An In Vitro Nerve Agent Brain Poisoning Transwell Model for Convenient and Accurate Antidote and Nanomedicine Evaluation

Background: Nerve agents (NAs) can irreversibly inhibit acetylcholinesterase (AChE). An effective NA antidote should permeate the blood–brain barrier (BBB) to reactivate the inhibited AChE in brain. There is an urgent requirement for the large-scale evaluation and screening of antidotes. Existing methods for evaluating reactivators in vitro can only examine the reactivation effect of drugs and not brain-target properties. The current Transwell BBB model can only evaluate the drug penetration performance for crossing the barrier, but not the pharmacodynamics. Methods: Highly puri�ed rat brain microvascular endothelial cells (RBMECs) from 2-week-old Sprague Dawley rats were inoculated into the upper chamber of Transwell plates to establish a BBB model. Three key parameters of AChE reactivation were determined by the Ellman method: the minimum detection limit of AChE, the effective dosage of NAs (70% enzyme inhibition rate), and the optimal dosage of reactivators. AChE and NAs were added to the lower pool of Transwell plates to simulate central poisoning, and antidotes of reactivators were added to the upper pool to simulate drug administration. The AChE activity of samples, collected from the lower pool, was measured. A liposomal nanomedicine loaded with the reactivator asoxime chloride (HI-6) was prepared using the extraction method and tested by the model. Results: The obtained RBMECs exhibited a typical monolayer “paving stone” morphology, and tight junctions were expressed among the RBMECs. The concentrations of AChE, sarin, and the reactivator were 0.07 mg/mL, 10 –6 v/v, and 0.03 mg/mL, respectively. The reaction rate of the reactivators obtained from the model was signi�cantly lower than that obtained from the non-model group. Furthermore, a nanomedicine loaded with HI-6 was synthesized. The �nal results and rules obtained from the model were in accordance with those evaluated in vivo. Conclusion: The therapeutic effect of antidotes can be rapidly and accurately evaluated using this model.


Background
Nerve agents (NAs) are the most toxic synthetic chemicals known [1,2]. Many countries are equipped with them, and they have been utilized in several instances of local warfare, such as in the Iraq-Iran War [3] and Syrian civil war [4,5], leading to great casualties. In recent years, these chemicals have been used in terrorism attacks, such as the Tokyo subway sarin attack [6], and assassination attempts, such as the poisoning of Sergei and Yulia Skripal in Salisbury, England [7,8] and the assassination of Kim Jong-nam in Malaysia [9]. NAs can rapidly and irreversibly inhibit AChE in the brain and prevent the hydrolysis of the neurotransmitter acetylcholine (ACh), leading to hyperfunction of the cholinergic system [10]. Without an effective antidote, exposure to NAs can cause immediate mortality [11]; thus, the synthesis and evaluation of an effective antidote is urgently needed.
First, they must cross the blood-brain barrier (BBB) and target the brain. Second, they must show good performance in attacking the toxic agent molecules and separating the agent from the AChE to reactivate hydrolysis. At present, a great number of synthesized candidate antidotes need to be evaluated for further research. For in vitro evaluation, these candidate antidotes are mixed with a certain dosage of NAs directly and AChE in tubes and measured by the date of activity of the enzyme AChE [14]. This in vitro evaluation method only considers the interaction between the antidote and NAs and the reactivation effect, but not the brain-targeting properties. Therefore, many of the candidate antidotes were effective in vitro, but invalid for animal detoxi cation in instances of real NA poisoning. The mature Transwell system of the BBB model inoculated with RBMECs in the membrane can successfully simulate the barrier function in vitro [15][16][17]. However, this model only re ects the ability of the drug to penetrate the barrier layer, but not the pharmacodynamics of detoxi cation.
The detoxi cation effect of the antidote is evaluated more generally in vivo [13,18]. However, several inherent bottlenecks restrict the utilization of animal experiments in real NA poisoning and detoxi cation evaluations, such as the high dosage NA for animal poisoning, marked differences between animal individuals, large numbers of animals required for experimentation, complex and time-consuming experimental processes, and inaccurate results. Therefore, in the preliminary stage of drug screening, it is necessary to develop a method to evaluate brain-targeted detoxi cation by the candidate antidote with safety, rapidness, accuracy, convenience, using high-throughput screening.
In this study, a model based on the Transwell system was proposed to simulate brain poisoning for antidote evaluation (Fig. 1a-b). First, RBMECs were cultured and planted on the membrane of the Transwell system to construct a barrier effect. Then, the antidote, nerve agent, and AChE were added into different chambers of the Transwell system to simulate systemic administration to detoxify brain poisoning. Using this modi ed model, the reactivation rates of existing common reactivators were tested and compared with the data obtained from traditional in vitro and in vivo methods to verify the effectiveness, accuracy, and convenience of the model. Furthermore, a nano-reactivator was synthesized and evaluated using a modi ed model to determine the real detoxi cation effect in mice. The in vitro and in vivo results obtained were consistent, suggesting that the modi ed model could truly and successfully re ect the therapeutic effect of the candidate antidote against nerve agent brain poisoning.

