Sage Journals: Discover world-class research

Abstract

The future of the life sciences is linked to automation and microfluidics. As robots start working side by side with scientists, robotic automation of microfluidics in general, and droplet microfluidics in particular, will significantly extend and accelerate the life sciences. Here, we demonstrate the automation of droplet microfluidics using an inexpensive liquid-handling robot to produce human scaffold-free cell spheroids at high throughput. We use pipette actuation and interface the pipetting tip with a droplet-generating microfluidic device. In this device, we produce highly monodisperse droplets with a diameter coefficient of variation (CV) lower than 2%. By encapsulating cells in these droplets, we produce cell spheroids in droplets and recover them to standard labware containers at a throughput of 85,000 spheroids per microfluidic circuit per hour. The viability of the cells in spheroids remains high throughout the process and decreases by >10% (depending on the cell line used) after a 16 h incubation period in nanoliter droplets and automated recovery. Scaffold-free cell spheroids and 3D tissue constructs recapitulate many aspects of functional human tissue more accurately than 2D or single-cell cultures, but assembly methods for spheroids (e.g., hanging drop microplates) have limited throughput. The increased throughput and decreased cost of our method enable spheroid production at the scale needed for lead discovery drug screening, and approach the cost at which these microtissues could be used as building blocks for organ-scale regenerative medicine.

Keywords

microfluidics microtechnology robotics and instrumentation engineering 3D microtissue droplet microfluidics high-throughput cell processing

Introduction

We believe that the future of the life sciences is linked to automation and microfluidics. With robots working side by side with scientists, robotic automation of microfluidics in general, and droplet microfluidics in particular, will significantly extend and accelerate the life sciences.

Throughout the years, microfluidics has developed a plethora of applications, particularly in the life sciences, and has been part of miniaturizing, increasing throughput, and lowering the cost of a large number of macroscale sample processing and analysis techniques, as well as making possible new processes relying on microscale phenomena. Microfluidic technologies have seen some general automation approaches, such as pressure-controlled valving¹ and programmatically reconfigurable devices.² Most of these, however, require substantial infrastructure outside of the microfluidic device,^3–7 and only very few (e.g., the Fluidigm integrated fluidic circuits) can be readily interfaced with general laboratory robotics such as liquid-handling robots.

One major microfluidic technology platform is droplet microfluidics. It is a high-throughput sample processing and analysis method enabled by miniaturizing the reaction vessels to nano- or picoliter droplets.⁸ Droplet microfluidics relies on structured compartmentalization of immiscible liquids (e.g., water and fluorinated oil in the form of a water-in-oil emulsion stabilized by a surfactant).^9,10 These tiny cell-scale compartments produced with high size accuracy allow for millions of individual biological events¹¹ or chemical reactions to take place simultaneously,¹² to a large extent chemically and physically separated.¹³ A set of microfluidic devices is used in droplet microfluidics to form, merge, inject into, analyze, and sort millions of droplets quickly and efficiently.⁸ Droplet microfluidics is used in multiple applications in biology, including metabolic pathway interaction studies,¹⁴ single-cell genomics^15,16 and transcriptomics,¹⁷ digital droplet PCR,¹⁸ directed microbial evolution,¹⁹ as well as cell spheroid formation or cell aggregation in confined environments^20,21 and cell encapsulation in gels.²² Immense droplet production rates as high as 1 trillion(!) per hour have been achieved by parallelized droplet-generating devices containing up to 10.000 droplet generation nozzles.^23,24

Automation of microfluidics in general and droplet microfluidics in particular is an obvious next step for technology development in this field. It is even more timely now, as more and more commercial applications are being developed based on droplet microfluidics. Automating even simple steps that are performed on a daily basis in droplet microfluidic workflows—such as droplet generation, as well as sample recovery/droplet breaking—is not a trivial task. For the automation to truly succeed, it has to rely on widely accessible, standardized solutions that are as simple as possible.

Liquid handling is an essential part of many experiments in the life sciences; therefore, large numbers of standardized solutions for liquid handling have been developed. These include the pipette and robotic pipetting workstations [also known as liquid-handling robots (LHRs)]. Robots can work without fatigue, increase throughput, perform consistently, and ensure accuracy and precision.²⁵ The pipetting robot has not previously been used as the sole automation tool to perform all the tasks needed for successful droplet generation or for droplet breaking and living content recovery. In this article, we present a fully automated workflow powered only by an LHR paired with a droplet microfluidic component. To demonstrate the approach, we use the LHR to automate the production of three-dimensional (3D) cell culture models, cell spheroids.

Uses for cell spheroids and other 3D tissue models are emerging areas of interest both in fundamental research and in industrial applications, because they provide more predictive in vitro models to study fundamental cell biology, disease pathophysiology, and the development of novel therapeutic agents.^26,27 Spheroids are also considered a central component of regenerative medicine, being one of the key elements of the rapidly growing precision regenerative medicine field.^28,29 As the demand for 3D cell culture models grows in the academic, medical, and industrial environments, there is an increasing need for affordable automation of spheroid production, because this process could eventually play a pivotal role in lowering the total expenses of spheroid fabrication and thus facilitating their use at a large scale. The increased throughput and decreased cost of our method enable spheroid production at the scale needed for lead discovery drug screening, and approach the price at which these microtissues could be used as building blocks for organ-scale regenerative medicine.

Materials and Methods

Droplet-Generating Microfluidic Chips

A droplet microfluidics circuit for production of nanoliter droplets was designed in AutoCAD ( Suppl. Fig. 1 ). The flow-focusing droplet-generating circuits were fabricated from a 2 mm thick poly(methyl methacrylate) (PMMA) plastic sheet using computer numerical control (CNC) micromilling (MDX-40A 3D Milling Machine; Roland, Irvine, CA) with a 200 µm milling tool. All the microchannels, including the nozzle, were designed to be 200 µm wide and 100 µm deep. The inlets and the outlet were drilled with a 1 mm drill by a CNC micromilling machine. Then, the chip was cleaned with a detergent using a cleaning brush, and carefully washed with water and Milli-Q water to remove any plastic residues from all microchannels. Subsequently, the chip was dried with compressed air, and it was sealed with a pressure-sensitive adhesive tape (PSA, ARcare 90445Q; Adhesives Research, Glen Rock, PA) by peeling off one side of the tape and sticking it to the side of the PMMA chip that contained the open microchannels. Then, a custom-made 3D-printed reservoir element was carefully aligned and attached to the other side of the microfluidic droplet-generating chip, also using an adhesive tape (top of the droplet-generating chip; tesa, Hamburg, Germany). The tape was precut—three holes were cut out using a 2 mm Harris Uni-Core biopsy puncher. The holes in the tape were aligned to meet the inlets and outlets of the droplet-generating chip, and also the corresponding outlets and inlets of the sample and oil reservoir of the 3D-printed part.

