The fluidic device is composed of a glass slide spin-coated by a thin polydimethylsiloxane (PDMS) layer, a PDMS reservoir for the droplet, and a holder for the electric motor (M20, DC1.5V ~ 3V) (Supplementary Figure S1A). Both the PDMS reservoir and holder are produced by casting PDMS on the top of the 3D-printed molds (CR-3040S, CREALITY, China; minimum printing line width of 0.4 mm and height of 0.1 mm) (Supplementary Figures S1B, C). In detail, to produce molds with a smooth surface, a printing material named PolysmoothTM (Polymaker, China) was used (Feng et al., 2020). The printed template was then positioned amid 75% ethanol or isopropyl alcohol for polishing, the duration of which depends on the structural features of the templates (30 min to 1 h). Following the standard protocol of PDMS casting, the mixer of the PDMS base and the curing agent at a ratio of 10:1 was poured into the 3D-printed molds, degassed, and baked at 45°C for at least 12 h for solidification.

Those PDMS parts are then bonded to the PDMS-coated glass slide after plasma treatment (Supplementary Figure S1D). The 3D-printed lever carrying two metal springs (i.e., wire diameter of 0.25 mm, outer diameter of 2.4 mm, and length of 9 mm) is connected to the electric motor. The elastic deformation of the spring while knocking on the glass slides ensures continuous rotation and thus controlled knocking frequency and amplitude. During the experiment, the whole device is mounted on the top of the microscope stage, where only four corners of the glass slide are fixed. Repeated knocking of the metal spring on the glass slide causes vibration, which propagates across the slides and reaches the droplet. Consequently, mechanically caused deformation of the droplet leads to variations in the internal flow patterns (Zeng et al., 2022).

All experimental procedures on zebrafish were approved by the Experimental Animal Management and Ethics Committees of Northwest University and carried out in accordance with the approved guidelines (NWU-AWC-20190617Z). Adult zebrafish of AB wild-type (wt) strain and Tg (elavl3:eGFP) transgenic line were raised with a standard light cycle (14 h light, 10 h dark) (Kim et al., 1996; Lu et al., 2020). Embryos were obtained by natural crosses and staged as described earlier (Mwaffo et al., 2017). Fertilized embryos were cultured in the E3 medium (NaCl: 0.29 g/L, KCl: 0.012 g/L, CaCl₂: 0.036 g/L, MgSO₄: 0.04 g, pH 7.4) at 28.5°C. For experiments in the vibrating droplet, all zebrafish were tested 72 h post-fertilization (hpf). Among them, wild-type fish were incubated without any treatment (Type I). The zebrafish with disrupted motility (Type II) was treated with 0.012% concentration of anesthetic [3-aminobenzoic acid ethyl ester methanesulfonate, SIGMA (A5040)] for 3 ~ 5 min to simulate the zebrafish with damaged motor nerves. For the recovery group (Type III), the anesthetized fish was put into the water and then waited for 5 min before tests.

A time-lapse of zebrafish within the miniature swimming pool, which is 12 mm in diameter and 1.5 mm in height, was recorded using a commonplace mobile camera at a frame rate of 20 fps. The videos were then extracted and processed using ImageJ software for better visualization and contrast. Trajectories of zebrafish were obtained by determining the fish centroids and tracking their movement using a customized Matlab program. The mean square displacement (MSD) of the zebrafish is calculated using the formula: MSD(τ) = < Δr(τ)²> = < [r(t+τ)-r(t)]²>, where r (t) is the position of the zebrafish centroid at time t, τ is the lag time between the two positions to calculate the displacement Δ r (τ) = r (t + τ) – r (t), and the average < ...> designates a time-average over t.

To assess zebrafish mobility in the flow field generated by the percussion of a droplet, the PDMS-based fluidic device was placed on a Nikon TiE inverted fluorescence microscope. The flow field at different substrate vibrating amplitudes and frequencies was examined by tracking the movement of fluorescent particles (polystyrene microspheres) at a frame rate of 30 fps, the trajectories of which were then analyzed using the software LabPIV. The macroscopic behavior (e.g., rotation) was imaged using 4 × objective. The capacity of the device in high-resolution fluorescent imaging was demonstrated using a 40 × objective. In all experiments, the temperature was maintained at 28.5 and 100% RH using a stage-top incubator.

We, herein, attempt to evaluate the mobility of different zebrafish groups, e.g., the ones with lipin-1 deficiency and anesthetized ones (Lu et al., 2021). Optimally, both the macroscopic fish movement and its neuron system at the single-cell level are monitored in real-time to establish direct connections. Even with the state-of-art motor stage and automated tracking program (Pardo-Martin et al., 2010; Choudhury et al., 2012), it is difficult to catch up with the so-called burst-and-coast swimming style of zebrafish and take fluorescent images in dynamic environmental conditions.

