Ontogenetic Differences in Swimming Behavior of Fish Exposed to the Harmful Dinoflagellate Cochlodinium polykrikoides
Abstract
Blooms of the ichthyotoxic dinoflagellate Cochlodinium polykrikoides present both lethal and sublethal threats to coastal marine organisms. Because prior studies of this harmful algal bloom (HAB) species have focused on its acute toxic effects on fish, there remains a limited understanding of the sublethal effects on the swimming behavior of fish when exposed to these blooms. We conducted a video-based laboratory assessment of the effects of C. polykrikoides exposures on larval and juvenile fish swimming behavior. Juvenile and larval life stages of three forage species common to the U.S. East Coast were examined: wild Atlantic Silversides Menidia menidia, hatchery-reared Inland Silversides M. beryllina, and hatchery-reared Sheepshead Minnow Cyprinodon variegatus. Results showed that juvenile Atlantic Silversides and Inland Silversides exposed to C. polykrikoides concentrations of 102 cells/mL swam significantly further distances in comparison with their baseline (control) behavior. In an ecosystem context, the results of this research support field-based studies demonstrating that when blooms of this HAB species are present, the relative abundance of fish declines (i.e., the fish presumably detect the blooms and swim away after exposure). Conversely, juvenile Sheepshead Minnow exhibited no increased swimming behavior when exposed, supporting previous studies indicating that this species is more resistant to C. polykrikoides toxicity. Importantly, no behavioral changes were found in experiments with larval conspecifics (i.e., Inland Silversides and Sheepshead Minnow), suggesting that younger life stages, which lack developed gills and olfactory systems, may be unable to detect the dinoflagellate. This clear ontogenetic difference implies that for certain fish species, reaching later life stages may provide a refuge to HAB toxicity.
Received March 26, 2017; accepted June 5, 2017 Published online August 9.2017
In many regions of the world, blooms of the ichthyotoxic dinoflagellate Cochlodinium polykrikoides now occur annually, presenting a significant threat to coastal marine organisms (Lee et al. 2002; Gobler et al. 2008; Tomas and Smayda 2008; Richlen et al. 2010; Kudela and Gobler 2012). Although the toxinology is not yet confirmed for this alga (Kim et al. 1999, 2002, 2009; Tang and Gobler 2009; Kim and Oda 2010), several studies have demonstrated that the toxic agents are consistent with reactive oxygen species: extracellular, short-lived (minutes), produced by physiologically active C. polykrikoides cells, and mitigated by free radical scavenging enzymes, such as peroxidase (Kim et al. 2002; Tang and Gobler 2009, 2010). In coastal ecosystems, blooms caused by C. polykrikoides can form dense, heterogeneous patches that vary in size (i.e., from m2 to km2 scales) and duration (i.e., weeks to several months; Gobler et al. 2008; Mulholland et al. 2009; Tang and Gobler 2010; Koch et al. 2014). This creates the potential for dynamic exposures to fish and other marine organisms (Gobler et al. 2008; Richlen et al. 2010; Koch et al. 2014). Previous understanding of the ichthyotoxicity of C. polykrikoides has been based almost entirely on static toxicity experiments or on observations of fish kill events (Onoue et al. 1985; Kim et al. 1999; Whyte et al. 2001; Landsberg 2002; Gobler et al. 2008; Tang and Gobler 2009; Richlen et al. 2010; Rountos et al. 2014). Although these studies clearly demonstrated the potent ichthyotoxicity of C. polykrikoides, they often failed to address the dynamic nature of exposures in the field or the potential for fish to display increased activity levels to avoid or escape these blooms.
An array of toxicity experiments and approaches has been used to understand the effects of contaminants on fish species and to determine acceptable exposure concentrations in the field (Sprague 1990; USEPA 2002). While traditional acute and chronic toxicity experiments are valuable for obtaining baseline toxicology data, they are often difficult to interpret in an ecological context (Kimball and Levin 1985; deVlaming and Norberg-King 1999). In this regard, behavioral toxicity experiments have been useful for providing ecologically relevant endpoints and toxicological information (Little et al. 1990; Weis and Weis 1995; Kane et al. 2005).
The impacts of anthropogenic contaminants on fish behavior have been explored for several decades (Rand 1985; Atchison et al. 1987; Little and Finger 1990; Webber and Haines 2003; Sloman and Wilson 2006). Many of these studies have demonstrated serious impairment of sensory abilities in fish after exposures to contaminants, ultimately leading to the compromised survival of exposed individuals (McPherson et al. 2004; Leduc et al. 2006, 2009; Tierney et al. 2010, 2011). Surprisingly, few studies have evaluated the effects of harmful algal blooms (HABs) or their toxins on fish behavior. Studies belonging to the relatively sparse literature on HABs and fish behavior (Lefebvre et al. 2001, 2004, 2012; Salierno 2005; Samson et al. 2008; Rountos et al. 2014) have found a variety of behavioral responses after exposures to HAB species, ranging from impaired swimming to reduced feeding ability. Unfortunately, for the vast majority of HAB species that plague coastal waters, only data from traditional toxicity assessments are available (Landsberg 2002).
