sensor failure were experienced during the experiment. Two mussels from the control group died, one on day 6 and the second on day 30 of the experiment. Also two mussels from the treatment group died during exposure scenario III. Two Hall sensors failed on two control mussels, one was replaced and unreliable data excluded for period 5 and 6. The Hall sensor on the second mussel failed to work correctly, but the fault was not noticed during the experiment. All data for this mussel were removed from the analysis. Out of the four linear mixed-effects models tested, Model 4 was the best-fitting statistical model for both Filtration Activity and
3. Results The A. anatina mussels demonstrated idiosyncratic behavior patterns which illustrates the challenges in behavior data analysis (Fig. 1). Over the 48 d of the experiment there was very little difference in the total filtration time of mussels, 47 ± 14% and 46 ± 8%
Table 1 ANVOA analysis results for the influence of de-icing salt on Anodonta anatina behavior patterns for two sublethal endpoints, Filtration Activity and Transition Frequency. ANOVA
Control Aquariums and All periods Control 3 aquaria Control period 1–11 Internal Standard 1 + 4 + 7 Treatment periods 1, 4 and 7 Control all v internal standard Control v treatment 1 + 4 + 7 Control all v exposure scenario I–III Control v treatment periods 2, 5, 8 Control v exposure I (p 2) Control v exposure II (p 5) Control v exposure III (p 8) Internal standard v exposure scenario I–III IS v treatment periods 2, 5, 8 IS v exposure I (p 2) IS v exposure II (p 5) IS v exposure III (p 8) Control all v after exposure I–III Control v treatment periods 3, 6, 9 Control v after exposure I (p 3) Control v after exposure II (p 6) Control v after exposure III (p 9) Internal standard v after exposure I–III IS v treatment periods 3, 6, 9 IS v after exposure I (p 3) IS v after exposure II (p 6) IS v after exposure III (p 9) Control all v long term effects Control v treatment period 9, 10, 11 Control v 48 h after exposure (p 9) Control v 7 d after exposure (p 10) Control v 14 d after exposure (p 11) Internal Standard v long term effects IS v treatment periods 9, 10, 11 IS v 48 h after exposure (p 9) IS v 7 d after exposure (p 10) IS v 14 d after exposure (p 11)
MANOVA
Filtration Activity
Transition Frequency
n
Df
F
Pr(>F)
F
Pr(>F)
137 137
2 10
0.541 0.790
0.583 0.639
0.800 0.500
45
2
0.267
0.767
182
13
0.991
181 152 152 151
3 1 1 1
4.72
89 60 60 59
3 1 1 1
180 152 152 150
Filtration Activity + Transition Frequency #
Sig.
F
Pr(>F)
0.451 0.888
0.657 0.649
0.975 0.873
1.46
0.243
0.891
0.473
0.462
0.722
0.740
2.07
0.130
0.003 0.127 0.087 0.308
33.6
ARTICLE IN PRESS
AQTOX-4150; No. of Pages 10
Aquatic Toxicology xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Aquatic Toxicology journal homepage: www.elsevier.com/locate/aquatox
Establishing mussel behavior as a biomarker in ecotoxicology Jason T. Hartmann a , Sebastian Beggel a , Karl Auerswald b , Bernhard C. Stoeckle a , Juergen Geist a,∗ a Aquatic Systems Biology Unit, Department of Ecology and Ecosystem Management, Technische Universitaet Muenchen, Muehlenweg 22, D-85350 Freising, Germany b Chair of Grassland Science, Department of Plant Science, Technische Universitaet Muenchen, Alte Akademie 12, D-85350 Freising, Germany
a r t i c l e
i n f o
Article history: Received 3 May 2015 Received in revised form 19 June 2015 Accepted 28 June 2015 Available online xxx Keywords: Bivalves Hall sensors Unionoida Sublethal endpoint Avoidance behavior NaCl
