Continuous Organelle Separation in an Insulator-based Dielectrophoretic Device

Abstract

Heterogeneity in organelle size has been associated with devastating human maladies such as neurodegenerative diseases or cancer. Therefore, assessing the size-based sub-population of organelles is imperative to understand the biomolecular foundations of these diseases. Here, we demonstrated a ratchet migration mechanism using insulator-based dielectrophoresis in conjunction with a continuous flow component that allows the size-based separation of sub-micrometer particles. The ratchet mechanism was realized in a microfluidic device exhibiting an array of insulating posts, tailoring electrokinetic and dielectrophoretic transport. A numerical model was developed to elucidate the particle migration and the size-based separation in various conditions. Experimentally, the size-based separation of a mixture of polystyrene beads (0.28 and 0.88 μm) was accomplished demonstrating good agreement with the numerical model. Furthermore, the size-based separation of mitochondria was investigated using a mitochondria mixture isolated from HepG2 cells and HepG2 cells carrying the gene Mfn-1 knocked out, indicating distinct size-related migration behavior. With the presented continuous flow separation device larger amounts of fractionated organelles can be collected in the future allowing access to the biomolecular signature of mitochondria sub-populations differing in size.

1. INTRODUCTION

Resolving subpopulations of organelles is important as size-heterogeneity in organelles has been linked to specific diseases such as Alzheimer’s disease and cancer.[1–4] One of the most vital organelles is the mitochondrion as it is involved in the regulation of many biomolecular and cellular functions such as energy production and cellular signaling.[5,6] Mitochondria size and morphology is regulated by fusion and fission processes in response to the cell environment.[7,8] Distributions of normally sized mitochondria average ~ 600 nm, while abnormally large mitochondria are typically > 1 μm and small mitochondria < 300 nm.[9,10] Abnormally large mitochondria have been linked to diseases such as Parkinson disease,[11] acute lymphocytic leukemia,[12] and microvascular alterations in renal allografts.[13] Fractionating mitochondria populations by size could thus provide an important tool to assess the biomolecular differences in organelle subpopulations.

Different separation techniques have been developed to fractionate mitochondria by size. Centrifugation techniques such as differential centrifugation[14,15] and density gradient centrifugation[15–18] have been the preferred extraction methods for organelles. However, these methods do not allow for fractionation into subpopulations by size and cannot remove fragmented organelles of smaller size or larger aggregates.[19] Electrokinetic (EK) techniques such as free-flow fractionation[20] and capillary electrophoresis[21] were developed to alleviate sample contamination (e.g., lysosomes) during centrifugation. An effective tool that has been recently exploited for the separation of biomolecules and bioparticles is dielectrophoresis (DEP). DEP refers to the movement of a polarizable particle in a nonuniform electric field. DEP combined with the versatility of microfluidics has resulted in synergistic advantages allowing to exploit the separation potentials of DEP. The DEP force scales with the radius of the particle to the third power and is thus advantageously suited for size-based separations of biological particles. DEP has been applied to manipulate different analytes from nanometer to micrometer scale.[22,23] For example, the separation of bacteria,[24–26] DNA,[27–30] proteins,[31–34] and carbon nanotubes[35–37] has been reported in insulator-based dielectrophoresis (iDEP) devices. The manipulation of mitochondria has also been successfully demonstrated in an iDEP microfluidic device in the low-frequency regime.[38]

Achieving continuous flow separation is required to recover an ample volume of sample for analysis to elucidate the role that organelle heterogeneity plays in human biology and disease pathways. Different DEP-based microfluidic devices have been developed previously for bioparticle separation, such as reported by Wu et al. for yeast cells and polystyrene beads [39] as well as Tajik et al. for Saccharomyces cerevisiae.[40] Dielectrophoresis has also been combined with other separation techniques. For example, Jiang et al. developed a microfluidic device that combined hydrophoresis and dielectrophoresis to separate neural stem cells.[41] This device achieved a throughput of 2.4 × 10⁵ cells/h at 3.5 μL/min. A drawback of these devices is that microfabrication is more cumbersome and resource-intensive since electrodes need to be microfabricated within the microfluidic devices. Most importantly, none of these methods allows bioparticle separation in sub-populations by size.

Recently, ratchet migration has been demonstrated to be a feasible candidate for the size-based separation of mitochondria. In this migration mechanism, a complex periodic electrical potential is applied to a post array within a microdevice to induce separation due to various electrokinetic forces including DEP. A periodic electrokinetic driving force is applied to alter the overall electrokinetic migration direction which is superimposed with a DEP force that may lead to DEP trapping for larger particles. Once the condition of DEP trapping is met, the larger particles are withheld during defined times allowing them a net migration into one direction only. The smaller particles are transported based on electrokinesis in the alternating electric field. However, their migration direction is biased determined by a small but non-negligible direct current offset and they are not affected by DEP trapping forces. As a result, the larger particles experiencing sufficiently high DEP trapping forces can only be transported into the direction opposing the one of the smaller particles. Such approach has been demonstrated for sorting and separation of different particle sizes.[42–44] Ratchet migration can be tuned for size-selectivity by coupling different separation techniques such as electrophoresis (EP),[45] hydrodynamic flow,[46] and DEP.[47] In the past, we designed a ratchet device capable of separating particles in the sub-micrometer regime. We demonstrated the separation of 0.28 and 0.87 μm polystyrene beads, mitochondria, and liposomes using this ratchet migration.[48] The separation was achieved by applying a complex waveform that introduced electrokinetic and dielectrophoretic forces to induce a migration mechanism that differs based on the size of the particles. In this work, we developed a continuous flow microfluidic device to pave the way for size-selective separation of organelles based on the above-mentioned ratchet migration in sufficiently large amounts to subsequently characterize their protein content with techniques such as mass spectrometry.

