High-throughput dynamical analysis of dielectrophoretic frequency dispersion of single-cells based on deflected flow streamlines

Abstract

Phenotypic quantification of cells based on their plasma membrane capacitance and cytoplasmic conductivity, as determined by their dielectrophoretic frequency dispersion, is often used as a marker for their biological function. However, due to the prevalence of phenotypic heterogeneity in many biological systems of interest, there is a need for methods capable of determining the dielectrophoretic dispersion of single-cells at high throughput and without the need for sample dilution. We present a microfluidic device methodology wherein localized constrictions in the microchannel are used to enhance the field delivered by adjoining planar electrodes, so that the dielectrophoresis level and direction on flow focused cells can be determined on each traversing cell in a high-throughput manner based on their deflected flow streamlines. Using a sample of human red blood cells diluted to 2.25 × 10⁸ cells/mL, the dielectrophoretic translation of single cells traversing at a flow rate of 1.68 μL/min is measured at a throughput of 1.1x10⁵ cells/min, to distinguish positive versus negative dielectrophoresis and determine their crossover frequency in media of differing conductivity for validation of the computed membrane capacitance to that from prior methods. We envision application of this dynamic dielectrophoresis (Dy-DEP) method towards high throughput measurement of the dielectric dispersion of single-cells to stratify phenotypic heterogeneity of a particular sample based on their DEP crossover frequency, without the need for significant sample dilution.

1. Introduction

Cellular systems often exhibit a degree of phenotypic heterogeneity [1] that has important implications on biological function and disease response [2]. Currently, phenotypic heterogeneity is quantified using fluorescent-activated flow cytometry methods. While this method is highly specific due to binding of cell receptors to fluorescently labeled antibodies and it gives multi-dimensional data on cell phenotypes [3], some drawbacks include its need for costly labeling steps, sample dilution, skilled technicians and rather sophisticated instrumentation. Furthermore, cell receptors are not often well-defined for various types of tumor [4] and stem cells [5]. Finally, since flow cytometry functions as an endpoint assay, it cannot be used repeatedly to analyze the same set of cells for kinetic monitoring of cell phenotype under different interventions [6]. Hence, there is much interest in alternate methods for biophysical analysis of single-cells, in a label-free manner based on their inherent properties [7].

Cell membrane capacitance is a label-free phenotype that can serve as a specific metric to identify cells based on their size and morphological characteristics [8]. Recent work has shown that membrane capacitance can serve as a marker to stratify vesicles based on their lipids [9], determine the viability of bacteria [10] and their adherence ability to host cells [11], identify parasite-infected red blood cells (RBCs) [12], quantify the morphological state of tumor cells in their adherent state [13] and predict the lineage of neural stem cells [14]. A common way to measure membrane capacitance is based on determining the dielectrophoretic crossover frequency of cells within media of varying conductivity [15], [16], [17]. For this purpose, the translation of polarized cells under a spatially non-uniform electric field is followed to determine the frequency at which the cells transition from negative dielectrophoresis (nDEP) or translation against the field gradient due to field screening by the cell, to positive dielectrophoresis (pDEP) or translation along the field gradient due to field termination at the cell [18].

DEP crossover frequency measurements are often carried out in a batch-mode using quadrupole or castellated electrode configurations [19], due to their well-defined field gradient direction owing to distinct regions of high field and low field within these device structures. However, for the purpose of effectively quantifying phenotypic heterogeneity, there is a need to measure a large number of events (10⁴-10⁶ cells) within a short time (< 1 h). Hence, there is a need for continuous-flow device configurations capable of rapidly detecting field screening on single-cells, as they flow past regions of field non-uniformity at high flow rate. A major limitation in this regard is that since field non-uniformities are usually highly localized (i.e. on the order of magnitude of cell size) for the purpose of enhancing DEP translation, due to its ∇E² dependence, the level of DEP translation falls off sharply within a few microns away from the field non-uniformity. Hence, a significant proportion of cells within the device do not often experience large enough alterations in translation for enabling facile distinction of the DEP cross frequency. Furthermore, even if the field non-uniformity were to be enhanced based on sharp features and/or enhanced voltage levels, the time period available for translating cells under DEP can drop off as the flow rate of cells through the device rises. Finally, DEP analysis of higher cell concentration levels has been limited by dipole-induced cell-cell interactions. As a result, DEP analysis has often been limited to relatively low throughput levels (well below 1 μL/min) and low cell number rates (<10³ cells/min), which are often not sufficient to quantify phenotypic heterogeneity with statistical certainty.

