An integrated microfluidic platform to fabricate single-micrometer asymmetric giant unilamellar vesicles (GUVs) using dielectrophoretic separation of microemulsions

Abstract

We present a microfluidic technique that generates asymmetric giant unilamellar vesicles (GUVs) in the size range of 2–14 μm. In our method, we (i) create water-in-oil emulsions as the precursors to build synthetic vesicles, (ii) deflect the emulsions across two oil streams containing different phospholipids at high throughput to establish an asymmetric architecture in the lipid bilayer membranes, and (iii) direct the water-in-oil emulsions across the oil–water interface of an oscillating oil jet in a co-flowing confined geometry to encapsulate the inner aqueous phase inside a lipid bilayer and complete the fabrication of GUVs. In the first step, we utilize a flow-focusing geometry with precisely controlled pneumatic pressures to form monodisperse water-in-oil emulsions. We observed different regimes in forming water-in-oil multiphase flows by changing the applied pressures and discovered a hysteretic behavior in jet breakup and droplet generation. In the second step of GUV fabrication, an oil stream containing phospholipids carries the emulsions into a separation region where we steer the emulsions across two parallel oil streams using active dielectrophoretic and pinched-flow fractionation separations. We explore the effect of applied DC voltage magnitude and carrier oil stream flow rate on the separation efficiency. We develop an image processing code that measures the degree of mixing between the two oil streams as the water-in-oil emulsions travel across them under dielectrophoretic steering to find the ideal operational conditions. Finally, we utilize an oscillating co-flowing jet to complete the formation of asymmetric giant unilamellar vesicles and transfer them to an aqueous phase. We investigate the effect of flow rates on properties of the co-flowing jet oscillating in the whipping mode (i.e., wavelength and amplitude) and define the phase diagram for the oil-in-water jet. Assays used to probe the lipid bilayer membrane of fabricated GUVs showed that membranes were unilamellar, minimal residual oil remained trapped between the two lipid leaflets, and 83% asymmetry was achieved across the lipid bilayers of GUVs.

INTRODUCTION

Synthetic vesicles have served as suitable models in biological science to study the properties of cell membranes,^1–3 for the in vitro encapsulation and incorporation of nanoparticles and biomolecules,^4–8 and as vehicles for drug delivery in engineered vaccines and antibiotics.^9–15 These structures can exist in different size ranges and have many interrelated characteristics, such as mechanical properties, curvature, permeability, and lipid bilayer unilamellarity and composition. Small unilamellar vesicles (SUVs) are typically below 100 nm, large unilamellar vesicles (LUVs) are between 100 nm and 1 μm, and giant unilamellar vesicles (GUVs) are above 1 μm.¹⁶ Although the role of the lipid bilayer in cell function has been extensively studied, the biological significance of more subtle membrane characteristics (i.e., compositional asymmetry and membrane curvature) remains unresolved. Discovering these interconnected roles requires the ability to fabricate complex model membrane systems that are physiologically more relevant to their natural systems (i.e., in smaller size ranges possessing asymmetry across their lipid bilayer, where the lipid composition in the inner leaflet differs from the outer leaflet). Conventional methods such as electroformation,^17–19 gel hydration,^17,20,21 bulk inverted emulsion,^22–25 extrusion,^26,27 and microfluidics^28–30 have addressed the needs in fabricating synthetic vesicles (e.g., giant unilamellar vesicles); however, the lipid architecture and composition, encapsulation efficiency, fabrication rate, and yield have not been satisfactory. For instance, GUVs made by the classical electroformation technique suffer from relatively low yield and polydispersity in size, while in gel hydration, the hydrogel residuals incorporated in the membrane significantly affect the properties of lipid bilayer membrane models.^17,19 Microfluidics has addressed some of these challenges; in particular, uniformity in size, controlled encapsulation, and high fabrication throughput. However, sophisticated technologies are yet to be developed to fabricate model membranes that are within the size range of natural organelle membranes and outer membrane vesicles³¹ (i.e., single- to sub-micrometer in diameter). Also, the ability to fabricate synthetic model membranes that possess complex compositions and architectures (i.e., asymmetry across the lipid bilayer)^32–34 is a paradigm shifting advancement in membrane biology research, and we look to extend this advancement to smaller vesicles.

In fabricating synthetic vesicles using droplet-based microfluidics, researchers have utilized water-in-oil-in-water double emulsions as precursors to accommodate the self-assembly of amphiphilic lipid blocks on the inner and outer oil–water interfaces. However, current techniques are limited by the channel geometries and can fabricate double emulsions in the size range of 5–150 μm in diameter.^35,36 Nevertheless, other active techniques in droplet generation have been deployed in microfluidics (i.e., electrospray and tip streaming) to reduce the emulsion size to single- to sub-micrometer droplets.^37–40 Deshpande et al. developed a microfluidic technique using alkane-based lipid-carrying phases to form unilamellar, monodisperse, and cell-sized liposomes. They fabricated double emulsions in the size range of 5–20 μm that went through a spontaneous solvent-extraction and detachment of the organic phase (i.e., dewetting mechanism) and left behind a fully maturated liposome within a few minutes. Although their technique was able to fabricate GUVs in the single-micrometer range, the architecture across the lipid bilayer of the GUVs remained symmetric.⁴¹ Matosevic and Paegel introduced a microfluidic method to build and immobilize water-in-oil emulsions to perform a systematic series of arbitrary phospholipid monolayer depositions.⁴² Using this technique, they demonstrated a layer-by-layer (LBL) membrane assembly of asymmetric unilamellar and multilamellar vesicles with an average diameter of 55 μm. However, they discussed that tuning the size of vesicles was constrained by the ability of their device to capture the emulsions for deposition.

To construct an asymmetric lipid architecture in synthetic vesicles and imitate the profile of cytoplasmic cell membranes more accurately, benchtop and hybrid on-chip techniques have been developed (i.e., $\beta$-cyclodextrin exchange,^33,43,44 hybrid microfluidics and centrifugation,^25,45,46 and hemifusion⁴⁷). Nevertheless, the fabrication rates using these techniques are limited. Continuous droplet-based microfluidic methods that are able to build asymmetric GUVs with control over lipid composition and lumen encapsulation are a potential option to increase throughput. To address this need, multiple steps for building GUVs using droplet-based microfluidics need to be integrated into one chip to autonomously fabricate and manipulate microscopic objects. Therefore, there are several passive and active techniques that have been developed to sort and separate cells, enrich biological samples, and manipulate soft compartments [e.g., deterministic lateral displacement (DLD),⁴⁸ pinched-flow fractionation (PFF),^28,49,50 and dielectrophoretic (DEP) steering^51–56]. We previously showed that passive PFF was able to steer water-in-oil emulsions across oil streams using an array of posts.^32,57,58 However, to achieve single-micrometer vesicles, a combination of active and passive manipulation techniques (e.g., PFF and DEP) is needed to establish asymmetry across the lipid bilayer of GUVs.

Here, we present a microfluidic chip integrated with an electric circuit to generate and manipulate femtoliter-to-picoliter water-in-oil emulsions. We use these microemulsions as precursors to build synthetic vesicles with lipid bilayer architectures that are physiologically more relevant to natural cytoplasmic membranes (i.e., possessing an asymmetric lipid bilayer). We modulate the emulsion size precisely by changing the pneumatic pressures in reservoirs using a pressure control unit (Fig. 1). We investigate a hysteretic behavior in a co-flowing jet breakup that generates emulsions. We separate the water-in-oil emulsions from the first carrier oil to another stream that contains a different phospholipid using DEP and PFF separation (Fig. 2). This provides the infrastructure to build an asymmetric lipid bilayer in fabricated GUVs. We also report a novel technique to encapsulate the inner aqueous lumen in a lipid bilayer membrane with minimal organic residual remaining trapped between the two lipid leaflets. To achieve this, we optimize the device operation to excite instabilities of a confined hydrodynamically focused oil-in-water jet by inducing viscous shear forces in the microchannels. This initiates a whipping mode that agitates the oil jet and causes the water-in-oil emulsions traveling in the jet to move across the oil–water interface. As the emulsions travel across the jet interface, the outer leaflet assembles on the surface of emulsions, completing the bilayer (i.e., GUV). We characterize the fabricated GUVs to ensure unilamellarity and asymmetry across the lipid bilayer membranes.

