Microswimmer Combing: Controlling Interfacial Dynamics for Open-Surface Multifunctional Screening of Small Animals

Abstract

Image-based screening of multicellular model organisms is critical for both investigating fundamental biology and drug development. Current microfluidic techniques for high-throughput manipulation of small model organisms, although useful, are generally complicated to operate, which impedes their widespread adoption by biology laboratories. To address this challenge, this paper presents an ultra-simple and yet effective approach, ‘microswimmer combing’, to rapidly isolate live small animals on an open-surface array. This approach exploits a dynamic contact line-combing mechanism designed to handle highly active microswimmers. The isolation method is robust, and the device operation is simple for users without a priori experience. The versatile open-surface device enables multiple screening applications, including high-resolution imaging of multicellular organisms, on-demand mutant selection, and multiplexed chemical screening. The simplicity and versatility of this method provides broad access to high-throughput experimentation for biologists and opens up new opportunities to study active microswimmers by different scientific communities.

Graphical Abstract

A dynamic contact-line combing mechanism enables a novel and ultra-simple approach to rapidly isolate active multicellular organisms on an open-surface microgel array. The open-accessible device offers a versatile platform for multifunctional screening of small animals. The approach can be easily adopted by the general scientific community for a wide range of applications from fundamental developmental biology to neuroscience studies to drug development.

1. Introduction

Multicellular small organisms, such as Caenorhabditis elegans and Danio rerio, offer excellent in vivo models to study system-scale biological mechanisms that connect gene expression to neural activities to whole-organism behavior.^{[1, 2]} High-throughput image-based screening of these active microswimmers are crucial to quantitatively characterize these complex and multiscale phenotypes. The resulting high-content information has proved invaluable for modeling human disease pathogenesis and developing drug candidates.^{[3, 4]} There are two major challenges of performing high-throughput whole-organism level screening: one is to handle and rapidly isolate the small live animals; the other is to precisely control the microenvironment that stimulates individuals. In the last decade, various microfluidic tools have been developed to meet these needs. One such well-studied microswimmer model organisms is the soil nematode C. elegans; different closed-channel microfluidic systems are designed and coupled with automated microscopy to isolate large numbers of C. elegans in microchannels for high-resolution imaging or in microchambers for behavior analysis.^[5–8] Comprehensive forward genetic screening by image-based deep phenotyping and large-scale drug screening are hence made possible.^{[3, 9]}

Despite demonstration of these powerful applications, the adoption of these microfluidic technologies into generic biology laboratories has been slow. This is primarily because most biology labs are not equipped with the necessary engineering expertise for microfluidic workflows, which includes a cleanroom environment and skills to fabricate microfluidic devices, and sophisticated off-chip controllers and trained personnel to operate these microfluidic systems.^{[10, 11]} These additional off-chip components introduce potential points of failure, which increase the challenge for non-experts. Further, different microfluidic device designs are usually optimized for only one specific screening function, such as mutant sorting through high-resolution imaging of immobilized animals or freely moving behavioral analysis in response to chemical stimuli.^{[6, 7]} Prototyping and fabricating a library of devices to perform different screening functions is time-consuming and can be expensive.^[11]

To lower the barrier for screening technologies, recent efforts have been reported for developing simpler platforms. One promising approach is open-surface microfluidics.^[12] By either engineering the surface wettability or the surface morphology of the device substrate, open-surface microfluidic approaches autonomously partition liquid buffers containing biological samples into nanoliter to microliter droplets on a predetermined array.^{[13, 14, 15]} These devices have found many successes in cell-based screening, largely due to their ease of use and the ability to parallelize device operation. Separation of passive entities such as cells or embryos can be performed with these devices by sedimentation and liquid partitioning without the need for off-chip equipment.^{[16, 17]} By eliminating enclosed channels, open-microfluidic devices are more robust against fouling and are more versatile in accommodating different screening functions. The open accessibility of individual droplets allows the customized selection and stimulation of cells,^{[15, 18]} while the multiwell array format is compatible with most standard biological assay workflows and liquid handlers.^{[11, 19]} Although these techniques have been successfully implemented in manipulating cells, adapting them for screening active multicellular microswimmers, like C. elegans, is not as straightforward. The challenge is that these small animals usually have an irregular body shape and are highly mobile, which make them difficult to handle and isolate by the existing sample loading mechanism through droplet partitioning.

To address this challenge while retaining the benefit of open microfluidics, we have developed ‘microswimmer combing’, a new approach that takes advantage of fast interfacial dynamics to isolate active small animals. By engineering a 2D microgel array device, we control local capillary pressures at the interface between a liquid film of microswimmer suspension and the heterogenous device substrate. The unique ‘combing’ approach drives a rapid movement of local contact lines, which dominates over the active locomotion of microswimmers, and hence achieves open-surface sample isolation. The device can be easily prototyped using common and readily available materials and, importantly, can be fabricated outside a cleanroom using machines that cost only a few hundred dollars. The sample loading operation is fast (under 30 seconds per device), simple, and robust without any external connections, pumps and controllers. General users with no engineering expertise can master the operation within minutes. The open-accessibility of our device allows for individual interrogation of microswimmers. Using C. elegans as a model organism, we demonstrated the versatile utility of our device through three common screening functions, including in vivo high-resolution imaging, on-demand mutant selection and retrieval, and multiplexed chemical screening. We thus envision this technique to be widely adopted in various whole-organism screening applications.