In the present study, we conducted a video-based laboratory assessment of the effects of C. polykrikoides exposures on the immediate swimming behavior of three forage fish species. A commercially available movement analysis system was used to examine behavioral changes in larval and juvenile fish after exposure to water containing C. polykrikoides. The experimental approach and laboratory framework developed here can inform future behavioral toxicity experiments with harmful algae or other toxicants, aiding in our understanding of the ecosystem effects of these blooms or other coastal hazards on fish populations.
METHODS
Bloom water and dinoflagellate cultures
Larval fish experiments used C. polykrikoides bloom water and nonbloom water (i.e., water collected at least 5 m away from a bloom patch of water) collected directly from the Stony Brook Southampton Marine Science Center on Old Fort Pond (a tributary of Shinnecock Bay, New York) during daytime hours. Water was collected with polyethylene carboys at a depth of 0.25 m from the surface and was taken into the laboratory immediately for experiments. Bloom and nonbloom patches were used, as Koch et al. (2014) demonstrated that aside from the high densities of this harmful dinoflagellate, the chemical composition and microbial community of the two patch types exhibited no major differences. In addition, sieved C. polykrikoides bloom water would not have been an appropriate control because residual toxicity from the HAB filtrate has been shown to impact survival of fish larvae (Tang and Gobler 2009).
Laboratory cultures of dinoflagellates were used in juvenile fish experiments to allow for the assessment of a nontoxic dinoflagellate control. The dinoflagellate C. polykrikoides (strain CP1) was isolated from bloom water collected in Flanders Bay (Peconic Estuary, New York) during 2006 (Tang and Gobler 2009). The similarly sized dinoflagellate Gymnodinium aureolum (strain Gym), which served as a nontoxic dinoflagellate control (Tang and Gobler 2009), was isolated from the Elizabeth River (Chesapeake Bay, Virginia) in 2006 (Tang et al. 2008). Cells of G. aureolum range from 27 to 34 μm in length and from 17 to 32 μm in width (Tang et al. 2008), while C. polykrikoides cells range from 21 to 35 μm in length and from 24 to 48 μm in width (Gobler et al. 2008). Clonal cultures of C. polykrikoides and G. aureolum were maintained in sterile GSe culture medium (defined below) that was prepared according to Tang and Gobler (2009), with autoclaved and filtered (0.22 μm) coastal Atlantic Ocean seawater (salinity = 30‰) supplemented with stock nutrients and an antibiotics solution with a final concentration of 2% (a mixture of penicillin [10,000 IU] and streptomycin [10,000 μg/mL]; Mediatech, Inc., Hemdon, Virginia). The “GSe medium” (sensu Doblin et al. 1999) is defined as culture medium that, in addition to macronutrients, contains selenium (Se), which is an important trace metal. Cultures of C. polykrikoides and G. aureolum were incubated at 21°C on a 12-h light: 12-h dark photoperiod with a light intensity of approximately 100 μmol quanta·m–2·s–1 (Tang and Gobler 2009). All cultures used for experimental exposures were in an active exponential growth phase, which was maintained by a constant replenishment of fresh media in cultures.
Cell densities of C. polykrikoides in collected bloom water, nonbloom water, and dinoflagellate cultures (C. polykrikoides and G. aureolum) were quantified before each experiment from a 10-mL aliquot fixed with Lugol's iodine solution. Using a Sedgewick–Rafter counting chamber, three 0.25-mL subsamples were enumerated, and an average cell density (cells/mL) was calculated. If necessary, denser dinoflagellate cultures were diluted with fresh GSe medium so that similar cell concentrations could be achieved between treatments for experiments. Table 1 provides an inventory of all experiments conducted and the cell densities used.
