Saikosaponin d causes apoptotic death of cultured neocortical neurons by increasing membrane permeability and elevating intracellular Ca²⁺ concentration

Abstract

Saikosaponins (SSs) are a class of naturally occurring oleanane-type triterpenoid saponins found in Radix bupleuri that has been widely used in traditional Chinese medicine. As the main active principals of Radix bupleuri, SSs have been shown to suppress mouse motor activity, impair learning and memory, and decrease hippocampal neurogenesis. In the present study, we investigated the effect of five SSs (SSa, SSb1, SSb2, SSc, and SSd) on neuronal viability and the underlying mechanisms in cultured murine neocortical neurons. We demonstrate that SSa, SSb1 and SSd produce concentration-dependent apoptotic neuronal death and induce robust increase in intracellular Ca²⁺ concentration ([Ca²⁺]_i) at low micromolar concentrations with a rank order of SSd > SSa > SSb1, whereas SSb2 and SSc have no detectable effect on both neuronal survival and [Ca²⁺]_i. Mechanistically, SSd-induced elevation in [Ca²⁺]_i is the primary result of enhanced extracellular Ca²⁺ influx, which likely triggers Ca²⁺-induced Ca²⁺ release through ryanodine receptor activation, but not SERCA inhibition. SSd-induced Ca²⁺ entry occurs through a non-selective mechanism since blockers of major neuronal Ca²⁺ entry pathways, including L-type Ca²⁺ channel, NMDA receptor, AMPA receptor, Na⁺-Ca²⁺ exchanger, and TRPV1, all failed to attenuate the Ca²⁺ response to SSd. Further studies demonstrate that SSd increases calcein efflux and induces an inward current in neocortical neurons. Together, these data demonstrate that SSd elevates [Ca²⁺]_i due to its ability to increase membrane permeability, likely by forming pores in the surface membrane, which leads to massive Ca²⁺ influx and apoptotic neuronal death in neocortical neurons.

1. Introduction

Saikosaponins (SSs) are a class of naturally occurring oleanane-type triterpenoid saponins found in Radix bupleuri (Yang et al., 2017), which has been widely used in traditional Chinese medicine for more than two thousand years. The therapeutic properties of SSs have been documented to include relieving fever stemming from infections, hemorrhoids and indigestion, as well as the treatment of liver diseases, such as stagnation (qi) and irritability (National Pharmacopoeia Committee, 2015). SSs have long been considered the active principals in Radix bupleuri products and their potential as therapeutics have been investigated in a variety of disease models. SSs have been reported to display anti-inflammatory (Du et al., 2018; Liu et al., 2014; Lu et al., 2012; Ma et al., 2015; Wang et al., 2018), anti-viral (Fang et al., 2017), anti-tumor (Chen et al., 2016; Li et al., 2017a; Wong et al., 2013; Zhong et al., 2016), and nephroprotective properties (Ma et al., 2015). Their anti-inflammatory activity primarily stems from inhibition of the NF-κB signaling pathway (Du et al., 2018; Fu et al., 2015; Liu et al., 2014; Ma et al., 2015). Among the SSs, saikosaponin d (SSd; Figure 1) appears to be the most active compound (Yuan et al., 2017).

Figure 1. Chemical structures of five Saikosaponins (SSs).

SSs are a class of natural occurring oleanane-type triterpene saponins widely distributed in Radix bupleuri. Type I saikosaponins (SSa, SSc and SSd) possess a cyclic ether bridge linking C13 and C28 while type II (SSb1 and SSb2) are not.

Results from additional studies have demonstrated the neuroactive properties of SSs. SSd has been reported to suppress stress-induced depressive-like behaviors and alter hippocampal neurogenesis (Li et al., 2017). One of the SSd analogs, saikosaponin a (SSa) significantly reduced seizure severity, duration and prolonged seizure latency in a pentylenetetrazol-induced epileptic seizure model; an action that could be mediated through inhibition of mTOR signaling (Ye et al., 2016). However, recent findings suggest that SSa also enhances glutamate uptake capacity of astrocytes (Gao et al., 2017). In hippocampal neuronal culture models of acquired epilepsy and status epilepticus, SSa effectively terminated spontaneous recurrent epileptic discharges and continuous high-frequency bursts at concentrations below micromolar through a number of mechanisms including inhibition of N-methyl-D-aspartate receptors (NMDAR) or voltage-gated sodium channels, and enhancement 4-aminopyridine-sensitive outward K⁺ current (Xie et al., 2013; Yu et al., 2012). In addition to its antiepileptic activity, SSa also displayed analgesic effects in a rat neuropathic pain model induced by chronic constriction injury of the sciatic nerve through a mechanism involving inhibition of p38 MAPK and NF-κB signaling pathways in the spinal cord (Zhou et al., 2014). Finally, SSa suppresses morphine, cocaine and alcohol self-administration, likely through modulation of GABA_B receptor (Maccioni et al., 2016; Yoon et al., 2012; Yoon et al., 2013).

Despite the clinical benefits of SSs in a variety of disease models, clinical use of Radix bupleuri products produce several severe adverse effects, including pain, allergy, low blood pressure, dizziness, convulsion as well as limb twitching (Ikegami et al., 2006), untoward effects that have precipitated a black box warning on injectable formulations of Radix bupleuri by China Food and Drug Administration on May 29^th, 2018 (http://cnda.cfda.gov.cn/WS04/CL2050/228214.html). Results from toxicological studies demonstrate that SSd induces cells death in human LO2 hepatocytes and pluripotent stem cell-derived hepatocytes, possibly through disrupting platelet-derived growth factor-β receptor/p38 and activating Fas death receptor within the concentration range capable of producing therapeutic effects (Chen et al., 2013; Li et al., 2017c; Smutny et al., 2018; Zhang et al., 2016). Most recently, SSd has also been demonstrated to produce neurotoxicity. Systematic administration of SSd suppresses motor activity of mice in the open field test, impairs learning and memory in Morris water maze test, and decreases hippocampal neurogenesis, possibly by interfering with the Akt/Foxg1 pathway (Xu et al., 2018). However, the primary molecular mechanism(s) of acute SSd neurotoxicity remains unclear.

