Investigating membrane nanoporation induced by bipolar pulsed electric fields via second harmonic generation

After the initial success of employing electric-field-induced membrane poration to perform gene transfection and enhance drug delivery, researchers began looking at ways of improving the process. One approach to enhancing this technique was an alteration in the pulse's construction.^1–4 Because monopolar (MP) pulses incur a polar response in the cell, preferentially porating the anodic-facing side of the membrane, one hypothesis was that a pair of temporally spaced pulses of opposite polarities could increase the insult's efficacy. Arena et al. predicted that these bipolar (BP) pulses could also achieve greater penetration depths in tissue while reducing joule heating, making it an ideal candidate for a number of therapies.^4,5

Experimentally, this hypothesis has been confirmed for longer pulses (>100 μs).¹ However, efforts to reproduce these effects using shorter, sub-μs pulse lengths have been unsuccessful and have even demonstrated a diminished effectiveness of BP nanosecond pulsed electric fields (nsPEFs) to induce membrane permeabilization and ion uptake.⁶ Even when using nsPEF parameters usually associated with induction of cell death, the opposite behavior was observed, and the survivability with respect to the MP case was improved.^6,7 Follow-on work showed that calcium uptake by cells exposed to BP pulses was also reduced as compared to a single MP pulse of equivalent amplitude.^8,9

There are several theories as to why this behavior might occur for shorter duration pulses.^6–8 Given the brevity of each pulse, it is likely that the second pulse mitigates the effects of the first, while the first pulse biases the cell membrane against the second. One method of elucidating the subject would be to analyze how the structural changes in the membrane vary depending on the refractory period between each constituent MP pulse. To best determine how differences in the pulse's construction may manifest themselves as structural changes in the membrane, though, we need an accurate means of gauging membrane disruption during the exposure.

Recently, we developed a measurement technique based on nonlinear optical probes that is capable of determining the extent and intensity of nanopore development throughout the cell membrane. It identifies the transition region in nsPEF-exposed cells where porated and unporated regions of the membrane meet and encapsulates this region in a metric called the critical angle, θ_c.^10,11 The measurement technique uses a lipophilic probe, Di-4-ANEPPDHQ (Di-4), and second harmonic generation (SHG) to determine if an area of membrane has been porated. The SHG signal is strongest when the probes are aligned in an intact membrane. However, if the cell's plasma membrane becomes porated, the probe's environment is no longer noncentrosymmetric and the probes would not be optimally aligned for SHG, resulting in a drop in signal. Therefore, we can map poration around the cell and determine the extent to which nanoporation has occurred as a function of the nsPEF exposure time and pulse structure.¹¹ More information is available in the supplementary material.

In this study, we use our SHG imaging method to characterize MP- and BP-induced nanoporation in Jurkat clone E6-1 human T-lymphocytes. A custom, laboratory-built pulse generator delivers the MP and BP nsPEFs to a sample stage containing specialized tungsten wire electrodes spaced approximately 150 μm apart (end-to-end). We measure the average value of the SHG signal across the membrane and compare its value before and after the pulse is applied. Complimentary control measurements (no electric field exposure) were also performed. More information on cell culture and a conceptual diagram of the experiment and image analysis are available in the supplementary material.

MP pulses with a duration of 300 ns are used as a benchmark for cellular effects. The standard BP exposure consists of a 300 ns positive-leading-edge square pulse followed by a 10 ns delay and then a 300 ns negative-leading-edge square pulse (shown in Figure 1(h)). To test our theory that the brevity of nsPEFs combined with a minimal refractory period plays a significant role in the dynamics of the biophysical response to BP pulses, we perform a systematic investigation into the dependence of nanopore density on the temporal spacing between pulses. The spacing is varied from 10 ns to 20 μs, covering over three orders of magnitude, and the response is compared to the monopolar case. The pulse amplitude is held constant at 16.2 kV/cm.