| Replicates | Treatment | Survival (%) | CC (cells/mL) | TL (cm) | WW (g) |
|---|---|---|---|---|---|
| Larval Inland Silversides | |||||
| 20 | FSW | 75 | 0 | 0.47 ± 0.07 | na |
| Nonbloom | 80 | 1.1 × 102 | 0.45 ± 0.08 | na | |
| CP bloom | 0 | 3.1 × 103 | 0.37 ± 0.06 | na | |
| Larval Sheepshead Minnow | |||||
| 20 | FSW | 100 | 0 | 0.53 ± 0.05 | na |
| Nonbloom | 100 | 29 | 0.54 ± 0.06 | na | |
| CP bloom | 100 | 2.4 × 103 | 0.54 ± 0.06 | na | |
| Juvenile Atlantic Silversides | |||||
| 10 | GSe | 91 | 0 | 1.92 ± 0.12 | 0.04 ± 0.01 |
| CP1 | 44 | 9 × 102 | 1.90 ± 0.12 | 0.03 ± 0.01 | |
| Juvenile Inland Silversides | |||||
| 15 | GSe | 98 | 0 | 2.08 ± 0.26 | 0.05 ± 0.02 |
| CP1 | 98 | 12 × 102 | 2.02 ± 0.17 | 0.05 ± 0.01 | |
| Gym | 98 | 13 × 102 | 2.17 ± 0.21 | 0.06 ± 0.01 | |
| FSW | 96 | 0 | 2.11 ± 0.18 | 0.05 ± 0.01 | |
| Juvenile Sheepshead Minnow | |||||
| 12 | GSe | 100 | 0 | 2.00 ± 0.21 | 0.11 ± 0.04 |
| CP1 | 100 | 5.3 × 102 | 1.95 ± 0.19 | 0.10 ± 0.03 | |
| Gym | 100 | 11 × 102 | 1.97 ± 0.19 | 0.10 ± 0.04 | |
Fish maintenance and care
Fish are ideal model organisms for behavioral toxicity experiments, as they are relatively easy to culture and have high ecological relevance (Scott and Sloman 2004; Kane et al. 2005). In addition, because of their sophisticated olfactory systems (Noakes and Godin 1988; Tierney et al. 2010), they are particularly useful for assessing sublethal toxicity (Noakes and Baylis 1990; Kane et al. 2005).
All fish culture, handling, experimental design, and final disposal in this study followed procedures approved by the Institutional Animal Care and Use Committee at Stony Brook University (Stony Brook, New York). Furthermore, the study organisms were approved for toxicity research (USEPA 2002). Fish were maintained at the Stony Brook Southampton wet laboratory facility (Southampton, New York) for at least 2 weeks before use in any experiment. The recirculating seawater system had a total volume of 5.68 m3, and water treatment included two commercial protein skimmers (My Reef Creations [MRC], Atlanta, Georgia), one 0.19-m3 de-gas/biofiltration tower, one double-length FLV Hayward cartridge filter (400-μm filter sock), one MRC fluidized bed filter, and seven 130-W ultraviolet sterilizers. The marine laboratory intake seawater originated from eastern Shinnecock Bay (salinity ≈ 30‰), and the system received an approximately 50% water change over the course of 1 month.
Three forage fish species common to the U.S. East Coast were used in experiments: the Atlantic Silverside Menidia menidia, Inland Silverside M. beryllina, and Sheepshead Minnow Cyprinodon variegatus. Larval fish used in experiments were approximately 2 weeks posthatch (~0.5 cm TL; Table 1), and juveniles were approximately 2 months posthatch (~2 cm TL; Table 1). No Atlantic Silverside larvae were available for experiments, but juveniles were obtained from Shinnecock Bay by using a fine-mesh beach seine and were kept in the laboratory for 1 month before the experiment. Larval and juvenile Inland Silversides were purchased from hatchery stocks maintained by Aquatic Resource Organisms (Hampton, New Hampshire), while Sheepshead Minnow of both life stages were obtained from Aquatic BioSystems (Fort Collins, Colorado).
Larval and juvenile life stages of fish were raised separately with conspecifics at a density of about 200 individuals per rearing container. Larval fish were reared in 19-L culturing buckets, each containing an air stone, that were placed in a flowing seawater table. Buckets had three holes covered with fine-mesh screens to allow for continuous water flow through the buckets but prevented fish escapes. Larvae were monitored several times daily for mortality and were fed an ad libitum ration of brine shrimp Artemia spp. nauplii. Any excess or accumulated organic material that settled to the bottom of the buckets was syphoned.
Juvenile fish were reared in separate, aerated glass aquaria (38 L; 33 × 19 × 11.5 cm) equipped with a filter and one automatic feeder (Petco Auto Fish Feeder; Petco, San Diego, California) containing Tetramin flake food (Tetra, Melle, Germany). Juveniles were also monitored several times daily for mortality and were automatically fed flakes at regular intervals during daylight (i.e., 0700, 1000, 1300, and 1600 hours). Water quality was maintained by weekly water exchanges (~20 L) with filtered seawater (FSW) from the wet laboratory. Laboratory rearing temperatures were maintained between 20°C and 25°C, and the photoperiod was maintained at 12 h light: 12 h dark. Mortality of fish during culture was low (<5%), and there were no visible signs of stress or fish disease in larval or juvenile rearing tanks. Fish that were used in experiments stopped receiving food approximately 12 h before the first trial of their respective experiments.
Experimental design
Experiments were designed so that fish served as their own behavioral controls, which greatly increased the statistical power by reducing interindividual variability (Kane et al. 2005). To accomplish this, postexposure activity levels of fish were compared with their baseline activity levels. All experiments followed the same general steps, but the treatments varied between experiments as described below (Table 1).