2. Results and Discussion

2.1. Working principle of ‘microswimmer combing’

To investigate model microswimmers with individual animal resolution, we designed the ‘microswimmer combing’ method to rapidly array microswimmers into isolated spots on a planar surface from a population in suspension. The open-surface substrate for housing individual animals consists of an array of separated hydrophilic microgel pads surrounded by a continuous, less-hydrophilic plastic surface on a regular microscope glass slide (Figure 1A and B(i)). To demonstrate the convenience of device prototyping, we used materials that are commonly available to fabricate the device in this study: Kapton tape with a polyimide film carrier for the continuous surface and a polyethylene glycol (PEG)-based hydrogel for microgel pads. The Kapton tape was directly laser cut and microgel pads were made simply via discontinuous dewetting (see Experimental Section/Methods, Figure S1 and Movie S1).^[20] These microgel pads provide a hydrated environment for the animals similar to the agarose pads frequently used during microscopy. Such a two-dimensional and yet heterogeneous surface enables our equipment-free animal isolation strategy.

Figure 1.

Overview of the ‘microswimmer combing’. (A) The image of an open-surface microgel array device. It is fabricated on a regular size microscope glass slide (3” * 1”) and contains 319 microgel pads. The microgel pads are dyed in green color only for illustration. Scale bar: 10mm. (B) The operation workflow. (i) The 2D, open-surface device consists of an array of separated hydrophilic microgel pads (shown in green) surrounded by a continuous, less-hydrophilic plastic surface (shown in orange) on a supporting glass slide (shown in light yellow). (ii) A drop of microswimmer suspension is dispensed on the substrate. In this example, the microswimmer is represented by active C. elegans. (iii) The ‘combing’ process is performed by squeezing the suspension drop into a thin liquid film and sliding it across the microgel array. Individual microswimmers can be separated and captured when the receding edge of the liquid film travels though the microgel pads. (iv) The array of live microswimmers are isolated on individual microgel pads after the suspension liquid film is combed through the entire device. (C) The micrograph of an open-surface microgel array loaded with C. elegans. Five microgel pads with successful single worm isolation are highlighted. Scale bar: 2mm.

The ‘combing’ isolation workflow is shown schematically in Figure 1B. To load active microswimmers, we dispensed a drop of animal suspension on the substrate (Figure 1B(ii)). To capture individual animals on microgel pads, we simply used a glass slide to squeeze the drop into a thin liquid film and slid it across the microgel array (Figure 1B(iii)). The thickness of the suspension film, which is important for capture efficacy, can be specified by a spacer on the glass slide. Due to the fast contact line dynamic on the heterogeneous substrate, individual microswimmers were separated and captured when the receding edge of the liquid film traveled through the microgel pads (Figure 1B(iii)–(iv)). This workflow is inspired by the ‘capillary assembly’ method, which integrates micro- and nanoscale objects into ordered arrays.^[21] However, in contrast to the ‘capillary assembly’ mechanism, which combines equilibrium thermodynamic forces and structured cavity confinements to trap static colloidal particles, ^[21] our method exploits controlling the dynamic movement of local menisci on a 2D surface to isolate actively moving microswimmers with a larger and variable body size. Thus, we name this technique “combing”. An example of the microgel array loaded with microswimmer C. elegans after the combing process is demonstrated in Figure 1C.

In contrast to existing cell-based open-surface microfluidics using gravity and droplet partitioning for sample loading,^{[13, 15, 17]} our ‘microswimmer combing’ is specifically designed to capture actively moving microswimmers on microgel pads. It does so by taking advantage of the moving interfacial contact line (Figure 2A). By design, there is a significant difference in wettability between the hydrogel surface and the plastic surface (Figure 2B); the imbalance of local capillary pressures on these two surfaces drives the rapid collapse of opposite menisci between two adjacent microgel pads as the receding edge of the liquid film sliding through the microgel pad (Figure 2A–B). Individual animals are thus pinned by the collapsing menisci at the boundary of microgel pads (highlighted by red arrows in the time-lapse image sequence in Figure 2C). When the receding edge passes, the swimmers are subsequently loaded onto the pads by their own locomotion, as well as by a weak fluid convection as the collapsed meniscus recedes to the hydrophilic pad (orange arrows in Figure 2C). The liquid film of animal suspension is easily slid by hand with a speed of a few millimeters per second. This simple action results in ~100 C. elegans isolated on the substrate within 30 seconds. In contrast, conventional methods of mounting animals would involve manually picking animals sequentially on a hydrogel pad with anesthetics, which is labor-intensive and time-consuming, especially when handling a large number of animals, potentially resulting in injury or death by dehydration during the process. In contrast, this fast combing method easily keeps animals hydrated.

Figure 2.

