A dinoflagellate exploits toxins to immobilize prey prior to ingestion

Abstract

Toxins produced by the harmful algal bloom (HAB) forming, mixotrophic dinoflagellate Karlodinium veneficum have long been associated with fish kills. To date, the perceived ecological role for toxins has been relief from grazing pressures. Here, we demonstrate that karlotoxins also serve as a predation instrument. Using high-speed holographic microscopy, we measure the swimming behavior of several toxic and nontoxic strains of K. veneficum and their prey, Storeatula major, within dense suspensions. The selected strains produce toxins with varying potency and dosages, including a nontoxic one. Results clearly show that mixing the prey with the predatory, toxic strains causes prey immobilization at rates that are consistent with the karlotoxins’ potency and dosage. Even prey cells that continue swimming slow down after exposure to toxic predators. The swimming characteristics of predators vary substantially in pure suspensions, as quantified by their velocity, radii of helical trajectories, and direction of helical rotation. When mixed with prey, all toxic strains that are involved in predation slow down. Furthermore, they substantially reduced their predominantly vertical migration, presumably to remain in the vicinity of their prey. Conversely, the nontoxic control strain does not alter its swimming and does not affect prey behavior. In separate experiments, we show that exposing prey to exogenous toxins also causes prey immobilization at rates consistent with potency. Clearly, the toxic predatory strains use karlotoxins as a means of stunning their prey, before ingesting it. These findings add a substantiated critical understanding for why some HAB species produce such complex toxin molecules.

Marine dinoflagellates produce a diverse suite of complex toxins (1, 2), yet their biological raison d’etre remains enigmatic. Among this group, Karlodinium veneficum is a small (∼8–12 μm) athecate phytoplankton, common in coastal aquatic ecosystems (3 –5). It has a mixed nutritional mode, relying on both photosynthesis and phagotrophy for growth (mixotrophy). It is frequently present in relatively low cell abundance (10²–10³ cells·mL⁻¹), but is capable of forming intense blooms of 10⁴–10⁵ cells·mL⁻¹ that have been associated with fish kills (6 –8). Karlotoxins (KmTxs, see description in SI Section S1.1 and Fig. S1), which are produced by K. veneficum (9, 10), generate pores in membranes with sterols (11) and increase the ionic permeability of cell membranes, resulting in membrane depolarization, disruption of motor functions (6), osmotic cell swelling, and lysis (SI Section S1.3). These measured effects have been based on purified toxins (Fig. S2). Karlotoxin type and cell quota vary with K. veneficum strain (5), culture conditions (12), and geographic location (5). Along the U.S. East Coast, K. veneficum strains from south of the Chesapeake Bay produce karlotoxin 2 (KmTx-2), whereas those from within the Bay produce karlotoxin 1 (KmTx-1). The hemolytic potency of KmTx-1 is 10-fold stronger than that of KmTx-2, and toxicity increases with toxin cell quota (5). Roles of karlotoxins have been conjectured to mediate grazing pressures and as an allelochemical (13, 14). However, on the basis of field data showing that cryptophyte’s abundance precedes K. veneficum blooms (15) and laboratory feeding experiments demonstrating that only toxic strains are mixotrophic (16, 17), Adolf et al. (14 –17) hypothesized that karlotoxins provide advantages in prey capture. Here, we investigate whether and how karlotoxins are used during predation by measuring changes in swimming behavior of predators and prey.

Using high-speed cinematic digital holographic microscopy (DHM), we characterize the three-dimensional motion of several K. veneficum strains and their cryptophyte prey, Storeatula major. As described briefly below, and in more detail in Sheng et al. (18), this method enables us to obtain simultaneous data on 3D swimming of thousands of cells in dense suspensions of 50,000–100,000 cells/mL that have substantial depth (e.g., 3 mm, ∼300 cell lengths). In mixtures, predators and prey cells are distinguished on the basis of their different shape and size. The black and gray tracks in Fig. 1 A and B are sample composite time series of in-focus images of S. major cells (only) shortly and 5 h after mixing them with predators, respectively. Note that each track consists of a series of (one of five) in-focus reconstructed DHM images over the entire depth of the sample volume. Samples of these S. major trajectories are also highlighted in green, to demonstrate the complexity of their swimming. When tracks are straight, the cells are not swimming, and consequently are advected, mostly vertically, by the background flow (which is subtracted when the swimming velocity is calculated). In addition, we also superimpose a few complex and representative trajectories of motile K. veneficum cells in red. Additional characteristic trajectories of the K. veneficum strains examined in this study, demonstrating the complexity of their helical trajectories (consistent with prior observations, ref. 19), are provided in SI Section S2, Fig. S4) and in ref. 18.

