Behavioral Evidence for a Chemical Disturbance Cue in Bat Rays (Myliobatis californica)

ABSTRACT

The release of waterborne chemical cues is a well‐documented anti‐predator strategy in teleost fish, allowing individuals to detect, and respond to, predation threats. However, little is known about whether an ancient group of fishes, elasmobranchs, employ similar mechanisms to communicate predation risk. Here, we experimentally investigate the behavioral and physiological responses of Bat rays (Myliobatis californica) following the manipulation of an upstream conspecific. Specifically, we examined whether Bat rays receiving water from an upstream conspecific that had been physically chased with a stick for 30 s, simulating a non‐injurious predation disturbance, exhibited changes in activity levels and physiological stress markers. Receiving Bat rays increased their velocity by approximately 3 cm/s (~21%) compared to their pre‐exposure state, whereas Control rays showed no significant change. Receivers also exhibited a significant shift toward active swimming behaviors, while Controls spent more time in passive behaviors. This shift was positively correlated with velocity, reinforcing that exposure to waterborne disturbance cues from a distressed conspecific triggered increased activity. Physiological measurements showed no significant differences between groups in glucose, β‐HB, lactate, or pH levels. We suggest the activity increase corresponds with a flight‐type response to predation risk. These findings reveal a previously undocumented communication mechanism in marine predators, contributing to a broader understanding elasmobranch behavior and physiology.

Chased Bat rays may release chemical cues that alert conspecifics of predator threats. In our experiment, Receiver rays exposed to water from a chased “Sender” ray increased swimming speed and shifted to more active behavioral states.

Summary

California Bat Rays exposed to waterborne cues from chased conspecifics increased swimming velocity by ~21%, providing evidence for a disturbance cue mediated behavioral response in elasmobranchs.

1. Introduction

Information exchanged between individuals or groups of animals facilitates and coordinates behaviors critical for survival (Klump et al. 1986; Klump and Shalter 1984). In public information use, a sender animal may directly or inadvertently relay auditory, visual, electrical, or chemical signals and cues (Klump et al. 1986) containing information about its behavioral or physiological state, the environment, or a combination of the two, to a receiver animal who decodes the information and can respond. This exchange of information can help animals coordinate to find food, mates, or avoid predators (Leavesley and Magrath 2005).

Disturbances, such as predation attempts, displacements, or dangerous environmental conditions, can drive natural selection, and many organisms respond to such threats by communicating the risk to nearby individuals through a disturbance or alarm cue (Klump et al. 1986; Klump and Shalter 1984; Leavesley and Magrath 2005; Verheggen et al. 2010), which are common in terrestrial (e.g., avian and mammalian) systems (Klump and Shalter 1984; Leavesley and Magrath 2005; Alatalo and Helle 1990). In aquatic systems, many teleosts (bony fishes) utilize chemical alarm cues (ACs), where an individual releases a chemical cue in response to a threat nearby (Spinella et al. 1993; Vadas and Elner 2003; Watson et al. 2005; Wisenden and Millard 2001). ACs are released when an individual sustains tissue injury during a predation attempt, as seen in Ostariophysan fishes (Verheggen et al. 2010). In these species, chemical extracts from macerated skin tissue trigger behavioral responses in conspecifics (Brown et al. 2000; Friesen and Chivers 2006; Pfeiffer 1978; Wisenden and Chivers 2006; Wisenden and Barbour 2005). In other species, the release of a chemical cue can occur upon threat recognition (e.g., predator) without injury; called disturbance cues (DC) (Wisenden et al. 1995; Jordão and Volpato 2000). For example, Iowa darters (Etheostoma exile) receiving water from upstream conspecifics chased by a model predator increased vigilance relative to control darters (Wisenden et al. 1995). Under both circumstances, chemical alarm cues and disturbance cues (AC/DC) are adaptive as they alert nearby individuals of potential danger and trigger a behavioral response.

AC/DC can also trigger a physiological response that includes activation of metabolic and stress axes (e.g., sympathetic or hypothalamic‐pituitary‐interrenal (HPI)). Activation of the HPI presumably supports the response behavior with energy mobilization (e.g., increased circulating glucose) and may occur in response to or in concert with fast, anaerobic production of ATP (as indicated by increased lactate and decreased blood pH) during intensive behaviors. For example, physiological changes in response to AC that indicate activation of the HPI have been observed in several teleost fish species, such as European minnows (Phoxinus phoxinu), Pearl dace (Semotilus margarita), and Coho salmon (Oncorhynchus kisutch) (Barcellos et al. 2011; Pfeiffer and Lamour 1976; Rehnberg et al. 1987; Rehnberg and Schreck 1987). Additionally, increased cortisol levels were observed in phylogenetically distant teleost species—Jundiá (Rhamdia quelen) and Nile tilapia (Oreochromis niloticus)—after exposure to DC released by predator‐stressed conspecifics (Barcellos et al. 2011). These studies of phylogenetically distant species of teleost fish suggest AC/DC may be part of an adaptive and ancient mechanism for teleost fish survival (Barcellos et al. 2011).

