ABSTRACT
Calanoid copepods, depending on feeding strategy, have different behavioral and biological controls on their movements, thereby responding differently to environmental conditions such as changes in seawater viscosity. To understand how copepod responses to environmental conditions are mediated through physical, physiological and/or behavioral pathways, we used high-speed microvideography to compare two copepod species, Acartia hudsonica and Parvocalanus crassirostris, under different temperature, viscosity and dietary conditions. Acartia hudsonica exhibited ‘sink and wait’ feeding behavior and typically responded to changes in seawater viscosity; increased seawater viscosity reduced particle-capture behavior and decreased the size of the feeding current. In contrast, P. crassirostris continuously swam and did not show any behavioral or physical responses to changes in viscosity. Both species showed a physiological response to temperature, with reduced appendage beating frequency at cold temperatures, but this did not generally translate into effects on swimming speed, feeding flux or active time. Both copepod species swam slower when feeding on diatom rather than dinoflagellate prey, showing that prey type mediates copepod behavior. These results differentiate species-specific behaviors and responses to environmental conditions, which may lead to better understanding of niche separation and latitudinal patterns in copepod feeding and movement strategies.
INTRODUCTION
As copepods are an important link in the marine food web (Cushing, 1989; Levinson et al., 2000; Turner, 2004) with influence over biogeochemical cycles (Jónasdóttir et al., 2015), their movements may have broad impacts on marine ecosystems. In the nutritionally dilute ocean, planktonic copepods must search approximately 106 body volumes of seawater each day to consume adequate food (Kiørboe, 2011b; Kiørboe and Jiang, 2013). From a fluid dynamical point of view, a copepod's feeding current depends on its excess weight (Jiang et al., 2002a,b; Kiørboe and Jiang, 2013; Strickler, 1982). However, there are also biological and physiological constraints on the feeding current, leading to a diversity of movement and feeding strategies (Kiørboe, 2011a,b). Copepods must beat their cephalic appendages to drive the feeding current, and appendage movement may be constrained by metabolism, temperature and seawater viscosity. Ingestion also depends on active feeding time, which is under behavioral control and may be altered according to environmental conditions. Additionally, copepods with different feeding strategies may have different behavioral responses. For example, copepods that can use multiple feeding strategies (Kiørboe et al., 1996) can switch to the most advantageous method, while copepods that use only one mechanism cannot show this response. These biological constraints on feeding have seldom been studied.
At 0°C, the viscosity of 36‰ salinity seawater is 1.89×10−3 kg m−1 s−1, whereas at 30°C, its viscosity is only 0.86×10−3 kg m−1 s−1 (Miyake and Koizumi, 1948). Biogenic compounds, such as the mucus released by some bloom-forming algae, may also increase seawater viscosity (Seuront et al., 2006). Because of their small size (∼0.5–10 cm), copepods are characterized by low Reynolds numbers of 10−2 to 103 (Yen, 2000) (Re=ρuL/μ, where ρ is seawater density, μ is seawater dynamic viscosity, u is the speed of the organism and L is the length of the organism). In this low-to-intermediate Re regime, movements are impacted by seawater viscosity.
Therefore, several temperature-related responses in the zooplankton can be attributed to a physical response to seawater viscosity. Seawater viscosity is responsible for reduced ingestion at cold temperature (Bolton and Havenhand, 1998; Podolsky, 1994; Tyrell and Fisher, 2019) and reduced swimming speed at cold temperature (Larsen et al., 2008). Yet, we are unaware of any laboratory studies that have used videography to directly investigate the physical and behavioral pathways through which seawater viscosity and temperature impact copepods.
Here, we used high-speed microvideography to study the swimming and movement behavior of two species of common temperate copepods, Acartia hudsonica and Parvocalanus crassirostris, which are both found along the eastern coastal USA (Milligan et al., 2011; Turner, 1981). Acartia spp. have a cephalothorax length of ∼0.8–1.0 mm and utilize a ‘sink and wait’ feeding strategy, with short feeding periods (<0.5 s) interspersed with longer periods of sinking (>1 s) (Kiørboe et al., 1996). In contrast, P. crassirostris have a cephalothorax length of ∼0.3–0.4 mm and spend more time creating a feeding current (A.S.T., personal observation; Bradley et al., 2013). The contrast between these two species may illuminate the advantages and trade-offs of each feeding strategy. Because diet influences copepod response to seawater viscosity (Tyrell and Fisher, 2019), we fed the copepods two unialgal diets: Thalassiosira weissflogii, a non-motile diatom, and Prorocentrum mimimum, a motile dinoflagellate. The natural ranges of both phytoplankton species overlap with the ranges of the two copepod species (Turner, 1981; Hargraves, 2002; Heil et al., 2005; Sorhannus et al., 2010; Milligan et al., 2011). Species-level resolution of copepod ingestion patterns in the field is scarce, but both P. crassirostris (Calbet et al., 2000) and the A. hudsonica congener Acartia tonsa (Bollens and Penry, 2003) consume small autotrophic cells, including diatoms and flagellates; therefore, T. weissflogii and P. minimum are representative dietary species. In all experiments, we separated the effect of viscosity from the effect of temperature by manipulating seawater viscosity with the non-toxic polymer polyvinylpyrrolidone (PVP) (Podolsky and Emlet, 1993; Riisgård and Larsen, 2007; Tyrell and Fisher, 2019) in addition to altering the temperature.
