Journal list menu

Volume 12, Issue 6 e03527
Freshwater Ecology
Open Access

Exposure to a common antidepressant alters crayfish behavior and has potential subsequent ecosystem impacts

Alexander J. Reisinger

Corresponding Author

Alexander J. Reisinger

Soil and Water Sciences Department, University of Florida, Gainesville, Florida, USA

E-mail:[email protected]

Search for more papers by this author
Lindsey S. Reisinger

Lindsey S. Reisinger

Fisheries and Aquatic Sciences Program, University of Florida, Gainesville, Florida, USA

Search for more papers by this author
Erinn K. Richmond

Erinn K. Richmond

Water Studies Centre, School of Chemistry, Monash University, Clayton, Victoria, Australia

Search for more papers by this author
Emma J. Rosi

Emma J. Rosi

Cary Institute of Ecosystem Studies, Millbrook, New York, USA

Search for more papers by this author
First published: 15 June 2021
Citations: 8

Corresponding Editor: Whitney S. Beck.

Abstract

Pharmaceuticals are ubiquitous in aquatic environments, yet little is known regarding their impacts on ecological processes. Selective serotonin reuptake inhibitors (SSRIs) are frequently prescribed human antidepressants and have been shown to alter crayfish behavior. These behavioral alterations are particularly relevant as crayfish play a central role in freshwater ecosystems and often reach high biomass in anthropogenically influenced environments commonly exposed to pharmaceutical contamination. Using a 14-d artificial stream experiment, we exposed spinycheek crayfish (Faxonius limosus) to citalopram, a common SSRI, at an environmentally realistic concentration (0.5 µg/L). We used a Y-shaped flume to quantify the effects of SSRI exposure on crayfish behavior and food/conspecific preference. We also tested the interacting effects of citalopram and crayfish on habitat-specific and whole-stream ecosystem functions and biomass. Crayfish exposed to SSRIs exhibited increased boldness (time to emerge from shelters; P < 0.05) and spent more time orienting to food resources than nonexposed crayfish. Crayfish increased water column chlorophyll a (P < 0.01) and benthic organic matter (P = 0.03). Furthermore, crayfish potentially increased water column respiration (P = 0.09) and potentially decreased nitrate uptake (P = 0.05). SSRI exposure exhibited a potential effect of decreasing benthic chlorophyll a (P = 0.07), but there were no significant CRAY+SSRI interactions. Neither crayfish nor SSRI treatments affected whole-stream metabolism. These results suggest that citalopram has the potential to affect algal biomass but did not affect ecosystem functioning. However, alterations to crayfish behavior driven by SSRI exposure could lead to subsequent ecosystem-level effects as crayfish did affect various response metrics. We were unable to detect the effects of altered crayfish behavior at the ecosystem scale during our study, likely due to the short time frame (2 weeks) of our experiment. Further work is needed to quantify longer-term ecosystem consequences of sublethal effects of pharmaceuticals, but these results show that ecological responses to pharmaceuticals should consider the entire ecosystem.

Introduction

Pharmaceuticals are ubiquitous in aquatic ecosystems globally and are recognized as environmental contaminants of concern (Kolpin et al. 2002, Boxall et al. 2012, Bernhardt et al. 2017). The presence of pharmaceuticals in surface waters is no longer a question, but less is known regarding the impacts of pharmaceuticals on ecological dynamics. Typically, when pharmaceuticals are found in the environment, they are at very low concentrations (ng/L range). These low concentrations are below the threshold at which most pharmaceuticals are lethal to model organisms tested using standard ecotoxicology methods (Brausch et al. 2012). However, both acute exposure and chronic exposure to a variety of pharmaceuticals can have nonlethal effects across a range of taxa and trophic levels and disrupt ecological processes (Brausch et al. 2012, Richmond et al. 2017).

Selective serotonin reuptake inhibitors (SSRIs) are prescribed as antidepressants and are commonly detected in surface waters (Kolpin et al. 2002, Writer et al. 2013). In humans, SSRIs block the reuptake of serotonin (5-HT), a neurotransmitter (Schafer 1999), by receptors in the brain, thereby increasing serotonin levels (Monteiro and Boxall 2010). Commonly prescribed SSRIs include fluoxetine (trade name Prozac) and citalopram (trade name Celexa). Although these SSRIs are designed for, prescribed to, and used by humans, inevitably some proportion of SSRIs make their way into the aquatic environment, either through incomplete metabolism and human excretion or through inappropriate disposal. Once in the environment, aquatic communities are exposed to these SSRIs, which can alter stream community and ecosystem dynamics. For example, exposure to fluoxetine has been shown to increase dipteran emergence (Richmond et al. 2016) and inhibit benthic metabolism (Richmond et al. 2016, 2019). Other pharmaceuticals, either alone or as mixtures, can affect ecosystem functions, although these effects are dependent upon the environmental context (Gallagher and Reisinger 2020, Robson et al. 2020).

In addition to their effects on ecosystem functioning, SSRIs and other pharmaceuticals accumulate in aquatic invertebrates (Richmond et al. 2018, Miller et al. 2019). The accumulation of pharmaceuticals in aquatic invertebrates poses a range of potential impacts to the invertebrates themselves and their predators, including altered growth, reproduction, and behavior (Brausch et al. 2012, Richmond et al. 2017, 2018). From an ecological perspective, it is critical to understand behavioral alterations induced by pharmaceutical compounds. This may be particularly important for SSRIs, as 5-HT levels affect aggression, locomotion, and fight intensity in crustaceans (Huber and Delago 1998, Tierney et al. 2004, Bacque-Cazenave et al. 2018). Depending on the dose and timescale of exposure, studies have found either decreases or increases in aggression in crustaceans exposed to serotonin or SSRIs. Most of this work has involved the injection of serotonin directly into crustaceans, and few studies have examined the effects of environmental exposure to SSRIs (but see Woodman et al. 2016). We hypothesized that by altering crayfish boldness, aggression, and/or locomotion, exposure to SSRIs would change crayfish impacts on ecosystem processes by altering the time crayfish allocate to interacting with conspecifics, using refuge, and foraging.