The mechanism of open-surface microswimmer isolation. (A) The schematics of the microswimmer isolation mechanism. When the receding edge of the liquid film passes through the microgel pad, the opposite menisci between two adjacent microgel pads collapse rapidly and pins the microswimmer on the surface, as shown in the magnified schematic on the right. (B) The top-view and side-view schematics of the collapsing menisci between two adjacent microgel pads. The cross-section schematics show the shape of the meniscus on the plastic surface (side view on the right), and the meniscus on the hydrogel surface (side view at the bottom), respectively. The local capillary pressures are defined by the surface contact angles and device geometrical parameters: the thickness of the swimmer suspension film “h”, and the spacing between the two microgel pads “d”. (C) A series of time-lapse pictures and corresponding schematics of the rapid open-surface isolation during the ‘microswimmer combing’. The microswimmers pinned by the collapsing menisci are highlighted by red arrows. Subsequently, the microswimmers loaded onto microgel pads are highlighted by orange arrows. Scale bar: 1mm. (D) The validation of the “collapsing requirement” $\frac{h}{d}<(1-\frac{\text{cos}{\theta }_{\mathit{\text{plastic}}}}{\text{cos}{\theta }_{\mathit{\text{gel}}}})$. The isolation of microswimmers onto microgel pads is successful when the ratio of geometrical parameters “h” and “d” is smaller than a critical value, shown by the blue dots and an example micrograph of successful sample isolation. When the ratio of design parameters “h” and “d” is larger than this critical value, the sample loading is unsuccessful and the microswimmers remain swimming unboundedly in the bulk droplets left on the surface, shown by the red crosses and the example micrograph on the right. Scale bar: 1mm.

The design of successful isolation of active microswimmers is guided by two principles. First, the autonomous collapse of the menisci between two adjacent microgel pads is insensitive to user operation, but rather determined by two geometrical parameters: the thickness of the microswimmer suspension film, “h”, and the spacing between these two microgel pads, “d” (Figure 2B). Given the differential wettability on the substrate, the meniscus collapsing is guaranteed as long as the local capillary pressure on the separated hydrogel surface, “ΔP_gel”, is larger than the one on the continuous plastic surface, “ΔP_plastic”. To meet this condition, the two aforementioned geometrical parameters need to satisfy the “collapsing requirement”, $\frac{h}{d}<\left(1-\frac{\text{cos}{\theta }_{\mathit{\text{plastic}}}}{\text{cos}{\theta }_{\mathit{\text{gel}}}}\right)$ (Figure 2B and Supporting Information Section S1). The inequality provides the driving force to effectively “break” the bulk suspension liquid film and move the opposite menisci towards each other, leaving a micro-scale thin film of liquid and the animal encapsulated around the microgel pad (Figure 2A&C).

We verified this collapsing requirement by testing microswimmer isolation on substrates with different spacing designs between microgel pads (“d”), and with different bulk film thicknesses during combing (“h”). Sample loadings were successful (blue dots in Figure 2D) and animals were isolated on microgel pads when $\raisebox{1ex}{$h$}\!\left/ \!\raisebox{-1ex}{$d$}\right.$ was smaller than a critical threshold. This threshold was empirically determined by the wettability of the polyimide and the PEG-hydrogel surface, shown as a solid blue line in the “d-h” phase diagram in Figure 2D. In contrast, when $\raisebox{1ex}{$h$}\!\left/ \!\raisebox{-1ex}{$d$}\right.$ was larger than this critical threshold, the bulk suspension film could not be broken after the top glass slide was slid through. This resulted in animals swimming freely in the remaining bulk liquid film instead of being pinned on microgel pads (red crosses in Figure 2D).

The second principle considers the speed of the meniscus collapsing compared to that of the microswimmers’ movement. To capture moving individuals from the suspension, the opposite menisci between two adjacent microgel pads need to collapse faster than the active locomotion of microswimmers. We estimated the latter empirically. As evident in Figure 2C, microswimmers were concentrated at the receding edge of the suspension film during the combing process; this indicates that the swimming speed of the microswimmer is generally smaller than the speed by which the liquid film sliding through the microgel array (~millimeter-per-second). Therefore, we scale the maximum swimming speed of microswimmers approximately with the sliding speed of the liquid film. To analyze the relative dynamics, we define a dimensionless “Capture number”, as the ratio between the two timescales: the local meniscus collapsing timescale, “t_c”, and the bulk suspension film sliding timescale, “t_sliding”. This Capture number, $\raisebox{1ex}{$t_c$}\!\left/ \!\raisebox{-1ex}{$t_{\mathit{\text{sliding}}}$}\right.$, is thus determined by $C{a}^{3l_g}{/}_{\alpha h}$, where Ca is the capillary number of the bulk liquid film during combing, l_g is the length of the microgel pad (Figure 2B), and $\alpha =\left(\text{cos}{\theta }_{\mathit{\text{gel}}}-\text{cos}{\theta }_{\mathit{\text{plastic}}}\right)-\frac{h}{d}\text{cos}{\theta }_{\mathit{\text{gel}}}$, representing the “collapsing requirement” (see Supporting Information Section S2). In our microscale combing process, we estimate that the Capture number is on the order of 10⁻². A very small Capture number (<<1) indicates that the local collapsing of menisci is the faster process. It thus guarantees the capture of the microswimmer in the vicinity of the meniscus, as shown in Figure 2C. Furthermore, under our experimental conditions, because the Capture number is in fact two orders of magnitude smaller than one, the successful combing does not require external equipment to precisely control the liquid film sliding speed and is robust against small perturbations due to different user operations. Other microswimmers with a higher swimming speed can also be isolated using this method by adjusting the liquid film sliding operation.