We hypothesized that temperature responses would be partly or fully explained as a viscous response. We hypothesized that copepod movements would be slowed and activity would be decreased by increased seawater viscosity. Because our previous research has shown that copepod response to viscosity differs depending on diet (Tyrell and Fisher, 2019), we hypothesized that copepod behavior and movement would be also affected by diet. Because A. hudsonica is found at cooler temperatures and is a less active swimmer than P. crassirostris, we hypothesized that A.hudsonica would be less sensitive to changes in viscosity.
MATERIALS AND METHODS
Phytoplankton cultures
Prorocentrum minimum (clone CCMP 696) and Thalassiosira weissflogii (clone CCMP 1336) cultures were used. Cultures were maintained in sterile filtered (0.2 µm) seawater from Woods Hole, MA, USA (salinity of 32–35‰) supplemented with f/2 nutrients (Guillard and Ryther, 1962). Cells were kept on a 14 h:10 h light:dark cycle at ambient temperature (18–21°C).
Copepod cultures
Acartia hudsonica Pinhey 1926 were collected from Woods Hole, MA, USA, 1–2 days before videography trials. AlgaGen ReefPods™ cultures of Parvocalanus crassirostris (F. Dahl 1894) were ordered from LiveAquaria (Rhinelander, WI, USA) in May 2018. Copepods were maintained in the laboratory before experiments and fed ad libitum with a mixture of P. minimum, T. weissflogii and Isochrysis galbana. All copepods were kept on a 14 h:10 h light:dark cycle at ambient temperature (18–21°C) until shortly before trials began.
Experimental treatments
The three experimental treatments used 0.2 μm filtered water (salinity of 32–35‰) collected from Woods Hole, MA, USA. Treatments were: (1) 10°C seawater, (2) 20°C seawater and (3) 20°C seawater with 0.12% w/v PVP (Sigma-Aldrich, SID 24899318) added to create a viscosity similar to that in 10°C seawater (Table S1). The PVP polymer is non-toxic and addition of 0.12% w/v enabled manipulation of seawater viscosity without affecting temperature or seawater density (Podolsky and Emlet, 1993; Riisgård and Larsen, 2007; Tyrell and Fisher, 2019).
Seawater samples were taken from each experiment and stored at 4°C in the dark until analysis. The kinematic viscosity of the experimental water was measured at the appropriate temperature using an Ubbelohde viscometer (Sigma-Aldrich UBBEL02UKC) and the equation supplied by the manufacturer. Kinematic viscosity was converted to dynamic viscosity using the equation: dynamic viscosity=density×kinematic viscosity.
Videography system
A high-speed microscale imaging system (HSMIS) was used to record high-resolution 2D digital videos at 2000 frames s−1. The HSMIS consists of a Photron FASTCAM SA3 120 K monochrome video camera (San Diego, CA, USA) that takes 1024×1024 pixel resolution images at frame rates up to 2000 frames s−1. The camera is mounted horizontally with a 150 mm focal length objective lens plus an infinity-corrected, long working distance microscope objective (4×/0.10 18.5 mm working distance) to yield a field-of-view of a vertically oriented area of ∼4.8×4.8 mm, or approximately 6 A. hudsonica body lengths or 12 P. crassirostris body lengths. A 1 W white LED light source was collimated to provide backlit illumination in which light was shone toward the camera through a prepared flask placed in front of the microscope objective. The field of view was focused at the center of the flask, which was at least 1 cm (∼12 or 25 copepod body lengths) away from flask walls. A HSMIS of different optical specifications has been previously used for quantitative microvideography and micro-particle image velocimetry (Jiang and Johnson, 2017; Jiang et al., 2018; Du Clos and Jiang, 2018; Jiang and Paffenhöffer, 2020). The HSMIS has the advantage of achieving sharp imaging under low illumination. A summary of video conditions can be found in Table S1.
Experimental measurements and analysis
One day prior to recording videos, 28–66 adult female copepods (Table S1) were placed into 20 ml (P. crassirostris) or 40 ml (A. hudsonica) treatment seawater. Within each experimental treatment, there were three dietary/particle conditions: (1) P. minimum cells only, (2) T. weissflogii cells only and (3) T. weissflogii cells with polystyrene tracking particles added (3 µm diameter) to help trace the seawater movement around the copepods (hereafter, these three treatments are referred to as the diet treatments). Copepods sometimes consumed a small portion of the cells and tracking particles during the 2–8 h of videography. The copepods were acclimated for >13 h in seawater of the appropriate temperature and viscosity with algal cells added (Table S1). The copepods were kept in indirect light on a 14 h:10 h light:dark cycle during this acclimation period.
