Pitfalls in assessing microvascular endothelial barrier function: impedance-based devices versus the classic macromolecular tracer assay

The most frequently used parameters to describe the barrier properties of endothelial cells (ECs) in vitro are (i) the macromolecular permeability, indicating the flux of a macromolecular tracer across the endothelium, and (ii) electrical impedance of ECs grown on gold-film electrodes reporting on the cell layer’s tightness for ion flow. Due to the experimental differences between these approaches, inconsistent observations have been described. Here, we present the first direct comparison of these assays applied to one single cell type (human microvascular ECs) under the same experimental conditions. The impact of different pharmacological tools (histamine, forskolin, Y-27632, blebbistatin, TRAP) on endothelial barrier function was analyzed by Transwell® tracer assays and two commercial impedance devices (xCELLigence®, ECIS®). The two impedance techniques provided very similar results for all compounds, whereas macromolecular permeability readings were found to be partly inconsistent with impedance. Possible reasons for these discrepancies are discussed. We conclude that the complementary combination of both approaches is highly recommended to overcome the restrictions of each assay. Since the nature of the growth support may contribute to the observed differences, structure-function relationships should be based on cells that are consistently grown on either permeable or impermeable growth supports in all experiments.


Impedance analysis of cell-covered gold-film electrodes using ECIS ® or xCELLigence ®
The electrical tightness of cell monolayers is commonly quantified by the area-specific transendothelial or transepithelial electrical resistance (TER), given in Ω⋅cm². In this study we used impedance measurements of cell-covered gold-film electrodes to measure the electrical tightness of the cell layers under study, given as normalized impedance |Z| (ECIS ® ) or as normalized cell index CI (xCELLigence ® ). The numerical values of both quantities are dependent on the AC frequency that is used for data recording. Moreover, they contain contributions from the electrode/electrolyte interface and the resistance of the bulk electrolyte. Thus, the measured impedance |Z| is never identical to the TER, but it may be correlated when the measurement parameters are properly selected. In this study we have chosen 32 kHz as monitoring frequency in ECIS assays as it provides a close-to-linear correlation between the measured impedance and TER values as shown below. Since it is difficult to show the correlation between TER and |Z| experimentally, we have performed model calculations to verify the correctness of the sampling frequency: There are two non-redundant physical models describing the impedance of cell-covered electrodes discussed in the literature. Both models describe the impedance of the electrode/electrolyte interface by a constant phase element (CPE) in series to a resistor (R bulk ) accounting for the solution resistance.
In the simplified model (A) the cell layer is represented by an RC-element representing transendothelial resistance (TER) and capacitance. 1 The second, more precise model of the cell layer (B) comprises a non-ideal impedance element to account for the resistance arising in the cell-electrode junction. This resistance is often referred to as cleft resistance and quantified by a parameter α or α 2 .
The resistance of the cell-cell junctions is referred to as R b (resistance between cells). The cell membrane is accounted for by the membrane capacitance. 2,3 The TER value from model A integrates over the contributions from R b and α in model B.
From the pool of our experimental data we have precise knowledge about the parameter values of either model for cell layers like the ones studied here. 2,3 To verify that impedance readings at 32 kHz are suitable for the monitoring of endothelial barrier function, we have calculated the impedance |Z|(32 kHz) as a function of TER and R b . The capacitance of the apical and basal membrane was set to 1 µF/cm 2 as it has been found throughout the literature and from the pool of our data. This value is a direct consequence of the unfolded membrane topography in endothelial cells which will not change along the experiments performed here. The cleft resistance (model B) was set to α 2 = 25 Ωcm 2 as derived from experimental spectra.
The figure below shows the correlation between |Z|(32 kHz) and TER (Model A) as well as the correlation between |Z|(32 kHz) and the resistance between cells R b (Model B). The data covers TERs and R b s from 0 to 10 Ωcm 2 , which is significantly more than the range of values that are relevant for our studies.
Both diagrams prove a close-to-linear correlation between TER/R b and the impedance measured at an AC sampling frequency of 32 kHz. Numerical values of |Z| and TER are, however, not identical. The cell index CI, reported by the xCELLigence ® device, reads the impedance of cell-covered electrodes at 10 kHz, 25 kHz or 50 kHz. In either case, the frequency is close enough to 32 kHz that the same conclusion applies to xCELLigence ® experiments. It is noteworthy that the frequency has to be evaluated on a log scale such that 10 kHz and 50 kHz are still appropriate. The figure below shows the calculated, frequency-dependent impedance magnitude for a cell-covered ECIS ® electrode as they were used in this study (blue) relative to the cell-free electrode (black). The parameters used for the calculation are based on published values. 2,3 The impedance spectrum for cells that only spread on the electrode but do not express cell-cell junctions is shown in red for comparison. The impedance data confirms that meaningful measurements are only possible in the frequency range above 10 kHz.