Even small changes in behavior may alter key ecosystem processes, as freshwater crayfish are often dominant consumers of leaf litter, periphyton, and benthic invertebrates in streams (Whitledge and Rabeni 1997, Evans-White et al. 2003) and can have large effects on ecosystem functions such as leaf litter decomposition (Creed and Reed 2004, Alp et al. 2016) and nutrient cycling (Evans-White and Lamberti 2005). The potential for SSRIs to influence crayfish behavior and subsequent ecosystem responses is particularly relevant given that some crayfish are tolerant to a range of physical and chemical stressors (Reynolds et al. 2013). Due to this stress tolerance, crayfish are commonly found in anthropogenically influenced environments such as urban streams (Schilderman et al. 1999, Olden et al. 2011, Reynolds et al. 2013), which are also likely to have elevated concentrations of SSRIs and other pharmaceuticals (Ellis 2006, Richmond et al. 2018, Rosi et al. 2018).

We investigated the effects of an environmentally realistic concentration of citalopram, a commonly prescribed SSRI, on crayfish behavior and biogeochemical dynamics within artificial streams. We performed this controlled experiment in an artificial stream facility that has previously shown the effects of various contaminants on stream community and ecosystem dynamics (Lee et al. 2016, Richmond et al. 2016, 2019). Our overarching question for this study was as follows: Does citalopram exposure change the effects of crayfish on stream ecosystem dynamics? We hypothesized that (1) the presence of crayfish in artificial streams will alter ecosystem metrics; (2) the addition of citalopram to artificial streams will decrease rates of various stream ecosystem functions; (3) crayfish behavior will be altered by citalopram exposure; and (4) there will be an interaction between crayfish and citalopram, such that artificial streams containing both crayfish and citalopram will have different rates of ecosystem functions than streams containing only crayfish or only citalopram.

Materials and Methods

Artificial streams

We conducted our experiment over 14 d in April 2016 at the artificial stream facility located on the campus of the Cary Institute of Ecosystem Studies in Millbrook, New York, USA. We filled 20 oval-shaped recirculating artificial streams with 60 L of low-nutrient groundwater with minimal pharmaceutical contamination (unpublished data that analyzed the same water source for 14 separate pharmaceuticals, but not citalopram, found all analytes to be below detection, with detection limits in ng/L range). Water in each stream was circulated by stainless steel paddle wheels with a constant rotational velocity set at 25 rpm.

To provide a realistic stream community of algae, bacteria, fungi, and invertebrates, we pre-colonized quartz rocks (Maryland River Rock, Ayres Supply, Pennsylvania, USA) for three months and deployed leaf packs containing 5 g of air-dried red maple (Acer rubrum) within plastic mesh bags for two weeks prior to the start of the experiment in Wappinger Creek, a local stream running through the Cary Institute campus. Immediately prior to the beginning of the experiment, 60 pre-colonized rocks and six leaf packs were added to each artificial stream. Leaves were removed from the plastic mesh bags but held together as individual leaf packs using plastic-coated metal clips tethered to the side of the artificial stream (to prevent leaves from building up on the paddle wheels of the artificial stream).

Experimental design

After adding water, rocks, and leaf packs to each artificial stream, streams were randomly assigned to one of four treatments (n = 5 streams per treatment): control (CONT) streams that received neither crayfish nor citalopram additions, crayfish (CRAY) streams that received crayfish but no citalopram additions, SSRI (SSRI) streams that received citalopram additions but no crayfish, and a combined CRAY + SSRI treatment that received both crayfish and citalopram additions. Crayfish were collected from Wappinger Creek using baited minnow traps. After collection, crayfish were housed in aerated buckets filled with the same groundwater used in the artificial streams until the beginning of the experiment (within a week of crayfish collection). Although Wappinger Creek is located in a rural area, it likely receives regular PPCP exposure due to the combination of septic tanks in the watershed and a small wastewater treatment facility located upstream. We did not analyze Wappinger Creek water for PPCPs in the current study, although concentrations of various PPCPs (not including citalopram) were low (10 ng/L range) when analyzed approximately a decade prior to this study. For this experiment, we used the spinycheek crayfish (Faxonius limosus; hereafter “crayfish”), the crayfish species most commonly found in traps. We added three male crayfish to each stream receiving either the CRAY or SSRI + CRAY treatments so that each stream had a small (30–34 mm carapace length; CL), medium (33–37 mm CL), and large (36–39 mm CL) individual. We added crayfish to artificial streams 12–24 h following the addition of rocks and leaf packs and allowed crayfish to acclimate to the artificial streams overnight.

The morning after adding crayfish to streams, we performed our initial nutrient and pharmaceutical additions to each of the artificial streams. Prior to these additions, we ensured all stream volumes were maintained at 60 L by replacing water lost due to evaporation using the same low-nutrient groundwater source originally used to fill the streams. After refilling streams, we added a concentrated solution of nutrients ( NO 3 - -N and PO 4 3 - -P) to every stream, with a target nutrient enrichment of 100 and 10 µg/L NO 3 - -N and PO 4 3 - -P, respectively. For streams receiving a citalopram treatment (SSRI and SSRI + CRAY), we added citalopram HBr to our chemical addition solution, with a target final concentration of 0.5 µg/L (500 ng/L) of citalopram in the artificial streams. To meet this target concentration, we created a concentrated stock solution of citalopram by adding 0.0374 g of citalopram HBr (AK Scientific, Union City, California, USA) to 1 L of ultrapure water for a target concentration of 30,000 µg citalopram/L. We then performed a serial dilution to create a 3000 µg citalopram/L secondary stock. We added 10 mL of this secondary stock to each stream receiving a citalopram treatment by pipetting 10 mL of the secondary stock into the water column of the recirculating streams. Previous work in these systems to estimate nutrient cycling (and the nitrate uptake methods described below) has shown that streams are well mixed and that solutes added in this fashion homogenize within the artificial streams in <30 min. Nutrient enrichment of all streams was necessary due to the low-nutrient content of the groundwater used to fill the artificial streams (below a detection limit of 10 µg/L for both NO 3 - -N and PO 4 3 - -P). We completed chemical additions to each stream every other day for the duration of the experiment. This regular addition of nutrients and citalopram minimized potential nutrient limitation and mimicked the pseudo-persistent nature of pharmaceuticals in the environment (Daughton and Ternes 1999). Chemical addition solutions were stored at 4°C in glass volumetric flasks for the duration of the experiment.