Fig. 1.

Superposition of reconstructed in-focus holographic images (only one of every five exposures is shown for clarity). Gray trajectories, tracks of prey, S. major (only) after introduction to a K. veneficum, BM1 suspension; green, highlighted samples of S. major trajectories; red, a few samples of K. veneficum BM1 (predator) trajectories. (A) Shortly after mixing; (B) 5 h later; (C and D) captured S. major cells (smaller ones) being ingested by a BM1 cell, (C) a reconstructed hologram and (D) SEM. (E and F) A pair of K. veneficum, BM1 cells interacting (possibly cell division): (E) reconstructed hologram, (F) SEM. Vertical linear tracks belong to immotile prey; convection by the background flow causes their linear motion, which is subtracted while calculating velocity. (Scale bars: A and B, 100 μm; C and E, 5 μm.) The complex motions of motile cells and increasing fraction of immotile ones with time are evident.

We show that in the presence of all toxic K. veneficum strains, S. major cells slow down and become immobilized (Fig. 1, details follow) at rates that are consistent with potency and cell toxin quotas. Additional experiments involving purified toxins confirm that these effects are a result of exposure to toxins (SI Section S3). To facilitate predation of immotile prey, toxic K. veneficum cells reduce their vertical migration and slow down in the presence of prey. In contrast, mixing of S. major and a nontoxic (and nonpredatory) strain of K. veneficum does not alter the swimming characteristics of either organism. Clearly, karlotoxins are mediating this behavioral change in prey.

Swimming Behavior of Predators.

Four strains of K. veneficum (7, 17) with identical genetic signatures (internal transcribed spacer sequence), two from Chesapeake Bay (MD5 and CCMP 1974) and two from estuaries south of Chesapeake Bay (BM1 and CCMP 2064), were examined. MD5 produces no detectable toxin and shows no mixotrophic behavior (15). Of the rest, 1974 produces KmTx-1 (0.3 pg/cell), whereas BM1 and 2064 produce KmTx-2 in a cell quota ranging from 1 to 2 pg/cell (5). The long-term ingestion rates of prey by the toxic strains decrease slightly in the order 1974 > BM1 > 2064 (15). We quantify the swimming of all of the above-mentioned organisms before, shortly after, and 5 h after mixing, along with effects of purified toxins without predator on S. major swimming.

Fig. 2 contains a series of joint probability distribution functions (PDF) that characterize the swimming behavior of various cells on the basis of velocity magnitude and radius of their helical trajectories, as illustrated in Fig. 2A. Corresponding ensemble averaged velocities, radii, and angular velocities are also summarized in Table 1. Each data point provides the mean value and its SD, the latter demonstrating the extent of variability in the statistics over time and among individual cells. As summarized in Table S1, the uncertainty in mean statistics, as estimated using the bootstrap method, is ∼1%. Each section of Fig. 2 also shows the total number of tracks analyzed to obtain the specific PDF, along with the fraction of cells that are motile. Only motile cells are included in characterization of swimming. Fig. 2 B presents the K. veneficum swimming statistics, the Top row before mixing with S. major, the Middle row shortly after mixing, and the Bottom row 5 h later. As is evident, in isolation, swimming velocity and helix radius of K. veneficum vary substantially among strains. Both MD5 and BM1 swim in left- and right-handed helices, whereas 1974 and 2064 are preferentially right-handed, with small fractions performing left-handed maneuvering at low speed. It should be noted here that unlike previous observations, in which dinoflagellates swam only in right-handed helices, e.g., ref. 20, two of the present strains are almost equally likely to swim in left-hand and right-hand trajectories.

Fig. 2.