Finally, the surface of all microchannels (made from PMMA and PSA tape) was treated with Aquapel. The surface-modifying chemical was removed from the channels by compressed nitrogen after 2 min. That step was followed by passing the hydrofluoroether (HFE) oil (Novec HFE-7500 Engineered Fluid; 3M, Two Harbors, MN) through all microchannels to wash out any remaining Aquapel. Finally, all the liquids were removed from the chip by compressed nitrogen.

To prevent sedimentation of the cells in the dispersed-phase/sample reservoir and clogging of the microfluidic microchannel, and to keep as much as possible equal numbers of cells entrapped in individual droplets, the whole droplet-generating chip was placed on a magnetic stirrer (IKA big squid; IKA, Staufen, Germany), and a sterile small magnet was placed in the dispersed-phase/cell-sample-containing reservoir. The magnetic stirrer was set at 150 rpm.

For collecting droplets, a modified 1 ml pipette tip was used (BRAND, 50–1000 µL; BrandTech, Essex, CT). A simple 4 mm latex sleeve [natural rubber latex tubing with outer diameter (OD) 2.3 mm and inner diameter (ID) 0.8 mm; Kent Elastomer Products, Inc., Kent, OH] was secured at the thinner end of the pipette tip. This rubber sleeve plays an important role, because it seals the connection between the output pipette tip being filled with generated droplets and the outlet of the droplet-generating chip.

Cell Sample

HEK293 cells were cultured in FreeStyle293 culture medium in a nonadhesive cell culture flask (125 mL Erlenmeyer flask, nonpyrogenic, polycarbonate; Corning, Corning, NY). RT4 and A-431 cells were cultured in McCoy’s 5A (Modified) Medium and RPMI culture medium, respectively, in a cell culture incubator (Heracell 150, 37 °C, 5% CO2; Thermo Scientific, Waltham, MA). Before encapsulating in droplets, the cell suspension was prepared to meet the desired concentration of cells (i.e., HEK cells: 3.88×106, 4.7×106, and 7×106 cells/mL; RT4 cells: 4.52×106 cells/mL; and A-431 cells: 7×106 cells/mL).

Droplet Generation Process

As a continuous phase, an HFE (Novec HFE-7500 Engineered Fluid) with 1% polyethylene glycol–perfluoropolyether (PEG-PFPE) amphiphilic block copolymer surfactant (RainDance Technologies, Billerica, MA) was used. One milliliter of HFE with 1% of the surfactant was added to the continuous-phase reservoir with the aid of a pipetting robot (Opentrons OT-One-S Hood; Opentrons, Brooklyn, NY). This primes the microchannels, limiting to the minimum the amount of air trapped inside the network of microfluidic channels. The initial step was followed by filling up the dispersed-phase/sample reservoir with 500 µL of the cell suspension. Then, a negative pressure was applied to the upper end of the droplet-collecting element (a modified 1 ml pipette tip) by a computer-controlled pipette with which the LHR is equipped (DragonLAB TopPette 100–1000 µL; DragonLAB, Split, Croatia). The pipette was programmed to transfer 1000 µL in 100 steps (10 µL each), generating an estimated 0.58 kPa of negative pressure in each step at a frequency of 0.5 Hz. Once the suction force applies, it removes all the remaining air trapped in the system and starts the negative pressure–driven, micropipette-based droplet generation process.

Droplet Content Recovery

To recover the cells from within the droplets (i.e., to separate the continuous phase from the dispersed phase), a polytetrafluoroethylene (PTFE) filter membrane (PTFE filter, 0.45 µm pore size; Sartorius Biolab Products, Göttingen, Germany) was used. About 400 µL of the emulsion (containing ca. 17,000 droplets) was placed using a pipetting robot on the surface of the PTFE filter membrane. Immediately, the HFE oil drained into the membrane, destabilizing and breaking the emulsion and allowing the droplets to merge, forming a large droplet full of spheroids floating freely in the culture medium. Alternatively, a chemical-based sample recovery method was used, in which 400 µL of droplets were merged using 30 µL of perfluorooctanol.^30,31 The recovered spheroids were subsequently analyzed for viability.

Cell Viability Test

HEK, RT4, and A-431 cell cultures

Before droplet generation, cell cultures were analyzed with a Bio-Rad TC20 Automated Cell Counter using Bio-Rad counting slides with dual chambers (cat. no. 145-0011; Bio-Rad, Hercules, CA). The cells were stained with Trypan Blue solution (cat. no. T8154-100ML; Sigma-Aldrich, St. Louis, MO), which selectively color dead cells blue. Live cells with intact cell membranes are not colored.

Ten microliters of the cell suspension were mixed with 10 µL of Trypan Blue, and finally 10 µL of the mixture was pipetted into one of the chambers of the counting slide. The automatic measurements were triplicated, and an average value of viable cells was calculated.

The viability of the used cells was 96% for HEK and A-431 cells, and 97% for RT4 cells.

Cell spheroids

To analyze the viability of the cells forming cell spheroids that were fabricated in droplets, a combination of Hoechst 3334 (cat. no. C47198; Life Technologies, Carlsbad, CA), propidium iodide (cat. no. C27858; Life Technologies), and Calcein AM green (cat. no. 56496-50UG; Sigma-Aldrich) was used. One microliter of each was added to 1000 µL of 1× phosphate-buffered saline (PBS). Then, 100 µL of the Hoechst 3334, propidium iodide, and Calcein AM mixture was added to a 50 µL solution containing the recovered spheroids (cell aggregates suspended in culture medium). This was followed by an incubation step of 30 min in the dark. Finally, fluorescent images of the stained spheroids were taken using fluorescent microscopy with a Nikon Eclipse Ti (Nikon, Tokyo, Japan) to measure the final viability of formed spheroids [in bright field, fluorescein isothiocyanate (FITC), tetramethylrhodamine (TRITC), and 4′,6-diamidino-2-phenylindole (DAPI)]. The number of dead cells was counted in a randomly selected image containing 40 spheroids. The viability of the cell spheroids (%) was calculated using the following formula: Number of dead cells / Estimated total number of cells forming a spheroid × 100.