Our experimental design aimed to mimic the pseudo-persistent nature of pharmaceuticals in the environment, with pharmaceuticals being regularly discharged from wastewater treatment facilities at low concentrations. Our approach of adding a low mass of citalopram to each SSRI-treated stream was designed to be similar to wastewater treatment effluent. We did not measure citalopram concentrations directly within the artificial stream. Without these data, it is possible that we could have increased citalopram concentrations to levels exceeding our target or exceeding environmental concentrations. However, based on previous studies regarding the fate of pharmaceuticals in these and similar artificial streams, we are confident that citalopram concentrations are within the range of environmentally measured concentrations. For example, if there were absolutely no environmental degradation and all citalopram added was conserved within the stream water, the overall mass of citalopram added to each stream would have raised concentrations by 3.5 µg/L by the end of the study. Although this concentration is higher than our target concentration of 0.5 µg/L, it would be well within the range of environmental concentrations recorded in surface waters (Fick et al. 2009, Cunha et al. 2019). Furthermore, this approach has been shown to provide environmentally realistic concentrations throughout the course of a study using fluoxetine, another SSRI antidepressant (Richmond et al. 2019). Pharmaceuticals are known to accumulate in biota (Richmond et al. 2018), indicating that it is likely that citalopram accumulated within crayfish tissue. A similar approach to the current study that investigated the effects of a mixture of pharmaceuticals on denitrification followed a similar approach of targeting a nominal concentration (Robson et al. 2020). We are unable to specify an exact concentration or exposure value from this study, and we only performed the study at one exposure level. Previous studies suggest that our design likely achieved environmentally realistic citalopram concentrations, and this study provides mechanistic support for establishing the interactions between citalopram and crayfish on ecosystem processes.

Crayfish behavioral endpoints

On day 13 of the experiment, we performed a crayfish behavioral assay to assess crayfish refuge use and orientation to chemical cues from food resources or conspecifics. We constructed a Y-maze flume out of clear plexiglass. The flume was 84 × 30 cm (length × width) with 12.5 cm high walls. A wall ran 53 cm along the center of the flume, connecting to the upstream flume wall, separating the entire flume into two different flow paths. There were two holes on the upstream end of the flume, one on each side of the wall, that were connected to 20-L buckets via plastic valves and tubing that served as the feedwater for the experiment. Four holes with valves on the downstream end of the flume served as the outlet for water that was directed into an outflow sink. The inflows were approximately 2/3 of the way up the walls, whereas the outflows were approximately 1/3 of the way up the walls. This spacing of inflows and outflows coupled with similar head pressure from both feed tanks allowed us to use gravity-induced flow rates for both treatments, providing a consistent flow rate between both sections of the maze. A visual schematic of the flume design is provided in Fig. 1. The wall in the center of the maze created two flow paths for the entire length of the flume, which was confirmed visually using a rhodamine WT visual tracer experiment prior to the initiation of the study.

Details are in the caption following the image
Schematic of Y-maze flume used for crayfish behavioral assays. Specific dimensions are listed in the figure. The different sections of the Y-maze are identified as neutral, conspecific, and food. The total amount of time spent in each section was recorded during behavioral assays. Inlets were connected to feedwater sources containing either conspecific or food chemical cues, and outlets fed out directly into a sink.

The bottom of the flume was marked to divide it into three sections, including the area behind the center wall (the neutral area, 31 × 30 cm) and each arm of the flume (areas that received chemical cue from a conspecific or food resource, 53 × 15 cm each). We covered the bottom and sides of the flume with black plastic to minimize visual disturbances to crayfish. Each flow path of the flume received water flowing from a 20-L bucket, which contained either a conspecific crayfish (collected from Wappinger Creek and not exposed to citalopram) or fish gelatin as a food resource. The conspecific crayfish treatment was used to measure aggression. Crayfish that approach conspecifics are considered to be more aggressive than those that avoid these encounters (Bruski and Dunham 1987). To make the gelatin, we combined 13 g of homogenized canned sardines, 7 g of unflavored gelatin, and 175 mL of boiling water. We poured the gelatin into a ceramic pan, cooled it in the refrigerator, and cut the gelatin into 1-cm2 blocks. Buckets were refilled with 20 L of aerated groundwater between trials. Two fresh gelatin blocks were used in each trial.

At the beginning of the trial, the flume was filled with fresh, aerated well water, and the valves to the buckets with chemical cue were closed. We placed one crayfish in a shelter (half of a bottle covered with black plastic with a trap door on one end) in the center of the neutral section with the shelter’s trap door facing toward the upstream end of the flume. We allowed the crayfish to acclimate in the shelter for 5 min with the trap door closed. After the acclimation period, we opened the valves so that water with chemical cue flowed into flume. Immediately after opening the valves, we removed the trap door to the shelter. We recorded the latency of crayfish to peek out of the front of the shelter and emerge completely from the shelter (as measures of refuge use). We also recorded the seconds the crayfish spent in the arm of the flume with crayfish cue (as a measure of interaction with conspecifics) and the arm with food cue (as a measure of foraging). Each trial lasted for 10 min.

Ecosystem endpoints

We quantified a suite of stream ecosystem endpoints within the artificial streams, including metabolic activity, biomass, and nitrate uptake. We deployed a miniDOT dissolved oxygen (DO) and temperature logger (PME, San Diego, California, USA) in each experimental stream and four light loggers (HOBO pendant loggers; Onset Computer, Bourne, Massachusetts, USA) spaced throughout the artificial stream facility that were pre-calibrated using a LI-COR PAR sensor (LI-COR Biosciences, Lincoln, Nebraska, USA) to estimate photosynthetically active radiation (PAR). Both DO and light loggers were set to log data every five minutes. Due to issues with sensor calibration and drift, DO data were unusable for two streams (one CRAY and one SSRI + CRAY). We therefore were only able to estimate whole-stream metabolism for 18 artificial streams. We used these DO, temperature, and PAR data to estimate daily stream metabolism from 18 artificial streams using the streamMetabolizer model framework (Appling et al. 2018a). A detailed description of the model framework is available in Appling et al. (2018a, b) Briefly, streamMetabolizer uses an inverse modeling approach with a Bayesian Markov chain Monte Carlo fitting procedure to model DO concentrations and the rate of oxygen change at each time point throughout the day by identifying the best estimates of daily gross primary production (GPPd; g O2·m−2·d−1), daily ecosystem respiration (ERd; g O2·m−2·d−1), and K600d, the daily average value of the standardized gas exchange rate coefficient (per day, scaled to a Schmidt number of 600) to best fit the DO data collected over the course of a day. We constrained the model by providing an estimate of K600 based on the rate at which water with initially low DO concentrations (due to its groundwater origin) reaches saturation under different stream velocities.