Larval fish experiments utilized four wells of clear, six-well polystyrene plates (individual well volume = 4 mL; well diameter = 3.7 cm; well height = 2.2 cm), while juvenile experiments utilized the bottom half of a polystyrene petri dish (total volume = 150 mL; diameter = 14 cm; height = 1.6 cm). These arenas were selected to encourage horizontal movement and prevent the vertical movement of fish, as the behavioral analysis software tracked fish in two dimensions. Before each trial, three fish were placed into each arena containing 2 mL of FSW (larval experiments) or 75 mL of GSe (juvenile experiments), and the fish were allowed to acclimate for 30 min without disturbance. Acclimation times (i.e., 30 min) were based on previous behavioral studies with these species and life stages (e.g., Weis and Weis 1974; Weltzien et al. 1999; Billerbeck et al. 2001; Lankford et al. 2001; Rosenfield et al. 2004). In addition, three fish were used in arenas so as to reduce the potential stress to these species when not in schools (Reinfelder and Fisher 1994; Billerbeck et al. 2001; Kane et al. 2005). Larval and juvenile fish were previously found to exhibit normal behavior, with no observable signs of stress and no mortality, at these volumes in the arenas for at least 96 h (i.e., length of the experiment; K. J. Rountos, unpublished observations). Normal behavior is defined here as fish exhibiting healthy swimming behaviors that were similar to those of fish in rearing tanks with no visible signs of stress. To minimize stress, larval fish were transferred to wells by means of a modified transfer pipet (Rountos et al. 2014); juveniles were transferred with a dip net and were only in contact with the net surface for a few seconds (Tierney 2011). No mortality occurred during fish transfer or acclimation.
Behavioral experiments were conducted on white video stages, each backlit with one light bulb (Ecosmart CFL spiral, 19 W [100 W equivalent], 120 V); a high-definition video camcorder (Sony Handycam Model HDR-CX) was mounted above the video stage to record fish dorsoventrally. Although experiments were conducted in a lit laboratory during daytime hours, video stages were illuminated from the bottom to increase video contrast and eliminate shadows (Skjaeraasen et al. 2008; Herbert et al. 2011). Preliminary trials demonstrated that the video stage and arenas did not warm from the light bulbs over the experimental periods. Dividers were also used to isolate fish arenas from each other and to prevent peripheral visual disturbances such as those from human motion in the room. After an acclimation period of 30 min and depending on the life stage being tested, an additional 2 mL of FSW or 75 mL of GSe were added to each arena (total volume of larval arena = 4 mL; total volume of juvenile arena = 150 mL), and video recording of the baseline portion of the trial began. This procedure was followed in order to ensure that the fish experienced the same level of volume addition and pipetting activity between both the baseline and experimental portions of a trial. Fish were recorded without further disturbance for 20 min. Once the baseline portion was complete, 50% of the total volume was removed from each arena with a pipet and was replaced by the experimental treatment (Table 1). Fish were then recorded for another 20 min to complete the experimental portion of the trial. A completed trial produced a video that was at least 40 min in length (i.e., baseline portion = 20 min; experimental portion = 20 min). Because only three video cameras could be used at a given time, treatments were randomized across trials throughout an experiment to prevent any temporal bias. Fish in completed trials were left undisturbed (i.e., no aeration or food) on a laboratory bench for 24 h so that survival could be assessed. At experimental termination, the TL (nearest 0.1 cm) was recorded for larval and juvenile fish, and the mass (wet weight, nearest 0.01 g) of juveniles was also taken (Table 1).
Larval experiments (experiments 1 and 2) included three treatments: (1) the FSW in which they were reared (0 cells/mL), (2) C. polykrikoides bloom water (103 cells/mL), and (3) nonbloom water (<120 cells/mL; Table 1). Juvenile experiments did not utilize C. polykrikoides bloom water but instead used a combination of treatments, including clonal cultures of C. polykrikoides (henceforth, “CP1 treatment”); a nontoxic dinoflagellate control, G. aureolum (henceforth, “Gym control”); and two non-dinoflagellate controls (i.e., a GSe culture medium control and an FSW control; Table 1). For the Atlantic Silverside experiment (experiment 3), only a GSe culture medium control (0 cells/mL) and a CP1 treatment (9 × 102 cells/mL) were used (Table 1), since a nontoxic Gym culture was not available at the time of experimental initiation and fish numbers precluded the addition of an FSW control. A Gym control and an FSW control were subsequently added to experiment 4 (Inland Silverside juveniles) along with the GSe control and CP1 clonal culture treatments. Experiment 5 (Sheepshead Minnow juveniles) consisted of a CP1 treatment, a Gym control, and a GSe control. An FSW control was not possible for experiment 5 due to the unavailability of appropriately sized fish for that treatment. Table 1 provides the concentrations of all treatments used in the experiments. Although we attempted to include treatments consistently among all experiments, we reiterate that the availability of appropriately sized fish, the densities in algal cultures, and the access to natural bloom water restricted the treatments that could be used in each experiment. These limitations should be considered when making comparisons across experiments.