A total of 519 videos were recorded according to the following procedure; a live video feed was monitored. When a copepod swam across the field of view, the previous 2.7 s of footage was captured by manually triggering the camera. The footage was then edited to include only the period of time during which the copepod was in the frame, and it was saved. The saving process took ∼5–10 min, after which the procedure was repeated. A target of 25–30 videos of each copepod species were saved for each of the nine diet–treatment combinations. Sample sizes of video data are constrained by the intensive data collection process, and it is not always possible to know whether a video will be suitable for all analyses at the moment it is collected. Through bootstrap analysis of 1000 sets of random data, we calculated that ≥5 suitable videos per diet–treatment combination would yield a power of ≥0.9 for an effect size of a 50% reduction in the mean if the standard deviation was ≤50% of the mean, while ≥15 suitable videos per diet–treatment combination would yield a power of ≥0.8 for an effect size of a 25% reduction in the mean if the standard deviation was ≤50% of the mean. Therefore, our experimental design was suitable to detect effects.
Most videos contained only one copepod, although a small portion of videos contained multiple copepods. Because the field of view was larger in terms of P. crassirostris body lengths over A. hudsonica body lengths, the larger-scale movements of P. crassirostris may have been better captured, while the individual-level details of A. hudsonica may have been better captured. The maximum video length was 2.7 s, and the mean (±s.d.) video time for both copepod species was 2.5±0.5 s. In all analyses, only adult females were considered.
Flow fields, copepod movements, appendage beating frequency, proportion of time spent swimming and swimming speed were analyzed. Analyses were done in R (version 3.6.1) or Python (version 2.7.17); specific software package details are shown in Table S2. Depending on the analysis, videos were included if the copepod was in focus, swimming in the plane of the camera and/or if its movements could be resolved by eye; analyses are described below.
Particle tracking: feeding flux
In the videos with tracking particles added, ≥100 frames (A. hudsonica) or ≥1000 frames (P. crassirostris) from videos with consistent, in-focus copepod feeding movements were analyzed using a micro-particle tracking velocimetry (µPTV) method in Python using the libraries trackpy and OpenCV2. All particles in an 800×800 pixel (∼3.8×3.8 mm) window (A. hudsonica) or 400×400 pixel (∼1.9×1.9 mm) window (P. crassirostris) around the copepod were tracked, and the particle trajectories were recorded. The trajectories were then centered on the copepod's frame of reference. The speed of each particle was calculated from its net displacement divided by the period of time that it was in the video.
For each video, the speed of the feeding current was calculated as the mean speed of all particles that crossed a line positioned one body length forward from the copepod's center of mass. Depending on the orientation of the copepod, the line's width was equal to the width of the antennae (dorsoventral orientation) or two body widths (sagittal orientation). This feeding current speed was then used to calculate the feeding flux (the amount of seawater moving past the copepod's mouthparts per unit time) by assuming an elliptical shape of the cross-section of the feeding current, with the long axis equal to the width of the antennae and the short axis equal to two body widths. The effect of temperature/viscosity treatment on feeding flux was determined with one-way ANOVA with Tukey's post hoc test.
Copepod movements
Copepod movements were classified as sinking, swimming, twitching of the first antennae or urosome, hopping or jumping. The copepod was classified as sinking when it was motionless. Swimming was defined as rhythmic beating of the feeding appendages (cephalic appendages). During swimming, the copepods sometimes twitched the urosome or first antennae, and this was noted as well. Large hops were defined as up to two (A. hudsonica) or three (P. crassirostris) rapid beats of the swimming appendages. Acartia hudsonica displayed more nuanced hopping behavior, so A. hudsonica small hops were categorized separately and defined as incomplete deployment of the swimming appendages in conjunction with movement of the first antennae and/or urosome that displaced the copepod a short distance. Jumping was defined as more than two (A. hudsonica) or three (P. crassirostris) beats of the swimming appendages. All movements where the copepod was relatively in focus, but not necessarily moving in the plane of focus, were assessed: 287 A. hudsonica copepods were analyzed, with copepods in the video for a mean (±s.d.) of 2.3±0.7 s (minimum length 0.1 s); 235 P. crassirostris copepods were analyzed, with copepods in the video for 2.4±0.7 s (minimum length 0.05 s).