In addition to whole-stream metabolism, we estimated habitat-specific metabolism in the water column and benthic zone of each stream on days 10 and 13 of the experiment using the light–dark incubation technique. Briefly, we used 250-mL glass canning jars to isolate specific habitats of the stream. We either filled a jar with water directly from the water column of the respective artificial stream (water column treatment), or we added a representative rock from the bottom of the stream to the jar first and then filled the remaining space with water from the stream’s water column (benthic treatment). We used a ProODO handheld DO and temperature probe (YSI; Yellow Springs, Ohio, USA) to measure the DO (mg/L and % saturation) of water in each stream prior to filling jars. Jars were then filled directly within the stream and capped under water to ensure no air bubbles were present. We then incubated jars (inverted to prevent lids from affecting light dynamics) within each stream for 60–120 min, exposed to ambient temperature and light conditions. At the end of this incubation period, we measured DO concentrations within each jar. The change in DO under light conditions represents net ecosystem productivity (NEP). We repeated the incubations using fresh site water, but instead of incubating in full light, we transferred jars into a cooler (partially full of groundwater to buffer any temperature fluctuations) that was kept closed throughout the 60–120 min incubation to ensure dark conditions. At the end of this dark incubation, we again recorded DO concentrations and calculated ER as the decrease in DO during the dark incubation. We then used our estimates of NEP and ER to estimate habitat-specific GPP as follows:
GPP = NEP + | ER |

We multiplied water column GPP (GPPwc) and ER (ERwc) by artificial stream depth (0.1 m) to express it areally (g O2·m−2·h−1), whereas benthic metabolic activity (GPPben and ERben) was converted to areal fluxes (g O2·m−2·h−1) by dividing the hourly rate by surface area of the rock in each individual incubation (Reisinger et al. 2015).

On day 14 of the experiment, we estimated NO 3 - -N spiraling by adding a pulse of NO 3 - to each stream (target enrichment concentration = 200 µg NO 3 - -N/L above background) and measuring the change in concentration over a two-hour period using a submersible ultraviolet nitrate analyzer (SUNA) fitted with a rubber collar around the flow cell to allow for bench-top analysis. Prior to enrichment, we collected 25 mL of ambient water from each stream using a 60-mL syringe (separate syringes were used for SSRI and non-SSRI streams to prevent contamination), with each sample being immediately analyzed for NO 3 - on the SUNA. We rinsed syringes and the SUNA with low-nutrient groundwater between samples. After the pulse addition, we measured NO 3 - from each stream using the same approach continuously for two hours. We were able to complete one round of sampling (i.e., sample all 20 streams) in approximately 30 min, giving us four post-enrichment sampling events. We estimated NO 3 - -N uptake rates as the decrease in concentration over time following the addition (g·m−3·h−1), which we multiplied by mean stream depth (0.1 m) to convert to areal uptake (U; g N·m−2 h−1).

On one of the last two days of the experiment (either day 14 or day 15), we collected a subsample of water from the water column of each stream and filtered each subsample through two separate pre-ashed glass fiber filters (Pall A/E, 1.0 µm nominal pore size; Pall, Port Washington, New York, USA). We used one filter to estimate seston chlorophyll a (chl a) via the cold methanol fluorometric approach (Wetzel and Likens 2001) and the second filter was used to estimate seston ash-free dry mass (AFDM; the total organic content of suspended material) via loss on ignition (Steinman et al. 2006). We scaled both water column chl a (mg chl a/m2 and AFDM (g AFDM/m2) per area by multiplying resulting slurry concentrations (mg chl a or g AFDM/m3) by mean stream depth (m). After collecting these water column subsamples, we removed all remaining water from the artificial streams.

Once streams were dewatered, we removed leaf packs and thoroughly scrubbed all remaining benthic surfaces (both the stream structure itself and any rocks), to make a benthic slurry. We collected a subsample of this slurry and analyzed it for chl a and AFDM using the same approach as for water column samples. However, instead of multiplying by mean stream depth, we scaled chl a or AFDM by the known surface area of the entire artificial stream. These benthic chl a and AFDM estimates represent all of the chl a or organic matter (as AFDM) >1.0 µm that was present on benthic rocks and the artificial stream structure itself (e.g., walls of the artificial stream). Finally, we added water column and benthic chl a and AFDM estimates to provide whole-stream estimates for both algal biomass (as chl a) and total organic matter (as AFDM). We used these whole-stream estimates to analyze autotrophic and heterotrophic efficiency by scaling the cumulative daily gross primary production and ecosystem respiration throughout the study by the total chl a and AFDM, respectively (e.g., GPP per unit chl a and ER per unit AFDM).

Statistical analyses

To test the effects of SSRIs on crayfish (n = 3 per stream; N = 60) behavior, we used a generalized linear mixed-effects model (GLMM) with a Poisson distribution and a log link function. In this model structure, the crayfish behavioral response (latency to peek from shelter, latency to emerge from shelter, and time spent in either the food or crayfish side of the flume) was the response variables and SSRI treatment of the home stream of the crayfish was the factor. Individual crayfish were treated as replicates in this analysis. However, as three crayfish were used per stream, we treated stream as a random effect to account for non-SSRI-related differences among artificial streams.

