Unraveling the dielectric heterogeneity response of cell lines under electroporation: a microfluidic single-cell on-chip system

Abstract

In the clinical application of irreversible electroporation for tumor therapy, individual differences in cellular electrical responses often lead to the reliance on empirically selected treatment parameters, which easily cause over/under-treatment in patients and hinder the development of precision tumor therapy. To clarify the response mechanism induced by single-cell individual differences, this study developed an experimental platform integrating microfluidic chip technology, dielectrophoresis (DEP), electrorotation (EROT), and electroporation (PEF) technologies, enabling the full process of single-cell “capture-rotation-electroporation-re-rotation” on one chip. By extracting four core dielectric parameters (conductivity and permittivity of membrane and cytoplasm) via rotation spectrum analysis and the size effect of cell, we proposed a semi-quantitative Electroporation Sensitivity Index (EPI). The EPI is grounded in the cumulative contribution of static intrinsic sensitivity and dynamic promotion to quantify cellular responses. This study used four cell lines (A549, HUH-7, U251, B16F10) for validation. The results showed that: the platform stably captured cells at the center of the electrodes and induced rotation, with the coefficient of determination (R²) of the rotation spectrum fitting curves all greater than 0.97, confirming the strong robustness of the method. The EPI ranking of the four cell lines (B16F10: 3.628 > A549: 2.678 > U251: 2.364 > HUH-7: 1.974) was completely consistent with the actual ranking of maximum instantaneous mortality (IR) (B16F10: 82.28% > A549: 81.87% > U251: 78.81% > HUH-7: 60.61%), verifying the effectiveness of EPI in analyzing cell electroporation responses. This study provides a quantitative framework for decoding tumor heterogeneity, laying a theoretical foundation for precision electroporation therapy.

Introduction

Electroporation technology induces the formation of transient nanopores in cell membranes via pulsed electric fields¹, and has emerged as a core tool in biomedical fields such as gene delivery^2,3, drug loading^4,5, and tumor ablation^4,6. In the field of irreversible electroporation (IRE) for tumor ablation, therapeutic efficacy and safety are highly dependent on the precise matching between electric field parameters (intensity, pulse width, waveform) and cellular dielectric responses.

Currently, over-treatment or under-treatment is prone to occur in clinical practice, with the core problem being the widespread single-cell heterogeneity within the tumor microenvironment^7,8. Significant variations in membrane potential and capacitance exist among cells within the same tumor^9,10, leading to distinct susceptibility to poration and heterogeneous death patterns¹¹. Consequently, conventional protocols with preset parameters often fail to achieve the ideal goal of “eradicating tumor cells while preserving normal tissues”.

To decode such cellular heterogeneity, recent advancements in microfluidic technologies have revolutionized single-cell analysis. Emerging platforms have achieved remarkable success in the high-throughput isolation and multi-dimensional profiling of living single cells. For instance, integrated microfluidic chips combined with antibody barcode arrays have enabled the multiplexed profiling of secreted biomarkers from living single cells, providing critical functional information for tumor cell classification and immunotherapy assessment^12,13. Furthermore, advanced isolation strategies have been developed to capture and analyze single-cell-derived small extracellular vesicles or exosomes from complex clinical samples, serving as non-invasive biomarkers for cancer screening^14–16. In parallel, planar DNA nanoball arrays have pushed the boundaries of throughput, facilitating imaging-integrated single-cell profiling that preserves spatial context¹⁷.

Despite these breakthroughs in biochemical and molecular profiling, the biophysical properties of single cells, particularly their dielectric response to electric fields, remain a distinct yet crucial dimension of cellular identity that is often overlooked. Unlike biochemical markers, dielectric properties offer a label-free, instantaneous snapshot of the cell’s structural integrity and physiological state, which is directly relevant to physical therapies like electroporation.

Several existing studies have improved tumor ablation efficacy by optimizing treatment parameters. Kim et al.¹⁸ enhanced the ablation range of pancreatic cancer by optimizing electrode arrangement. Wang et al.¹⁹ proposed a novel waveform called Synergistic Bipolar Irreversible Electroporation (SBIRE), which can expand the ablation area while reducing muscle contractions. Although these methods have improved ablation efficacy to a certain extent, they fail to explain the mechanism underlying the differences in electroporation responses caused by tumor heterogeneity. The enhanced efficacy is solely based on experimental results, without elucidating the electrical mechanism of efficacy improvement from the perspective of electrical response.

Therefore, to address the core bottleneck of unclear mechanisms underlying heterogeneous electroporation responses induced by single-cell heterogeneity in the tumor microenvironment and the lack of electrical-mechanism support in existing studies. Ion channels exhibit heterogeneous expression in cancer, and variations in membrane potential exacerbate differences in electroporation responses²⁰. Poderyte et al.²¹ first applied scanning electrochemical microscopy to the study of electroporation in living yeast cells, monitoring the process of electric field-induced membrane changes through impedance data analysis of cells before and after poration. Li et al.²² combined self-powered triboelectric nanogenerators with electrical impedance spectroscopy to measure cellular impedance spectra, enabling real-time evaluation of electroporation efficiency.

However, these methods are limited to population-level analysis across multiple cells and fail to distinguish the inherent specific electrical differences between individual cells. Consequently, it is impossible to establish a direct correlation between single-cell dielectric parameters and electroporation responses, making it difficult to reveal the underlying electrical mechanisms of heterogeneous electroporation responses.

In recent years, the combination of dielectrophoresis (DEP) and electrorotation (EROT) on microfluidic chips has emerged as a mainstream strategy for high-precision dielectric characterization^23–25. DEP is a physical technique that manipulates charged or polarizable particles by exerting force through non-uniform electric fields^26,27. Similarly, when charged or polarizable particles are placed in a non-uniform rotating electric field, the field generates a controllable torque on the particles, driving their rotation. This phenomenon known as EROT^27,28. Trainito et al.²⁹ developed a microfluidic chip that uses DEP electric fields to capture MOSE cells at different stages, measure their rotation speeds under rotating electric fields, fit the electrical parameters of MOSE cells, and quantifies the individual-specific differences among MOSE cells at different stages. Moisescu et al.³⁰ investigated the effects of electroporation on cellular electrical parameters and distinguished electroporated from non-electroporated cells using DEP technology, but failed to accurately measure the cellular electrical parameters. Trainito et al.³¹ integrated DEP capture, EROT dielectric measurement, and pulsed electric field (PEF) electroporation in a microfluidic chip, enabling the comparison of single-cell dielectric parameters before and after electroporation. However, it did not establish a quantitative correlation between dielectric parameters and cell viability. Since then, few studies have been reported on single-cell personalized dielectric differences and the dielectric response mechanisms of cells to pulsed electric fields.

In summary, establishing a direct correlation between single-cell dielectric parameters and electroporation responses, and revealing the underlying electrical mechanisms of heterogeneous electroporation responses, are critical for achieving more precise ablation efficacy and advancing pulsed electric field-based tumor therapy.

To address the above issues, this study aims to develop a high-precision single-cell dielectric parameter model and explore the dielectric response mechanisms of single cells under pulsed electric fields. As shown in Scheme1. We propose an integrated “microfluidics-DEP-EROT-PEF” platform. Specifically, we fabricated microfluidic electrodes suitable for three electric field functions: DEP, EROT, and PEF. The electrodes are fabricated on glass substrates using wet etching technology, with an inter-electrode distance of 60 μm and a microchannel width of 100 μm. A workflow of “capture–rotation–electroporation–re-rotation” was adopted to measure the electrorotation spectra. Genetic algorithms and other methods were used to automatically fit and monitor four core electrical parameters before and after electroporation. Meanwhile, cell suspension experiments were conducted under the same electric field parameters to determine the cell mortality rate. Four cancer cell lines (HUH-7, U251, A549, B16) were tested in this study. As shown in Scheme 1b. Based on the four electrical parameters and physical morphology of the cells, a semi-quantitative analytical index (Electroporation Sensitivity Index, EPI) was proposed. EPI consists of static intrinsic sensitivity and dynamic promotion coefficient, integrating five specific heterogeneous parameters involved in the electroporation effect: shunting inhibition coefficient, shielding inhibition coefficient, size promotion coefficient, membrane polarization coefficient, and morphological parameter. As shown in Scheme 1c. This index can be directly used to compare the relative electroporation sensitivity among different cell lines, realizing the transformation from qualitative description to quantitative expression. The proposed method enables accurate measurement of cellular electrical parameters and quantifies the electrical response mechanisms of heterogeneous single cells to pulsed electric fields, which is expected to lay a theoretical foundation for cancer research and precision medicine.