Ca²⁺ dynamic controls neuronal development, activity-dependent neurite outgrowth and synaptogenesis (George et al., 2012; Saneyoshi et al., 2008). Significant increase in the intracellular Ca²⁺ concentration ([Ca²⁺]_i) leads to neuronal death (He et al., 2017; Nayler, 1983; Zieminska et al., 2017). SSd has been demonstrated to increase the [Ca²⁺]_i level which leads to autophagic cell death in HeLa and MCF-7 cancer cells through activation of Ca²⁺/calmodulin-dependent pathway (Wong et al., 2013). Computational ligand docking and coupled enzyme assays suggest that sarco/endoplasmic reticulum Ca²⁺ ATPase (SERCA) pump may be a molecular target responsible for elevated [Ca²⁺]_i leading to cell death in cardiomyocytes and hepatocytes (Wang et al., 2017; Wong et al., 2013). It has also been reported that saponins, a class of compounds containing amphiphilic molecular structure with a lipophilic aglycone and a hydrophilic sugar side chain were able to complex with cholesterol to form non-selective cation/anion channels in both biological and artificial membranes (Sudji et al., 2015; Li et al., 2005) which may also mediate extracellular Ca²⁺ influx.

In the present study, we therefore investigated the neurotoxicity of SSs in primary neocortical neuronal cultures and their underlying mechanisms. We demonstrate that SSd produces concentration-dependent apoptotic neuronal death as well as induces robust increase in [Ca²⁺]_i within the low micromolar concentration range. We also demonstrate that elevation in [Ca²⁺]_i primarily results from enhanced extracellular Ca²⁺ influx, which likely triggers Ca²⁺-induced Ca²⁺ release through ryanodine receptor (RyR) activation, but not SERCA inhibition. Further experiments demonstrate that SSd induces elevation in [Ca²⁺]_i due to its ability to increase the membrane permeability, possibly through forming pores in the surface membrane that leads to massive Ca²⁺ influx and apoptotic neuronal death in neocortical neurons.

2. Materials and Methods

2.1. Materials

Saikosaponins (SSa, SSb1, SSb2, SSc, SSd, purity ˃ 98%) were from Chengdu Purify Technology Development Co., Ltd. (Chengdu, Sichuan, China). Calcein-AM was from Shanghai Aladdin Bio-Tech Co., Ltd. (Shanghai, China). Trypsin, DNase I, soybean trypsin inhibitor, neurobasal medium, fetal bovine serum, glutamax, and leupeptin were obtained from Life Technology (Grand Island, NY, USA). Fluorescein diacetate (FDA), propidium iodide (PI), 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide (MTT), and Hoechst 33342 were from Beyotime BioTech Institute (Shanghai, China). Nifedipine, cyclopiazonic acid (CPA), dizocilpine maleate (MK-801) and 2-[2-[4-(4-nitrobenzyloxy)phenyl]ethyl]isothiourea mesylate (KB-R7943) were purchased from Tocris Bioscience (Ellisville, MO, USA). Poly-L-lysine was from Peptide Institute, Inc. (Osaka, Japan). Fluo-4/AM was from ATT Bioquest (Sunnyvale, CA, USA). Ryanodine was from Ascent Scientific, Inc. (Bristol, UK). [³H]Ryanodine was purchased from PerkingElmer (San Jose, CA, USA; specific activity: 56.6 Ci/mmol). Nicotinamide adenine dinucleotide (NADH) was from Roche Molecular Systems (Pleasanton, CA, USA). D-(+)-glucose, Na₂ATP, MgATP, glycine, ruthenium red (RR), 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), phenol red, ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), cytosine β-D-arabinofuranoside, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), 4’-Chloro-3-methoxycinnamanilide (SB366791), thapsigargin (TG), phosphocreatine, creatine phosphokinase, Arsenazo III, Hoechst 33342, n-decane and 3-morpholinopropanesulfonic acid (MOPS) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Lipid mixture was from Avanti Polar Lipids, Inc. (Alabaster, AL, USA).

2.2. Primary cortical neuron culture

All animal experiment protocols were carried out according to the National Institutes of Health guideline for the care and use of Laboratory animals (NIH Publications No. 8023, revised 1978) and approved by the China Pharmaceutical University and UC Davis Institutional Animal Care and Use Committees. Primary cultures of neocortical neurons were obtained from embryonic day 16–17 C57BL/6 mice (both male and female) as described previously (Cao et al., 2014). Briefly, dissected cerebral cortex was minced and digested with trypsin for 25 min at 37 ^°C. The dissociated neurons were re-suspended in neurobasal medium supplemented with NS21 (Chen et al., 2010), 0.2 mM glutamax and 5% fetal bovine serum. The dissociated cortical cells were plated onto poly-L-lysine pre-coated (0.5 mg/mL) clear-bottom, black-wall, 96-well imaging plates (BD, Franklin Lakes, NJ, USA) or 12-well plate at densities of 1.5 × 10⁵ cells/well and 1.2 × 10⁶ cells/well, respectively. The medium was changed at days in vitro (DIV) 5 and 7 by replacing half the volume of culture medium with serum free neurobasal complete medium. The neurons were maintained at 37 ^°C with 5% CO₂ and 95% humidity. A final concentration of cytosine β-D-arabinofuranoside (10 μM) was added to the culture medium 24–36 h after plating to prevent the astrocytes proliferation.

