Polystyrene nanoparticle exposure induces ion-selective pores in lipid bilayers

Abstract

A diverse range of molecular interactions can occur between engineered nanomaterials (ENM) and biomembranes, some of which could lead to toxic outcomes following human exposure to ENM. In this study, we adapted electrophysiology methods to investigate the ability of 20 nm polystyrene nanoparticles (PNP) to induce pores in model bilayer lipid membranes (BLM) that mimic biomembranes. PNP charge was varied using PNP decorated with either positive (amidine) groups or negative (carboxyl) groups, and BLM charge was varied using dioleoyl phospholipids having cationic (ethylphosphocholine), zwitterionic (phosphocholine), or anionic (phosphatidic acid) headgroups. Both positive and negative PNP induced BLM pores for all lipid compositions studied, as evidenced by current spikes and integral conductance. Stable PNP-induced pores exhibited ion selectivity, with the highest selectivity for K⁺ (P_K/P_Cl ~ 8.3) observed when both the PNP and lipids were negatively charged, and the highest selectivity for Cl⁻ (P_K/P_Cl ~ 0.2) observed when both the PNP and lipids were positively charged. This trend is consistent with the finding that selectivity for an ion in channel proteins is imparted by oppositely charged functional groups within the channel’s filter region. The P_K/P_Cl value was unaffected by the voltage-ramp method, the pore conductance, or the side of the BLM to which the PNP were applied. These results demonstrate for the first time that PNP can induce ion-selective pores in BLM, and that the degree of ion selectivity is influenced synergistically by the charges of both the lipid headgroups and functional groups on the PNP.

1. Introduction

Engineered nanomaterials (ENM) exhibit desirable properties that make them useful for a wide range of applications such as drug delivery systems, gene carriers [1, 2], biosensors [3, 4], imaging reagents [5, 6] and consumer products [7]. The increasingly widespread use of ENM raises questions about possible toxic effects. ENM toxicity has traditionally been investigated using animal (in vivo) or cell (in vitro) platforms that measure parameters such as lactic dehydrogenase (LDH) release and immune response [8, 9]. Moreover, it has been shown that inhaled ambient ultrafine particles can be found in heart, bone marrow, blood vessels and other organs [10–12]. These observations suggested that the particles could penetrate through the alveolar epithelium. Mechanisms by which ENM interact with, injure, and are transported across the alveolar epithelium are important in understanding health effects related to ENM [13]. However, for ENM to exhibit toxic effects, they must first interact with cell membranes, which are composed of a bilayer lipid membrane (BLM) with a variety of biomolecules that impart desired molecular functions [14]. Consequently, ex vivo assays that directly measure molecular interactions between ENM and model BLM can also provide important insights into ENM translocation across epithelial cell barriers and toxicity [15].

A principal role of cell membranes is to provide a selectively permeable barrier that defines cell boundaries and maintains the essential environment for cytoplasm and organelles. ENM have been shown to penetrate biomembranes dependent in part on ENM properties [16–18]. For example, more efficient cellular uptake has been reported for negatively charged nanoparticles as compared with positively charged nanoparticles [16] and negatively charged nanoparticles displayed a less efficient rate of endocytosis than positively charged nanoparticles [17]. A recent review of the influence of surface properties of ENM-cell interactions has indicated that, in general, uncharged ENM interact less aggressively with cells and positively charged ENM are most effective in crossing cell membranes [18]. Consistent with this trend, Yacobi et al. [19, 20] observed that transcellular trafficking of positively charged, amidine-terminated polystyrene nanoparticles (amidine-PNP) across primary rat alveolar epithelial cell monolayers was 20–40 times faster than that of negatively charged, carboxyl-terminated PNP (COOH-PNP), and that trafficking did not occur via endocytic pathways, suggesting that such translocation may take place by diffusion of PNP through the lipid bilayer of cell plasma membranes.

A second role of cell membranes is to regulate transport of specific ions and molecules. Ion-selective transport generates ion gradients, which in turn play essential roles in energy generation, signaling and other cell functions [21]. Ion-selective transport may involve membrane protein ‘channels’ with a specific sequence of amino acids, and extensive research has been conducted to identify molecular mechanisms that impart selectivity [22–26]. Besides this class of highly selective channels, there are also channels with mild selectivity. For example, members of the BCL-2 protein family that were reconstituted into BLM formed partially selective channels. A mild cation selectivity was observed for antiapoptotic BCL-2 (P_K/P_Cl = 2.4) and mild anion selectivity was found for proapoptotic molecule BAX (P_K/P_Cl = 0.3). These selectivities may reflect the positively charged residues of BAX and negatively charged residues of BCL-2 [27]. The permeability ratio calculated for voltage dependent anion-selective channel (VDAC) from outer mitochondrial membrane is usually quoted as 2:1 Cl⁻:K⁺ [28]. It has also been found that VDAC has a cation selective open state [29]. Although the mechanism of switching selectivity is unclear, charged side chains that face the lumen of the channel must influence the selectivity [28]. Moreover, evidence exists that highly polar complexes of polyphosphate and polyhydroxybutyrate may impart ion selectivity in certain bacterial channels [30]. In general, selectivity for cations may involve local negative charge, and selectivity for anions may involve local positive charge, within the channel. Such local charge characteristics are likely generated by specific amino acids as well as inorganic compounds (polyphosphate).