Scheme 1. Derivation and validation of the Electroporation Sensitivity Index (EPI) based on single-cell dielectrophoretic response and dimensional analysis.

a The “Capture-Rotation-Electroporation-Re-rotation” experiment using quadrupole electrodes. b Construction of the EPI model, which synthesizes static intrinsic sensitivity and dynamic structural promotion. c Comparative analysis confirms that the EPI ranking matches the hierarchy of treatment efficacy (B16F10 > A549 > U251 > HUH-7), verifying its predictive capability for tumor heterogeneity

Results

Cell viability results

The viability experiments of the four cell lines were conducted in accordance with the sequence designed in Table 4, and the mortality rate of the cell suspensions are shown in Fig. 1. Experiments show that the cell suspensions exhibited differential mortality rates under different pulse parameters. Statistically significant (P < 0.05) differences in mortality rates existed at both the intra- and inter-cell line levels. The results indicate that the observed gradient in mortality rates across the cell suspensions was reasonable.

Table 4.

Pulsed stimulation parameters of the experimental group

Experimental number	Waveform	Number of pulses
No. 1	IRE	2
No. 2		5
No. 3		8
No. 4	SBIRE19	2
No. 5		5
No. 6		8

Fig. 1.

Quantitative analysis of cell mortality rates across four distinct cancer cell lines under varying pulsed electric field parameters. a HUH-7, b U251, c A549, and d B16 cell lines corresponding to the experimental groups defined in Table 1. The bar charts illustrate the mortality percentages, with statistical significance indicated by asterisks (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001)

Composition of the single-cell dielectric measurement platform

Figure 2 shows the composition of the single-cell dielectric measurement platform. The signal generation system comprises three distinct units: a dual-channel signal generator (Label 1) producing the pulsed electric field (PEF) for electroporation, a second dual-channel generator (Label 2) providing the dielectrophoretic (DEP) field for cell capture, and a four-channel generator (Label 3) generating the electrorotation (EROT) field to induce rotational torque. These components are coordinated by a host computer (Label 4). A custom-fabricated signal combiner (Label 5) synchronizes and superimposes the signals, subsequently transmitting the output to the electrodes of the microfluidic chip (Label 6). The chip is mounted on an optical microscope (Label 9) integrated with a CMOS camera for visualization. For sample delivery, a microfluidic pump (Label 8), regulated by a controller (Label 7). The chip features an inlet and an outlet, allowing all single-cell dielectric measurements and manipulations to be performed.

Fig. 2.

Schematic diagram of the integrated microfluidic single-cell dielectric measurement and electroporation platform. The system architecture includes: 1–3 Signal generation units for PEF, DEP, and EROT fields; 4 Host computer for process control and data analysis; 5 Signal combiner; 6 Microfluidic chip interface; 7-8 Microfluidic pump system; and 9 Optical microscopy system for real-time monitoring

Cell dynamics throughout the procedure are monitored via microscopy, with image data systematically acquired and processed by the host computer. Through image analysis, the relationship between rotation speed and frequency is established, enabling the quantification of variations in rotational velocity before and after exposure to varying PEF parameters. Post-electroporation cell movement is monitored for a duration of 10 minutes to statistically analyze the dynamic evolution of rotation speed. Consequently, alterations in single-cell dielectric parameters are characterized by applying a fitting algorithm to correlate experimental data with theoretical spectra.

Working principle of cell manipulation

The cell manipulation process is divided into four sequential modes, relying on the precise regulation of three electric fields (DEP, EROT, and PEF) to achieve cell capture, rotation, electroporation, and post-electroporation re-rotation. The specific operating principles and mathematical descriptions are as follows:

MODE1: DEP cell capture

The prepared cell suspension is injected into the microchannel via a microfluidic pump. At this stage, the signal generator (Fig. 2 Lable 2) is controlled to produce a DEP electric field, as shown in Fig. 3a. Cells become polarized under the electric field, and their internal charges redistribute to form a dipole moment. For spherical cells, the dipole moment can be expressed as:

$$\mathbf{p}=4\pi {\epsilon }_0{\epsilon }_m{R}^3{K}_{\mathrm{CM}}\left(\omega \right)\mathbf{E}$$ 1

Where ε₀ is the vacuum permittivity, ε_m is the medium permittivity, R is the cell radius, K_cm(ω) is the Clausius-Mossotti (CM) factor, and E is the electric field intensity.

Fig. 3.

Working principles and experimental validation of single-cell manipulation modes. a–c Schematic representations of electrode configurations and signal applications for a DEP capture, b EROT rotation, and c PEF electroporation. Finite Element Method (FEM) simulations of the electric field distribution for (d) the DEP field and (e) the EROT field at various phases. Time-lapse microscopy sequences demonstrating (f) the trajectory of cell capture and (g) the induced cell rotation dynamics

The CM factor describes the relative polarization ability between the cell and the medium, and its complex form is:

$$K_{CM}=\frac{{\epsilon }_c^{*}-{\epsilon }_m^{*}}{{\epsilon }_c^{*}+2{\epsilon }_m^{*}}$$ 2

where the complex permittivity of the cell ε₀* is determined using the single-shell dielectric model of cells:

$${\epsilon }_c^{*}={\epsilon }_{mem}^{*}\frac{{\epsilon }_{cyto}^{*}+2{\epsilon }_{mem}^{*}+2\left({\epsilon }_{cyto}^{*}-{\epsilon }_{mem}^{*}\right){\left(\frac{R-d}{R}\right)}^3}{{\epsilon }_{cyto}^{*}+2{\epsilon }_{mem}^{*}-\left({\epsilon }_{cyto}^{*}-{\epsilon }_{mem}^{*}\right){\left(\frac{R-d}{R}\right)}^3}$$ 3

Here, d is the cell membrane thickness, and ((R−d)/R)³ represents the ratio of the cytoplasmic volume to the total cell volume. The complex permittivities of the three regions (cytoplasm, cell membrane, and external medium) are defined as follows:

• Cytoplasm: ${\epsilon }_{cyto}^{*}{=\epsilon }_{cyto}{\epsilon }_0-i\frac{{\sigma }_{cyto}}{\omega }$

• Cell membrane: ${\epsilon }_{mem}^{*}={\epsilon }_{mem}{\epsilon }_0-i\frac{{\sigma }_{mem}}{\omega }$

• External buffer medium: ${\epsilon }_m^{*}{=\epsilon }_m{\epsilon }_0-i\frac{{\sigma }_m}{\omega }$

The magnitude of the DEP force acting on cells is positively correlated with the electric field intensity, and its direction depends on the sign of the real part of the CM factor. The DEP force is described by:

$${\mathit{F}}_{DEP}=2\pi {\epsilon }_0{\epsilon }_m{R}^3\mathrm{Re}\left[K_{CM}\left(\omega \right)\right]\nabla {\mid \mathit{E}\mid }^2$$ 4

When Re[K_cm(ω)] > 0, cells move toward regions of high electric field intensity (positive dielectrophoresis, pDEP); conversely, they move toward regions of low electric field intensity (negative dielectrophoresis, nDEP). Thus, the position of cells can be regulated by adjusting the electric field. As shown in Fig. 3f, a cell moves along the microchannel with an initial velocity at t = 0 s. As it approaches the electrode region at t = 0.6 s, it decelerates slowly under the action of the DEP force, and finally stops at the center of the electrodes at t = 1.5 s.

