Nanomedicine is an attractive field involving the production and clinical application of size-controlled nanoparticles, usually employed in therapy as drug delivery systems or in diagnostics as contrast agents [1
]. Amongst the different types of nanoparticles, liposomes have attracted considerable interest for their application as drug delivery systems. Liposomes are artificial spherical vesicles generally composed of natural phospholipids, which performance depends on different physico-chemical variables including chemical composition, size and method of production [2
]. Different techniques have been developed for producing size-controlled liposomes with reproducible physical properties; the preparation methods can be generally divided into two groups: (i) “bulk” methods, comprising common macroscale or batch techniques; and (ii) microfluidic methods. The first group includes lipid film hydration, solvent (ethanol or ether) injection, and reverse-phase evaporation [3
In recent years, there has been a growing interest in the use of microfluidic-based production of liposome formulations. This approach has proven to be particularly effective, offering several advantages compared to macroscale techniques; these include small amount of reagents required, potential for in-situ analysis at high temporal and spatial resolution, devices’ portability and cost effectiveness [5
In microfluidics, the formation of liposomes is caused by the unfavorable interactions between lipids and water, causing the self-assembly of lipids (a process often defined as nanoprecipitation) to form spherical vesicles [7
]. In a typical microfluidic method, phospholipids are dissolved in a polar solvent (e.g., ethanol or isopropanol) and injected in the central channel of a microfluidic hydrodynamic focusing (MHF) device. The solvent is subsequently focused by water streams coming from two lateral channels [8
], leading to controlled mixing between chemical species.
Therefore, the formation of liposomes in microfluidic devices is often governed by diffusive mass transfer of compounds at the liquid interface between solvent (i.e., alcohol) and non-solvent (i.e., water). The alcohol, in which the lipids are initially solubilized, diffuses into the water until it decreases down to a critical concentration [9
]. The alcohol diffusion thus governs the formation of vesicles, by a mechanism described as “self-assembly”. Specifically, it has been postulated that the reduction of lipids’ solubility associated with water and alcohol diffusion across the fluid streams causes the formation of intermediate structures, in the form of oblate micelles, which finally enclose to form liposomes.
It is well known that microfluidic systems are characterized by steady laminar flow, which typically occurs when the Reynolds number is lower than a critical value of ~2000, due to the stronger contribution of viscous forces compared to inertial forces at the micrometer scale [10
]. The laminar flow regime has two main implications: (i) the flow in microchannels is typically characterized by parabolic fluid velocity profile; and (ii) the transfer of chemical species is dominated by diffusion, due to the low fluid velocity magnitude (resulting in low Péclet number) [11
The diffusion of chemicals (i.e., solvents, solutes and suspended particles) depends on the contact area between the fluids flowing in the microchannels. The diffusion coefficient scales approximately with the inverse of the molecular size (i.e., the hydrodynamic radius) and also depends, to some extent, on the shape of the molecule [12
]. Therefore, smaller molecules have higher diffusion coefficient and will move a longer average distance per unit time, compared to larger molecules that have a smaller diffusion coefficient.
On one hand the mixing of chemical species in microfluidic channels is therefore highly controllable (i.e., being governed by diffusion) and reproducible (i.e., due to the laminar flow conditions), but on the other hand, it is associated with low throughput and in some cases full mixing may not be achieved within the limited length typical of microfluidic devices.
Different methods for quantifying mixing in microfluidics have been presented; these are generally based on the acquisition of microscopic images of two or more colored or fluorescently labelled liquids, followed by quantification of mixing efficiency using simple mathematical functions. Examples of dyes employed are food dyes or stains for biological microscopy [13
], or fluorescent dyes such as fluorescein [14
Usually, mixing is quantified by processing a set of microscope images to yield a meaningful index, frequently defined as ‘mixing index’ that is representative of the extent of mixing. Different fluids are usually distinguished based on differences in the light intensity and spectral properties received by a charge-coupled device (CCD) camera.