Elasmobranchs (cartilaginous fishes) represent an ancient, early‐diverging group of fishes that often occupy a higher trophic position than teleost fishes. However, elasmobranchs can also be prey, particularly at early life stages (Gervais et al. 2021; Dedman et al. 2024). As such, similar to teleosts, AC/DC may allow for a behavioral and physiological response. Studies of elasmobranch fishes that provide insights into the evolutionary development of AC/DC are lacking (Crane et al. 2021). In fact, only one study has investigated an elasmobranch AC mechanism under controlled conditions. Port Jackson sharks (Heterodontus portusjacksoni) were found to decrease oxygen uptake rates when exposed to necrotic tissue extracts derived from dead conspecifics (Gervais and Brown 2021). Similarly, a single field study found concentrated putrefied shark tissue administered via canister deployments resulted in decreased feeding activity in Caribbean reef (Carcharhinus perezi) and Blacknose sharks (C. acronotus) (Stroud et al. 2014). While these studies have evaluated responses triggered by necrotic tissue, no studies have investigated a DC mechanism in elasmobranchs in a non‐injurious context.

Although targeted research on AC/DC in elasmobranchs remains largely unexplored, behavioral evidence suggests even large predatory elasmobranchs may utilize an AC/DC strategy to avoid predation. Acoustic tagging data of White sharks (Carcharodon carcharias) in California demonstrated relocation and a flight response from the Southeast Farallon Island in response to brief visits from a transient group of Orca whales (Orcinus orca), which are known to specialize in eating sharks (Jorgensen et al. 2019). Though the Orcas were present for only 2.5 h, the sharks continued to depart for ~8 h and did not return until the following year. The authors hypothesized a chemical cue as the most parsimonious driver of the flight response (Jorgensen et al. 2019).

Inspired by this hypothesis, and the lack of chemical cue studies in this evolutionarily distinct group, we conducted a laboratory experiment to test for the release of DCs in a meso‐predatory elasmobranch, the California Bat ray (Myliobatis californica; CBR). CBR are good models for investigating a DC strategy, being commonly predated by marine mammals and other larger species of elasmobranch (Matern et al. 2000). CBR aggregate in areas with considerable distributional overlap with their known predators (Odenweller 1975; McInturf et al. 2023), suggesting chemical cues could be ecologically relevant, and the species can be successfully housed in captivity.

We hypothesized CBR utilize DCs similar to teleost fish when chased in simulated predation disturbance. If CBR use DCs to alert conspecifics to disturbance, then we predict CBRs housed in a tank downstream (Receivers) of an uninjured, but disturbed animal (Sender), will show measurable behavioral and physiological changes. Specifically changes in swimming speed and behavioral activity budgets as well activation of the HPI. Unlike teleost fishes, in which cortisol functions as the dominant glucocorticoid associated with stress‐axis activation, a glucocorticoid equivalent to cortisol has not yet been unequivocally identified in elasmobranchs. As such, we assessed secondary physiological indicators of stress‐axis activation (e.g., increased circulating glucose, BHB, lactate, and decreased pH) that prepare them to cope with or avoid a potential predator interaction rather than direct corticosteroid responses. We chase CBR with a stick to simulate a predation attempt rather than utilize a real predator (e.g., mount) or induce injury to ensure that chemical cues originate from the Sender, rather than from the predator or from tissue damage. This is the first study to experimentally investigate the release of DC and the associated behavioral and physiological responses in an elasmobranch, thereby providing additional insights on elasmobranch behavioral and physiological responses during a simulated predation disturbance.

2. Materials and Methods

We used a repeated measures design to determine if chemical cues from an upstream CBR could alter behavior or blood physiology of downstream CBR. Juvenile CBR ( ~ 1 year old, seven male and eight female, wingspan range 31–38 cm) were sourced from the Oregon Coast Aquarium's captive breeding program.

2.1. Animal Care

Experiments were conducted at Hatfield Marine Science Center (HMSC), Newport, Oregon, USA (44°37′ N, 124°02′ W), in the Experimental Seawater Facility under Oregon State University animal care protocol IACUC‐2022‐0251. Air stones were placed in each tank to ensure adequate aeration. All CBRs were fed daily during the experiment to continue their normal feeding regime and avoid incidental activation of the HPI through fasting. On experimental days, CBRs were fed following the completion of all experimental trials.

2.2. Experimental Housing and Water Flow

Experimental housing and water flow was based on similar studies with teleost fish (e.g., Wisenden et al. 1995). Specifically, CBR were housed individually in 8 circular, 2.5 m diameter tanks. All tanks were filled with seawater pumped at high tide from Yaquina Bay, Oregon at ambient conditions from the seawater reservoir at HMSC. Water flowed naturally from the head tank (Tank 1) to two downstream tanks (Tanks 2 and 3) where it drained back into the recirculating life‐support station (Figure 1). See supporting materials for additional animal monitoring and housing design details.

Figure 1.

A diagram of tank configuration showing Treatment and Control set configurations for trials 1 (a) and 2 (b) within the first experimental Round. (a) In the Treatment set for trial 1 water flowed from the head tank (1‐s) to the two receiving tanks (2‐r and 3‐r); (b) in trial 2, the signal tank in the Treatment set was switched to Tank 2 (2‐s), and tanks 1 and 3 were receiving tanks (1‐r and 3‐r). An additional set of Treatment tanks in each experimental Round, not shown here (Supporting Information Figure 1S), was configured similarly. During all trials, the Control set configuration was unchanged and water flowed from the empty tank (ET) to the two receiving controls (7‐c and 8‐c).