Figure 1 contains representative SHG images of a Jurkat cell before and after exposure to a MP pulse, a BP pulse with a 10 ns refractory period, and a BP pulse with a 10 μs refractory period. In the monopolar case, the signal loss can clearly be seen on the anodic-pole. The same cannot be said for the bipolar case, despite doubling the energy deposited. There are no concentrated, easily discernable losses in signal but rather an identically muted effect on both sides of the cell. As the refractory period of the bipolar pulse is increased, the effect of the pulse is increased, again, equally on both sides of the cell.

A more quantitative analysis can be performed by using the percent difference before and after exposure to determine θ_c and the relative increase in the number of pores at either pole. As shown previously, this change can be measured against the average standard deviation of control exposures (no field application) to determine where appreciable nanoporation has occurred.^10,11

In previous work, we compared our experimental results to a circuit model of the cell to determine how much of the cell's response could be attributed to the electrical behavior of the membrane.^11,12 We employ the same approach here. Since the selection criteria of SHG eliminates out-of-plane noise and our image is focused at the widest part of the cell, our model simulates this spherical zone. By further partitioning this central region, we generate spherical caps that can be approximated as planar membrane patches. Each patch can then be modelled to describe the evolution of the transmembrane voltage in that region as the pore density changes during nsPEF exposure. Additional details on the model are available in the supplementary material. Because we are modeling only isolated patches of the membrane, the simulation is susceptible to an overestimation of pore density. While this has not proven to be a significant source of error in our previous work, other research studies using similar models have cautioned that an isolated patch model could overestimate poration by 20%.^11,13

The results of this analysis and the corresponding experimental data are available in Figure 2 with the anodic-facing pole of the cell in Figures 2(a) and 2(b) and the cathode-facing pole in Figures 2(c) and 2(d).

As expected, in the MP case, the anodic-facing pole of the cell sustains the most membrane damage in both θ_c and maximum pore density. This preference could be attributed to the resting transmembrane potential, which, at −55 mV for Jurkat cells, effectively gives this pole a “head start” on its way to hyperpolarization.^13–15 The experimental data also confirms findings that BP pules with limited delay (∼ns) between the polarity reversal are less effective in destabilizing the membrane than their MP counterparts.^6–9 Even with a refractory period of 20 μs, the damage on the anodic pole of the cell never reaches that of the MP case. When these results are taken together, it appears that the second pulse inhibits, rather than enhances, the cell's response to the first pulse.

The model data matches well with the experimental data in both the MP case and the BP case when the refractory period is 10 ns. However, the results differ once the refractory period increases beyond 10 ns. After this spacing, the model predicts that the anodic-facing pole should experience more nanoporation as the refractory period is increased, eventually matching (and slightly exceeding) that of the MP case. It is important to note that the model assumes that the system is unchanged throughout the pulse, which the experimental data shows this is clearly not the case.

Interestingly, the experimental data indicates that nanoporation is mitigated at an increased rate until the refractory period reaches 10 μs. At this point, both θ_c and maximum pore density peak and approach values more closely resembling the model's idealized prediction of increased nanoporation with increased refractory period. The suppressed nanoporation response could be attributed to a biochemical process initiated by the first pulse that prevents the cell from experiencing further membrane damage. This idea is bolstered by the fact that 10 μs is enough time for K⁺, Na⁺, Cl⁻ and Ca²⁺, four ions of critical importance to transmembrane potential, to diffuse across the ∼10 nm cell membrane. Conversely, the frequency oscillations associated with a 10 μs spacing could match a resonance of either the membrane itself or the intracellular/extracellular solutions.

Shifting to the other side of the cell, our experimental data confirms that the cathodic-facing pole experiences minimal nanoporation when exposed to a MP pulse. Again, this result could be due to the resting membrane potential, as the pole of the cell facing the cathode must depolarize before poration.^13,15 But from the first BP pulse on, this pole experiences marked nanoporation. The effect does not occur to the same extent that a single MP pulse has on the anodic-facing pole, but a diminished effect is expected given the preconditioning of the first pulse.