Here, we present a continuous flow microfluidic device to separate polystyrene beads and mitochondria inducing ratchet and normal migration by applying a complex electrical waveform that induces simultaneous electrokinetic and dielectrophoretic forces overlaid by a pressure driven component. This device offers the ability of size-tuning by simply modifying the electrical parameters and takes advantage of steering particles into opposite directions with similar electrokinetic parameters. The size-based separation of 0.28 and 0.87 μm polystyrene beads was initially investigated as a validation of the device functionality. Furthermore, the ability of size-based separation of organelles was demonstrated with a mixture of Hep G2 mitochondria and from a knock-out (KO) HeLa cell line, where fusion was inhibited. A numerical model was adapted from our previous work to study the optimum electrical parameters in conjunction with the pressure-driven flow (PDF).[48] Our work provides a critical parameter study for continuous flow DEP separations and collection of organelles for biomolecular assessment.

2. Materials and Methods

2.1. Materials and Chemicals

Fluorescently-labeled polystyrene beads, size of 0.28 μm (FP-0262–2, purple fluorescent particles) and 0.87 μm diameters (FP-0852–2, yellow fluorescent particles), were purchased from Spherotech (Lake Forest, IL, USA). Polydimethylsiloxane (PDMS, SYLGARD 184 silicone elastomer kit) was purchased from Dow Corning Corp (Midland, MI, USA). 4-(2-Hydroxyethyl)- piperazine-1-ethanesulfonic acid (HEPES), poly(ethylene glycol)-block-poly(propylene glycol)-block-poly(ethylene glycol) (Pluronic F108), potassium hydroxide (KOH), dimethyl sulfoxide (DMSO), and sucrose were purchased from Sigma-Aldrich (St. Louis, MO, USA). Pure deionized (DI) water was collected from an Elga water purification system (Woodridge, USA). Mitotracker Red and Green were purchased from Life Technologies (Carlsbad, CA, USA). Microscope glass slides (75 mm × 25 mm × 1.0 mm) were purchased from Thermo Fisher Scientific Inc. (Waltham, MA, USA). Platinum wire (0.3 mm) was purchased from Alfa Aesar (Ward Hill, MA, USA).

2.2. Sample Preparation

A 10 mM HEPES buffer solution containing 250 mM of sucrose and 1 mM of F108 was prepared. The tri-block copolymer F108 reduced electroosmotic flow[49] through a dynamic coating procedure and avoids non-specific adsorption of polystyrene beads.[50] The pH was adjusted to 7.4 by 1 M of KOH with the conductivity of the solution resulting in 0.03 S/m. Prior to being used, the buffer solution was filtered by a 0.2 μm filter. A suspension polystyrene beads containing 0.1 % w/v of 0.88 μm and 0.03 % w/v 0.28 μm polystyrene beads was prepared by 10 s vortexing. The bead suspension was further diluted by a factor of 30, and the bead suspension was incubated overnight to ensure F108 coating on the particles.

Mitochondria were isolated from HepG2 cells and HepG2 cells with the gene Mfn-1 knocked out, via a procedure described in previous work without the tissue disruption and homogenization steps.[38] In this manuscript, we refer to these mitochondria as HepG2 mitochondria and Mfn-1 KO mitochondria, respectively. The isolated mitochondria samples were frozen and shipped from the University of Minnesota – Twin Cities on dry ice. Upon arrival at Arizona State University, the samples were stored in a liquid nitrogen Dewar. For the experiment, the mitochondria sample was transferred to a −80 °C freezer for one night and thawed in an ice-containing box for 30 min. Then, the buffer solution was replaced to exchange the mitochondria storage buffer with the HEPES buffer containing sucrose and F108. After replacing the buffer solution, the mitochondria isolated from HepG2 cells (HepG2 mitochondria) were stained with Mitotracker Green, and the mitochondria isolated the HepG2 cells with the gene Mfn-1 knocked out (Mfn-1 mitochondria) were stained with Mitotracker Red, as described next. Mitrotracker stock solution in DMSO was diluted to 1 mM Mitotracker and added to a mitochondria sample incubating at room temperature (25°C) for 1 hour with gentle shaking (120 rpm). After incubation, the stained mitochondria sample was centrifuged at 10,000g for 10 min, and the buffer was replaced by 1000 μL fresh HEPES buffer. This process was repeated three times to remove Mitotracker dye in the supernatant. Finally, the resulting pellet was suspended in 600 μL of buffer solution. Typically, in these preparations mitochondria remain morphologically stable for up to four hours.