Exposure to Gym served as a nontoxic dinoflagellate control for experiments, as G. aureolum is similar in size, shape, and swimming behavior to C. polykrikoides (Tang et al. 2008; Tang and Gobler 2009). Additionally, an FSW control was used in the Inland Silverside experiment to assess whether fish behavior in the GSe culture medium control was different than behavior in FSW, which was used to rear the fish at the wet laboratory. To ensure that dissolved oxygen was at normoxic levels in all treatments before experiments, flasks containing the different treatment stocks (i.e., GSe, CP1 clonal culture, etc.) were measured with a self-stirring biochemical oxygen demand probe connected to a YSI Model 5100 benchtop dissolved oxygen meter (YSI, Inc., Yellow Springs, Ohio).
Video processing and analysis
All video files in MPG format (.mpg; frame rate = 30 Hz) were converted to AVI format (.avi) for analysis with LoliTrack version 4 software (Loligo Systems, Viborg, Denmark). Videos were converted to AVI by means of video conversion freeware (Any Video Converter version 5.0.6; AnvSoft, Inc.), and each 20-min portion (i.e., baseline or experimental) of the video was cropped to 10 min by removing the first 5 min and last 5 min of each video. Video cropping was done to remove the influence of pipetting activities on fish behavior. Ten-minute assessments of fish behavior were used to determine the immediate behavioral response of fish to HABs, given the knowledge that longer-term exposures (i.e., hours) to C. polykrikoides have caused significant mortality in these fish species (e.g., Gobler et al. 2008; Tang and Gobler 2010; Rountos et al. 2014).
Edited videos were loaded into LoliTrack software, and pixel distances were calibrated with the diameter of the fish arenas. This ensured that distances calculated between videos were standardized. Fish were detected in each video by adjusting the red, blue, and green color scales. To minimize errors in fish detections within and between videos, consistent general settings for masking, dilations, filtering, and color scales were used to track fish in all videos from each experiment. Videos were then tracked and analyzed with the combination of settings that produced the fewest tracking errors based on preliminary studies with the LoliTrack error diagnostic tracker. Tracking errors can occur for a variety of reasons, such as when the fish disappears from the tracking program by blending into the background color, when a piece of detritus is large enough and similar in color to the fish, or when the fish are very close together (i.e., causing the software to detect them as one fish).
After tracking, the software quantified a summary of the swimming behavior of each individual fish in the video, including the total distance swum (cm), total time for which the fish was active (s), mean speed (cm/s), and mean acceleration (cm/s2). These variables were selected because they are appropriate measures of general fish behavioral characteristics and have been used in previous behavioral experiments employing LoliTrack software (Skjaeraasen et al. 2008; Herbert et al. 2011; Poulsen et al. 2011). Figure 1 provides example images from the tracked videos. Fish were identified by the software as being “active” as long as their movement in one video frame exceeded a threshold distance of approximately 2% of their mean body length in any direction (~0.01 × 0.01 cm for larvae; ~0.03 × 0.03 cm for juveniles). The software calculated fish speed only when a fish was considered “active.” All video experiments were analyzed with a laptop computer (Lenovo T410 with the Windows 7 operating system).
Examples of fish tracks from the experiment with juvenile Inland Silversides: (A)–(D) initial detection of fish by LoliTrack software, (E)–(H) tracking of fish during the baseline portion of the trial, and (I)–(L) tracking of fish during the experimental portion of the trial. Treatments included a filtered seawater control (FSW); a GSe culture medium control (GSe); Gymnodinium aureolum clonal culture as a nontoxic dinoflagellate control; and Cochlodinium polykrikoides clonal culture as a harmful dinoflagellate treatment. [Color figure available online.]
Data analysis and statistical approach
Behavioral data for the three fish in each arena were averaged to obtain one mean value of fish behavior for each video segment (baseline and experimental). Replicates (i.e., independent arenas) were then averaged for each treatment and video portion. All statistical analyses were conducted with R statistical software (version 3.1.3; www.R-project.org). Two-way, repeated-measures ANOVAs were used to determine statistical significance between time periods (i.e., baseline versus experimental) and treatments; these were followed by a post hoc Tukey's honestly significant difference test for multiple comparisons if significant differences were found for each experiment. Statistical tests were run separately for each experiment and each behavioral variable. Results from all tests were considered statistically significant at P-values less than 0.05. Data that did not meet assumptions for multivariate normality (i.e., Henze–Zirkler's multivariate normality test, MVN package; Korkmaz et al. 2014) and homogeneity of variance (i.e., Levene's test, “car” package; Fox and Weisberg 2011) were transformed until these assumptions were met (Sokal and Rohlf 1995). Sphericity was assessed via Mauchly's test (Mauchly 1940), and if assumptions of sphericity were violated, a Greenhouse–Geisser correction was applied (Girden 1992).