Based on recent observations that ENM can induce stable pores in BLM [15], and recognizing that both ENM and phospholipid headgroups of BLM can impart localized changes in charge distribution, we hypothesized that ENM-induced pores could exhibit ion selectivity, and that this selectivity could be influenced by the charge of the ENM and/or the phospholipid headgroup. To test these hypotheses, electrophysiology methods were used to characterize ion currents flowing across BLM exposed to either positively charged amidine-PNP or negatively charged COOH-PNP. The phospholipid headgroup charge was varied by forming BLM from dioleoyl phospholipids having zwitterionic phosphocholine, cationic ethylphosphocholine, and anionic phosphatidic acid headgroups. Our findings establish for the first time that ENM can induce ion-selective pores in BLM and that the ion selectivity can be influenced by the surface charge of both ENM and phospholipid headgroups. These novel findings provide insights into the mechanisms by which PNP may create pores in, and penetrate through, cell plasma membranes.

2. Materials and methods

2.1. Materials

The two types of PNP (20 nm diameter) used in this study were purchased from Invitrogen (Eugene, OR): Fluorescent Yellow-Green amidine-PNP (catalog number C27370) and Fluorescent Red COOH-PNP (catalog number F8786). Phospholipids (with the same lipid tail of 1,2-dioleoyl (DO) but different headgroups) were obtained from Avanti Polar Lipids (Alabaster, AL): 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycero-3-ethylphosphocholine (DOEPC) and 1,2-dioleoyl-sn-glycero-3-phosphate (DOPA), N-decane was purchased from Aldrich (St. Louis, MO).

2.2. Preparation of liposomes

Unilamellar liposomes were obtained by a conventional extrusion technique. Three lipid compositions were used for liposome preparation: DOPC alone, a 3:1 (w/w) mixture of DOPC:DOEPC, and 3:1 mixture of DOPC:DOPA. Dry lipids (1 mg) were hydrated in 1 ml of 10 mM KCl solution for 2 h with periodic stirring in a vortex mixer. After three cycles of freezing/thawing at −10/+30 °C, the suspension of multilamellar liposomes was passed 11 times through a 0.1 μm polycarbonate membrane (Whatman, Buffalo Grove, IL) using Avanti microextruder (Avanti Polar Lipids). All operations, excluding freezing/thawing, were carried out at room temperature. Liposomes were stored at +4 °C during the experiments.

2.3. Hydrodynamic diameter and zeta potential measurements

PNP size (hydrodynamic diameter) and surface charge (zeta potential) were determined at room temperature on a 90 Plus Particle Size Analyzer (Brookhaven Instruments, Holtsville, NY). Particle size distribution was analyzed by dynamic light scattering using 90Plus/BI-MAS software. Electrophoretic mobility of PNP and liposomes was measured with a ZetaPALS (Phase Analysis Light Scattering) software. Data were collected from 10 runs of 10 cycles per run in the presence of 10 mM KCl and either 100 μg/ml PNP or liposomes (with a lipid concentration of 13 μg/ml).

2.4. Characterization of PNP-induced pores in planar bilayer lipid membranes

Planar BLM were formed from a 10 mg/ml lipid solution in n-decane (Aldrich). The solution was painted across the 200 μm aperture of a Delrin cup (Warner Instruments, Hamden, CT). Both cis (voltage command side) and trans (virtual ground) compartments of the cup contained 10 mM KCl (unless otherwise noted). Either COOH-PNP or amidine-PNP were added to the cis side of BLM to a final concentration of 100 μg/ml or 50 μg/ml, respectively. All measurements were performed at room temperature.

The approach commonly used to characterize selectivity of ion channels was adapted for determination of ion selectivity of PNP-induced pores in BLM. The approach entails measuring reversal potentials under the influence of a transmembrane ion gradient. For this purpose, a 10:1 KCl concentration gradient in the trans-to-cis direction was established by bathing the cis compartment with 10 mM KCl and the trans compartment with 100 mM KCl. The liquid junction potential offset was compensated for prior to membrane painting [31]. PNP were then added to the cis compartment in order to minimize electrolyte-induced PNP aggregation.

2.5. Recording and data analysis

Currents flowing across BLM were recorded with an integrating patch-clamp amplifier (Axopatch 200A, Axon Instruments, Sunnyvale, CA). The cis compartment was connected to the CV 201A head stage input and the trans compartment was held at virtual ground via a pair of matched Ag/AgCl electrodes. Currents flowing through voltage-clamped BLM were low-pass-filtered at 10 kHz (−3 dB cutoff) using an eight-pole Bessel filter (902LPF, Frequency Devices, Ottawa, IL) and digitized at 1 kHz using pClamp9 software (Axon Instruments) and recorded after digitization through an analog-to-digital converter (Digidata 1322A, Axon Instruments). Using standard voltage conventions, positive clamping potentials are listed as potentials with respect to ground and positive currents are shown as upward deflections.

Noise in the conductance signal of unmodified BLM had an amplitude of 0.2 pA. Conductance events were identified automatically and analyzed using Clampfit9 software. Total charge transfer through BLM was estimated by integrating area under the current trace over time. Average charge transfer rate was calculated as the total charge divided by the time interval over which the charge was estimated.