2.3. Device Fabrication

The microfluidic device (see Figure 1) was designed via AutoCad (Ver. 2018, AutoDesk, Mill Valley, CA, USA), and the photomask was purchased through Advance Reproduction Corp. (North Andover, MA, USA). Next, the master mold was fabricated by a standard photolithographic process using a negative photoresist. Briefly, on a 4” silicon wafer (University Wafer, South Boston, MA, USA), an approx. 17 μm thick layer of photoresist (SU-8 2020, Kayaku Advanced Materials, Westborough, MA, USA) was spin-coated, which was then exposed to UV light through the photomask using a Suss MJB4 Mask Aligner (Suss MicroTech, Germany). After exposure, the photoresist was developed using SU-8 developer (Kayaku Advanced Materials, Westborough, MA, USA) and baked at 150 °C for 30 min prior to use.

Figure 1.

A Schematic of the ratchet device. The device has an inlet and two outlets for sample collection. Platinum wires were inserted in each outlet. The distance between the two outlets was 1 cm, the width of the channel is 2 mm, and the main channel contains an array of insulating posts. B SEM image shows the insulting posts in the main channel. The dimensions of the well-resolved posts are summarized next to the SEM image. C Waveforms applied for the electrokinetic components inducing ratchet migration. Waveform A switches applied ± U_{ac_1} with a DC potential offset, U_dc (here for U_dc> 0). Waveform B is a sinusoidal signal applied in every second half of the driving period of U_{ac_1} at 10 kHz. τ indicates the length of one complete driving period. D Electric field distribution obtained from the COMSOL model. The arrow points to the regions of low electric fields, where nDEP occurs for polystyrene beads and mitochondria. E-F Illustration of forces and the migration direction they induce during the first E and second half driving period F. G-J Exemplary particle trajectories in the two sections of the device. G-H top section and illustration of the particle migration in the first G and second H half driving period. Dashed lines indicate particle trajectories, and arrows indicate the direction migration based on the respective force. In the first half period of waveform A, both particles migrate towards the top electrode since the EOF and PDF overcome the EP force. In the second half driving period, waveform B is superimposed on waveform A. In this case, the large particle is trapped by the DEP force, and the small particle migrates towards the bottom section of the device. I-J On the bottom section of the device, the two particles migrate towards the top section of the device since the EOF is overcoming EP and PDF in the first half driving period. In the second half driving period, the large particles are trapped by DEP, and the small particles migrate towards the bottom, overcoming EP forces.

The PDMS was prepared by mixing base prepolymer and curing agent at a w/w ratio of 10:1 (base prepolymer: curing agent), which was poured on the master mold. Then, the PDMS was degassed using a desiccator and cured in an oven at 70°C overnight. Next, the cured PDMS was peeled off from the master mold, and 3 holes for the reservoirs were punched using Biopsy punches (Henry Schein, Phoenix, AZ, USA), one for the inlet (1.5 mm diameter), and the other two for outlets (3 mm diameter). The imprinted PDMS was cleaned with isopropanol (IPA) and dried in an oven at 70°C overnight. The cleaned PDMS piece and a glass slide (25mm × 75mm × 1mm) were bonded by treatment with oxygen plasma using a plasma cleaner (PDC-001; Harrick Plasma, Ithaca, NY, USA). Then, the channel was incubated with 10 mM HEPES buffer solution containing 1mM F108 and 250 mM sucrose, and the devices were stored in 100% humidity conditions prior to being used for experiments to prevent evaporation.

2.4. Device Set-up and Experimental Conditions

The common inlet was connected to a 50 μL syringe (Hamilton, NV, USA) via a silica capillary (Molex, Phoenix, AZ, USA), and the sample was injected into the device using a syringe pump (Harvard Apparatus, MA, USA) at 20 nL/min for an hour (for both PS particles and mitochondria), thus about 1.2 μL of sample was processed per run. The particle concentration was assessed by particle counting in a defined volume section of the device during the separation through video microscopy and resulted in ~2.7*10⁵ particles/μL). A platinum electrode was inserted in each outlet, and one electrode was connected to the ground, and the second was connected to the high voltage amplifier (AMT-3B20, Matsusada Precision Inc., Japan) using micro-clamps (LabSmith, Livermore, CA, USA). Additionally, in each outlet reservoir, 10 μL of HEPES buffer solution was added, and 5 μL of mineral oil was applied on top of the buffer solution to prevent the drying of outlet reservoirs.