a b s t r a c t Most freshwater mussel species of the Unionoida are endangered, presenting a conservation issue as they are keystone species providing essential services for aquatic ecosystems. As filter feeders with limited mobility, mussels are highly susceptible to water pollution. Despite their exposure risk, mussels are underrepresented in standard ecotoxicological methods. This study aimed to demonstrate that mussel behavioral response to a chemical stressor is a suitable biomarker for the advancement of ecotoxicology methods that aids mussel conservation. Modern software and Hall sensor technology enabled mussel filtration behavior to be monitored real-time at very high resolution. With this technology, we present our method using Anodonta anatina and record their response to de-icing salt pollution. The experiment involved an environmentally relevant ‘pulse-exposure’ design simulating three subsequent inflow events. Three sublethal endpoints were investigated, Filtration Activity, Transition Frequency (number of changes from opened to closed, or vice versa) and Avoidance Behavior. The mussels presented a high variation in filtration behavior, behaving asynchronously. At environmentally relevant de-icing salt exposure scenarios, A. anatina behavior patterns were significantly affected. Treated mussels’ Filtration Activity decreased during periods of very high and long de-icing salt exposure (p < 0.001), however, increased during short de-icing salt exposure. Treated mussels’ Transition Frequency increased during periods of very high and long de-icing salt exposure (p < 0.001), which mirrored the Avoidance Behavior endpoint observed only by mussels under chemical stress. Characteristics of Avoidance Behavior were tighter shell closures with repeated and irregular shell movements which was significantly different to their undisturbed resting behavior (p < 0.001). Additionally, we found that mussels were sensitive to a chemical stressor even when the mussel’s valves were closed. Due to the effects of de-icing salt pollution on freshwater mussel behavior, we suggest better management practices for de-icing salt use be implemented. Our experimental method demonstrated that, with the application of current technologies, mussel behavioral response to a chemical stressor can be measured. The tested sublethal endpoints are suitable for mussel ecotoxicology studies. Avoidance Behavior proved to be a potentially suitable endpoint for calculating mussel behavior effect concentration. Therefore we recommend adult mussel behavior as a suitable biomarker for future ecotoxicological research. This method could be applied to other bivalve species and for physical and environmental stressors, such as particulate matter and temperature. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Pollution is a major driver towards the endangerment of freshwater bivalves (Unionoida) which are the fastest diminishing taxa with all of the European species listed on the International Union for Conservation of Nature’s (IUCN) ‘Red List’ (Bogan, 1993; Geist, 2011; IUCN, 2014). Unionoida conservation is a high priority. As keystone species they contribute to a number of essential functions
∗ Corresponding author. Fax: +49 8161 71 3477. E-mail address: [email protected] (J. Geist).
in aquatic ecosystems (Geist, 2010; Vorosmarty et al., 2010). Due to their high exposure pathway as filter feeders and their limited evasion capabilities, mussels are especially susceptible to organic and inorganic chemicals (Strayer et al., 2004). The number of registered chemical substances is over 100 million (CAS, 2015). This therefore presents a substantial risk to Unionoida (Dudgeon et al., 2006). The consequences of further decline in filter-feeder biodiversity is likely to have far reaching implications towards the failure of lotic and lentic ecosystem services (Cardinale et al., 2012; Chapin et al., 2000; Vaughn, 2010). In recognition of this, Eggen et al. (2004) appealed for the ecotoxicology research community to develop new concepts, tools,
http://dx.doi.org/10.1016/j.aquatox.2015.06.014 0166-445X/© 2015 Elsevier B.V. All rights reserved.
