**1. Introduction**

Compartmentalization of cells or biomolecular samples into monodisperse reaction vessels via droplet microfluidics offers unique benefits to chemical and biological analysis and screening. Key among these benefits are improved throughput, enhanced sensitivity of detection due to the reduced dilution in these small reaction volumes, and the minimization of reagent and sample consumption [1].

Droplet microfluidics, a versatile and high-throughput platform technology, employs pico- to nanoliter-sized aqueous droplets in an immiscible fluorinated oil phase [2]. Fluorinated oils, such as Novec HFE-7500, have several advantageous properties, such as high gas solubility [3] and the availability of biocompatible surfactants [4], making them the preferred choice for biological assays. These biocompatible surfactants usually consist of a fluorophobic-hydrophilic polyethyleneoxide head group and fluorophilic-hydrophobic tails. These amphiphilic characteristics lead to them populating the interface and stabilizing the metastable droplet by lowering the surface tension, preventing coalescence [5]. The droplets are stable over extended incubation periods [6] and at higher temperatures [7,8].

A plethora of applications and assays have been developed using droplet microfluidics. Single cell assays [9,10], droplet PCR [11,12] and viral detection [13], among many others, demonstrate the versatility of the platform and the potential for customizable workflows. Entire microfluidic pipelines are created by combining several unit operations, e.g., combining droplet production, splitting, or sorting, in sequence. Aided by the control offered by microfluidics, a specifically optimized microfluidic device has been developed for virtually every individual unit operation. Moreover, there have been advances in chip design and flow rate management to enable an even wider range of applications [14,15].

**Citation:** Trossbach, M.; de Lucas Sanz, M.; Seashore-Ludlow, B.; Joensson, H.N. A Portable, Negative-Pressure Actuated, Dynamically Tunable Microfluidic Droplet Generator. *Micromachines* **2022**, *13*, 1823. https://doi.org/10.3390/ mi13111823

Academic Editor: Pingan Zhu

Received: 23 September 2022 Accepted: 21 October 2022 Published: 25 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Recently, microfluidic droplets have been explored as vehicles for tissue engineering, e.g., in spheroid assembly [16] or gel encapsulation applications [17,18], and are even gaining traction in research studying stem cell differentiation [19,20]. The droplets used in these applications are usually larger than droplets used for single cell studies to allow for encapsulation of sufficient cell numbers [6] or ensure adequate nutrient supply to maintain cell viability for multi-day droplet culture [21]. Addressing the need for scaling, we previously published an automated workflow for high-throughput spheroid production in microfluidic droplets using a liquid handling robot [6], and presented a production optimization pipeline employing a deep neural network to characterize spheroid morphology [22].

There are numerous active and passive methods for producing aqueous droplets in a continuous phase of which flow-focusing droplet generation is most widely used [23,24]. In negative-pressure droplet generation, a pressure source set below ambient pressure is applied at the device outlet to actuate droplet generation. While positive-pressure or positive flow-driven droplet generation using flow-focusing techniques have been extensively studied, including analysis of the impacts of geometry and flow regimes on droplet size [25–29], negative-pressure operation has received less attention. Conceivably, the general tenets still apply, but conditions such as hydraulic resistances become more impactful [30]. Crucially, while the flow rates of the dispersed and continuous phase are directly controllable in positive-pressure or flow-driven droplet generation, they are not independent input parameters for negative-pressure droplet generation.

However, negative-pressure droplet generation has some key advantages over the more conventional positive-pressure-driven methods. Actuated from the collection outlet, these methods allow for continuous access to the oil and sample inlets and for interfacing with liquid handling robots, enabling automation. Compared to syringe pump-driven droplet generation, these advantages are even more pronounced, and with the added advantage of drastically reduced setup and turnaround time. Hundreds of thousands of droplets, each a compartmentalized reaction vessel, can be produced in a matter of minutes while the samples can be agitated in the inlet reservoir, thus preventing particle aggregation, striation, and sedimentation. This is particularly advantageous for sensitive biological samples, such as mammalian cells.

Mammalian cell culture is dependent on sterile conditions and is often sensitive, as some cells, especially primary or stem cells, react negatively to prolonged exposure to lower-than-physiological temperatures [31–33]. In addition, some antibiotics are lightsensitive, which adds to the need for fast experimental processing for steps outside of the incubator [31]. It follows that any cell processing step should be rapid, and the instrument should be deployable in a constrained space, such as a biosafety cabinet, allowing for a fast turnaround time. Naturally, this not only applies for normal cell culture steps, such as freezing, thawing, and upkeep, but for microfluidic operations as well—possibly even more so, as the cells have to resist additional stresses, such as shear forces. Several commercial providers offer relatively compact pressure controller devices that are used extensively within the microfluidic community. However, they still require additional infrastructure, such as an external power supply and pressure source. This might not pose a challenge to research groups focusing on microfluidics, but it adds a layer of complexity and thus a hurdle to overcome for less expert laboratories wanting to adopt microfluidic techniques.