Platform Setup

For the robot to operate smoothly, the device has to go through the initiation round. The operator has to make sure that the 1 mL pipette tip rack ( Fig. 1D_1 ) has clean tips loaded. The first row of the rack should have modified pipette tips (equipped with a special rubber O-ring). The heat block ( Fig. 1D_2 ) has to be switched on and set at 37 °C. Depending on the number of samples, one or several Eppendorf tubes filled with 0.5 mL of mineral oil have to be placed in the heat block’s well to receive droplets at a certain point of the protocol. A 96-well plate ( Fig. 1D_3 ) or an additional Eppendorf tube has to be placed on the deck or in an Eppendorf tube rack ( Fig. 1D_4 ), respectively—here, the formed and recovered spheroids will finally be deposited. Then, an Eppendorf tube with the HFE oil supplemented with 1% of surfactant should be located in the first (A1) position of the Eppendorf tube rack ( Fig. 1D_4 ), followed by another one with the cell sample in the neighboring (B1) position of the same tray ( Fig. 1D_4 ). Next, the magnetic stirrer ( Fig. 1D_6 ) should be centered under the microfluidic chip. Finally, a small PTFE-coated magnet should be placed into the cell sample reservoir, and the magnetic stirrer should be switched on.

Figure 1.

The automated workflow for generating cell spheroids using a robotized droplet microfluidic platform consists of five steps: (A,B) sample and reagent loading, (C) droplet generation, and (D) spheroid formation, cell incubation, and spheroid recovery. All steps are performed independently by the liquid-handling robot (LHR) without assistance from an operator. (E) The full process, which transforms a single-cell suspension into spheroids, takes between 5 and 24 h, with actual work performed by the robot lasting only 22 min (12 min droplet formation and 10 min sample recovery, once the spheroids were formed).

Robot Operations

Setup

The robot’s head, equipped with a 1 mL pipettor, grabs a regular 1 mL pipette tip from the tip rack ( Fig. 1D_1 ) by locking the pipettor’s lower end into the pipette tip—just like a human would do—and moves it toward the Eppendorf tube that contains HFE oil supplemented with the surfactant ( Fig. 1D_4 ). Then, it lowers itself into the vessel, aspirates 950 µL of the HFE oil, and transfers it into the continuous-phase (smaller) reservoir of the droplet-generating microfluidic chip ( Fig. 1D_6 ). The robot’s head then moves back to the initial position ( Fig. 1D_1 ) and returns the tip into the same well of the tip rack, so that it could be used later for the same purpose. Similarly, but with a new pipette tip collected from a neighboring position, the robot transfers 450 µL of the cell sample to a slightly bigger reservoir (dedicated to the dispersed phase) that contains at this point also a rotating magnet. Once the sample is pipetted out, the robot moves its head toward the waste box ( Fig. 1D_7 ) and drops contaminated tips there.

Droplet generation

From the A1 position of the 1 mL pipette tip rack ( Fig. 1D_1 ), the robot collects a modified pipette tip that has a seal (an O-ring) and moves its head above the outlet of the droplet-generating microfluidic chip. Before reaching its destination, it drops the plunger of the pipettor to the minimum, removing any remaining air from the pipette’s pumping mechanism. Then it lowers itself, reaching the level of the actual outlet of the droplet-generating microfluidic chip (the PMMA part; Fig. 1C and 2C). The connection between the pipettor and the microfluidic chip becomes truly sealed when the bottom of the O-ring reaches the level of the PMMA and compresses against it. At the same time, the conical shapes of both the pipette’s tip (covered with the rubber seal) and the 3D-printed well lock, enhancing the connection by making it air- and liquid-tight. Then, the controlling software activates generation of the negative pressure (which value was estimated at 0.58 kPa/step) in the microfluidic system by sequentially lifting the plunger and aspirating 10 µL with a frequency of 0.5 Hz, every 2 s. The robot repeats this step 100 times. It is essential to generate negative pressure gradually, for example by a sequential lift of the plunger, because this helps in maintaining a sealed connection between the rubber O-ring and the microfluidic chip. Otherwise, the airtight connection might fail as the vigorously sucked air bursts through the seal, preventing a constant buildup of the pressure difference between the pipettor and the 3D-printed reservoirs. If that happens, bubbling inside the pipette tip is observed and droplets are no longer generated, because not strong enough negative pressure is transferred through the channels of the microfluidic chip.³² Both of the liquids, the cell suspension, and the HFE oil are passed through a network of separate microchannels until they reach the flow-focusing part of the microfluidic chip. There, the cell suspension (dispersed phase) is being “pinched away” by the HFE oil (continuous phase), forming the droplets ( Fig. 2F , Suppl. Fig. 2, and Suppl. Movie 1) with a similar number of cells inside them (Fig. 3A and 3B, and Suppl. Figs. 3 and 4).^33,34 A fluorosurfactant was used to stabilize the emulsion at a concentration of 1% (w/v) and was dissolved in the continuous phase before the experiment.

Figure 2.

(A) The droplet-generating chip is made from a 3D-printed element bonded to a micromilled poly(methyl methacrylate) (PMMA) sheet in which a network of 200 µm wide and 100 µm deep microchannels was engraved and sealed with pressure-sensitive adhesive tape. (B,C) A novel interface between a pipetting robot and a microfluidic chip was necessary to maintain an air- and liquid-tight connection. The interface consists of a rubber O-ring at the lower opening of the pipette tip, and a conically shaped outlet well in the 3D-printed part in which the modified pipette tip is locked during droplet generation. (D) The robotized droplet microfluidic platform produces highly monodisperse 290 µm droplets. (E) The measured size distribution of generated droplets [diameter coefficient of variation (CV)% is 1.7]. (F) The process of droplet generation actuated by the sequential movement of the computer-controlled pipette observed in the droplet-generating chip. Scale bar = 100 µm.

Figure 3.