2.5. Detection and Data Analysis

Fluorescence images were obtained using an inverted microscope (IX71, Olympus, Center Valley, PA, USA) equipped with a 100 W mercury burner (URFL-T, Olympus) with 40× (NA = 0.60) objective, an Optosplit element (Oxford Instruments, UK), and a multiband fluorescence filter set (exciter Brighline 468–553, dichroic FF493/574-Di-01, emitter Brightline 512–630, Semrock, USA). Videos were recorded by a CCD camera (QuantEM:512SC, Photometrics, Tucson, AZ, USA) using Micro-Manager software (version 1.4.7, Vale Lab, UCSF, CA, USA). The Optosplit enabled recording fluorescence signals of two different wavelengths simultaneously. Recorded videos of both mitochondria and beads were processed by ImageJ software[51] (version 1.53, NIH) and Mosaic particle tracker plugin[52] was used to trace the particle migration. Experiments at any particular electrical driving conditions were repeated three times, and at least 40 particles were tracked in each run for both PS beads and mitochondria, respectively. Then, the migration velocity of particles was calculated by an in-house script written in MATLAB (R2020, MA, USA).

2.6. Numerical Modeling

The numerical model was built using COMSOL Multiphysics 5.6 according to a model previously described by Kim et al.[48] further detailed in the Supporting Information. Briefly, in the model, electrokinetic (EK) migration (the sum of electroosmotic flow (EOF) and EP), pressure-driven flow (PDF), and DEP acting on various particles were simulated, and their movement was traced. The modules employed were the Electric Current, Creeping Flow, and Particle Tracing modules and used to simulate the same conditions as applied experimentally. The model geometry and electric field were scaled to a section of the ratchet device, and the electric field was adapted across the 1,000 μm long post array section. The electric field was calculated based on the periodic waveform signal, and the stationary value of the pressure-driven force from the applied flow rate was also calculated with the respective module. Then, the migration of particles was tracked in the Particle Tracing module based on the previously calculated forces and the DEP force which was applied on the particles in every second half period of the electrical waveform for 80 seconds (8 periods). A total of 120,000 particles (280 nm and 870 nm, respectively) was released for 60 seconds at a rate of 200 particles per 0.1 seconds.

3. Results and Discussion

3.1. Working principle of Migration Mechanism

We designed a continuous flow separation device based on ratchet migration in a post array under the combined action of electrokinetic, dielectrophoretic, and hydrodynamic flow. The microfluidic device (Figure 1A) contained a periodic array of insulating microposts (Figure 1B) similar to a previously reported ratchet device.[48] However, particles were introduced into the post array region via two inlets employing continuous pressure-driven flow (PDF). In addition to the PDF, the applied periodic waveform (see Figure 1C) induced electrokinetic migration and DEP. Large particles experiencing negative DEP (nDEP) were trapped at the flat surface of the microposts where the electric field was weakest (Figure 1D). nDEP occurs, when the Clausius-Mossotti (CM) factor is negative (see equation 1), which is the case for both polystyrene beads and mitochondria, as previously observed with similar medium composition.[38,48] On the contrary, smaller particles were not trapped because they were not experiencing sufficient nDEP. This ratchet separation principle was overlaid here with the PDF component, and we investigated suitable parameters for particle separation by size in continuous mode.

In Figure 1E–F, the working principle of the ratchet migration mechanism is explained based on the case of PS particles. During ratchet migration, the large particles experienced ratchet migration (assigned negative sign for migration velocity) and the smaller particles experienced normal migration (assigned positive sign for migration velocity) when subjected to the periodic driving force. This effect was induced through a periodic electrical driving force U_{ac_1} with defined U_dc offset, which was coupled to a higher frequency component in the second half period (U_{ac_2}), see Figure 1C, to induce DEP trapping and the migration direction reversal for large particles. Overlaid with the PDF component, the various forces on the particles were schematically outlined in Figure 1E–F in the two half-driving periods. While the PDF component was the same in each driving period, the switch of the sign in the electrical potential in the second half period reverses the electrokinetic driving forces. The EK force resulted from the sum of electroosmotic and electrophoretic contributions.

For the case of PS particles in the first half-driving period, the EOF was the stronger force dominating the migration direction, therefore, the net direction of the small and large particles was towards the positive electrode. In the second half of the electrical driving force U_{ac_1}, a DEP force was generated, which is sufficiently large to trap the larger particle species by nDEP. The DEP force for a spherical particle can be expressed with the equation 1[22,53,54]:

$${\text{F}}_{\text{DEP}}=2\pi r^3{\epsilon }_m\text{Re}\left[CM\right]\nabla {\left|\text{E}\right|}^2$$ (1)

where, r is the radius of the particle, ${\epsilon }_m$ is the permittivity of the medium surrounding the particle, Re[$CM$)] refers to the real part of the Clausius-Mossotti (CM) factor and $\text{E}$ is the electric field. Thus, the magnitude of DEP force acting on a small particle is weaker than that on a larger particle.

In combination with the PDF, the migration directions of two differently sized particles can now be described for each half driving period (Figure 1G–J). We first discuss the situation in the top section of the device (Figure 1G–H). In the first half-driving period, since EOF was the stronger force dominating the EK migration direction, the net direction of the small and large particle migration was toward the top outlet. PDF accelerates the migration of both particles toward the top outlet because the direction of PDF and EOF components were the same (Figure 1G). As the DEP force was generated in the second half driving period, the large particles were trapped at the insulating posts. However, the strength of DEP on the small particles was not sufficient to induce trapping, thus, the small particles traveled toward the bottom outlet. PDF decelerates the migration of small particles because the migration direction is opposite to the overall electrokinetic contribution, but the strength of PDF was weaker than that of the EK force (Figure 1H). Overall, the migration of the large particles in the top section resulted in migration towards the top outlet, whereas the migration of the small particles resulted in the opposite direction. The velocities at which these processes taking place strongly depend on the applied flow conditions and the magnitude of the applied potentials. We assessed several conditions in this manuscript carefully, as demonstrated below.