Data were averaged using results from at least five independent experiments with the total duration of all records for each type of PNP lasting at least 160 min. Quantitative data are presented as mean ± standard error (n = number of observations). Student’s t-test was used for comparison of two group means, with P < 0.05 indicating a statistically significant difference. Minitab 15 software (Minitab, State College, PA) was used to perform locally weighted scatter smoother (LOWESS) analysis of the scatter plots.

3. Results

3.1. PNP size, aggregation and zeta potential

To rule out artifacts due to PNP aggregation, a sample of PNP-containing electrolyte (10 mM KCl) solution was removed from the cuvette after each BLM experiment and hydrodynamic size measured. No PNP aggregation was observed under the conditions tested. The average size measured for COOH-PNP was 16.8 ± 0.4 nm (n = 13) and for amidine-PNP was 20.5 ± 1.2 nm (n = 11). The size of COOH-PNP following our BLM studies was slightly different from 20 nm, the size specified by Invitrogen. However, the difference is probably due to the different methods used for the estimation of particles size, i.e. transmission electron microscopy (Invitrogen) and dynamic light scattering performed in our case. Nevertheless, the particle size analysis indicates that coalescence did not take place during the BLM experiments. Zeta potential measured in the same solution as used for BLM experiments was −26.5 ± 1.1 mV (n = 20) for COOH-PNP and +32.2 ± 0.8 mV (n = 40) for amidine-PNP. For comparison, the zeta potentials of COOH-PNP and amidine-PNP in water provided by Invitrogen are −60 mV and +37 mV, respectively, indicating that the presence of 10 mM KCl decreases the magnitude but maintains the polarity of zeta potentials of the PNP utilized.

3.2. PNP-induced pore formation in DOPC BLM

In the absence of PNP, DOPC BLM were stable with a noise level of 0.2 pA. After PNP exposure, a variety of transient current patterns indicative of pore formation were observed, including brief current spikes that returned to baseline (perhaps representing rapid resealing of induced pores) and relatively stable/persistent ion flow (perhaps representing induced pores without complete membrane resealing (integral conductance)). Conductance events varied considerably, both in terms of temporal pattern (transient spike vs integral or persistent conductance) and magnitude. Patterns of fluctuating integral conductance could, in principle, arise from either a single pore having a fluctuating diameter or superimposed multiple conductance events. A higher concentration (100 μg/ml) of negatively charged COOH-PNP was required to induce ion flow through BLM than that for positively charged amidine-PNP (50 μg/ml). Fig. 1 shows representative current records depicting PNP-induced conductance over several minutes for both COOH-PNP and amidine-PNP. Increasing the concentration of amidine-PNP to 100 μg/ml led to rapid BLM disintegration. At this concentration, noisy conductance would begin a few minutes after PNP exposure and BLM would rupture within seconds of the onset of conductance. Moreover, at this higher PNP concentration, it was more difficult to manually reestablish (repaint) BLM after breakage than at the lower concentration. For both types of PNP, ENM exposure induced relatively high amplitude currents (Fig. 1C). However, such high amplitude currents were observed in about 20% of the total length of records in the presence of amidine-PNP but in only about 4% of the total length of records in the presence of COOH-PNP. Statistical analyses of the low amplitude currents, which were observed in most experiments, are presented in Fig. 2. Average conductance induced by COOH-PNP was 116.1 ± 11.8 pS (n = 246), while that induced by amidine-PNP was 194.4 ± 10.5 pS (n = 130). Similarly, the average charge transfer rate in the presence of COOH-PNP (7.3 ± 1.3 pC/s (n = 60)) was less than half that in the presence of amidine-PNP (18.1 ± 2.9 pC/s (n = 92)), even though COOH-PNP concentration was two-fold higher than amidine-PNP concentration. These results indicate that negatively charged COOH-PNP are considerably less potent at inducing pores in BLM than are the positively charged amidine-PNP.

Fig. 1.

Current traces induced by interactions of PNP with DOPC BLM. Symmetric solutions of 10 mM KCl bathed both sides of BLM and a transmembrane potential of +100 mV was applied. Currents were induced by addition of 100 μg/ml of COOH-PNP (A) or 50 μg/ml amidine-PNP (B, C) to the cis side. Transition of current induced by amidine-PNP from low to high amplitude state is shown in C. Dotted line indicates zero current.

Fig. 2.

Analysis of low amplitude currents resulting from cis exposure of DOPC BLM to PNP having different surface functionalization. Average conductance (left) and average charge transfer rate at +50 mV (right) for pores induced by COOH-PNP or amidine-PNP are shown. * indicates significantly lower conductance and charge transfer rate.

3.3. Ion selectivity of BLM pores induced by PNP

Experiments were conducted under a 10:1 KCl gradient (trans/cis) to explore if PNP-induced pores might also show preference between the principal cation (K⁺) and principal anion (Cl⁻). Two main patterns of current traces were observed following exposure of the DOPC BLM to COOH-PNP under a 10:1 KCl gradient. In about two-thirds of the experiments, currents were observed at negative voltages (−100 mV), but changing the sign of potential led to resealing of BLM with no current observed over prolonged periods (Fig. 3A). In the remaining third of the experiments, pore(s) exhibiting high-amplitude currents were formed (Fig. 3B). Unlike COOH-PNP, positively charged amidine-PNP formed stable currents through BLM under both negative and positive voltages (Fig. 3C).