The simulation and experimental results do match fairly well, but there is a discrepancy between the two in the MP case. While the simulation predicts a smaller θ_c and smaller maximum SHG signal loss on the cathode-facing pole than the anode, the difference is not as pronounced as it is in the experimental case. Other simulations have experienced similar problems in that the pore density, although diminished as compared to the anode-facing pole, does not demonstrate a prominent (greater than 15%) differential.^13–15 We attribute part of this variance to the isolated nature of our membrane patch model, but it is also likely that this process is more complex than one pole or the other receiving an electric “head start.”

In the BP cases, the model again predicts that nanoporation should increase in both extent and intensity as the refractory period is increased. The results agree very well for exposures with refractory periods of 10 ns and 1 μs. Beyond this temporal spacing, nanoporation increases to a plateau that the experimental results never reach. However, as mentioned earlier, the longer the refractory period, the more time there is for the system to change before experiencing the second pulse, something that cannot be accounted for in the model. Therefore, the discrepancy is somewhat expected. Overall, the experimental data for the cathode-facing pole agrees with our simulation, suggesting that its behavior is primarily governed by its electric characteristics. This response is in a stark contrast to the anode-facing pole. It is worth noting here that additional control experiments were conducted where the electrode polarity was reversed and the effect remained.

The discrepancy in membrane damage between the two poles could be due to the absence of lower frequency components in BP pulses, as, theoretically, nanoporation is directly related to the electrical characteristics of the pulse and membrane.^13,14,16 A Fourier analysis of representative pulses used in our experiment demonstrates that the majority of the energy lies near DC for the MP pulse (Figure 3). However, in the BP case with a 10 ns refractory period, the dominating frequencies are shifted roughly 1.5 MHz higher. This analysis indicates that the DC component of the pulse may be necessary for nanoporation. One way to satisfy this constraint may be to increase the temporal spacing between pulses. As shown in Figure 3(a), when the refractory period is increased, the average frequency content shifts again, returning towards DC.

It is worth noting, however, that this shift is not a perfect reconstitution of the MP spectrum. Oscillations are introduced by the second pulse that correspond with the temporal spacing. This concept can be demonstrated by comparing the frequency components of a BP pulse with a refractory period of 1 μs (Figure 3(b)) to that of a 2 μs refractory period (Figure 3(c)). As previous studies have employed 600 ns BP, we have also included a comparison showing that, while these longer constituent pulses have a larger frequency content nearer to DC, there is still a depression of the components within 0.5 MHz.

The frequency analysis and experimental results bolster the hypothesis that nanoporation proliferation is dependent on both the pulse parameters and the relaxation time after nsPEF exposure.⁸ The theory could play a large role in explaining the behavior of the anodic-facing pole. The time between pulses must be increased to a point where the second pulse does not interfere with the first, but not so long that the system has changed sufficiently as to diminish the effect of the second pulse. The cathode-facing pole, however, is the last to charge and therefore does not experience any mitigation of its relaxation time, which explains why the behavior of the cathode-facing pole more closely resembles model predictions.

In conclusion, we have employed a previously developed optical imaging technique for measuring membrane structural integrity to investigate and quantify cell membrane nanoporation associated with shorter pulse width (100 s of ns) BP pulses. The results provide key insight on why BP pulses have proven less effective at inducing cell death and provide direction for future work in a number of fields, both experimentally and theoretically. A comparison of the theoretical and experimental data shines a light on several biophysical and electrical properties governing membrane potential during exposure to electrical stimuli. This has clear implications for electroporation research, but it is also important for understanding how active cell types like neurons, muscle cells, and some endocrine cells can be manipulated to treat diseases and develop more advanced human machine interfaces.

See supplementary material for detailed information on the experimental methodology, cell culture protocol, and cell circuit model.

E.K.M. would like to thank Colin Vaz for helpful conversations. This work was supported by the Air Force Office of Scientific Research's Laboratory Research Independent Research (LRIR) Grant (No. 16RHCOR348). E.K.M. performed this research under the 711 HPW Repperger Summer Research Internship program led by Dr. Morley Stone and Dr. David Luginbuhl. E.K.M. also thanks the Oak Ridge Institute for Science and Education (ORISE) and the Achievement Rewards for College Scientists (ARCS) Foundation for their funding and support.