Joint probability density functions (PDFs) of velocity magnitude (V) and radius of helical trajectories (R). (A) Sample trajectory of a BM1 cell, color coded with velocity magnitude. Only one of every four frames is shown for clarity. In this example, the cell slows down at the end of the time series, when it is located very close to a prey. (Inset) Definitions of velocity, radius of a helix, and angular velocity quantifying the 3D motion of dinoflagellates. (B) Several strains of the dinoflagellate K. veneficum before (Top), shortly after (Middle) and 5 h after (Bottom) introducing them to prey (S. major). R < 0 and R > 0 indicate left-handed and right-handed helices, respectively. Total number of analyzed motile (only) K. veneficum tracks and their fractions of the total K. veneficum population are shown. (C) Response of S. major, swimming velocity and helix radius, before and shortly after introducing them to several K. veneficum strains, as indicated above each frame. Numbers in frames: Upper Left, the total number of analyzed tracks of motile cells and their fraction of the total S. major population; Upper Right, average total number of S. major cells in the sample volume during the test. (D) Response of S. major 5 h after exposure to low-concentration exogenous karlotoxins. Numbers in frames: same as in C. (E) A sample image showing a KmTx-1-producting strain 1974 capturing a cryptophyte prey. The prey cell is evidently swollen from its original size of 6–8 μm to 10 μm.

Table 1.

Summary of motilities of motile K. veneficum and S. major

	K. veneficum (motile)					S. major (motile)
Strain	V ± σV (μm·s−1)	R ± σR (μm)	ω ± σω (rad·s−1)	Dzz/ν	Dzz/ (i = x, y)	V ± σV(μm·s−1)	R ± σR (μm)	ω ± σω (rad/s)	Dzz/ν	Dzz/(i = x, y)	Clearance rate (no. cells−1·h−1)
S. major						86.4 ± 47.0	5.8 ± 6.1	7.3 ± 4.0	0.45	1.8
MD5	81.3 ± 44.9	4.57 ± 4.9	6.98 ± 3.7	2.67	9.1
MD5 + S. major (h0)	84.5 ± 48.6	4.6 ± 5.3	6.9 ± 2.2	2.51	8.9	85.2 ± 46.1	5.1 ± 5.9	7.2 ± 3.8	0.45	1.9
MD5 + S. major (h5)	82.3 ± 50.1	4.7 ± 5.1	6.8 ± 3.0	2.57	9.0	86.8 ± 40.1	6.1 ± 4.8	7.8 ± 2.5	0.45	1.7	0.03 ± 0.12
1974	102.3 ± 56.4	9.2 ± 8.6	5.67 ± 2.9	1.01	2.0
1974 + S. major (h0)	160.4 ± 59.6	16.2 ± 5.7	8.7 ± 6.1	0.85	1.6	42.7 ± 37.7	2.9 ± 4.3	8.1 ± 4.1	0.28	1.5	0.39 ± 0.13
BM1	111.2 ± 55.15	9.3 ± 8.8	6.7 ± 3.1	2.05	2.6
BM1 + S. major (h0)	81.8 ± 55.5	6.5 ± 7.4	6.4 ± 3.0	1.15	2.0	65.1 ± 41.9	4.1 ± 4.8	6.9 ± 3.2	0.45	6.3
BM1 + S. major (h5)	92.7 ± 43.6	8.7 ± 7.5	5.6 ± 2.7	1.28	2.5	69.8 ± 44.6	5.1 ± 6.0	6.7 ± 3.3	0.5	1.7	0.36 ± 0.06
2064	80.9 ± 38.9	6.5 ± 6.8	5.0 ± 2.7	0.78	3.2
2064 + S. major (h0)	37.8 ± 40.1	3.76 ± 5.0	6.8 ± 3.8	0.64	3.1	81.7 ± 44.4	4.7 ± 5.2	6.9 ± 3.5	0.72	4
2064 + S. major (h5)	59.4 ± 35.1	4.7 ± 5.1	5.5 ± 3.0	0.61	3.1	63.2 ± 41.4	4.2 ± 5.6	6.4 ± 3.0	0.25	1.4	0.30 ± 0.11
S. major + methanol (h5)						78.85 ± 29.7	3.58 ± 3.63	6.51 ± 2.92
S. major + KmTx-1, 2.5 ng·mL−1 (h5)						60.77 ± 36.7	3.09 ± 3.03	7.92 ± 3.56
S. major + KmTx-2, 2.8 ng·mL−1 (h5)						70.3 ± 28.44	5.17 ± 4.74	5.7 ± 4.3

The mean and SDs of swimming are characterized by velocity (V), helical radius (R), and helical angular velocity (w) that indicates the speed of an organism completing a helical full circle. Swimming-induced dispersion coefficients (D_ii, i = x, y, z) are computed using autocorrelation functions of swimming velocity components. The dispersion coefficient associated with the preferred vertical direction, D_zz, normalized with kinematic viscosity of seawater at 20 °C, ν = 1.05 × 10⁻⁶ m²s⁻¹, suggests that swimming organisms disperse at the same rate as the momentum diffusion of background flow. The preferences of vertical dispersion by difference strain are indicated by the ratio between vertical component and the mean of horizontal directions. h0, immediately after introducing prey; h5, 5 h after introducing prey. The bottom three rows present the motilities of S. major 5 h after they were exposed to low dosage exogenous karlotoxins. Clearance rate is defined as the number of prey cells that disappeared per predator cell per time.