2.2. Robustness and simplicity of ‘microswimmer combing’

The two mechanistic principles described above enable isolation of microswimmers on the microgel array. The method is robust and does not rely on user expertise; the results are largely guaranteed a priori by design. We next characterized the sample loading performance with different device designs by varying the microgel pad geometry and the array pattern (Figure 3A). To standardize the collapsing criterion, the inter-pad spacing “d” and the thickness of the animal suspension film “h” were held the same across all designs, with d = 1 mm and h = 330 μm. The individual size of the microswimmers used in the experiments was controlled by using synchronized C. elegans larvae (L4s, ~ 800 μm in body length and 50 μm in diameter). Based on previous studies,^[22] the excluded volume of individual worms in a worm suspension can be mainly attributed to steric hindrance. We thus tuned the size of each microgel pad to be comparable to the projected 2D area of an individual’s excluded volume in the suspension. Such a design is aimed to achieve single animal isolation on most microgel pads that capture microswimmers.

Figure 3.

Robustness and simplicity of ‘microswimmer combing’. (A) The schematics of four different device designs. The inter-pad spacing “d” (1 mm) and the thickness of the microswimmer suspension film “h” (330 μm) were held the same across all designs. The model microswimmers used in this characterization were synchronized C. elegans larvae (L4s). The suspension concentration was standardized to ~10 worms per microliter for all experiments except in (C). (B) The sample loading is characterized by the overall device occupancy (left) and single-animal isolation rate (right). The ‘microswimmer combing’ mechanism is effective to isolate individual animals across all designs. (C) Single-animal isolation rates are compared between combing with 10 worms / microliter suspension and combing with 20 worms / microliter suspension. Blue / red bars show the results using the large / small microgel pads. (D) Single-animal isolation rates are compared between combing on the 2D microgel pad array and combing on the 165 μm-deep microwell array. The single-animal isolation rate is much higher on microgel pad arrays. Because animals are pinned by the collapsing menisci on microgel pads compared with loaded by gravity into microwells. Blue / red bars show the results using the large / small microgel pads or microwells. (E) Both the overall device occupancy and the single-animal isolation rate are comparable between loading by novice users and by the expert user. N=3 for each condition shown in (B) to (D) and the “Expert average” in (E).

We quantified sample loading by the overall device occupancy and the single-animal isolation rate among all loaded microgel pads. To do so, we standardized the concentration of the worm suspension to ~10 worms per microliter. Approximately 30–40 percent of the microgel pads can consistently capture worms from the bulk suspension with 60–80% single-animal isolation rate across all four designs (Figure 3B). The substrate design with small microgel pads and staggered array pattern shows a slightly higher single-animal isolation rate. In a simple 30-second combing operation, this provides ~100 worms successfully isolated on chip, which is sufficient for most screening applications. The individual isolation number per chip can also be easily scaled up by increasing the array size to meet application needs. Both the device occupancy and the single animal isolation rates are robust across different microgel pad geometries and array patterns. This robustness makes it convenient to tailor the substrate design for different screening applications, for example, smaller microgel pads in a dense array for high-throughput and high-resolution imaging, or larger microgel pads in a sparse array for behavioral screening assays.

The single-animal isolation by the combing method is also not overly sensitive to different initial concentrations of the worm suspension (Figure 3C). Although a lower suspension concentration results in higher single-animal isolation rates in general, the effect is a small one. This is likely because microswimmers are concentrated at the receding edge of the suspension liquid film during combing, as shown in Figure 2C. The local worm concentration above the microgel pads where capturing events happen is much higher regardless of the initial worm concentration in bulk suspension. Therefore, we expect that changes in initial suspension concentration do not have a strong effect on the sample loading performance, even if the initial concentration is outside the range used in our experiments (10–20 worms per microliter). Such robustness further simplifies the sample preparation for screening applications. For behavioral applications, the width of a microgel pad being 5–10 times larger than that of a microswimmer is a key advantage. This allows for free movement of the isolated animal, which is especially important for behavior screenings.

We hypothesize that the flattened 2D substrate is the key factor that facilitates single-animal isolation despite this dimension mismatch. To test this hypothesis, we compared the sample loading by combing on the Kapton substrate with structured microwells (165 μm in depth) before and after fabricating microgel pads inside the wells. The single-animal isolation rate is significantly lower on the empty microwell array (~20–30% single-animal isolation rate) regardless of the geometry of the microwell (Figure 3D). We reasoned that by using microgel pads to flatten the surface, we only capture worms in the vicinity of the collapsing menisci, while preventing extra worms from being trapped by gravity. In comparison, microwells trap animals by sedimentation and cannot exert much control in the process.

To demonstrate the simplicity and usability of our method, we asked six novice volunteers to perform the ‘microswimmer combing’ for three trials and compared their sample loading performance with the “expert average” done by the lead author. After a single self-guided training trial, both the device occupancy and the single-animal isolate rate performed by the non-expert volunteers were comparable to, and in some instances better than, the expert average (Figure 3E). This result strongly demonstrates that the “combing” operation is mainly determined by the designed mechanisms rather than the user expertise. We thus believe that our method is highly transferrable to biology research laboratories without engineering expertise in microfluidics and instrumentation.

