Hippocampal Proteomic Analysis in Male Mice Following Aggressive Behavior Induced by Long-Term Administration of Perampanel.

Antiepileptic drugs have been shown to be associated with inducing or exacerbating adverse psychotropic reaction, including aggressive behavior. Perampanel, the first pharmacological compound approved by the FDA in 2012, is an effective antiepileptic drug for intractable epilepsy but induces severe aggression. So far, the underlying molecular mechanisms of aggression induced by perampanel remain incompletely understood. In the present study, a model of aggressive behavior based on the clinical use of perampanel was established and resident-intruder test and open field test were performed. Changes in hippocampal protein profiles were detected by tandem mass tag (TMT) proteomics. The behavioral results indicated that long-term use of perampanel increased the aggressive behavior of C57BL/6J mice. Proteomic analysis revealed that 93 proteins were significantly altered in the hippocampus of the perampanel-treated group (corrected p < 0.05), which were divided into multiple functional groups, mainly related to synaptic function, synaptogenesis, postsynaptic density protein, neurite outgrowth, AMPA-type glutamate receptor immobilization, and others. Bioinformatic analysis showed that differentially expressed proteins were involved in synaptic plasticity and the Ras signaling pathway. Furthermore, validation results by western blot demonstrated that glutamate receptor 1 (GluA1) and phosphorylation of mitogen-activated protein kinase (ERK1/2) were notably up-regulated, and synaptophysin (Syn) and postsynaptic density 95 (PSD95) were down-regulated in perampanel-treated mice. Therefore, our results provide valuable insight into the molecular mechanisms of aggressive behavior induced by perampanel, as well as potential options for safety treatment of perampanel in the future.

Introduction

Antiepileptic drugs have been shown to be associated with inducing adverse psychotropic reaction, including aggressive behavior.[1] Perampanel (PER), as the latest generation of antiepileptic drug, is a negative allosteric (noncompetitive) α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor antagonist with a novel pharmacological mechanism of action,[2] which is licensed as adjunctive treatment of partial-onset seizures in patients with epilepsy aged 12 years and older.[3] Pooled data from phase III studies showed an increased rate of hostile/aggressive adverse events in patients who received PER compared with those who received placebo.[4] There is a black box warning by the FDA about serious psychiatric and behavioral reactions in the prescribing information, including “aggression, hostility, and irritability”.[5] There is increasing evidence suggesting that irritability and aggression are the most serious mental responses that occurred at dose-dependent rates.[6] A pharmacokinetic–pharmacological study confirmed a significant correlation between the blood concentration of PER and the incidence of aggressive behavior.[7] According to a 4 year clinical extension study, the most common adverse events that led to drug discontinuation were irritability (1.2%) and fatigue (1.1%).[8] Moreover, aggressive behavior caused by PER not only impacts the use of drugs but also brings difficulty with continuing the treatment of epilepsy. Consequently, studying the molecular mechanisms of aggressive behavior induced by PER is of paramount importance for safer application of this promising agent.

Aggression is a complex phenomenon, involving the interaction among genetic, neurobiological, and psychosocial factors.[9] So far, the underlying molecular mechanisms of aggression remains incompletely understood. Early animal studies had shown brain regions implicated in aggressive behavior including the prefrontal cortex, nucleus accumbens, hypothalamus, amygdala, and striatum.[10] Indeed, structural abnormalities in any of these brain regions could increase the risk of aggressive behavior.[11] Additionally, studies have shown that neurotransmitters and hormones also affect the occurrence of aggressive behavior.[12] However, aggressive behavior is a multifactorial disease that involves abnormal molecular alterations.[13] A recent study has implicated that aggressive behavior is involved in neurodevelopmental processes and synaptic plasticity.[14] It may indicate that aggressive behavior does not only involve neurotransmitter function but also involve molecules in establishing neuronal circuits and brain plasticity.[15] The hippocampus is a key anatomical brain region involved in learning and memory, anxiety, and depression.[16] It is highly susceptible to drugs and environmental toxins.[17] A structural magnetic resonance imaging study found that a decreased volume of the hippocampal cortex and structural brain abnormalities were strongly associated with aggressive behavior.[18] In addition, another study revealed that the hippocampal CA2 region to the lateral septal circuit disinhibits social aggression.[19] Although the hippocampus is known to be important for declarative memory, how the hippocampus regulates aggression is less well understood. Proteomics, as a high-throughput technique, can indiscriminately study the protein response and behavior pattern under different stress conditions and is widely applied to gain a deeper understanding of drug action, which contributes to the exploration of the induction mechanism of drug-mediated toxicity and provision of potential biomarker candidates.[20−22] Of note, to date, very few studies have employed a proteomic approach in revealing protein markers of aggressive behavior induced by PER. Therefore, the aim of our study was to determine whether administration of perampanel to mice would induce alterations in hippocampal protein profiles and to elucidate the possible neurobiological mechanisms of aggressive behavior.