Various flow through configurations for DEP analysis and separations have been used to assess cells based on their crossover frequency, including various methods based on planar electrodes, such as cell-levitation by DEP field flow fractionation (DEP-FFF) [20] and cell-deflection based on gradients in media conductivity [21]. In order to address the problem that planar electrodes have a limited spatial extent of field non-uniformity over the device depth, various sidewall electrode configurations have been developed [22], but these are often difficult to fabricate reproducibly. Easier fabrication strategies for sidewall electrodes has been demonstrated in recent work by filling PDMS channels with conducting composites [23], so that hydrodynamic focusing can be used to place cells in the vicinity of the non-uniformity. However, filling of dead-end PDMS channels with conducting composites can cause poor definition of the metal interface to the fluidic channel [24], whereas strategies based on lead-in channels with an inlet and outlet that better define this interface can limit the spacing between the DEP electrodes to > 100 μm, thereby reducing the field levels. Alternatively, there are various electrode-less strategies based on the field non-uniformity created by insulators, with the field applied by global external electrodes [25], [26], [27]. However, this configuration restricts the level of applied field, especially at high MHz frequencies, wherein voltage amplifiers exhibit losses [28]. Additionally, since only one set of electrodes can be used per channel in this configuration, the lack of voltage addressability limits the ability to spatially modulate fields for carrying out electrically-independent downstream separations.

To address these deficiencies, we present a device configuration combining 3D insulator constrictions with a set of addressable planar electrodes so that the net spatial extent of the field non-uniformity exceeds that of a configuration with 3D electrodes extending over the entire channel depth, which is difficult to fabricate. As a result, hydrodynamically focused cells traversing at high flow rates (>1 μL/min) over a range of streamlines in the vicinity of the field non-uniformity can be deflected by dielectrophoresis at high cell number rates (~10⁶ cells/min). This study is focused on presenting the field profiles and particle tracing simulations for this so-called dynamic-DEP (Dy-DEP) device configuration to enable its comparison to the equivalent device with 3D electrodes, as well as present its application towards determining the crossover frequency of red blood cells (RBCs) based on spatially distinct streamlines for nDEP, pDEP and no DEP over a range of media conductivities. Based on independent validation of the determined membrane capacitance of RBCs for device operation at high flow rates and high cell number rates, we envision the application of this device configuration in future work for the purpose quantifying phenotypic heterogeneity.

2. Device Design and Operation Principle

Figure 1 shows a schematic of the overall microfluidic Dy-DEP device and its operating principle for separating cells into differing streamlines based on magnitude and direction of DEP response. Per Fig. 1a, our overall objective is to develop a device that can deflect each cell traversing the field non-uniformity regions created by consecutive insulator constriction tips, due to a balance of dielectrophoretic trapping force (F_nDEP away from tip or F_pDEP towards the tip) versus the drag force (F_drag), so that cells can be separated along differing streamlines based on their dielectrophoresis levels (Fig. 1b). For this purpose, the sample with cells is focused along the streamlines close to the channel wall using a sheathing flow of much higher flow rate (3x of sample flow rate), so that each cell in the sample has the opportunity to interact with the high field points at the constriction tips. Under this spatial field non-uniformity, cells experience dielectrophoretic translation based on a magnitude and direction that depends on the frequency dispersion of their polarization response versus that of the surrounding media. In the situation wherein F_DEP (pDEP or nDEP) just exceeds F_drag, the cells undergo translation across flow streamlines to continue along the particular streamline wherein the net DEP and drag force are equal. While Fig. 1d shows the schematic for separation of streamlines based on nDEP level, our subsequent results demonstrate separation of the cell streamlines based on their DEP behavior; i.e. pDEP, no DEP at crossover frequency (f_xo) and nDEP level, while the high net flow rate of the cells (sample plus sheathing flow rate of 1.68 μL/min) ensures continuous particle deflection with no DEP trapping across the length of the device. Table 1 lists the distinguishing characteristics of the current study versus prior work. While the prior work has been focused on engineering separations using dilute cell samples operated at a low enough flow rate to ensure a significant time period for action of the DEP force, our work is focused on high throughput cell analysis to determine frequency and media conductivity ranges for different levels of nDEP, no DEP at crossover, and pDEP, due to cell deflection across streamlines to characteristic positions based on their DEP behavior. The reported device is validated using red blood cells (RBCs) obtained from human blood samples diluted to concentration levels of 2.25 × 10⁸ cells/mL for demonstrating DEP analysis at a throughput of 1.1x10⁵ cells/min, so that the determined spectra and membrane capacitance can be compared to prior work. Hence, the device is particularly suited for the purpose of high throughput characterization of the DEP dispersion behavior to stratify phenotypic heterogeneity of a particular sample based on their DEP crossover frequency, without the need for significant dilution.

