Exploratory Cortex Metabolic Profiling Revealed the Sedative Effect of Amber in Pentylenetetrazole-Induced Epilepsy-Like Mice

Epilepsy is a common clinical syndrome characterized by sudden and recurrent attacks and temporary central nervous system dysfunction caused by excessive discharge of neurons in the brain. Amber, a fossilized organic substance formed by the resins of conifers and leguminous plants, was prescribed to tranquilize the mind in China. In this paper, the antiepileptic effect of amber was evaluated by a pentylenetetrazole (PTZ)-induced epileptic model. An untargeted metabolomics approach was applied to investigate metabolic changes in the epileptic model, which was based on HILIC-UHPLC-MS/MS multivariate statistical analysis and metabolism network analysis. The outcome of this study suggested that 35 endogenous metabolites showed marked perturbations. Moreover, four metabolism pathways were mainly involved in epilepsy. After treatment by amber, the endogenous metabolites had a marked tendency to revert back to the situation of the control group which was consistent with phenobarbital. This study characterized the pentylenetetrazole-induced epileptic model and provided new evidence for the sedative effect of amber.


Introduction
Epilepsy is a common clinical syndrome characterized by sudden and recurrent attacks and temporary central nervous system dysfunction caused by excessive discharge of neurons in the brain. This can result from a variety of endogenous or exogenous factors, often occurring months or years after a sudden injury [1]. A seizure is a form of epilepsy characterized by abnormal movement or behavior caused by unusual electrical activity in the brain. About 70% of epileptic patients who take antiepileptic drugs have complete remission from seizure [2]. The drugs most commonly used for antiepileptic therapy are the benzodiazepines diazepam (oral or rectal), clobazam, buccal or nasal midazolam, lorazepam, phenobarbital, valproic acid, nitrazepam, acetazolamide, chloral hydrate, pyridoxine, and antipyretics [3,4]. Continuous administration of classical antiepileptic drugs has been argued against because of potentially toxic, sedative and cognitive side effects [5,6]. Thus, new antiepileptic medicines are still needed for the remaining one third of epileptic patients that accounted for approximately 1% of the world population [7,8].

Brain Tissue Nissl Staining
The Nissl staining, a classic nucleic acid staining method to observe the damage degree in the cortex and hippocampal neurons of mice, was widely used in the study of epilepsy [35,36]. To evaluate the effect of amber on cell death in the PTZ-induced kindling model, Nissl staining was performed. The neuronal cells of the control group were found to be round or conical while the model group was impaired. The average density of intact surviving neurons was lower in the PTZ group compared to the control group, while pre-treatment with amber reversed the damage to cell morphology ( Figure 2). Under the intervention of amber, the incubation period was significantly prolonged, and the level of seizures was reduced as well (p < 0.01, Figure 1A,B).

Brain Tissue Nissl Staining
The Nissl staining, a classic nucleic acid staining method to observe the damage degree in the cortex and hippocampal neurons of mice, was widely used in the study of epilepsy [35,36]. To evaluate the effect of amber on cell death in the PTZ-induced kindling model, Nissl staining was performed. The neuronal cells of the control group were found to be round or conical while the model group was impaired. The average density of intact surviving neurons was lower in the PTZ group compared to the control group, while pre-treatment with amber reversed the damage to cell morphology ( Figure 2).

Brain Tissue Nissl Staining
The Nissl staining, a classic nucleic acid staining method to observe the damage degree in the cortex and hippocampal neurons of mice, was widely used in the study of epilepsy [35,36]. To evaluate the effect of amber on cell death in the PTZ-induced kindling model, Nissl staining was performed. The neuronal cells of the control group were found to be round or conical while the model group was impaired. The average density of intact surviving neurons was lower in the PTZ group compared to the control group, while pre-treatment with amber reversed the damage to cell morphology ( Figure 2).

Analysis of Metabolite Profiling
The cortex samples were analyzed by ESI-MS under positive and negative ion modes. Base peak chromatograms (BPCs) for different groups are shown in Figure 3. In order to reveal the differences, peak extraction, peak alignment, background deduction, and elimination of missing values by the 80% rule of zero were carried out for the data of each group. After data filtering, 408 metabolites from positive and negative ion modes were used to build multivariate models separately.

Analysis of Metabolite Profiling
The cortex samples were analyzed by ESI-MS under positive and negative ion modes. Base peak chromatograms (BPCs) for different groups are shown in Figure 3. In order to reveal the differences, peak extraction, peak alignment, background deduction, and elimination of missing values by the 80% rule of zero were carried out for the data of each group. After data filtering, 408 metabolites from positive and negative ion modes were used to build multivariate models separately.