To meet the challenges, we developed a simple-to-use fluidic device, which was produced using 3D printing (Figure 1A and Supplementary Figure S1). In brief, the device is composed of a glass slide and a small electric motor (M20, DC1.5V~3V), which is mounted on a glass slide via a holder made by PDMS. The electric motor is connected to two levers carrying a metal spring. Rotation of the levers drives the metal springs to repeatedly knock on the glass substrate and causes vibration, which propagates to the droplet housing zebrafish. The actuation amplitude and frequency of substrate vibration are controlled by adjusting the spring length and electric power, respectively (Supplementary Video S1). To ensure that the droplet is sitting on a substrate with controllable stiffness and hydrophobicity, a PDMS reservoir is placed on the top of the glass slide. The droplet shape can then be adjusted by tuning the hydrophobicity of the substrate (i.e., the contact angle) via polyvinyl alcohol processing following plasma treatment (Trantidou et al., 2017). To prevent unwanted evaporation, the droplet fluidic device was positioned in a sealed box at 100% RH, and a temperature of 28.5°C to maintain a healthy environment for zebrafish (Zeng et al., 2022).

Using fluorescent particles of 1 μm in diameter as a tracker, our results demonstrate that the various flow patterns can be generated in the droplet by gradually changing the droplet size (via liquid volume), shape (via hydrophobicity of the glass substrate), the vibration amplitude (via lever length), and frequency (via motor rotation). When the vibration amplitude is large (i.e., > 0.2 mm underneath the droplet), the droplet easily deforms, resulting in an uncontrollable and unstable flow pattern (Supplementary Videos S2, S3). It is difficult to determine the frequencies at which flow patterns with two and four recirculating zones were generated (Supplementary Figures S2E–H, M–P). In contrast, at low vibration amplitude (i.e., < 0.1 mm), the flow pattern shows dependency on the vibration frequency (Supplementary Figures S2A–D, I–L). For example, when the droplet volume is 10 μL and is positioned under a vibration amplitude of 0.1 mm, the flow pattern gradually transits from random to four recirculating flow patterns with increasing vibration frequencies, i.e., random → one recirculating zone → four recirculating → two recirculating → random (Figures 1B–D and Supplementary Figure S2). The transitioning process varies with changing droplet shape, determined by liquid volume and substrate hydrophobicity. Notably, the velocity distributes ununiformly in the droplet (Figures 1E–G), showing considerably lower velocity at the center of each recirculating zone.

No regular flow pattern was generated when vibration is induced to bulk solution in a 96-well plate (Supplementary Video S4). These results indicate that the vibration of the substrate induces droplet deformation, which subsequently causes internal flow. The distinctive conformational transition of the droplet at various experimental conditions (i.e., the droplet size, shape, vibration amplitude, and frequency) leads to various internal flow patterns. Consistently, studies by Yang et al. reveal that shape deformation of the droplet leads to variations in the internal flow patterns (Yang et al., 2019). In contrast, the hydrophobicity of the substrate plays no crucial role in regulating the flow pattern in the droplet (Supplementary Figure S2).

Assuming that a zebrafish is an inextensible object within the droplet, the mechanical forces, which force the object to move, include gravitational forces pulling zebrafish down to the bottom, viscous force f_v forcing zebrafish to go with the flow, and centrifugal forces f_c causing zebrafish to bend (Figure 2). When droplet dimension is comparable to zebrafish, the movement in longitude direction is trivial, and therefore, the effects of gravitational force are neglected. In a droplet with a single recirculating flow, viscous force f_v at both ends provides torque forcing the object to rotate with the flow (Figure 2A). When the object rotates about its own central point and is unbendable, the torque of centrifugal forces f_cis zero. Considering also the fact that the flow velocity is considerably lower in the central region of the recirculating flow (Figure 1E), the total torque can, then, be simplified as Torque = f_v1·r+f_v2 nbvcxb·r = 2·f_v·r. Deformation (i.e., bending) of zebrafish in the flowing environment causes deviation of zebrafish from the droplet center (Figure 2B), leading to unbalanced centrifugal forces and driving the object to the marginal region, where f_v is maximized (Supplementary Video S5) (Ha and Gary Leal, 2001).

In two recirculating flows, the torque on zebrafish is zero at the balance position (Figure 2C). Active zebrafish movement leads to increased torque, i.e., Torque= fv1·r-fv2·r′+fv3·r′-fv4·r = (fv1-fv4)·r+(fv2-fv3)·r′, where r and r′ denote the actuation points of viscous forces of two recirculating flow patterns (Figure 2D). Obviously, the action of one recirculating flow pattern (i.e., f_v1 and f_v2) provides counterforce for another one (i.e., f_v3 and f_v4). Therefore, forces on the zebrafish in the droplet are mostly balanced. The viscous force in two recirculating flow patterns causes zebrafish to bend.