Tanks were arranged into two sets: one Treatment set and one Control set (Figure 1a,b). Each tank within a set housed a single CBR, except for the empty head tank (ET) in the Control set. Tanks within a set were connected by directional recirculating water. In the Treatment set, water flowed passively (i.e., gravity‐fed) from an upstream head tank (Sender; Tanks 1‐s or 2‐s; Figure 1) to a splitter that directed water to two downstream receiving tanks (Receivers; Tanks 2‐r, 3‐r, or 1‐r, 3‐r, depending on the trial; Figure 1). Tank configuration allowed for the head tank to be alternated between trials by adjusting valves, thus switching which CBR acted as the Sender or Receiver without moving the animals. This manipulation of water flow direction enabled testing of different individuals as Senders across trials. In the Control set (7‐c and 8‐c; Figure 1), water flow direction and experimental disturbance was identical to the Treatment sets with the only difference being the head tank (ET) did not contain a CBR. Tanks were insulated and lifted off the floor with 7.5 cm thick closed cell polyurethane foam and separated from each other by approximately 0.5 m distance to minimize sound and vibrational cues. A black opaque barrier was drawn around each tank to eliminate potential visual cues coming from another tank or the room (Michael‐Romero 2002). Water entering downstream tanks fell from a height of 15 cm to eliminate any transmission of acoustic cues between tanks. A single video camera (1080p 60fps web camera) was attached to the ceiling directly above each tank and operated remotely to avoid disturbing animals during experimental trials. Each tank fit within the view frame of the camera.

2.3. Procedures

Experimental trials were conducted from August to December 2022 in two rounds (Round1, n = 8; Round2, n = 7). CBR were acclimated for approximately 1 month to allow all rays to adjust to housing conditions at HMSC, as is typical in captive studies in elasmobranchs [31]. After this initial acclimation period, a baseline blood sample was collected from each animal (pre‐baseline1). There was an unexpected power outage on the day of baseline sample collection during Round1. While all measured water quality parameters and feeding rates were within normal ranges, and therefore, physiological responses were likely not altered due to this outage, we allowed animals to acclimate another 17 days, after which a second baseline blood sample was collected (pre‐baseline2). The first experimental trial began 1 week after the second baseline sample. Each animal was a Sender once and a Receiver twice for a total of 3 experimental trials per round (except for Tanks 7/8 in Round2 because the system only allowed for 2 animal tanks; Figure 1). Animals were allowed to reacclimate for 1 week between trials to allow a return to baseline physiological values [32]. Order of Sender for each individual was included in the analysis to account for potential habituation or sensitization effects. Acclimation time and baseline blood sampling periods were mirrored for CBR in Round2 to keep procedures consistent between Rounds (see Supporting Information Figure S1 for experimental timeline).

For each experimental trial, all animals were video recorded in three phases: 15 min immediately before the simulated predation disturbance, during the simulated predation disturbance, and 15 min following the disturbance. A 15 min recording period was chosen to match the time interval used in a similar experiment performed on Iowa Darters (Wisenden et al. 1995). Blood sample collection during trials for senders, receivers and controls is described below (see Section 2.4 Blood Sample Collection).

Following the conclusion of a trial, water direction was changed so a different CBR became the Sender (e.g., Figure 1a,b) and animals were allowed to re‐acclimate for 1 week prior to starting the next trial. Note size limitations on one of the Receiver tanks in Round2 prevented us from holding a second Receiver animal in one treatment set. Following the third trial in an experimental round, animals were allowed to re‐acclimate for 1 week before a post‐experiment baseline blood sample was collected from all CBR. All experimental procedures took place late morning to early afternoon (10:00–14:00).

Sender: During each trial, the Sender (note we reference labeling from Figure 1a as an example throughout), was physically chased with a stick for 30 s to simulate a predation disturbance. The Receiver CBR received the water flowing from the signal tank to the receiving tank (potentially containing chemical cues). Five min after the start of the chasing event, the Sender CBR was removed from the tank and a 1 mL blood sample was collected to determine how chasing impacted physiological metrics (see Blood Processing Methods below). The Sender CBR was then placed in a separate, disconnected, holding tank to avoid any additional chemical cues from the blood sampling event being transmitted to Receiver CBR.

Receiver: After the 15 min post‐chase period of behavioral observation, Receivers were captured. Water was simultaneously shut off on the outgoing lines during capture to prevent transmission of any materials released from capturing to the other Receiver. Blood samples were immediately collected using the same methods as for the Sender animal (see blood processing methods below). Following blood sampling of the Receiver CBR, the trial was considered complete, and all CBR were returned to their original tanks.

Control: The repeated measures design allows for each Receiver to act as its own control, but we included at least one Control set per trial to ensure behavioral or physiological changes were not induced by the simulated predation disturbance protocol. The Control was adjacent to the experimental Treatment sets to mimic experimental trials, but the head tank did not contain a Sender animal (Figure 1). In controls, an empty head tank was stirred with the stick for 30 s and the other two CBR received water from the stirred, empty tank. All other sampling and recording methods mirrored the experimental procedures. To minimize the chances of pseudoreplication, we alternated which tank set functioned as the Treatment and Control between rounds (Supporting Information Figures S2 and S3).