Reagents
All reagents were purchased from Sigma-Aldrich (Darmstadt, GER) unless otherwise stated. Endothelial cell medium (ECM) was purchased from ScienCell (Carlsbad, CA, USA). Fetal bovine serum and phosphate-buffered saline (PBS) were purchased from Gibco (Waltham, MA, US). Bovine serum albumin was purchased from VWR (Monroeville, PA, USA). Cell Counting Kit-8 (CCK-8) was purchased from Dojindo (Kumamoto, Japan). Dipalmitoylphosphatidylcholine (DPPC) was purchased from TCI (Tokyo, Japan). RBMEC study High activity RBMECs were isolated from 2-week-old Sprague Dawley rats by two-step enzyme digestion, density gradient centrifugation, and differential adhesion in sequence [18,19]. The extracted primary cells were cultured in T25 culture asks at a seeding density of 2 × 10 5 cells/mL, and the medium was changed every 2-3 days.
The growth curve and morphology of the cells were monitored by microscopy and CCK-8 assays daily after seeding in 96-well plates and T25 culture bottles at a density of 1 × 10 5 cells/mL. Six days later, the cells were collected by separating them from the bottle with a cell scraper and centrifuging for 5 min (1000 rpm, 25 ℃), xed, dehydrated using 3% paraformaldehyde and alcohol, and observed via transmission electron microscopy (TEM).

Establishment And Characterization Of The Transwell System
The isolated RBMECs were planted on the membrane of the Transwell system at a seeding density of 2 × 10 5 cells/mL [20]. The growth curve of the cell barrier layer was monitored daily via the transendothelial electrical resistance (TEER) assay using an epithelial volt/ohm meter (Millicell® ERS-2; Millipore, Bedford, MA, USA). The effective resistance was calculated using the following formula: Effective resistance (Ω·cm 2 ) = (cell pore resistance -blank hole resistance) × membrane area (4.67 cm 2 ).
After a period of time, the barrier layer formed, and a certain amount of ECM was added to both the upper and lower pools of the Transwell chamber to form a level difference of more than 0.5 cm. After 4 h, the liquid level difference was measured to con rm the barrier function of the cell layer.