Fig. 3.

Interactions of PNP with DOPC BLM in the presence of 10:1 KCl gradient (trans/cis). Low (A) and high (B) amplitude currents induced by 100 μg/ml of COOH-PNP are shown for two applied potentials (+100 and −100 mV in A and +75 and 0 mV in B). Currents induced by 50 μg/ml amidine-PNP (C) are shown for different applied potentials between −50 and +100 mV.

The effect of applied voltage gradients on current magnitude (I) flowing across BLM was explored by varying transmembrane voltage (V) in 10 mV steps. Current was plotted as a function of voltage to construct I-V relationship and reversal potential (E_rev, the voltage at which current changed direction) was determined. To investigate the possibility of hysteresis, transmembrane voltage was stepped both from negative to positive (positive direction) and from positive to negative (negative direction). In the presence of COOH-PNP, E_rev obtained in the positive direction was 30.6 ± 2.5 mV (n = 13) and in the negative direction was 32.3 ± 3.1 mV (n = 11) (Fig. 4A). For positively charged amidine-PNP, E_rev obtained in the positive and negative directions were 19.2 ± 2.0 mV (n = 24) and 25.3 ± 4.2 mV (n = 9), respectively (Fig. 4B). For both types of PNP, the difference in reversal potentials measured in opposite directions was not statistically significant, indicating that hysteresis does not play a significant role. Averaging all data for each type of PNP gave reversal potentials of 31.4 ± 2.0 mV (n = 24) for COOH-PNP and 20.9 ± 1.9 mV (n = 33) for amidine-PNP. Even though these values are both positive, they are significantly different from each other, indicating that PNP surface charge influences its corresponding reversal potential.

Fig. 4.

Determination of ion preference for pores induced by COOH-PNP (A) or amidine-PNP (B). Current-voltage (I-V) relations were obtained during voltage transitions shown in Fig. 3C with the direction of applied voltage steps shown. Insets show expanded views of the current reversal regions, in which squares and solid line indicate voltage shifts from negative to positive, while triangles and dashed line indicate voltage shifts from positive to negative. The direction of the voltage shift did not significantly affect reversal potentials (E_rev) for either COOH-PNP or amidine-PNP.

The Goldman-Hodgkin-Katz equation (Eq. 1) gives the equilibrium potential across a cell membrane, taking into account all ions that can pass through the membrane:

$$E_{\mathit{rev}}=\frac{RT}{F}ln\frac{P_K{[K^{+}]}_{\mathit{trans}}+P_{Cl}{[{Cl}^{-}]}_{\mathit{cis}}}{P_K{[K^{+}]}_{\mathit{cis}}+P_{Cl}{[{Cl}^{-}]}_{\mathit{trans}}}$$ (1)

where E_rev is reversal potential or zero-current potential, P_j permeability for ion j, R the ideal gas constant, T absolute temperature and F Faraday’s constant [32]. Equation 1 was used to calculate the permeability ratio (P_K/P_Cl) from experimentally measured E_rev and known ion concentrations. Resulting P_K/P_Cl values were 5.0 for COOH-PNP and 2.8 for amidine-PNP, indicating that pores induced by negatively charged COOH-PNP exhibit greater preference for K⁺ than pores induced by positively charged amidine-PNP.

3.4. Surface potential of BLM

To investigate the influence of BLM surface charge on selectivity of PNP-induced pores, we tested different combinations of three phospholipids that had identical dioleoyl (DO) lipid tails but varied in headgroup charge: zwitterionic phosphocholine (PC), cationic ethylphosphocholine (EPC) and anionic phosphatidic acid (PA). Three phospholipid compositions were tested: 3:1 ratio of DOPC:DOEPC, DOPC alone, and 3:1 ratio of DOPC:DOPA. Liposomes prepared from these three compositions exhibited statistically different zeta potentials of +31.5 ± 1.4 mV (n = 60), −10.6 ± 0.5 mV (n = 139), and −47.9 ± 1.2 mV (n = 60), respectively (Fig. S1). The slightly negative zeta potential for zwitterionic PC is consistent with a previous report [33].