2.3. MTT assay

Neocortical neurons cultured in 96-well plate at 7–9 DIV were used to measure the cell viability using MTT assay as described previously (Levitz and Diamond, 1985). Briefly, neurons were continuously exposed to different concentrations of SSs or vehicle control (0.1% DMSO) for 24 h in the culture medium. After aspirating the medium, a volume of 100 µL medium containing 0.5 mg/mL MTT was added to each well and incubated for 10 min at 37 ^°C. After discarding the medium, a volume of 150 μL DMSO was added to each well. The absorbance was measured at 570 nm and 650 nm in a microplate reader (Tecan Group Ltd., Männedorf, Switzerland) to measure the content of formazan generated in each well.

2.4. FDA/PI and Hoechst 33342 staining

FDA/PI double staining and Hoechst 33342 staining were performed as described previously (Cao et al., 2010; He et al., 2017). Briefly, after exposure to different concentrations of SSs or vehicle control (0.1% DMSO) for 24 h in the culture medium, neurons cultured in 96-well plate were washed for three times using Locke’s buffer (in mM: HEPES 8.6, KCl 5.6, NaCl 154, D-glucose 5.6, MgCl₂ 1.0, CaCl₂ 2.3, and glycine 0.1, adjust pH to 7.4 with NaOH). A volume of 100 μL Locke’s buffer containing FDA (5 μg/mL) and PI (1 μg/mL) was then added to each well and incubated in room temperature for 10 min to visualize the live and dead cells. To determine the apoptotic cells, a concentration of 5 µg/mL Hoechst 33342 was added to visualize nuclei. Images were taken with a Nikon eclipse fluorescence microscope (Nikon, Tokyo, Japan) by using DAPI, FITC or Texas Red filters. The numbers of live, dead cells, and condensed nuclei were manually counted in each picture. Data were represented as % dead cells or % apoptotic neurons.

2.5. Intracellular Ca²⁺ monitoring

Neocortical neurons at 7–9 DIV were used to investigate the influence of SSs on [Ca²⁺]_i as described previously (Cao et al., 2015). In brief, after 45 min loading of Fluo-4/AM at 37 ^°C, neurons were washed four times with fresh Locke’s buffer and loaded into a Fluorescent Imaging Plate Reader (FLIPR^Tetra; Molecular Devices, CA, USA) incubation chamber. Emission fluorescence at wavelength of 515–575 nm was recorded at a sampling rate of 1 Hz with an excitation wavelength of 470–495 nm. After recording the basal fluorescence for about 200–300 s, different concentrations of SSs were added to corresponding wells using a programmable 96-channel robotic pipetting system and the fluorescent signals were recorded for an additional 400 s. Data were presented as F/F₀, where F is the fluorescent signal at different time point whereas F₀ is the basal fluorescence. The area under curve (AUC) obtained from an epoch of 5 min after addition of SSs was used to compare the influences of SSs on [Ca²⁺]_i.

2.6. Preparation of skeletal muscle (RyR1) and cardiac muscle (RyR2) preparations

Microsomal membrane vesicles enriched RyR1 and SERCA1, and ventricle membrane homogenate preparations enriched in RyR2 were prepared from skeletal muscle of New Zealand White rabbit and C57BL/6 mouse heart ventricles, respectively, as described previously (Cherednichenko et al., 2004; Pessah et al., 1986). Each membrane preparation was kept at −80 ^°C until use. Protein concentrations were determined using BCA assay kit (Life Technology, Grand Island, NY, USA).

2.7. Measurement of microsomal vesicle Ca²⁺ uptake and efflux

The active uptake and passive release of Ca²⁺ ion from junctional sarcoplasmic reticulum (JSR) vesicles was determined by measuring the differential absorbance of metallochromic Ca²⁺ indicator Arsenazo III at 650–700 nm in an Agilent 8453 UV-visible Spectroscopy System (Agilent Technologies, Santa Clara, CA, USA) at 37 ^°C as recently described (Feng et al., 2017). The assay solution contained 100 μg/mL of rabbit skeletal muscle JSR, 2 mM MgATP, 10 mM phosphocreatine, 20 μg/mL creatine phosphokinase, and Arsenazo III (250 μM) in reaction buffer (in mM: KCl 100, MOPS 20, Na₄P₂O₇ 6, pH7.4). Calcium loading of JSR vesicles was accomplished by three sequential additions of 45 nmole CaCl₂. After the final bolus of Ca²⁺ was accumulated into the vesicles, vehicle control (0.1% DMSO) or test compounds was added to determine the Ca²⁺ efflux. To quantified the SSd influence on Ca²⁺ efflux, the Ca²⁺ release rate (nmol Ca²⁺/sec/mg JSR) was calculated from 60 s right after addition of compounds.

2.8. Measurement of SERCA activity

Activity of SERCA from each membrane preparation was measured using a coupled enzyme assay that monitors the rate of oxidation of NADH at 340 nm using an Agilent 8453 UV-Visible diode array spectrophotometer (Agilent Technologies, Santa Clara, CA, USA) as described previously (Ta et al., 2006). After the baselines were established in about 1–2 min, 0.3 mM NADH was added to reaction mixture followed by addition of 0.4 mM Na₂ATP to initiate the oxidation reaction. The vehicle control (0.1% DMSO), TG, TG + CPA or test compounds were introduced before the establishment of baseline. TG or TG + CPA were used as positive controls to determine SERCA1 and SERCA2-independent ATPase activity and verification of the assay system, respectively. To quantify the influence of SSd on SERCA activity, NADH oxidation rate (nmol/sec/mg protein) was calculated from epochs of 3 min and 5 min right after addition of Na₂ATP for SERCA1 and SERCA2, respectively.

2.9. Ca²⁺ imaging in HEK293 cells

Single-cell intracellular cytosolic Ca²⁺ was measured in null HEK293 or IP₃R triple knock out (KO) HEK293 cells stably expressing Ca²⁺ indicator, GCaMP6m (Kovacevic et al., 2013). Ca²⁺ imaging was performed at room temperature using a ZEISS oberserver-Z1 microscope (Carl Zeiss AG, Jena, Germany). The imaging solution contained (in mM): NaCl 107, KCl 7.2, MgCl₂ 1.2, D-glucose 11.5, and HEPES 20 (pH 7.2). Fluorescence signals were recorded at a sampling rate of 2 Hz using an FITC filter. After recording the basal fluorescence for approximately 30 s, test compound was applied by bath perfusion and the signals were recorded for additional 600 s. Fluorescence data were presented as F/F₀. To quantify the influence of SSd on null HEK293 or IP₃R KO HEK293 cells, the AUC values were quantified from an epoch of 5 min right after addition of SSd.