2.4. Blood Sample Collection

CBR were captured within 15 s using a rubberized net and transferred to a sampling station, which consisted of a smaller (0.5 × 1 × 0.25 m) aerated tank. CBR were then manually restrained upside down with their head and spiracles submerged underwater and blood samples were collected from the caudal vein. Blood samples were collected within 3 min from initiating capture with the net to limit capture‐induced impacts on circulating HPI metrics (Hoffmayer and Parsons 2001; Lawrence et al. 2018). In some cases (9 of 45), the 3 min time was exceeded and noted, though within these cases, only five samples exceeded 3.5 min. Time‐since‐capture was considered in analyses of each response variable.

2.5. Blood Physiology

Unlike teleost fishes, elasmobranchs do not produce or utilize cortisol as their primary stress hormone. Instead, elasmobranchs produce a unique dominant corticosteroid, 1α‐hydroxycorticosterone (1αOHB), for which a complete steroidogenic pathway has not yet been determined (Wiens et al. 2017). However, quantification of 1αOHB is hindered by an extremely limited supply of available standard (Bouyoucos et al. 2021). Due to the ambiguity surrounding the functional role of 1αOHB and challenges associated with its quantification, we measured several secondary HPI characteristics, including blood pH, lactate, glucose and ketones to determine if CBR respond physiologically to chemical cues (Bouyoucos et al. 2021).

A microprobe with a pH sensor (Accumet model AE 150, Fisher Scientific, USA) was used to measure pH from 20 μL of plasma. The probe was calibrated prior to use. All enzymatic assays were performed at the University of Manitoba, Canada, following (Schoen et al. 2021). Plasma glucose concentrations were measured using an enzymatic hexokinase assay following (Treberg et al. 2007). Plasma lactate concentrations were measured using an enzymatic lactate dehydrogenase (LDH) assay following (Morrison et al. 2020). Plasma 3‐HB concentrations were measured using an enzymatic 3‐hydroxybutyrate dehydrogenase (3‐HBDH) assay following (Schoen et al. 2021).

An initial validation was performed for all assays using six to eight randomly chosen samples. As determined by validations, a dilution factor was used to dilute all samples. All samples were run in duplicate and were rerun when the coefficient of variation between duplicate samples was above 15%. If a sample value fell below the negative control (water) or above the top standard detection range of an assay, the sample was rediluted and rerun. Every plate was run with serially diluted standards (1000 μM–31.25 μM). Interassay variation was assessed among assay plates (2.8%, 8.9%, & 4.1%, for 3‐HB, lactate, and glucose, respectively). All assays were run at room temperature (20°C) for 50 min and conducted using a plate reader (Power wave XS2, Biotek) with Gen5 software (Biotek, Winooski, VT, USA).

2.6. Kinematics

Video data were analyzed using a combination of manual and automated data coding. DeepLabCut (DLC) was used to label a single point on the CBR heads in each video to allow for tracking their position frame by frame (Mathis et al. 2018). The DLC pose estimation model was validated using 250 manually labeled frames (containing frames from all CBR used in the experiment) for 100,000 iterations and default parameters with a test error of 9.5 pixels and train error of 2.49 pixels (image size was 1920×1080 pixels). The trained model was then used to analyze all video segments to track animal movements frame by frame in (x, y) pixel dimensions.

A raw velocity calculation was performed using the nose_x and nose_y position data using Euclidean distance and the distance between labeled points on CBR over time. Velocity was then smoothed using a rolling mean over five frames. Improbable velocities (e.g., errors due to water surface disturbance caused by the air stones) were removed if they exceeded a predefined threshold of 40 cm·sec⁻¹ ( > 1.5× maximum published CBR swimming speed (Matern et al. 2000); Klimley et al. 2005). Data were then downsampled to 1 frame per second (fps) to reduce sampling error. Pixel distance was converted to centimeters using the known diameter of the tank each animal was housed in and the video's resolution. Water flow from the sending tank to receiving tanks took approximately 30 s, as determined by behavioral response of receiving CBR to food cue stimuli introduced to the sender's tank. Therefore, 30 s of animal behavior starting from the moment the simulated predation disturbance began was removed from the analysis to give time for the water traveling from the signaler's tank to receiving tanks. Similarly, the first 30 s of animal behavior was removed from the 15 min before period for consistency between before and after periods. Velocity (cm·sec⁻¹) was subsequently averaged for 14.5 min before and 14.5 min after the simulated predation disturbance (Wisenden et al. 1995).

2.7. Behavioral Coding

We developed an ethogram in the event‐logging Behavioral Observation Research Software (BORIS version 7.13.9) (Friard and Gamba 2016) to code CBR behaviors (Table 1). The ethogram was developed to encapsulate observed behaviors for CBR in captivity (Supporting Information Figures S4–S8) using distinct categories for each continuous behavioral state that were coded as a duration in seconds. Behavioral data were assessed using a randomized blind protocol to avoid bias in the annotation. Behavioral data were grouped by average time spent demonstrating each behavior minute⁻¹ for the recoding periods. Coded behaviors were then averaged for 14.5 min before/after the simulated predation disturbance (Wisenden et al. 1995).

Table 1.

Descriptions of observed behavioral states in CBR during tank observations.