RESULTS
Larval Behavior Experiments
No significant differences were found in the total distance swum by fish, the total time for which the fish were active, the mean speed, or the mean acceleration among treatments in the larval Inland Silverside or Sheepshead Minnow experiments (Table 2). Survival of Inland Silversides at 24 h was 75% for the FSW control, 80% for the nonbloom treatment, and 0% for the C. polykrikoides bloom treatment. No mortality occurred in the Sheepshead Minnow larval experiment (Table 1).
| Treatment | Distance swum (cm) | Total time active (s) | Swim speed (cm/s) | Acceleration (cm/s2) | ||||
|---|---|---|---|---|---|---|---|---|
| B | E | B | E | B | E | B | E | |
| Larval Inland Silversides | ||||||||
| FSW | 139 ± 62 | 115 ± 67 | 123 ± 46 | 93 ± 43 | 1.10 ± 0.15 | 1.24 ± 0.50 | 52 ± 11 | 57 ± 14 |
| Nonbloom | 105 ± 43 | 110 ± 52 | 97 ± 37 | 97 ± 45 | 1.08 ± 0.13 | 1.14 ± 0.15 | 60 ± 16 | 60 ± 13 |
| CP bloom | 115 ± 62 | 114 ± 45 | 99 ± 52 | 105 ± 37 | 1.21 ± 0.43 | 1.09 ± 0.14 | 62 ± 28 | 60 ± 16 |
| Larval Sheepshead Minnow | ||||||||
| FSW | 333 ± 79 | 353 ± 129 | 169 ± 35 | 172 ± 45 | 1.96 ± 0.22 | 2.02 ± 0.45 | 38 ± 9 | 42 ± 23 |
| Nonbloom | 378 ± 107 | 446 ± 134 | 179 ± 47 | 204 ± 47 | 2.11 ± 0.18 | 2.15 ± 0.22 | 42 ± 9 | 41 ± 6 |
| CP bloom | 392 ± 96 | 425 ± 137 | 189 ± 44 | 210 ± 49 | 2.08 ± 0.25 | 2.00 ± 0.33 | 41 ± 8 | 41 ± 14 |
| Juvenile Atlantic Silversides | ||||||||
| GSe | 1,289 ± 594 | 1,148 ± 747 | 335 ± 85 | 292 ± 106 | 3.65 ± 1.09 | 3.54 ± 1.31 | 75 ± 22 | 71 ± 20 |
| CP1 | 1,249 ± 666 | 2,233 ± 953 | 334 ± 110 | 440 ± 103 | 3.54 ± 1.05 | 4.85 ± 1.16 | 74 ± 19 | 82 ± 19 |
| Juvenile Inland Silversides | ||||||||
| GSe | 876 ± 522 | 678 ± 419 | 119 ± 62 | 94 ± 52 | 7.18 ± 0.66 | 6.98 ± 0.54 | 181 ± 20 | 183 ± 21 |
| CP1 | 745 ± 704 | 1,688 ± 934 | 99 ± 83 | 206 ± 92 | 7.00 ± 0.89 | 7.82 ± 1.13 | 178 ± 12 | 181 ± 23 |
| Gym | 1,625 ± 734 | 2,130 ± 881 | 196 ± 68 | 245 ± 71 | 7.86 ± 1.20 | 8.41 ± 1.25 | 174 ± 17 | 174 ± 17 |
| FSW | 771 ± 459 | 473 ± 372 | 109 ± 55 | 69 ± 49 | 6.75 ± 0.82 | 6.45 ± 0.90 | 173 ± 14 | 170 ± 16 |
| Juvenile Sheepshead Minnow | ||||||||
| GSe | 2,983 ± 878 | 3,293 ± 1,074 | 460 ± 64 | 466 ± 105 | 6.36 ± 1.23 | 7.32 ± 1.61 | 115 ± 24 | 112 ± 26 |
| CP1 | 3,696 ± 963 | 4,136 ± 1,366 | 501 ± 30 | 507 ± 44 | 7.31 ± 1.49 | 8.01 ± 2.07 | 124 ± 17 | 130 ± 19 |
| Gym | 3,879 ± 1,065 | 4,604 ± 1,699 | 503 ± 30 | 500 ± 72 | 7.64 ± 1.65 | 8.96 ± 2.37 | 122 ± 18 | 122 ± 15 |
Juvenile Behavior Experiments
Atlantic Silverside juveniles
The juvenile Atlantic Silversides that were exposed to CP1 at a concentration of 9 × 102 cells/mL swam significantly greater distances on average (mean = 2,233 cm; SD = 953) than during their baseline control period (mean = 1,249 cm; SD = 666; F1, 9 = 20.11, P = 0.03; Figure 2). Fish in the GSe control swam similar distances in both the baseline and experimental portions (i.e., mean = 1,289 and 1,148 cm, respectively; F1, 9 = 20.11, P = 0.97; Figure 2). No significant differences were observed in mean speed, the total time the fish were active, or mean acceleration (Table 2). Survival at 24 h was 91% for Atlantic Silverside juveniles in the GSe control and 44% for those in the CP1 treatment (Table 1).