2.10. Equilibrium [³H]Ryanodine binding analysis

Equilibrium binding analyses of [³H]Ryanodine ([³H]Ry) to skeletal muscle RyR1 (0.05 mg/mL) and cardiac muscle RyR2 (0.5 mg/mL) were measured at 37 ^°C for 3 h with constantly shaking in a binding buffer containing (in mM): KCl 250, NaCl 14, HEPES 20, pH 7.4, and 2 μM free Ca²⁺ (obtained by the addition of EGTA calculated using software Bound-and-Determined (Brooks and Storey, 1992). The concentrations of [³H]Ry were used at 1 nM and 2 nM for skeletal and cardiac preparations, respectively. Nonspecific [³H]Ry binding was measured in the presence of 2 µM unlabeled ryanodine. Vesicles were harvested by rapid filtration through Whatman GF/B glass fiber filters (Whatman, Gaithersburg, MD, USA) using a 48-sample Cell Harvester (Brandel, Gaithersburg, MD, USA). Bound [³H]Ry was quantified using a Beckman scintillation counter (Beckman, Indianapolis, IN, USA).

2.11. Single Channel Voltage Clamp

Bilayer lipid membranes (BLM) were formed by mixing 30 mg/mL phosphatidylethanolamine, phosphatidylserine and phosphatidylcholine (lipid mixture) in n-decane. RyR1 channel incorporation into BLM was accomplished by induced vesicle fusions with BLM as previously described (Holland et al. 2017). Once channel fusion was successful, the recording buffer was adjusted to 500 mM (cis-cytoplasmic side) and 50 mM (trans-luminal side) CsCl, buffered with 20 mM HEPES, pH 7.4 to establish a 5:1 Cs⁺ gradient. Fluctuations in channel current were recorded at a holding potential of −40 mV applied to the trans side adjusted to 100 μM Ca²⁺. The Ca²⁺ in the cis side was 2 μM. Baseline was recorded for 1–1.5 min and extended for a time course of 3–15 min after addition of SSd. Data acquisition was made using digitizer DD1440A (Axon-Molecular Devices), digitized at 10 kHz and filtered with 1 kHz low-pass filter (Low-Pass Bessel Filter 8 Pole). Data were quantified from a duration of 1 min after addition of SSd for 1 min. The parameters including channel open probability (P_o), mean open time (τ_o) and mean close time (τ_c) were performed using Clampfit 10.4 (Molecular Devices).

2.12. Calcein leak assay

Neocortical neurons cultured in 12-well plate at 7–9 DIV were used to examine the effect of SSd on the membrane permeability by measuring the leakage of calcein as described previously (Debiton et al., 2004; Gauthier et al., 2009). After loading with calcein-AM (5 μM) in Locke’s buffer for 30 min at 37 ^°C, cells were exposed to SSd (3 µM). Media harvested at different time points were used to measure the fluorescence at 540 nm with excitation wavelength of 480 nm using a fluorescence microplate reader (Thermo Fisher Scientific, Waltham, MA, USA).

2. 13. Measurement of Caspase-3 activity

Caspase-3 activity was determined by Caspase-3 activity assay kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, Jiangsu, China). Briefly, cortical neurons at DIV 7–8 were lysed with lysis buffer provided by the kit, and lysate was centrifuged at 16,000 g for 15 min at 4 ^°C. The supernatant was collected for determining the protein concentration using BCA assay and measuring the activity of Caspase-3 according to the protocols provided by the manufactures.

2.14. Whole-cell voltage-clamp recording

Cultured neocortical neurons at 7–9 DIV were voltage-clamped by whole-cell patch-clamp using an EPC-10 amplifier and Pulse software (HEKA Elektronik, Lambrecht/Pfalz, Germany) as described previously (Paternain et al. 1996). Cells were bathed in extracellular solution containing (in mM): NaCl 140, KCl 5, D-glucose 10, CaCl₂ 2, MgCl₂ 1, HEPES 10, pH adjusted to 7.4 with NaOH. Pipettes were pulled from 1.5-mm capillary tubing and filled with intracellular solution containing (in mM): KCl 140, CaCl₂ 0.4, MgCl₂ 2, EGTA 1, HEPES 20, Na₂ATP 3, pH adjusted to 7.2 with KOH. Pipette-tip resistances were 2–3 MΩ. The cells were continuously clamped at a holding potential of −70 mV. Vehicle control (0.1% DMSO) and SSd were administrated by fast perfusion. Experimental data were collected and analyzed with Patchmaster (HEKA Elektronik) and Origin (version 8.0, OriginLab, Northampton, MA, USA).

2.15. Data analysis

Data analysis was performed using GraphPad Prism software (Version 5.0; GraphPad Software Inc., San Diego, CA, USA). Concentration-response relationship curves were fit by a non-linear regression equation with variable slope. Statistical significance was determined by One-way ANOVA analysis, where appropriate, a Dunnett’s multiple comparison test or T- test; a p value below 0.05 was considered to be statistically significant.