Nontoxic K. veneficum MD5 shows no change in swimming behavior upon introduction of S. major for up to 5 h. Thus, we use it as a control for noninteracting species. Conversely, PDFs of 1974 (Fig. 2B, second and third rows) become bimodal, with one peak at low speed and small radius and the other at high speed and large radius. Of the low speed cells, 37 (38%) of the 98 tracked are located near, or connected to, a prey, i.e., appearing similar to the BM1 and prey cell in Fig. 1C, indicating that successful feeding commences immediately. Those that are not located very close to prey are positioned within five body lengths. On the basis of nearest neighbor distance PDFs (SI Section S4 and Fig. S5), the 1974 cells that continue swimming fast have a slight tendency toward being located near prey in comparison with randomly distributed cells. Initially, both KmTx-2 strains slow down appreciably in response to introduction of prey. However, very few (9%) start immediately feeding. After 5 h, 60% of the BM1 cells still swim, mostly in right-handed helices and at 83% of prepredation speeds (Table 1). Of them, 18%, mostly slow ones, are feeding (e.g., Fig. 1 C and D). A few immotile cells are also actively ingesting. The 2064 strain follows the same trend, but with different percentages (14%). Nearest neighbor distance statistics for 2064, presented in ref. 18, reveal that predator and prey cells are randomly distributed relative to themselves, but in a mixed culture, predators are preferentially clustered around prey cells.

Furthermore, ensemble averaged Lagrangian autocorrelation functions for each velocity component (Material and Methods and SI Section S5), whose time integral is the swimming-induced dispersion rates, D_ii (21 –23), show conclusively (Table 1) that in the phototrophic mode, all strains preferentially spend more energy in vertical migration (Fig. S6), as compared to horizontal motions. This trend monotonically decreases with increasing toxicity. Visually, helices of all strains are preferentially aligned vertically, as shown in Fig. 1 and Fig. S4. The implication is that assemblages would spread vertically faster than horizontally. Introduction of prey reduces the dispersion rates of all toxic strains (Fig. S7). In two cases, 1974 and BM1, the ratio of vertical to horizontal diffusion also decreases, by 20 and 27%, respectively, but preference toward vertical migration largely remains (see red trajectories in Fig. 1 for BM1), especially for 2064.

Response of Prey to Predators and Purified Karlotoxins.

Before mixing with predators, motile S. major cells swim in both left- and right-handed helices. The PDF (Fig. 2C, left panel) shows a fraction swimming at high speed and another significant group moving slowly. We quantify the response to toxic strains on the basis of both changes in swimming behavior (Fig. 2C and Table 1) and the fraction of cells that become immotile (Fig. 3 and straight tracks in Fig. 1 A and B). Statistics on clearance rate, i.e., the rate of prey disappearance per predator cell, are also provided in Table 1. In the control sample, which is mixed with MD5, all PDF peaks persist with a shift in probability toward high speed. The immotile fraction remains unchanged, at ∼30%. Conversely, exposures to toxic strains and purified toxin extracts, the latter shown in Fig. 2D, substantially reduce the speed of motile cells and increase the fractions of cells that become immobilized (Fig. 3). KmTx-1-producing strains cause an immediate reduction in velocity of motile cells (by >50%, see Table 1), and a substantial increase in the fraction of immotile ones, from 31 to 54%. In 2.5 ng·ml⁻¹ extracts of this toxin, a small quantity that would be generated by ∼1,000 cells·ml⁻¹, 52% become immobile immediately and >90% 5 h later. During exposure to this toxin, as illustrated by the sample microscopic image presented in Fig. 2E, the prey (left, orange one) swells from its original 6-μm size to ∼10 μm. Note that because the exogenous toxin is water insoluble, it is dissolved in methanol and then mixed with the culture. Whereas the karlotoxins are fully recoverable upon addition to media, once cryptophytes are added, the toxins become undetectable. To rule out the effect of carrier addition, we have also measured the motility of S. major cells in the same concentration of pure methanol (Fig. 2D, second panel). As is evident, the exogenously added methanol has little effect on the prey’s motility.