3.5. Influence of lipid charge on selectivity of pores induced by amidine-PNP

Addition of positively charged amidine-PNP (50 μg/ml) to the bathing media of BLM caused ion conductance in the form of spikes and integral conductance (Fig. S2) for all BLM tested. During periods of relatively stable integral conductance, we used two protocols to determine E_rev: changing applied voltage continuously at a fixed rate (Fig. 5) or changing voltage in a series of 10 mV steps (Fig. S3). Recorded currents (I) were plotted as a function of applied voltage (V) to determine I-V relationships, and E_rev was determined as the voltage at which the I-V curve intersected the zero current line. The E_rev values obtained from the voltage step applications were −31.3 ± 2.4 mV (n = 27), 20.9 ± 1.9 mV (n = 33) and 29.8 ± 1.9 mV (n = 24) for BLM formed from DOPC:DOEPC (3:1), DOPC alone and DOPC:DOPA (3:1), respectively. The E_rev values calculated from the continuously applied voltage protocol were −31.9 ± 2.5 mV (n = 8), 20.4 ± 1.7 mV (n = 39), and 28.7 ± 1.2 mV (n = 36) for BLM formed from DOPC:DOEPC (3:1), DOPC alone and DOPC:DOPA (3:1), respectively. These values of reversal potential are significantly different from each other for the three types of lipid composition. However, for each lipid composition, there was no significant difference between the values obtained using the voltage steps vs the continuous voltage ramp. The E_rev values were used with the Goldman-Hodgkin-Katz equation to estimate permeability ratios (P_K/P_Cl) of 0.2, 2.8 and 4.6 for BLM prepared from DOPC:DOEPC (3:1), DOPC alone and DOPC:DOPA (3:1), respectively. These results indicate that ENM-induced pores in BLM exhibit selectivity for K⁺ over Cl⁻ when BLM are formed from lipids with negatively charged headgroups, and that the pore selectivity for K⁺ is greater when BLM are formed from lipids having more negative headgroups. In the case of BLM composed of lipids with positive headgroups, ion selectivity of BLM pores showed preference for Cl⁻.

Fig. 5.

Reversal potentials assessed from I-V relationships due to amidine-PNP interactions with BLM as a function of BLM surface charge under 10:1 (trans/cis) KCl gradient. Currents were recorded as voltage was increased continuously over a period of 15 s from −100 to +100 mV. Arrow indicates the corresponding reversal potential for each I-V relationship. Dotted lines indicate zero current.

3.6. Influence of lipid charge on ion selectivity of pores induced by COOH-PNP

Exposure to 100 μg/ml COOH-PNP induced pores in BLM, as evidenced by current spikes as well as relatively stable integral conductance having various current amplitudes (Fig. S4). Fig. 6 shows current profiles recorded while the applied potential was continuously increased from −100 mV to +100 mV. Arrows indicate E_rev values for the three BLM lipid compositions studied. Reversal potentials determined using this protocol were 20.2 ± 0.4 mV (n = 12), 31.7 ± 0.8 mV (n = 43) and 37.9 ± 2.3 mV (n = 15) for the membranes composed from DOPC:DOEPC (3:1), DOPC alone and DOPC:DOPA (3:1), respectively. E_rev values were also determined by stepping the applied voltage in 10 mV increments, as shown in Fig. 3C. Reversal potentials were 19.4 ± 2.0 mV (n = 22) (P_K/P_Cl ~ 2.6), 31.4 ± 2.0 mV (n = 24) (P_K/P_Cl ~ 5.0) and 38.9 ± 1.1 mV (n = 19) (P_K/P_Cl ~ 8.3) for BLM composed of DOPC:DOEPC (3:1), DOPC alone and DOPC:DOPA (3:1), respectively. These E_rev values are all significantly different (p < 0.05) from each other. However, the E_rev values obtained for the same lipid composition but using different voltage application protocols (voltage steps vs the continuous voltage ramp) were not significantly different.

Fig. 6.

Reversal potential assessed as a function of BLM surface charge. Currents were induced by COOH-PNP interactions with BLM under a 10:1 (trans/cis) KCl gradient. Currents were recorded as voltage was increased continuously over a period of 15 s from −100 to +100 mV. Arrow indicates the corresponding reversal potential for each I-V relationship. Dotted lines indicate zero current.

3.7. Effects of conductance on ion selectivity

PNP-induced ion conductances varied in magnitude from pS to nS. The hypothesis that ion selectivity is controlled by electrostatic interactions between the ion and charges lining the narrow pores suggested that E_rev values might be correlated with pore size, and hence conductance. To investigate a possible correlation, we plotted E_rev as a function of conductance calculated as the slope of the current vs voltage (I-V) curves. Several simple correlation models (linear, exponential and polynomial) were tested, but none gave a good fit. Next, we used locally weighted scatter plot smoother (LOWESS) analysis to fit an arbitrarily shaped interpolation curve to the data. At each of the n points in the data set, a low-degree polynomial is fit to a subset of the data, with explanatory variable values near the point whose response is being estimated. The polynomial is fit using weighted least squares, giving more weight to points near the point whose response is being estimated and less weight to points further away. The value of the regression function for the point is then obtained by evaluating the local polynomial using the explanatory variable values for that data point. The smoothing process is considered local because each smoothed value is determined by neighboring data points defined within the span (in our case 50%). The LOWESS fit is complete after regression function values have been computed for each of the n data points, so n different lines are fit. Finally, the fitted points are connected to produce a LOWESS curve [34, 35]. Fig. 7 shows an example LOWESS curve for amidine-PNP interactions with DOPC BLM. A peak in the LOWESS curve was apparent in five cases out of six (encompassing two types of PNP and three lipid compositions). However, the LOWESS curve for the sixth case, COOH-PNP and BLM composed of DOPC:DOPA (3:1), did not yield a recognizable peak.

Fig. 7.

Reversal potential (E_rev) vs conductance data for stable pores induced in a DOPC BLM by amidine-PNP under a 10:1 (trans/cis) KCl gradient; E_rev values were estimated from I-V curves; (●) denotes experimental data, and solid line denotes LOWESS interpolation curve generated by Minitab software (degree of smoothing = 0.5).