The number of twitches, hops and jumps was analyzed by generalized linear models (GLMs), with temperature/viscosity treatment, diet and length of video as covariates. The negative binomial distribution provided a better fit of the data over the Poisson distribution as determined by comparison of Akaike's information criterion with small sample size correction (AICc) and so was used for all analyses, except in the case of P. crassirostris jumps, which were fitted with a Poisson model because the negative binomial model did not converge. For each movement, the statistical importance of treatment, diet and the interaction between treatment and diet was determined by χ2 likelihood ratio tests comparing full and reduced models (comparisons detailed in Table 1). Groupwise differences for each statistically significant factor were determined by creating single-factor zero-intercept GLMs and comparing coefficient estimates; if the 95% confidence interval of coefficient A did not contain the estimate of coefficient B, and vice versa, coefficients A and B were considered statistically different.
Appendage beating frequency
Appendage beating frequency was counted manually for all swimming copepods with visible and identifiable feeding appendages (210 A. hudsonica copepods and 169 P. crassirostris copepods). The method of counting was tailored to the specific movement patterns of each copepod species.
For P. crassirostris, the second antenna and the maxilliped beat in time, and the frequency of this beating was counted by recording the frame numbers at the start of five consecutive beating cycles; average appendage beating frequency for one cycle was then calculated. The start of the appendage beating cycle was arbitrarily defined as the frame when the second antenna and/or the maxilliped (whichever was most visible) returned to a recognizable, cyclic position. Beating frequency was counted in this way three times per video, at manually selected periods approximately evenly distributed throughout the full amount of time during which the copepod was swimming. Appendage beating frequencies were only recorded during periods when the copepod was not twitching the urosome, appendages or first antennae; these behaviors may indicate particle capture or rejection and may result in non-uniform beating of the feeding appendages. Because the interaction between temperature/viscosity treatment and diet was statistically significant, the effect of the three diets and three temperature/viscosity treatments on P. crassirostris appendage beating frequency was analyzed by two-way ANOVA with Type III error.
For A. hudsonica, the appendage beating frequency was measured by recording the beating of the maxilliped. As A. hudsonica beat its appendages less rhythmically and for a shorter average duration compared with P. crassirostris, the beginning frame of all beat cycles was recorded and defined as the frame when the maxilliped first moved away from the body. Then, the first frame when the maxilliped returned to the body and the last frame when the maxilliped returned to the body were both recorded. Appendage beating frequency was calculated as the time between the first return of the maxilliped and the final return of the maxilliped, divided by the number of beats in between (one fewer than the total number of beats). The average appendage beating frequency for each A. hudsonica in each video with a visible maxilliped was determined. Because the interaction between temperature/viscosity treatment and diet was not statistically significant, the effect of the three diets and three temperature/viscosity treatments on appendage beating frequency was analyzed by two-way ANOVA with Type II error.
Proportion of time spent swimming
The total number and duration of swimming periods were determined for each copepod in each video. Swimming periods were classified as times when the copepods were rhythmically beating their feeding appendages, and did not include hops or jumps. The proportion of each video spent swimming was calculated by dividing the swimming time by the total time that the copepod was in the video. Videos with no swimming were included as zeroes in the analysis.
For P. crassirostris, the first and last frames with movement of the feeding appendages were recorded for each swimming period; 20 out of 235 P. crassirostris copepods did not swim. For A. hudsonica, swimming was recorded in conjunction with appendage beating frequency measurements (see above). The total time spent swimming was calculated as the difference in time between the first movement of the maxilliped and the final return of the maxilliped, summed over all feeding periods in the video; 72 out of 285 A. hudsonica copepods did not swim. The proportion of time spent swimming was compared across the three diets and three temperature/viscosity treatments with the length of the video included using ANCOVA using Type II error if the interaction between temperature/viscosity treatment and diet was not statistically significant and Type III error if the interaction was statistically significant. Tukey's post hoc test was performed following ANCOVA.
Swimming speed
The coordinates of copepods swimming in the plane of focus were obtained using ImageJ (versions 1.52a and 1.52q) and corrected for background movement by subtracting the movement of a randomly selected background particle. The coordinates were then smoothed using a fourth-order Savitsky–Golay filter with a 23-point window (Jiang and Kiørboe, 2011) and speed was calculated by taking the first derivative of the copepod's position. The average swimming speed was calculated for feeding periods, defined as periods when the copepod was rhythmically beating its feeding appendages. Only periods longer than 100 frames (0.05 s; A. hudsonica) or 450 frames (0.2 s; P. crassirostris) with no twitching were considered. Speed was broken into three categories: (1) copepod oriented upwards with net movement upwards [mean±s.d. video lengths of 0.8±0.6 s (99 P. crassirostris feeding periods) and 0.1±0.07 s (98 A. hudsonica feeding periods)]; (2) copepod oriented upwards but with net movement downwards [mean±s.d. video lengths of 1.0±0.7 s (79 P. crassirostris feeding periods) and 0.1±0.06 s (94 A. hudsonica feeding periods)]; and (3) copepod oriented downwards with net movement downwards [mean±s.d. video lengths of 0.7±0.5 s (24 P. crassirostris feeding periods) and 0.1±0.07 s (31 A. hudsonica feeding periods)]. The copepods were never oriented downwards with net movement upwards.