3. Results

3.1. Influence of SSs on the cell viability in cultured neocortical neurons

The chemical structures of five SSs studied have been heuristically divided into two major classes depending on the presence (Type I: SSa, SSc and SSd) or absence (Type II: SSb1 and SSb2) of a cyclic ether linkage between C13 and C28 as illustrated in Figure 1. Both SSa and SSd induced significant neocortical neuron death as reflected by the increase in PI positive neurons, enhanced formazan formation in MTT assay, and increased number of condensed nuclei (Figure 2). The EC₅₀ values for SSd induced neuronal death were 2.92 µM [1.14–5.74 µM, 95% confidence intervals (CI)], 2.56 µM (2.34–2.68, 95% CI), and 3.80 µM (1.92–7.53 µM, 95% CI) measured by the MTT assay (Figure 2A), the FDA/PI double staining assay (Figure 2B & C), and nuclei condensation assay (Figure 2D), respectively. Consistent with its ability to induce nuclei condensation, SSd at 3 µM significantly increased Caspase-3 activity (Figure 2E). SSa was ~ 5-fold less potent than SSd regardless of the neuronal death assay used, with EC₅₀ values of 16.16 µM (13.55–19.26, 95% CI), 15.58 µM (13.76–17.64, 95% CI) and 14.49 µM (13.41–15.65 µM, 95% CI) in MTT, FDA/PI double staining and nuclei condensation assays, respectively. SSb1 appeared to be much less toxic, at the concentration of 30 µM, SSb1 only produced ~ 20% and 45% neuronal death in MTT and FDA/PI double staining assays, respectively (Figure 2). The other two SSs, SSb2 and SSc, were without detectable effect on neuronal death with the assays tested in the present study (Figure 2).

Figure 2. SSs induce cell death in primary cultured neocortical neurons.

(A) Concentration-response relationship curves for five SSs-induced neocortical neuronal cell death measured by formazan formation in MTT assay. (B) Representative images of FDA/PI labelled neocortical neurons treated with vehicle and SSd (3 µM). FDA in green; PI in red. (C) Concentration-response relationship curves for SSa-, SSb1-, and SSd-induced neuronal death quantified from FDA/PI staining. (D) Concentration-response relationship curves for SSa- and SSd-induced nuclei condensation visualized by Hoechst staining (Inlets are the representative images of Hoechst labelled nuclei treated with vehicle or 3 µM SSd). (E) SSd increased Caspase-3 activity. Each data point represents the Mean ± SEM from three independent experiments, each in five replicates. ^**, p < 0.01, SSd vs. Veh. Veh indicates 0.1% DMSO.

3.2. Influence of SSs on the [Ca²⁺]_i in cultured neocortical neurons

We next examined the influence of five SSs on the Ca²⁺ dynamics in neocortical neurons. Addition of SSa or SSd produced sustained elevation of [Ca²⁺]_i in a concentration-dependent manner with EC₅₀ values of 8.1 µM (6.0–10.9 µM, 95% CI), or 2.9 µM (2.1–4.0 µM, 95% CI), respectively (Figure 3A & B). The rise in [Ca²⁺]_i with SSb1 addition was only detectable at concentrations ≥ 20 µM, suggesting that SSb1 was much less potent (Figure 3C). SSb2 and SSc were inactive on the [Ca²⁺]_i overloading up to 30 µM (Figure 3C).

Figure 3. SSs induce elevation in [Ca²⁺]_i in primary cultured neocortical neurons.

SSa (A) and SSd (B) -induced [Ca²⁺]_i elevation in neocortical neurons as a function of time. (C) Concentration-response relationship curves for SSs-elevated [Ca²⁺]_i. Each data point represents the Mean ± SEM from three independent experiments, each in six replicates. Arrow indicates the addition of Veh or test compounds. Veh indicates 0.1% DMSO.

3.3. SSd triggered Ca²⁺ release from microsomal vesicles

The increase in neuronal cytoplasmic [Ca²⁺]_i could result from enhanced Ca²⁺ entry across the plasma membrane, release of intracellular Ca²⁺ stores, or both. We therefore tested whether SSd triggers the release of Ca²⁺ from microsomal vesicles in the presence of active SERCA pump activity. After Ca²⁺ was loaded into the vesicles by three bolus sequential additions of extravesicular Ca²⁺, addition of SSd elicited release of Ca²⁺ from the vesicles in a concentration-dependent manner. SERCA inhibitor TG (30 µM) also produced Ca²⁺ release from the vesicles although the maximum rate of release was ~ 11-fold smaller than SSd (10 µM) (Figure 4).

Figure 4. Effects of SSd on calcium release in junctional sarcoplasmic reticulum (JSR) microsomal preparation.

(A) Representative traces of Ca²⁺ fluxes from JSR, a total of 135 nmole Ca²⁺ was sequentially loaded to the vesicles. 2–3 min following the accumulation of final bolus of CaCl₂ was completed, Veh or appropriate concentration of SSd was added to the system. (B) Quantification of Ca²⁺ release rate from the initial 60 s upon the introduction of SSd. *, p < 0.05, ^**, p < 0.01, Drugs vs. Veh. Thapsigargin (TG, 30 μM) was used as a positive control. Each data point represents the Mean ± SEM (n = 5). Arrow indicates the addition of compounds. Veh indicates 0.1% DMSO.

3.4. SSd negligibly inhibited SERCA activity

Virtual ligand docking studies and coupled enzyme assay have suggested that SSd may cause [Ca²⁺]_i overloading by suppressing SERCA activity (Wong et al., 2013). We therefore investigated whether SSd can directly suppress SERCA activity. Coupled enzyme assay was used to evaluate the inhibitory effect of SSd on SERCA activity. In contrast to previous published results we found that even at as high as 30 µM SSd only negligibly suppressed SERCA1 (~ 9.8% inhibition) and SERCA2 (~17.6% inhibition) activity (Figure 5).

Figure 5. Lack of effect of SSd on SERCA1 and SERCA2 activities.