Optical density (OD) of AChE
The OD of each component (AChE, nerve agent, and antidote) was tested and calculated before it was added to the modi ed Transwell system. AChE solutions with different concentrations were prepared using an ECM-phenol red free basic solution. The AChE activity was measured with the Ellman method, as previously described [21]. Samples were added to 96-well plates mixed with 8.6% acetylthiocholine (ATCh) and PBS and incubated at 37°C for 30 min. 5,5′-Dithiobis-(2-nitrobenzoic acid) (DTNB) was added to each well for the chromogenic reaction in the last step. ATCh would be degraded by the enzyme of AChE, and the degradation product can react with DTNB for chemiluminescence. The standard curve was calculated using the OD value from the sample plate, and the linear range of the enzyme solution was con rmed.
Therapeutic effect of the classical reactivator by modi ed Transwell model AChE (0.07 mg/mL) and sarin (10 − 6 v/v) were simultaneously added to the lower chamber of the Transwell board, and then, the reactivators (0.03 mg/mL) were added to the upper chamber. After allowing time for the reactivator to penetrate the barrier and react with the inhibited enzyme, samples were collected from the lower chamber and detected by the Ellman method for enzyme activity, as previously described. Preparation And Characterization Of The Liposomal Nanomedicine (LipoHI-6) The proposed antidote (LipoHI-6) was prepared using an extrusion process [23]. Cholesterol (Chol) and DPPC mixed in a 2:1 molar ratio were dissolved and dispersed in chloroform and dried in a ask to produce a homogeneous lipid lm by rotary evaporation. After adding to the HI-6 solution (2.2 mg/mL), ultrasonication was performed for 5 min to separate the lm from the ask and encapsulate the HI-6 droplets to form a suspension. The suspension was then extruded successively 10 times through 200-nm and 100-nm polycarbonate lters using an Avanti Mini Extruder (Alabaster, AL, USA). Finally, the unencapsulated free HI-6 was removed by dialysis with PBS. The size and zeta potential of LipoHI-6 were evaluated by dynamic light scattering (DLS) with a particle analyzer (Nano ZS90; Malvern Panalytical, Malvern, UK). The morphology of the LipoHI-6 nanoparticles was determined using TEM. The concentration of HI-6 loaded into the liposomes was quanti ed at 280 nm using a C 18 column equipped with a UV detector in a high-performance liquid chromatography (HPLC) system (L7100; Hitachi, Tokyo, Japan) [18]. The mobile phase in the HPLC system comprised 15% acetonitrile and 85% H 2 O at a ow rate of 2.0 mL/min. The encapsulation e ciency (EE%) and loading e ciency (LE%) of LipoHI-6 were evaluated using ultra ltration/centrifugation. LipoHI-6 solution (1 mL) was sealed in a dialysis bag and suspended in 800 mL of PBS at pH 7.4 in a shaking incubator at 37°C to evaluate the release behavior of the nanomedicine. Samples (1 mL) were collected for the system at a predetermined period and quanti ed by HPLC.
Permeability and therapeutic effect of LipoHI-6 according to the classical/modi ed Transwell model The permeability of LipoHI-6 across the BBB was calculated using the classical Transwell model. LipoHI-6 (0.03 mg/mL) was added to the upper chamber, and 10 min later, samples collected from the lower chamber were quanti ed using HPLC. The permeability of HI-6 across the BBB was also tested using the same procedure as that for the control.
To evaluate the therapeutic effect of LipoHI-6, it was again added (0.03 mg/mL) to the upper chamber. At the same time, AChE was mixed with sarin in the lower chamber, as described previously. The activity of AChE after 10 min was analyzed using the Ellman method. both speci c pathogen-free) were acquired from Beijing Charles River Laboratory Animal Technology Co, Ltd. (Beijing, China). Rodents had free access to sterilized food and distilled water and were maintained in stainless steel cages lled with hardwood chips in an air-conditioned room on a 12:12 h light/dark cycle.

Therapeutic effect of LipoHI-6 in vivo
The activity of central and peripheral AChE in KM mice was measured after sarin exposure and antidote administration. KM mice were randomly divided into four groups according to their body weight, with 12 mice in each group: normal group (no administration), poisoned group (only sarin poison), control group (poisoned and administered HI-6 [2.2 mg/mL]), and the liposomal nanomedicine group (poisoned and administered LipoHI-6 [2.2 mg/mL]). Mice were injected subcutaneously with sarin (180 µg/kg), followed by a subcutaneous antidote injection (10 µL/g) in the neck. The blood and brain were collected 10 min later, after which the brain tissue supernatant was obtained by centrifugation and diluted 500 times with PBS, and the activity of AChE in the sample was detected using the Ellman method.

Statistical analysis
Data are presented as means ± standard deviation (SD). An unpaired t-test was performed for two groups using GraphPad Prism 7.00 (GraphPad Software; San Diego, CA, USA). Statistical signi cance was set at p < 0.05.