FIG. 1.

Illustration of the experimental setup and fluid flows in our microfluidic device. Two amber vials were used as fluid reservoirs containing the inner aqueous and Oil + Lipid₁ solutions. The pressure control unit supplied pressurized nitrogen gas into the reservoirs that was precisely controlled with computer software. Three syringes were loaded on pumps; the ones containing Oil + Lipid₂ and outer aqueous solutions were dispensing while the control syringe was in the withdrawal mode (red arrows indicate the direction of flows). Waste and GUVs collection drained into the centrifuge tubes exposed to the atmosphere.

FIG. 2.

Schematic of the microfluidic device that used three steps to build asymmetric giant unilamellar vesicles (GUVs). (a) The formation of water-in-oil emulsions and self-assembly of Lipid₁ molecules on the water–oil interface of emulsions. The green aqueous phase breaks into water droplets (containing calcein dye) that are dispersed in the Oil + Lipid₁ continuous phase (i.e., red stream). Water-in-oil emulsions proceed further into a serpentine channel and approach the dielectrophoretic steering region. The black rectangle is a portion of the aluminum electrode that electrically shields the droplet generation from the applied electric field in the DEP steering region. (b) An applied DC voltage (400 V) on the black electrodes creates an electric field in the vicinity of the triangular post in the microchannels. As the water-in-oil emulsions travel with the flow in the electric field gradient, the dielectrophoretic force and pinched-flow fractionation steer them to the Oil + Lipid₂ phase (i.e., the yellow stream in the schematic). Consequently, water-in-oil emulsions travel across the two oil streams (i.e., red and yellow) and proceed to the hydrodynamic focusing stage. (c) The induced shear stress on the oil–water interface of the narrowed yellow jet derives instabilities and results in an oscillating stream of Oil + Lipid₂ surrounded by the outer aqueous phase. As the result of oscillation, the giant unilamellar vesicles (GUVs) pinch off from the interface and continue into the side channels for collection. In all fluorescent and bright field images, pressures in the Aq_in and Oil + Lipid₁ reservoirs were set at 420 and 448 mbar, and the flow rates of Oil + Lipid₂, withdrawal control, and Aq_out were set at 350, 300, and 820 μl h^–1, respectively. The scale bars denote 100 μm.

MATERIALS AND METHODS

Materials

1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC) and 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) were purchased in chloroform from Avanti Polar Lipids. Lissamine™ Rhodamine B 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt (Rhodamine-DHPE), and NBD-PE [N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt] (NBD-DHPE) were purchased from Thermo Fisher Scientific and stocked in chloroform at 1.25 mg ml^–1 at −20 °C. Dow SYLGARD™ 184 silicone elastomer clear kit was purchased from Ellsworth. 100 nm aluminum-coated 75 × 25 mm² glass slides were purchased from Dynasil. Negative SU-8 photoresist, SU-8 developer, positive S1813 photoresist, MCC Primer 80/20, and the MF-26A developer were purchased from MicroChem Chorp. Oleic acid, calcein, and poly(ethylene glycol)-block-poly(propylene glycol)-block-poly(ethylene glycol) (Synperonic® F-108) were purchased from Sigma Aldrich. To confirm the unilamellarity of the fabricated GUVs, we purchased $\alpha$-hemolysin from Staphylococcus aureus (Sigma) and stored in stock solution at 0.5 mg ml^–1 at 5 °C. All other chemicals were purchased from VWR.

Experimental setup and device operation

In operating our microfluidic device, the inner aqueous solution and the first carrier oil (i.e., Oil + Lipid₁) were loaded in two 40 ml amber glass vials and transferred to the device through 50 cm long polytetrafluoroethylene (PTFE) tubing (1/32 in. i.d.) (Fig. 1). As is shown in this figure, the nitrogen gas pressure inside the vials was precisely controlled by a pressure control unit (MFCS™-EZ, Fluigent). This was to eliminate the induced noise from syringe pumps in the microchannels.⁵⁹ This makes pressure-driven flows more reliable when using flow-focusing junctions. The second carrier oil and the outer aqueous solution were loaded into 2.5 ml glass syringes (Hamilton). We used syringe pumps (Fusion 200, Chemyx) to generate flow of the control, Oil + Lipid₂, and Aq_out solutions in the channels. All the solutions were supplied to or collected from the ports on the device through translucent PTFE tubing (1/32 in. i.d., Cole-Parmer).

We used a high voltage DC power supply (PS325/2500V-25W, Stanford Research Systems, Inc.) to apply an electric field near the integrated aluminum electrodes (depicted in the schematic in Fig. 2). Typically, DC power supplies provide a constant current or constant voltage to the device under test (DUT) and monitor the resulting voltage drop or current draw. The positive power output (i.e., positive polarity, when both voltage and current have same signs) is called sourcing, and negative power output (i.e., negative polarity, when voltage and current have opposite signs) is called sinking. To optimize the performance of the dielectrophoretic steering, we tested both sourcing and sinking modes of the DC power supply at positive and negative polarities, respectively. When using the sourcing mode, we observed significant fission of the water-in-oil emulsions and chaotic separation. The sinking mode was more stable and provided more reliable DEP steering.

Microfabrication and surface patterning of the integrated microfluidic chip

We cast the polydimethylsiloxane (PDMS) block and the patterned aluminum electrodes of our microfluidic device by utilizing conventional lithography and wet etching processes (Fig. 2 illustrates the design).^60–63 We spin coated a 90 μm thick layer of SU-8 2050 on a 4-in. silicon wafer, soft baked it for 5 min at 65 °C and 12 min at 95 °C, and patterned it by exposure to UV light at 55 mJ cm^–2 through a polyester photomask (Fineline Imaging, Inc.). Following the exposure, we post-baked the wafer for 3 min at 65 °C, 9 min at 95 °C, and then 3 min at 65 °C. Finally, the photoresist layer was developed for 7 min in the SU-8 developer and hard baked at 150 °C for 12 min. We silanized the SU-8 mold by chemical vapor deposition (CVD) for 30 min at ∼17 psi. PDMS was mixed at a ratio of 9:1 of pre-polymer base to the curing agent and cured at 65 °C, overnight. We patterned the aluminum-coated glass slide using photolithography. We first covered the aluminum surface with a layer of MCC primer for 1 min following by spin coating at 3500 rpm for 1 min. We then spin coated positive photoresist S1813 at 3500 rpm for 45 s and baked the glass slide at 95 °C for 1 min. We patterned the electrode design by exposure to UV light at 55 mJ cm^–2, baked at 95 °C for 1 min, and etched in MF-26A for 12 min. We rinsed the residual photoresist with acetone and isopropyl alcohol. The PDMS block was bound to the glass slide with patterned aluminum electrodes by using oxygen plasma (PDC-32G, Harrick Plasma) at 18 W for 2 min.