In the bottom section of the device (Figure 1I–J), PDF and overall EK migration direction were opposite in the first half-driving period. As a result, both particles were migrating toward the top outlet (Figure 1I). In the second half driving period, the large particles were trapped at the microposts due to nDEP, and the small particles were migrating toward the bottom outlet because the overall EK migration and PDF were in the same direction (Figure 1J). Again, the magnitude of the resulting migration velocities strongly depends on the interplay of PDF with EK components, which was assessed below. Since the ultimate goal was to steer particles under continuous flow conditions into opposing directions, we also followed the sign convention introduced above. The migration direction of the small particles was denoted as normal migration because the small particles were not affected by the ratchet migration. The goal was to collect them at the bottom outlet and a positive sign was assigned to the normal migration velocity. On the contrary, a negative sign was assigned to the ratchet migration which is opposite to the normal migration and should occur for the larger particles, which we aimed to collect at the top outlet.

3.2. Numerical Model and Migration Characteristics

The numerical model allows to assess the migration mechanism quantitatively and compare observed experimental particle migration with the theoretical model. To build the model, a geometry mimicking the post array in the main channel was designed, however, with reduced length resulting in suitable modeling times (Figure 2). The dimension of the microposts was adapted to the experimental cross section and distances of the posts in the 2D model. Similar to the experimental set-up, the potential was applied through electrodes at the top and the bottom outlets. The particles were released at a line in each inlet channel near the main channel (Figure 2A). The conditions simulated in the numerical model were equivalent to the experimental conditions: U_{ac_1} = 70, 80, 90V and U_dc = 10, 20, 30 V and U_{ac_2} = 1200 V at 10kHz. Further details of the model parameters were summarized in the Supporting Information. Figure 2B shows a snapshot of the particle distribution for U_{ac_1} = 70 V, U_dc= 20 V, U_{ac_2} = 1200 V at 10 kHz after 6 driving periods (60 s) for both 870 nm and 280 nm particles exemplarily.

Figure 2.

A Illustration of the geometry used in the numerical model. The distance between the two outlets was 1000 μm and the channel width was 500 μm. B Snapshot of particle migration at t= 60 s. The small particles (280 nm) are displayed as blue dots and the large particles (870 nm) as red dots. The simulated conditions were U_{ac_1} = 70V, U_dc = 20V, and U_{ac_2} = 1200V at 1200Hz and a pressure driven flow of 20 nL/min.

To count the number of particles migrating into the top or bottom section of the device, the main channel was divided into sections of 40 μm length (see Figure 3A). Since the particles were transported from the release line to the center of the main channel from two opposing inlets, the particles in the center of the device were neglected (90 μm from the center in each direction). The number of particles in the top and bottom section of the device were assessed separately within the 40 μm sections outside of the inlet section. Figure 3B summarized the result of particle counting for U_{ac_1} = 80 V, U_dc = 10 V, U_{ac_2} = 1200 V at 10 kHz and 20 nL/min PDF after 8 driving periods (= 80 s). As observed from this figure, the larger particles (0.87 μm) were primarily distributed in the top section of the main channel, whereas the smaller particles (0.28 μm) were concentrated at the bottom section. After the 8 driving periods, a negligible quantity of the smaller particles was found in the top outlet, and none of the large particles were found in the bottom outlet, indicating successful separation of the two particle species. This behavior of migrating particles is also observed in Supplementary Video 1 for U_{ac_1} = 90 V, but otherwise similar conditions. We note that the number of particles in the large and small case varies, based on the differences in their average migration velocities, as outlined below in Figure 5. Moreover, some particles get lost during the simulation, which reduced the total number of particles for each particle size below the number released. The loss of particles can be accounted for and does not influence the outcome of this numerical particle study negatively.

Figure 3.

Particle counting with the numerical model. A Device with vertical lines, showing slicing in 40 μm sections for particle counting. The area from −90 μm to +90 μm was excluded in the particle counting as no substantial separation is occurring in this region. In total, 120000 particles (0.28 μm and 0.87 μm, respectively) were released for each case. B Result of particle counting for U_{ac_1=} 80 V, U_dc= 10 V, U_{ac_2}= 1200V at 10kHz, and 20 nL/min PDF operated for 8 periods (80s) is demonstrated.

Figure 5.