For whole-stream metabolism, we used GLMMs with normal distributions, with crayfish (present or absent) and SSRI (added or not) as the two factors, an interaction term, and we treated date and stream as random effects. We used the same GLMM structure for benthic and water column metabolism as for whole-stream metabolism, but only had two dates to include in the model (rather than an estimate from each day of the experiment). We performed GLMMs using the lmer function in the lme4 package (Bates et al. 2015) and provide P values as estimated from the lmerTest package (Kuznetsova et al. 2017) in the statistical software program R (R Core Team 2019). For each GLMM, we also report the marginal ( R GLMM m 2 ) and conditional ( R GLMM c 2 ) r-squared values as originally proposed by Nakagawa and Schielzeth (2013) and modified by Nakagawa et al. (2017). The R GLMM m 2 estimates the overall variance explained by fixed effects only in a GLMM, whereas the R GLMM c 2 estimates the total variance explained by fixed and random effects of a GLMM. Furthermore, we calculated R β 2 for individual fixed effects. The R β 2 term is a semi-partial R2 value that estimates the variance explained by an individual fixed effect within a GLMM (Jaeger et al. 2017). We calculated R β 2 values using the Nakagawa and Shielzeth (2013) method implemented with the r2beta function in the r2glmm package (Jaeger 2017). These R β 2 terms can be thought of as effect sizes or individual R2 values for each specific fixed effect in GLMM frameworks.

For response variables that had only one estimate per stream, which included NO 3 - uptake (day 13), benthic, water column, and whole-stream chl a and AFDM (end of experiment), as well as whole-stream autotrophic and heterotrophic efficiency, we used two-way ANOVA to test the effects of treatments, with the relevant endpoints as response variables, and we included CRAY, SSRI, and their interaction as factors in the models. For significant ANOVA results, we performed pairwise comparisons using Tukey’s HSD tests and we also calculated effect sizes for individual factors using partial eta-squared ( η p 2 ) values (Richardson 2011). The η p 2 describes the amount of variance in a response variable that is described by a fixed effect after accounting for the variance described by other fixed effects and interactions. ANOVA tests were performed using the aov function within the base R program (R Core Team 2019), whereas partial η2 values were calculated manually from ANOVA results within R. For all statistical analyses, we set α = 0.05 as our critical value, although we report results with P values of 0.05 < P < 0.10 as potential effects. The strength of these effects can be assessed using the various effect size measures presented along with P values.

Results

Crayfish behavior

Crayfish behavior was significantly affected by SSRI exposure. Crayfish exposed to SSRIs peeked (GLMM; df = 1,25; P = 0.005; Fig. 2A) and emerged (GLMM; df = 1,25; P = 0.020; Fig. 2B) from shelter at the start of the flume trial more quickly (mean latency to peek = 39 s; mean latency to emerge = 67 s) than crayfish from control streams (peek = 80 s; emerge = 119 s). SSRI-exposed crayfish also spent more time in the portion of the Y-maze that received food chemical cues (mean = 275 s) than control crayfish (mean = 145 s; GLMM; df = 1,25; P = 0.009; Fig. 2C), and SSRI-exposed crayfish spent less time in the portion of the Y-maze that received conspecific chemical cues (mean = 85 s) than the control crayfish (mean = 179 s; GLMM; df = 1,25; P = 0.003; Fig. 2D). The difference between treatments was driven by SSRI-exposed crayfish spending >3× more time in the food resource section than the conspecific section, whereas control crayfish showed no preference for either section, spending similar amounts of time in both sections.

Details are in the caption following the image
Crayfish exposed to citalopram (Selective serotonin reuptake inhibitor [SSRI] treatment, blue boxes) peeked (A) and emerged (B) more quickly than crayfish not exposed to citalopram (CONT treatment, orange boxes). SSRI-exposed crayfish spent more time in the food section (C) of Y-maze flume than CONT crayfish and less time in the section with conspecific chemical cues (D). Each panel contains individual data points (solid points) overlain on box-and-whisker plots. The lower and upper edges of the box represent the first and third quartiles, the solid line is the median, and the lower and upper whiskers extend to the smallest or largest value within 1.5× the innerquartile range of the box. Crayfish behavior significantly differed (P < 0.05) between SSRI and non-SSRI streams for all panels. Full statistical results are provided in the text.

Algal biomass and organic matter

Benthic chl a at the end of the experiment averaged 22.24 (range: 6.45–43.60) mg chl a/m2 across all sites and treatments (Fig. 3A). There were no significant effects of CRAY, SSRI, or SSRI + CRAY (two-way ANOVA; df = 1,14; P = 0.118, 0.068, and 0.272, respectively) on benthic chl a, although there was a potential negative effect of SSRIs on benthic chl a (P = 0.068). The SSRI treatment ( η p 2  = 0.22) explained more variance than the CRAY treatment ( η p 2  = 0.17) or the SSRI + CRAY interaction ( η p 2  = 0.09). Water column chl a at the end of the experiment averaged 0.08 (range: 0.02–0.22) mg chl a/m2 across all sites and treatments (Fig. 3B). CRAY significantly increased water column chl a (two-way ANOVA; df = 1,16; P < 0.001), but there were no significant SSRI (two-way ANOVA; df = 1,16; P = 0.812) or SSRI + CRAY interaction (two-way ANOVA; df = 1,16; P = 0.742) effects. The CRAY treatment ( η p 2  = 0.70) explained a considerable amount of variation in seston chl a, whereas SSRI ( η p 2  < 0.01) and interaction ( η p 2  < 0.01) effects did not.

Details are in the caption following the image
Benthic (A, C) and water column (B, D) chlorophyll a (chl a; A, B) and ash-free dry mass (AFDM; C, D) from the end of experiment in control (CONT, light blue boxes), crayfish-only (CRAY, dark blue boxes), Selective serotonin reuptake inhibitor [SSRI]-only (SSRI, light green boxes), and SSRI + crayfish (SSRI + CRAY, dark green boxes)-treated streams. Each panel contains individual data points (solid points) overlain on box-and-whisker plots. Solid lines in the boxes represent the median value, the boxes denote the first and third quartiles, and whiskers extend to the most extreme value within 1.5× the interquartile range (IQR) of the boxes. Inset panels show pooled data when independent effects of either CRAY or SSRI treatments exhibited P values < 0.10. Exact P values are provided in inset panels, and full statistical results are provided in the text.