Please cite this article in press as: Hartmann, J.T., et al., Establishing mussel behavior as a biomarker in ecotoxicology. Aquat. Toxicol. (2015), http://dx.doi.org/10.1016/j.aquatox.2015.06.014
G Model AQTOX-4150; No. of Pages 10 2
ARTICLE IN PRESS J.T. Hartmann et al. / Aquatic Toxicology xxx (2015) xxx–xxx
and approaches to assess the effects of complex chemical mixtures on all levels of the ecosystem, i.e., organisms, communities and ecosystem functions. The use of standard ecotoxicology methods, exposing standard organisms (i.e., algae, daphnia, and fish) to measure standard endpoints (i.e., mortality, growth and fertility) is known to often underestimate the toxic effects on the diverse range of aquatic organisms, particularly for filter feeders (Connon et al., 2012). The sublethal effects of chemical exposure, such as behavior response, can be 10–100 times more sensitive than standard lethal exposure parameters (Hasenbein et al., 2015; Robinson, 2009). The behavior response of mussels is the result of physiological and environmental factors. Behavioral measurements reflect the combination of conditions representing a series of acute cumulative effects (Bae and Park, 2014). Thus, the development of sublethal ecotoxicological methods, by monitoring mussel filtration behavior, may provide the opportunity to measure the consequence of chemical stress on bivalves. This could offer an important addition to the suite of currently used ecotoxicological tools and explain an element causing freshwater mussel decline. The measurement of mussel valve movements (valvometry) using remote sensing technologies were first implemented by Kramer et al. (1989) as a tool for early warning alarms for aquatic pollution. Recent technology developments (e.g., Halls sensor technology and software) provide the ability to collect valvometric measurements with very high resolution in real-time (Robson et al., 2009; Sow et al., 2011). A Hall sensor is a transistor that, with the application of a controlled current, creates an output voltage proportional to the strength of a magnetic field (Nagai et al., 2006). The attachment of a Hall sensor and a magnet to the valves of an adult mussel allows the recording of their valve gap movements as a change in voltage output. The recording of gap changes provides evidence of a mussel’s response to disturbances as the closing of a mussel’s shells is an indicative evasion behavior (Lorenz et al., 2013). To our knowledge, Hall sensor technologies have not been used to improve the understanding of bivalves’ filtration behavior or evasive behavioral response to environmentally relevant chemical exposure scenarios. Mussel behavior could be established as a biomarker in ecotoxicology studies and provide the tool to assess mussel filtration behavior as a sublethal endpoint. To discuss and demonstrate the use of mussel behavior as a biomarker, we use the example of de-icing salt pollution. Sodium chloride (NaCl) is the most common component of de-icing salt and a substantial contaminant to freshwater ecosystems, particulary during winter and spring (Kaushal et al., 2005). Across the northern hemisphere millions of tons of NaCl are dispersed on hard surfaces to maintain community safety and economic activity during periods of freezing temperatures (Environment Canada, 2004; Thunqvist, 2004; US EPA, 2010). Chloride mass balance studies by Howard and Haynes (1993) and Perera et al. (2013) of the Highland Creek watershed east of Toronto (Canada) revealed that up to 60% of the de-icing salt, applied to hard surfaces, runs into surface water bodies almost immediately after application. Consequently, Cl− concentrations in urban freshwater streams have been reported to peak as high as 5000 mg/L in North America from de-icing salt contaminated snow melt (Corsi et al., 2010; Kaushal et al., 2005). This Cl− concentration is in the order of brackish conditions (i.e., Cl− concentrations between 500 mg/L to 5000 mg/L). The effects of de-icing salt on mussel larvae (glochidia) mortality has been investigated in a few studies, but not for adult mussel behavior (Beggel and Geist, 2015; Blakeslee et al., 2013; Gillis, 2011). NaCl is used in toxicology as a reference toxicant (ASTM, 2006), making it a suitable chemical to demonstrate our methodology. In our study, we use the European Unionoida species Anodonta anatina (Linnaeus 1758) as the model species. The purpose of this study was to establish mussel behavior patterns as a biomarker for ecotoxicology research for Bivalvia. The