To validate the trapping mechanism, FEM simulations were performed. As shown in Fig. 3d, the quadrupole electrode geometry generates a characteristic electric-field cage. We set the grid size to the COMSOL default normal grid size. The radius of the semi-circular electrode is consistent with the actual value of 25 µm. The conductivity of the medium solution is set to 0.1 s/m. The electric field intensity is minimal at the center and increases rapidly towards the electrode tips. Since the cells exhibit nDEP behavior in the low-conductivity buffer, they experience a repulsive force from the high-field regions, effectively confining them within the central low-field potential well. This simulation confirms that the experimental cell capture is driven by the spatial gradient of the electric field.

MODE 2: EROT-induced cell rotation

Once cells are stabilized at the center of the electrodes, the signal generator (Fig. 2 Lable 3) is activated to produce an EROT electric field, as illustrated in Fig. 3b. When a rotating electric field is applied, the induced dipole moment of cells lags behind the electric field variation, resulting in a phase difference. The magnitude of the torque is proportional to the square of the electric field intensity and the imaginary part of the CM factor:

$${\mathbf{T}}_{\mathrm{EROT}}=2\pi {\epsilon }_0{\epsilon }_m{R}^3\mathrm{Im}\left[K_{\mathrm{CM}}\left(\omega \right)\right]{\mid \mathbf{E}\mid }^2$$ 5

The angular velocity of cell rotation is expressed as:

$${\mathrm{\Omega }}_{ROT}\left(\omega \right)=-\frac{{\epsilon }_0{\epsilon }_m{\mid \mathit{E}\mid }^2}{2\eta }Im\left[K_{CM}\left(\omega \right)\right]$$ 6

Where Ω_ROT(ω) is the rotational angular velocity, ω is the angular frequency of the electric field, and η is the dynamic viscosity of the medium.

The direction of cell rotation is determined by the sign of the imaginary part of the CM factor. As shown in Fig. 3g, cells rotate clockwise under the EROT field in this study.

For the EROT measurements, quadrature sinusoidal signals were applied to the four electrodes to generate a rotating electric field. FEM simulations (Fig. 3e) confirmed that the electric field distribution in the central region is spatially uniform in magnitude while rotating in direction. This verified field characterization ensures the accuracy of the subsequent rotation spectrum analysis.

MODE 3: Single-cell electroporation

The EROT electric field is turned off while the DEP field is maintained to keep cells in place. At this stage, the signal generator (Fig. 2 Lable 1) is controlled to produce a PEF electric field, as shown in Fig. 3c. Cells undergo electroporation under the action of the pulsed electric field, and the degree of membrane permeabilization can be precisely controlled by adjusting the PEF parameters.

MODE 4: Re-rotation

The PEF electric field is turned off, and the DEP field is retained to stabilize the cells. The EROT electric field is re-applied to measure the rotational behavior of cells after electroporation. This step enables the acquisition of post-electroporation rotation spectra, which are used to invert the changes in dielectric parameters (e.g., membrane conductivity and permittivity) induced by electroporation.

Setup of the single-cell dielectric measurement platform

Based on the platform’s working principle, an experimental platform for single-cell dielectric measurement was established, as shown in Fig. 4 (labels correspond exactly to those in Fig. 2). To ensure accurate measurements, the instruments were configured according to the platform’s operational mechanism:

Fig. 4.

Photograph of the assembled experimental setup for single-cell dielectric measurement. The numerical labels correspond to the hardware components detailed in the schematic diagram shown in Fig. 2, illustrating the integration of signal generators, the microscope, and the microfluidic control system

DEP electric field for cell capture

As illustrated in Fig. 3a, the DEP field was generated by two sinusoidal waves (peak-to-peak voltage: 7 V_pp, frequency: 40 kHz, phase difference: 180°) provided by the signal generator AFG31022 (AFG31022, Teledyne LeCroy Inc., New York, USA) (Fig. 4, Label 2).

EROT field for cell rotation

As shown in Fig. 3b, the EROT field consisted of four sinusoidal waves (amplitude: 3 V_pp, phase difference between adjacent voltages: 90°, angular frequency range: 100 kHz–20 MHz) applied via the function generator AWG2184 (AWG2184, Active Technologies SRL., Ferrara, Italy) (Fig. 4, Label 3).

PEF for electroporation

To induce electroporation in pre-captured single cells, the signal generator DG2102 (DG2102, Rigol Technologies Co., Ltd., Suzhou, China) (Fig. 4, Label 1) provided manually programmable complex waveforms (Fig. 3c).

Visualization and acquisition of image data

The microfluidic chip is mounted on a microscope (CKX53, Olympus Corporation, Tokyo, Japan) (Label 9) equipped with a CMOS camera (E3ISPM12000KPA, Hangzhou ToupTek Photonics Co., Ltd., Hangzhou, China). The CMOS camera is interfaced with the host computer to facilitate real-time visualization and data acquisition.

Precise microfluidic flow control

A microfluidic syringe pump controller (Label 7) is used to control the microfluidic pump (Label 8) (Jianmi Zhikong Technology Co., Ltd., Wuhan, China), which injects the cell suspension sample into the microchannel of the microfluidic chip through a PTFE tube. The controller features a minimum flow rate precision of 10 μL/h.

Synchronization and control

All devices were synchronized and controlled by a host computer (Fig. 4, Label 4) via a LabVIEW interface (NI LabVIEW, Austin, Texas, USA). The three types of electric fields were superimposed using a self-made signal combiner (Fig. 4, Label 5).

Prior to each experiment, the signals applied to the chip electrodes were calibrated using an oscilloscope. The experimentally measured waveforms are shown in Fig. 5. It can be observed that the voltage signals delivered to the chip maintained high fidelity with negligible waveform distortion. Specifically, a frequency of 1 MHz was selected for the EROT waveform shown here. Likewise, identical calibration procedures were performed across the entire frequency range from 100 kHz to 20 MHz.

Fig. 5.

Oscilloscopic verification and calibration of voltage waveforms loaded on the microfluidic chip electrodes, demonstrating high-fidelity waveform transmission. a Dual-channel sinusoidal signals for the DEP trapping field (Mode 1), confirming that opposing electrodes receive completely identical signals as required by the DEP field configuration. b Four-channel quadrature sinusoidal signals generated for the EROT field. c Composite waveforms during Mode 2. Since the DEP field is maintained to strictly prevent random cell displacement, the waveform represents the superposition of EROT and DEP signals

Changes in the electrical parameters of cells before and after electroporation

The established single-cell dielectric measurement platform was used to conduct experimental studies on four different cell lines. As shown in Fig. 6a-d, all experimental results demonstrated that the four cell lines were successfully captured at the center of the electrodes and subsequently induced to undergo electrorotation under the action of the EROT field. Subsequently, changes in the four core electrical parameters of single cells were acquired before and after PEF stimulation.

Fig. 6.

Microscopic visualization of single-cell capture via DEP and morphological evidence of electroporation. a–d Representative images demonstrating the stable trapping of single cells at the electrode center for a B16, b HUH-7, c A549, and d U251. e Morphology of a U251 cell prior to electroporation. f Post-electroporation morphology of the same U251 cell, exhibiting visible cytoplasmic leakage indicative of membrane permeabilization

The PEF stimulation parameters to single cells were calculated based on the experimental electric field intensity in Table 1 and the electrode distance of the microfluidic chip. As illustrated in Fig. 6e, f, morphological changes occurred in the cells after PEF application, with substances observed to flow out of the cells. This indicates that the PEF stimulation was effectively applied to the cells, validating the reliability of the platform’s electric-field loading capability.

Table 1.