Preparation of solutions

To prepare the Oil + Lipid₁, we mixed the stock chloroform solutions of phospholipid and dye at a molar ratio of DOPC:Rhodamine-DHPE (200:1) and DOPC:NBD-DHPE (100:1), dried under a gentle nitrogen stream for 20 min, and held the solution under vacuum overnight. Following the complete removal of chloroform, we added oleic acid to the dry lipid film at the final concentration of 10 mg ml^–1 and sonicated for 3 h at 40 °C. We followed the same procedure to prepare oil solutions containing POPC with no fluorescent lipid and POPC:NBD-DHPE (100:1) at 10 mg ml^–1 (i.e., Oil + Lipid₂). Polyvinyl alcohol (PVA) (MW ∼57 000–66 000, 86%–89% hydrolyzed) was added to the inner and outer aqueous solutions at 1 vol. % to prevent coalescence of the fabricated GUVs. We also added sucrose to the inner and outer aqueous solutions at 200 mM and sonicated for 1 h at 40 °C to balance the osmotic pressure across the GUVs. The non-ionic surfactant F-108 was added to the outer aqueous solution at 2 wt. %. Calcein was added to the inner aqueous solution at 1 mM.

Microscopy and image processing

Images were taken using an inverted microscope (Eclipse Ti, Nikon) and observed through the microscope objective. Image sequences were recorded using a supersensitive camera (Prime 95B, Photometrics) in both fluorescent and bright field modes and processed with Micro-Manager software. The excitation wavelengths for the fluorescent probes were ex = 463 nm (NBD-PE) and ex = 560 nm (Rhodamine-DHPE). The emission of fluorescence was measured at em = 536 nm (NBD) and em = 580 nm (Rhodamine-DHPE). For postprocessing the gray scale images in the fluorescent mode, we developed a custom-built MATLAB code that reads and converts the gray scale images into matrices with the pixel values. We then used the Savitzky–Golay filter (a moving average method) to smooth the intensity profiles in each column of the image matrix.

Dielectrophoresis theory

The use of dielectrophoresis is fast becoming a proven technique for manipulating particles and macromolecules in microfluidic systems. Dielectrophoresis is a phenomenon by which mobile electrically neutral particles polarize in a non-uniform electric field.⁶⁴ The electric field can be formed by applying an AC voltage that varies in time and space or by applying a DC voltage that only changes in space and its gradient depends on the electrode design. In dielectrophoresis, dielectric particles move in a continuous medium with a different dielectric constant in the presence of an inhomogeneous electric field. The non-uniform electric field induces a net force on the polarizable particles. The DEP force pulls the particles toward the greater electric field density if the dielectric constant of particles is larger than that of medium (positive DEP), whereas particles with a smaller dielectric constant than the medium move to the opposite direction (negative DEP). The first order or dipole contribution to the time-averaged dielectrophoretic force is given by

$${\mathit{F}}_{DEP}=2\pi {\epsilon }_m{r}^3\mathrm{Re}[K(\omega )]\mathrm{\nabla }(\mathit{E}\cdot \mathit{E}),$$ (1)

where ${\epsilon }_m$ is the absolute permittivity of the medium, r is the radius of suspended particle, $\mathrm{Re}[K(\omega )]$ is the real part of the Clausius–Mossotti (CM) factor, and $\mathrm{\nabla }(\mathit{E}\cdot \mathit{E})$ is the gradient of the square of the applied electric field. The direction of the applied force can be determined by the CM factor as

$$K(\omega )={\displaystyle \frac{{\epsilon }_{\mathrm{particle}}^{\ast }-{\epsilon }_{\mathrm{medium}}^{\ast }}{{\epsilon }_{\mathrm{particle}}^{\ast }+2{\epsilon }_{\mathrm{medium}}^{\ast }},}$$ (2)

where $(\epsilon \ast =\epsilon -i\sigma /\omega )$ is the complex permittivity, $\sigma$ is the electrical conductivity, and $\omega$ is the frequency of the AC electric field.

Jet instabilities and droplet formation in microfluidic devices

In designing microfluidic channels, three main hydrodynamic-based categories have been developed to generate droplets: (i) a co-flowing junction, in which the continuous and dispersed phases flow in the same direction; (ii) a flow-focusing junction, in which the dispersed phase is squeezed into a jet by the immiscible continuous fluid; and (iii) a T-junction where the dispersed phase flows perpendicular to the continuous flow direction.⁶⁵ Most of the droplet-based devices run at low Reynolds numbers. Therefore, the viscous and interfacial tension forces are dominant over inertial forces in governing droplet generation. Also, inherent instabilities and relatively high interfacial tensions of two immiscible fluids (typical oil–water systems have interfacial tensions of $\gamma =1\text{–}20\mathrm{mN}{\mathrm{m}}^{\text{–}1}$) cause one fluid to break into droplets dispersed in the continuous phase. For instance, in flow-focusing geometries, a column of liquid that is squeezed by an immiscible liquid tends to break into monodisperse droplets to minimize the interfacial area/energy of the two immiscible fluids. Depending on the fluid properties, flow rates, and channel geometries, the balance between the viscosity and interfacial tensions can result in different regimes of droplet formation (e.g., dripping, microdripping, jetting, tip-streaming, and tip-multi-breaking).⁶⁶

In addition to droplet generation regimes in flow-focusing geometries, high shear stress on the oil–water interface of the column of liquid may initiate the transition from the varicose to the whipping (bending, kink) mode for large Weber numbers.⁶⁷ In whipping mode, surface tension has a stabilizing effect (i.e., maintains the oil ligament in the form of a long jet), and the destabilizing factor is purely hydrodynamic. Hydrodynamic whipping is enhanced by the jet's viscosity, which dampens the varicose mode in favor of bending perturbations. Acero et al. have reported that the evolution of lateral perturbations in whipping mode is the result of the competition between the destabilizing hydrodynamic factors associated with the relative velocity between the two immiscible flows and the restoring capillary stress due to the surface tension.⁶⁸ Their stability analysis of a jet emitted from a fluid meniscus in air revealed the existence of three transitions that resulted in the whipping mode: (i) the meniscus instability where the liquid meniscus oscillates laterally while continuously emitting a microjet (global whipping), (ii) the convective/absolute instability transition in the jet where the whipping oscillations are convected downstream and upstream and spread over the entire jet at moderately high Weber numbers, and (iii) the local instability at the meniscus tip due to surface tension effects where the lateral oscillations may originate in the meniscus and not in the emitted jet.

RESULTS AND DISCUSSION

Summary of device operation

In fabricating single-micrometer asymmetric GUVs, we utilized the spontaneous self-assembly of amphiphilic phospholipids on the water–oil interface of emulsion precursors. As is shown in the schematic of our microfluidic device (Fig. 2), we started the fabrication process by forming water-in-oil emulsions at high throughput. The two fluorescent images in Fig. 2(a) confirm the encapsulation of the inner aqueous solution [tagged with calcein (green)] that is dispersed in the continuous carrier oil phase containing Lipid₁ [DOPC mixed with Rhodamine-DHPE at 200:1 molar ratio (red)]. Phospholipids that self-assemble on the water–oil interface of the emulsions will comprise the inner leaflet lipid layer of the GUVs. To provide sufficient time for the lipids to assemble on the emulsion interface, the flow was directed into a serpentine channel. Travel through the serpentine channel provided enough time to the lipid molecules to self-assemble on the water–oil interface and was in accordance with other studies using microfluidics.³² In addition, we confirmed the assembly of Lipid₁ under fluorescence microscopy after the separation step [Fig. 2(b) and Movie S1 in the supplementary material].