In the droplet with four recirculating flow zones, there are two balanced positions, where the total viscous torques are zero (Figures 2E, F). To illustrate, we simplify the flow pattern to a 2D straining flow (u,v)=(ε•x,-ε•y), where ε• is the strain rate of the straining flow simplified from the velocity field within the droplet (Figure 3A). The flow field can be written in polar coordinates as follows:

where u_r is the radial velocity; u_θ is the circular velocity; u is the velocity in the x direction; and v is the velocity along the y axis. The parallel and perpendicular drag coefficients of a slender object, with length L and cross-section radius a, are c_|| = 4πμ/c and c_⊥ = 8πμ/c, respectively, where μ is the liquid viscosity and c = 2ln(L/a)-1 is the drag coefficient correction (Batchelor, 1970). The relative velocity of the object can then be written as ur′=ur, and uθ′=uθ-θ•·s, where θ•=dθdt is the radial velocity of the object and s is the arc length. When the object is positioned at angle θ, the torque is then

It is demonstrated that the orientation of torque varies with the rotation angle of zebrafish in a small droplet, i.e., with a size comparable to zebrafish. There are two balanced positions, where the torque on zebrafish is zero, i.e., position-1: π/2 and position-2: π (Figure 3B). Deviation of the object from both position-1 (π/2) and position-2 (π) leads to increased torque (Figure 3C). The difference is that direction of the torque is the same as zebrafish rotation at position-1, forcing the object to go further way, which is indicated by the arrows in Figure 3C. At position-2, the torque direction is opposite to zebrafish movement and thus drives the object to return to its original position.

By generating four recirculating flow patterns surrounding a zebrafish (Figure 4A), we observed that zebrafish with high mobility tend to stay at position-1, where the heads point against the flow direction (i.e., π/2 in Figure 4B) (Chu et al., 2004), which is consistent with the previous observation that wild-type zebrafish tend to swim against the current (Mwaffo et al., 2017; Oteiza et al., 2017). The narrow region, where wild-type zebrafish (type I) stabilize themselves within the time span of 5 min, ranges approximately from 80 to 100. As indicated in the previous section, the deviation of zebrafish from position-1 induces a torque forcing the zebrafish to move toward position-2. Therefore, the stabilization of the zebrafish at position-1, where the head of the zebrafish is against the current, suggests good mobility, balance, and stability (Figure 4A) (Mwaffo et al., 2017; Oteiza et al., 2017). In contrast, the zebrafish with disrupted mobility (type II) is forced to position-2 (angle of ~0±10°) (Supplementary Video S6), suggesting the loss of mobility to escape from the mechanical trap (Figures 3C, 4B, C).

These results demonstrate that zebrafish with intact mobility can distinguish themselves from disabled ones. Unlike the zebrafish in the miniature swimming pool (Supplementary Figure S3), whose responses depend on the zebrafish state and may sometimes require poking with a stick, the ones in droplets are constantly exposed to environmental stimulation of continuous flow. Wild-type zebrafish have to be active in order to stay at position-1, which eliminates the possibility of false results. For example, wild-type Fish04 in Supplementary Video S7, by lacking stimulation, remains immobile for a relatively long period of time as compared to others. No similar phenomena were observed in the vibrating droplet. Even though the recovered fishes position themselves in a relatively broader region than the wild-type fishes (~90 ± 40°), these fishes can be easily distinguished from disabled ones (type II). The positioning of different fishes is stable within the duration of experiments (Figure 3C), indicating that the proposed device is a suitable platform for assessing zebrafish mobility.

The conclusion is further supported by statistical analysis of 5–10 zebrafish under each experimental condition (Supplementary Section 1). It is demonstrated that the anesthetized zebrafish (i.e., type II with compromised mobility) and the ones with motor dysfunction (i.e., lipin-1 deficiency) can be better distinguished from the wild-type zebrafish (type I) in the four recirculating flow patterns as compared to the two recirculating flow patterns (Figure 5A). In the two recirculating flow patterns, all five type I (i.e., wild-type) zebrafish stay in a narrow region of ~30 (Figure 5C), where their heads point against the current (Figures 5A, C, D). The probability can be calculated using p=(30360)5·C121, which gives a p-value of 4.8 × 10⁻⁵. In the four recirculating flow patterns, the head of four type I zebrafish points to both 90 and one toward 270 (Figure 5D), all of which fluctuate in a range of ~30. This gives a p-value of ~0.01. Even though zebrafish with disrupted mobility (type II and lpin-1 morphants) position their heads at different angles ranging from −20 to 20 and 180 to 220 in four recirculating flow patterns, the p-value remains as small as 0.024, which is smaller than the threshold for statistical significance 0.05. In contrast, zebrafish recovering from anesthetization show a broad distribution of angles in the vibrating droplet (Figures 5G, H). A clear dependence of zebrafish position distribution on the recovery time is observed, indicating that the relative angle of zebrafish is closely associated with its mobility (Figure 5I).

Another key feature of the device is that zebrafish are maintained in a droplet of comparable size, where the movement of zebrafish is restricted. The stabilization of zebrafish in the four recirculating flow patterns allows us to study zebrafish behavior in dynamic environmental conditions during high-resolution fluorescence imaging. Our results demonstrate that using 4 × objective, the distribution of neurons (Figure 6A) and movement of internal organs (Supplementary Video S8) can be clearly imaged at an exposure time of ~50 ms. At higher magnification, i.e., 40 × , neurons and their axons were observed (Figures 6B–D), showing a similar resolution and clarity as those obtained from the fixed samples. We, therefore, conclude that the droplet-based fluidic device provides opportunities for us to quantitatively investigate the behavior of zebrafish larvae and perform high-resolution fluorescence imaging.