Images of HEK cells in droplets: (A,B) (at 4× and 10× magnification) immediately after the droplets were formed and (D,E) (at 4× and 10× magnification) after 16 h of incubation in droplets. (C) Fluorescent image of the viability-stained spheroids. Main image: All channels merged; insets: (top) fluorescein isothiocyanate (FITC) corresponding to Calcein AM, (middle) 4′,6-diamidino-2-phenylindole (DAPI) corresponding to Hoechst, and (bottom) tetramethylrhodamine (TRITC) corresponding to propidium iodide. (F) The viability of cells before encapsulation and in an aggregated, spheroidal form.

After this step, the robot pauses for 300 s until a full 1 mL of liquid transfers through the network of the microchannels into a modified pipette tip. The pause in the workflow is needed for the pressure difference to equalize between the pipettor and the atmospheric pressure through the network of microchannels of the chip.33 The pause also limits sucking an excess amount of air into the droplet-containing pipette tip, which prevents the formation of a polydisperse emulsion.

Now, the tip attached to the pipettor doubles as a temporary droplet reservoir. Once all the droplets are collected, the robot lifts its pipettor and moves it above the heat block. The head is lowered, and the droplets are transferred into an Eppendorf tube, where the cells will be incubated for at least 16 h and will transform into spheroids. To prevent evaporation of droplets during the extended incubation, the emulsion is stored in Eppendorf tubes under a protective, immiscible, and nonvolatile layer formed by mineral oil (Suppl. Movie 2).

Spheroid recovery

Once the overnight/16 h lasting pause is finished, the robot activates itself and grabs a clean tip from the 1 mL pipette tip rack ( Fig. 1D_1 ). Then, depending on the method chosen for the droplet recovery, it travels to the Eppendorf tube rack ( Fig. 1D_4 ) and transfers 30 µL of the droplet-breaking chemical (perfluorooctanol) into an Eppendorf tube, or it moves directly to the heat block. At the heat block site, the robotic arm lowers a 1 mL pipettor into an Eppendorf tube in which the droplets were incubated overnight and collects its content (1000 µL), leaving—if present—the mineral oil behind. Then, the robot pauses for an additional 60 s, allowing the droplets to cream at the top of the HFE oil (now stored again in the pipette tip).

Finally, the robot lifts its head a few millimeters up and removes 600 µL of the HFE oil that remains in the lower part of the tip, saving only the packed emulsion. Then, depending on the selected recovery method, the robot moves the pipettor either to the Eppendorf tube rack ( Fig. 1D_4 ) or to a neighboring position that has a recovery membrane.

Results and Discussion

The robotically automated droplet microfluidic platform consists of four major components: a liquid-handling robot, a droplet-generating microfluidic chip, an interface connecting them, and a computer that controls every movement and action of the robot. The platform is built on a commercially available LHR. Opentrons OT-One Hood was chosen because of its open source, simple design, and relatively low price ($4000). The Opentrons OT-One Hood robot is essentially a standard 1 mL or eight-channel 300 µL pipettor attached to the robotic arm, which can move in all directions according to simple Python code. The LHR can be controlled either by a Python script or “manually” using a Python console and XYZ coordinates, as well as using Opentrons software—in all cases, an external computer with appropriate open-source software has to be connected to the robot.

Although the robot has 15 slots for the well plates on its deck, the design of the device allows accommodating only eight of them, meaning that only part of the surface of the deck (about 30×40 cm) can be used for fully automated operations. The rest of the surface is out of reach for both or just one of the pipettors attached to the robotic arm.

To fit the necessary equipment—a 1 mL tip rack, a waste box, a 12-slot Eppendorf tube rack, a 96-well plate, a full-size magnetic stirrer with a droplet-generating microfluidic chip on top of it, and a heat block for the Eppendorf tubes—we have used all of the eight available slots.

To generate relatively large quantities of highly monodisperse nanoliter droplets, a pressure source (negative^32,35 or positive^33,36) and a microfluidic chip based on specifically organized microchannels are needed.³⁷ To fulfill those requirements, we chose a simple flow-focusing design as the droplet-generating element, which we fabricated in a 2 mm thick PMMA plastic sheet using a CNC micromill.³⁸ The CNC-engraved PMMA sheet was sealed with pressure-sensitive adhesive tape, forming a closed network of microfluidic channels. All microchannels, including the nozzle, were 200 µm wide and 100 µm deep ( Fig. 2A ). To minimize the complexity of the platform, we decided to actuate the microfluidic device using the 1 mL pipette that the robot is equipped with.³² In this way, we eliminate the need for expensive syringe pumps or pressure controllers as the means of actuating liquids in droplet generation. This decision also greatly simplified the control protocol, because only a single device, the LHR equipped with a 1 mL pipette, needs to be programmed. To use a pipettor as a negative pressure source for droplet microfluidics,³² we devised a sealing interface that connects the pressure source, the pipettor and pipette tip, with the microfluidic chip without losses in suction force generated by the pipettor. For microfluidic chips fabricated from polydimethylsiloxane (PDMS), the polymer itself can act as a seal.³⁹ Unfortunately, PDMS chips lack reproducibility because they are usually manufactured in low quantities, and by hand—introducing human error. Reproducibility is key to successful lab work as well as automation.⁴

To seamlessly connect a 1 mL pipette tip with the outlet of the microfluidic channel, we added a 4 mm long rubber O-ring at the lower part of the pipette tip. The conical shape of the 1 mL pipetting tip prevents the rubber O-ring from sliding along the tip and locks it in place. This modification allows the robot to use the same modified pipetting tip multiple times in a series of workflows ( Fig. 2B ). To further increase the quality of the connection between the negative pressure source—the pipettor—and the outlet of the microfluidic chip, we have added a 3D-printed part with a 5 mm deep and 3 mm diameter outlet well and two inlet reservoirs on top of the plastic microfluidic chip. The two inlet reservoirs are intended for the continuous and dispersed phases—HFE oil supplemented with 1% surfactant and the cell sample, respectively ( Fig. 2A ). We have dedicated a smaller reservoir to the continuous phase, because its volume is 1.15 mL, supporting droplet generation in a 1 mL pipette. A slightly larger reservoir, with a capacity of 1.65 mL, also houses a small magnetic stirrer submerged in the cell sample to prevent the sedimentation of the cells during droplet formation ( Fig. 2B ).