Changes in cellular electrical parameters before and after pulsed stimulation

Cell Line	Electrical Parameters	After pulses						Before pluses
		No. 1	No. 2	No. 3	No. 4	No. 5	No. 6
HUH-7	σmem[S/m]	5.89 × 10−5	8.22 × 10−5	9.99 × 10−5	6.89 × 10−5	7.62 × 10−5	8.74 × 10−5	5.88 × 10−5
	εmem	8.54	8.37	7.38	8.50	8.45	7.93	8.81
	σcyto[S/m]	0.112	0.078	0.070	0.091	0.088	0.075	0.119
	εcyto	77.4	52.2	50.6	74.4	72.7	52.9	78.8
U251	σmem[S/m]	6.74 × 10−5	7.96 × 10−5	9.27 × 10−5	7.02 × 10−5	7.54 × 10−5	8.77 × 10−5	6.69 × 10−5
	εmem	13.4	5.91	4.62	8.47	6.36	4.63	13.6
	σcyto[S/m]	0.095	0.053	0.039	0.087	0.063	0.051	0.103
	εcyto	69.9	41.1	30.2	63.3	47.2	39.3	72.7
A549	σmem[S/m]	4.70 × 10−5	6.81 × 10−5	8.67 × 10−5	5.20 × 10−5	6.32 × 10−5	7.41 × 10−5	4.53 × 10−5
	εmem	5.25	7.36	8.70	5.86	6.30	8.01	5.19
	σcyto[S/m]	0.078	0.065	0.061	0.071	0.069	0.064	0.079
	εcyto	42.2	49.8	54.2	45.7	48.8	52.4	41.3
B16	σmem[S/m]	5.96 × 10−5	7.55 × 10−5	9.29 × 10−5	6.17 × 10−5	6.76 × 10−5	7.9 × 10−5	5.48 × 10−5
	εmem	5.83	7.39	8.58	6.31	7.25	8.32	5.82
	σcyto[S/m]	0.129	0.092	0.087	0.127	0.095	0.089	0.150
	εcyto	57.3	66.3	70.8	60.7	65.2	67.8	57.0

Rotation spectra of the four cell lines before electrical stimulation were first plotted, as shown in Fig. 7. The coefficient of determination (R²) for the fitted rotation spectra of all four cell lines exceeded 0.97, indicating that the fitting model can accurately describe the experimental data and exhibits excellent robustness. The experimentally measured spectra were in good agreement with the corresponding theoretical predictions. From the electrorotation spectra of each cell line, it can be observed that the frequencies corresponding to the peak rotational speeds differed among the four cell lines: B16 cells exhibited the highest frequency at the peak, while U251 cells showed the lowest. Additionally, the experimental results revealed that U251 cells had the smallest peak rotational speed. These findings confirm the distinct properties of the four cell lines and highlight their heterogeneous dielectric characteristics.

Fig. 7.

EROT spectra characterizing the frequency-dependent rotational velocity of four cancer cell lines. a HUH-7, b U251, c A549, and d B16. Experimental data points are presented as mean ± standard deviation, and the solid lines represent the theoretical fitting curves. The high coefficient of determination (R² > 0.97) for all fits confirms the robustness of the dielectric modeling

As indicated in Eq. 6, ε₀, ε_m, η, and E are known constants; thus, the rotational speed of cells is proportional to the imaginary part of the Clausius-Mossotti factor K_CM(ω). PEF stimulation induces changes in the four core dielectric parameters of cells. Macroscopically, this leads to alterations in the frequency response of cells, manifested as changes in rotational speed at different frequencies.

The same method was used to measure the rotational spectrum data of cells under different electrical stimulations, as shown in Fig. 8. Compared with the cell suspension mortality results in Fig. 1, under the same stimulation waveform, a higher cell mortality rate induced by electrical parameters was associated with a more significant decrease in the peak rotational speed of cells after electrical stimulation. The peak rotational speed under clockwise (CW) torque is dominated by the cytoplasmic conductivity, validating the trend of decreased cytoplasmic conductivity before and after electroporation. The polarization between single cells and the buffer medium underwent changes: some cells exhibited a reversal in polarization polarity around 8 MHz after electrical stimulation, transforming from CW rotation before stimulation to counterclockwise (CCW) rotation. In contrast, the peak rotational speed under CCW torque is dominated by the cytoplasmic relative permittivity. The CCW peak rotational speeds of U251 and HUH-7 cell lines increased, reflecting a decrease in their cytoplasmic relative permittivity. Conversely, the CCW peak rotational speeds of A549 and B16 cell lines decreased, indicating an increase in their cytoplasmic relative permittivity. The heterogeneous dielectric responses of different cell lines to pulsed electric fields were further confirmed in our experiments.

Fig. 8.

Comparative analysis of post-electroporation electrorotation spectra under varying pulse stimulation parameters. a HUH-7, b U251, c A549, and d B16. The shifts in rotation frequency response curves illustrate the dielectric changes induced by different pulsed electric field conditions

Determination of dielectric parameters and correlation with cell mortality

Once the electrorotation spectra were measured, the four core dielectric parameters before and after electrical stimulation were determined by fitting analysis (Table 1). Table 1 summarizes the mean values. Meanwhile Fig. 9 takes the experiment of Group No.3 as an example to further visualize the complete single-cell parameter distributions to transparently illustrate the intrinsic dielectric heterogeneity. The overlaid scatter plots demonstrate that the electric field stimulation induced highly significant and mathematically robust alterations in cellular dielectric properties. For instance, the electric-field-induced decrease in cytoplasmic conductivity for U251 cells demonstrated an exceptionally large effect size (p < 0.001, Cohen’s d = 11.21, 95% CI: [0.059, 0.069] S/m). Similarly, the cytoplasmic relative permittivity of the B16 cell line increased significantly with a large effect size (p < 0.001, Cohen’s d = 3.23, 95% CI: [10.18, 17.42]). These consistently large effect sizes (Cohen’s d > 0.8) strongly substantiate the statistical reliability of the observed dielectric shifts under the current single-cell sample size.

Fig. 9.

Single-cell parameter distributions demonstrating dielectric heterogeneity and robust electroporation responses. a Membrane conductivity. b Membrane relative permittivity. c Cytoplasmic conductivity. d Cytoplasmic relative permittivity. Data across four cancer cell lines before (Control) and after pulsed electric field stimulation (PEF No.3 group) are presented as overlaid scatter points to transparently illustrate the true extent of single-cell variance, accompanied by distribution boxes denoting the median and interquartile ranges. Each experimental condition contains 12 independent single-cell measurements (n = 12) derived from three biological replicates. The pronounced statistical separation despite intrinsic biological heterogeneity highlights the robust dielectric shifts induced by the electric field

These data are presented in Fig. 10, illustrating the alterations in dielectric parameters induced by electrical stimulation, which correspond to the physico-biological changes in single cells caused by PEF. Notably, all four parameters showed exclusively monotonic relationships with cell mortality.

• The membrane conductivity (σ_mem) was positively correlated with cell mortality. In electroporation theory, increased membrane conductivity directly indicates membrane poration. A higher σ_mem suggests an increase in the number and/or size of membrane pores, thereby elevating the probability of cell death. And the Cytoplasmic conductivity (σ_cyto) was negatively correlated with cell mortality.

• The correlations of membrane and cytoplasmic relative permittivity (ε_mem and ε_cyto) with mortality varied across cell lines. Both permittivity parameters are negatively correlated with mortality in U251 and HUH-7 cells, whereas they are positively correlated with mortality in A549 and B16 cells.

Fig. 10.

Correlation analysis between cell mortality rates and variations in four core single-cell dielectric parameters. a Membrane conductivity. b Membrane relative permittivity. c Cytoplasmic conductivity. d Cytoplasmic relative permittivity. The distinct trends for A549, B16, HUH-7, and U251 highlight the heterogeneity in dielectric responses

Discussion

To enable a correlative analysis between the single-cell electroporation state and the suspension-based cell mortality assay, consistency in experimental parameters was essential. Therefore, the cell concentration was standardized at 1 × 10⁶ cells/mL for both the electroporation treatment and the subsequent single-cell electrorotation measurements.