Multivariate Data Analysis
Representative HILIC-UHPLC-ORBITRAP-MS cortex metabolic profiles from a control and a PTZ-treated animal, in ESI + and ESImodes, are shown in Figure 3. The use of quality control (QC) samples and evaluation of data quality have been detailed previously for metabolomics analyses of biological samples [21]. Clustering of QC samples was assessed using principal component analysis (PCA) to reveal if platform stability had been achieved. A PCA scores plot (PC1 vs. PC2) of all study cortex and QC samples analyzed in ESI + mode and ESI − mode are shown in Figure 4A (ESI + ) and 4B (ESI − ). The QC samples are clustered, indicating good reproducibility of the data.
PCA analysis in positive and negative TIC was used to evaluate the PTZ-induced epileptic model. In the positive ion mode, a model with two principal components was obtained (R2X cum = 0.641, Q2 cum = 0.461) while three principal components were obtained in negative mode (R2X cum = 0.836, Q2 cum = 0.449). PCA scores showed that all samples in the positive ion mode were distributed in the 95% confidence interval ellipse while one sample was distributed out of the ellipse in negative mode.
As shown in the PCA scores scatter plot, significant separation between the three groups in the unsupervised mode could be observed, indicating the difference in the metabolites between the different groups.

Multivariate Data Analysis
Representative HILIC-UHPLC-ORBITRAP-MS cortex metabolic profiles from a control and a PTZ-treated animal, in ESI + and ESI − modes, are shown in Figure 3. The use of quality control (QC) samples and evaluation of data quality have been detailed previously for metabolomics analyses of biological samples [21]. Clustering of QC samples was assessed using principal component analysis (PCA) to reveal if platform stability had been achieved. A PCA scores plot (PC1 vs. PC2) of all study cortex and QC samples analyzed in ESI + mode and ESI − mode are shown in Figure 4A (ESI + ) and 4B (ESI − ). The QC samples are clustered, indicating good reproducibility of the data.
PCA analysis in positive and negative TIC was used to evaluate the PTZ-induced epileptic model. In the positive ion mode, a model with two principal components was obtained (R2X cum = 0.641, Q2 cum = 0.461) while three principal components were obtained in negative mode (R2X cum = 0.836, Q2 cum = 0.449). PCA scores showed that all samples in the positive ion mode were distributed in the 95% confidence interval ellipse while one sample was distributed out of the ellipse in negative mode.
As shown in the PCA scores scatter plot, significant separation between the three groups in the unsupervised mode could be observed, indicating the difference in the metabolites between the different groups.

Analysis of Metabolite Profiling
The cortex samples were analyzed by ESI-MS under positive and negative ion modes. Base peak chromatograms (BPCs) for different groups are shown in Figure 3. In order to reveal the differences, peak extraction, peak alignment, background deduction, and elimination of missing values by the 80% rule of zero were carried out for the data of each group. After data filtering, 408 metabolites from positive and negative ion modes were used to build multivariate models separately.

Multivariate Data Analysis
Representative HILIC-UHPLC-ORBITRAP-MS cortex metabolic profiles from a control and a PTZ-treated animal, in ESI + and ESImodes, are shown in Figure 3. The use of quality control (QC) samples and evaluation of data quality have been detailed previously for metabolomics analyses of biological samples [21]. Clustering of QC samples was assessed using principal component analysis (PCA) to reveal if platform stability had been achieved. A PCA scores plot (PC1 vs. PC2) of all study cortex and QC samples analyzed in ESI + mode and ESI − mode are shown in Figure 4A (ESI + ) and 4B (ESI − ). The QC samples are clustered, indicating good reproducibility of the data.
PCA analysis in positive and negative TIC was used to evaluate the PTZ-induced epileptic model. In the positive ion mode, a model with two principal components was obtained (R2X cum = 0.641, Q2 cum = 0.461) while three principal components were obtained in negative mode (R2X cum = 0.836, Q2 cum = 0.449). PCA scores showed that all samples in the positive ion mode were distributed in the 95% confidence interval ellipse while one sample was distributed out of the ellipse in negative mode.
As shown in the PCA scores scatter plot, significant separation between the three groups in the unsupervised mode could be observed, indicating the difference in the metabolites between the different groups.

Screening and Identification of Metabolic Differences
In this study, loading S-plots ( Figure 6) and the variable importance (VIP) generated by OPLS-DA analysis in the projection were used to select potential biomarkers. VIP values larger than 1 were considered to be more important on the classification than average. Ions with p < 0.05 (using an independent sample t-test) showing significant changes in the model group compared to the control group were taken as candidate biomarkers [37]. For those differential features, theoretical database searching and manual spectrum confirmation were used for identification. Thirty-five metabolites differentially expressed between the control and model groups (Table 1) were identified.