To establish an asymmetric lipid architecture across the lipid bilayer membrane of the GUVs, we utilized a DEP steering separation technique combined with pinched-flow fractionation to deflect the water-in-oil emulsions from the (red) oil stream containing DOPC into the Oil + Lipid₂ stream containing POPC [Fig. 2(b)]. The triangle post condenses the streamlines of the two oil streams above its apex. This forces the emulsions to travel across the streamlines under the DEP force toward the black pin electrodes (positive DEP). Afterward, we collected the separated emulsions and sent them into a hydrodynamic focusing region (depicted in the schematics in Fig. 2). To obtain the hydrodynamic focusing, we used high flow rates of the outer aqueous solution (Aq_out) to form a thin oil thread (Oil + Lipid₂, depicted as the yellow stream in Fig. 2). The difference between the flow velocities in Aq_out and Oil + Lipid₂ streams induced shear stresses on the oil–water interface of the jet that resulted in jet oscillation. As is shown in Fig. 2(c), the separated water-in-oil emulsions containing calcein traveled inside the oscillating jet and eventually broke through the jet interface. We note that the jet interface was decorated with a self-assembled layer of POPC and the F-108 surfactant. The emulsions were able to pass across this interface and were enclosed with an asymmetric lipid bilayer consisted of Lipid₁ and Lipid₂ (i.e., DOPC inside and POPC outside) to form vesicles. The fabricated vesicles traveled along the two side channels and were collected in a centrifuge tube (Figs. 1 and 2). We tested their characteristics to ensure the formation of asymmetric unilamellar vesicles.

Generating monodisperse water-in-oil emulsions as GUV precursors

As the first step of GUV fabrication, we developed a robust pressure-driven technique capable of forming water-in-oil emulsions at different regimes of droplet generation (Fig. 3). Our goal was to eliminate the induced noise from the syringe pumps and to achieve a size range an order of magnitude smaller than emulsions made by other microfluidic methods.^{32,35,36,46,69} As is shown in Fig. 1, a pressure control unit was used to precisely control the pneumatic pressures in two reservoirs connected to the device (i.e., we used amber vials as pressurized reservoirs containing Aq_in and Oil + Lipid₁). Changes in the reservoirs’ pressures were ultimately translated, through the tubing, into kinetic flow velocities in the channels that intersected at the first flow-focusing region [Figs. 2(a) and 3]. We defined the Reynolds number as $Re=d_{\mathrm{nozzle}}\sqrt{2P\rho }/\mu$, where $d_{\mathrm{nozzle}}$ is the nozzle opening in the first flow-focusing region [Fig. 3(a)], P is pressure in the inner aqueous or Oil + Lipid₁ reservoirs, $\rho$ is the density, and $\mu$ is the dynamic viscosity. We note that for the range of induced pressures in the oil and aqueous reservoirs, the Reynolds numbers were in the ranges of $0.3<R{e}_{\mathrm{Oil}}+{\mathrm{Lipid}}_1<0.7$ and $10<R{e}_{\mathrm{A}{\mathrm{q}}_{\mathrm{in}}}<22.5$, which are typically smaller than the values for the original flow-focusing studies.^70,71 By maintaining a constant pressure in the Aq_in reservoir ($P_{\mathrm{Aq}}$) and increasing or decreasing the pressure in the Oil + Lipid₁ reservoir ($P_{\mathrm{Oil}}$), we acquired different multiphase flows (e.g., water-in-oil emulsions or a water-in-oil jet, Fig. 3). At relatively low values of $P_{\mathrm{Oil}}$, the aqueous phase formed into a solid jet (jetting) at the flow-focusing region where viscous effects dominate over inertia and surface tension [Fig. 3(a)]. In contrast, by increasing $P_{\mathrm{Oil}}$, the jet broke into an array of dispersed water-in-oil emulsions (dripping) close to the nozzle exit. By comparing Figs. 3(b)–3(d), we found that higher $P_{\mathrm{Oil}}$ shrank the thin oil jet in the nozzle that broke into smaller emulsions.

FIG. 3.

The effect of pressure in the Oil + Lipid₁ reservoir on the size of the water-in-oil emulsions. The pressure in the Aq_in reservoir remained at 400 mbar. Pressures in the Oil + Lipid₁ reservoir were (a) 398, (b) 410, (c) 428, and (d) 432 mbar. We note that the asymmetric bulge on the top side of the junction could be due to nonhomogeneous surface wettability of the top and bottom PDMS surfaces. This could happen during the SU8 mold fabrication that ultimately resulted in different flow resistances in the top and bottom channels and an asymmetric meniscus using the pressure-driven flow control unit. The black rectangle is a portion of the aluminum electrode that electrically shields the droplet generation from the applied electric field in the DEP steering region. The scale bar denotes 100 μm.

To tune the operational condition in forming emulsion precursors, we investigated the jet breakup in the first flow-focusing region for a wide range of pressures and discovered a hysteretic behavior in the transition from a water-in-oil jet to dispersed water-in-oil emulsions, and vice versa. Figures 4(a) and 4(b) show the phase diagram indicating the two modes: a stable jet (i.e., yellow crosses demonstrating jetting mode) and water-in-oil emulsions (i.e., blue circles demonstrating dripping mode). Comparing Figs. 4(a) and 4(b) reveals that the transition between a stable jet and the emulsion formation (i.e., the boundary between the yellow and blue parameter spaces) does not reoccur at the same relative pressure (i.e., $P_{\mathrm{Oil}}-P_{\mathrm{Aq}}$). We propose that this hysteretic behavior is due to two different transitions: (i) jet fragmentation into emulsions when increasing $P_{\mathrm{Oil}}$ [i.e., from jetting to dripping, Fig. 4(a)] and (ii) jet formation when decreasing $P_{\mathrm{Oil}}$ [i.e., from dripping to jetting, Fig. 4(b)]. In the former transition, increasing the pressure in the Oil + Lipid₁ reservoir results in higher velocities in the oil phase (the viscous phase). Consequently, the induced viscous stresses excite the varicose instability at the interface that breaks the jet into emulsions. However, in the latter transition, decreasing $P_{\mathrm{Oil}}$ results in thicker jets in the nozzle and consequently larger emulsion, as is shown in Fig. 3. The resultant increase in flow rate and velocity of the aqueous phase (i.e., inside the jet) convects the induced instabilities in the nozzle downstream and inhibits the emulsion formation. These two transitions do not happen at the same relative pressure (i.e., $P_{\mathrm{Oil}}-P_{\mathrm{Aq}}$) and therefore cause the hysteretic behavior. The two red dashed lines demonstrate the boundaries between the two jetting and dripping regimes—a larger slope of 0.11 in Fig. 4(a) as opposed to 0.05 in Fig. 4(b). The boundaries were specified by taking the average values of the two adjacent blue and yellow data points. In addition, plotting $P_{\mathrm{Oil}}-P_{\mathrm{Aq}}$ vs the dimensionless $P_{\mathrm{Oil}}/P_{\mathrm{Aq}}$ ratio reveals that the hysteretic behavior occurred at $P_{\mathrm{Oil}}>P_{\mathrm{Aq}}$ [Figs. 4(c) and 4(d)]. In other words, both jetting and dripping modes existed at $P_{\mathrm{Oil}}>P_{\mathrm{Aq}}$ when increasing $P_{\mathrm{Oil}}$ [Fig. 4(c)], while decreasing $P_{\mathrm{Oil}}$ only resulted in the dripping mode when $P_{\mathrm{Oil}}>P_{\mathrm{Aq}}$, [Fig. 4(d)]. This validates our hypothesis that in order for the varicose instability to grow, the velocity in the Oil + Lipid₁ stream should be greater than the aqueous phase (i.e., $P_{\mathrm{Oil}}>P_{\mathrm{Aq}}$).

FIG. 4.