Velocity of 0.28 and 0.88 μm diameter polystyrene beads investigated experimentally and as obtained from the numerical model. A-C large particles assessed in the top portion of the device: A velocities assessed at U_{ac_1} = 70 V, U_dc varying from 10 to 30 V, B velocities assessed at U_{ac_1} = 80 V, U_dc varying from 10 to 30 V, C velocities assessed at U_{ac_1} = 90 V, U_dc varying from 10 to 30 V. D-F small particles assessed in the bottom portion of the device: D-F same conditions as A-C. For all conditions, U_{ac_2} was 1200 V at 10 kHz and the driving period 10 s. At least 40 particles were traced in each case. Note that velocities for the large particles exhibit negative signs corresponding to ratchet-based migration and small particles exhibit positive velocities. The opposing sign also indicated opposing migration directions and thus separation of the two particle species.

To assess the fractionation capability of the device, the recovery efficiency ($\varphi$) for the small and large particles was calculated according to equations 2 and 3:

$${\varphi }_{\mathit{\text{top}},s/l}=\frac{N_{\mathit{\text{top}},s/l}}{N_{\mathit{\text{top}},s/l}+N_{\mathit{\text{bottom}},s/l}+N_{\mathit{\text{center}},s/l}}$$ (2)

$${\varphi }_{\mathit{\text{bottom}},s/l}=\frac{N_{\mathit{\text{bottom}},s/l}}{N_{\mathit{\text{top}},s/l}+N_{\mathit{\text{bottom}},s/l}+N_{\mathit{\text{center}},s/l}}$$ (3)

where $N_{\mathit{\text{top}},s/l}$ refers to the number of either small, $s$, or large, $l$, particles in the top section of the device. Similarly, $N_{\mathit{\text{bottom}},s/l}$ accounts for the number of particles (either small or large particles) in the bottom section, and $N_{\mathit{\text{center}},s/l}$ is the number of either small or large particles in the center section of the device. The center fraction contains particles that have not yet experienced significant separation, as they just left the inlets. We investigated a series of DC offsets for various driving conditions, as summarized in Figure 4. Particles in the entire bottom and top section were counted and summed.

Figure 4.

Overview of particle counting results based on the numerical model for different electrical parameters. The model was conducted for 8 driving periods in each case.

Figure 4 allows to ascribe a trend in particle migration caused by the change of DC offset, while the magnitude of U_{ac_1} only marginally affects the recovery efficiency. An increase in DC offset increased the number of both large and small particles found at the bottom section, while the trend was stronger for the larger particles, likely because of EK forces counteracting DEP trapping in the second half period. In the top section, an increased DC offset resulted in a decrease of recovery efficiency, likely also due to the EK forces counteracting DEP trapping in the second half driving period, which is necessary to induce the opposite migration of larger particles according to the ratchet mechanism. We note that the goal would be to selectively migrate only small particles to the bottom, and only large particles to the top.

To assess the separation behavior of the two-particle mixture further, the selectivity (S) of the particle separation was calculated by equation 4 and 5:

$$S_{\mathit{\text{top}},l}=\frac{{\varphi }_{\mathit{\text{top}},l}}{{\varphi }_{\mathit{\text{top}},l}+{\varphi }_{\mathit{\text{bottom}},l}}$$ (4)

$$S_{\mathit{\text{bottom}},s}=\frac{{\varphi }_{\mathit{\text{bottom}},s}}{{\varphi }_{\mathit{\text{top}},s}+{\varphi }_{\mathit{\text{bottom}},s}}$$ (5)

The resulting selectivity for the top and bottom sections, $S_{\mathit{\text{top}},l}$ for large particles and $S_{\mathit{\text{bottom}},s}$ for small particles, are shown in Table 1. It is apparent that an increase of U_dc increased the selectivity of large particles in the top section and decreased the selectivity of small particles in the bottom section. The latter can be explained in the following way. The augmented U_dc induced stronger electrokinetic migration toward the bottom outlet, hence, fewer small particles were observed in the top section.

Table 1.

Selectivity, $S$, for all conditions investigated in Figure 4.

	${\text{U}}_{\text{dc}}$
	10 V	20 V	30 V	10 V	20 V	30 V
${\text{U}}_{\text{ac}\_\mathbf{1}}$	Top (large)			Bottom (small)
70 V	0.96	0.57	0.55	0.75	0.96	1.00
80 V	1.00	0.48	0.34	0.70	1.00	1.00
90 V	0.87	0.72	0.33	0.92	0.87	0.98

For an optimized fractionation experiment, both $S_{\mathit{\text{top}},l}$ and $S_{\mathit{\text{bottom}},s}$ are equal to unity. Consequently, we define the product of $S_{\mathit{\text{top}},l}$ and $S_{\mathit{\text{bottom}},s}$ as the separation efficiency, P, which again, should result in a value of 1 for optimized separation conditions:

$$P=S_{\mathit{\text{top}},l}*S_{\mathit{\text{bottom}},s}$$ (6)

Table 2 list the resultant P values and shows that the highest P, 0.80, was reached at U_{ac_1} = 90 V and U_dc = 10 V. The interplay of all migration effects based on PDF and electrokinetic phenomena is a complex situation that is difficult to assess experimentally. The theoretical treatment of the separation of the two-particle system via the numerical model thus allows to assess separation conditions in a quantitative approach.