objective was to test the suitability of three sublethal endpoints; Filtration Activity, Transition Frequency (change in activity status) and, Avoidance Behavior, in an innovative ecotoxicological experiment using environmentally relevant potential exposure scenarios simulating winter storms. With a ‘pulse-exposure’ experimental design and Hall sensor technology, we tested the hypothesis that exposure of adult A. anatina to de-icing salt (NaCl) alters their filtration behavior patterns. 2. Materials and methods 2.1. Study animals and experimental setup Anodonta anatina mussels were collected from ponds in Bavaria (Germany) and were kept outdoors in an 8000 L tank supplied with groundwater until experimentation. Following DNA extraction as described by Geist and Kuehn (2005), the mussels were genetically validated as A. anatina following the molecular identification method in Zieritz et al. (2012). Thirty mussels were used in the experiment with the mean length, height, and thickness of 9.02 ± 0.73 cm (mean ± SD), 5.26 ± 0.47 cm, and 3.2 ± 0.42 cm, respectively. Five mussels were placed randomly in one of six 25 L aquaria (25 cm × 40 cm × 25 cm with 4 cm of gravel substrate). Three aquaria were used for the treatment replicates and three aquaria for control replicates. All aquaria were kept in a water bath 12.5 ◦ C ± 0.5 ◦ C and filled with local groundwater (pH: 8.03, electrical conductivity at 20 ◦ C: 1072 S/cm, and 460 mg/L CaCO3 ). The mussels were fed 5 mL of Shellfish Diet 1800TM (Reed Mariculture Inc., U.S.A.) per aquarium once a week. A semi-closed system was designed to control the Cl− concentration within each aquarium without directly disturbing the mussels. A peristaltic pump (Ismatec MCP Standard, IDEX Health & Science GmbH, Germany) with Tygon® tubing (ID 3.2 mm, SaintGobain Performance Plastics Corporation, France) was used to create a flow-through exposure system. A stock solution of 10 g/L NaCl (99.5% purity, Merck KGaA, Germany) or fresh groundwater was pumped into the aquaria to achieve the desired Cl− concentration. Control aquaria received fresh groundwater during treatment phases. Two aeration stones were placed in each aquarium to ensure even mixing of the saline solution and to maintain dissolved oxygen levels. 2.2. NaCl exposure scenarios The mussels in the treatment aquaria were exposed to three dynamic NaCl exposure events designed to simulate possible deicing salt runoff events, each a week apart (Fig. 1). The three scenarios included: I A simulated single heavy runoff event after extensive application of road salts. Exposure duration of 30 h and a peak Cl− concentration of 3125 mg/L. II Three simulated runoff events within a short time frame. Total exposure duration of 48 h, with three peak Cl− concentrations up to 1750 mg/L and with low concentrations of 1000 mg/L between peaks. III A simulation of a period of cold conditions with salt application but without runoff of de-icing chemicals followed by a heavy runoff event with numerous de-icing salt applications. Exposure duration of 72 h and a peak Cl− concentration of 3840 mg/L. 2.3. Hall sensors, data measurement and data collection Hall sensor technologies were used in this experiment similar to the methods described by Wilson et al. (2005) and Robson et al.
Please cite this article in press as: Hartmann, J.T., et al., Establishing mussel behavior as a biomarker in ecotoxicology. Aquat. Toxicol. (2015), http://dx.doi.org/10.1016/j.aquatox.2015.06.014
G Model AQTOX-4150; No. of Pages 10
ARTICLE IN PRESS J.T. Hartmann et al. / Aquatic Toxicology xxx (2015) xxx–xxx
(2009). A Honeywell SS495A linear positions Hall sensor (Honeywell, U.S.A.) and a 0.2 g N42 neodymium magnet were glued to the opposite valves of each mussel using a cyanoacrylate adhesive, Dupla Plant Fix (Dohse Aquaristik GmbH, Germany). The Hall sensors were water proofed using synthetic rubber protective coating (Performix PLASTI DIP® , U.S.A.). Mussel movements were recorded as voltage readings (1 V) every minute for each mussel for the 48 days of the experiment using National Instruments hardware (module NI-9923 and cRIO-9074 controller) and a program developed in LabVIEW 2013TM (National Instruments, Austin U.S.A.). 2.4. Behavior data analysis Mussel behavior was analyzed for significant differences between the treatment group and control group (each group consisting of the three replicate aquaria containing 5 mussels each), between exposure and non-exposure phases within each group, and the individual behavior response to de-icing salt exposure. Three behavior endpoints were tested, Filtration Activity, Transition Frequency (per day) and Avoidance Behavior. The Filtration Activity was measured as the fraction of time a mussel’s shells were open and considered to be filtering over a specific period of time. The Transition Frequency was the number of observations where a mussel’s status changed from open to closed and vice versa. The Avoidance Behavior was a mussel’s evasive response to the chemical stressor observed when the mussel’s valves were closed. This behavior is distinctly different to the mussel’s undisturbed resting behavior. The status of each individual was determined as open (active filtering) or closed (not filtering) by using the mean voltage of the experimental period. Voltage measurements greater than the mean were considered open and values less were considered closed. To reduce noise a 10 min moving average was calculated for each data point. Eleven periods were extracted from the data for statistical analysis of significant effects for Filtration Activity and Transition Frequency endpoints (Fig. 1). The periods were selected before, during, and after each NaCl exposure scenario plus 7 and 14 d after the final exposure. Each period covered a duration of 48 h (i.e., 2880 data points) except for exposure scenarios I and III which lasted 30 h and 78 h, respectively. Each period was analyzed for normal distribution using Shapiro–Wilks Normality Test and Levene’s Test for heterogeneity of the variance. The values for Transition Frequency were normalized with a square root transformation.