Furthermore, the solution medium should ideally be identical between these two experimental stages, as procedural variations in the electrorotation assay can readily influence cellular rotation speed. However, standard high-conductivity culture medium could not be used for the electrorotation experiments. Its use would induce significant Joule heating, risking thermal damage to individual cells and confounding the observed correlation between post-pulse mortality and the four derived dielectric parameters.

To precisely position cells at the center of the electrodes for EROT measurements, a stable DEP trapping field must be continuously applied during the experiment. When setting the voltage parameters for this DEP field, careful attention must be paid to the appropriate selection of the peak-to-peak voltage, as the prolonged application of excessively high voltages leads to the corrosion of bare gold electrodes. To mitigate electrode corrosion, an insulating layer such as silicon dioxide (SiO₂) can be fabricated on the electrode surface to achieve physical isolation from the solution medium. However, such insulating coatings are prone to causing poor sealing of the microfluidic chamber and subsequent liquid leakage. These related technical challenges remain to be further addressed in future research. To ensure the practical operational lifespan and robustness of the microfluidic platform without introducing insulating coatings or increasing the complexity of the chip fabrication process, preliminary experiments were conducted. The results demonstrated that irreversible electrode corrosion is triggered when the DEP trapping signal at 40 kHz reaches or exceeds 11 Vpp. Therefore, the capture voltage was strictly maintained below this critical threshold during all cell manipulation procedures. Furthermore, the electrorotation process, characterized by dynamic frequency sweeping, exerts a minimal impact on electrode degradation. This study independently optimized this process to ensure stable cellular rotation without inducing any morphological alterations in the cells. By strictly adhering to the aforementioned optimized parameters, the electrodes maintained their structural integrity and exhibited stable electrical impedance characteristics across multiple repeated experimental cycles. This targeted parameter optimization scheme effectively resolves the electrode corrosion issue while circumventing the fabrication and sealing difficulties associated with insulating coatings, ultimately guaranteeing the continuous and stable operation of the platform during high-throughput single-cell measurements.

Establishment of the Electroporation Sensitivity Index

To quantify the differential electroporation responses observed across tumor cell lines, we compared the changes in the four core electrical parameters before and after electrical stimulation at the single-cell level. Significant differences in electroporation sensitivity exist among different tumor cell lines, leading to difficulties in standardizing clinical therapeutic outcomes³². To address this, we established the Electroporation Sensitivity Index (EPI) based on the four cell lines studied, aiming to semi-quantitatively analyze the response mechanism of heterogeneous dielectric parameters to electroporation.

The key mechanism of electroporation lies in the accumulation of transmembrane potential (Δφ) across the cell membrane under an external electric field. The efficiency of Δφ formation is jointly determined by the impedance ratio of the cell membrane to the cytoplasm (Z_m/(Z_m + Z_c))³³, cell size (R)³⁴, and electric field intensity. Therefore, EPI was constructed by quantifying the synergistic effects of shunting effect, shielding effect, size effect, and membrane polarization, enabling the exploration of single-cell electroporation response mechanisms.

To quantify cellular dielectric heterogeneity, four key dielectric parameters were extracted. Given the disparities in parameter magnitudes, a ratio-based max-scaling normalization method was adopted. For each parameter, the value was normalized by dividing it by the global maximum within the current experimental cohort: X_std = X/X_max. This approach ensures that all normalized coefficients are confined within the range of [0, 1], thereby preserving the relative physical magnitudes while guaranteeing mathematical robustness for subsequent calculations. Specifically, the cell radii R were derived from experimentally measured microscopy data obtained during microfluidic experiments to accurately reflect the actual geometric dimensions of the distinct cell lines, as presented in Table 2:

Table 2.

Radius measurements for each cell line

Cell Line	HUH-7	U251	A549	B16
Radius (μm) (mean ± SD)	9.27 ± 1.36	8.69 ± 0.43	6.20 ± 0.57	7.20 ± 0.46

Definition of key coefficients

Shunting inhibition coefficient (S_h)

S_h quantifies the current-shunting effect attributable to cytoplasmic conductivity (σ_cyto). Conceptually, a cell with higher σ_cyto provides an easier parallel conductive path for the applied electric field, “shunting” current away from the charging process of the cell membrane. Thus, a larger σ_cyto yields a larger S_h value, denoting stronger inhibition of transmembrane potential induction. It is calculated directly from the normalized cytoplasmic conductivity of the control cells: S_h = σ_{cyto_std}.

Shielding inhibition coefficient (S_c)

S_c quantifies the shielding effect of the cytoplasm, which is determined by its relative permittivity (ε_cyto). A higher ε_cyto results in a larger S_c, denoting a stronger attenuation of the penetrating electric field. It is defined as: S_c = ε_{cyto_std}.

Size promotion coefficient (S_R)

S_R captures the size-dependent promotion of the induced transmembrane potential (Δφ), governed by the cell radius (R). A larger radius yields a larger SR value, consistent with the theoretical proportionality between transmembrane potential (Δφ) and cell size. It is calculated as: S_R = R_std, where R_std are the normalized values of membrane radius.

Membrane polarization coefficient (P)

P represents the membrane’s overall effectiveness in polarizing under an electric field. It combines the membrane’s insulating quality, inversely related to its conductivity (1-σ_{mem_std}), with its intrinsic polarizability (ε_{mem_std}). The composite formula is: P = (1 − σ_{mem_std}) × ε_{mem_std}. where σ_{mem_std} and ε_{mem_std} are the normalized values of membrane conductivity and permittivity in the control group, respectively.

Static intrinsic sensitivity (S)

The coefficient S integrates the membrane polarization capacity (P) with the three effects described above (S_h, S_c, S_R) to reflect a cell’s intrinsic electroporation sensitivity prior to stimulation. It is calculated as: $S=\frac{P\times S_R}{\left(1+S_h\right)\times \left(1+S_c\right)}$, a larger S indicates higher static intrinsic sensitivity.

Dynamic promotion coefficient (D)

The coefficient D quantifies how changes in cellular properties induced by the electrical pulse further promote electroporation. It is primarily based on the post-stimulation decrease in cytoplasmic conductivity (∆σ_cyto) and the change in membrane permittivity (∆ε_mem). D is defined as: D = (1 + ∆σ_{cyto_std})×(1 + ∆ε_{mem_std}), where ∆σ_{cyto_std} and ∆ε_{mem_std} are the normalized magnitudes of these changes. A larger D indicates a stronger dynamic, stimulation-induced promotion effect.

EPI calculation

Finally, to comprehensively quantify cellular susceptibility to electroporation, the EPI was formulated as the arithmetic sum of the static and dynamic components: EPI = S + D. This additive model is grounded in the cumulative contribution hypothesis of electroporation-induced lethality. In contrast to multiplicative models, this additive framework underscores the independence and complementarity of the two distinct biophysical processes; specifically, the ultimate cellular lethality is determined by the cumulative effect of the efficiency of transmembrane potential establishment (S) and the severity of membrane structural collapse (D). A higher EPI value predicts a higher overall cellular sensitivity to electroporation, which corresponds to the measured instantaneous mortality (IR).

Validation of EPI effectiveness and mechanism analysis

The validation results demonstrated a strong concordance between the EPI ranking and the actual maximum instantaneous mortality (IR). The EPI order (B16: 3.628 > A549: 2.678 > U251: 2.364 > HUH-7: 1.974) perfectly mirrored the IR ranking (B16: 82.28% > A549: 81.87% > U251: 78.81% > HUH-7: 60.61%). This close alignment confirms that the EPI index effectively quantifies differential electroporation sensitivity across cell lines.

To validate the robustness of the proposed EPI, a Leave-One-Out Cross-Validation approach was employed. In each iteration, one cell line was excluded from the cohort, and the EPI model parameters were recalculated based on the data from the remaining three cell lines. Subsequently, the excluded cell line was evaluated using this newly established EPI formulation. A total of four validation rounds were conducted by systematically excluding A549, HUH-7, U251, and B16, respectively. The resulting EPI indices are presented in Table 3:

Table 3.