Within each video, each copepod's swimming speed was averaged across all swimming periods with the same orientation and displacement directions. Because no interactions between temperature/viscosity treatment and diet were statistically significant, two-way ANOVA with Type II error was used to compare the effects of the three diet and three temperature/viscosity treatments on copepod speed for each orientation/displacement combination.
Interpretation of results
In all of our analyses, we interpreted the effects of seawater viscosity and temperature according to the following patterns. If there was no difference between the 20°C treatment and the high viscosity treatment, it was determined that viscosity did not have an effect. If there was no difference between the 10°C treatment and the high viscosity treatment, it was determined that temperature did not have an effect.
RESULTS
Particle tracking: feeding flux
Fig. 1 shows examples of the flow fields around A. hudsonica and P. crassirostris. The feeding flux of A. hudsonica was statistically decreased by viscosity but unaffected by temperature (Tukey's post hoc, P<0.05, following one-way ANOVA, F2,3=24.2, P=0.014) (Table 2). After multiplying by the proportion of time spent swimming, A. hudsonica cleared 0.1×106–0.2×106 body volumes (4.7–8.7 ml) of seawater per day at high viscosity (regardless of temperature); at 20°C with unchanged viscosity, A. hudsonica cleared 0.9×106 body volumes (32.5 ml) per day (Table 2). The feeding flux of P. crassirostris was not affected by seawater viscosity or temperature (ANOVA, F2,10=0.40, P=0.68) (Table 2). Parvocalanus crassirostris cleared 1.4×106–1.6×106 body volumes (6.4–9.4 ml) per day after multiplying by the proportion of time spent swimming (Table 2).
Movements
Examples of copepod movements can be viewed in Movies 1–6. These videos replay at 20 Hz (100× slower than real time).
Acartia hudsonica movements primarily included hops, sinking and short swimming periods (mean±s.d. length of 0.089±0.068 s, n=871 feeding periods) (Table 3). Acartia hudsonica displayed fewer twitches at high viscosity, regardless of temperature (median number of twitches per video reduced from 2 to 1; Tables 1 and 3). Acartia hudsonica small hops, large hops, and jumps were not affected by temperature/viscosity treatment (Tables 1 and 3). Diet did not affect any A. hudsonica movements (Tables 1 and 3).
Parvocalanus crassirostris behavior primarily included long swimming periods (mean±s.d. length of 1.55±0.96 s, n=293 swimming periods; this is an underestimate because many swimming periods continue beyond the duration of the video) with twitches of the first antennae and/or urosome (Table 3). There was no effect of temperature/viscosity treatment or diet on P. crassirostris twitching, hopping or jumping (Tables 1 and 3).
During the swimming periods, copepods of both species rhythmically moved their feeding appendages and typically were displaced through the water (i.e. no ‘hovering’).
Ten randomly selected videos of each copepod species were re-checked for consistency, which showed that movements were reliably scored. There was small variation in the classification of hops and twitches of A. hudsonica (3/31 classified differently); 8/10 P. crassirostris videos counted the same number of twitches when re-checked and 2/10 videos varied. Hop counts for P. crassirostris and jump counts for A. hudsonica and P. crassirostris were all 100% repeatable.
Appendage beating frequency
Acartia hudsonica appendage beating frequency (Fig. 2) was affected by temperature/viscosity treatment (two-way ANOVA, F2,201=57.41, P<1×10−7). Diet also had an effect (two-way ANOVA, F2,201=4.62, P=0.011), and there was no interaction between diet and temperature/viscosity (two-way ANOVA, F4,201=0.87, P=0.48). Appendage beating was significantly faster at 20°C than at 10°C regardless of viscosity (Tukey's post hoc following two-way ANOVA, P>0.05), and appendage beating was significantly faster when feeding on T. weissflogii with tracking particles versus T. weissflogii without tracking particles (Tukey's HSD following two-way ANOVA, P>0.05).
Parvocalanus crassirostris appendage beating frequency (Fig. 2) was affected by temperature/viscosity treatment (two-way ANOVA, F2,160=160.29, P<1×10−7), but diet had no effect (two-way ANOVA, F2,160=1.78, P=0.17). The interaction between diet and temperature/viscosity was statistically significant (two-way ANOVA, F4,160=3.32, P=0.012). Appendage beating was significantly faster at 20°C than at 10°C, regardless of viscosity (Tukey's post hoc following two-way ANOVA, P>0.05).
A random selection of 12 A. hudsonica videos and 10 P. crassirostris videos were re-checked, which showed that appendage beating frequency measurements were consistent and repeatable (paired t-test, A. hudsonica: t11=0.14, P=0.89; P. crassirostris: t9=0.26, P=0.80).