Figure 1.

Schematic of microfluidic device for dynamic dielectrophoresis (Dy-DEP): (a) Functioning principle based on balance of nDEP versus drag forces; (b) overall chip design; (c) focusing effect of the sheath flow pushes cells in the sample away from electrodes and towards the constriction regions of the device; (d) example differences in fluid flow streamlines of cell types with differing DEP response.

Table 1.

Distinguishing characteristics of device in current study versus prior work

Citation	Device	Sample Type (DEP type)	Initial Cone. (# per mL) †	Flow rate (μl/min)*	Frequency	Media σm (S/m) ††	Throughput (cells/min)
[29]	X pattern insulating structure with 3 types of electrodes: planar, dual-planar and 3D electrodes	Live vs. dead HeLa cells (pDEP level)	1 × 107	0.54	1 kHz	0.00176	5×103–4×104
[21]	Diagonal top-bottom planar electrodes in channel under a conductivity gradient	Live vs. dead yeast cells (crossover)	5 × 106	3	100kHz-10MHz	0.0093-0.047	1.5×104
[30]	Planar slanted electrodes with sheathing flow to focus sample on sidewalls	Platelets from diluted whole blood (nDEP level)	5 × 108	2.5	1 MHz	0.05	1.32×106
[31]	Planar electrodes with asymmetric orifices to generate the non-uniform field on sheath flow focused sample	Live vs. dead yeast cells (pDEP level)	Not reported	0.225	1kHz-10MHz	5.5 × 10−6	Not reported
[20]	DEP-FFF in channel with planar interdigitated electrodes	Various types of cells (nDEP level)	1 × 106	20 (diluted sample)	5-60 kHz	0.01-0.05	104 – 105
[23], [24]	3D AgPDMS electrodes along channel with sheath flow focusing of sample	Live vs. dead yeast cells, as well as beads (pDEP level)	107-106	0.1 (S) + 0.9 (F)§	0.1-1 MHz	0.02-0.05	Up to 103§ (using highest sample level)
[32]	Set of recessed planar electrodes with sheath flow focusing of sample	Platelets from RBC & WBC (pDEP level)	~108	0.02 (S) + 0.08 (F)§	0.1 MHz	5 × 10−4	<2.5 × 103§ (using highest sample level)
[33]	3D ionic electrodes create funnel shaped field non-uniformity (pDEP level)	Various cell types and viability	0.5 × 106	0.83 (S) + 4.17 (F)	10 kHz – 1 MHz	10−3-10−4	4.1 × 103
[34]	Serpentine channel for inertial focusing with interdigitated planar electrodes and sample focus by sheath flow	Size-based separation of polystyrene beads (pDEP level)	1.4-8.5 × 105	100 (S) + 200 (F)	0.1-30 MHz	1.5 – 2.4 × 10−4	1.4–8.5 × 104
Current device	DEP-induced deflection of flow streamlines at 3D insulator constrictions with focused sample	RBCs from whole blood (nDEP level, crossover & pDEP)	2.25 × 108	0.48 (S) + 1.20 (F)	10 kHz -10 MHz	0.0017 to 0.0525	1.1x105