Growth and morphology of RBMECs
The morphological characterization of the RBMECs is shown in Fig. 1c. The RBMECs showed a bead-like appearance immediately after inoculation. After 16 h, the cells adhered to the wall and started growing.
After 3 days of growth, the growth area of the endothelial cells started expanding continuously. After 8 days of culture, the cells were covered with a culture ask, and the shape of the paving-stone-like arrangement was observed. As shown by the RBMEC growth curve measured by CCK-8 (Fig. 1d), the logarithmic growth phase was observed on day 5, and peak growth occurred on days 7-8.

Evaluation of BBB model in vitro
After the RBMECs were incubated in T25 culture asks for 6 days, tight junctions (TJs) were formed between cells, as observed by TEM (Fig. 1e). Throughout the 4-h leakage experiment, it was observed that the liquid level difference between the upper and lower pools of the Transwell plate could be maintained beyond 0.5 cm in the cell chambers (Fig. 1f). After 9 days of culture, the TEER reached 275 Ω·cm 2 (Fig. 1g). Dosage of each component of the model Different concentrations of AChE were detected by the Ellman method. The OD value obtained from Ellman method and the concentration of AChE showed good linearity in the range from 0.01 to 0.1 mg/mL, with a correlation coe cient (R 2 ) of 0.9946 (Fig. 1h). Furthermore, the inhibition rates (Fig. 1i) were investigated by the mixing of different concentrations of AChE (0.05-0.07 mg/mL) and sarin (10 − 3 -10 − 7 v/v). The maximum inhibition rates were in the order of 60%, 61% and 76% with AChE concentrations of 0.05, 0.06, and 0.07 mg/mL, respectively, and the inhibition rate increased with the amount of sarin. Detoxi cation effects of the four classical reactivators were evaluated to determine the amount of the reactivator used (Fig. 1j). The reactivation rates at 0.03 mg/mL AChE were 13%, 39%, 77%, and 53% for HI-6, LüH, 2-PAM, and MMB-4, respectively.
Characterization of LipoHI-6 in vitro TEM results showed that LipoHI-6 had a granular morphology and an obvious lipid bilayer structure at the nanoscale (Fig. 2a). The DLS results revealed that the average particle size was 147.8 ± 5.47 nm, with a polydispersity index (PDI) of 0.17 ± 0.03 (Fig. 2b). The charge of the liposome was close to neutral, and the average potential was 1.14 ± 1.31 mV ( Table 1). The drug-releasing behaviors of LipoHI-6 and HI-6 solutions were measured using the rotating basket method. HI-6 was completely released within 40 min; in the same time, only about 25% of the HI-6 was released from LipoHI-6 (Fig. 2c). The entrapment and loading e ciencies of liposomes were 73.18% ± 3.26% and 18.92% ± 0.84%, respectively, as determined by HPLC (Table 1). In the experiments on brain-target property, the amount of HI-6 delivered by the LipoHI-6 and HI-6 solutions and transported across the layer of the RBMECs into the lower chamber was determined by HPLC (Fig. 2d); within 10 min, the drug concentrations were 0.2 µg/mL and 0.6 µg/mL in the HI-6 and LipoHI-6 solutions, respectively, and there was a signi cant difference between the two concentrations (p < 0.01).

Therapeutic effect of LipoHI-6 in vitro and in vivo
The reactivation rates in the modi ed Transwell model are shown in Fig. 2e. The reactivation rate of the HI-6 group was only 6%, whereas that of the LipoHI-6 group was signi cantly different at 19% (p < 0.001).
In further experiments on nerve agent detoxi cation in mice, the central reactivation rate of the animal model group treated with HI-6 was approximately 2%, and that of the group treated with LipoHI-6 was 20% (Fig. 2f). There was no signi cant difference between the reactivation rates of HI-6 and LipoHI-6, either in the traditional reactivation without the BBB in vitro (Fig. 2g) or the peripheral blood in vivo (Fig. 2h).