3.8. Effects of side of BLM to which PNP were applied on ion selectivity

Experiments were conducted to determine whether E_rev values varied with the side of the BLM to which PNP were added (trans vs cis side) with a 10:1 KCl gradient. To avoid artifacts due to KCl-induced PNP aggregation, stability of the PNP dispersions was measured for 10 and 100 mM KCl and duration of 5 h. Neither amidine-PNP nor COOH-PNP aggregated significantly in 10 mM KCl, and COOH-PNP did not aggregate significantly in 100 mM KCl, whereas the average particle size of the amidine-PNP increased significantly in 100 mM KCl. Based on these observations, only the COOH-PNP were deemed stable against aggregation at both concentrations of KCl. Ion selectivity was then measured when the COOH-PNP were added to the trans side (100 mM KCl) vs the cis side (10 mM KCl). E_rev measured with COOH-PNP added to the trans (100 mM KCl) compartment was 29.7 ± 1.2 mV (n = 33) (Fig. 8), which is not statistically different from that with COOH-PNP added to the cis (10 mM KCl) compartment (31.4 ± 2.0 mV (n = 24)).

Fig. 8.

Effects of COOH-PNP exposure on the cis or trans side of DOPC BLM under a 10:1 (trans/cis) KCl gradient. (A) I-V plot obtained from analysis of currents resulting from protocol described in Fig. 3C. (□) denotes PNP on cis side; (○) denotes COOH-PNP on trans side. (B) Average reversal potentials assessed from I-V relationships due to COOH-PNP on cis or trans side of BLM.

4. Discussion

The mechanisms by which ENM induce BLM pore formation are not well understood. Efforts to characterize effects of ENM on intact biomembranes are hindered by the inherent complexity and variability of biomembranes, as well as the diversity of ENM across many dimensions, including size, shape, stiffness, net charge, reactivity, spatial arrangement of hydrophobic and hydrophilic groups, tendency to aggregate and degree of particle heterogeneity within the sample. The possible involvement of many molecular species and/or interaction mechanisms makes it difficult to unambiguously test hypotheses and interpret experimental results mechanistically using intact biomembranes.

In contrast, the bottom-up biomimetic approach used in this study allowed exploration of the interactions of two homogeneous populations of well-characterized PNP with positive (amidine) vs negative (COOH) termination on their surfaces with synthetic BLM having three different known compositions. Because the only significant difference between the two PNP populations is the surface functional group (and therefore surface charge), the effects of this variable on BLM poration could be investigated in the absence of other complicating factors. Application of sensitive electrophysiology methods historically used to characterize properties of biomembrane pores induced by ion channel proteins provided detailed current-time “fingerprints” (Figs. 1 and 3) whose single-pore sensitivity and rapid dynamics yielded a rich source of information about underlying molecular interactions. Statistical analyses of these data (Figs. 2 and 4) allowed differences in current flows induced by different PNP to be quantified. Methods to extract additional information from these current profiles are still being refined.

Amidine-PNP induced pores at half the concentration required for COOH-PNP. In addition, relatively high-amplitude currents were observed more frequently for amidine-PNP, even at a lower concentration. Quantitative analysis of the low-amplitude currents indicated that average conductance and charge transfer rate for positively charged amidine-PNP were about twice those for negatively charged COOH-PNP, despite the lower concentration of amidine-PNP. These trends indicate that surface charge is an important variable in ENM-induced poration of BLM, and that positively charged amidine-PNP are more potent in this regard than negatively charged COOH-PNP. These conclusions are consistent with results from other studies that positively charged nanoparticles are more likely to cause toxicity than negatively charged nanoparticles [36], including the observation that positively charged PNP have greater flux across primary rat alveolar epithelial cell monolayers [19] and Madin-Darby canine kidney II cell monolayers [37] than negatively charged PNP. Our results also suggest that nanotoxicity could be triggered by direct molecular interactions between PNP and the cell plasma membrane, without the need for PNP uptake via endocytosis or other protein-mediated transport mechanisms that cannot occur in our simple biomimetic platform. This finding supports our previous assertion that PNP may pass through epithelial cell membranes by ‘diffusion’ [38]. However, direct biomembrane disruption by ENM could trigger complex cascades of events leading to a variety of endpoints associated with nanotoxicity, including cell death, LDH release, cytokine induction, DNA mutation, expression of genes in toxicity-associated pathways and production of reactive nitrogen/oxygen species.

In general, simple pores formed in BLM would not be expected to exhibit ion selectivity due to the absence of the precise structures of the type responsible for ion selectivity in ion-selective channels comprised of specific protein(s). However, Antonov et al. [39] reported the selectivity order of monovalent cations for soft perforation pores generated at the BLM’s phase transition temperature. They suggested that the hydration force is responsible for cation selectivity of planar BLM at soft perforation. Our results demonstrate that some PNP-induced pores do exhibit ion selectivity, and that this selectivity is influenced by both PNP surface charge and BLM phospholipid headgroup (Table 1), confirming our original hypothesis. To the best of our knowledge, this study is the first to show that ENM can induce ion-selective pores in biomembranes and that both ENM surface charge and phospholipid headgroup charge can significantly influence the selectivity.