Benthic AFDM averaged 11.45 (range: 7.85–15.16) g/m2 at the end of the experiment (Fig. 3C). There was a significant increase in benthic AFDM in response to CRAY (two-way ANOVA; df = 1,16; P = 0.029) but no effect of SSRI or SSRI + CRAY treatments (two-way ANOVA; df = 1,16; P = 0.607 and 0.772, respectively). The CRAY treatment ( η p 2  = 0.26) had a larger effect size than either the SSRI treatment ( η p 2  = 0.02) or the SSRI+CRAY interaction ( η p 2  = 0.01). Water column AFDM averaged 0.40 (range: 0.30–0.69) g/m2 at the end of the experiment (Fig. 3D). There were no significant treatment effects on water column AFDM (two-way ANOVA; df = 1,16; P = 0.079, 0.693, and 0.643, respectively, for CRAY, SSRI, and SSRI + CRAY), although the CRAY treatment exhibited a potential positive effect (P = 0.079). The potential positive effect of the CRAY treatment is further supported by the fact that CRAY ( η p 2  = 0.19) exhibited a larger effect size than the SSRI treatment ( η p 2  = 0.01) and the SSRI + CRAY interaction ( η p 2  = 0.01).

Metabolism

Whole-stream metabolism was more influenced by variation among streams and days (random effects) than the specific experimental treatments (Table 1, Fig. 4). There were no significant effects of CRAY, SSRI, or their interaction on GPP (GLMM; df = 1,14 for each; P = 0.572, 0.640, and 0.565, respectively) or ER (GLMM; df = 1,14 for each; P = 0.777, 0.899, and 0.437, respectively). Although the overall model described a large amount of variation in whole-stream GPP and ER ( R GLMM c 2  = 0.93 for both), nearly all of this variance was explained by random effects (Table 1). Of the variation explained by fixed effects, crayfish ( R β 2  = 0.008) explained the most variance in GPP and the SSRI + CRAY interaction ( R β 2  = 0.007) explained the most variation in ER (Table 1, Fig. 4). There were no differences in autotrophic (cumulative GPP/total chl a) or heterotrophic (cumulative ER/total AFDM) efficiency across treatments (data not shown).

Table 1. Summary of generalized linear mixed-effects model results for whole-stream and habitat-specific gross primary production (GPP) and ecosystem respiration (ER).
Habitat Function R GLMM c 2 R GLMM m 2 CRAY SSRI SSRI + CRAY
P R β 2 P R β 2 P R β 2
Whole stream GPP 0.93 0.01 0.572 0.008 0.64 <0.001 0.565 0.004
ER 0.93 0.01 0.777 0.006 0.899 0.005 0.437 0.007
Water column GPP 0.83 0.03 0.377 0.008 0.361 0.006 0.935 <0.001
ER 0.51 0.11 0.093 0.07 0.482 0.02 0.651 0.01
Benthic GPP 0.59 0.03 0.336 0.01 0.899 <0.001 0.904 <0.001
ER 0.66 0.03 0.229 0.03 0.999 0.01 0.234 0.01

Notes

  • Fixed effects include crayfish (CRAY) and selective-serotonin reuptake inhibitor (SSRI) treatments and their interaction (SSRI + CRAY). Artificial stream identity and sample date were included as random effects.
Details are in the caption following the image
Mean daily whole-stream gross primary production (GPP; A) and ecosystem respiration (ER; B) were not affected by crayfish or citalopram treatments. Each point represents the daily mean value across five streams from control (CONT; light blue squares), crayfish (CRAY; dark blue circles), citalopram (Selective serotonin reuptake inhibitor [SSRI]; light green triangles), or citalopram plus crayfish (SSRI + CRAY; dark green diamonds) treatments.

Habitat-specific metabolic activity was not affected by the fixed effects of CRAY, SSRI, or their interaction. After accounting for the random effects of date and stream, CRAY, SSRI, and SSRI + CRAY did not significantly affect benthic (GLMM; df = 1,16 for each; P = 0.336, 0.899, and 0.904, respectively) or water column (GLMM; df = 1,16 for each; P = 0.377, 0.361, and 0.935, respectively) GPP (Table 1; Fig. 5). The overall models described considerable amounts of variation in benthic ( R GLMM c 2  = 0.59) and water column ( R GLMM c 2  = 0.83) GPP, but nearly all of this variation was explained by the random effects of stream and date (Table 1).

Details are in the caption following the image
Benthic (A, C) and water column (B, D) rates of gross primary production (GPP; A, B) and ecosystem respiration (ER; C, D) did not differ among control (CONT, light blue boxes), crayfish (CRAY, dark blue boxes), citalopram (Selective serotonin reuptake inhibitor [SSRI], light green boxes), and citalopram plus crayfish (SSRI + CRAY, dark green boxes) treatments, although there was a potential CRAY effect on water column ER (P = 0.093). Each panel contains individual data points overlain on box-and-whisker plots. Solid lines in the boxes represent the median value, the boxes denote the first and third quartiles, and whiskers extend to the most extreme points within 1.5× the interquartile range above or below the boxes. The inset panel (D) shows the potential effect of CRAY regardless of SSRI treatment with the P value provided. Full statistical results are provided in the text.

Similar to habitat-specific GPP, benthic ER was not affected by any treatment (Fig. 5C). After accounting for random effects of stream and date, there were no significant differences across treatments (GLMM; df = 1,16; P = 0.229, 0.999, and 0.234 for CRAY, SSRI, and SSRI + CRAY, respectively). The overall model described a substantial portion of the variability in benthic ER ( R GLMM c 2  = 0.66), but nearly all of this was due to random effects, as the fixed effects explained a small amount of the total variation ( R GLMM m 2  = 0.03; Table 1). Water column ER also was not significantly affected by any CRAY, SSRI, or SSRI + CRAY effects after accounting for random effects of stream and date (GLMM; df = 1,16; P = 0.093, 0.482, and 0.651, respectively), although water column ER exhibited a nonsignificant increase in CRAY treatments (P = 0.093; Fig. 5D). The amount of variation explained by the overall water column ER model ( R GLMM c 2  = 0.51) was driven primarily by random effects, although fixed effects explained a nontrivial amount of this variation ( R GLMM m 2  = 0.11; Table 1).