Discussion
It is necessary to evaluate the drug detoxi cation effect in animal models in the nal phase of research. The experimental procedure includes nerve agent poisoning, antidote administration, sample (brain and blood) collection, treatment (blood removal, dilution, and centrifugation), and measurement by the Ellman method, which requires a long time. Only one or two drugs could be evaluated in real time, because three animal groups (HI-6 treatment as the positive control, no antidote to determine the poison level, and no treatment or poisoning as the blank control) need to be considered and tested before drug evaluation. Moreover, ensuring the correct reactivation dates requires a re ned nerve agent subcutaneous poisoning operation because small deviations in poison dosage and the direction and depth of insertion would cause a large error in the experimental results. In addition, nerve agent poisoning and antidote administration for each animal should be completed within 1 min of operation and repeated nearly 70 times. All these technical di culties limit research into animal antidotes, especially with regard to the brain. Therefore, it is not appropriate to evaluate candidate drugs in vivo during the initial stage of drug screening, and in vitro drug evaluation might be more appropriate at this stage.
Based on these requirements, a modi ed simulated brain detoxi cation model integrating the Transwell system and reactivation evaluation required to be designed and constructed. In this model, the Transwell board serves as a basic framework, and the RBMECs are planted in the middle membrane, forming a compact cell layer to simulate the BBB structure and achieve cell-level isolation. A reactivator was added to the upper chamber to simulate detoxi cation via intravenous administration. NAs and AChE were added and mixed in the lower chamber to simulate inhibition of AChE by NAs in the central nervous system (Fig. 1a). Only the drug molecules permeating the middle cell layer into the lower chamber could effectively combine with the nerve agent and peel it off from the enzyme to restore the activity. Finally, the therapeutic effect of drugs on sarin-inhibited AChE in the center could be directly determined by monitoring AChE activity in the lower chamber.
Original RBMECs were selected and extracted to build a barrier layer for the core of the system owing to the increased formation of dense TJs by the original cells compared to the subcultured ones (Fig. 1b). The cells showed a "paving stone" growth state 5 days after inoculation (Fig. 1c) and entered the logarithmic growth phase (Fig. 1d). During the logarithmic growth phase, stable and dense TJs (Fig. 1e) were formed between cells. The TEER of the Transwell system reached 275 Ω·cm 2 after cell inoculation on the membrane for 9 days (Fig. 1g), which indicated that the cells formed a stable and dense barrier layer on the membrane when the TEER exceeded 200 Ω·cm 2 [24,25]. Furthermore, the compact barrier layer composed of RBMECs effectively prevented the liquid from penetrating the upper chamber, as demonstrated by the stable liquid level difference in the 4-h leakage experiment (Fig. 1f).
In the next stage, each component in the Ellman method was integrated into the Transwell system, which met two basic requirements: maintaining the activity of the AChE chromogenic reaction and avoiding damage to the cell barrier layer. Therefore, the use of each component was monitored and evaluated to ensure the minimum effective dosage and protection of cells. First, the concentration of AChE chosen for integration in the Ellman method must be within the linear concentration range to ensure accuracy of the AChE chromogenic reaction. However, at an AChE concentration as low as 0.07 mg/mL, the inhibition rate was lower than 75% regardless of the amount of sarin used (Fig. 1i). One prerequisite for reactivator evaluation was that the modi ed model be in a severely acute poisoning state, implying that the inhibition rate be higher than 70% [26]. Therefore, the lowest dose of AChE was determined to be 0.07 mg/mL. Second, to reduce stimulation and ensure the integrity of the endothelial vascular layer, the lowest dose of sarin (10 − 6 v/v) was chosen when the inhibition rate reached 75% (Fig. 1i). Third, the antidote effectively counteracted the nerve agent when the reactivation rate of central AChE reached 10%, as shown in previous reports [27]; therefore, drugs with a reactivation rate of more than 10% in vitro were added to the upper pool. The reactivation rates of the four classic reactivators at different concentrations ( Fig. 1j) could exceed 10% when the concentration was higher than 0.03 mg/mL, which was selected as the screening drug concentration for preferential consideration. Thus, the appropriate concentration and dosage of each component (enzymes, NAs, and candidate drugs) were determined individually to satisfy the requirements.