Thus, while a host of operations is available to a specialized lab with appropriate equipment and sufficient expertise—including, but not limited to, generation, incubation, injection, splitting and sorting—the lack of standardization and use of bulky, customized experimental setups limit the dissemination of these powerful techniques to a wider range of laboratories. This creates a disconnect between what is technically possible today and what is easily available and readily used—both limiting advances in research areas that could benefit from microfluidics tools and hampering awareness of biological users' needs in the droplet microfluidics community.

In order to enable non-specialized labs to make use of droplet microfluidics, we compiled and thoroughly characterized a minimal droplet generation setup using only low-cost, commercially available equipment, with a low barrier to entry in mind. Actuation of liquids by exerting variable (negative) pressures at the outlet and inlets of the system enables wide-ranging tuning of droplet sizes and dynamic changes in droplet composition, using a single circuit geometry. The presented setup can produce a range of droplet rates, sizes, and compositions on demand, providing versatility while maintaining a minimal footprint. We further demonstrated the utility of the setup by encapsulating primary hepatocytes in droplets to generate cell spheroids for screening purposes. As shown, this setup addresses the needs of the wider biomedical research community and encourages the adoption of microfluidic workflows and furthers the democratization of healthcare by, e.g., decreasing costs for screening campaigns.

#### **2. Materials and Methods**

## *2.1. Droplet Generation Setups*

Three different experimental setups were tested and configured as shown in Figure 1. Briefly, a single pump allows for limited size tuning by adjusting the applied pressure at the collection pump; an additional second pump attached to the continuous phase reservoir increases the accessible droplet size (Figure 1A). *Micromachines* **2022**, *13*, x 4 of 14

**Figure 1.** Schematic depictions of the minimal droplet generator setup. (**A**) Illustration of single- and dual-pump configuration. The depiction shows how the components are assembled for operating with one or two pumps; when a second pump is added to the continuous phase reservoir, larger droplets can be generated. (**B**) Schematic depiction of three-pump configuration. The illustration shows how the components are assembled for operating with three pumps. The numbers refer to the part list (see Supplementary Information). Tubing length is customizable, however, we used lengths of 7) 3 cm and 8) 5 cm, with enough overlap to ensure seal. The collection syringe can be any syringe with a Luer lock. **Figure 1.** Schematic depictions of the minimal droplet generator setup. (**A**) Illustration of single- and dual-pump configuration. The depiction shows how the components are assembled for operating with one or two pumps; when a second pump is added to the continuous phase reservoir, larger droplets can be generated. (**B**) Schematic depiction of three-pump configuration. The illustration shows how the components are assembled for operating with three pumps. The numbers refer to the part list (see Supplementary Information). Tubing length is customizable, however, we used lengths of 7) 3 cm and 8) 5 cm, with enough overlap to ensure seal. The collection syringe can be any syringe with a Luer lock.

The microfluidic setup was employed in the same way as described in Section 2.3. To transfer the generated emulsion afterwards, a syringe attached at the Luer T connector is manually aspirated, emptying the collection tank. The emulsion was then imaged using a Ti-E Eclipse (Nikon) and droplet sizes were manually measured using ImageJ [34]. The microfluidic device, the Fluidic 163 (microfluidic ChipShop), features a channel depth of 175 µm, and 140 µm-wide channels for the aqueous and oil phases, respectively. The channel widens to 420 µm 200 µm downstream of the junction (see also Supplemental Figure S1). Attaching two pumps at the dispersed phase reservoirs allows for on-demand tuning of the droplet composition. All necessary microfluidic components

The microfluidic setup was employed in the same way as described in Section 2.3.

To numerically study droplet generation using flow-focusing and negative pressure applied solely at the collection outlet, a 2D approximation of the nozzle region was implemented in COMSOL Multiphysics (COMSOL AB), using the respective densities and dynamic viscosities for water and Novec HFE-7500. Virtual channel boundaries were made using acrylic plastic, and the contact angle was set to superhydrophobic 0 rad to exclude any wetting behavior. The simulation encompassed the first 1.25 s after application of a

and disconnected to remove any pressure differential. The collected emulsion was broken using an antistatic gun, Zerostat 3 (Milty), according to published protocols [35], and the phases were separated into different tubes. The mass of the samples was measured using a high-precision scale (Ohaus) and their volume was calculated considering the densities

of both phases (aqueous phase: 1000 kg/m3, continuous phase: 1614 kg/m3).