Phase diagram of the two regimes for water-in-oil multiphase flows when pressure in the oil reservoir ($P_{\mathrm{Oil}}$) increases [(a), (c), and (e)], and pressure in the oil reservoir ($P_{\mathrm{Oil}}$) decreases [(b), (d), and (f)] while the pressure in aqueous reservoir ($P_{\mathrm{Aq}}$) remains constant. (a) illustrates the parameter space where the water-in-oil jet broke into droplets as we increased the pressure in the oil reservoir ($P_{\mathrm{Oil}}$). The slope of the fitted line is 0.11 that determines the boundary between the two regimes of jetting and dripping. (b) illustrates the parameter space where water-in-oil droplets converted to a solid jet as we decreased the pressure in the oil reservoir ($P_{\mathrm{Oil}}$). The slope of the fitted line is 0.05 that determines the boundary between the two regimes of jetting and dripping. (c) and (d) represent the relationship between the pressure difference ($P_{\mathrm{Oil}}-P_{\mathrm{Aq}}$) and the pressure ratio ($P_{\mathrm{Oil}}/P_{\mathrm{Aq}}$) in the two reservoirs. In (e) and (f), the relationship between the dimensionless size (i.e., of the water-in-oil emulsions or the water-in-oil jet) and the corrected pressure ratio ${[(P_{\mathrm{Aq}}\cdot {\mu }_{\mathrm{Oil}})/(P_{\mathrm{Oil}}\cdot {\mu }_{\mathrm{Aq}})]}^{1/3}$ is shown. The data in (e) and (f) are normally distributed around the fitted line within the 95% prediction interval. In all plots, the blue circles represent the dripping mode of water-in-oil emulsion generation and the yellow crosses represent the jetting mode where no emulsions were produced. Flow rates in the Oil + Lipid₂, withdrawal control, and Aq_out were set at 280, 300, and 820 μl h^–1, respectively. The waste and GUVs collection ports were exposed to the atmosphere. Data points were collected in 3 min time intervals. All fitted lines possess R-square values of larger than 90%.

We note that in the absence of the pulsatile flow of the syringe pumps (with using the pressure control unit), the varicose instability gives rise to the emulsion formation and determines their size. Changes to the tubing used in the experiments or channel geometries, while modifying the pressure in the flow-focusing region, are not expected to alter the hysteretic behavior. In addition, changing the flow rates downstream in the Oil + Lipid₂ and outer aqueous channels did not have any effect on the jet breakup transition. This is an additional advantage of the serpentine channel, which creates distance and, therefore, hydrodynamic resistance between the emulsion generation at the first flow-focusing region and the steering and hydrodynamic focusing regions downstream. To shield the electric field and prevent charge accumulation during jet breakup, we incorporated an electrically grounded electrode between the first flow-focusing and DEP steering regions (the black rectangle shown in Fig. 3).

We constructed a physical model to characterize the relationship between the size of the jet or emulsion precursors and the applied pressures in the Aq_in and Oil + Lipid₁ reservoirs. In our model, the volume of the water-in-oil jet or emulsion is proportional to the inner aqueous solution flow rate and inversely proportional to the Oil + Lipid₁ flow rate,

$${\left({\displaystyle \frac{d}{d_{\mathrm{nozzle}}}}\right)}^3\propto {\displaystyle \frac{Q_{\mathrm{A}{\mathrm{q}}_{\mathrm{in}}}}{Q_{\mathrm{Oil}+\mathrm{Lipi}{\mathrm{d}}_1}},}$$ (3)

where d is the diameter of the water-in-oil emulsions or the jet of inner aqueous solution surrounded by the Oil + Lipid₁ stream. The steady state flow rates are defined by $Q=Uwh$, where U is the average velocity in the microchannels, w is the channel width, and h is the channel height.

Simplifying the Navier–Stokes equation in the flow ($x$) direction gives

$${\displaystyle \frac{\mathrm{\partial }P}{\mathrm{\partial }x}=\mu {\displaystyle \frac{{\mathrm{\partial }}^{2}u}{\mathrm{\partial }{y}^2},}}$$ (4)

where u is the flow velocity and y is the direction normal to the flow along the channel height. Combining Eqs. (3) and (4), we conclude

$${\displaystyle \frac{d}{d_{\mathrm{nozzle}}}=k{\left({\displaystyle \frac{P_{\mathrm{Aq}}.{\mu }_{\mathrm{Oil}}}{P_{\mathrm{Oil}}.{\mu }_{\mathrm{Aq}}}}\right)}^{1/3},}$$ (5)

where $P_{\mathrm{Aq}}$, ${\mu }_{\mathrm{Aq}}$, $P_{\mathrm{Oil}}$, and ${\mu }_{\mathrm{Oil}}$ are the absolute pressures and dynamic viscosities of the inner aqueous and Oil + Lipid₁ reservoirs, respectively. Figures 4(e) and 4(f) are plots of dimensionless size ${\pi }_1=d/d_{\mathrm{nozzle}}$ vs the corrected pressure ratio ${\pi }_2={[(P_{\mathrm{Aq}}.{\mu }_{\mathrm{Oil}})/(P_{\mathrm{Oil}}.{\mu }_{\mathrm{Aq}})]}^{1/3}$. The data points from Figs. 4(a) and 4(b) collapsed onto two single lines (i.e., for decreasing and increasing $P_{\mathrm{Oil}}$) and confirmed the linear relationship predicted by our theoretical model in Eq. (5). The constant k (measured experimentally) was 33 and 12 for the dripping and jetting modes, respectively. This implies that there is a linear relationship between the characteristic size and the corrected pressure ratio within a given mode (i.e., dripping and jetting). However, the transition between modes exhibits a hysteretic behavior. In addition, the intersection of the two fitted lines in Fig. 4(e) passed through the point where $P_{\mathrm{Oil}}=P_{\mathrm{Aq}}$; however, the two fitted lines in Fig. 4(f) did not cross. This implies that emulsions and jets of the same size were possible when increasing $P_{\mathrm{Oil}}$, while this was not possible when decreasing $P_{\mathrm{Oil}}$. Also, by comparing the areas above the polynomial red dashed boundaries in Figs. 4(c) and 4(d), we expect to achieve hysteretic behavior only at pressures above the intersection point. We note that the relationship in Eq. (5) is of great importance in tuning the operational conditions when using the pressure control unit to achieve the required size in fabricating water-in-oil emulsion precursors.

Establishing an asymmetric lipid bilayer architecture in GUVs using dielectrophoretic steering of water-in-oil emulsions across streams

To build an asymmetric architecture across the lipid bilayer membranes in fabricated GUVs, we separate water-in-oil emulsions (coated with a layer of Lipid₁ on their interface) from the Oil + Lipid₁ carrier oil phase into a second oil phase that contains a different phospholipid composition (i.e., Oil + Lipid₂). We have previously shown that pinched-flow fractionation (PFF) can hydrodynamically transfer water-in-oil emulsions from one carrier stream into another.^{32,57,58,61,72} This was achieved using a dual-pinching strategy where streamlines were pinched at the leading and trailing edges of a series of triangular posts. The hydrodynamic centers of emulsions in the size range of 20–120 μm were forced across streamlines. However, smaller emulsions (<20 μm) need additional external body forces to redirect their hydrodynamic center across oil streamlines. The left column of images in Fig. 5 demonstrates the control case where we used PFF at different Oil + Lipid₂ flow rates without DEP (i.e., at zero volts) and water-in-oil emulsions (black dots in the red stream) remained in the Oil + Lipid₁ stream. This confirmed the need to incorporate the DEP force as an additional body force that improved the separation of emulsions with smaller radii. Therefore, we utilized DEP forces in addition to PFF in our microfluidic chip to steer water-in-oil emulsions from one carrier oil stream to another in a confined geometry. As is shown in Fig. 2(b), we applied a DC voltage with negative polarity across the black electrodes to generate a non-uniform electric field in the vicinity of the triangle post. The position of the three pin electrodes was selected to achieve the greatest electric field gradient above the triangle post apex where the pinched-flow fractionation occurs (i.e., the middle electrode was aligned with the right edge of the triangle post).