In the present study, a model of aggressive behavior was established to simulate the clinical use of perampanel for 2 months, and changes in hippocampal protein profiles were detected using tandem mass labeling (TMT) proteomics. Bioinformatic analysis showed that differentially expressed proteins (DEPs) were involved in synaptic plasticity and the Ras signaling pathway. In addition, changes in synaptic marker proteins and key molecules of the Ras signaling pathway were further verified by western blot. These results provide profound implications of the molecular mechanisms for PER-induced aggressive behavior, as well as the disturbance of proteomic profiles supporting PER-induced aggressive behavior.

Materials and Methods

Animals and Treatment

This experiment was performed in accordance with the Guidelines for The Care and Use of Laboratory Animals of the National Institutes of Health and approved by the Ethics Committee for the Use of Laboratory Animals of Ningxia Medical University (Z2019/023, approved 10 November 2019). Sixty 8 week-old male C57BL/6J mice (21 ± 0.5 g) were selected from the Experimental Animal Center of Ningxia Medical University. All mice were reared under standard feeding conditions, including temperature (22 ± 2 °C), relative humidity (35–50%), light/dark cycle for 12 h, and free food and water. All mice were grouped in transparent plastic cages (24 × 13.5 × 13 cm) with five mice per cage. Mice were randomly divided into two groups: resident mice and intruder mice (30 mice in each group). The intruder mice were not given any treatment, only free feeding and water. Resident mice were assigned to two experimental groups: drug “perampanel” group (PER) and control “vehicle” group (Con) (n = 15 each group). The Con group was intragastrically given 0.5% carboxymethyl cellulose-sodium (CMC-NA) solution (0.1 mL/g), mice in the PER group were intragastrically given PER/CMC-Na suspension (0.015 mg/kg, PER dissolved in CMC-NA), and the dose of PER was selected according to the current clinically published literature.[23] The mice were administered intragastrically continuously for 60 days, which took place between 3 pm and 5 pm per day. The initial dose of PER was 0.015 mg/kg at the 1st month, and the dose was increased to 0.03 mg/kg at the 2nd month. The resident–intruder test was performed at the 61st day.

Open Field Test

Mice were placed in a closed template area (50 cm × 50 cm × 40 cm) and allowed to explore freely for 10 min, and the mouse movement was monitored using Smart 3.0 video tracking software (Panlab, Spain). The experiment was carried out in the behavioral laboratory with sound insulation, light intensity, temperature, and humidity suitable and consistent. At the beginning of the experiment, the experimental equipment was cleaned to avoid the residual information of the last animal affecting the results. The mice in each group were placed in the box in turn, and the experimenter quickly evacuated the video area. The system automatically records the moving distance (cm) and time (s) of the mice in the central and surrounding areas and analyzes the data. The experiment began at the 61st day and 1 h before the beginning of the resident–intruder test.