*2.6. Numerical Simulation of Negative Pressure Droplet Generation* 

*2.4. Droplet Size Measurements* 

*2.5. Flow Rate Measurements* 

are shown, and a detailed list of parts can be found in the Supplementary Material (see Supplemental Figure S2 and Supplemental Table S1). The reservoir container lids connected to pumps were sealed using circular Parafilm cut-outs, prohibiting air leakage. The entire setup is portable (see Supplemental Figure S3) with a compact footprint.

### *2.2. Pump Characterization*

The mp-gas+ membrane pump was connected to a Pressure Unit S (Fluigent) via microfluidic PEEK tubing (Zeus) and the pressure was recorded using OxyGEN software (Fluigent). The frequency response was assessed in 10 Hz increments between 50 and 800 Hz; the amplitude response was assessed in 2 AU increments between 18 and 250 AU.

#### *2.3. Droplet Rate Measurements*

Fluorinated oil (Novec HFE-7500, 3M) was supplemented with 2% *w*/*v* 008 surfactant (Ran Biotechnologies) for the continuous phase. Fluorescein (Sigma) at a concentration of 20 µM in phosphate buffered saline (PBS, Medicago) was used as the aqueous phase.

The microfluidic chip was connected to tubing as shown in Figure 1A. For the singlepump setup, the mp-gas+ (Bartels Mikrotechnik GmbH) pump at the collection outlet was set to 4.61–13.78 kPa of negative pressure. For the dual-pump setup, the collection pump was set to 13.78 kPa of negative pressure and the mp-gas+ pump attached to the oil inlets was set to 0.3–1.4 kPa of negative pressure.

We focused a 514 nm laser just after the widening of the channel downstream of the nozzle, collected the fluorescence emission using a photomultiplier tube (Hamamatsu), and recorded the fluorescent signal using a custom LabView program (National Instruments).

#### *2.4. Droplet Size Measurements*

The microfluidic setup was employed in the same way as described in Section 2.3. To transfer the generated emulsion afterwards, a syringe attached at the Luer T connector is manually aspirated, emptying the collection tank. The emulsion was then imaged using a Ti-E Eclipse (Nikon) and droplet sizes were manually measured using ImageJ [34].

#### *2.5. Flow Rate Measurements*

The microfluidic setup was employed in the same way as described in Section 2.3. The pumps were started and the emulsion collected. After 2 min, the pumps were stopped and disconnected to remove any pressure differential. The collected emulsion was broken using an antistatic gun, Zerostat 3 (Milty), according to published protocols [35], and the phases were separated into different tubes. The mass of the samples was measured using a high-precision scale (Ohaus) and their volume was calculated considering the densities of both phases (aqueous phase: 1000 kg/m<sup>3</sup> , continuous phase: 1614 kg/m<sup>3</sup> ).

#### *2.6. Numerical Simulation of Negative Pressure Droplet Generation*

To numerically study droplet generation using flow-focusing and negative pressure applied solely at the collection outlet, a 2D approximation of the nozzle region was implemented in COMSOL Multiphysics (COMSOL AB), using the respective densities and dynamic viscosities for water and Novec HFE-7500. Virtual channel boundaries were made using acrylic plastic, and the contact angle was set to superhydrophobic 0 rad to exclude any wetting behavior. The simulation encompassed the first 1.25 s after application of a pressure differential in 5 ms increments. The simulation was carried out on a 2020 M1 16 GB RAM Macbook Pro (Apple).

#### *2.7. Droplet Content Manipulation*

We used the same continuous phase as described in Section 2.3. Fluorescein (Sigma) at a concentration of 20 µM in PBS (Medicago) was used as aqueous phase 1 and 15 µm Fluoro-Max Red Dry Fluorescent Particles (ThermoFisher Scientific) dispersed in Milli-Q water were used as aqueous phase 2.

The microfluidic chip was connected to tubing as shown in Figure 1B. The mp-liq pumps (Bartels Mikrotechnik GmbH) connected to the aqueous phase reservoirs were dynamically controlled, we regulated the amplitude parameters of one pump from 24 AU up to 118 AU, while regulating the other from 118 AU down to 24 AU in a corresponding, opposed fashion. A mp-gas+ pump (Bartels Mikrotechnik GmbH) was connected at the collection outlet and set to −13.8 kPa. The emulsion was imaged using a Ti-E Eclipse (Nikon) and ImageJ [34]. Median droplet fluorescence and particle number were manually measured using ImageJ.