Screening and Identification of Metabolic Differences
In this study, loading S-plots ( Figure 6) and the variable importance (VIP) generated by OPLS-DA analysis in the projection were used to select potential biomarkers. VIP values larger than 1 were considered to be more important on the classification than average. Ions with p < 0.05 (using an independent sample t-test) showing significant changes in the model group compared to the control group were taken as candidate biomarkers [37]. For those differential features, theoretical database searching and manual spectrum confirmation were used for identification. Thirty-five metabolites differentially expressed between the control and model groups (Table 1)

Metabolic Pathway Analysis
Pathway analysis of the discriminating metabolites was performed with MetaboAnalyst 4.0, a web-based tool for pathway analysis and visualization of metabolomics [38]. As shown in Figure 7, biological pathway analysis revealed that the identified metabolites important for epilepsy are mainly responsible for the following metabolism pathways: (A) glycerophospholipid metabolism; (B) Figure 6. Loading S-plots generated by OPLS-DA analysis in positive mode (A) and negative mode (B). The x-axis is a measure of the relative abundance of ions, and the y-axis is a measure of the correlation of each ion to the model.

Metabolic Pathway Analysis
Pathway analysis of the discriminating metabolites was performed with MetaboAnalyst 4.0, a web-based tool for pathway analysis and visualization of metabolomics [38]. As shown in Figure 7, biological pathway analysis revealed that the identified metabolites important for epilepsy are mainly responsible for the following metabolism pathways: (A) glycerophospholipid metabolism; (B) nicotinate and nicotinamide metabolism; (C) alanine, aspartate and glutamate metabolism; and (D) pyruvate metabolism. The trends of the metabolites associated with the above four pathways are shown in Figure 8. nicotinate and nicotinamide metabolism; (C) alanine, aspartate and glutamate metabolism; and (D) pyruvate metabolism. The trends of the metabolites associated with the above four pathways are shown in Figure 8.  The variation of phosphorylcholine, choline, acetylcholine, phosphatidylcholine (PC) and lysophosphatidylcholine (LysoPC) in the cortex could potentially indicate that the glycerophospholipid metabolism was disrupted and played a major role in seizures. Increased choline may reflect myelin breakdown, increased cell density, or gliosis, which may indicate Alzheimer's disease or epilepsy. Phosphatidylcholine is the main component of the cell membrane and usually exists on the surface of the ectoplasmic membrane. Excitotoxic events enhance the hydrolysis of phosphatidylcholine in the brain, which was evidenced caused by a concomitant increase in the levels of choline and free fatty acids [39].   The variation of phosphorylcholine, choline, acetylcholine, phosphatidylcholine (PC) and lysophosphatidylcholine (LysoPC) in the cortex could potentially indicate that the glycerophospholipid metabolism was disrupted and played a major role in seizures. Increased choline may reflect myelin breakdown, increased cell density, or gliosis, which may indicate Alzheimer's disease or epilepsy. Phosphatidylcholine is the main component of the cell membrane and usually exists on the surface of the ectoplasmic membrane. Excitotoxic events enhance the hydrolysis of phosphatidylcholine in the brain, which was evidenced caused by a concomitant increase in the levels of choline and free fatty acids [39]. The variation of phosphorylcholine, choline, acetylcholine, phosphatidylcholine (PC) and lysophosphatidylcholine (LysoPC) in the cortex could potentially indicate that the glycerophospholipid metabolism was disrupted and played a major role in seizures. Increased choline may reflect myelin breakdown, increased cell density, or gliosis, which may indicate Alzheimer's disease or epilepsy. Phosphatidylcholine is the main component of the cell membrane and usually exists on the surface of the ectoplasmic membrane. Excitotoxic events enhance the hydrolysis of phosphatidylcholine in the brain, which was evidenced caused by a concomitant increase in the levels of choline and free fatty acids [39].
The variation of gamma-aminobutyric acid (GABA) in the cortex could potentially reflect that the metabolism of alanine, aspartate and glutamate is disrupted, which plays a major role in seizures. GABA, a key inhibitory neurotransmitter, is synthesized through the decarboxylation of glutamate via alanine, aspartate and glutamate metabolism [1]. Studies have shown that epileptic-related brain damage is caused by the release of excitatory amino acid neurotransmitters from over-discharged presynaptic terminals that eventually reach neurotoxic concentrations [40]. Evidence indicated that there are certain regions of the brain where enhanced GABA transmission is anticonvulsant [41]. The GABA level in the cortex was lower in the epileptic group than those in the healthy group which supported the literature [1,42]. After the amber intervention, the level of GABA rebounded and approached the control group.