Table 2.

Separation efficiency, P, for all investigated U_{ac_1} and U_dc cases.

${\text{U}}_{\text{ac}\_\mathbf{1}}$	${\text{U}}_{\text{dc}}$
	10 V	20 V	30 V
70 V	0.72	0.55	0.54
80 V	0.70	0.48	0.34
90 V	0.80	0.63	0.32

3.3. Experimental Observation of the Migration Mechanism with Model Particles

We further assessed the observed migration velocities from the numerical model and compared it to experimentally obtained particle migration under identical driving conditions. Figure 5 compared the average velocities for 0.28 μm and 0.87 μm polystyrene beads as assessed through particle tracking. The velocities were studied for U_{ac_1} amplitude from 70 V to 90 V and varying DC offsets (U_dc = 10 – 30 V). Supplementary Video S-2 and S-3 show corresponding video sequences for large and small particles stained with different fluorophores applying simultaneous two-color fluorescence detection. Figure 5A–C shows the average velocity of large particles assessed in the top section of the device. Negative velocities were apparent for the large particles indicating ratchet migration. The numerical model supports the ratchet migration of the large particles on the top section of the device in good agreement with experimental migration velocities showing similar trends. It is noted that the increase of U_{ac_1} had a negligible effect on the migration velocity for the 0.87 μm particles presumably due to a negligible increase of only 20 V. However, the increase in U_dc decreases the amplitude and decelerates the migration of the particles towards the top outlet of the device. This trend is apparent both in the model and experiment.

Figure 5D–F shows the average velocities of the smaller 0.28 μm PS beads at different U_{ac_1} and U_dc assessed on the bottom section of the device. Similar trends in the variation of U_dc are also apparent in the model and experiment. In contrary to the large beads in the top section, the migration velocities now increase with increasing U_dc. This behavior is caused by the PDF accelerating the normal, EK migration direction in the bottom region of the device. In addition, the average velocities at U_{ac_1} = 70 V showed excellent agreement with the numerical model, whereas the experimental values at U_{ac_1} = 80–90 V deviated slightly from the numerical model. The discrepancy between model and experiment for the here demonstrated conditions are attributed to deviations in the magnitude of the PDF in experiments, which – over the course of an experiment – might have deviated from 20 nL/min due to evaporation effects and fluctuations in the applied flow rates.

3.4. Mitochondria Migration

Next, we assessed the migration behavior of mitochondria in the continuous flow separation device. We employed two different mitochondria samples exhibiting different size distributions. The size of the mitochondria isolated from HepG2 cells was assessed with DLS measurement to exhibit a mean organelle size of 1280 ± 210 nm. The size of the mitochondria isolated from Mfn-1 knockout cells was smaller since the gene knockout prevented fusion of mitochondria, resulting in an average size of 610 ± 53 nm. The average size of the mitochondria mixture (HepG2 mitochondria + Mfn-1 mitochondria) before the separation was measured to be 714 ± 183 nm which was injected into the ratchet device. The mitochondria were stained with different Mitotracker fluorescent dyes (HepG2 mitochondria were stained with Mitotracker Green, and Mfn-1 mitochondria were stained with Mitotracker Red) to distinguish them during the migration studies based on fluorescence video microscopy. Example videos can be found in Supplementary videos S-4 and S-5. Figure 6A demonstrates the outcome of the DLS measurements after staining each mitochondria sample and the corresponding size distribution of the mixture.

Figure 6.

The DSL measurement of the size of mitochondria sample, and the experimental assessment of mitochondria migration velocities. A DLS measurements of mean size of mitochondria before separation. The mean size of HepG2 mitochondria was 1280 ± 210 nm, the mean size of the Mfn-1 mitochondria was 610 ± 53 nm, and the mean size of the mixture of these two mitochondria was 714 ± 183 nm. B The velocity of HepG2 mitochondria was measured at the bottom section of the device. C The velocity of Mfn-1 mitochondria was measured at the top section. The migration direction of mitochondria was opposite to the PS particles, as a result, the ratchet migration was observed at the bottom side, and the negative sign was assigned. On the contrary, the positive sign was assigned to the velocity of mitochondria on the top section. D DLS measurement of mean size of mitochondria after separation. The separation condition was U_{ac_1}= 80 V, U_dc = 10V and U_{ac_2} = 1200 V at 10 kHz with flow rate of 20 nL/min applied for one hour. The mean size of the mitochondria collected from the top outlet after 1 h of separation was 640 ± 347 nm. The mean size of the mitochondria collected from the bottom outlet was 857 ± 201 nm.

The direction of the EK force inducing migration of mitochondria was observed to be opposite to the EK migration of PS particles. In the first half driving period - unlike polystyrene particles - mitochondria migrated toward the bottom outlet. The wall coating and the buffer composition were the same for both mitochondria and polystyrene particles; thus, the strength of EOF should be equal in both cases. We assume that the reversal in migration direction is due to a higher electrophoretic contribution in the case of mitochondria counteracting EOF. To confirm this, the Zeta potential of polystyrene beads and mitochondria was assessed by Zeta potential measurements (see supplementary Table S3). The Zeta potential of mitochondria was found to be 2-fold higher than that of polystyrene particles which implies that the surface charge on mitochondria surface was more negative. This difference caused different strengths of EP that resulted in the opposite migration direction in the mitochondria case.