2.3. High-resolution imaging in multicellular model organisms

Our combing method can benefit several widely used whole-animal screening applications because it can separate microswimmers on an open-access surface where each animal experiences an individualized microenvironment. For instance, we first demonstrate for C. elegans the capability of our platform to perform high-resolution imaging. C. elegans is well known for its stereotypical cell lineage, which can be visualized by transgenic labeling of specific cells with fluorescent proteins. Image-based screening of cellular and sub-cellular features in C. elegans is broadly applied for studies such as cell development, neuronal signaling and gene expression.^[2]

To image a live C. elegans with cellular resolution, one needs to immobilize the moving animal rapidly. Existing high-throughput microfluidic techniques use either physical entrapment,^{[5, 23]} on-chip cooling,^[6] or CO₂ exposure,^[24] for animal immobilization. Most of these techniques require connection of the microfluidic platform to off-chip pneumatic controllers, cooling systems, or CO₂ gas supplies. These external components may practically limit the use of microfluidic devices on standard microscopes. Our device, in contrast, takes advantage of the microgel environment to immobilize microswimmers. Before loading the worms, we pre-soak all microgel pads in a buffer solution containing tetramisole, an acetylcholine agonist that acts as chemical anesthetic (Figure S2).^[25] After worms are captured on the microgel pads, tetramisole (which has a small molecular weight) quickly diffuses out from the hydrogel substrate and acts on the animal. The array of fully immobilized C. elegans can then easily be transferred and imaged on common high-resolution microscopes, such as confocal microscopes.

We tested the imaging of cellular features in C. elegans by imaging animals carrying a transgene expressing green fluorescent protein (GFP) in the seam cells (top panel of Figure 4A). The individual seam cells were clearly visualized. Interestingly, since the microgel pad is larger than the worm and does not physically constrain it (inset in top panel of Figure 4A), the animal could freely move and orient itself into its preferred lateral body orientation before being fully immobilized. We quantified the orientation distribution of the dorso-ventral body axis by comparing the relative location between the seam cells and the animal’s body outline (bottom panel of Figure 4A and Figure S3). More than 90 percent of the animals on the microgel pads oriented themselves within 35 degrees of the lateral (“L-R”) body orientation, which is sufficient to align different animals for cross-inspection of morphological features between them. Our microgel array approaches the orientation control accuracy of a previous microfluidic device that uses flow in a curved channel to orient C. elegans.^[26] Furthermore, our method is more convenient and is gentler to animals.

Figure 4.

High-resolution imaging in multicellular model organisms. (A) Top panel: the composite bright-field and fluorescent image of a transgenic C. elegans carrying transgene wIs51 with seam cells expressing GFP. Inset shows the same animal immobilized on the microgel pad. Scale bar: 50 μm. Bottom panel: the orientation distribution of the dorso-ventral body axis of isolated C. elegans on microgel pads. The inset shows the definition of the lateral (“L-R”) body orientation by comparing the relative location between the GFP labeled seam cells and the animal’s body outline. N = 44 in this experiment. (B) 60X fluorescent image of a GRASP punctum between IL1 and IL2 neurons, indicating a potential synapse connection between these two neurons. The IL1 and IL2 neurons are labeled with red fluorescent markers. Scale bar: 10 μm. The bright-field image of the same animal is shown in the top right inset.

Many neurobiology studies require investigation of not only cells, but also the organization of subcellular components, such as synapses. We next demonstrate the compatibility of our system with high-resolution imaging. We used a strain expressing GRASP (GFP Reconstitution Across Synaptic Partners) in the IL1 and IL2 neurons to visualize cell-specific synapses.^[27] The strain also carries a cytoplasmic cell-specific red fluorescent protein marker in both of these two neurons. The images were acquired using a spinning-disk confocal microscope under 60X oil objective. We can clearly distinguish sub-cellular features such as synapses, as indicated by the GFP puncta, as well as the dendrites of the neurons (Figure 4B), with image quality comparable to that from conventional agarose pads. Hence, our platform is compatible with most screening applications that rely on collecting high-resolution fluorescent images.

2.4. On-demand sample selection and retrieval

Next, we demonstrate the ease of on-demand live animal selection on our open-accessible device. The microgel array enables a simple workflow to selectively recover animals after they have been isolated and characterized (Figure 5A). To keep animals and microgel pads hydrated during the screening and selection process, we covered the array surface with a thin layer of bio-compatible oil (Figure 5A(i)). Both halocarbon oil for microinjection and silicone oil for droplet microfluidics can be used in this workflow. Highlighted by the orange arrow, the animal of interest is the worm on the center microgel pad showing fluorescence in the pharynx and body wall muscle cells. To recover this animal, we dispensed a micro-droplet of S-buffer solution on top and let it displace the oil and sink to the device surface (Figure 5A(ii)). Due to the differential wettability of the device surface, the micro-droplet self-aligned and anchored on the hydrophilic microgel pad (Figure 5A(iii)). After the animal of interest swam into the micro-droplet (Figure 5A(iv)), we simply recovered the animal by pipetting away the micro-droplet containing the animal (Figure 5A(v)–(vi)). Using this method, users do not need to build customized sorting and control systems, nor do they need to have precision and dexterity of handling the swimming organisms, such as in using a wire-pick. During both the ‘microswimmer combing’ and on-demand selection process, the animal is handled only in liquid droplets and not directly touched, for example, by a wire-pick; this process is gentle and further eliminates any potential physical damage of the animal due to mishandling.