FIG. 5.

All panels illustrate the DEP separation of the water-in-oil emulsions (black dots in the red stream) under different applied DC voltages (columns) and Oil + Lipid₂ flow rates (rows), in the negative mode. The red stream indicates the Oil + Lipid₁ stream that contained DOPC mixed with Rhodamine-DHPE (200:1). The dark Oil + Lipid₂ stream did not contain fluorescently labeled lipid. All other applied pressures and syringe pumps flow rates were kept constant as we changed the Oil + Lipid₂ flow rate and applied DC voltage. White dashed lines were added to specify the PDMS microchannel walls. The scale bar denotes 100 μm.

Equations (1) and (2) show that the magnitude and direction of the DEP force depends on the absolute permittivity of the continuous oil medium, radius of the water-in-oil emulsions, and relative complex permittivity and conductivity of the inner aqueous solution with respect to the oil medium. The dielectric constant (relative permittivity) of water and oleic acid are 78.4 and 2.3 at 25 °C, respectively.⁷³ We note that the presence of other constituents in the inner aqueous (i.e., sucrose, PVA, and calcein) and in oleic acid (i.e., phospholipids) may slightly affect the dielectric constants. We measured the electric conductivity of the inner aqueous phase (∼400 μS cm^–1 using CON 150 Meter Kit, Oaklon) and neglected the electric conductivity of oleic acid. This resulted in $\mathrm{Re}[K(\omega )]=0.99$, which establishes a positive DEP force when using DC voltages. Therefore, emulsions were attracted toward the sharp pin electrodes, where the electric field was highly concentrated, into the Oil + Lipid₂ stream (yellow stream as shown in the schematic in Fig. 2). Consequently, the combination of DEP force and PFF steered the water-in-oil emulsions from the fluorescent (red) oil stream to the (dark) oil stream with no fluorescent dye, Fig. 5. We recall that the triangular post constricts the two oil streams above its apex and facilitates the transverse motion of water-in-oil emulsions across the streamlines (i.e., by pinched-flow fractionation). Also, the length from the trailing end of the triangle post to the downstream bifurcation of the channel provides sufficient travel distance for the emulsions to migrate across the microchannel subject to the DEP force.

During the DEP separation, we observed mixing of the two oil streams as emulsions were transferred between the two (Fig. 5). This may be due to the chaotic motion of the emulsions as they pass the apex of the triangular post under DEP. Therefore, some traces of the red stream containing Lipid₁ were transferred to the dark stream and contaminated the Oil + Lipid₂ flow (Fig. 5). Also, water-in-oil emulsions coalesced into larger emulsions at higher voltages due to the dipole–dipole interaction.⁷⁴ Mixing and coalescence reduced the separation efficiency and consequently disrupted the desired asymmetric architecture across the lipid bilayer membranes in fabricated GUVs. Mixing between the two oil streams was detrimental to controlling the architecture of the membrane (i.e., asymmetry) and needed to be minimized to enhance the performance of the device. Therefore, we defined a metric to quantitatively measure the degree of mixing between the two oil streams in the DEP region. A custom-built MATLAB code was developed to perform image processing and to detect the boundary between the two oil streams [Figs. 6(a)–6(c)]. We then summed all the pixel intensities and averaged within the integration domain {i.e., the area enclosed by the detected boundary between the two oil streams [i.e., green border in Fig. 6(a)] and the upper channel walls}. Finally, we normalized the integrated intensity with respect to the average pixel intensity value upstream in Oil + Lipid₁.

FIG. 6.

Procedure to determine the boundary between the two oil streams (shown by the green border). (a) The DEP steering region in which we applied the electric field. The white dashed lines represent the edges of the channels. The yellow vertical line is an arbitrary column of the image matrix (1200 × 1200 pixels) where we extracted the gray scale values of every pixel. (b) Smoothed intensity profile using the Savitzky–Golay filter. (c) The first centered derivative of the smoothed intensity profile. The red dashed line sets the threshold where we picked the first data point specifying the boundary between the two oil streams on the arbitrary yellow line. We applied this threshold across all the acquired image matrices at different flow rates and applied DC voltages. (d) Plot representing the degree of mixing between the two oil streams (i.e., Oil + Lipid₁ and Oil + Lipid₂) in the DEP steering region. It shows the relationship between the average normalized gray scale intensity per pixel and the Oil + Lipid₂ flow rate for different applied voltages. Each data point indicates the average value of 50 frames. The error bars are ±1 standard deviation.

Figure 6(d) demonstrates the effect of applied DC voltage and Oil + Lipid₂ flow rate on the normalized integrated intensity to quantify mixing between the two oil streams. We also confirmed that negligible mixing occurred between the oil streams when no electric field was applied. This confirmed that the mixing was only a result of DEP steering and transversal motion of water-in-oil emulsions across the streamlines of the two oil streams. Figure 6(d) shows that as we increased the applied DC voltage, the average normalized gray scale intensity in the integration domain increased up to 6% at 400 V, which means that more mixing occurred between the two oil streams. This is due to a stronger DEP steering of water-in-oil emulsions that were separated and dragged from Oil + Lipid₁ into Oil + Lipid₂ stream. Also, the average normalized gray scale intensity decreased with the Oil + Lipid₂ flow rate. This confirms that higher Oil + Lipid₂ flow rates squeeze the Oil + Lipid₁ [fluorescently labeled flow in Fig. 6(a)] stream and suppress mixing between the two oil flows under the DEP separation. These measurements confirm that the rate of mixing between the two oil carrier phases (i.e., concentration of lipid₁ contaminating the Oil + Lipid₂ stream) did not exceed 6% under the strongest DEP at 400 V. Also, the effect of DEP separation can further be diminished by increasing the Oil + Lipid₂ flow rate, as is displayed in Fig. 6(d).

The results in Figs. 5 and 6(d) can guide one to find the optimal operational conditions to separate emulsions under the dielectrophoretic steering, while preventing cross-contamination between the carrier streams to establish satisfactory asymmetry in the lipid bilayer of fabricated GUVs. As is shown in Figs. 2(b) and 5, the separated water-in-oil emulsions (green) continued traveling in the dark Oil + Lipid₂ stream with no fluorescently tagged lipids (Movie S2 in the supplementary material) and approached the hydrodynamic focusing region to complete formation of the lipid bilayer in GUVs (Fig. 2 and Movie S3 in the supplementary material).