Despite the overall reversal of migration directions in the case of mitochondria, the ratchet migration of the larger mitochondria in the second half driving period was still observed. The experimentally determined migration velocities for HepG2 and Mfn-1 mitochondria were summarized in Figure 6B–C. The studied conditions were U_{ac_1} = 70, 80, 90V and U_dc = 10, 20, 30 V and U_{ac_2} = 1200 V at 10kHz. We note that these velocities were assessed in a mixture of the two mitochondria species, each stained with a different fluorophore. According to Figure 6B–C, the effect of U_{ac_1} on the migration velocity of the mitochondria was negligible. Similar to the result of polystyrene particles, the increase in DC offset decelerated the migration of HepG2 mitochondria (larger) and accelerated the migration of Mfn-1 mitochondria (smaller). HepG2 mitochondria had a higher tendency to migrate towards the bottom outlet (ratchet migration), and Mfn-1 mitochondria mainly migrated towards the top outlet (normal migration).

Moreover, the observation from Figure 6D showed a distinguishable size difference between the samples collected at the top and the bottom outlet compared to the initial mixture (Figure 6A). Fractions of separated mitochondria were collected at each outlet, and the size was measured using DLS after 1 h of separation. The separation condition was U_{ac_1} = 80 V, U_dc = 10V and U_{ac_2} = 1200 V at 10 kHz with a flow rate of 20 nL/min applied to the inlets. The mean size of the mitochondria collected from the top outlet was 640 ± 347 nm, and the mean size of the mitochondria collected from the bottom outlet was 857 ± 201 nm. This showed that the larger mitochondria were preferably collected at the bottom outlet, and smaller mitochondria were preferably collected at the top outlet. However, a wider range of size distributions was observed from the fractions collected from both outlets, especially samples collected from the bottom outlet.

In both collected fractions, very large particles (>2 μm) were found in the DLS measurement. This could result from the aggregation of mitochondria during the experiment or sample preparation. In the case of the fraction collected at the top, DLS also indicated some smaller < 100 nm species, which we attribute to impurities originating from cell fragments of the original sample or damaged mitochondria. Based on the processing times (1h) and the particle concentrations with which the separation was carried out (2.7*10⁵ particles per μL), we estimate that ~3*10⁵ particles were processed. This is close to the value reported by Jiang et al.[41] with a continuous flow DEP technique, however, does not allow to access organelle sub-populations based on size. In summary, the applied driving conditions allowed for size-based selectivity in the continuous flow separation as observed by the difference in size distribution of the two samples collected from opposite ends of the device.

4. Concluding Remarks

In this work, a continuous flow DEP-based ratchet device was designed and thoroughly investigated for the continuous flow separation of polystyrene beads and mitochondria of varying size at flow rates of 20 nL/min. The numerical model provided a good understanding of the polystyrene particle migration in the microfluidic device and the effect of the driving conditions on the migration velocities and directions. With the numerical model, the separation efficiency was quantitatively assessed to find the best separation conditions for the non-intuitive and complex migration phenomenon. The migration velocity of a heterogeneous polystyrene bead mixture (0.87 and 0.28 μm) was also investigated experimentally and was found to be in good agreement with the numerical model. Due to the continuous nature of the separation, we defined the separation selectivity and separation efficiency, which we found to be largest at driving conditions of U_{ac_1} = 90 V and U_dc = 10 V through modelling.

In addition, the separation of a size-heterogeneous mixture of mitochondria isolated from HepG2 cells with the Mfn-1 gene knockout and HepG2 cells also showed that the migration direction under the complex applied driving parameters resulted in size-based separation. A distinct difference in the mean size of mitochondria collected at the top and bottom outlet was observed after separating a mixture of Mfn-1 and HepG2 mitochondria. Overall, the here demonstrated device and driving parameters demonstrated that size-based particle and organelle separation was achieved under continuous flow conditions concomitantly with ratchet migration. We note that while several protocols and methods for purification of mitochondria were described and optimized in the past, however, there is no single method allowing mitochondria fractionation into subpopulations based on size.[55] While we have previously demonstrated the size-based separation of mitochondria with a ratchet migration mechanism, ⁵³ the here presented continuous fractionation into distinct mitochondria sub-populations is an important step towards further assessing the molecular content of normal and enlarges mitochondria. To the best of our knowledge, we provide the first evidence of mitochondria fractionation by size with this continuous flow method, which has the potential to be scaled up for higher throughput through device design and dimension alterations. For example, the sample collection from the outlets can be further optimized to collect larger sample volumes needed for subsequent biomolecular mitochondria assessment. Moreover, the active device component (the post array region) can be increased and overall the device can be designed to process larger volumes, even on a microchip. The work demonstrated here represents a new approach for continuous size-based fractionation of diagnostically relevant organelle subpopulations.

Data Availability Statement

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