The automated workflow for cell spheroid generation using the robotized droplet microfluidic platform consists of five steps: sample and reagent loading, droplet generation, cell incubation for spheroid formation, spheroid recovery, and spheroid dispensing. All steps are performed independently by the LHR without assistance from an operator. Throughout all steps, the LHR exactly follows the experimental protocol, which was uploaded as a Python script to the LHR-controlling computer. The LHR automatically transfers samples and generates, collects, and deposits droplets using the standard and modified pipette tips, respectively. Even though a regular 1 mL pipette operated by a robot was used as a negative pressure source, the measured size distribution of generated droplets [coefficient of variation (CV)%] was less than 2% ( Fig. 2E ).

Cell Incubation for Spheroid Formation

We have tested two methods for cell incubation. Cells in droplets were incubated either in an external incubator or on the heat block in an Eppendorf tube, where the HFE oil surrounding droplets acts as the gas reservoir for the spheroid-forming cells.^20,21 Both methods yielded viable spheroids. During incubation, encapsulated cells are stored at 37 °C. It takes several hours for the cells to form loose aggregates that subsequently convert into spheroids.^20,21 The spheroids are essentially tightly packed cell aggregates that do not disassemble when exposed to an external force.⁴⁰ In our case, a properly formed spheroid should survive the process of sample recovery as well as physical relocation performed by a pipetting robot. Because our method does not rely on any hydrogel that may maintain the cell aggregates together,^34,41-43 without sufficient physical interaction between cells, the aggregates disassemble as droplets are broken, and the physical micro-compartmentalization disappears. Since our platform currently does not allow for real-time observation of the biological processes that appear in droplets during the spheroid formation step, to explain what happens in the droplets, we have to rely on the hypothesis that was given by Baroud et al.³⁴ on how the cells sediment and aggregate into spheroids in droplets. It is worth noting that although our platform uses similar droplet size, it does not lock the droplets in wells, and thus it does not prevent the HFE oil from freely flowing around the droplets, allowing the flow to be transferred through the water–oil interface into the droplets. This additional internal flow could potentially change the behavior of the cells in droplets and at least delay their sedimentation. The minimum time needed for the cells to aggregate, such that they maintain a spheroidal form after the recovery, was about 5 h, which is consistent with previously published observations.^20,34 For convenience, we decided that the cells will be incubated overnight (16 h), giving the cells enough time to interact with each other so they could form a stable spheroid.

Chemical-Based Sample Recovery

In the case of the chemical-based sample recovery method, 400 µL of the packed emulsion is pipetted into an Eppendorf tube containing 30 µL of perfluorooctanol, a detergent known to destabilize emulsions; the emulsion is pipetted there directly following the 16 h incubation period ( Fig. 1D_4 ). Then, the LHR automatically exchanges the pipetting tip for a new one and initiates a 5 min pause to allow the perfluorooctanol to break down the emulsion entirely and for the droplets to merge into a single aqueous droplet that contains all of the spheroids. Once the emulsion breaking is finished, the robot aspirates 500 µL from the spheroid-containing Eppendorf tube (to be sure that the whole of the sample is collected). This step is followed by pipetting out 125 µL, which contains the air and all of the waste (the HFE oil mixed with perfluorooctanol). At this point, the pipettor’s tip holds all of the recovered spheroids suspended in culture medium.

Membrane-Based Sample Recovery

In the membrane-based sample recovery method, the robot transfers the remaining 400 µL of the packed emulsion onto a PTFE membrane located next to the Eppendorf tube rack ( Fig. 1D_8 ). The PTFE membrane allows the two phases to separate. Meanwhile, the robot changes its pipetting tip and initiates a 5 min pause required for all of the droplets to merge. After 5 min, the LHR collects all spheroids by aspirating directly from the PTFE membrane. Finally, the robot begins the sample relocation step. Here, it optionally dilutes the spheroid-containing sample with culture medium and, if programmed, distributes 75 µL segments of the sample into a preprogrammed set of wells in the 96-well plate ( Fig. 1D_3 ). This step finishes the automated workflow, which can now be repeated with a new set of samples.

Sample Recovery by Physical Phase Separation on a PTFE Membrane

The process of membrane-based sample recovery described above relies on the physical separation of two phases by a fluorinated oil–attracting (fluorophilic) and water-repelling (hydrophobic) PTFE membrane.⁴⁴ Membrane separation of immiscible liquids is commonly used for oil and water separation.^44–46 This process is also called de-emulsification, and it is essential in industries involving the recovery of solvents and desalination of oils, as well as in the petroleum industry, especially during crude oil production.^47–49 Membranes made of PTFE, composites based on PTFE, and other hydrophobic-oleophilic membranes are incredibly useful in filtering out oil from water.^44–46,50 Fluorinated oil extraction from droplet microfluidics emulsions in mass spectrometry sample preparation has been demonstrated,⁵¹ but the phenomenon has not been applied to cell or spheroid recovery from droplet microfluidics emulsions. The difference in surface energies between aqueous media and HFE-7500 oil, when applied to a PTFE membrane, is evident from the differences in contact angles with the PTFE membrane: That of pure HFE-7500 oil has been reported to be 0°,⁵² while the water contact angle on PTFE ranges between 90° and 110° depending on the pH.⁵³

Once emulsion is deposited on the PTFE membrane, the HFE oil wets the surface of the PTFE and spreads around and across the membrane immediately ( Fig. 4A ). With the removal of the HFE oil also comes the withdrawal of the surfactant. These two factors destabilize the emulsion, forcing the droplets to merge (Fig. 4B and 4C). This process is rapid and straightforward.

Figure 4.

A membrane-based sample recovery process based on physical separation of two phases by an oil-attracting (oleophilic) and water-repelling (hydrophobic) membrane made of polytetrafluoroethylene (PTFE). (A) The hydrofluoroether (HFE) oil wets the surface of the PTFE and spreads around and across the membrane immediately after the emulsion hits its surface, because the oleophilic nature of PTFE attracts it. (B) The physical removal of the HFE oil from the emulsion (partly aided by HFE evaporation) destabilizes the emulsion, forcing the droplets to merge. (C) Spheroids released from droplets gather in one big, oil-free droplet. (D) The process is automation ready and easy to perform solely with a liquid-handling robot (LHR).