^†- Initial sample concentration of cells

^*- S=sample and F=focus flow

^††- σ_m is media conductivity

^§- estimated

3. Experimental and Theoretical Methods

3.1. Microfluidic device fabrication and assembly

The device was microfabricated by standard photolithography methods using SU-8® photoresist (2025, MicroChem) and a mask aligner (EVG 620, EV Group) to generate a patterned master. Following this, PDMS (SylgardTM 184, Dow Corning) was cast into the master and crosslinked at 60°C overnight. PDMS chips were then cut and a biopsy punch was used to create the inlets and outlets. Separately, electrodes were patterned on a glass wafer (University Wafer) by first patterning an underlying resist (AZ-1512, MicroChem) followed by electron beam deposition of an overlayer of Au (100 nm) over a Ti adhesion layer (5 nm), so that the lift-off technique with acetone can be used to remove excess resist to pattern Au on glass. The glass wafer was diced (DISCO DAD 3240, Kiru-Kezuru-Migaku Technologies) to obtain microchips with the patterned electrode features on glass. Following this, the electrode features were aligned to the PDMS channel features using a stereoscope and clamped for bonding under a low energy plasma system (Tergeo, Pie Scientific) for 30s and 20 W power. The channel depth was 30μm and the constriction gaps were designed to be 60 μm, with at least 30 μm spacing between the respective electrode edge and constriction tip along the y-axis (i.e. space for flow passage was at least 5-times the size of RBCs), as obtained using a stereomicroscope for alignment of the PDMS channel to electrodes on the glass chip (<± 5 μm mis-alignment). The likelihood of channel clogging due to trapping of sample particles between the constriction tip and electrode edge was lowered by using a sheath flow to focus the sample particles close to the channel wall neighboring the constriction. Since aggregates of PDMS posed a bigger risk to clogging, we included an array of posts just before the region of sample entry into the microchannel to filter such aggregates that were in the several ten micron size range, thereby avoiding their transport to the constricted region of the channel for preventing clogging. The microfluidic device was assembled into a 3D printed holder (FDM) with an embedded custom PCB (Printed Circuit Board) for the required electrical connections to the electrodes.

3.2. Microfluidic device operation

Syringe pumps (Nemesys, Cetoni GmbH) were used for driving the sample and focusing flow through the chip. Electric fields were applied using a signal generator (33220A LXI, Agilent technologies) coupled to an amplifier (A400DI, FLC Electronics) to deliver the final peak-to-peak voltage (~60 V_pp). Dielectrophoretic deflection of cells was imaged on an inverted microscope (Axio Observer 7, Zeiss) with a CMOS camera (Orca flash 4.0 V2, Hamamatsu). Post processing of the images was accomplished with an open source image processing software (Fiji, National Institute of Health).

3.2. Biological sample preparation

The biological samples for these studies was a stock solution of human red blood cells (hRBCs) (Malaria Lab, University of Virginia) in albumin (HSA, Sigma Aldrich) diluted to a concentration level of 2.25 × 10⁸ cells/mL. The sample was spun down for 5 min at 1000 rpm (5430 centrifuge, Eppendorf) and washed twice with DEP buffer (8% Sucrose, 1% BSA & 1X PBS for the higher media conductivities), so that the net media conductivity could be adjusted to: 17 μS/cm, 150 μS/cm and 525 μS/cm, as per three independent measurements using a conductivity meter (LAQUAtwin, Horiba).

3.3. Electric field simulations and fits to the shell dielectric model

Computational Fluid Dynamic (CFD) simulations were conducted for the purpose of device design and optimization, using the COMSOL Multiphysics software (COMSOL Inc.), to simulate field profiles, flow streamlines and particle transport under the force fields. The DEP response of model RBCs was performed using custom MATLAB code [35], [36], and verified using the MyDEP package [37].