Completing the lipid bilayer assembly of GUVs: Depositing the outer leaflet on emulsion templates

Following the DEP steering region, water-in-oil emulsions [i.e., the green droplets traveling in the Oil + Lipid₂ stream in Fig. 2(b)] coated with a layer of Lipid₁ (red) travel into the Oil + Lipid₂ stream (Movie S1 and S2 in the supplementary material). To complete the lipid bilayer formation and transfer the vesicles to the aqueous phase [Fig. 2(c)], we utilized the whipping instability⁶⁸ of an oil-in-water jet (i.e., Oil + Lipid₂ stream co-flowed in the outer aqueous phase) to deposit the outer leaflet lipid layer on the emulsion templates. We employed this method to overcome the geometric limitations of encapsulating single-micrometer water-in-oil emulsions in an oil layer using the conventional encapsulation methods in flow focusing. The other objective was to prevent excess oil from remaining trapped between the two lipid monolayers (i.e., leaflets) and to eliminate the oil extraction step⁷⁵ (to convert double emulsions to the GUVs) from the fabrication process. As is shown in the exaggerated schematic image in Fig. 7(a), the Oil + Lipid₂ stream (yellow) containing the green water-in-oil emulsions is oscillating in a flow of the outer aqueous phase. The water-in-oil emulsions are coated with a layer of Lipid₁ (red) and the oil–water interface of the oscillating oil thread is coated with Lipid₂ (purple). While the jet oscillates laterally, the emulsions are forced across the interface and wrap into a lipid bilayer forming asymmetric GUVs. This is in addition to the non-inertial lift force that is due to the lateral migration of deformable emulsions toward and possibly through the jet interface (i.e., from the high viscous phase (Oil + Lipid₂ stream) to the low viscous phase (outer aqueous phase).⁷⁶ We propose that the dispersed lipids and surfactants in Oil + Lipid₂ and outer aqueous streams, respectively, lowered the interfacial tension on the oil–water interface of the jet and prevented the jet breakup. In particular, the effect of F-108 on lowering the surface tension ($\gamma =0.04\mathrm{mN}{\mathrm{m}}^{\text{–}1}$ for an oil–water interface with 10 mg ml^–1 soy PC in oleic acid and 1 wt. % Pluronic F-108 in the aqueous phase) has been confirmed previously by Krafft et al. in initiating a dewetting transition and forming oil lenses attached to GUVs.⁷⁷ In addition, Jayaprakash et al. reported on the dynamic migration of rigid polystyrene microparticles across an interface of co-flowing streams of two aqueous and oil phases at low Reynolds numbers.⁷⁸ They investigated the role of wetting behavior and the presence of surfactants on the full detachment of microparticles, when the width of the inner aqueous phase approached the diameter of the suspended microparticles. They discussed that the spreading parameter in both the oil- and water-side [i.e., $S_o={\gamma }_{sw}-({\gamma }_{so}+{\gamma }_{wo})$ and $S_w={\gamma }_{so}-({\gamma }_{sw}+{\gamma }_{wo})$, where ${\gamma }_{sw}$, ${\gamma }_{so}$, and ${\gamma }_{wo}$ are the solid–aqueous, solid–oil, and aqueous–oil interfacial energies, respectively] should be negative to achieve full migration of polystyrene microparticles in the presence of a minimum external force. This minimum external energy can be from the intrinsic thermal energy at the interfaces and flow perturbations due to syringe pumps or presence of particles on the interface. We propose that the same process is involved in the detachment of green-labeled water-in-oil emulsions from the oil–water interface of the whipping jet as is pointed by the red arrows in Fig. 2(c). In the detachment/encapsulation of the green-labeled water-in-oil emulsions, a similar dewetting transition occurs as the emulsions come into contact with the jet interface. The interfacial tension between the inner and outer aqueous phases (${\mathit{\gamma }}_{\mathrm{Aq},\mathrm{in}-\mathrm{Aq},\mathrm{out}}$) is greater than the interfacial tension between inner aqueous and oil phase (${\mathit{\gamma }}_{\mathrm{Aq},\mathrm{in}-\mathrm{Oil}}$) and outer aqueous and oil phase (${\mathit{\gamma }}_{\mathrm{Aq},\mathrm{out}-\mathrm{Oil}}$),

$${\mathit{\gamma }}_{\mathrm{Aq},\mathrm{in}-\mathrm{Aq},\mathrm{out}}>{\mathit{\gamma }}_{\mathrm{Aq},\mathrm{in}-\mathrm{Oil}}+{\mathit{\gamma }}_{\mathrm{Aq},\mathrm{out}-\mathrm{Oil}}.$$ (6)

FIG. 7.

(a) Schematic of an oil jet oscillating at the whipping mode. Green water-in-oil emulsions coated with a layer of fluorescent-labeled phospholipid (red) travel in the oil phase. A different phospholipid (purple) is dispersed in the oil phase and is self-assembled on the oil–water interface of the jet. The whipping oil jet agitates the emulsions and forces them to travel across the interface, wrapped in a purple lipid layer and complete the bilayer formation. (b) Phase diagram illustrating three possible multiphase modes, formed in the flow-focusing region, based on the Oil + Lipid₂ and outer aqueous flow rates. Each mode occurred at a specific range of flow rate ratios starting with no meniscus/jet formed to whipping to a straight jet as we increased the oil to aqueous flow rate ratio. (c) The bright-field microscopic image shows that the induced shear stress on the oil–water interface of the narrowed oil jet initiated an instability and resulted in an oscillating stream of Oil + Lipid₂ surrounded by the outer aqueous phase. The flow rates of Oil + Lipid₂, withdrawal control, and Aq_out were set at 350, 300, and 820 μl h^–1, respectively. The scale bar denotes 100 μm.

The dewetting transition for detaching the green emulsions and fabricating the GUVs required the use of Pluronic F-108, which may also affect the membrane properties (e.g., mechanical stabilization,⁷⁹ protection of vesicles against peroxidation,⁸⁰ and increasing the permeability for small molecules⁸¹). Generally, hydrophobic copolymers act as permeants, while hydrophilic ones seal the membrane. Increased permeability should not necessarily be considered a drawback for synthetic GUVs because this could provide a feasible mechanism for membrane transport. Also, the conventional methods for reconstitution of membrane proteins involve the use of surfactants, which perturb the bilayer structure and are subsequently removed.⁸²

As is shown in Fig. 2(c), the Oil + Lipid₂ stream (with a viscosity of 27.6 mPa s at 25 °C and a negligible electrical conductivity surrounded by the outer aqueous solution with electrical conductivity of ∼400 μS cm^–1 and a viscosity of 0.9 mPa s) enters a hydrodynamic focusing region and tapers to a thin oil-in-water jet. To accomplish this, we set a strong shear stress on an oil meniscus and stretched it until a thin ligament was ejected using a parallel co-flowing aqueous stream inside a confining microfluidic geometry. We also deployed a suction force downstream from the control (withdrawal) port to facilitate the oil ligament formation, as shown in Fig. 2. The oil ligament (jet) was precisely controlled by the outer aqueous flow rate and the applied suction. The large relative velocity between the outer aqueous and Oil + Lipid₂ streams (with the optimum value of 13.4 mm s^–1 that results in $Re=2.9$) induced a strong shear stress on the oil–water interface that triggered the transition from the varicose to the whipping (bending, kink) mode for large enough Weber numbers ($We=20.2$ for the stretched oil ligament in this study) [Fig. 2(c)]. In addition, we note that the high viscosity ratio of the oil phase compared to the continuous aqueous phase (${\mu }_{\mathrm{Oil}}/{\mu }_{\mathrm{Aq}}=30.7$) inhibits droplet generation and maintains the jet. This implies that the hydrodynamic whipping is enhanced by the jet's viscosity, which dampens the varicose mode in favor of bending perturbations. Therefore, the convective/absolute instability transition occurs in the jet where the whipping oscillations are convected downstream and upstream and spread over the entire jet.⁶⁸

To better understand the behavior of the whipping mode, the effect of Oil + Lipid₂ and outer aqueous flow rates on the oscillating jet wavelength and amplitude were investigated. We found that the amplitude and wavelength of the whipping jet were insensitive to the Oil + Lipid₂ and outer aqueous flow rates and varied in the range of 88 ± 21 and 764 ± 184 μm, respectively. To achieve the whipping mode and deposit the outer leaflet on emulsion templates, we tuned the flow rates and found the phase diagram in which the device needs to be operated [Fig. 7(b)]. Three regions were observed: (i) at relatively low oil to aqueous flow rate ratios, the outer aqueous stream pushed the upcoming flow of Oil + lipid₂ phase backward and stopped formation of the oil-in-water meniscus and no jet was formed; (ii) at higher ratios, the meniscus was formed and a thin oil thread was ejected and oscillated at the whipping mode; (iii) higher oil to aqueous flow rate ratios resulted in a straight oil-in-water jet that was stable. The whipping mode provided a unique capability to encapsulate the emulsions into the outer leaflet (completing the formation of lipid bilayer membrane) and transfer the GUVs to the outer aqueous solution. We believe that while oscillating, the off-axis motion triggered the dewetting transition by mechanically agitating water-in-oil emulsions on the interface and forcing them to detach from the jet.