The method presented here allows for a simple chemical-free sample de-emulsification for droplet microfluidics, which is inert to the cells.⁵⁴ This method stands in contrast to commonly used chemicals such as perfluorooctanol, which may interfere with the results of the biological experiments performed in droplets.⁵⁵ The membrane-based method is easy to scale up, is ready for automation, and does not affect the biological sample entrapped in droplets.

Spheroid Viability

Published results indicate that droplet microfluidics used in spheroid formation has limited impact on the viability of the cells entrapped and the spheroids formed in droplets.^20,34 The majority of published droplet microfluidics–based spheroid-forming protocols rely on double emulsions^34,41–43 or on adding hydrogels^34,41–43 to the culture medium in which the cells aggregate.

To test whether the phase separation technique (membrane-based sample recovery) used to release the dispersed phase from the emulsion influences the overall viability of the cells that form spheroids, we used fluorescent viability stains. Cell viability was assessed by adding Hoechst 3334, propidium iodide, and Calcein AM mixture to the recovered spheroids in culture medium. Following 30 min of incubation in the dark, fluorescence microscopy images of the stained spheroids were acquired (in FITC, TRITC, and DAPI; Fig. 3C ), and the number of live, dead, and nuclear-stained cells was counted on randomly selected images containing ≥40 spheroids. The overall viability of the HEK cell–based sample after an overnight incubation, during which the cell aggregation took place and the cell spheroids formed, was 92% ( Fig. 3F ). The initial viability of the HEK cells used to form spheroids was 96%, indicating that the whole process of the transformation of cell suspension into cell spheroids and the PTFE membrane–based recovery using our automated droplet microfluidic platform resulted in a 4% decrease in HEK cell viability. We have also observed that with the increasing number of HEK cells entrapped in droplets, the percentage of viable cells forming spheroids decreased to 83% and 68% for the 4.7×10⁶ HEK cells/mL and 7×10⁶ HEK cells/mL samples, respectively.

In the case of spheroids created with RT4 and A-431 cells (Suppl. Figs. 3 and 4), the initial viability of the sample was 97% and 96%, respectively. The final measured viability of the cells, which went through the processes of spheroid formation and PTFE membrane–based recovery using the automated droplet microfluidic platform, was 96.5% for RT4 and 90.5% for A-431 cell–based spheroids. These latter values reflect decreases of 0.5% and 5.5%, respectively, in the viability of the cells (this counts for all of the cells that were in droplets and were recovered: both the cells that have not aggregated and also those that formed spheroids).

We also have observed differences in the viability of the cells that have aggregated and were part of the formed and recovered spheroids.

In case of HEK cell–based spheroids, most of the dead cells were in the spheroids and not free-floating in the culture medium. In contrast, RT4 and A-431 cell–based spheroids were composed mainly of the cells that were alive: The measured viability of the cells forming spheroids was 99% and 98.8% for RT4 and A-431 cell–based spheroids, respectively. In both cases, the dead cells that were recovered from droplets were mainly free-floating in the culture medium.

The method used for spheroid recovery may also influence the overall viability of the biological content recovered from droplets. As a proof-of-concept experiment, we have recovered A-431 cell–based spheroids entrapped in droplets using either perfluorooctanol or the PTFE membrane–based method paired with the automated droplet microfluidic platform. All of the measured viability parameters were lower for the chemically recovered sample.

The overall viability of the recovered cell content was 87%, with a 9% drop in viability when compared to the 96% viability of the initial sample entrapped in droplets. At the same time, the A-431 cell–based sample recovered with a PTFE membrane–based method experienced only a 5.5% drop in viability. Also, the percentage of viable cells that were forming spheroids was 1% lower when compared to the sample recovered with an alternative PTFE membrane–based method (97.8% vs. 98.8%).

From the results of the proof-of-concept experiments, it seems that the newly proposed PTFE membrane–based method for sample recovery has a lower negative impact on the viability of the formed spheroids.

Characterization of the Cell Spheroids Produced Using Automated Droplet Microfluidics

To characterize cell spheroids, we have chosen to use the circularity and aspect ratio parameters of the recovered spheroids, because those parameters are somewhat independent of the number of cells forming cell aggregates. The circularity of the spheroids was analyzed using ImageJ’s Measure command, which computes circularity of an object according to the following formula: 4π × area / perimeter2, in which circularity of 1.0 indicates a perfect circle.46 The mean value of circularity of analyzed HEK, RT4, and A-431 cell–based spheroids was similar: 0.7, 0.69, and 0.68 [standard deviation (SD), 0.07, 0.09, and 0.11], respectively. The mean aspect ratio (height vs. width) of the spheroids of the three different tested cell lines was 1.37 (SD, 0.25, 0.25, and 0.24).

Both parameters indicate that the spheroids produced are not perfect spheres, as is evident from images (Fig. 3D and Suppl. Figs. 2 and 3). This variability could be due to cell aggregation before droplet formation or the way that cells sediment and aggregate in a spherical shape at the bottom of a droplet.³⁴

Droplets form the closest packing of undeformed spheres at a volume fraction of 0.74.⁴⁵ Based on that estimation, about 26% of the densely packed emulsion generated by the automated droplet microfluidic platform was the continuous phase (HFE oil), and the remaining 74% was the dispersed phase (the spheroids suspended in culture medium). With a volume of 12.7 nL and a constant average number of cells in the generated droplets, we estimate that in one run, the robotically automated droplet microfluidic platform produces more than 17,000 droplets filled with HEK cells/spheroids. Since the duration of one run is 12 min, the robot can cycle five different samples in 1 h, generating up to 85,000 spheroids. This puts the estimated maximum capacity of the platform, at this stage of its development, at 120 different samples per day, which would result in producing 2.04×106 spheroids in 48 h, because similar time is needed for the robot to recover the spheroids from the emulsion after overnight incubation.

Conclusion

This article presents a robotically automated droplet microfluidic platform that allows for production of cell spheroids independent from a human operator. The new platform was tested as an automated assembly line for scaffold-free HEK, RT4, and A-431 cell spheroids, allowing for generation of more than 17,000 spheroids in a single 12-min run. The system’s maximum throughput is 85,000 spheroids per hour. We also present a scalable method for sample recovery from the droplets, which eliminates the need for the use of chemicals for disrupting the emulsion and supports process automation. We believe that the low price of the main component of the robotically automated droplet microfluidic platform, the high production throughput, and the full automation of the process decrease the generation cost of a single spheroid to the scale needed for it to lead drug discovery screening, matching the costs at which spheroids could be used as building blocks for organ-scale regenerative medicine.