Aggressive Behavior Test

Aggressive behavior is assessed using the resident–intruder standard procedure.[24] Before the behavior test, the sawdust of the residential home cage was not cleaned for 1 week. The entire behavioral room was lit at 40 lux, and the mice were acclimated to the environment for 1 h. At the beginning of the resident–intruder test, one mouse was left in the home cage, while four mice were moved to a new cage. One intruder mouse was placed into the home cage containing one resident mouse. The intruder mouse weighed about 10% less than the resident mouse. Behavior was videotaped with an iPhone mounted directly above the cages for 10 min. The video recordings were later analyzed by two trained observers who were blinded to the treatment groups. The latency to the first attack, the total attack number, and duration were measured. Mice with the aggression score of more than 3 points were screened and entered the sampling stage. The aggressive intensity was assessed based on the aggression scale: 0 points, no aggressive behavior; 1 point, attack number ≤ 3, sniffing, wagging their tails, chasing; 2 points, attack number ≤ 10, attacking the head, flanking, suppressing, squeaking, and violently attacking the opponent; 3 points, attack number ≤ 20, in addition to the above performance, there is mutual bite, falling or climbing on the intruder; 4 points, attack number > 20, bite violently, form a ball, bite each other, and continue to attack after yielding.

Protein Extraction and Digestion

Three minutes after the end of the last aggressive behavior test, mice were anesthetized by intraperitoneal injection of 3% pentobarbital sodium solution at a dose of 50 mg/kg. After that, the mice were directly killed by neck cutting. The hippocampal tissues were removed quickly on ice and frozen in liquid nitrogen before being stored in a refrigerator at −80 °C for further analysis. SDT (4% SDS, 100 mM Tris–HCl, pH 7.6) buffer was used for sample lysis and protein extraction. The amount of protein was quantified with the BCA Protein Assay Kit (Bio-Rad, USA). Protein digestion by trypsin was performed according to the filter-aided sample preparation (FASP) procedure described by Mann et al.[25] The peptide content was estimated by UV light spectral density at 280 nm using an extinction coefficient of 1.1 of 0.1% (g/L) solution that was calculated on the basis of the frequency of tryptophan and tyrosine in vertebrate proteins.

Tandem Mass Tag (TMT) Labeling

One hundred micrograms of peptide mixture of each sample was labeled using TMT reagent according to the manufacturer’s instructions (Thermo Scientific). Peptides of the two groups were labeled with different TMT labels: three biological repeats of the control group were labeled with TMT-126, TMT-127, and TMT-128 and three biological repeats of the PER group were labeled with TMT-129, TMT-130, and TMT-131.

High-pH Reversed-Phase Fractionation

Labeled peptides were fractionated into 10 different fractions by the High pH Reversed-Phase Peptide Fractionation Kit (Thermo Scientific). Briefly, the dried peptide mixture was reconstituted and acidified with 0.1% TFA solution and loaded to an equilibrated, high-pH, reversed-phase fractionation spin column. The collected fractions were desalted on C18 Cartridges (Empore SPE Cartridges C18 (standard density); bed I.D., 7 mm; volume, 3 mL; Sigma) and concentrated by vacuum centrifugation.

LC–MS/MS Analysis and Protein Identification

Samples were analyzed using a Q Exactive mass spectrometer (Thermo Scientific) that was coupled to an Easy nLC1200 (Thermo Fisher Scientific). The peptides (1 μg) were loaded onto a reversed-phase trap column (Thermo Scientific Acclaim PepMap100, 100 μm × 2 cm, nanoViper C18) connected to a C18 reversed-phase analytical column (Thermo Scientific Easy Column, 10 cm long, 75 μm inner diameter, 3 μm resin) in buffer A (0.1% formic acid) and separated with a linear gradient of buffer B (84% acetonitrile and 0.1% formic acid) at a flow rate of 300 nL/min controlled by Intelli Flow technology. The MS raw data for each sample were searched using the MASCOT engine (Matrix Science, London, UK; version 2.2) embedded into Proteome Discoverer 1.4 software for identification and quantitation analysis. The false discovery rate was set as ≤0.01. The protein ratios were calculated as the median of only unique peptides of the protein. All peptide ratios were normalized by the median protein ratio, and the median protein ratio should be “1” after the normalization. Fold changes were based on the 95% confidence interval of the variance between the two replicate standard TMT-plex reporter ion ratios (n = 6849, 95%, 1.007–1.011). By considering that the samples were deviated from 1-fold expression (no change in TMT reporter ion intensities vs control intensity values), statistical analysis was used with p < 0.05 as a criterion for significance.