The four classic reactivators were further evaluated using modi ed Transwell and normal models. Without the cell barrier layer, all drugs could effectively attach to the inhibited enzyme to achieve good treatment effect (up to 72% [the minimum HI-6 was only 26%]) ( Table 2). With the cell barrier, the reactivation rate of all drugs reached only 5.09%. The reactivation rate of all drugs gradually increased (Fig. 1k) but did not reach 10% throughout administration, while HI-6 displayed the lowest therapeutic effect among the four reactivators. The results revealed that most of the hydrophilic reactivators were blocked by the barrier cell layer to counteract the NAs; only very small amounts of reactivators permeated the layer, and the amount only increased slightly with extended periods of administration. It has been proven that extremely serious errors would occur in the absence of a simulated barrier layer in drug evaluation models [14]; drug screening in a reaction model without a barrier layer cannot really re ect the effect of drug treatments. Mean ± SD; n = 4; ***p < 0.001 compared to the control group In addition to evaluating small-molecule drugs, the modi ed Transwell model was used to evaluate the reactivation rate of the nanomedicine. The treatment effect of the nanomedicine in vivo was also processed at the same time to verify the effectiveness of the modi ed Transwell model. Liposomes with sizes < 200 nm (Fig. 2a) were successfully synthesized by using the classic blown lm extrusion method, which consisted of a shell of DPPC and Chol and a core of HI-6 solution. Owing to the advantage of the high lipolytic action of the liposome shell, the nanoparticles are able to permeate the BBB [28 -30]. The DLS results showed that the particle size distribution of the synthesized liposomes was uniform (Fig. 2b).
Compared to the pro le of the HI-6 solution, HI-6 was released more slowly from the liposomes (Fig. 2c).
In the classical Transwell system, HI-6 carried by the nanoparticles effectively penetrated the cell barrier layer of the Transwell system into the lower chamber, as quanti ed by HPLC. The drug concentration in the liposome group was more than three times higher than that in the HI-6 group (Fig. 2d), indicating that the synthesized liposome nanoparticles had a certain central targeting ability.
Further evaluation using the modi ed Transwell model showed that the reactivation rate of the LipoHI-6 group was 19%, and that of the HI-6 group was 6% (Fig. 2e), which proved that the liposomal nanoparticles could more effectively permeate the cell barrier layer of the Transwell system and successfully reactivate sarin-inhibited AChE. Meanwhile, in the reactivation experiments using the animal poisoning model (Fig. 2f), the central reactivation rate of LipoHI-6 was signi cantly higher than that of the HI-6 solution. In the peripheral environment of blood circulation, which has no physiological barrier, the reactivation rates of LipoHI-6 and HI-6 showed no signi cant difference (Fig. 2h), which was similar to the results in the traditional in vitro model without the cell barrier layer representing the BBB (Fig. 2g). However, all of the data accessed from the in vitro and in vivo models led to the same conclusion, which means that the modi ed in vitro model could realistically re ect the drug treatment effect.
Therefore, the three models were comprehensively compared (Table 3). We found that the traditional model in vitro (without a cell layer) had the advantages of simplicity, low time requirement, and low dosages of toxic agents; however, it can only re ect the interaction between drugs and poisons without characterization of the central targeting ability. The result obtained from the animal poisoning model was closest to the true fact. However, the complex operation and requirement of more manpower and higher dosages of NAs limit the application of the method. The modi ed model detailed in this study combines the advantages of the two previously described models and can achieve rapid, accurate, and large-scale drug screening. Furthermore, a new brain-targeting nanomedicine was synthesized, and this modi ed model was evaluated. The results obtained from the modi ed model showed that the nanomedicine allowed for central targeting and effectively reactivated the central AChE, which was consistent with the data obtained from the animal poisoning model. All results show that the modi ed model is suitable for the evaluation of nanomedicine in addition to small molecules, and the nano-delivery system had good prospects for the application of brain disease treatments.  Characterization of LipoHI-6 and its therapeutic effect in vitro and in vivo. a. Transmission electron microscopy (TEM) photographs of LipoHI-6. The scale bar indicates 100 nm. b. Size of LipoHI-6. c.