Calculated values of EPI

Cell Line	B16	A549	U251	HUH-7
Remove A549	4.07	3.01	2.36	2.01
Remove HUH-7	3.66	2.68	2.35	2.00
Remove U251	3.73	2.74	2.52	2.11
Remove B16	3.64	2.67	2.32	1.98

As evidenced by the results in Table 3, the constructed EPI model demonstrates significant robustness while rigorously preserving the biophysical significance of each constituent parameter.

Furthermore, Table 2 highlights that cell size is not a uniform constant but exhibits natural biological heterogeneity. Given that the size effect is a critical determinant of the electroporation response, we employed a Monte Carlo simulation approach (10,000 iterations) to quantify the impact of this variability. In each iteration, the radius for each cell line was randomly sampled within the range defined by its mean and standard deviation. The analysis of the 10,000 simulated outcomes indicates that: despite the fluctuations in radius affecting the static sensitivity, the overall EPI ranking (B16 > A549 > U251 > HUH-7) remains statistically robust (Probability > 99%).

To ensure the cross-cohort transferability of the EPI and eliminate its dependence on batch-specific maxima, we propose a standardized absolute scaling protocol for future independent cell lines or primary tumor cells. The maximum values obtained from the current four cell lines encompass a broad spectrum of dielectric heterogeneity and are therefore permanently fixed as Universal Reference Denominators X_URD. Furthermore, the A549 cell line is designated as the system calibration standard. When evaluating any novel primary tumor cohort in the future, standard A549 cells must be tested in parallel on the same microfluidic chip batch as an internal reference.

To account for minor variations in chip electrode aging or buffer conductivity across different laboratories, a system correction factor, denoted as k, is first calculated based on the static cytoplasmic conductivity of the calibration cells:

$$k=\frac{{\sigma }_{cyto\_A549\_new}}{{\sigma }_{cyto\_A549\_original}}$$

Subsequently, the raw dielectric parameters of the new cell line are corrected by dividing them by k. These corrected values are then normalized strictly using the fixed X_URD values rather than the new cohort’s internal maxima. To prevent instances where an exceptionally high membrane conductivity in a new cell line exceeds the universal reference, which would result in a mathematically invalid negative value for the membrane insulation term (1 − σ_{mem_std}), a mathematical truncation is introduced. If the normalized coefficient X_{std_new} exceeds 1, it is capped at 0.99, representing a physical state in which the membrane insulation has degraded to its absolute extreme limit.

Correlational analysis between single-cell EPI and suspension mortality

To further substantiate the mechanistic link between single-cell dielectric alterations and suspension mortality, a comprehensive correlational analysis was conducted across diverse pulsed electric field conditions. Specifically, the individual Electroporation Sensitivity Index for single cells within the control group and the six experimental groups (G1 to G6) was systematically calculated and mapped against the corresponding macroscopic instantaneous mortality rates.

As illustrated in Fig. 11, the microscopic single-cell EPI values exhibit a highly consistent tracking trend with the macroscopic lethality across all four cell lines. The data distribution reveals that the single-cell experiments maintain excellent intra-group stability. More importantly, whether under standard irreversible electroporation waveforms or synergistic bipolar pulse stimulations, any variation in the applied electric field parameters that induces an elevation in the median single cell EPI invariably corresponds to a proportional increase in the suspension mortality. This strict qualitative and quantitative synchrony provides robust evidence that the microfluidic-derived EPI model accurately captures the underlying biophysical mechanisms governing cell death. It definitely proves that the magnitude of single-cell dielectric parameter shifts can serve as a highly reliable predictor for macroscopic electroporation outcomes.

Fig. 11.

Correlational analysis mapping single cell EPI distributions to macroscopic suspension mortality across varying pulsed electric field conditions. a B16F10. b A549. c U251. d HUH-7. The left vertical axis illustrates the microscopic single-cell EPI distributions, represented by box plots and individualized scatter points, alongside the median EPI trend line. The right vertical axis displays the macroscopic instantaneous mortality rate. Across the control and all six experimental pulse groups (G1 to G6), the median single-cell EPI exhibits a strict synchronous tracking trend with the suspension mortality. This profound qualitative and quantitative alignment explicitly demonstrates that the microfluidic-derived EPI accurately captures the underlying biophysical mechanisms governing cell death and possesses robust predictive capabilities for macroscopic electroporation therapeutic outcomes

Common response mechanisms of the four cell lines

All four cell lines shared two key dielectric responses: an increase in membrane conductivity (σ_mem) and a decrease in cytoplasmic conductivity (σ_cyto).

The rise in σ_mem is a direct signature of electroporation, where pulsed electric fields induce transient or permanent pores, transforming the insulating membrane into a more conductive structure^35,36. The sustained transmembrane potential expands these pores by reorganizing phospholipid polar heads, further enhancing permeability and conductivity³⁷. Consistently, the σ_mem of the control group was significantly lower than that at maximum IR (Table 1) which serves as the physical basis for membrane insulation (1 – σ_{mem_std}) in the membrane polarization coefficient (P). Cells with lower initial (σ_mem) (e.g., A549) exhibit stronger membrane insulation and higher static intrinsic sensitivity.

Membrane poration disrupts the distribution and balance of ions inside and outside the cell, allowing abnormal transport of ions (e.g., (K⁺), (Na⁺), (Ca²⁺), (Cl⁻)) and causing rapid changes in cytoplasmic ion concentration, which leads to a decrease in cytoplasmic conductivity. Cytoplasmic conductivity is closely correlated with ion concentration³⁸. This change is reflected in the EPI index as a positive contribution from (∆σ_{cyto_std}): B16 and U251 cells showed the largest decreases in cytoplasmic conductivity, which became the core factor enhancing their dynamic promotion coefficient (D).

Cell-type-specific differences in relative permittivity changes

In contrast, the changes in membrane relative permittivity (ε_mem) and cytoplasmic relative permittivity (ε_cyto) exhibited two distinct patterns among the cell lines:

A549 and B16 cell lines

In these cell lines, electroporation increased ε_mem. Pore formation facilitates massive influx of ions and water^39,40, while phospholipid rearrangement enhances interfacial polarization⁴¹, collectively raising the membrane’s relative permittivity. Consequently, positive ∆ε_{mem_std} values significantly boosted their dynamic promotion coefficient (D).

In the cytoplasm, ion loss results in a relative increase in the proportion of water molecules, and the dielectric response of certain polar molecules or protein structures may be enhanced, collectively manifesting as an increase in cytoplasmic relative permittivity. Reflected in the EPI index, A549 and B16 cells had low shielding inhibition coefficients (S_c), indicating weak static inhibition effects, and thus maintained high static intrinsic sensitivity (S).

HUH-7 cell line

Some researchers suggest that HUH-7 cells typically possess high initial cholesterol content, which could potentially stabilize the membrane structure and might result in the formation of fewer or smaller water channels during electroporation^42,43. Therefore, we hypothesize that such compositional characteristics might influence overall membrane stability, thereby restricting ion transport during the electroporation process⁴⁴. In the EPI index, HUH-7 had the highest shielding inhibition coefficient (S_c = 1.000). Concurrently, its moderate membrane permittivity and insulation properties failed to confer a significant static advantage (S = 0.159). More critically, the proposed inherent stability of the membrane structure likely constrained the dynamic response, resulting in the lowest dynamic promotion coefficient (D = 1.815). This dual limitation on both static and dynamic factors ultimately led to the lowest overall EPI score.