A dye is often used to absorb transmitted light, reflect incoming light, or emit light. The mixing index is computed using intensities of pixels over a cross-section of a grayscale image that delineates a mixing event or region. The simplest index is calculated by taking the standard deviation of the pixel intensities. This method, however, may not be suitable for comparing mixing efficiency across different studies, from the moment that it is sensitive to different lighting conditions that may be difficult to standardize [16
One approach to measure the concentration of chemical species in microfluidic mixers is based on the use of fluorescent probes, where mixing is assessed from changes in the fluorescence intensity distribution along the device [17
]. Three-dimensional characterization of the mixing performance could be performed with these methods, by using confocal microscopes. Alternative techniques based on changes in the fluorescence lifetime of viscosity-sensitive molecular rotors have also been reported [18
]. They, however, require expensive equipment, including sensitive detectors, suitable microscope optics, and specific software/hardware, which hinders their adoption from the broader microfluidic and lab-on-a-chip community.
However, methods based on the use of a dye or fluorescent probe typically do not provide a direct quantification of the mixing between a solvent and a non-solvent, but rather a quantification of the transport of a selected dye or probe molecule. The physical and chemical properties of the probe may therefore have a significant impact on the measured mixing performance.
In this study, we describe and critically analyze a new method for studying mixing processes in different microfluidic chip architectures for nanoparticle production (i.e., MHF or Y-junction), which is based on the use of the pH indicator bromoxylenol blue (BB). The method provides a direct quantification of the exchange between solvent and non-solvent, and it relies on the color shift of a pH sensitive molecule, rather than on color or fluorescence intensity changes.
2. Experimental and Numerical Methods
Highly pure phosphatidylcholine (PC) 90% from soybean (Phospolipon 90G) was purchased from Lipoid GmbH (Ludwigshafen, Germany). Dimethyldioactdecylammoniumbromide (DDAB), bromoxylenol blue (BB), trichloro(1H,1H,2H,2H)-perfluorooctylsilane, hydrofluoric acid, and ammonium fluoride were purchased from Sigma-Aldrich Co. Ltd (Irvine, UK). Polydimethylsiloxane (PDMS) monomer Sylgard®184 and curing agent were purchased from Dow Corning Corporation (Auburn, AL, USA), and SU-8 photoresist from Chestech Ltd (Rugby, UK). All other regents and solvents were supplied by Sigma-Aldrich Co. Ltd (Irvine, UK). The water employed was ultrapure water (Merck Millipore, Billerica, MA, USA).
2.2. Microfluidic Devices Design and Fabrication
Two different microfluidic architectures were employed in the present study (see Figure 1
). #chip1-MHF is characterized by a cross flow geometry, in which the oblique side channels (length: 9.3 mm) intersect the central channel (length: 30 mm) with an angle of 40°. The channels have a rectangular cross section with a width of 0.15 mm and a depth of 0.10 mm. They were produced via soft lithography. Briefly, a SU-8 mold with the designed microchannel architecture was fabricated following a standard procedure [19
]. The mold was subsequently covered with a layer of a 10:1 (w
) polydimethylsiloxane (PDMS) monomer and curing agent liquid mixture, and heated for 1 hour at 80°. The solidified PDMS sheet, with the microchannel architecture on one surface, was then removed from the mold and permanently bonded to a glass slide after surface treatment with a plasma asher (PVA TePla AG, Wettenberg, Germany).
#chip2-YJ is characterized by a “Y” shape geometry in which the 2 inlets intersect with a 120° angle; the mixing channel (length: 66 mm) has a serpentine geometry. Channels have a squared cross-section with width and depth of 0.32 mm. The device is made of cycloolefin copolymer (COC) and was obtained from Thinxxs Microtechnology (Zweibrücken, Germany).