Acartia hudsonica frequently reversed the direction of movement of its appendages while feeding, without a consistent pattern (Table S3, Movie 1). This switching movement was not affected by temperature/viscosity or dietary treatment (temperature/viscosity treatment and diet analyzed as a single factor; Pearson's χ2=17.88, d.f.=16, P=0.33). When appendage movement direction was included in the factorial ANOVA, movement direction did have a statistically significant effect on appendage beating frequency (F2,340=40.82, P<1×10−7), but the interpretation of temperature/viscosity treatment and diet was unchanged (temperature/viscosity: F2,340=106.26, P<1×10−7 and same Tukey's results; diet: F2,340=8.25, P=8.32×10−4 and copepods feeding on T. weissflogii without tracking particles beating their appendages more slowly than the other two diets). Therefore, we do not distinguish between movement direction in our analysis of appendage beating frequency. The movement direction was sometimes difficult to classify; 12 videos were re-analyzed for consistency, and 18% of appendage movements were classified differently from the original. Parvocalanus crassirostris only exhibited appendage reversal when rejecting large particles, and these events were not included in the calculation of appendage beating frequency.
Proportion of time spent swimming
Acartia hudsonica spent 7.9–19.7% of the time swimming (Fig. 3). The proportion of time spent swimming was affected by temperature/viscosity treatment (ANCOVA, F2,275=6.42, P=1.9×10−3). Copepods spent 27–41% less time swimming when the viscosity was higher, regardless of temperature (Tukey's post hoc following ANCOVA, P<0.05). Diet did not affect the proportion of time spent swimming (ANCOVA, F2,275=0.78, P=0.46), and there was no interaction between temperature/viscosity treatment and diet (ANCOVA, F4,275=1.85, P=0.12).
Parvocalanus crassirostris spent 57.9–90.8% of the time swimming (Fig. 3). The proportion of time spent swimming was not affected by temperature/viscosity treatment (ANCOVA, F2,225=1.32, P=0.27) or diet (ANCOVA, F2,225=0.29, P=0.75), though the interaction was statistically significant (ANCOVA, F4,225=2.77, P=0.028).
A random selection of 12 P. crassirostris and 12 A. hudsonica videos were re-checked, which showed that the proportion of time swimming was consistently and repeatably recorded in the videos (paired t-test, A. hudsonica: t11=−0.95, P=0.36; P. crassirostris: t11=−1.60, P=0.14).
Swimming speed
Acartia hudsonica swimming speeds were generally 2–4 mm s−1 (Fig. 4A–C), corresponding to Re of 1–3. Parvocalanus crassirostris swimming speeds were generally 0.4–0.6 mm s−1 (Fig. 4D–F), corresponding to Re of 0.1–0.5. In five out of the six species–orientation combinations, there was a dietary effect on copepod swimming speed, with copepods feeding on T. weissflogii swimming more slowly than copepods feeding on P. minimum (Fig. 4, Table 4). Acartia hudsonica swimming was sometimes slowed by high viscosity, while P. crassirostris swimming was sometimes slowed by cold temperature (Table 4).
DISCUSSION
The observed movements of the two copepod species clearly differed, with P. crassirostris swimming steadily, while A. hudsonica had short active feeding periods. The A. hudsonica movement patterns match those observed in an Acartia congener (Kiørboe et al., 1996), while the P. crassirostris movement patterns consisted of more continuous swimming than previously observed (Bradley et al., 2013).
Based on the differences in their movements, these copepod species may therefore target different prey species and be targeted themselves by different predators; copepods with continuous-swimming movement strategies are more susceptible to predation compared with copepods that use an ambush strategy (van Someren Gréve et al., 2017). Additionally, the swimming and feeding flux of these two copepod species was impacted very differently by changes in temperature and viscosity, with A. hudsonica being affected by viscosity while P. crassirostris was unaffected.
Our hypotheses about the effects of seawater viscosity on copepod movements were partially upheld. Acartia hudsonica feeding flux was reduced by high viscosity but not temperature, and P. crassirostris feeding flux was not affected by either temperature or viscosity (Table 2). Similarly, high viscosity reduced the number of A. hudsonica twitches, possibly indicating a reduction in particle capture, while P. crassirostris twitching was not affected by treatment (Tables 1 and 3). Furthermore, A. hudsonica swimming time and swimming speed were both reduced by high viscosity, while P. crassirostris was not affected (Figs 3, 4, Table 4). Correspondingly, feeding of the congener A. tonsa is severely reduced by increased seawater viscosity, and P. crassirostris feeding is less sensitive to viscosity (Tyrell and Fisher, 2019). Rather than reducing P. crassirostris feeding flux, high seawater viscosity may diminish particle capture efficiency (Koehl, 1981), thus explaining reduced feeding at high viscosity (Tyrell and Fisher, 2019). Therefore, our hypothesis that temperature effects can be explained by viscosity was partially supported by the data. However, our hypothesis that A. hudsonica would be less sensitive to viscosity because of its more northward range was not supported. Future research should broadly investigate other copepod and zooplankton species to determine how they respond to changes in seawater viscosity and temperature and whether latitudinal and feeding mechanism-specific response patterns exist.