Behavioral state	Description
Wall swimming	The CBR swims while maintaining close proximity to the tank wall and is traversing the perimeter with one wing angled upward toward the ceiling, while the other is directed toward the tank bottom.
Wall climbing	The CBR positions its ventral side against the tank wall, with its nose pointed vertically upward and its tail directed toward the bottom of the tank.
Laying on the bottom	The CBR rests, without moving, ventral side down against the bottom of the tank.
Bottom swimming	The CBR swims ventral side against the bottom of the tank.
Sub‐surface swimming	The CBR swims between the surface and the bottom of the tank.

A principal components analysis (PCA) was performed on behavioral state variables to identify relationships between all measured behavioral states and reduce them to two dimensions. To evaluate the behavioral interpretation of the principal components, Spearman's rank correlation was used to assess the relationships between average velocity and the first two principal components (PC1 and PC2).

2.8. Data Analysis

Differences in the behavioral PC1 and PC2 and velocity measurements before and after the simulated predation disturbance were tested using linear mixed effects models (LMM) in JMP Pro v17.0.0 with sex, treatment (i.e., Control vs. Receiver), repeated measurements (before vs. after the simulated predation disturbance), and the interaction of treatment and repeated measurements as fixed effects; and with animal ID, round, and experimental trial as random effects (Supporting Information Table S3). Random effects that were insignificant and explained < 5% of the error were removed.

Physiological parameters were assessed using an identical model structure using pre‐ and post‐ baseline values for before and after experiment measurements. LMM residual distributions were considered normal if skewness was less than 2 and kurtosis less than 7 (Kim 2013). All model residuals were normal.

3. Results

3.1. Velocity and Behavior

In total, 16 h of video data were manually coded using BORIS and 15 h of video were analyzed in DLC for 15 CBR. Across all individuals and treatments, observed velocities ranged from 4.2 to 28.6 cm·s⁻¹. The interaction between treatment group and time significantly impacted velocity (β = 0.79, p = 0.05), with Receiver CBR increasing their average velocity by approximately 3 cm·s⁻¹ (from 13.75 ± 0.95 cm·s⁻¹ to 16.63 ± 0.95 cm·s⁻¹; +~21%) following chasing the upstream CBR. In contrast, Control CBR showed a non‐significant small decrease of −0.29 cm·s⁻¹ in velocity (from 10.71 ± 0.83 cm·s⁻¹ to 10.43 ± 0.83 cm·s⁻¹; −2.7%) (Figure 2). CBR ID explained a large proportion of the total error variance that is not explained by the fixed effects ( ~ 68%), suggesting individuals differ in velocity. No differences were observed between male and female CBR for velocity, PC1, or PC2. Video recordings failed on 6 of 8 tanks (4 Receiver and 2 Control) during the first experimental trial in Round1 and for 2 Control tanks in the third experimental trial in Round2, resulting in a loss of velocity quantification for 8 of 45 total individual trials.

Figure 2.

Mean change in velocity (cm·s⁻¹) by treatment group (Control vs. Receiver). The dashed red horizontal line at y = 0 represents no change in velocity. Mean changes for each treatment group are displayed in the top left: Receivers increase velocity (2.88 cm·s⁻¹), whereas Controls showed a slight decrease (−0.29 cm·s⁻¹).

Principal component 1 (PC1, Eigenvalue = 2.05) captured the most variance (34.24%) in the data and was interpreted as a metric of activity level as it loaded passive behaviors (i.e., Laying on the Bottom and Wall Climbing) against more active swimming behaviors (i.e., Wall Swimming, Sub‐surface Swimming, Bottom Swimming; Supporting Information Figure S9). PC1 likely represents a gradient from Wall Climbing (negative loadings) to Wall Swimming/Sub‐surface Swimming (positive loadings). Principal component 2 (PC2, Eigenvalue = 1.48) captured an additional 25% of the variance for a cumulative total of 59% and loaded Laying on the bottom and Bottom swimming (positive loadings) against Wall Swimming, Sub‐surface Swimming, and Wall Climbing (negative loadings) and therefore indicate a switch from behavioral states associated with the bottom of the tank to behaviors higher in the water column (Supporting Information Figure S9).

PC1 was significantly impacted by treatment group and time (β = 0.31, p = 0.02). Receivers showed a larger increase in PC1 (mean change of 0.29 from 0.17 ± 0.28 to 0.46 ± 0.28) (Figure 3), indicating a shift toward more active behaviors, whereas Control CBR exhibited a mean decrease of −0.95 in PC1 (from 0.09 ± 0.41 to –1.04 ± 0.41) following the stirring of an empty upstream tank. Wall Swimming and Wall Climbing loaded strongly against one another for PC1, suggesting the decrease in Controls was driven by more time spent in the more passive Wall Climbing behavior whereas Receivers spent more time Wall Swimming, which is associated with faster swimming speeds. Indeed, PC1 was positively associated with velocity (Spearman's ρ = 0.57, p < 0.001; Supporting Information Figure S10) and visualizations of Receiver movement heatmaps using the DLC outputs reflect this, showing a behavioral shift to occupy and circle the perimeter of their tank (Supporting Information Figure S17).

Figure 3.

Mean change in PC1 by treatment group (Control vs. Receiver). The dashed red horizontal line at y = 0 represents no change in PC1. Mean changes for each treatment group are displayed in the top left: Receivers showed an increase in PC1 (0.29), whereas Controls exhibited a decrease (−0.95).