Figure 5.

On-demand sample selection and retrieval. (A) The workflow to selectively recover individual animals based on their phenotype. The micrograph in each step shows the action corresponding to the schematic. The transgenic C. elegans carrying fluorescence marker in its pharynx and body wall muscle cells can be recovered using only a standard pipette, while animals on the neighboring microgel pads are undisturbed. (B) Proof-of-concept mutant selection. The rare mutant labeled with GFP markers can be isolated from a mixed population in which the majority is labeled with whole-body RFP markers, regardless the mixing ratio. The isolated GFP labeled mutant can be recovered by the workflow shown in (A). The mixing ratio between population densities of two mutants is 20:1 (RFP labeled animal : GFP labeled animal) in the top panel, and 10:1 in the bottom panel. Scale bar: 1mm.

We next performed a proof-of-concept mutant selection experiment (Figure 5B). We mixed two different fluorescently labelled transgenic strains of C. elegans in different ratios (20:1 and 10:1). By combing the mixed population, we could isolate the rare GFP expressing animals in the array regardless the mixing ratio and recover them from the abundant mutant strain labeled with whole-body red fluorescent reporters. Because isolated animals are associated with the on-chip location of the microgel pad, the selection criteria in an actual screening can be determined even after the entire chip is imaged. Animals can also be sorted into multiple cohorts based on various phenotypes, such as readouts from animal’s behavior or the intensity from the same kind of fluorescent markers, with our device and this selection workflow.

2.5. Multiplexed chemical screening with individual behavior analysis

The open-surface accessibility of our platform not only enables us to select samples as needed, it also allows us to individually address and precisely modify the chemical microenvironment of each microgel pad. Since the microgel pad can be much larger than the isolated microswimmer, it is straightforward to keep the animal freely moving and to analyze its behavioral dynamics. It has been shown that isolating C. elegans in an array of microdroplets and monitoring their behavior are critical to study various biological phenomena, such as quiescent behavior and adverse olfactory learning. ^[28] However, current methods require labor-intensive manual picking to place individual C. elegans in microdroplets, which greatly limits the experimental throughput. In contrast, the ease of isolating tens to hundreds of animals on-chip using our ‘combing’ method provides a sufficient sample size to understand the heterogeneity in animal behavior. To demonstrate the simplicity of screening individual behavioral responses to multiplexed chemical environments with our device, we performed a paralysis assay of C. elegans at different concentrations of tetramisole. To achieve multi-concentration chemical environments, we modified the microgel pads in different regions on a single chip by incubating them for 10 minutes in three separated drops of solution (akin to the schematic in Figure 6A). The detailed multiplexed chemical modification method can be found in Experimental Section/Methods. We used S-basal solution as control, 1mM tetramisole in S-basal solution as a low concentration condition, and 10mM tetramisole in S-basal solution as a high concentration condition. Animals were then loaded onto these modified microgel pads and the paralysis dynamic determined by monitoring the activity of isolated worms in 10-second intervals over the span of 200 seconds (Figure 6B). All animals in the control group moved normally throughout the duration of the experiment without being paralyzed (solid black dots and hollow black stars in Figure 6B), demonstrating the viability of animals after loading.

Figure 6.

Multiplexed chemical screening with individual behavior analysis. (A) The chemical microenvironment of microgel pads in different regions can be modified by simultaneously incubated in different solutions. The bottom picture shows the multi-zone microgel array device used for experiments in (B) to (D). Microgel pads dyed in green representing the S-buffer control, in blue representing the low-concentration environment (1mM tetramisole), and in red representing the high-concentration environment (10mM tetramisole) of results shown in (B) and (D). (B) The population average of the paralysis dynamic in different concentrations of tetramisole. Two strains of C. elegans were used: N2 the wild-type strain and CB211 the tetramisole-resistant strain. The sample loading by ‘microswimmer combing’ was less than 10 seconds. N = 42 (N2 10mM), 48 (N2 1mM), 57 (N2 Buffer), 43 (CB211 10mM), 50 (CB211 1mM), 60 (CB211 Buffer). (C) The individual posture dynamic of a wild-type worm (N2) isolated in 10 mM tetramisole microenvironment. (D) The individual activity traces of wild-type animals (N2) upon exposure to either high or low concentration tetramisole microenvironments. The yellow color indicates the animal moved actively during the 10-second observation window, and the blue color indicates the animal showed no movement during the 10-second observation window.