Bioinformatic Analyses

Cluster 3.0 and Java Treeview software were used to perform the hierarchical clustering analysis. Functional annotation of Gene Ontology (GO) was derived from the software program Blast2GO (https://blast.ncbi.nlm.nih.gov/). Proteins were classified based on three categories: biological process, cellular component, and molecular function. Following annotation steps, the studied proteins were blasted against the online Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/kegg/) to retrieve their KEGG anthology identifications and were subsequently mapped to pathways in KEGG. Enrichment analysis was applied based on Fisher’s exact test. Benjamini–Hochberg correction for multiple testing was further applied to adjust derived p-values. Also, only functional categories and pathways with p-values under a threshold of 0.05 were considered significant. The protein–protein interaction (PPI) information of the studied proteins was retrieved from the IntAct molecular interaction database by their gene symbols and STRING software (http://string-db.org). The results were imported into Cytoscape software (version 3.6.0) to visualize and further analyze functional protein–protein interaction networks.

Western Blot

Isolation of hippocampal proteins was described above. The samples (20 μg for each sample) were separated on 12% polyacrylamide SDS–PAGE gel (83 × 73 × 15 mm) and then transferred to polyvinylidene difluoride membranes, which were incubated with anti-synaptophysin monoclonal antibody (Abcam, cat. #: ab32127, 1:1000), anti-PSD95 monoclonal antibody (Abcam, cat. #: ab2723, 1:200), anti-ERK1(phospho T202) + ERK2(phospho T185) monoclonal antibody (Abcam, cat. #: ab214036, 1:1000), anti-glutamate receptor 1 (AMPA subtype) monoclonal antibody (Abcam, cat. #: ab109450, 1:2000), or anti-alpha tubulin polyclonal antibody (Abcam, cat. #: ab7291, 1:5000) overnight at 4 °C and then incubated with horseradish peroxidase (HRP) conjugated anti-rabbit or anti-goat secondary antibody (Proteintech, 1:10000) for 3 h at room temperature. Protein bands were visualized using a Supersignal West Pico Chemiluminescence kit (Pierce).

Statistics

For histograms, volcano plots, and clustering analysis, proteomic data are log-transformed and normalized when appropriate. Statistical analysis was performed using SPSS software version 13.0 (IBM Software, New York, USA). Data are expressed as mean ± standard deviation (SD). Comparative analysis of the quantitative data was performed using Student’s t-test with statistically significant difference at p < 0.05.

Results

Perampanel Exposure-Induced Aggressive Behavior

Aggressive behavior is considered to be one of clinical phenotypes resulting from patients with long-term use of PER. Based on these reports, we developed a mice model to study the aggressive behavior induced by PER. Before assessing the level of aggression, we measured the locomotor activity in an open field test. The total distance traveled and central duration of PER-treated mice were not different from the Con group (Student’s t-test, p > 0.05,Figure ), indicating that PER did not affect the locomotor function of mice. Regarding aggressive behavior, compared with the control group, PER-treated mice showed a decrease in latency to the first attack and an increase in the number of attacks and total duration of attacks (Student’s t-test, p < 0.001, Figure ). These results indicated that long-term exposure to PER increased the aggressive behavior of C57BL/6J mice.

Figure 1

Locomotor activity of PER and CON groups in the open field test. No significant difference in the total distance and center time was found between Con and PER groups in mice, respectively, which was shown in a computer-generated trace of the animal’s movements over 10 min (n = 6 mice per group; Student’s t-test). (A–D) Data are presented as mean ± SD.