U251 cell line

For U251, it is speculated that specific biological characteristics, potentially including the expression of certain stem cell markers or active receptor signaling pathways, might contribute to enhanced membrane stability^45,46. We hypothesize that these unspecified biological factors could hinder extensive pore expansion during electroporation, which offers a possible explanation for the negligible change in membrane permittivity (∆ε_mem ≈ 0). However, our EPI model reveals that U251 actually possesses the highest static intrinsic sensitivity (S = 0.308). This is predominantly driven by its exceptionally high membrane relative permittivity (ε_mem = 13.6). Although U251 exhibits slightly elevated membrane conductivity, the substantial magnitude of its relative permittivity dominates the calculation of the polarization factor (P), thereby sustaining its superior static polarization capability. Conversely, its dynamic promotion coefficient remains relatively low (D = 2.056). This disparity between high static potential and restricted dynamic evolution resulted in its final ranking falling to third place.

Summary and future work

In summary, the electroporation response of cells is the result of the synergistic effect of their static intrinsic properties and dynamic response laws:

• B16 exhibited the lowest static intrinsic sensitivity (S = 0.10) due to their smaller geometric dimensions. However, they demonstrated the most significant dynamic response capability (D = 3.56). This dominant dynamic membrane instability effectively compensated for the geometric limitation, resulting in the highest EPI score (3.65), which is consistent with the highest observed mortality rate (IR = 82.28%).

• A549 maintained effective membrane insulation due to lower initial conductivity, which, combined with a relatively strong dynamic promotion coefficient (D = 2.56), yielded a high EPI score (2.67) and a correspondingly high mortality rate (IR = 81.87%).

• U251 displayed a distinct response pattern: they possessed the highest static intrinsic sensitivity (S = 0.29) driven by their high membrane relative permittivity. Nevertheless, the stability of their membrane structure constrained the further evolution of dynamic damage (D = 2.06), resulting in an intermediate EPI score (2.34) and mortality rate (IR = 78.81%).

• HUH-7 were constrained by both high cytoplasmic shielding effects and membrane stability, exhibiting low levels of both static sensitivity (S = 0.18) and dynamic promotion (D = 1.81). This dual inhibition led to the lowest EPI score (2.00), corresponding to the lowest mortality rate observed in the experiments (IR = 60.61%).

Although our single-cell dielectric measurements successfully captured the biophysical phenotypic changes induced by electroporation, the aforementioned biological mechanisms regarding lipid composition, cholesterol content, and specific receptor pathways remain entirely speculative. Furthermore, given that this microfluidic platform revealed the patterns of cellular heterogeneity under low-conductivity conditions, caution must be exercised when directly extrapolating these specific dielectric parameter thresholds or absolute EPI values to clinical high-conductivity environments. To overcome these limitations and deepen the understanding of the underlying mechanisms, future work must integrate the microfluidic platform with targeted molecular biological assays, including cholesterol staining, membrane order probes, and dynamic ion-flux analyses. Such a multidisciplinary approach is essential for directly validating these mechanistic hypotheses. Meanwhile, subsequent research will focus on exploring detection conditions that closely mimic physiological environments and establishing a comprehensive database linking diverse pulse parameters, single-cell dielectric responses, and macroscopic mortality outcomes, thereby providing a more robust translational foundation for precision electroporation therapy.

Conclusions

To address the core challenge in tumor electroporation therapy—significant intercellular heterogeneity in electrical responses leading to inconsistent ablation outcomes—this study developed and validated a microfluidic-based single-cell dielectric monitoring platform. By establishing the EPI, we achieved a mechanistic and quantitative understanding of single-cell electroporation responses. The main conclusions are as follows:

1. The integrated platform, combining DEP capture, EROT measurement, and PEF electroporation on a single chip, enables stable single-cell manipulation and precise dielectric characterization. Validation across four cancer cell lines (A549, HUH-7, U251, B16F10) demonstrated that the electrorotation spectra (pre- and post-electroporation) were fitted with a coefficient of determination (R²) exceeding 0.97 using a genetic algorithm, confirming the platform’s high accuracy and robustness.

2. The study elucidated the dynamic evolution of dielectric parameters in cancer cell lines before and after electroporation. Simultaneous cell suspension mortality experiments verified that the four core dielectric parameters (membrane conductivity, membrane relative permittivity, cytoplasmic conductivity, and cytoplasmic relative permittivity) exhibit strong correlations with cell mortality. This association provides a quantitative basis for predicting macroscopic therapeutic outcomes from microscopic dielectric heterogeneity.

3. To quantify the impact of single-cell heterogeneity, the EPI was constructed based on the cumulative contribution of static intrinsic sensitivity and dynamic promotion. The calculated EPI scores showed a robust positive correlation with the measured maximum instantaneous mortality (IR). Crucially, the model reveals distinct response mechanisms: it elucidates how dominant dynamic membrane instability can compensate for geometric disadvantages, while membrane stability can constrain therapeutic efficacy despite high static potential. This work successfully transforms the qualitative description of electroporation responses into a quantitative framework, providing a theoretical foundation for precision ablation through correlated micro-macro models.

Future work will focus on building a comprehensive database correlating a wider range of pulse parameters (intensity, width, waveform) with dielectric responses across multiple cell lines. By integrating big data and artificial intelligence, we aim to develop a universal predictive model to optimize treatment planning, thereby accelerating the clinical translation of electroporation for precise tumor therapy.

Materials and methods

Cell preparation

A549, HUH-7, U251, and B16F10 cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37 °C in a humidified atmosphere containing 5% CO₂. Cells were passaged at 80–90% confluence using 0.25% Trypsin-EDTA (Gibco, USA). Identical protocols were used for sample preparation across all cell lines.

For single-cell electrorotation experiments, cell suspensions were washed three times and resuspended in a low-conductivity isotonic buffer. The buffer composition was as follows: sucrose (242 mM), Na₂HPO₄ (5.5 mM), NaH₂PO₄ (3 mM), and MgCl₂ (1.7 mM). The measured specific conductivity of the buffer was 0.1 S/m at 25 °C.

Electroporation parameter settings

To establish a quantitative relationship between the specific electrical parameters of individual cells and their electroporation responses, the electric pulse parameters were maintained consistent throughout the experiments. The parameters of the experimental group are shown in Table 4.

• IRE waveform: Electric field strength = 1.5 kV/cm, pulse width = 100 µs;

• Synergistic Bipolar Irreversible Electroporation (SBIRE) waveform¹⁹: Simultaneous application of nanosecond bipolar pulses ( ± 8 kV/cm, ±500 ns, “+” to “−” delay = 1 µs, “−” to “+” delay = 200 µs, 5 cycles) and microsecond bipolar pulses (±1.5 kV/cm, 2 µs, “+” to “−” delay = 2 µs, “−” to “+” delay = 1000 µs, 25 cycles).

Fabrication of the microfluidic chip

The microfluidic chip was fabricated on a glass substrate through the following sequential processes: First, as shown in Fig. 12a, the glass substrate was cleaned via acid washing and deionized water rinsing to remove surface impurities, followed by drying and sputtering of a chromium (Cr) protective layer. Subsequently, a positive photoresist (Az1500) with good adhesion and appropriate thickness was spin-coated onto the treated substrate (thickness: 1 µm, rotation speed: 3000 rpm for 30 s), and pre-baked at 95 °C for 90 s. Then, as shown in Fig. 12b, using a custom-fabricated photomask placed between the substrate and the light source, the photoresist was selectively irradiated with ultraviolet (UV) light (wavelength: 365 nm, energy: 100–150 mJ/cm²) through the mask. Next, the exposed photoresist was removed using a developer solution, with a typical development time of 1–3 min. Following this, the microchannel was etched via a chemical reaction between hydrofluoric (HF) acid and glass, where the regions protected by the photoresist remained unetched at an etching rate of approximately 3 µm/min. Afterward, inlet and outlet holes were drilled on the glass chip using ultrasonic machining. The positive photoresist on the substrate surface was then stripped with acetone, and the Cr protective layer was cleaned off using a chromium etchant, followed by secondary cleaning of the substrate with hydrogen peroxide (H₂O₂). Subsequently, as shown in Fig. 12c, a gold (Au) layer (thickness: 70 nm) and an adhesion layer (Cr: 10–20 nm) were deposited via magnetron sputtering. To pattern the Au electrodes, the aforementioned steps (spin-coating, pre-baking, UV lithography, and development) were repeated using an electrode-specific photomask. As shown in Fig. 12e, the Au and Cr layers were then etched via wet etching to form four Au electrodes, which were used to apply the DEP field, EROT field, and PEF for electroporation, with a key inter-electrode distance of 60 µm and an electrode arc radius of 25 µm. The width of the microchannel is 100 µm. For bonding, low-temperature bonding technology was adopted: prior to bonding, oxygen plasma treatment was performed to enhance hydrophilicity, and the etched glass substrate (with microchannels) was bonded to the electrode-deposited substrate via alignment bonding (alignment accuracy ≤5 µm) at a temperature of 200–300 °C for 2–4 h. Finally, as shown in Fig. 12d, the bonded glass chip was diced, and PEEK fittings were attached to the inlet/outlet holes using AB epoxy. The completely assembled device is shown in Fig. 12f.