The PC2 of Receivers (mean change of –1.03 in PC2 scores from 0.46 ± 0.25 to –0.57 ± 0.25) tended to be lower than Controls (mean change of –0.42 from 0.31 ± 0.36 to –0.11 ± 0.36) (Figure 4) after stirring, though there was no interactive effect of treatment group and time on PC2 (β = −0.15, p = 0.57). However, there was a significant effect of time (β = −0.36, p = 0.02), where CBR tended to swim higher in the water column after stirring regardless of treatment. Unlike PC1, PC2 had no association with velocity (ρ = 0.13, p = 0.355; Supporting Information Figure S11).

Figure 4.

Mean change in PC2 by treatment group (Control: −0.42 vs. Receiver: −1.03). PC2 is associated with a switch from behavioral states associated with the bottom of the tank to behaviors higher in the water column. The dashed red horizontal line at y = 0 represents no change in PC2. Mean changes for each treatment group are displayed in the top left: Receivers showed a greater decrease in PC2 (−1.03) compared to controls (−0.42).

3.2. Physiological Results

Each of the 15 CBR had six blood samples collected across approximately 2 months (pre‐baseline1, pre‐baseline2, trial 1, trial 2, trial 3, and post‐baseline). The treatment order for each trial (i.e., Receiver vs. Sender vs. Control) was dependent on tank location and varied for each ray (Figure 1). No samples were obtained for CBR 4 baseline‐post or CBR 5 pre‐baseline1, for a total of 73 samples (see Supporting Information Table S2). Four physiological parameters for CBR plasma were assessed in this study (glucose, 3‐HB, lactate, and pH). In some instances, sample volumes were not sufficient to analyze all parameters (Supporting Information Table S2).

No significant differences were observed between Receivers and Controls for any of the physiological parameters assessed (Supporting Information Table S1), however there were some changes across time. Baseline glucose concentrations were significantly different (β 0.75, p = 0.009) from post‐baseline values (Supporting Information Figure S12). Glucose tended to be lower immediately after experimental trials for both Receivers (mean = 1.79 mM ± 0.75) and Controls (2.12 mM ± 1.04) and had less variation compared to post experiment baseline values (mean = 3.05 mM ± 1.67, 3.18 mM ± 1.45, respectively; Supporting Information Figure S12). Similar to glucose, 3‐HB levels decreased between the pre‐baseline1 and post‐baseline samples (β = −0.23, p = 0.04). Overall, average 3‐HB concentrations were low ( < 1 mM) and consistent between Receivers (mean = 0.43 mM ± 0.29) and Controls (mean = 0.44 mM ± 0.43) and between pre‐ and post‐baseline values (mean = 0.70 mM ± 0.65, 0.32 mM ± 0.22, respectively; Supporting Information Figure S13). Lactate concentrations were also low ( < 1 mM) and consistent between Receivers (mean = 0.17 mM ± 0.20) and Controls (mean = 0.22 mM ± 0.20) and between pre‐ and post‐baseline values (mean = 0.17 mM ± 19, 0.21 mM ± 0.26; Supporting Information Figure S14). Similar to lactate, no differences in pH were observed between Receivers (mean = 7.86 ± 0.42) and Controls (mean = 7.77 ± 0.44) and between pre‐ and post‐baseline values (mean = 7.8 ± 0.3, 8.1 ± 0.36; Supporting Information Figure S15).

4. Discussion

This study provides the first empirical evidence of chemical disturbance cues in elasmobranchs. CBR downstream (Receiver) of conspecifics exposed to a simulated predation disturbance (Sender) changed their behavior relative to Controls. We found a shift from more passive to more active swimming behaviors in Receivers following the manipulation of a Sender animal. The behavioral response of Receivers was also associated with an increase in velocity, whereas Controls showed no change in velocity. This suggests Receivers detected and responded to a chemical cue released by the Sender, as other sensory cues were either unavailable or shared amongst Control and Receiver animals. Notably, this behavioral response occurred without measurable changes in secondary stress traits, suggesting the potential perception of chemical disturbance cues and the subsequent behavioral response did not activate a secondary stress response in CBR.

4.1. Behavioral Parameters

In the context of predator–prey interactions, behavioral changes are often adaptive (Howland 1974). In this study, Receiver CBR increased movement following the manipulation of a Sender CBR, consistent with an antipredator behavior response (Howland 1974). Increased velocity is a primary mechanism for avoiding predators and the increased speed can help prey escape or outmaneuver the threat (Howland 1974). While our estimated average velocity measurements are comparable to average velocities of CBR in the wild (Matern et al. 2000; Klimley et al. 2005), it is important to note aeration stones distorted the water surface in the video recordings and may have introduced variability in the measurements. Nonetheless, the relative increase in velocity and the changes to more active qualitative behavioral states after manipulation, provides strong evidence of a behavioral shift. In a confined environment, faster swimming speeds while maintaining proximity to the tank wall (i.e., Wall Swimming) may indicate CBR are attempting to escape a perceived threat, similar to a flight response in the wild (Jorgensen et al. 2019); Towner et al. 2023). These results suggest Receiver CBR are detecting and responding to cues released by the manipulated Sender CBR with behavioral changes consistent with an antipredator strategy that is evasive in nature, characterized by increased movement and escape‐oriented behaviors rather than concealment or freezing.