(A) Representative oxidation traces of NADH in coupled enzyme assay were monitored by measuring absorbance at 340–400 nm. Test compounds or Veh was added separately to each cuvettes followed by the addition of NADH and Na₂ATP to initiate the reaction. TG (30 μM) was used as a positive control. (B) Quantification of NADH oxidation rate in the absence and presence of SSd. ^**, p < 0.01, TG vs. Veh. (C) Time-response relationship of SSd on SERCA2 activity in heart ventricle homogenate. TG + cyclopiazonic acid (CPA) was used as a positive control. (D) Quantification of NADH oxidation rate in the absence and presence of SSd. Each data point represents the Mean ± SEM (n = 3). ^**, p < 0.01, TG + CPA vs. Veh. Veh indicates 0.1% DMSO.

3.5. RyR only partially contributes to SSd-induced Ca²⁺ release

Given the massive Ca²⁺ release from JSR vesicle flux model, we next investigated whether Ca²⁺-releasing channels were engaged in SSd-induced Ca²⁺ release. In the presence of RR, which suppresses RyR but not inositol trisphosphate receptor (IP₃R) function, SSd-triggered efflux rate was inhibited by ~ 59% in the presence of RR after release was initiated (Figure 6A & B). Equilibrium [³H]Ry binding analysis demonstrated that SSd increased [³H]Ry binding to JSR vesicles (RyR1) at concentrations greater than 3 µM but had no measurable influence on cardiac ventricle homogenate preparations (RyR2) (Figure 6C). The EC₅₀ value for SSd-evoked increase in [³H]Ry binding to RyR1 was 18.1 μM (13.9–30.1 μM, 95% CI) and the maximal increase reached 290% of control (Figure 6C). In lipid bilayer single channel recording, SSd (1 µM) significantly increased RyR 1 channel activity by increasing the channel open probability (Po), the mean open time (τ_o) as well as by decreasing the mean close time (τ_c) (Figure 6D and E). To investigate the effect of SSd on IP₃R, SSd-induced Ca²⁺ response was investigated in null HEK293 cells and IP₃R triple KO HEK293 cells. SSd (3 µM) induced similar Ca²⁺ response between these cells excluding activation of IP₃R by SSd (Figure 6F & G). Similarly, a phospholipase C inhibitor, U73122 (3 µM), failed to prevent SSd (3 µM)-induced Ca²⁺ overloading in neocortical neurons, which further excluded the engagement of PLC-IP₃R pathway on SSd-induced elevation of [Ca²⁺]_i (Figure 6H).

Figure 6. SSd directly activates ryanodine receptor.

(A) Time-response relationship for ruthenium red (RR, 4 μM) on SSd-stimulated Ca²⁺ release in JSR. (B) Quantification of RR effect on SSd-stimulated Ca²⁺ release. ^**, p < 0.01, SSd vs. Veh; ^#, p < 0.05, SSd + RR vs. SSd (n = 3). (C) Concentration-response curves to SSd linked by line showing [³H]Ryanodine ([³H]Ry) binding to RyR1 and RyR2. All the data points were normalized to the mean value of Veh group. (D) Representative traces of RyR1 channel activity in lipid bilayer single channel recording in the absence and presence of 1 µM SSd. (E) Quantification of SSd (1 µM) on RyR1 channel gating parameters including channel open probability (P_o), mean open time (τ_o) and mean close time (τ_c). Each data point represents the Mean ± SEM from 3 independent channels. Statistical significance between groups was calculated by T-test. ^**, p ＜ 0.01, SSd vs. Veh. (F) Representative traces of SSd-stimulated [Ca²⁺]_i changes in null HEK293 and IP₃R triple KO HEK293 cells as a function of time. Arrow indicates the addition of SSd (3 µM). (G) Quantification of SSd-stimulated [Ca²⁺]_i overloading in null HEK293 and IP₃R triple KO HEK293 cells (n = 6). The area under curve (AUC) values were calculated from an epoch of 5 min after addition of SSd. (H) Influence of U73122 (3 μM) on SSd-stimulated [Ca²⁺]_i elevation. The first arrow indicates the addition of Veh or U73122 and the second arrow indicates the addition of Veh or SSd (3 µM). Veh indicates 0.1% DMSO.

3.6. SSd-induced [Ca²⁺]_i overload in primary neurons independent of Ca²⁺ release from intracellular Ca²⁺ store

Given the direct but weak activation of RyR by SSd, we next investigated whether SSd-induced Ca²⁺ release in JSR vesicle translated into endoplasmic reticulum (ER) store depletion in intact neurons. Neocortical neurons were treated with 10 µM TG for 50 min to deplete the intracellular Ca²⁺ store and then were challenged with 3 µM SSd. Depletion of intracellular Ca²⁺ store had no effect on the rise in [Ca²⁺]_i triggered by SSd (3 µM), thereby excluding Ca²⁺ release from ER/sarcoplasmic reticulum (SR) Ca²⁺ store as the primary source for SSd-elevated [Ca²⁺]_i in intact neocortical neurons (Figure 7A & B)

Figure 7. Depletion of intracellular Ca²⁺ store by TG (10 μM) did not affect SSd (3 μM)-induced [Ca²⁺]_i increase.

(A) Representative traces of SSd-stimulated [Ca²⁺]_i elevation in the absence and presence of TG. The first arrow indicates the addition of Veh or TG, and the second arrow indicates the addition of Veh or SSd (3 µM). (B) Quantification of TG effect on SSd-induced [Ca²⁺]_i overloading. The area under curve (AUC) values were obtained from an epoch of 5 min after second addition. ^**, p < 0.01, Drugs vs. Veh. Each data point represents the Mean ± SEM from three independent cultures, each in triplicates. Veh indicates 0.1% DMSO.

3.7. SSd-induced [Ca²⁺]_i overloading was dependent on extracellular Ca²⁺ influx

Given the lack of detectable influence of ER Ca²⁺ store depletion on SSd-induced [Ca²⁺]_i rise, we investigated whether the rise in [Ca²⁺]_i in neocortical neurons was partially or wholly dependent on influx of extracellular Ca²⁺. After loading the Ca²⁺ fluorescence dye, Fluo-4/AM, the cells were bathed in normal (2.3 mM), medium (0.2 mM) and 0 mM Ca²⁺ (chelated with 1 mM EGTA) Locke’s buffer. Lower extracellular Ca²⁺ concentration significantly attenuated SSd-induced Ca²⁺ response observed in the physiological buffer (Figure 8A & B).