Fig. 12.

Fabrication process flow and structural characterization of the microfluidic chip. a Substrate preparation and cleaning. b Photolithography and wet etching for microchannel formation. c Magnetron sputtering for Au/Cr electrode deposition. d Bonding of PEEK connectors. e Geometric dimensions of the microchannel (width:100 μm) and quadrupole electrode array (gap: 60 μm). f Photograph of the fully assembled microfluidic device

Image data acquisition and processing

A low-conductivity buffer was injected into the microfluidic chip channel to thoroughly clean the internal channel, removing impurities and wetting the channel surface. The resuspended cell suspension was aspirated and slowly injected into the channel using a microfluidic pump. During injection, residual air bubbles in the pipeline were eliminated to avoid interference with subsequent single-cell manipulation.

The signal generator AFG31022 was activated to generate the DEP electric field for stable cell capture, ensuring the cell was positioned at the center of the electrodes. After successful capture, the AWG2184 was adjusted to output the EROT field to drive cell rotation. Images of cell rotation at each frequency were recorded synchronously, with three replicates performed for each cell under each experimental condition.

Image data processing was performed using Python 3.13: firstly, an adaptive contrast enhancement algorithm was applied to preprocess raw microscopic images, where contrast was adjusted adaptively in blocks to effectively improve the signal-to-noise ratio and suppress noise amplification; subsequently, Otsu’s adaptive thresholding method was used for globally optimal binarization of the enhanced images, achieving preliminary separation of the target cell region from the background based on the criterion of maximizing the between-class variance of pixel gray-level distribution; then, morphological composite operations (opening operation first to remove discrete noise and fine artifacts, followed by closing operation to fill voids in the cell region) were conducted to suppress background noise and enhance features of the cell region, ensuring the integrity of the cell contour; the cv2.minAreaRect() function was applied to the segmented target contour for minimum bounding rectangle fitting, where the rotation angle of the rectangle was used to obtain the cell’s principal axis direction while synchronously extracting the geometric center coordinates; further, an angle variation threshold was set to eliminate instantaneous perturbations, and robust estimation of the motion direction was realized based on the consistency judgment of directions across consecutive frames; finally, during the velocity calculation stage, a Hampel filter was used to smooth the instantaneous angle variation sequence, removing abrupt outliers to significantly reduce random errors and improve the long-term stability and repeatability of dynamic parameters.

The rotational speed of the cell was calculated as the total rotation angle divided by the rotation time over a 20 s video sequence. A genetic algorithm was used to fit the experimental and theoretical rotation spectra. Since genetic algorithms minimize the objective function, the optimization function was constructed as follows:

$$R^2=1-\frac{\sum {\left(y_{\exp }-y_{fit}\right)}^2}{\sum {\left(y_{\exp }-{ȳ}_{\exp }\right)}^2}$$

where y_exp is the experimental data and y_fit is the fitted data.

Setup of the cell suspension mortality assay platform

To establish a quantitative relationship between cell viability and single-cell electrical parameters, cell suspension mortality experiments were conducted using a self-made bipolar pulse generator. The system adopted a dual Marx generator configuration, consisting of: A nested high-voltage short-pulse module (maximum output voltage: ±12 kV, adjustable pulse width: 100 ns–1 µs); A low-voltage long-pulse module (maximum output voltage: ±3 kV, adjustable pulse width: 1–100 µs). High voltage pulses were applied across a 2 mm electrode cup containing the cell suspension. The voltage across the electrode cup was measured using a WavePro 760Zi-A oscilloscope (Teledyne LeCroy Inc., New York, USA) and a PPE-5 kV high-voltage probe.

Cell viability assay

100 µL of cell suspension was pipetted into a 2 mm electrode cup and subjected to electrical stimulation with either IRE or SBIRE waveforms. After stimulation, the suspension was transferred into 500 µL of complete medium and mixed thoroughly. The mixture was seeded into a 96-well plate with the following setup: 6 columns for experimental groups, 1 column for the control group, and 1 column for the blank group. Each group included 6 replicate wells, with 100 µL of the mixed cell suspension added to each well.

The 96-well plate was incubated at 37 °C in a humidified atmosphere with 5% CO₂ for 3–4 h. Subsequently, cell viability was assessed using the Cell Counting Kit-8 (CCK-8, Beyotime Biotechnology Ltd., Shanghai, China). 10 µL of CCK-8 reagent was added to each well, and the plate was incubated at 37 °C for 1–3 h in the dark. Absorbance was measured at 450 nm using a multimode microplate reader (Varioskan LUX, Thermo Fisher Scientific Inc., Massachusetts, USA). For each group, data were expressed as the mean value (excluding maximum and minimum values), and the experiment was repeated three times.

The percentage of cell death was calculated using the following formula:

$$Percentage=\frac{\mathrm{O}{\mathrm{D}}_{\mathrm{C}}-\mathrm{O}{\mathrm{D}}_{\mathrm{S}}}{\mathrm{O}{\mathrm{D}}_{\mathrm{C}}-\mathrm{O}{\mathrm{D}}_{\mathrm{M}}}\times 100\%$$

Where OD_S, OD_C, and OD_M represent the absorbance values of the experimental, control, and blank groups, respectively.

Statistical analysis

To ensure statistical robustness and minimize biological variability, all experiments were performed using three independent biological replicates. This involved the use of three separately fabricated microfluidic chips and three distinct cell culture batches. For each of the seven experimental groups, a total of 12 single cells were analyzed. Specifically, to avoid clonal bias, four single cells were randomly selected from each of the three independent culture batches. Only cells exhibiting intact morphology were captured for subsequent dielectric characterization.

All results are presented as mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism 10.1.2 software with one-way analysis of variance (ANOVA). Significance levels are indicated as follows: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. A p-value < 0.05 was considered statistically significant.

To rigorously evaluate the statistical power given the targeted sample size of single-cell manipulations (n = 12), effect sizes using Cohen’s d and 95% confidence intervals (95% CI) were calculated for key pre-electroporation and post-electroporation comparisons. A Cohen’s d value greater than 0.8 is considered a large effect, confirming that the magnitude of the electric-field-induced dielectric shift significantly outweighs intrinsic biological variance.

Author contributions

L.Z., Y.W., Y.S. and S.X. conceived the study, designed the experiments, and wrote the manuscript. B.Z. performed data processing and analysis. Y.Z. and H. Liu prepared the figures and tables and contributed to manuscript writing. L.Z. and Y.W. conducted the biological mechanism analysis. C.Y. and S.D. supervised the study, guided the experimental design, and edited the manuscript. All authors reviewed and approved the manuscript.

Data availability

Due to the confidentiality agreement with the collaborating institution, the raw experimental data cannot be publicly shared. Researchers interested in accessing the data should contact the corresponding author (dsl@cqu.edu.cn) with a detailed research plan, and access will be granted following the completion of a data transfer agreement.

Competing interests

The authors declare no competing interests.