Membrane asymmetry and unilamellarity characterization of fabricated GUVs

Following the final step for assembling the lipid bilayer membrane, the GUVs were collected from the two side channels in the outer aqueous phase, as shown in Figs. 1 and 2. The fluorescent image in Fig. 8(a) confirms the full assembly of the lipid bilayer membranes and the formation of GUVs in the size range of 2–14 μm (4 fl–1.4 pl) as indicated in Fig. 8(d). The red florescent signal shows the assembly of DOPC:Rhodamine-DHPE (200:1) as the inner leaflet.

FIG. 8.

Confirmation of fabricating/collecting synthetic vesicles and characterizing their lipid bilayer membrane unilamellarity and asymmetry by a transmembrane protein (α-hemolysin) insertion assay and a fluorescence quenching assay, respectively. (a) illustrates the vesicles that were collected from the chip. The red fluorescence confirms the assembly of DOPC:Rhodamine-DHPE (200:1) as the inner leaflet. (b) shows the average normalized fluorescence intensity for two configurations during the quenching assay: when NBD was deposited in the inner leaflet (red boxes) vs when deposited in the outer leaflet (green boxes). The fluorescence quenching solution (0.5M sodium dithionite) was added to the observation Petri dish until no further changes were observed in vesicle fluorescence after 20 min. Average fluorescence intensities across the vesicles (n > 4) for all the images were measured using ImageJ. The background information was subtracted from each image to set the field outside of the vesicle to an intensity of zero. (c) Vesicles with α-hemolysin embedded into their lipid bilayer displayed an abrupt decay in calcein fluorescence when observed for 1 h. However, vesicles without the transmembrane protein showed a moderate decay in fluorescence intensity under the same exposure time and fluorescent light. This confirms the membrane unilamellarity and minimal oil phase residuals in the lipid bilayer. (d) displays the size distribution from 120 GUV measurements. The scale bar denotes 20 μm.

To assess the asymmetric lipid distribution across the lipid bilayer membranes of the fabricated GUVs, we used fluorescently labeled targeted lipids (NBD-DHPE) to perform a quenching assay on individual GUVs immediately following vesicle formation. Two configurations were tested: (i) inner leaflet: DOPC mixed with NBD-DHPE (100:1), outer leaflet: POPC; and (ii) inner leaflet: DOPC, outer leaflet: POPC mixed with NBD-DHPE (100:1). When NBD is incorporated into the inner leaflet, membrane-impermeable dithionite (NBD quencher, 0.5M aqueous solution of Na₂S₂O₄) is unable to quench the fluorophore.^83–85 As is shown in Fig. 8(b), 83% of the average fluorescence intensity remained unquenched when only the inner leaflet was labeled with NBD. However, when the outer leaflet was labeled with NBD, 82% of the fluorescence was quenched after 20 min. This assay confirmed that the lipid bilayer membrane was 83% asymmetric across the two leaflets in the GUVs. We note that in the former configuration, the 17% reduction in the fluorescence intensity may be attributed to the replacement efficiency (i.e., mixing between the two carrier oil phases) and the outward translocation (i.e., flip-flop) of the phospholipids across the lipid bilayer membranes.

The unilamellarity of the lipid bilayer membranes was probed using a transmembrane protein assay ($\alpha$-hemolysin).²² The membrane protein α-hemolysin is known to span only a single bilayer and oligomerize to form water-filled transmembrane pores.²² We confirmed the formation of these pores by their permeability to membrane-impermeable calcein molecules (encapsulated in the GUVs at 1 mM). Changes in the fluorescence intensity of individual GUVs in time indicated the passive diffusion of calcein molecules to the extracellular domain [Fig. 8(c)]. However, in the absence of embedded $\alpha$-hemolysin in the lipid bilayer membrane [the control case in Fig. 8(c)], the fluorescence intensity inside the GUVs decayed at a much slower rate that was attributed to photobleaching. The comparison between the embedded $\alpha$-hemolysin in the lipid bilayer membrane and the control case confirms that GUVs were permeable to calcein molecules in the presence of transmembrane pores. This assay provides evidence that the membrane is unilamellar (two leaflets the width of $\mathrm{a}$-hemolysin), and there was minimal residual oil trapped between the leaflets. In addition, this assay demonstrates that a protein can be successfully incorporated into the fabricated GUVs using our technique.

CONCLUSIONS

A microfluidic chip integrated with an electric circuit was developed and used to fabricate single-micrometer asymmetric vesicles. We started with generating femtoliter-to-picoliter water-in-oil emulsions as precursors using a pressure-driven flow control unit. We defined a parameter space based on the induced pressures that determined the breakup dynamics of an aqueous phase into an oil phase to form femtoliter-to-picoliter water-in-oil emulsion. We modulated the jet breakup and resultant emulsions size with the induced pressures in the reservoirs and discovered a hysteretic behavior. We showed that there was a linear relationship between the characteristic size and the pressure ratio within a given mode of dripping or jetting. However, the transition between modes displayed a hysteretic behavior while increasing or decreasing the pressure in the oil reservoir. Following emulsion generation, we deployed dielectrophoretic (DEP) and pinched-flow fractionation (PFF) separation to steer water-in-oil emulsions across two oil streams containing different lipid compositions to establish an asymmetric lipid architecture in fabricated GUVs. We observed that as the emulsions travel across streamlines of the oil flows, the two oil streams mix and disrupt the separation efficiency and asymmetric lipid bilayer in GUVs. We investigated the effect of applied DC electric field and the oil flow rates on the degree of mixing between the two oil streams using image processing and found the ideal operational conditions. As the last step in building asymmetric GUVs, we reported a novel technique to assemble the outer leaflet on the water-in-oil emulsion templates and complete the vesicle formation with minimal organic residual remaining trapped between the two lipid layers. To achieve this, we applied strong shear forces on a thin oil-in-water jet containing the separated emulsions to excite natural convective/absolute instabilities on the interface. This initiated a whipping mode that oscillated the jet and forced the emulsions inside the jet to transverse through the jet interface and complete the GUVs formation. We proposed that the lipid/surfactant composition assembled on the jet interface initiated a dewetting transition that captured the emulsions, encapsulated them in a lipid bilayer, and formed GUVs. A membrane protein insertion assay and a fluorescence quenching assay were conducted to evaluate unilamellarity and asymmetry across the lipid bilayer in vesicles, respectively. Fabricated GUVs in the size range of 2–14 μm possessed unilamellar membranes (minimal residual oil remained trapped between the two lipid leaflets) and 83% asymmetry was achieved across the lipid bilayer. The synthetic asymmetric vesicles built using our strategy display properties physiologically more relevant to the natural cytoplasmic membranes (i.e., asymmetry and size range) and hold the potential to be used as drug delivery vehicles in biological systems.

SUPPLEMENTARY MATERIAL

See the supplementary material for the assembly of Lipid₁ under fluorescence microscopy after the separation step (Movie S1), the separated water-in-oil emulsions (green) traveling in the dark Oil + Lipid₂ stream with no fluorescently tagged lipids (Movie S2), and the hydrodynamic focusing region where we complete formation of the lipid bilayer in GUVs (Movie S3).

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding author upon reasonable request.