In Vivo Experiments Protocol
Specific Pathogen-Free (SPF) male ICR mice were purchased from the Experimental Animal Center of Qinglongshan (Nanjing, China, license number: 2018-0001). All the mice were kept in the Specific Pathogen Free Center of Nanjing University of Chinese Medicine, Nanjing, China. All animal studies were in accordance with the guidelines of the Animal Ethics Committee of Nanjing University of Chinese Medicine (201810A016).
After an initial acclimation period of 7 days in cages, 40 mice were randomly allocated into 4 groups: Control, model (oral administration of water for 14 days followed by intraperitoneal injection of PTZ at the dose of 60 mg/kg), PB (PTZ + phenobarbital, abdominal injection with phenobarbital at a dose of 40 mg/kg followed by an intraperitoneal injection of PTZ at the dose of 60 mg/kg with 30 min interval), and amber (oral administration of amber for 14 days at the dose of 0.9 g/kg followed by an intraperitoneal injection of PTZ at the dose of 60 mg/kg).
After the intraperitoneal injection of pentylenetetrazole, observation of the seizures lasted for 30 min, and the severity, latency and duration of seizures were recorded. The degree of epileptic behavior was divided into 6 grades according to the Racine standard of neurology: Level 0, no response; Level I, ear and facial twitch; Level II, myoclonus, but no upright position; Level III, myoclonus, with axial position; Level IV, systemic tonic-clonic seizure; and Level V, systemic tonic-clonic seizure and loss of postural control [43].
The mice were sacrificed after the observation. Three brains of each group were taken and immersed in 4% paraformaldehyde, and paraffin sections of coronal plane were used for Nissl staining. The rest of the brain was divided into cortex and hippocampus, and kept in liquid nitrogen.
After thawing in the fridge at 4 • C, 40 mg of the cortex sample was precisely weighed and used for the following process: 160 µL of extract solution, vortexed for 1 min, sonicated for 5 min in an ice bath, centrifuged for 10 min under 13,000 rpm at 4 • C, and then 100 µL of supernatant was collected and concentrated to dry. The extract was then reconstituted with 40 µL of mobile phase, vortexed for 1 min, sonicated for 5 min in ice bath and centrifuged for 10 min under 13,000 rpm at 4 • C. The supernatant was then finally collected for analysis. Quality control (QC) samples were prepared by pooling aliquots (2 µL) of each sample.
For LC separation, UHPLC Dionex Ultimate 3000 (Thermo Scientific, San Jose, CA, USA) and an ACQUITYTM UPLC BEH Amide column (1.7 µm, 2.1 mm × 100 mm) were used. Water and acetonitrile modified with 5 mM ammonium formate, 5 mM ammonium acetate, and 0.1% formic acid were used as mobile phase A and B, respectively. The column was eluted with a program as follows: The percentage of B was decreased from 95% to 55% at the first 13 min, and then held for 2 min. The flow rate and injection volume were set at 0.4 mL/min and 2 µL, respectively.

Data Analysis
The experimental data were analyzed by R and Compound Discoverer 2.1 software (Thermo Fisher Scientific, Waltham, MA, USA), including peak extraction, peak alignment, background deduction and compound identification. After removing exogenous component interference, the extracted ion fragment peak area was normalized by SIMCA 14.1 software (Umetrics AB, Umea, Sweden), and multivariate statistical analysis was conducted after standardization. The mean-centering method and pareto-scaling method were used to transform the data, and the importance of low-abundance ions was increased, while the noise was not obviously amplified. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were used for classification. The s-plot was generated to discover the significant components between groups as potential markers.
The potential endogenous biomarkers were identified based on accurate molecular mass, MS/MS fragments, and retention behavior by searching online databases. The potential markers were identified within 5 ppm. Moreover, the MS/MS spectrum match was searched in the METLIN database. In this study, the Compound Discoverer (Thermo Fisher Scientific, Inc., Waltham, MA, USA) was used to search KEGG, HMDB and LIPID MAPS database. Metabolism pathway analysis was performed with MetaboAnalyst 4.0 (McGill University, Montreal, QC, Canada), a web-based tool for pathway analysis and visualization metabolomics.

Conclusions
Metabolomics, as a systematic method, could systematically identify, quantify or reveal the metabolites of diseases, provide a basis for the diagnosis, biomarkers and/or monitoring tools of diseases, and provide potential targets for the treatment and prevention of diseases. In this study, the metabolic changes and potential biomarkers of epileptic models were studied by using the metabolomics method of LC-MS technology and metabolic network analysis. After intervention of amber, the incubation period was significantly prolonged, and the level of seizures was reduced as well. The damage to the cortex and hippocampal neuron cells was reversed, the fluctuating composition metabolites had a marked tendency to revert back to the control group which was consistent with phenobarbital. This study characterized the PTZ-induced epileptic model and provided new evidences for amber sedative effect. Our work enhanced not only the understanding of the pathology of epilepsy, but also revealed that amber could effectively inhibit seizures with a similar mechanism to that of phenobarbital.