Fig. 3.

Fraction of immotile cells in suspensions before, shortly after mixing predator and prey, and 5 h later, as well as fraction of immotile prey cells 5 h after exposure to exogenous karlotoxins and toxin solvent. (Upper) K. veneficum strains; (Lower) S. major. (Left) S. major and K. veneficum mixtures; (Right) S. major exposed to exogenous karlotoxins and solvent alone. Error bars: SD of instantaneous data from the time average (mean values that deviate by more than twice the SD are significantly different at the 5% confidence level). Many of the immotile prey cells are attached to predator cells. Substantial fractions of S. major cells become immotile shortly after exposure to K. veneficum 1974. In the presence of K. veneficum BM1 and 2064 the immotile fraction increases only after 5 h. 0h, immediately after S. major is introduced; 5h, 5 h later.

Short-term responses to KmTx-2 strains are much weaker, and levels are consistent with toxin quotas produced by the predators. In a BM1 suspension, S. major slows down by ∼25%, but there is no appreciable change in its motility fraction initially. In the presence of 2064, behavior modifications are mild. Five hours later, the velocity of swimming S. major cells is lower than control values by 20–25% for both KmTx-2 cases, and the fraction of immotile cells almost doubles, to 55 and 59% in BM1 and 2064 suspensions, respectively. In 2.8 ng·ml⁻¹ KmTx-2 toxin extract, 70% become immotile after 5 h. Figure 1 A and B provides a striking graphical demonstration of the decreased fraction of motile S. major cells with time. Whereas very few tracks are straight shortly after mixing the prey with KmTx-2-producing predator, 5 h later, most of the S. major cells are immobile. Also, in Fig. 1B, 13% of the vertical linear tracks (immotile cells) involve attached cell pairs. Some involve a K. veneficum cell predating on S. major, as in the magnified example and corresponding SEM image presented in Fig. 1 C and D, respectively, but others consist of predator pairs, possibly cell division or mating, as shown in Fig. 1 E and F.

Before mixing within the control strain, dispersion coefficients of S. major also display preference toward vertical migration (Table 1). The dispersion decreases immediately by ∼40% in all directions upon exposure to the KmTx-1 strain. When mixed with KmTx-2 strains, in both cases, the magnitude of more than doubles initially and then returns to premixed ratios after 5 h. Is the initial enhanced preference toward vertical migration an “escape strategy” before the less potent karlotoxin takes effect?

Discussion

Mixotrophy substantially enhances the growth rate of K. veneficum (SI Sections S1.4, Fig S3 and ref. 15). Here we show that karlotoxins are a vital instrument in the predation process. Clearly, in the presence of toxic K. veneficum strains (only), prey cells slow down and become immobilized at rates that are consistent with phenotypic toxicity and dosage of toxins. Predation, appearing as predator–prey attachment in holographic and microscopic images (Movie S1), occurs concurrently with reduction in prey motility. Exposure to the KmTx-1 strain causes immediate reaction, whereas the response to KmTx-2 strains is slow, but effective in a period of hours. Immobilization occurs also when prey is introduced to toxins without predators, confirming the causality between motility reduction and presence of toxins. Furthermore, prior studies have shown that adding exogenous KmTx-2 (25 ng·mL⁻¹) to feeding cultures of 2064 and 1974 enhances their ingestion efficiency by nearly 3-fold (14). The presently measured trends in clearance rates of prey cells are consistent with toxin potency and cell quota, and the actual values agree with previously published ingestion rates by the various strains (15). In contrast, S. major does not alter its swimming behavior in the presence of a nontoxic strain. Whereas few expect that harmful algal bloom (HAB) species make toxins specifically to kill fish, fewer still have been able to demonstrate why HABs make toxic substances in the first place. Use of KmTx to alleviate grazing, as shown before (13), and to capture prey, as comprehensively demonstrated here, would serve to increase the population growth rate of K. veneficum in nature and may foster bloom formation for this cosmopolitan species.