Figure 2

Aggressive behavior test in two groups. (A) Latency to the first attack (p < 0.001); (B) number of attacks (p < 0.001); (C) attack duration (p < 0.01). Data represent mean ± SD (n = 6 per group). **p < 0.01 and ***p < 0.001 indicate significant statistical difference between groups.

Differential Expression of Hippocampal Proteins

In the current study, PER exposure disrupted hippocampal proteomic profiles of mice. A total of 6849 proteins were identified by one or more unique peptides from the protein database (false discovery rate < 1%) by TMT-based quantitative proteomic analysis. Corresponding to 6849 proteins, 93 proteins were significantly altered in the hippocampus tissue of the PER group (corrected p < 0.05); we used GO annotation to identify the functional significances of all identified proteins, which were divided into multiple functional groups, mainly related to synaptic function, synaptogenesis, synaptic plasticity, postsynaptic density protein, vesicular transport, neurite outgrowth, AMPA-type glutamate receptor immobilization, Ras GTPase, and others. The detailed information of the differentially expressed proteins (DEPs) is listed in Table . Among the DEPs, 59 proteins were up-regulated and 34 proteins were significantly down-regulated in perampanel-treated mice compared with the control group. We also conducted DEP clustering using the hierarchical cluster analysis and generated heatmap in which the two comparable groups can be easily recognized (Figure A,B), highlighting the notable differences between PER and Con groups.

Table 1

List of Quantitatively Modulated Differentially Expressed Proteins in the Hippocampus Identified by TMT Proteomics Following PER Exposurea