Mean distance swum (cm; ±95% confidence interval) by juvenile Atlantic Silversides in the GSe culture medium control (GSe) and the Cochlodinium polykrikoides clonal culture treatment (CP1) during baseline (B) and experimental (E) portions of the experiment. Different lowercase letters indicate statistically significant differences (P < 0.05) based on two-way, repeated-measures ANOVA and post hoc Tukey's honestly significant difference test for multiple comparisons.
Inland Silverside juveniles
The distance swum (mean = 1,688 cm; SD = 934) by Inland Silverside juveniles when exposed to a CP1 concentration of 12 × 102 cells/mL was more than twice their baseline distance (mean = 745 cm; SD = 704; Figure 3; Table 2). The difference between baseline and experimental portions was statistically significant (loge transformed data; F3, 42 = 17.39, P = 0.004). No significant differences in the total distance swum were found between baseline and experimental portions for the GSe culture medium control (loge transformed data, F3, 42 = 17.39, P = 0.99), FSW control (loge transformed data, F3, 42 = 17.39, P = 0.92), or Gym control (loge transformed data, F3, 42 = 17.39, P = 0.42). Fish that were exposed to CP1 were also significantly more active than their baseline control activity (Greenhouse–Geisser correction is denoted as ê = 0.71, F2.13, 29.77 = 12.77, P < 0.001). No significant differences in mean swimming speed or acceleration were detected (Table 2), and Inland Silverside juveniles belonging to all treatments exhibited high survival (≥96%) after 24 h (Table 1).
Mean distance swum (cm; ±95% confidence interval) by juvenile Inland Silversides in the GSe culture medium control (GSe), the Cochlodinium polykrikoides clonal culture treatment (CP1), the nontoxic dinoflagellate control (Gymnodinium aureolum clonal culture; Gym), and the filtered seawater control (FSW) during baseline (B) and experimental (E) portions of the experiment. Different lowercase letters indicate statistically significant differences (P < 0.05) based on two-way, repeated-measures ANOVA and post hoc Tukey's honestly significant difference test for multiple comparisons.
Sheepshead Minnow juveniles
No mortality of Sheepshead Minnow juveniles was observed in any treatment 24 h after exposure (Table 1). No statistically significant differences were found for the total distance swum (F2, 22 = 0.46, P = 0.64), mean swimming speed (Greenhouse–Geisser correction applied, ê = 0.68, F1.35, 14.88 = 0.53, P = 0.53), mean total time for which the fish were active (F2, 22 = 0.05, P = 0.95), or mean acceleration (F2, 22 = 0.46, P = 0.53; Table 2).
DISCUSSION
This study provides the first assessment of the effects of C. polykrikoides exposures on larval and juvenile fish behavior. A clear ontogenetic difference was found in the behavior of Inland Silversides after exposure to C. polykrikoides, but no differences in the behavioral response were apparent between life stages of Sheepshead Minnow. Juvenile Inland Silversides exposed to concentrations of CP1 suffered little mortality and swam more than twice as far as their baseline after exposure to C. polykrikoides concentrations. In contrast, larval Inland Silversides exposed to significantly higher bloom concentrations (i.e., approximately three times greater) did not show any immediate behavioral effects but suffered complete mortality (Table 1). Juvenile Atlantic Silversides demonstrated a significant immediate behavioral response (i.e., increase in distance swum) similar to that of Inland Silverside juveniles, but they did not exhibit substantial mortality until after 24 h of exposure (Figure 2; Table 1). To date, only one other study (Rountos et al. 2014) has assessed temporally dynamic exposures to C. polykrikoides and the sublethal impacts on fish.
Previous studies have demonstrated rapid ichthyotoxicity of this HAB species for these fishes regardless of life stage (Tang and Gobler 2009; Rountos et al. 2014); therefore, it is unlikely that the observed differences in the behavior between larval and juvenile Inland Silversides were the result of differential toxicity of C. polykrikoides. In fact, smaller individuals are more susceptible to toxicity than larger fish (Tang and Gobler 2009), yet larval Inland Silversides and larval Sheepshead Minnow displayed no immediate behavioral effects after exposure. Importantly, this behavioral response was also not simply due to the presence of dinoflagellates, as no statistically significant differences were found in the distance swum for juvenile Inland Silversides and Sheepshead Minnow exposed to the nontoxic dinoflagellate G. aureolum (Gym control; Table 2). This suggests that a factor related to C. polykrikoides exposure—likely its toxicity—elicited the behavioral responses we observed and that the responses were not just attributable to dinoflagellate presence.