Figure 8. SSd-induced [Ca²⁺]_i increase is dependent on the extracellular Ca²⁺ influx.

(A) Representative traces of SSd-elevated [Ca²⁺]_i as a function of time in 2.3 mM, 0.2 mM, and 0 mM Ca²⁺ buffers (containing 1 mM EGTA). Arrow indicates addition of SSd (3 µM). (B) Quantification of SSd Ca²⁺ response in 2.3 mM, 0.2 mM, and 0 mM Ca²⁺ buffers (containing 1 mM EGTA). The area under curve (AUC) values were obtained from an epoch of 5 min addition of SSd. ^**, p < 0.01, 0 mM or 0.2 mM Ca²⁺ vs. 2.3 mM Ca²⁺; ^##, p < 0.01, 0 mM Ca²⁺ vs. 0.2 mM Ca²⁺. Each data point represents the Mean ± SEM from three independent cultures, each in triplicates. Veh indicates 0.1% DMSO.

3.8. SSd-induced elevation in [Ca²⁺]_i was not dependent on the major routes of Ca²⁺ influx

Given the involvement of extracellular Ca²⁺ influx on SSd-elevated [Ca²⁺]_i, we next investigated whether known inhibitors/blockers of Ca²⁺ regulatory proteins expressed and targeted to plasma membrane of neurons were potential targets of SSd. Inhibitors tested included nifedipine, an inhibitor of L-type calcium channels, KB-R7943, an antagonist of Na⁺-Ca²⁺ exchanger, MK-801, an inhibitor of NMDAR, CNQX, an α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor/kainic acid receptor (AMPAR/KAR) inhibitor, SB366791, a TRPV1 inhibitor, and RR, a non-selective cation channel blocker. However, none of these inhibitors was effective in mitigating SSd-elevated [Ca²⁺]_i in neocortical neurons (Figure 9).

Figure 9. Major Ca²⁺ influx routes are not involved in SSd-elevated [Ca²⁺]_i.

Representative traces of nifedipine (A), KB-R7943 (B), MK-801 (C), CNQX (D), SB366791 (E), and RR (F) effect on SSd-stimulated [Ca²⁺]_i elevation as a function of time. This experiment was performed in three independent cultures, each in triplicates with similar results. The first arrow indicates the addition of Veh or inhibitor and the second arrow indicates the addition of Veh or SSd (3 µM). Veh indicates 0.1% DMSO.

3.9. SSd enhanced membrane permeability in primary cultured neocortical neurons

Given that the SSd-triggered increase in [Ca²⁺]_i seemed non-selective, we more closely examined actions of SSd on the plasma membrane. After loaded with calceinAM, neurons were washed and exposed to vehicle control (0.1% DMSO) or SSd and the temporal release of calcein was measured. SSd produced a significant efflux of calcein at as early as 8 min post-exposure, which continued to gradually increase and saturated over 90 min (Figure 10A). To further examine the membrane effect, neocortical neurons were voltage-clamped at a holding potential of −70 mV. Addition of SSd (3 µM) but not vehicle (0.1% DMSO) induced an inward current (Figure 10B).

Figure 10. SSd increases membrane permeability in cultured neocortical neurons.

(A) SSd (3 µM)-induced calcein efflux in neocortical neurons as a function of time. Each data point represents the Mean ± SEM from three independent cultures, each in triplicates. (B) Representative trace of SSd-induced inward current. The first arrowhead indicates the addition of Veh and the second arrowhead indicates the addition of SSd (3µM) perfusion. This experiment was performed in 5 neocortical neurons with similar results. Veh indicates 0.1% DMSO.

4. Discussion

As one of the main anti-fever and anti-inflammation traditional Chinese medicines used in the clinic, preparations of Radix bupleuri have been reported to display severe side effects in many organ systems (Itoh et al., 1995; Lee et al., 2011; Li et al., 2017c; Xu et al., 2018). Clinical safety issues have prompted the China Food and Drug Administration to require a black box warning for products containing Radix bupleuri. Recent studies have demonstrated that SSd suppresses mouse motor activity, impairs learning and memory, and decreases hippocampal neurogenesis (Xu et al., 2018). In the present study we demonstrate that SSa, SSb1 and SSd induce neurotoxicity in primary cultured neocortical neurons with rank order of SSd > SSa > SSb1, whereas SSb2 and SSc have no significant effect on the cell viability of neocortical neurons. The rank order of SSs on neurotoxicity is consistent with their hemolytic and anti-inflammatory activity, as well as their inhibitory action on drug-induced hepatic injury (Abe et al., 1981b; Ashour and Wink, 2011; Yuan et al., 2017). SSa and SSd produce significant neurotoxicity while SSb1 is much less toxic and SSb2 is without toxic effect on the neocortical neurons, suggesting the importance of the cyclic ether linkage between C13 and C28. In addition, the conformation of the hydroxyl group at C16 also appears to be important for SS-induced neurotoxicity.

Previous studies have demonstrated that SSd elevated [Ca²⁺]_i level in Hela and MCF-7 cells and this elevated [Ca²⁺]_i level was responsible for the autophagic cell death, posited to be mediated by over-activation of the Ca²⁺/calmodulin-dependent kinase kinase-AMP-activated protein kinase-mammalian target of rapamycin (CaMKK-AMPK-mTOR) pathway (Wong et al., 2013). Tight spatial and temporal regulation of [Ca²⁺]_i levels controls neuronal development, including activity-dependent neurite outgrowth and synaptogenesis, whereas uncontrolled changes in [Ca²⁺]_i lead to neuronal death (George et al., 2012; He et al., 2017; Zieminska et al., 2017). Here we demonstrate that SSs trigger robust elevation of [Ca²⁺]_i in neocortical neurons. The ability of SSs to elevate [Ca²⁺]_i parallels its acute neurotoxicity with respect to potency.