We also found both Receiver and Control groups switched from behaviors associated with the bottom of the tank to behaviors higher in the water column following treatment, indicating CBR may have been aware of experimental treatment procedures. While Controls may have increased their alertness, which brought them higher into the water column, their response are not necessarily anti‐predator behavior. At the Oregon Coast Aquarium, CBR were observed by staff rising to the water surface in response to certain stimuli, such as the movement of a stool used during feeding times (Boyt, pers comm). While constant ambient noise in the tank (e.g., circulation pumps) and efforts by investigators to be quiet made such stimuli less likely in our experimental set up, it is possible CBR in the Control group were aware of stimuli resultant from our experimental procedures. In a similar study design in Iowa Darters, control fish increased movement from baseline following stirring of an upstream empty tank with a model predator in comparison to receiver fish. This response was likely cause by sediment disturbance that introduced a chemical stimulus when stirring the empty tank with a model predator (Wisenden et al. 1995). In our system, there was no substrate added, though residue naturally accumulated on the floor of the tanks. Therefore, stirring the tank may have lifted a chemical stimulus off the floor of the tank that resulted in the observed PC2 response where Controls and Receivers both moved higher into the water column.

No other significant behavioral differences were observed between Receivers and Controls, suggesting the behavioral changes were specific to certain strategies (e.g., Wall Swimming and increased velocity) and did not affect all aspects of the CBR' behavior. Both Receivers and Controls were exposed to identical experimental procedures (outside of the presence of a conspecific Sender CBR in experimental treatment tanks) and, to our knowledge, all other communication pathways (e.g., sound, visual, or electro/mechanosensory) were either unavailable or shared with Controls, suggesting CBR released chemical communication signals.

4.2. Physiological Parameters

We did not find evidence that behavioral changes due to DC were associated with a physiological response (e.g., change in energy metabolites or anaerobic byproducts). There were no significant differences observed during experimental trials between Receivers and Controls for glucose, 3‐HB, lactate, or pH. Interestingly, post‐experimental baseline glucose measurements had greater variance and tended to be higher than pre‐baseline and experimental values for both Controls and Receivers. This possibly indicates a response to cumulative manipulations after multiple experimental trials. Glucose, lactate, and pH were similar, though lower, to previously published results measuring in vivo resting CBR hematological characteristics (Hopkins and Cech 1994). In that study, CBR glucose values were overall lower, yet comparable, to other batoid species—American Cownose ray (Rhinoptera bonasus), Brazillian Cownose ray (Rhinoptera brasiliensis), Spotted Eagle ray (Aetobatus narinari), and Longnose Stingray (Hypanus guttanus)—after 5 min capture and handling times (Rangel et al. 2021). However, in air exposed Atlantic Stingrays (Hypanus sabinus), baseline glucose concentrations were considerably lower (0.71 mM) and decreased further after 15 min of air exposure (Lambert et al. 2018). The large variation in baseline and stress‐induced glucose levels across species and experiments complicates interpretation. A more thorough understanding of how glucose changes relate to corticosteroid hormones, life history or environmental contexts, and species‐specific physiology is necessary to draw broader conclusions about stress and metabolic responses in elasmobranchs.

Lactate and pH are byproducts of an anaerobic metabolic response, and lactate has been used as a reliable indicator of stress responses in elasmobranchs (Lambert et al. 2018). The absence of changes in lactate and pH in the presence of behavioral changes suggests our observed behavioral response was within the CBR's aerobic capacity. Indeed Hopkins and Cech (Hopkins and Cech 1994) found CBR exhibit high blood oxygen‐carrying capacity, large Bohr factors, and significant non‐bicarbonate buffering capacity, all of which support sustained high activity levels and efficient recovery from stressors.

We also measured levels of 3‐HB, a ketone body that may play an important role in elasmobranch energy balance. Ketone bodies, like 3‐HB, in elasmobranchs have been found to increase during periods of fasting, prolonged exercise, or metabolic stress, and may be used as an alternative energy source when glucose is depleted (Anderson et al. 2002; Cliff and Thurman 1984; Richards et al. 2003). Our study represents the first measured 3‐HB levels in CBR. Overall, 3‐HB levels remained low and were similar across all treatment groups. This aligns with previous findings in other elasmobranchs where well‐fed Spiny dogfish (Squalus suckleyi) displayed low 3‐HB levels, indicating glucose was likely meeting their energy demands (Schoen et al. 2021; Schoen 2023). The lack of change in 3‐HB suggests CBR did not require significant mobilization of fat reserves nor undergo significant metabolic adjustment during the experimental period or CBR do not predominately use 3‐HB under these conditions. Frequent feeding is known to suppress ketogenesis in elasmobranchs (e.g., Treberg et al. 2007; Ballantyne 1997), supporting the interpretation energetic demands were adequately met through carbohydrate metabolism. These findings suggest under well‐fed conditions, CBR may not rely heavily on 3‐HB for energy, even in the presence of increased activity.