Fish use a variety of sensory modalities to detect toxicants in their environment (Hara 1994; Tierney 2016). Additionally, there are important differences between larval and adult life stages in terms of the routes of toxicant uptake (Kane et al. 2005) and sensory system functioning and capacity (Fuiman and Magurran 1994; Kasumyan 2011). The observed ontogenetic differences in behavior may be the result of several factors and not limited to differences between life stages in terms of the surface area of gill epithelial tissues (Kane et al. 2005). Adult and juvenile fish have larger gill surface areas for respiration than larvae, which utilize the skin as the major respiratory interface (Kane et al. 2005). Previous histological assays have confirmed that C. polykrikoides causes significant hyperplasia and fusion of gill lamellae in adult fish (Gobler et al. 2008); thus, it is possible that the differential ontogenetic behavioral responses observed in this study may have resulted from the reduced gill tissue surface area in larval conspecifics. Regardless of the sensory modality employed, the inability of larval Inland Silversides or larval Sheepshead Minnow to detect the toxicity of C. polykrikoides, and subsequently to display greater activity, could have substantial ecological implications for this life stage. Short-term and ecologically relevant exposure times have already been demonstrated to cause significant mortality and the loss of swimming ability (Rountos et al. 2014), so the lack of a behavioral response in exposed larvae is likely to have high ecological relevance.
Toxic dinoflagellate blooms have been shown to alter the relative abundances and distribution of fish in estuaries (Bauman et al. 2010; Friedland et al. 2011; Hallett et al. 2016). For instance, Friedland et al. (2011) found that the CPUE of Atlantic Menhaden Brevoortia tyrannus, an abundant neritic forage species in the York River estuary (Virginia), was inversely related to the C. polykrikoides concentration in the water column during bloom events. Atlantic Menhaden CPUE was negligible during blooms relative to the CPUE obtained when blooms were absent, suggesting that these fish might avoid and actively swim out of blooms in the estuary (Friedland et al. 2011). A similar response was detected during an extensive (>500-km2) bloom of C. polykrikoides in the Gulf of Oman. Bauman et al. (2010) found substantial declines (i.e., 72–86%) in fish abundance (number of fish per 150 m2) when comparing surveys of coral reef fish before and during the C. polykrikoides bloom. Furthermore, anecdotal evidence from fishers also suggests that C. polykrikoides blooms may affect the distributions of wild fish populations. Fishers in Rhode Island reported that they no longer fished a particular cove during the summer months when C. polykrikoides blooms were occurring in that area, as baited traps set during blooms were always empty (Tomas and Smayda 2008). Similarly, pound-net fishers in Shinnecock Bay have reported negligible or no catch of fish when C. polykrikoides blooms occur in close proximity to their nets (J. Semlear, Southampton Town Trustee, Southampton, New York, personal communication). Our results generally support these field observations by providing the first evidence that juvenile Atlantic Silversides and Inland Silversides change their swimming behavior when exposed to concentrations of HABs. However, this response is not ubiquitous among all fish species, as no behavioral response was seen for Sheepshead Minnow juveniles that were exposed to lower concentrations of C. polykrikoides.
Exposures to the harmful dinoflagellate C. polykrikoides caused juvenile Atlantic Silversides and Inland Silversides to swim significantly further distances compared with their pre-exposure baseline activity. Atlantic Silversides had a similar behavioral response when exposed to an insecticide (Weis and Weis 1974) and to copper (Koltes 1985). Although fish species possess a diversity of sensory systems covering a variety of modalities (Hara 1994), the majority of responses to natural contaminants are aversive (Tierney 2016). Swimming greater distances in response to C. polykrikoides exposures may be a physiological effect due to toxicity or an effective response evoked by a sensory cue to remove themselves from exposure, as these blooms are heterogeneous in their spatial extent (Gobler et al. 2008; Mulholland et al. 2009; Tang and Gobler 2010; Koch et al. 2014). Sheepshead Minnow juveniles, on the other hand, were not affected by exposures to C. polykrikoides based on the behavioral variables tested, although these fish were exposed to lower concentrations. This finding was not overly surprising, as Rountos et al. (2014) demonstrated that Sheepshead Minnow were more resistant to C. polykrikoides than were Atlantic Silversides and Inland Silversides of the same life stage.
Ultimately, a greater understanding of the sublethal behavioral effects on fish from exposures to HAB species like C. polykrikoides is needed, both in the laboratory and in the field. A thorough assessment of the fish behavioral effects has important implications for improving our ability to sustainably manage fisheries in regions with these recurrent blooms (Burkholder 1998; Kudela and Gobler 2012).
ACKNOWLEDGMENTS
We are grateful to R. M. Cerrato, M. G. Frisk, and T. E. Essington for fruitful discussions that improved the manuscript. We acknowledge financial support provided by the Institute for Ocean Conservation Science, the Suffolk County Department of Health Services, and the Shinnecock Bay Restoration Program. The Shinnecock Bay Restoration Program received major funding from the Laurie Landau Foundation and the Jim and Marilyn Simons Foundation.