Nitrate uptake

Background concentrations of NO 3 - -N on day 14 prior to the NO 3 - -N pulse addition averaged 0.19 mg N/L, showing that the continual addition of NO 3 - -N did indeed raise concentrations above ambient groundwater levels (below detection). Following the pulse addition of NO 3 - -N, NO 3 - -N uptake averaged 12.86 (range: 9.95–14.58) g N·m−2·h−1 across all streams at the end of the experiment (Fig. 6). There were no significant effects of CRAY, SSRI, or SSRI + CRAY on NO 3 - -N uptake (two-way ANOVA; df = 1,16; P = 0.051, 0.216, and 0.605, respectively). However, the CRAY treatment exhibited a potential effect of decreasing NO 3 - -N uptake (P = 0.051). This potential effect of CRAY decreasing NO 3 - -N uptake is further supported by the CRAY treatment exhibiting a higher effect size ( η p 2  = 0.21) than the SSRI treatment ( η p 2  = 0.09) and the SSRI+CRAY interaction ( η p 2  = 0.02).

Details are in the caption following the image
Nitrate ( NO 3 - -N) uptake did not significantly differ among control (CONT, light blue box), crayfish (CRAY, dark blue box), citalopram (Selective serotonin reuptake inhibitor [SSRI], light green box), or citalopram plus crayfish (SSRI + CRAY, dark green box) treatments, although there was a potential decrease in NO 3 - -N uptake due to the CRAY treatment observed (P = 0.051). Individual points represent individual stream results overlain on box-and-whisker plots. Solid lines in the boxes represent the median value, the boxes denote the first and third quartiles, and whiskers extend to the most extreme data points within 1.5× the interquartile range above or below the boxes. The inset panel shows the potential effect of CRAY regardless of SSRI treatment with the P value provided. Full statistical results are provided in the text.

Discussion

Despite having known for almost two decades that pharmaceuticals are commonly found in aquatic environments (Kolpin et al. 2002), we are only beginning to understand the potential implications of these synthetic chemicals on aquatic organisms and ecosystem processes. Our study found that exposure to an environmentally realistic concentration of citalopram significantly affected crayfish behavior by increasing crayfish boldness and altering chemical cue preferences. These behavioral changes were coupled with significant crayfish effects on habitat-specific biomass of algae and organic matter, as well as a potential crayfish effect on NO 3 - -N uptake. In contrast, we were only able to detect a potential effect of the citalopram treatment related to a decrease in benthic algal biomass. There were no other direct effects of citalopram evident at the ecosystem scale. Effects of either treatment on metabolism at both the habitat-specific and whole-stream scale were less apparent. Overall, these results support two of our four initial hypotheses: (1) Crayfish affected stream ecosystem metrics, and (3) crayfish behavior was altered by citalopram exposure. Our other hypotheses, (2) citalopram would decrease stream ecosystem functioning, and (4) there would be an interaction between the crayfish and citalopram treatments on ecosystem functioning, were not supported. The addition of citalopram to artificial streams on its own did not have measurable effects on ecosystem functions, and there were no significant interactions.

SSRI effects on crayfish behavior

Previous research has found that altering serotonin levels within crayfish led to altered aggression and boldness (Huber and Delago 1998, Bacque-Cazenave et al. 2018), but these studies were typically performed by direct serotonin injection into crayfish (Huber et al. 2001, but see Woodman et al. 2016). In contrast to these previous studies, here we show altered crayfish behavior in response to exposure to an environmentally realistic concentration of citalopram, a common SSRI, in the water. These results suggest that the SSRI mode of action, inhibition of serotonin reuptake, can lead to similar behavioral responses as direct serotonin injection, despite low SSRI concentrations. Similarly, Buřič et al. (2018) found that marbled crayfish (Procambarus virginalis) activity was significantly reduced when exposed to ~0.9 µg citalopram/L. By altering key behavioral traits in crayfish, environmental concentrations of SSRIs are likely to affect the fitness of individuals and their impacts on the environment.

In our study, crayfish exposed to citalopram were bolder and were more attracted to food odor and less attracted to conspecific odor than crayfish that were not exposed to citalopram. The additional time that SSRI-exposed crayfish spent in the food resource section of the flume could result from a greater attraction toward food or decreased aggression toward conspecifics. Overall, these behavioral changes suggest that citalopram decreases predator avoidance behavior and increases the time that crayfish allocate to foraging. Therefore, crayfish exposed to SSRIs in the environment may be more vulnerable to predation. For example, previous studies have found that crayfish that are actively foraging and outside of refuge are more likely to be consumed by predatory fish (Soderback 1994).

While reduced predator avoidance and increased foraging may increase predation risk, these behaviors also likely increase the per capita impacts of crayfish on lower trophic levels. Previous studies have found that the presence of predatory fish can dramatically alter macrophyte and macroinvertebrate abundance by depressing crayfish feeding activity (Hill and Lodge 1995). However, predator presence does not have the same effect on feeding behavior if crayfish are especially bold. For example, parasitism can increase crayfish boldness, and this negates the cascading effects of predators on crayfish impacts (Reisinger and Lodge 2016). Overall, the behavioral changes caused by citalopram exposure could reduce crayfish population sizes by increasing vulnerability to predation, but they could also increase per capita impacts of crayfish on the environment by increasing foraging rates.

Experimental duration and SSRI effects on crayfish impacts

We exposed artificial stream ecosystems to the different experimental treatments for 14 d. This experiment length was based on a previous study performed in these same artificial streams that found insect emergence increased in SSRI-exposed streams relative to controls (Richmond et al. 2016). Indeed, we were able to detect significant differences in certain response metrics during this two-week time frame. However, it is possible that the effects of SSRIs or crayfish may take longer to become detectable. Previous work that has found that crayfish affect ecosystem processes (Lodge et al. 1994) and that behavioral variations can modify these effects (Reisinger and Lodge 2016) have been performed over 4- to 12-week time periods, rather than 14 d. The effects of SSRIs are not likely to be immediate, as it takes time for the drug to accumulate within an organism. It is possible that the behavioral response of crayfish was still developing at the end of our experiment, so a longer study duration may be needed to document the cascading effects of crayfish exposure to citalopram eventually leading to changes in ecosystem processes. While it is possible that there are minimal ecosystem effects of altered crayfish behavior due to citalopram exposure, we expect that if we had allowed our experiment to run for longer the behavioral differences induced by citalopram exposure would have led to more pronounced interactive effects of citalopram and crayfish on stream ecosystem metrics.