3
To determine whether NaCl exposure had a significant effect on A. anatina’s Filtration Activity and Transition Frequency, we analyzed the data with linear mixed-effects models from the lme4 package in R version 2.12.1 (Bates et al., 2014; R Core Team, 2014). We tested four models for each dependent variable (Filtration Activity and Transition Frequency). Model 1 included the individual as random factor and period as fixed factor (period + (1/individual)). Model 2 contained aquaria as additional fixed factor (aquaria). In Model 3, group (treatment and control) as a fixed effect was added (group differences + period). In Model 4, the interaction between group differences and period (group differences × period) were included. We compared the models using the ANOVA function in the lme4 package in R. The Akaike Information Criterion (AIC) was then used to evaluate the most suitable model. Analysis of variance (ANOVA) was used to test for significant effects for the Filtration Activity and Transition Frequency endpoints with the aov() function in R version 2.12.1 (R Core Team, 2014). To verify the experimental method, the control was tested for significant differences between the three aquaria and between the eleven periods. The three periods before each of the exposure scenarios I, II, and III (periods 1, 4, and 7) were tested for their suitability to be used as an ‘Internal Standard’ for the treatment group. The internal standard was then then validated by comparison with the control mussels. To determine if de-icing salt exposure had an effect on A. anatina filtration behavior (i.e., Filtration Activity and Transition Frequency endpoints), the treatment group was compared against both the control and the internal standard. The exposure scenarios I, II and III (periods 2, 5 and 8) were compared together to test if mussel behavior changed during chemical exposure. The periods immediately after each exposure scenarios (periods 3, 6, and 9) were compared together to test for delayed or ongoing behavior changes. To detect long term effects, the three periods after the final exposure (time period 9, 10 and 11) were tested together. If ANOVAs identified significant effects, post-hoc tests (Tukey’s) were computed. Additionally, each group was also tested for difference with a multivariate ANOVA (MANOVA) approach combining the dependent variables, i.e., Filtration Activity and Transition Frequency using the lm() function implemented in the R package CAR (Fox and Weisberg, 2011). These statistical analyses were conducted using R version 2.12.1 (R Core Team, 2014). Significance was accepted at p < 0.05.
Fig. 1. An example of an Anodonta anatina’s individual behavioral responses to NaCl exposure. The roman numerals I–III represent environmentally relevant NaCl exposure scenarios, which simulate winter exposure to de-icing salt pollution over 48 days. For statistical analysis of measurable effects from NaCl exposure, eleven periods were selected from the Hall sensor readings (grey boxes). The letters A–C are three examples of closed valve behavior, disturbed, avoidance, and undisturbed, respectively. Photographs on the left illustrate mussels with Hall sensors attached to their valves, with the upper picture presenting an open (i.e., filtering) and the lower picture representing a closed (i.e., non-filtering) specimen.