Both A. hudsonica and P. crassirostris cleared volumes similar to the 106 body volumes per day required to satisfy energetic requirements (Kiørboe, 2011b; Kiørboe and Jiang, 2013), and A. hudsonica clearance rates match previous measurements of its congener A. tonsa (Kiørboe et al., 1996) (Table 2). Parvocalanus crassirostris clears slightly more body volumes of water per day than A. hudsonica; this higher clearance rate is reflected in a higher respiration per unit body volume (Tyrell et al., 2020).
To our knowledge, we are the first to use the Python libraries OpenCV2 and trackpy to track fluid flow around zooplankton, although other groups have used these libraries in different contexts (Bianco et al., 2013; Urmy and Warren, 2017; Wolf and Heuschele, 2018). These libraries offer an open-source alternative fluid analysis method that works even under conditions when particle image velocimetry (PIV) analysis is not suitable as a result of low particle densities, and we encourage future work to use OpenCV2 and trackpy to track fluid flow.
Increases in seawater viscosity reduced A. hudsonica swimming speed by 23–28% in limited cases (Table 4), compared with previous studies showing a 50% decrease in swimming speed over a comparable viscosity range (Larsen et al., 2008). Acartia hudsonica swimming speeds of 2–4 mm s−1 (Fig. 4A–C) match previously published speeds of Acartia spp. (Buskey et al., 1983; Larsen et al., 2008). In contrast, there was no effect of viscosity on P. crassirostris swimming speed, though cold temperature sometimes caused a 25–26% decrease in swimming speed (Table 4). Parvocalanus crassirostris speeds near 0.5 mm s−1 (Fig. 4D–F) are similar to previous studies (Bradley et al., 2013).
A viscous speed response may have been masked or diminished because of the relatively high cell densities. Copepod swimming speeds slow at a cell density of 104–105 cells ml−1 (van Duren and Videler, 1995), which is the range of cell densities used in our study. We chose a high density of copepods with a high cell density to maximize the number of videos while ensuring that the copepods would have sufficient cells remaining after an acclimation period. Future video studies could investigate the effect of seawater viscosity and temperature on copepod swimming speeds at lower cell densities.
The measured percentage of time that A. hudsonica spent swimming was low (7.9–19.7%) (Fig. 3) compared with a previous study that reported up to 40% active swimming time in a congener (Kiørboe et al., 1996). Previous studies have been limited to frame rates as low as 30–50 frames s−1 (0.02–0.03 s frame−1) (Kiørboe et al., 1996; Turner et al., 1993), which could cause multiple short periods to be classified as one longer period, thereby artificially inflating the measured active swimming time. The percentage of time that P. crassirostris spent swimming (57.9–90.8%) was higher than the 33% previously reported (Bradley et al., 2013).
Overall, the lack of response of P. crassirostris to viscosity may lead to more energy being spent at high viscosities and cold temperatures, while A. hudsonica’s reduction in swimming speed and activity at high viscosity and cold temperature may result in less energy being spent. However, the total amount of energy spent on movement is generally <2% of the total metabolic rate (Vlymen, 1970; Kiørboe et al., 1985; van Duren and Videler, 2003), so this difference in response may instead stem from differing muscle strengths or sensitivity to stimuli rather than from an energy conservation requirement. Furthermore, a sensitivity to seawater viscosity may not necessarily propagate into ecological effects if the growth, reproduction and survival of the copepod is not affected. When considering the energetic balance of ingestion and respiration, the A. hudsonica congener A. tonsa has optimal temperatures lower than those of P. crassirostris (Tyrell et al., in press), and A. hudsonica is found at colder temperatures than P. crassirostris (Lonsdale and Coull, 1977; Sullivan et al., 2007), indicating that A. hudsonica is more adapted to cold environments than P. crassirostris. Future research should investigate these issues, as well as how temperature-related differences in movement, such as the A. hudsonica reduction in the proportion of time spent swimming and swimming speed, translate to ecological impacts on predation and reproduction; less movement of A. hudsonica at colder temperatures may result in less predation under these conditions.