Table 1.

Influence of both PNP surface charge (listed in left hand column) and BLM phospholipid headgroup charge (listed in top row) on selectivity (P_K/P_Cl) of the pores induced by PNP^a

	DOPC:DOEPC (ζ = +31.5 ± 1.4 mV)	DOPC (ζ = −10.6 ± 0.5 mV)	DOPC:DOPA (ζ = −49.7 ± 1.2 mV)
Amidine-PNP (ζ = +32.2 ± 0.8 mV)	0.2	2.8	4.6
COOH-PNP (ζ = −26.5 ± 1.1 mV)	2.6	5.0	8.3

^aZeta potential (ζ) is shown in parentheses.

Discussion of possible mechanism(s) responsible for ion selectivity of PNP-induced pores can draw a common theme from research on ion-selective channel proteins that selectivity for an ion is due to a small pore having a precise spatial orientation of oppositely charged groups. For ion channel proteins, the spatial orientation and charges responsible for selectivity are provided by amino acids and inorganic polyphosphate complexes [30]. By analogy, we hypothesize that for PNP-induced pores in BLM, charges responsible for ion selectivity are contributed by both the PNP and the lipid headgroups. The consistent trend of increasing P_K/P_Cl values with more negative PNP and lipid headgroup charges supports this hypothesis. Moreover, the effects of the PNP and lipid headgroup charges appear to be synergistic. The highest selectivity for K⁺ (P_K/P_Cl ~ 8.3) was obtained when both the lipids and PNP were negatively charged, and the highest selectivity for Cl⁻ (P_K/P_Cl ~ 0.2) was obtained when both the lipids and PNP were positively charged. The observed ranges of P_K/P_Cl values provide further insight into the relative impact of PNP charges vs lipid headgroup charges in determining ion selectivity of PNP-induced pores. For positively charged amidine-PNP, both anion and cation selectivity were observed, in that the permeability ratio ranged from P_K/P_Cl ~ 0.2 for positively charged DOPC:DOEPC, indicating anion preference, to P_K/P_Cl ~ 4.6 for negatively charged DOPC:DOPA, indicating cation preference. On the other hand, for negatively charged COOH-PNP, the P_K/P_Cl values ranged from 2.6 to 8.3, indicating cation preference for all lipids tested. In this latter case, the influence of PNP surface charge appears to dominate over the lipid headgroup charge in determining ion selectivity.

Whereas many ion-selective channel proteins have a highly specific and conserved amino acid sequence and channel architecture that results in consistent and reproducible conductance and selectivity traits, PNP-BLM interactions result in a wide variety of conductance events that vary significantly in their duration and conductance magnitude. This variety is to be expected, given the fluidity of BLM and the possibility of multiple modes of interaction between PNP and BLM, including PNP aggregation at, or in, BLM to form an ion-transporting complex [40]. Alternatively, PNP could cause a phase transition in BLM at the point of contact [41], which could trigger pore formation due to structural rearrangement from the liquid crystalline state to the gel state [39].

Results of this study were not able to establish a simple, statistically significant correlation between pore size (conductance) and the ion selectivity. The occurrence of a peak in several of the LOWESS curves (e.g., Fig. 7) suggests a possible weak connection between these variables. However, interpretation of conductance data is complicated, because what appears to be a single integral-conductance event could in fact be due to multiple small pore operating in parallel. Even with ion-selective protein channels, there is no uniform correlation between conductance and ion selectivity. There are selective channels with low conductance of 5–9 pS for epithelial sodium channel [42], 4–14 pS for Ca-activated potassium channel (SK), 8–25 pS for different types of Ca²⁺ channels [32] and 40 pS for Cl⁻ channels [43], but there are also channels with high conductance of 300 pS for potassium BK channel [44] and 430 pS for Cl⁻ channel [45]. Among the class of porin proteins with mild selectivity there is also no noticeable correlation between conductance level and selectivity. A BAX channel reconstituted in a BLM displayed conductances of 330 pS and 730 pS at different pH values and was reported as mildly Cl⁻ selective (P_K/P_Cl = 0.3). In contrast, BCL-2 formed mildly K⁺ selective (P_K/P_Cl = 3.9) channels, with a most prominent initial conductance of 80 pS that increased to 1.9 nS [27]. Outer membrane protein A (OmpA) reconstituted into a BLM showed weak anionic selectivity [46]. Electrophysiological experiments have demonstrated two major conductance states for (OmpA) 50–80 pS and 320–450 pS [47, 48]. The voltage-dependent anion-selective channel (VDAC) has been reported to display transitions from a fully open (650 pS) anion-selective state to a set of partially open substates (300 pS) that are generally described as cation-selective [49]. Moreover, it has been shown that VDAC has a cation-selective state with conductance approximately equal to that of the canonical open state (680 pS) [29].