Supplemental Material

DS_TECH877376 – Supplemental material for Rapid Production and Recovery of Cell Spheroids by Automated Droplet Microfluidics

Supplemental material, DS_TECH877376 for Rapid Production and Recovery of Cell Spheroids by Automated Droplet Microfluidics by Krzysztof Langer and Haakan N. Joensson in SLAS Technology

Footnotes

Supplemental material is available online with this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received the following financial support for the research, authorship, and/or publication of this article: The authors acknowledge funding from VINNOVA, the Knut and Alice Wallenberg Foundation, and the Novo Nordisk Foundation.

ORCID iD

Haakan N. Joensson

References

Thorsen

Maerkl

S. J.

Quake

S. R.

Microfluidic Large-Scale Integration. Science. 2002, 298, 580–584.

Husser

M. C.

P. Q. N.

Sinha

; et al An Automated Induction Microfluidics System for Synthetic Biology. ACS Synth. Biol. 2018, 7, 933–944.

Cole

R. H.

Tang

S.-Y.

Siltanen

C. A.

; et al Printed Droplet Microfluidics for On Demand Dispensing of Picoliter Droplets and Cells. Proc. Natl. Acad. Sci. U.S.A. 2017, 114, 8728–8733.

Tran

T. M.

Kim

S. C.

Abate

A. R.

Robotic Automation of Droplet Microfluidics. bioRxiv. 2018.

Boitard

Cottinet

Bremond

; et al Growing Microbes in Millifluidic Droplets. Eng. Life Sci. 2015, 15, 318–326.

Ahmadi

Samlali

P. Q. N.

; et al An Integrated Droplet-Digital Microfluidic System for On-Demand Droplet Creation, Mixing, Incubation, and Sorting. Lab Chip. 2019, 19, 524–535.

Shih

S. C. C.

Gach

P. C.

Sustarich

; et al A Droplet-to-Digital (D2D) Microfluidic Device for Single Cell Assays. Lab Chip. 2015, 15, 225–236.

Sharma

Srisa-Art

Scott

; et al Droplet-Based Microfluidics. In Microfluidic Diagnostics: Methods and Protocols; Jenkins

G.;

Mansfield

C. D.

, Eds. Humana Press: Totowa, NJ, 2013, pp. 207–230.

Baret

J.-C.

Surfactants in Droplet-Based Microfluidics. Lab Chip. 2012, 12, 422–433.

10.

Wagner

Thiele

Weinhart

; et al Biocompatible Fluorinated Polyglycerols for Droplet Microfluidics as an Alternative to PEG-Based Copolymer Surfactants. Lab Chip. 2016, 16, 65–69.

11.

Dagkesamanskaya

Langer

Tauzin

A. S.

; et al Use of Photoswitchable Fluorescent Proteins for Droplet-Based Microfluidic Screening. J. Microbiol. Methods. 2018, 147, 59–65.

12.

Kim

J. H.

Jeon

T. Y.

Choi

T. M.

; et al Droplet Microfluidics for Producing Functional Microparticles. Langmuir. 2014, 30, 1473–1488.

13.

Debon

A. P.

Wootton

R. C. R.

Elvira

K. S.

Droplet Confinement and Leakage: Causes, Underlying Effects, and Amelioration Strategies. Biomicrofluidics. 2015, 9, 024119.

14.

Rouzeau

Dagkesamanskaya

Langer

; et al Adaptive Response of Yeast Cells to Triggered Toxicity of Phosphoribulokinase. Res. Microbiol. 2018, 169, 335–342.

15.

Demaree

Weisgerber

Lan

; et al An Ultrahigh-throughput Microfluidic Platform for Single-Cell Genome Sequencing. J. Vis. Exp. 2018, (135).

16.

Terekhov

S. S.

Smirnov

I. V.

Stepanova

A. V.

; et al Microfluidic Droplet Platform for Ultrahigh-Throughput Single-Cell Screening of Biodiversity. Proc. Natl. Acad. Sci. U.S.A. 2017, 114, 2550–2555.

17.

Klein

A. M.

Mazutis

Akartuna

; et al Droplet Barcoding for Single-Cell Transcriptomics Applied to Embryonic Stem Cells. Cell. 2015, 161, 1187–1201.

18.

Hindson

B. J.

Ness

K. D.

Masquelier

D. A.

; et al High-Throughput Droplet Digital PCR System for Absolute Quantitation of DNA Copy Number. Anal. Chem. 2011, 83, 8604–8610.

19.

Agresti

J. J.

Antipov

Abate

A. R.

; et al Ultrahigh-Throughput Screening in Drop-Based Microfluidics for Directed Evolution. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 4004–4009.

20.

Chan

H. F.

Zhang

Y.-P.

; et al Rapid Formation of Multicellular Spheroids in Double-Emulsion Droplets with Controllable Microenvironment. Sci. Rep. 2013, 3, 3462.

21.

X. Tomasi

R. F.

Sart

Champetier

; et al Studying 3D Cell Cultures in a Microfluidic Droplet Array under Multiple Time-Resolved Conditions. bioRxiv. 2018.

22.

Allazetta

Hausherr

T. C.

Lutolf

M. P.

Microfluidic Synthesis of Cell-Type-Specific Artificial Extracellular Matrix Hydrogels. Biomacromolecules. 2013, 14, 1122–1131.

23.

Headen

D. M.

García

J. R.

García

A. J.

Parallel Droplet Microfluidics for High Throughput Cell Encapsulation and Synthetic Microgel Generation. Microsys. Nanoeng. 2018, 4, 17076.

24.

Yadavali

Jeong

H.-H.

Lee

; et al Silicon and Glass Very Large Scale Microfluidic Droplet Integration for Terascale Generation of Polymer Microparticles. Nat. Commun. 2018, 9, 1222.

25.

Kong

Yuan

Zheng

Y. F.

; et al Automatic Liquid Handling for Life Science: A Critical Review of the Current State of the Art. J. Lab. Autom. 2012, 17, 169–185.

26.

Eglen

R. M.

Klein

J.-L.

Three-Dimensional Cell Culture: A Rapidly Emerging Approach to Cellular Science and Drug Discovery. SLAS Discov. 2017, 22, 453–455.