2.3. Liposome Preparation and Characterization
Liposomes were prepared using both #chip1-MHF and #chip2-YJ. The lipid mixture (containing PC 90G at 90 mM, and DDAB at 10 mM) was dissolved in ethanol and injected into the central inlet channel of #chip1-MHF or one inlet of #chip2-YJ; water was instead injected into the two side inlet channels of #chip1-MHF or the second inlet of #chip2-YJ. Teflon® tubes with an internal diameter of 750 µm (Sigma Aldrich, Irvine, UK) were employed to connect the inlets of the devices with syringe pumps (Pump Systems Inc., Farmingdale, PA, USA) for fluids’ delivery.
Liposome formation at different flow regimes was investigated by changing the flow rate ratio between water and ethanol (FRR) in the range 0.5–40, and the total flow rate (TFR) in the range 18.75–75.00 µL/min. The liposome samples were collected from the outlet tube (a 30 mm long Teflon® tube with an internal diameter of 750 µm) in a 1.5 mL microcentrifuge tube. Liposomes were analyzed for size and size distribution by DLS Zetasizer Nano-ZS (Malvern Instruments Ltd, Malvern, UK) with a backscattering detection angle of 173°, a He/Ne laser that emits at 633 nm, and a 4.0 mW power source. The data were used to report the intensity mean diameter (Z-average) and the dispersity of the liposome formulations. The mean particle size was obtained from the results of three independent experiments, carried out at 21 °C in water, without sample dilution (sample volume: 1 mL). Cryo-Transmission Electron Microscopy (cryo-TEM) images of liposomes were also acquired for morphological characterization. For this purpose, a 3 mL aliquot of a liposome sample was applied on plasma-treated (Gatan Solarus Model 950 Advanced Plasma System, pressure = 70 mTorr, H2 flow = 6.4 sccm, O2 flow = 27.5 sccm, forward RF target = 50 W, exposure time = 30 s) carbon copper grids (Quantifoil R 3.5/1), in the environmental chamber of a fully automated vitrification device for plunge freezing (FEI Vibrot). The relative air humidity was equal to 100% and temperature to 22 °C. The excess solution was removed by blotting with filter paper for 2 s, followed by 1 s draining and plunging of the samples into a 1:1 mixture of liquid ethane and liquid propane, which was cooled to 170 °C. Vitrified samples were cryo-transferred into a Jeol JEM3200FSC cryo-TEM, operating at 194 °C. The temperature of the samples was 187 °C during image acquisition. The microscope was operated in bright field mode, using a 300 kV acceleration voltage; the in-column energy filter was set to 0–20 eV energy-loss range (zero-loss imaging). Micrographs were recorded with a Gatan Ultrascan 4000 CCD camera.
2.4. Analysis of Mixing in Microfluidic Chips
The effect of FRR and TFR on the mixing of solvents and solutes in microfluidic channels were studied using the pH indicator BB and NaOH, which were added to the lipid solution and water respectively. BB was added to the ethanol lipid solution until saturation, after adjusting the pH by 0.1 M acetic acid; the concentration of NaOH in water was 0.1 N.
BB is a weak acid, and appears in yellow (below pH 6) or light blue (above pH 7.6) color when it is in the protonated or deprotonated state, respectively. It has a green color in the interval of pH comprised between 6 and 7.6, as an intermediate of the deprotonating mechanism in neutral solution. Therefore, the mixing between ethanol containing BB and water containing NaOH within the microfluidic devices, causes an increase in pH resulting in a change in BB color.
Different regions of the main channel within the two chips were imaged by an optical microscope (Hund® Wilovert 30, Helmut Hund GmbH, Wetzlar, Germany) equipped with a CCD camera (GXCAM-HICHROMESII, GT-Vision®, Haverhill, UK), at 4× magnification.
Images were taken nearby the junction between inlet channels, and at a more distal location along the main channel (in close proximity to the device outlet). The latter position was selected in order to provide a quantification of the overall mixing performance of the devices, at fixed flow dynamic boundary conditions.
Images were processed using ImageJ (NIH, Bethesda, MD, USA), to measure the width of the regions in which BB is either yellow, blue or green.