4. Results and discussion

4.1. Simulations of electric field profiles and particle flow trajectories in the microfluidic device

In order to assess the microfluidic device with 3D constrictions energized by planar electrodes, as presented in this work (Fig. 2a), we present its comparison to an equivalent device comprising a straight channel with 3D electrodes (Fig. 2c and Electronic Supplementary Material (ESM) Fig. S1) that has been widely studied for dielectrophoretic deflections in prior work [22], [38], but is difficult to fabricate. Focusing on simulation of the field profiles along the red boxed region of interest in Fig. 2a and 2c, we plot the field profiles for the two device types across the indicated probe-lines of Fig. 2b to quantify variations along the x-coordinate (Fig. 2d) and y-coordinate (Fig. 2e). Based on this, it is apparent that the 3D constriction design presents greater spatial extent of the high field region than obtained for the straight channel design with the electrodes. As a result, the chance for traversing cells to interact with the field non-uniformity is increased at a number of flow streamlines, whereas for the case of the straight channel with electrodes, the E-field profile is strongly damped in the region between the electrodes, which significantly reduces the effective area for high field. The plot of the E-field norm variation along the y-coordinate shows that the effect of the high E-field due to the constriction tip is expanded to cover the area between electrodes in the y axis. Hence, coupling of the 3D constriction with the planar electrode leads to an enhanced region of high field in x and y directions, so that it is comparable or higher than the extent of the high field region obtained for the straight channel device with 3D electrodes. As a result, we anticipate that cells interact with the field non-uniformity, not only at the constriction tips, but also over a more extended area than obtained in the case of the straight channel with electrodes.

Figure 2.

2D simulations of the electric field (V/m) profiles for: (a) constriction channel of Dy-DEP design versus (c) channel design with electrodes only. The field profiles for the two devices across the probe-lines per (b) horizontal probe-lines (top), vertical probe-lines (bottom) are shown in (d) for E-field norm variation in the x-coordinate for the 3D constriction channel: A-A’, B-B’, C-C’ versus on the equivalent straight channel with electrodes: a-a’, b-b’, c-c’. (e) E-field norm variation in the y-coordinate, per the inset for the E-field variation on the tip (C-C’).

Comparison of the 3D field profiles of the design in current work of 3D constrictions coupled to planar electrodes (Fig. 3a) versus the straight channel design with 3D electrodes (Fig. 3b) further illustrates the above inference. It is apparent that the design of the current work interacts with the field from the planar electrodes to result in a 3D spatial field distribution across the device depth, with an enhanced E-field magnitude versus that created by 3D electrodes in the straight channel. Furthermore, since the 3D constrictions are spread over an array of wide area, a wider microchannel with a higher sample volume can be used and cells over a larger number of fluid streamlines are able to interact with the field non-uniformity, thereby enhancing the throughput of cells analyzed (i.e. higher analyzed cells per min).

Figure 3.

3D simulations of the Electric field norm (V/m) distribution in: (a) Dy-DEP design with 3D constrictions coupled to planar electrodes (see inset for 3D E-field distribution between the planar electrodes and the constriction tip) compared to (b) the equivalent straight channel design with 3D electrodes. The colors are adjusted for equivalent field levels to present the relative differences in field extent, but (a) extends to a higher level of maximum field versus (b).

With the finalized device design, simulations were performed to optimize flow rates and number of constrictions by tracking the movement of particles under the force fields in the device of the current work (Fig. 4). Simulations (COMSOL) were used to determine the minimum ratio of sheath flow to sample flow required to focus cells to within 100 μm of the upper channel wall, so that upon further focusing in the constriction region due to enhanced velocity, the particles would pass along streamlines that were within a distance of ~20 μm from the constriction tip. This distance to set the limit for particle streamline from the constriction tip was based on the simulated high field region that is indicated as shaded in Fig. 2d and 2e. Using this minimum ratio of sheath flow to sample flow of 3, the maximum net flow rate level at which cells would continue to be deflected from their streamlines by dielectrophoresis was experimentally determined to be 1.68 μL/min. This sheath flow to sample flow level and the net flow rate level were subsequently used to study the dynamic dielectrophoretic deflection of human red blood cells (RBCs) at various media conductivity and frequency conditions of the applied field. As an example, the separation of RBCs (colored red in Fig. 4 and approximated to model cell of 5 μm radius) from platelets (colored blue in Fig. 4 and approximated to model cells of 1.8 μm radius) can be used to optimize the device design and operating conditions. In the absence of applied voltage (Fig. 4a), there is no separation of streamlines and the respective cells appear further scattered in the subsequent flow expansion region. On the other hand, in the presence of an applied voltage (Fig. 4b; 50 V_pp at 100 kHz within media of conductivity of 550 μS/cm), the far higher nDEP level on RBCs versus that on platelets causes a separation of their respective streamlines to a spatial extent of ~100 μm, with a further spatial separation to ~755 μm within the subsequent flow expansion region. Furthermore, it is apparent that the cells are progressively deflected over each of the consecutive field non-uniformities, for up to 16 constrictions (see ESM Fig. S2).