27.

Cui

Hartanto

Zhang

Advances in Multicellular Spheroids Formation. J. R. Soc. Interface. 2017, 14.

28.

Mironov

Visconti

R. P.

Kasyanov

; et al Organ Printing: Tissue Spheroids as Building Blocks. Biomaterials. 2009, 30, 2164–2174.

29.

Laschke

M. W.

Menger

M. D.

Spheroids as Vascularization Units: From Angiogenesis Research to Tissue Engineering Applications. Biotechnol. Adv. 2017, 35, 782–791.

30.

Mazutis

Gilbert

Ung

W. L.

; et al Single-Cell Analysis and Sorting Using Droplet-Based Microfluidics. Nat. Protoc. 2013, 8, 870–891.

31.

Thiele

Abdelmohsen

; et al Biocompatible Macro-Initiators Controlling Radical Retention in Microfluidic On-Chip Photo-Polymerization of Water-in-Oil Emulsions. Chem. Commun. 2014, 50, 112–114.

32.

Langer

Bremond

Boitard

; et al Micropipette-Powered Droplet Based Microfluidics. Biomicrofluidics. 2018, 12, 044106.

33.

Clausell-Tormos

Lieber

Baret

J.-C.

; et al Droplet-Based Microfluidic Platforms for the Encapsulation and Screening of Mammalian Cells and Multicellular Organisms. Chem. Biol. 2008, 15, 427–437.

34.

Sart

Tomasi

R. F. X.

Amselem

; et al Multiscale Cytometry and Regulation of 3D Cell Cultures on a Chip. Nat. Commun. 2017, 8, 469.

35.

Abate

A. R.

Weitz

D. A.

Syringe-Vacuum Microfluidics: A Portable Technique to Create Monodisperse Emulsions. Biomicrofluidics. 2011, 5, 14107.

36.

Umbanhowar

P. B.

Prasad

Weitz

D. A.

Monodisperse Emulsion Generation via Drop Break Off in a Coflowing Stream. Langmuir. 2000, 16, 347–351.

37.

Thorsen

Roberts

R. W.

Arnold

F. H.

; et al Dynamic Pattern Formation in a Vesicle-Generating Microfluidic Device. Phys. Rev. Lett. 2001, 86, 4163–4166.

38.

Guckenberger

D. J.

de Groot

T. E.

Wan

A. M. D.

; et al Micromilling: A Method for Ultra-Rapid Prototyping of Plastic Microfluidic Devices. Lab Chip. 2015, 15, 2364–2378.

39.

McDonald

J. C.

Whitesides

G. M.

Poly(dimethylsiloxane) as a Material for Fabricating Microfluidic Devices. Acc. Chem. Res. 2002, 35, 491–499.

40.

Breslin

O’Driscoll

Three-Dimensional Cell Culture: The Missing Link in Drug Discovery. Drug Discov. Today. 2013, 18, 240–249.

41.

Kumacheva

Hydrogel Microenvironments for Cancer Spheroid Growth and Drug Screening. Sci. Adv. 2018, 4, eaas8998.

42.

Otsuji

T. G.

Bin

Yoshimura

; et al A 3D Sphere Culture System Containing Functional Polymers for Large-Scale Human Pluripotent Stem Cell Production. Stem Cell Reports. 2014, 2, 734–745.

43.

Huang

; et al Generation and Manipulation of Hydrogel Microcapsules by Droplet-Based Microfluidics for Mammalian Cell Culture. Lab Chip. 2017, 17, 1913–1932.

44.

Feng

Zhang

Mai

; et al A Super-Hydrophobic and Super-Oleophilic Coating Mesh Film for the Separation of Oil and Water. Angew. Chem. Int. Ed. Engl. 2004, 43, 2012–2014.

45.

Chen

Weng

; et al Robust Superhydrophobic Polytetrafluoroethylene Nanofibrous Coating Fabricated by Self-Assembly and Its Application for Oil/Water Separation. ACS Appl. Nano Mater. 2018, 1, 2632–2639.

46.

Qing

Shi

Deng

; et al Robust Superhydrophobic-Superoleophilic Polytetrafluoroethylene Nanofibrous Membrane for Oil/Water Separation. J. Memb. Sci. 2017, 540, 354–361.

47.

Ding

; et al One-Step Preparation of Highly Hydrophobic and Oleophilic Melamine Sponges via Metal-Ion-Induced Wettability Transition. ACS Appl. Mater. Interfaces. 2018, 10, 6652–6660.

48.

Huang

Q.-L.

Xiao

C.-F.

X.-Y.

; et al Study on the Effects and Properties of Hydrophobic Poly(tetrafluoroethylene) Membrane. Desalination. 2011, 277, 187–192.

49.

Ezzati

Gorouhi

Mohammadi

Separation of Water in Oil Emulsions Using Microfiltration. Desalination. 2005, 185, 371–382.

50.

Zhang

Shi

Zhang

; et al Superhydrophobic and Superoleophilic PVDF Membranes for Effective Separation of Water-in-Oil Emulsions with High Flux. Adv. Mater. 2013, 25, 2071–2076.

51.

Pereira

Niu

deMello

A. J.

A Nano LC-MALDI Mass Spectrometry Droplet Interface for the Analysis of Complex Protein Samples. PLoS One. 2013, 8, e63087.

52.

Ochoa

Hatzikiriakos

S. G.

Paste Extrusion of Polytetrafluoroethylene (PTFE): Surface Tension and Viscosity Effects. Powder Technol. 2005, 153, 108–118.

53.

Preočanin

Selmani

Lindqvist-Reis

; et al Surface Charge at Teflon/Aqueous Solution of Potassium Chloride Interfaces. Colloids Surf. A Physicochem. Eng. Asp. 2012, 412, 120–128.

54.

Henry

B. J.

Carlin

J. P.

Hammerschmidt

J. A.

; et al A Critical Review of the Application of Polymer of Low Concern and Regulatory Criteria to Fluoropolymers. Integr. Environ. Assess. Manag. 2018, 14, 316–334.

55.

Karbaschi

Shahi

Abate

A. R.

Rapid, Chemical-Free Breaking of Microfluidic Emulsions with a Hand-Held Antistatic Gun. Biomicrofluidics. 2017, 11, 044107.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

3.73 MB