2.5. Numerical Simulation of Fluid and Species Transport
The transport of fluids and chemical species within both microfluidic devices was characterized numerically, using computational fluid dynamic (CFD) simulations. Firstly, the geometry of the microfluidic channels was designed using Inventor Pro 2016 (Autodesk Inc., San Rafael, CA, USA), and then transferred to ICEM CFD 17.0 (Ansys Inc., Concord, MA, USA) for meshing. The fluidic domain was discretized into finite volumes of tetrahedral shape. A mesh dependence study was performed to identify a compromise between solution accuracy and computational cost, leading to an optimal number of mesh volumes of 7′474′063 (#chip1-MHF) and 4′762′651 (#chip2-YJ). These corresponded to a mesh volume edge size of 0.012 mm (#chip1-MHF) and 0.03 mm (#chip2-YJ). Ansys® Fluent 17.0 (Ansys Inc., Concord, MA, USA) was employed to solve for mass and momentum conservation equations (i.e., Navier-Stokes equations at laminar flow regime), and species transfer (i.e., advection-diffusion equations). Boundary conditions were defined so as to replicate the experimental ones; a mass flow boundary condition was imposed at the device inlets, atmospheric pressure was imposed at the outlets, and a no-slip boundary condition was imposed at the channel walls. The experimental values of TFR and FRR were simulated numerically.
Fluids were assumed incompressible and Newtonian, and the ethanol-water diffusion coefficient was set to 1 × 10−9
]. The effect of solvents’ mixing on fluid density and viscosity was taken into consideration in the simulations. In order to compare the results of numerical simulations with the experimental images, the numerical contours of ethanol mass fraction were transferred to ImageJ for analysis. Stacks of RGB contour images at selected regions of interest (ROI) within the microfluidic devices were converted to 8-bit format, and subsequently thresholded to obtain a binary image. Reference lines were defined in agreement with the experimental image processing protocol, in order to obtain the width of fluid layers of specific relevance for characterizing the mixing process. The physical width of these layers was determined upon appropriate dimensional calibration of the images.
Microfluidic-based production of vesicular systems has proven to be an effective technique, offering several advantages compared to macroscale methods, particularly in terms of control over the physical properties of the end-product. These properties are usually highly dependent on the mixing between a solvent (i.e., ethanol) and a non-solvent (i.e., water). Thus, the design of a microfluidic architecture for vesicular systems’ production requires an in-depth characterization of the mixing within the microfluidic environment. In this study, we report on the development of a novel method based on the use of a pH indicator, and we demonstrated its utility by charactering the transport of solvent and non-solvent within two different microfluidic mixers typically used for the production of vesicular or micellar systems. Numerical simulations were performed to validate the experimental findings. With these methods, we evaluated the effect of the hydrodynamic boundary conditions (specifically the ratio between inlet flow rates, FRR) on the mixing performance of the selected microfluidic architectures, which had distinct geometrical and fluid dynamic characteristics. Our findings suggest that, in MHF devices, particular attention must be paid to the length of the main channel in order to achieve efficient mixing within the microfluidic device. The presence of a serpentine in the main channel was observed to significantly improve the mixing performance, and complete mixing was achieved for the large majority of FRRs and TFRs investigated. The latter device architecture may provide the benefit of efficient mixing at a larger spectrum of FRRs.
Compared to other mixing characterization methods based on changes in color or fluorescence intensity of a dye or probe, our proposed technique relies on the color-shift of a pH sensitive molecule, and may therefore be less sensitive to the lighting conditions employed in the experiment. Furthermore, it provides a route for qualifying and quantifying the solvent exchange process, which is postulated to govern the formation of vesicular systems in microfluidic devices.
The proposed mixing characterization method also presents advantages of cost-effectiveness and easiness of implementation in non-specialised laboratory settings, including those lacking in adequate computational facilities or expertise for performing numerical studies.