The distinct physiological effect of temperature on copepod appendage beating (Fig. 2) was not reflected in any temperature effect on copepod movements, activity, feeding flux or swimming speed of A. hudsonica, though P. crassirostris swimming speed was sometimes reduced by cold temperature (Fig. 4, Table 4). The lack of a viscous response is unexpected, given that an appendage has a maximum Re of approximately 1 (Cheer and Koehl, 1987), which implies that changes in viscosity should affect movement. The mechanisms that control copepod appendage movement are conserved across life stages, as previous data on copepod nauplii show similar results (Gemmell et al., 2013). Our results match the pattern of larger copepods having slower appendage beating frequencies (Price and Paffenhöfer, 1986; Dagg and Wasler, 1986); A. hudsonica (0.8 mm cephalothorax length) had a slower appendage beating frequency than P. crassirostris (0.35 mm cephalothorax length) (Fig. 2). Parvocalanus crassirostris appendage beating frequency was more severely reduced by cold temperature than that of A. hudsonica (Fig. 2). Acartia hudsonica has amyelinated nerves, while P. crassirostris has myelinated nerves (Buskey et al., 2017). Myelinated nerves respond more quickly to stimuli (Buskey et al., 2017; Lenz et al., 2000) but are more sensitive to low temperatures (Franz and Iggo, 1968). Based on copepod reaction times (2–10 ms) (Waggett and Buskey, 2008), copepod nerves transmit at 100–500 Hz, compared with appendage beating frequencies of 12–75 Hz; a temperature-induced reduction in nerve transmission speed of 3× (Franz and Iggo, 1968) would reduce nerve transmission to 30–170 Hz, possibly lowering nerve transmission frequency below optimal appendage beating frequency and thereby slowing appendage beating.
Most copepod movements were not affected by diet, contrary to our hypothesis. In contrast to the temperature/viscosity treatment, the diet treatment did not affect copepod swimming time (Fig. 3) or movements (Tables 1, 2), or appendage beating frequency of P. crassirostris (Fig. 2), although the appendage beating frequency of A. hudsonica was increased when feeding on T. weissflogii with tracking particles added compared with when feeding on T. weissflogii without tracking particles (Fig. 2). This may reflect an influence of particle size on the appendage beating frequency, as the tracking particles were smaller than the algal cells (3 µm diameter compared with ∼10–20 µm diameter). This lack of behavioral response to diet contrasts with some recent findings showing that food quality impacts behavior (Herstoff et al., 2019), although it is likely that both diets were nutritionally replete. Copepods prefer motile prey over non-motile prey (Atkinson, 1996; Verity and Paffenhofer, 1996) and tend to have higher ingestion rates when feeding on motile prey (Jakobsen et al., 2005; Henriksen et al., 2007). Future studies should investigate the movements and swimming speeds associated with selectivity when feeding on multi-algal diets of differing nutritional quality.
Diet distinctly impacted copepod swimming speeds, with copepods feeding on T. weissflogii consistently swimming 22–39% more slowly than copepods feeding on P. minimum (Fig. 4, Table 4). Changes in swimming speed may serve to maximize capture efficiency depending on cell characteristics (Koehl, 1981). This diet-induced change in swimming speed may propagate into ecosystem-level effects, as swimming speed influences both mating (Kiørboe and Bagøien, 2005) and risk of predation (Buskey, 1994; van Someren Gréve et al., 2017), and ecological models should be used to investigate these endpoints.
Conclusions
The swimming speeds and behaviors of two copepod species are influenced by both temperature and viscosity in different and biologically meaningful ways. Notably, only A. hudsonica responded to increased seawater viscosity. These distinct behaviors highlight the diversity of the zooplankton movement strategies and the differing ways that zooplankton respond to environmental conditions. A better understanding of calanoid feeding and movement behavior may help explain copepod distributions and trophic interactions. Broadly, the individual-level details resolved in videography studies are an important contribution to our understanding of marine zooplankton ecology; videography studies should be expanded to include more copepod and zooplankton species. Future studies should also investigate latitudinal patterns in copepod feeding mechanisms to determine how seawater viscosity and temperature may have influenced the evolution and development of different feeding mechanisms.
Acknowledgements
We thank P. Alatalo, S. Baines, N. Chatterjee, B. Colon, T. Gaylor, E. Herstoff, J. Kraemer, Y. C. Lu, M. Mace, B. Michel, V. Mikros, M. Niemisto, J. Padilla, N. Tyrell and L. Wong for assistance.
Footnotes
Author contributions
Conceptualization: A.S.T., H.J., N.S.F.; Methodology: A.S.T., H.J.; Software: A.S.T., H.J.; Validation: A.S.T.; Formal analysis: A.S.T.; Investigation: A.S.T.; Resources: A.S.T., H.J., N.S.F.; Data curation: A.S.T.; Writing - original draft: A.S.T.; Writing - review & editing: A.S.T., H.J., N.S.F.; Visualization: A.S.T.; Supervision: H.J., N.S.F.; Project administration: H.J., N.S.F.; Funding acquisition: H.J., N.S.F.
Funding
This study was supported by the National Science Foundation [OCE1634024 to N.S.F.; OCE-1433979 and OCE-1559062 to H.J.]; and by Stony Brook University [Graduate Council Fellowship and Turner Fellowship to A.S.T.].
Data availability
Data are archived at the Stony Brook University Library Academic Commons, indexed under the School of Marine and Atmospheric Sciences, SoMAS Research Data 7: https://commons.library.stonybrook.edu/somasdata/7/
References
Competing interests
The authors declare no competing or financial interests.