Spinycheek crayfish had several ecological impacts in our experiment that could become more pronounced with chronic citalopram exposure. For instance, crayfish increased the concentrations of chlorophyll a in the water column, which could be the result of algal particles sloughing off from crayfish grazing on benthic biofilms or a positive effect of the nutrients excreted by crayfish on phytoplankton biomass. There was also a trend for lower nitrate uptake in experimental streams containing crayfish, potentially because crayfish excrete more bioavailable forms of nitrogen. In addition, the presence of crayfish in our experiment increased benthic organic matter, probably because crayfish break down leaf litter (Creed and Reed 2004). While our behavioral results suggest that crayfish foraging rates were greater in the presence of citalopram, we did not observe stronger effects of crayfish in the citalopram treatment. However, if these behavioral effects are enhanced over longer time periods, citalopram may intensify crayfish effects in streams, potentially leading to greater rates of leaf litter decomposition, higher turnover of benthic biofilms, and altered nutrient dynamics.

Effects of exposure to pharmaceuticals at low concentrations likely operate under a chronic exposure framework. Assessing chronic, sublethal effects of pharmaceuticals on stream ecosystems is difficult in a short time period, as sublethal effects may not be evident until completion of a full organismal life cycle. For example, Hoppe et al. (2012) used an 83-d incubation to find that cimetidine, an antihistamine, can affect freshwater invertebrate population size structure and growth rates, which would not have been evident in a short-term study. Observing chronic effects was beyond the scope of this study, but previous work suggests that if our experiment had lasted longer, we may have observed effects of citalopram not only on behavior but also on other aspects of crayfish ecology. For example, citalopram and other SSRIs have been found to affect both behavior and population dynamics of other crustaceans (Sehonova et al. 2018 and references therein), although the vast majority of these longer-term ecotoxicology studies have focused on organisms with shorter life spans such as daphnids or amphipods (Fong and Ford 2014 and references therein). Despite the fact that crayfish were only exposed to SSRIs for 14 d during our experiment, we observed significant changes in behavior near the end of the study. These behavioral changes, coupled with crayfish-driven impacts on ecosystem dynamics observed in previous studies (Lodge et al. 1994, Hill and Lodge 1995), suggest that citalopram exposure can not only affect crayfish themselves, but also these effects may translate into impacts on ecosystem dynamics. These ecosystem impacts, however, may be delayed relative to behavioral alterations. We recommend further studies investigating these types of drug–animal interactions on ecosystem processes, as this reflects real-world conditions.

Environmental concentrations of SSRIs and lethal vs. ecological effects

We selected 0.5 µg/L as our target exposure concentration for citalopram, which is in the range commonly used for environmentally realistic concentrations, despite concentrations being found up to 76 µg/L in surface waters and 840 µg/L in wastewater effluent (Fick et al. 2009, Cunha et al. 2019). In this study, we chose to test the effects of an environmentally realistic concentration of citalopram rather than a dose–response experimental approach. Therefore it is not possible for us to establish a threshold concentration (i.e., an LC50 or EC50) of ecological disruption (sensu Richmond et al. 2017), but our exposure concentration is orders of magnitude lower than established citalopram LC50s for daphnids (Ceriodaphnia dubia; 48h LC50 = 3900 µg/L; Henry et al. 2004) or freshwater algae (Pseudokirchneriella subcapitata; 48h LC50 = 1600 µg/L; Christensen et al. 2007). Although ecotoxicology studies using ecosystem functions as endpoints are rare, Richmond et al. (2016) found that citalopram exposure elicited significant reductions in biofilm GPP and ER in a 14-d artificial stream study. However, this previous study used a much higher (20 µg/L) concentration of citalopram than in the current study (0.5 µg/L).

The drastic difference in concentrations between lethal effects found in previous studies relative to the concentrations found to have sublethal organismal and/or ecosystem effects here highlights the need for further focus on sublethal, ecosystem-level effects of pharmaceuticals. Pharmaceuticals are biologically active compounds designed to elicit behavioral and physiological changes in those to whom they are prescribed. More than 90% of ecotoxicology studies of pharmaceuticals have focused on lethal effects, and <2% of these studies have quantified ecosystem-level effects (Richmond et al. 2017). Pharmaceuticals are often neglected as a research foci given that environmental concentrations are typically much lower than established LC50s (Kolpin et al. 2002, Richmond et al. 2017). However, studies addressing sublethal effects (i.e., behavioral responses or ecosystem-scale endpoints) are increasingly showing that environmentally realistic concentrations of various pharmaceuticals (such as citalopram) can significantly affect fundamental ecological processes.

Conclusion

It is clear that pharmaceuticals, including citalopram and other SSRIs, are present in freshwater environments (Kolpin et al. 2002, Pal et al. 2010), can have various sublethal impacts on nontarget organisms (Fong and Ford 2014, Sehonova et al. 2018, Cunha et al. 2019), and these impacts can affect ecological processes (Lee et al. 2016, Richmond et al. 2016, Rosi et al. 2018). Ultimately, SSRIs and other pharmaceuticals appear to affect nontarget organisms, altering food web dynamics and top-down effects of organisms on ecosystem processes. Improving our understanding of the sublethal effects of pharmaceuticals (individually and as mixtures) on nontarget aquatic organisms at different trophic levels, and the subsequent ecological disruption that ensues, will allow for a more accurate assessment of the impacts of exposure to low concentrations of pharmaceuticals on ecosystem dynamics.

Acknowledgments

We thank Don Uzarski for giving LSR and AJR time to run the experiment, Paul Moore for advice on crayfish behavioral assays and valuable discussions on crayfish neurobiological responses to SSRIs, Heather Malcom for logistical help preparing and running the experimental system, and Paul DeBonis for help constructing the Y-maze flume. This manuscript was improved by comments from two anonymous reviewers. Author contributions: Alexander J. Reisinger contributed to study conceptualization, data curation, formal analysis, investigation, methodology, visualization, and writing of the original draft of the manuscript. Lindsey S. Reisinger contributed to study conceptualization, data curation, formal analysis, investigation, methodology, and reviewing and editing of the manuscript. Erinn K. Richmond contributed to study conceptualization, methodology, and reviewing and editing of the manuscript. Emma J. Rosi contributed to study conceptualization, investigation, methodology, project administration, securing resources for the project, supervision, and reviewing and editing of the manuscript. The authors declare no conflicts of interest.