The E_rev values obtained when COOH-PNP were added to the trans side (100 mM KCl) vs the cis side (10 mM KCl) were not statistically different, indicating that ion selectivity of pores induced by COOH-PNP does not vary with the side of the membrane to which the PNP were added. The direction of the KCl gradient relative to location of the COOH-PNP might have influenced ion selectivity through various mechanisms. First, the salt gradient could influence the driving force for COOH-PNP migration to, adsorption on, and penetration through, BLM [19, 38]. Thickness of the electric double layer surrounding both COOH-PNP and BLM is strongly affected by the salt concentration. The electric double layer, in turn, influences zeta potential and migration rate of COOH-PNP under the influence of an electric field. Second, net charge transfer across BLM carried by PNP as they traverse BLM could contribute to overall current. Transported ions would include both ionized carboxyl functional groups (COO⁻) bound to PNP groups and associated counterions. However, transient processes such as these may not be important in determining ion selectivity. The conductance events during which E_rev is measured are stable on the order of a minute, whereas most transient spikes induced by PNP have a duration on the order of a millisecond. Thus, factors that control the observed ion selectivity are likely to result from the structure of stable PNP-induced BLM pore assemblies as opposed to the rate of molecular transport or interaction processes.

The findings of this study, including that ENM can create ion-selective pores in biomembranes, and that the selectivity (E_rev value) varies with the charge of both the ENM and the lipid headgroup, are important because maintenance of transmembrane ion gradients is essential for cellular functions including energy generation, transport and signaling. For example, ENM-induced modulation of intracellular ion concentrations could alter immune cell activation, proliferation, differentiation and effector function [50]. As a result, significant perturbation in ion gradients could result from cell exposure to ENM. These findings indicate a new mechanism by which ENM may influence cell behavior or induce toxicity, and they also suggest the possibility of designing ENM that work synergistically with the lipids in a target biomembrane to achieve a desired modulation of transmembrane ion gradients or flux.

5. Conclusions

This study demonstrates the use of electrophysiology methodologies to characterize PNP-induced dynamic pore formation in BLM. Experiments conducted with PNP that differed only in their surface functionalization revealed that positively charged amidine-PNP were more potent at disrupting DOPC BLM than negatively charged COOH-PNP, as evidenced by significant differences in average pore conductance and charge transfer rate. This finding is consistent with previous reports that positively charged ENM are often more toxic than negatively charged ENM, and that amidine-PNP traffic across primary rat alveolar epithelial cell monolayers more rapidly than COOH-PNP. Ion selectivity of PNP-induced pores was demonstrated for the first time by measuring reversal potentials under the influence of a transmembrane KCl gradient. Ion selectivity depended on both PNP surface charge and lipid headgroup charge. More negative PNP and lipid headgroup charges resulted in greater selectivity for positive potassium ions, and more positive PNP and lipid headgroup charges resulted in greater selectivity for negative chloride ions. Selectivity was unaffected by current amplitude and the side of BLM to which PNP were introduced. Although the molecular mechanism by which ENM induce ion-selective pores in BLM is not well understood, results of this study are consistent with the hypothesis arising from ion-channel protein research that selectivity for an ion is imparted by oppositely charged functional groups lining the pore.

Supplementary Material

01

Fig. S1. Zeta potential of liposomes made from different lipid mixtures. Measurements were performed in 10 mM KCl. Addition of positively charged (DOEPC) or negatively charged (DOPA) lipids to the primary (DOPC) lipid provided positive or enhanced negative charge on the surface of lipid bilayers, respectively.

* indicates significantly different zeta potential for liposomes formed from DOPC alone than for liposomes formed from a DOPC:DOEPC (3:1) lipid mixture.

** indicates significantly different zeta potential for liposomes formed from a DOPC:DOPA (3:1) lipid mixture than for liposomes formed from DOPC alone.

02

Fig. S2. Currents induced by interactions of 50 μg/ml amidine-PNP with DOPC:DOEPC (3:1) BLM. The BLM was suspended between 10:1 (trans/cis) KCl gradient, and a transmembrane potential of +50 mV was applied. Dotted line indicates zero current.

03

Fig. S3. Estimation of ion selectivity from I-V relationships observed due to amidine-PNP interactions with DOPC:DOEPC (3:1) BLM under a 10:1 (trans/cis) KCl gradient. Currents were recorded while voltage was stepped from −100 to +100 mV in 10 mV increments.

04

Fig. S4. Currents induced by interactions of 100 μg/ml COOH-PNP with DOPC:DOEPC (3:1) BLM. A 10:1 (trans/cis) KCl gradient was imposed across the BLM, and currents were recorded under a transmembrane voltage of +100 mV. Dotted line indicates zero current.

Highlights.

• Polystyrene nanoparticles (PNP) form ion-selective pores in lipid bilayers

• Positive PNP are more potent at inducing bilayer pores than negative ones

• Charges of both nanoparticles and lipid headgroups affect pore ion selectivity

• Selectivity is unaffected by current amplitude and side of nanoparticle addition

Abbreviations

ENM

engineered nanomaterials

BLM

bilayer lipid membranes

LDH

lactic dehydrogenase

DOPC

1,2-dioleoyl-sn-glycero-3-phosphocholine

DOEPC

1,2-dioleoyl-sn-glycero-3-ethylphosphocholine

DOPA

1,2-dioleoyl-sn-glycero-3-phosphate

PNP

polystyrene nanoparticles

amidine-PNP

amidine-terminated polystyrene nanoparticles

COOH-PNP

carboxyl-terminated PNP

PC

phosphocholine

EPC

ethylphosphocholine

PA

phosphatidic acid

E_rev

reversal potential