We next characterized the dynamic drug responses of the N2 wild-type and the tmr-1 mutant. The mutation in tmr-1 gene affects acetylcholine receptor activity and subsequently results in resistance to tetramisole.^[25] Not surprisingly, almost all wild-type worms were completely paralyzed after 60-second exposure to the high-concentration tetramisole environment, while only about half of the wild-type population showed a lack of activity even after 190-second exposure to the low-concentration environment. In contrast, the tmr-1 mutant population displayed a reduced paralysis percentage and a slower paralysis rate in both high and low concentration environments. This delayed paralysis dynamic of tmr-1 animals is consistent with the previous literature,^[25] which validates the efficacy of our platform. In addition, the longitudinal measurements of separated individuals on our device (Figure 6C) can further elucidate the heterogeneous behavioral responses to different environments, which may be obscured by averaging the population data. We analyzed individual activity traces of 90 wild-type animals upon exposure to different tetramisole environments (Figure 6D). The individual activity in the low-concentration environment suggests that animals not only start to react to tetramisole at a later time, and exhibit a more gradual reduction in locomotion, but also are more heterogeneous in responses, compared with animals exposed to high-concentration environment. Therefore, by coupling the small animal isolation via ‘microswimmer combing’ and the independent modification of hydrogel micro-pads, we can achieve parallel multiplexed chemical screening with single-organism resolution in one chip, without complicated external controls.

3. Conclusion

In this work, we present ‘microswimmer combing’, an ultra-simple approach to isolate active microswimmers via a dynamic contact line combing mechanism, on a novel open-surface microgel array, which is amenable to multifunctional screening applications. The ‘microswimmer combing’ approach possesses several unique advantages. First, the method can be easily adopted by non-engineering scientific communities. In contrast with most mainstream microfluidic techniques, this method does not require any sophisticated controllers or user expertise to operate. The isolation process is mechanistically determined, resulting in robust performance even when used by novices such as undergraduate students with no prior training. Second, the fabrication process of the ‘microswimmer combing’ platform is straightforward, which only needs commonly available materials and low-budget equipment. Although made using polyimide-based Kapton tape and PEG-based hydrogel in our demonstration, the device can be designed for any plastic and hydrogel materials that have proper surface wettability. This expands the repertoire of applications and enables the potential for rapid adaptation and dissemination. Third, unlike closed-channel systems, the open-surface nature of our microgel array allows isolated active animals to be not only visually assessed, but also independently perturbed and manipulated with ease.

Due to its versatility, our approach may benefit a wide range of screening applications using model microswimmers. As proof-of-principle examples, we demonstrated three image-based screening functions with different purposes. First, we showed high-resolution imaging of cellular and sub-cellular features, which is a common mode of image-based screening. Second, we developed an equipment-free on-demand mutant retrieval workflow which could enable customized phenotype-based selection of animals in large-scale forward genetic screenings. Third, we performed multiplexed chemical screening through analysis of individual animal behavioral responses. This provides a straightforward approach to conduct multiplexed chemical screening assays with single-animal resolution, which is instrumental for various applications from elucidating neuronal mechanisms to evaluating drug efficacy. The parallel array format of our device is similar to the multiwell plate format used in most regular biological assays, and thus, can be incorporated into standard scientific workflows with minimal modifications. For example, coupled with a robotic liquid handler, our device may be directly applied to high-throughput genetic screening, directed evolution experiments, and large-scale drug screening.^{[4, 29]} Although we demonstrated the utility of this method using the nematode C. elegans as an example, we envision that our approach, with a proper scaling of the device, can be adapted to study other microswimmer model systems, such as D. rerio and Ciona larvae and motile bacteria. To apply this ‘combing’ method to studies in different fields, future users need to pay attention to the chemical composition of the microswimmer suspension to ensure that certain macromolecules in the solution, such as surfactant-like molecules, do not drastically alter the wettability of the plastic surface, which is necessary for the isolation of microswimmers.

The current limitation of our proof-of-concept device is that a fraction of microgel pads are not available to capture microswimmers due to non-uniform fabrication of the plastic Kapton surface by laser cutting, which lowers the device occupancy. However, because of the robust combing mechanism, the number of animals isolated on a regular microscope-glass-slide-size chip (~100) is sufficient for most screening applications. By using other industrially available materials and processes that meet the surface wetting criteria, the device can be further scaled up into a larger microgel array to isolate more samples, or to provide a higher number of multiplexed experimental conditions, should they be required for large-scale applications. Therefore, we envision this approach to offer a user-friendly technique for high-throughput and precision experimentation for general biology researches, and makes studying model animals accessible even in k-12 laboratories.

4. Experimental Section/Methods

Fabrication of the microgel array:

Kapton polyimide tape with silicone adhesive that is 165 μm-thick (KaptonTape.com, KPT5–3) was structured through laser cutting (Epilog Legend 36EXT) to create a microwell array pattern as the continuous plastic surface of the microgel array. To structure the microwell array pattern on one device shown in Figure 1A (319 microwells within a 2” * 1” area), laser cutting took about two minutes. The structured Kapton tape was cleaned by sonication in ethanol bath for five minutes to remove debris and then transferred onto a microscope glass slide (Fisher Scientific, 12-544-1). The PEG-based microgels consisted of 5% 4-arm poly(ethylene glycol) thiol (Mw 20000, Laysan Bio. Inc) and 5% 4-arm poly(ethylene glycol) norbornene (Mw 10000, Sigma-Aldrich, 808474). The photoinitiator used was 1% lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP, Sigma-Aldrich, 900889). The gel prepolymer solution was prepared in sterile S-basal medium. To fabricate the microgel pad, we dispensed 100 μL gel prepolymer solution droplet on the structured Kapton tape microwell array and brushed the solution droplet by a pipette tip across the array to fill microwells through discontinuous dewetting ^[20] (Figure S1 and Movie S1). These PEG-based microgels were cured immediately by exposure to 365nm light for 10 seconds to form microgel pads. The entire microgel array device can be fabricated within ten minutes. The completed microgel array was stored in sterile S-basal medium for 12 hours before each experiment, which allows microgel pads to properly swell to reach the same thickness of the surrounding Kapton tape. To prepare the top glass slide used in ‘microswimmer combing’, we stacked multiple layers of 165 μm-thick Kapton tape on a microscope glass slide as the spacer to define the thickness of the suspension liquid film (“h”). The Spacer containing two layers of Kapton tape was used in all experiments, while spacers containing one, two, four and eight layers of Kapton tape were used in the experiment shown in Figure 2D.