We caution the reader that given the substantial diversity in dinoflagellate toxins and modes of feeding, we do not intend to imply here that all toxins are used for predation. However, for cases with structures that are similar to that of karlotoxin, e.g., amphidinol (SI Section S1,1), produced by the mixotrophic Amphidinium sp., one may hypothesize a similar role in predation. We also have not addressed the method of delivering toxins from predator to prey. Presumably, although it is unknown where karlotoxins are located in the cell, they must be readily available for release because exposure to mild shear during filtration or centrifugation causes almost complete release of toxins (10). Conversely, in pure (unforced) suspensions, 95% of the toxin remains cell bound (10); i.e., very little is released without stimulation. When prey is introduced, cell quotas are reduced by >50% (SI Section S1.4 and Fig. S3), indicating release, and they become prey bound and undetectable. Due to its low solubility, it is reasonable to assume that the initial karlotoxin administration, before predator and prey are connected, requires close proximity, but we do not know whether it involves direct contact. We hypothesize why toxic K. veneficum slows down during predation in spite of very different swimming and migration patterns in isolation. Our argument is that once administered, there is no advantage for the predator to leave the proximity of an intoxicated prey that is in the process of being immobilized.

Materials and Methods

Materials.

All strains were maintained phototrophically as unialgal cultures in 15 ppt ESAW culture medium as described (5). On the day of observations (7 h into the light period), equal volumes of K. veneficum strains and S. major cells were mixed to give a starting cell density of 35,000 cells·mL⁻¹ and predator to prey ratio of 2.5. All experiments were performed with laboratory ambient lighting, and samples were returned to a lit incubator between observations. Because K. veneficum does not ingest cryptophytes or grow when maintained in the dark for several days (17), suggesting that this organism is an obligate phototroph, we opted to perform the measurements during the light phase of the normal photoperiod. Holograms were recorded for 20 s before mixing predators with prey, shortly after mixing, and 5 h later. The cells were also observed under a microscope to confirm feeding.

Exposure to Purified Toxins.

To measure effects of purified karlotoxin on cryptophyte cells, an exponential phase culture of S. major (15 ppt ESAW culture medium) was diluted to 35,000 cells·mL⁻¹. The PSII inhibition assay was conducted on 1-mL aliquots held in the cuvette of a ToxY PAM dual channel photosynthesis analyzer (Heinz Walz). In independent, paired assays with a volume of <6 μL of purified KmTx1-1 stock, (25 or 250 μg·mL⁻¹) were added to a 1-mL S. major aliquot to produce toxin levels between 50 and 1,000 ng·mL⁻¹. For each exposure to KmTx1-1, the same volume of MeOH (carrier) was added to a control. Samples were monitored in the ToxY PAM fluorometer and the percentage of PSII inhibition (relative to control) was recorded after 10 min, using ToxYWin software. Samples were then immediately diluted 10-fold and analyzed on a Coulter Multisizer II for cell abundance and diameter. A swelling index was calculated for each exposure as the cell diameter of the experimental treatment divided by the diameter of the control. Solutions of purified (10) karlotoxins (KmTx-1–3 and KmTx-2) in methanol at concentrations of 2.5, 25, and 250 ng·mL⁻¹ for KmTx-1–3 and 2.8, 28, and 280 ng·mL⁻¹ for KmTx-2, were prepared. These exogenous concentrations correlate to ≈10³–10⁶ cells·mL⁻¹. Swimming of S. major after toxin addition and toxin concentrations were monitored after 0.25, 2.5, and 5 h.

DHM Setup and Analysis.

Measurements were performed using in-line, cinematic (120 frames/s) digital holographic microscopy (18, 24, 25), using the optical setup described in ref. 18. The sample volume was illuminated by a collimated He-Ne laser beam (red, 0.6328-μm wavelength), and holograms magnified at 20× were recorded on a 1,024 × 1,024 pixels camera, at a resolution of 0.975 μm/pixel. The samples were placed in a 3 × 3 cross section, 40-mm high cuvette, and the sample volume was 0.8 × 0.8 × 3 mm, the latter being the depth. Digital reconstruction and analysis in-house developed software enabled simultaneous tracking of the particle within each suspension. Predator and prey cells were distinguished on the basis of their shape and size by examining in-focus images of each track. Velocity and helix radius were computed from the 3D position vector, , using

where is a unit vector aligned with the vector indicated in the subscript, the number of dots indicates order of time derivatives, k is radius of curvature, and τ is torsion. Further details along with associated uncertainties are provided in supporting information in ref. 18. Using extensions to Taylor’s theory for scalar and particle diffusion (21 –23, 26), swimming-induced dispersion coefficients were determined from an ensemble average of the Lagrangian autocorrelation function of particle velocity:

Here, ξ and η are time, u_i is the fluctuating component of velocity component, and denotes ensemble averaging over trajectories of the same species. The asymptotic value of yields the Fickian diffusion coefficient.