The physiological data in this experiment suggest the behavioral differences observed were not extreme enough to push an animal beyond its aerobic metabolic capacity, when compared to more extreme methods employed in prior studies (i.e., capture, thermal, and air exposure stress; Hoffmayer and Parsons 2001; Schoen et al. 2021; Lambert et al. 2018). Evidence from teleosts suggests the magnitude of a chemical cue must exceed a certain threshold to elicit a measurable response (Brown et al. 2006). For example, behavioral reactions to chemical signals in Matrinxã (Brycon amazonicus) did not coincide with changes in plasma glucose or cortisol. The authors postulated the lack of physiological response was due to insufficient amount of conspecific skin extract used as the stimulus (Ide et al. 2003). It is possible the disturbance event of the Sender CBR was not long enough, or disturbance cues elicit a different type of response than cues released upon injury of a conspecific. Alternatively, metabolic scope in elasmobranchs appears to be coupled with the level of swimming activity (i.e., faster, pelagic, elasmobranch species have a higher aerobic scope compared to slower, benthic, species; Graham et al. 1990). Therefore, active (and well‐fed) swimming elasmobranchs, such as CBR, may be better able to tolerate and recover from increased activity without sacrificing aerobic demands (Hopkins and Cech 1994; Graham et al. 1990).

Another potential factor influencing our physiological results was feeding. Typically, experiments investigating HPI responses in elasmobranchs have historically been conducted on fasted animals to standardize nutritional state. However, our CBR were fed daily to prevent any impacts of food restriction on chemical signaling; though feeding took place after experimental trials because food consumption is known to impact circulating glucose and 3‐HB in elasmobranchs (Treberg et al. 2007; Ballantyne 1997; Wood et al. 2010). Studies on the physiological stress response in routinely fed elasmobranchs are sparse; one caveat of our study is daily feeding may have masked or muted potential metabolite signals. Such differences highlight the need for studies on a wider array of more naturalistic physiological states in elasmobranchs.

Lastly, an unexplained pattern in the physiological data was the second baseline glucose measurement collected from animals in Round2 trended higher than other pre‐baseline values. There is no obvious explanation for the seemingly unusually high glucose values from this sampling event. Measured water quality parameters and feeding rates for that date, and the dates immediately surrounding it, were all reported within typical ranges. We note there was a small (3.5 magnitude) earthquake off the Oregon coast (approx. 180 km sw) that occurred the day before this baseline sample collection event. Animals are known to respond to low frequency pulses, but we cannot say for certain this explains the pattern ($author1$ et al. 2024).

4.3. Implications for Elasmobranch Ecology

As marine predators, elasmobranchs play important ecological roles. Thus, understanding how disturbance cues may influence their behavior and physiology has broad ecological relevance (Hammerschlag et al. 2025; Pace et al. 1999; Myers et al. 2007). Although behavioral responses may vary among elasmobranch species and ecological contexts, there is increasing recognition that chemical communication may mediate predator–prey interactions in this group (Gervais et al. 2021; Gervais and Brown 2021). For example, the flight response of White sharks to the presence of Orca whales was associated with decreased foraging in the area, thereby releasing intermediate predators (e.g., pinnipeds) from top‐down predation pressure (Jorgensen et al. 2019; Hammerschlag et al. 2025; Myers et al. 2007). Removal or displacement of elasmobranch predators has been linked to trophic cascades with far‐reaching consequences for marine ecosystems, influencing species composition and abundance (Hammerschlag et al. 2025; Myers et al. 2007). Displacement of predators can also alter foraging pressure if other predators fill the niche. These shifting dynamics are important to consider as the distribution and identity of predators can have profound impacts on natural resources. Our findings support the idea that chemically mediated risk assessment in elasmobranchs may be more widespread than previously recognized.

We did not detect a clear link between increased activity and changes in measured energy metabolites. However, this relationship remains poorly understood in elasmobranchs and warrants further investigation. Our study was limited to secondary stress indicators due to logistical constraints that limited our physiological assessment to blood sample collection, though future studies could more definitively gauge HPI activation by measuring neurohormones such as corticotropin‐releasing hormone, adrenocorticotropic hormone, adrenaline, or noradrenaline. It is also possible our physiological data underestimate the energetic impacts of these responses, given captive animals in confined spaces have different swimming conditions compared to animals in the wild. Even in the absence of a measurable physiological stress response, behavioral modifications can impose fitness costs (Schreck et al. 2001). Indirect costs of disturbance behaviors may extend beyond physical exertion to include lost foraging opportunities, potentially reducing overall energy intake (Pérez‐Tris et al. 2004). The tertiary responses to stress (i.e., fitness consequence) have not been as robustly studied in elasmobranchs as in teleosts, in which acute and chronic stressors can impact growth rates, reproductive outputs, and disease resistance (Schreck et al. 2001; Barton and Iwama 1991; Portz et al. 2006; Van Weerd and Komen 1998). If the energy needs exceed energy availability, then it may induce a metabolic response that could lead to secondary, tertiary, or chronic stress responses (Barton and Iwama 1991). Unnecessary and frequent exposure to AC/DC can affect other biologically important processes, such as development, growth, and reproduction if energy is allocated to a disturbance response instead. Elasmobranchs are of particular conservation concern given their slow life history characteristics (late age to maturity and low fecundity) and complex role in marine ecosystems (Dedman et al. 2024).

Ethics Statement

All animal procedures were performed in accordance with the guidelines and regulations of the Institutional Animal Care and Use Committee (IACUC) at Oregon State University and were approved under protocol IACUC‐2022‐0251.

Supporting information

Supporting File 1

Supporting File 2

Supporting File 3

Supporting File 4

Supporting File 5

Supporting File 6

Data Availability Statement

The data that support the findings of this study are available in the Supporting Material of this article.