C. elegans strains and maintenance:

Strains used were obtained from the Caenorhabditis Genetics Center or a gift from the laboratories that created the strains. The strains used in visualization of animal orientation and for demonstrating image quality include JR667 (wIs51[SCMp::GFP]; unc-119(e2948)), OH15089 (otIs657[klp-6p::mCherry + flp-3p::mCherry + klp-6p::NLG1::spGFP1–10 + flp-3p::NLG1::spGFP11]). For the on-demand sample selection and retrieval experiments, we used the strain carrying transgene [eft-3p::tdTomato::H2B::unc-54 3’UTR] as the red fluorescence labelled worms (gift of P. McGrath), and the strain carrying transgene [myo-2p::gfp] as the green fluorescence labelled worms (gift of G. Benian), respectively. We used N2 and CB211 (lev-1(e211)) in the multiplexed chemical screening of behavioral responses experiments. Animals were cultured following standard protocols on NGM agar plates (60 mm in diameter) with Escherichia coli (E. coli) OP50 lawns and maintained at 20 °C. We synchronized the animals by leaving 100 hermaphrodites per plate, laying eggs for 2 hours before removing them. Age-synchronized L4 larval stage worms were washed and suspended in S-basal medium before each experiment.

Image acquisition and sample preparation for imaging:

We used a whole-field illumination fluorescence dissecting scope (Leica, MZ16F) equipped with a CMOS camera (ThorLabs, DCC1545M) to collect videos and images for characterization of ‘microswimmer combing’ (Figure 2 and Figure 3), on-demand sample selection (Figure 5), and multiplexed chemical screening experiments (Figure 6). Images and videos were taken immediately after sample loading. For high-resolution imaging experiments (Figure 4), we used a spinning disk confocal microscope (PerkinElmer UltraVIEW VoX) equipped with a Hamamatsu FLASH 4 sCMOS camera. 20mM tetramisole hydrochloride (Sigma-Aldrich) in S-basal medium was used to immobilize the active worms following a protocol shown in Figure S2. To prevent the microgel pads and the microswimmers from dehydration during imaging, we covered the microgel array after sample loading with a bio-compatible silicone oil (1000 cSt, Clearco Products Co). After oil coverage, the microgel array can be imaged for more than 2 hours without significant evaporation.

Chemical modification of the microgel array for multiplexed chemical screening:

To perform the multiplexed chemical screening assay, we fabricated a microgel array, which consists of three smaller sub-arrays on a single chip. These three sub-arrays are made on one piece of Kapton tape and separated by 5 mm from the adjacent ones (Figure 6A). To achieve parallel and multiplexed modification of the chemical environment in each microgel pad, we first placed two polydimethylsiloxane (PDMS) blocks between adjacent sub-arrays fixed by double-sided tapes on the Kapton surface, and then pipetted 100 μL liquid drops containing drugs with different concentrations onto different sub-arrays to incubate microgel pads. The solutions used in the paralysis assay (Figure 6) were S-basal buffer as the control, 1mM tetramisole in S-basal solution as a low drug concentration condition, and 10mM tetramisole in S-basal solution as a high drug concentration condition. Figure S4 illustrates the multiplexed chemical modification and incubation set-up: Drop A (dark green) represents the S-basal buffer control; Drop B (dark blue) represents the 1mM tetramisole condition; Drop C (red) represents the 10mM tetramisole condition. The PDMS blocks separate different liquid drops to prevent them from mixing. The liquid drops were left on the microgel pads for 10 minutes to allow the drug to diffuse into microgel pads and to reach equilibrium. The incubation set-up was kept in a humid environment to prevent evaporation of liquid drops. After the 10-minute incubation, we tilted the chip to its side and let the liquid drops flow out of the microgel array region. A wipe was placed underneath the chip to absorb excess liquid. Because the Kapton surface is not hydrophilic, no excess liquid was left on the continuous plastic domain. We then removed PDMS blocks from the chip. The final device after modification is shown in Figure 6A. Microgel pads in different sub-arrays contain different concentrations of drug, as illustrated by different dyes. Animals were loaded by ‘microswimmer combing’ from the long side of the chip immediately after the chemical modification to start the paralysis assay. The worm loading was completed within 5 seconds. Therefore, we expect that the chemical environment of each microgel pad is not disturbed by the combing process.

Statistical Analysis:

All quantitative data were normalized to the respective control and presented as mean ± standard deviation. For each experimental condition shown in Figure 3, at least three independent experiments were conducted and used for statistical analysis. All data were analyzed and plotted using OriginPro (OriginLab Corporation).