Figure 4.

Particle tracing simulations with model cell types for optimizing design and operating conditions for the separation of RBCs (red of 5 μm) versus platelets (blue of 1.8 μm). (a) No applied Voltage (no DEP) causes the undeflected cells to be scattered at the outlet (right inset). (b) Applied Voltage (50 V_pp) shows significantly higher nDEP deflection of RBCs versus platelets (at 100 kHz with a media conductivity of 550 μS/cm), causing spatial separation in their flow streamlines (per inset).

4.2. Measurement of flow trajectories of dielectrophoretic deflected RBCs

To demonstrate the ability of the device to easily distinguish DEP translation direction and level in a high throughput (large number of cells per minute) and dynamical (high-flow rate) manner based on deflected particle streamlines, we use a sample of human red blood cells (RBCs) obtained from diluted human blood to a starting concentration of: 2.25 × 10⁸ cells/mL. Using a total flow rate 1.68 μL/min, obtained due to sample flow at 0.48 μL/min that is focused using a sheathing flow of 1.2 μL/min, we study the ability to measure dielectrophoretic deflections of varying level and direction, at a throughput of 1.1xl0⁵ cells/min at a voltage of ~60 V_pp applied across planar electrodes (spaced 150 μm) over a 10 kHz to 1 MHz frequency range and within media of conductivity levels of 17 μS/cm, 154 μS/cm and 525 μS/cm (Fig. 5). Based on the broad distribution of RBCs obtained under no field conditions (Fig. 5a), pDEP behavior at high frequencies of the applied field (500 kHz) within media of low conductivity (17 μS/cm) causes the RBCs to be focused right at the edge of the channel wall (Fig. 5d). As the frequency of applied field is lowered to 100 kHz, the pDEP focusing close to channel wall continues to be apparent (Fig. 5c), down to 10 kHz frequency of applied field wherein crossover begins to be apparent based on broader dispersion of cells across streamlines (Fig. 5b). At the higher media conductivity of 154 μS/cm, nDEP is apparent at 30 kHz of applied field based on focusing of RBCs away from edge of the channel wall (Fig. 5e), whereas crossover is apparent at 100 kHz of applied field based on the dispersed RBC streamlines (Fig. 5f) and pDEP is apparent at even higher frequencies of 500 kHz based on focusing of RBCs close to edge of the channel wall (Fig. 5g). Finally, at the highest media conductivity used in this work (525 μS/cm), nDEP is apparent at the lower frequencies of 30 kHz (Fig. 5h) and 100 kHz (Fig. 5i), based on focusing of RBCs at a critical distance away from edge of the channel wall, right up until a frequency of 400 kHz wherein crossover is apparent based on the dispersed RBC streamlines (Fig. 5j). It is noteworthy that in order for flowing particles to experience significant levels of pDEP trapping due to the electrodes, they would need to traverse in a streamline within 10 μm of the electrode edge in the y-direction, per the simulations of Fig. 2c. This situation is avoided by the sheath flow to focus the sample particles to within ~20 μm from the constriction tip, which places the particles at greater than 10 μm from the electrode edge along the y-axis. Furthermore, when the electrodes are at a frequency corresponding to pDEP behavior, the particles are pulled towards the constriction, thereby pushing the particles further away from the electrodes to avoid pDEP trapping. When the electrodes are at a frequency corresponding to nDEP behavior, the particles can be deflected closer to the electrodes, but the frequency level used ensures no pDEP trapping at the electrodes. For the case of operating the device at the crossover frequency wherein particle dispersions are at their maximum level, the images in Fig. 5b, 5f and 5j show that particle streamlines are at least 20 μm away from the electrodes, thereby avoiding any significant level of pDEP trapping. Finally, due to the high flow rates used in this study, the time period for pDEP at the electrodes is further reduced to obviate pDEP trapping.