Please cite this article in press as: Hartmann, J.T., et al., Establishing mussel behavior as a biomarker in ecotoxicology. Aquat. Toxicol. (2015), http://dx.doi.org/10.1016/j.aquatox.2015.06.014
G Model
ARTICLE IN PRESS
AQTOX-4150; No. of Pages 10
J.T. Hartmann et al. / Aquatic Toxicology xxx (2015) xxx–xxx
4
The Avoidance Behavior endpoint was analyzed first graphically by inspecting the filtration behavior pattern of each individual. Then six examples of the closed behavior, selected at regular intervals along the timeline, were investigated for each living mussel. The control group yielded a total of 78 observations. The observations for the treatment group were separated into periods of non-exposure to de-icing salt (71 observations) and the periods during exposure (16 observations). The closed behavior observations during the exposure scenarios occurred mostly during exposure scenario III (13 observations). Exposure scenarios I and II were too short for more entire closure examples to occur during these periods. A log-linear regression coefficient (R2 value) for each closed example and the time required for each movement from open to closed phase and reverse were compared. Homogeneity in the frequency of different closed behaviors between the control group and the treatment group were analyzed using Chi-Square (2 ) test.
for control and treatment mussels, respectively. However, considerable variations for the duration of each mussel’s open and closed phases were found within individuals and between individuals’ behavior patterns. Over the experimental period, the mean of mussels’ open behavior duration was 13.7 ± 8.1 h, while the closed behavior duration was 17.4 ± 11.5 h. The maximum active filtration duration of a mussel was 139 h (5.8 d) while the maximum resting duration was 159 h (6.6 d) with a coefficient of variation across all mussels of 43% and 59%, for open and closed, respectively. Based on the duration of filtering and resting phases and the large variation in filtration behavior patterns, no circadian rhythm or synchronism of the mussel’s filtration behavior patterns were observed. Mortality and Hall sensor failure were experienced during the experiment. Two mussels from the control group died, one on day 6 and the second on day 30 of the experiment. Also two mussels from the treatment group died during exposure scenario III. Two Hall sensors failed on two control mussels, one was replaced and unreliable data excluded for period 5 and 6. The Hall sensor on the second mussel failed to work correctly, but the fault was not noticed during the experiment. All data for this mussel were removed from the analysis. Out of the four linear mixed-effects models tested, Model 4 was the best-fitting statistical model for both Filtration Activity and
3. Results The A. anatina mussels demonstrated idiosyncratic behavior patterns which illustrates the challenges in behavior data analysis (Fig. 1). Over the 48 d of the experiment there was very little difference in the total filtration time of mussels, 47 ± 14% and 46 ± 8%
Table 1 ANVOA analysis results for the influence of de-icing salt on Anodonta anatina behavior patterns for two sublethal endpoints, Filtration Activity and Transition Frequency. ANOVA
Control Aquariums and All periods Control 3 aquaria Control period 1–11 Internal Standard 1 + 4 + 7 Treatment periods 1, 4 and 7 Control all v internal standard Control v treatment 1 + 4 + 7 Control all v exposure scenario I–III Control v treatment periods 2, 5, 8 Control v exposure I (p 2) Control v exposure II (p 5) Control v exposure III (p 8) Internal standard v exposure scenario I–III IS v treatment periods 2, 5, 8 IS v exposure I (p 2) IS v exposure II (p 5) IS v exposure III (p 8) Control all v after exposure I–III Control v treatment periods 3, 6, 9 Control v after exposure I (p 3) Control v after exposure II (p 6) Control v after exposure III (p 9) Internal standard v after exposure I–III IS v treatment periods 3, 6, 9 IS v after exposure I (p 3) IS v after exposure II (p 6) IS v after exposure III (p 9) Control all v long term effects Control v treatment period 9, 10, 11 Control v 48 h after exposure (p 9) Control v 7 d after exposure (p 10) Control v 14 d after exposure (p 11) Internal Standard v long term effects IS v treatment periods 9, 10, 11 IS v 48 h after exposure (p 9) IS v 7 d after exposure (p 10) IS v 14 d after exposure (p 11)
MANOVA
Filtration Activity
Transition Frequency
n
Df
F
Pr(>F)
F
Pr(>F)
137 137
2 10
0.541 0.790
0.583 0.639
0.800 0.500
45
2
0.267
0.767
182
13
0.991
181 152 152 151
3 1 1 1
4.72
89 60 60 59
3 1 1 1
180 152 152 150
Filtration Activity + Transition Frequency #
Sig.
F
Pr(>F)
0.451 0.888
0.657 0.649
0.975 0.873
1.46
0.243
0.891
0.473
0.462
0.722
0.740
2.07
0.130
0.003 0.127 0.087 0.308
33.6