Previous studies have suggested that SSd is a SERCA inhibitor and responsible for elevated [Ca²⁺]_i in Hela and MCF-7 cancer cells, as well as cardiomyocytes and hepatocytes (Wang et al., 2017; Wong et al., 2013). However, the significant differences between SSd concentrations needed to influence [Ca²⁺]_i and those needed to inhibit SERCA make it unlikely that SERCA inhibition is a primary mechanism leading to the rise in [Ca²⁺]_i and loss of cell viability. SSd at 10 µM produced a robust [Ca²⁺]_i elevation in Hela and MCF-7 cells but had negligible inhibition of SERCA activity (~ 15% inhibition) (Wong et al., 2013), consistent with our data that indicated SSd at 30 µM minimally suppressed SERCA1 (~ 10%) and SERCA2 (~ 18%). Although RyR1, but not RyR2, appeared to represent a direct target of SSd in equilibrium [³H]Ry binding assays and single channel recording, direct activation of RyR1 by SSd may be only partially responsible for Ca²⁺ efflux from JSR vesicle preparations since release was only partially suppressed by RR, a potent and complete blocker of RyR1-mediated Ca²⁺ release (Mack et al., 1992; Xu et al., 1998).

Our results showed that SSd-triggered elevation of [Ca²⁺]_i was primarily mediated through the entry of extracellular Ca²⁺ in intact neocortical neurons. Depletion of Ca²⁺ stores with TG (10 µM) did not mitigate the influences of SSd on [Ca²⁺]_i suggesting that depletion of the intracellular Ca²⁺ store was not a necessary step for SSd-elevated [Ca²⁺]_i. Moreover, lower the extracellular Ca²⁺ concentrations proportionally reduced SSd Ca²⁺ response suggesting the major role of extracellular Ca²⁺ influx for SSd-elevated [Ca²⁺]_i.

SSd-induced Ca²⁺ influx was not attenuated by addition of several blockers of major neuronal Ca²⁺ entry pathways, including L-type Ca²⁺ channels, NMDAR, AMPAR, Na⁺-Ca²⁺ exchanger, and TRPV1. These data suggest that SSd acts though a non-selective mechanism, possibly as a potent Ca²⁺ ionophore. A nonselective ionophore mechanism is consistent with previous results showing that SSs caused a significant decrease in the negative charge on the cell surface and the destruction of microvilli (Abe et al., 1981a). Saponins are, in fact, amphiphilic molecules possessing a lipophilic aglycone and a hydrophilic sugar side chain. Digitonin, a steroidal saponin disintegrates biological and artificial membranes, increases membrane permeability and causes membrane rupturing which is dependent on the presence of cholesterol (Sudji et al., 2015). In addition to steroidal saponins, avicins, a class of triterpenoid saponins with structural similarities to SSd, have been shown to form non-selective cation/anion channels with an estimated pore size of 1.1 nm in radius (Li et al., 2005). Considered together, these data strongly suggest that SSd-increased [Ca²⁺]_i is likely mediated through incorporation of ionophore that enhances Ca²⁺ influx. This interpretation is further supported by calcein efflux and patch-clamp recording experiments in which SSd produces a significant efflux of calcein and an inward current in neocortical neurons. Intercalation of SSd with the plasma membrane most likely accounts for its potent cytotoxicity towards various cancer cells, as well as primary cells in culture (Li et al., 2017b; Wong et al., 2013; Yao et al., 2014; Zhong et al., 2016). The mechanism of how SSd increases the membrane permeability and whether it possesses ion selectivity need further exploration. In addition, whether SSd neuronal toxicity is solely dependent on increased [Ca²⁺]_i needs further investigation.

In summary, we demonstrate that SSd induces apoptotic neuronal death in primary cultured neocortical neurons. This neuronal death is parallel to massive extracellular Ca²⁺ influx as a consequence to the enhanced membrane permeability, most likely by interacting with membrane bilayers to form pores permeable to Ca²⁺.

Highlights:

• SSd produces concentration dependent apoptotic cell death in neocortical neurons

• SSd elevates intracellular Ca²⁺ concentration ([Ca²⁺]_i) in neocortical neurons

• SSd-elevated [Ca²⁺]_i is not dependent on SERCA activity

• SSd Ca²⁺ response depends on enhanced Ca²⁺ influx but not main routes of Ca²⁺ entry

• SSd increases the membrane permeability leading to Ca²⁺ influx and neuronal death

Abbreviations

AMPAR

α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor

CI

Confidence intervals

CNQX

6-Cyano-7-nitroquinoxaline-2,3-dione

CPA

Cyclopiazonic acid

DIV

Days in vitro

DMSO

Dimethyl sulfoxide

EGTA

Ethylene glycol bis(β-amino-ethylether)tetraacetic acid

ER/SR

Endoplasmic/sarcoplasmic reticulum

FDA

Fluorescein diacetate

HEPES

4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid

IP₃R

Inositol trisphosphate receptor

JSR

Junctional sarcoplasmic reticulum

KB-R7943

2-[2-[4-(4-Nitrobenzyloxy)phenyl]ethyl]isothiourea

MK-801

Dizocilpine maleate

MOPS

3-Morpholinopropanesulfonic acid

MTT

3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

NADH

Nicotinamide adenine dinucleotide

NMDAR

N-methyl-D-aspartate receptor

PI

Propidium iodide

RR

Ruthenium red

RyR

Ryanodine receptor

SB366791

4’-Chloro-3-methoxycinnamanilide

SERCA

Sarco/endoplasmic reticulum Ca²⁺-ATPase

SSs

Saikosaponins

TG

Thapsigargin

TRP

Transient receptor potential cation channel

[³H]Ry

[³H]Ryanodine

[Ca²⁺]_i

Intracellular Ca²⁺ concentration.