Figure 5.

Effect of dielectrophoretic translation on flow trajectories of human red blood cells (hRBCs) at a sample concentration of 2.25 × 10⁸ cells/mL at a total flow rate of 1.68 μL/min (sample flow of 0.48 μL/min plus focusing sheath flow of 1.2 μL/min) for measurement at a throughput of 1.1x10⁵ cells/min.: (a) No applied voltage. (b-j) with applied voltages of ~60 V_pp across 150 μm spaced electrodes at indicated media conductivities (vertical axis) and frequencies (horizontal axis), with the DEP level and direction indicated by labels.

To quantify the flow trajectories of RBCs, we applied an image threshold method to assess the ability to distinguish direction of DEP deflection and its relative level based on position of the cells. The summary data of Fig. 6a is a box plot of the histogram range in position of traversing RBCs from edge of channel wall (y-direction) under the conditions investigated within Fig 5. For the ideal case of RBCs deflected under pDEP versus under nDEP, the respective histograms can be clearly distinguished based on lateral separations in streamlines of > 20 μm (Fig. 6b). The strong pDEP behavior at media of low conductivity (17 μS/cm) causes the RBCs to be focused to within 15 μm of wall edge versus the highly dispersed profile under no DEP (Fig. 6c). At intermediate media conductivity (154 μS/cm), weak nDEP at 30 kHz focuses RBCs to be at least 60 μm away from wall edge (Fig. 6d), and weak pDEP at 500 kHz focuses RBCs to within 30 μm of the wall edge (Fig. 6e), in comparison to the respective highly dispersed profiles under no DEP. Similarly, at higher media conductivity (525 μS/cm), strong nDEP is also distinguished well versus no DEP behavior. Based on the quantification presented here and the quantitative limits set for displaced RBC streamlines from the channel wall edge (see ESM Fig. S5), we infer that pDEP, nDEP and crossover behavior of single-cells can be discerned based on their deflected streamlines. The crossover frequency levels for the RBCs at the three measured media conductivity levels that is obtained from the current Dy-DEP device (see ESM Fig. S4) are validated by comparing the calculated membrane capacitance (C_mem) and dielectrophoretic dispersion versus that obtained in prior work [20], [37], as presented in ESM Table S1 and the computer dispersion in ESM Fig. S3. The computed C_mem of 11.7 ± 1.2 mF/m² is close to the ~10 mF/m² reported in prior work and the crossover values match to the computed dispersion based on established dielectric properties of RBCs using the MyDEP program.

Figure 6.

Intensity threshold plots obtained from phase contrast microscopy images are used to assess the ability to discern differences in dielectrophoresis level and direction based on the flow streamlines: (a) summary data box plot with range of histograms in displaced position of traversing RBCs from edge of channel wall (y-direction) under the conditions from Fig 5, including: (b) pDEP versus nDEP deflection is clearly distinguished based on lateral separations in streamlines of > 20 μm at 154 μS/cm and ~40 μm comparing pDEP at 17 μS/cm to nDEP at 154 μS/cm (see Fig. 6a); (c) strong pDEP causes focusing of RBCs to within 15 μm of wall edge versus the highly dispersed profile under no DEP; (d) weak nDEP focuses RBCs at least 60 μm away from wall edge, and (e) weak pDEP focuses RBCs to within 30 μm off the wall edge, in comparison to the highly dispersed profile under no DEP; (f) strong nDEP is also distinguished well versus no DEP behavior. For comparison, the displacement range for the FIELD OFF condition is also indicated as an arrow in (a) (95% confidence level).

4. Conclusions and outlook

We present a microfluidic device capable of high throughput dynamical analysis to determine the dielectrophoretic translation level and direction of single-cells over a wide frequency range based on their deflected flow streamlines. Using electric field simulations, the device with 3D insulator constrictions that is energized by a set of planar electrodes is shown to have a spatial extent of field that exceeds the equivalent straight channel device with 3D electrodes. Hence, cells focused along streamlines in the vicinity of the constriction region and traversing through the device at high flow rates have a high likelihood of experiencing significant levels of deflection due to varying levels of pDEP and nDEP, as confirmed by simulations of particle tracking using a set of 16 high field points. Based on measurements of particle deflection on such a device using human red blood cells at a high initial concentration, we show the ability to distinguish between strong nDEP versus strong pDEP; weak nDEP versus no DEP at the crossover frequency; and weak pDEP versus no DEP at the crossover frequency. The quantification ability of the current Dy-DEP device was validated by comparing the obtained membrane capacitance (C_mem) and dielectrophoretic dispersion to that obtained within prior work. Hence, based on the ability to discern dielectrophoresis-induced deflections in cell streamlines at a high flow rate and a high sample concentration, we suggest that the device can be used to determine the dielectrophoretic dispersion of a sample of cells at a high throughput, single-cell sensitivity and with no need for significant sample dilution. It is noteworthy that since each traversing cell is individually displaced based on its electrical phenotype and measured based on the position of its deflected flow streamlines, the reported method does not average across the population and is capable of quantifying the DEP frequency dispersion of single-cells. Furthermore, since cells are focused away from the electrodes and traverse through the consecutive high-field regions of the device at high flow rate (i.e. just milliseconds at high field points), we suggest that their viability is likely not adversely affected by the field. Future work is focused on validating viability effects on the cells within the device, determining the upper limit of cell concentration for DEP analysis in the device and measuring ability to quantify heterogeneity in the cell capacitance phenotype.

Biography

Karina Torres is a Ph.D. student at University of Virginia’s Charles L. Brown Department of Electrical and Computer Engineering. She received her B.E. in Chemical Engineering from the University of Costa Rica and her M.S. in MEMS from the Costa Rica Institute of Technology. Her areas of research interest are microfluidics, microfabrication and electrokinetics for lab-on-chip applications. She is also interested in new ways of harnessing energy from biological entities and bio-inspired soft-robots. Karina is a member of the American Association for the Advancement of Science (AAAS) and the Society of Women Engineers (SWE).

Carlos Honrado received his Ph.D. from the School of Electronics and Computer Science at the University of Southampton (UK), where his work focused on label-free single particle analysis and separation. He is now further pursuing these research topics, currently holding a Postdoctoral Research Associate position in the Department of Electrical and Computer Engineering at the University of Virginia (USA). His research interests are focused on the development of microfluidic devices for biomedical applications, including label-free microfluidics, single-cell sorting and analysis, AC electrokinetics, dielectric characterization of cells and technology integration. He is a member of the Biomedical Engineering Society (BMES).

Walter Varhue received his Ph.D. in electrical engineering from the University of Virginia and is currently serving as a research associate within the cross-disciplinary Center for Advanced Biomanufacturing at the University of Virginia. His research interests include biofabrication approaches to engineer cellular microenvironments and cellular analysis based on microfluidic separation and cytometry. He is a member of the Biomedical Engineering Society (BMES).

Vahid Farmehini is a Ph.D. student at the University of Virginia, with research interests centered on the application of analog electronic circuits for electrical stimulation and signal transduction within bio-analytical microdevices. His key contributions have been in the development of wideband amplifiers for frequency-selective particle enrichment by electrode-less dielectrophoresis, impedance-based measurement of cells during microfluidic manipulation, circuit designs for electrical stimulation with electromyography for measuring muscle regeneration and circuits to enable portable systems for forensic analysis in resource-poor settings. He completed his BS and MS degrees in Electronics at Sharif University in Iran.

Nathan Swami serves as Professor of Electrical & Computer Engineering at the University of Virginia (UVA), Charlottesville, VA. His research group specializes in label-free microfluidic techniques for biofabrication, electrophysiology-based single-cell analysis and nano-confined systems for biomolecular analysis. Prior to University of Virginia, he served on the scientific staff of the MEMS group at Motorola Labs and at Clinical Microsensors, Inc., a Caltech start-up interfacing microelectronics to bio-analysis. He seeks to impact diagnostic systems within point-of-care and resource-poor settings for precision medicine.