accession	description	Cov	Uniq	M/C	t-test p-value
Q80YS6	actin filament-associated protein 1, GN = Afap1	1	1	2.771190952	0.000318871
Q8K3K8	optineurin, GN = Optn	2	1	1.392116859	0.009811607
P07903	DNA excision repair protein ERCC-1, GN = Ercc1	7	1	1.310416285	0.007013598
P97328	ketohexokinase, GN = Khk	2	1	1.280888319	0.028568005
Q9CWQ3	mitochondrial inner membrane protease ATP23 homolog, GN = Atp23	4	1	1.218091972	0.026972971
P29351	tyrosine-protein phosphatase non-receptor type 6, GN = Ptpn6	3	2	1.211774969	0.002227745
O08967	cytohesin-3, GN = Cyth3	12	1	1.20656536	0.033370933
Q99K95	replication termination factor 2, GN = Rtf2	3	1	1.194115843	0.036390341
Q91ZE0	trimethyllysine dioxygenase, mitochondrial, GN = Tmlhe	2	1	1.190253971	0.030854053
Q8VBT0	thioredoxin-related transmembrane protein 1, GN = Tmx1	23	8	1.187900082	0.029794138
Q61333	tumor necrosis factor alpha-induced protein 2, GN = Tnfaip2	1	1	1.167683579	0.024939168
Q61249	immunoglobulin-binding protein 1, GN = Igbp1	21	6	1.165655296	0.008190139
P70460	vasodilator-stimulated phosphoprotein, GN = Vasp	9	3	1.157812095	0.022705772
Q5DQR4	syntaxin-binding protein 5-like, GN = Stxbp5l	21	18	1.14660469	0.019967117
Q0VBL3	RNA-binding protein 15, GN = Rbm15	3	3	1.146047039	0.029562494
P15208	insulin receptor, GN = Insr	4	5	1.140066392	0.003800028
Q8BGA3	leucine-rich repeat transmembrane neuronal protein 2, GN = Lrrtm2	6	3	1.123499392	0.049398248
P56393	cytochrome c oxidase subunit 7B, GN = Cox7b	10	2	1.120774032	0.028593212
Q8BKI2	trinucleotide repeat-containing gene 6B protein, GN = Tnrc6b	1	2	1.119896434	0.048432502
Q8BLU0	leucine-rich repeat transmembrane protein FLRT2, GN = Flrt2	6	4	1.119183137	0.031456573
P70372	ELAV-like protein 1, GN = Elavl1	35	10	1.109275956	0.043608033
Q99MP8	BRCA1-associated protein, GN = Brap	4	2	1.108830808	0.022466448
Q99JF5	diphosphomevalonate decarboxylase, GN = Mvd	13	5	1.108657253	0.048063462
F6SEU4	Ras/Rap GTPase-activating protein SynGAP, GN = Syngap1	45	49	1.108032821	0.015276438
O35144	telomeric repeat-binding factor 2, GN = Terf2	1	1	1.102550161	0.017687406
Q68FF0	uncharacterized protein KIAA1841, GN = Kiaa1841	2	2	1.100563703	0.040458368
Q8JZP2	synapsin-3, GN = Syn3	30	11	1.098423376	0.021453436
Q9CZN4	protein shisa-9, OS = Mus musculus, OX = 10090, GN = Shisa9	15	6	1.093834958	0.027522023
Q8BLB7	lethal(3)malignant brain tumor-like protein 3, GN = L3mbtl3	1	1	1.08984348	0.018365628
P05622	platelet-derived growth factor receptor beta, GN = Pdgfrb	3	3	1.089357506	0.002160346
Q8R2H9	phosphoethanolamine/phosphocholine phosphatase, GN = Phospho1	19	6	1.087790485	0.018061833
Q4ACU6	SH3 and multiple ankyrin repeat domains protein 3, GN = Shank3	31	41	1.076513927	0.022212105
P52189	inward rectifier potassium channel 4, GN = Kcnj4	9	4	1.07594632	0.002652225
Q91YQ3	cold shock domain-containing protein C2, GN = Csdc2	25	5	1.069415813	0.001378977
Q8VHQ3	protein phosphatase 1 regulatory inhibitor subunit 16B, GN = Ppp1r16b	7	4	1.068784484	0.043779232
Q3TC72	fumarylacetoacetate hydrolase domain-containing protein 2A, GN = Fahd2	57	12	1.067599347	0.044817035
Q78RX3	small integral membrane protein 12, GN = Smim12	27	3	1.065361562	0.04945411
Q9D009	putative lipoyltransferase 2, GN = Lipt2	16	4	1.063020604	0.04778066
Q9JK45	potassium voltage-gated channel subfamily KQT member 5, GN = Kcnq5	1	1	1.061013369	0.049625226
Q9WTU6	mitogen-activated protein kinase 9, GN = Mapk9	25	5	1.060252203	0.021940713
Q3UIU2	NADH dehydrogenase [ubiquinone] 1 beta subcomplex subunit 6, GN = Ndufb6	41	6	1.059345305	0.048990551
P46735	unconventional myosin-Ib, GN = Myo1b	20	21	1.056683157	0.049709811
Q91YX5	Acyl-CoA:lysophosphatidylglycerol acyltransferase 1, GN = Lpgat1	21	8	1.048716656	0.040300128
Q9ER39	torsin-1A, GN = Tor1a	2	1	1.047414076	0.022220112
Q99J56	derlin-1, GN = Derl1	11	3	1.046398534	0.015974799
Q69ZZ6	transmembrane and coiled-coil domains protein 1, GN = Tmcc1	20	9	1.0450555	0.049090563
Q62108	disks large homolog 4, OS = Mus musculus, OX = 10090, GN = Dlg4	54	30	1.041440349	0.015860967
Q811I0	ATP synthase mitochondrial F1 complex assembly factor 1, GN = Atpaf1	34	10	1.041126873	0.011658994
Q8BFR4	N-acetylglucosamine-6-sulfatase, GN = Gns	11	6	1.038349909	0.042754672
Q3UH99	protein shisa-6, OS = Mus musculus, OX = 10090, GN = Shisa6	19	9	1.036913837	0.023913209
Q9D0M1	phosphoribosyl pyrophosphate synthase-associated protein 1, GN = Prpsap1	47	11	1.034563714	0.00578867
Q9JM52	misshapen-like kinase 1, OS = Mus musculus, OX = 10090, GN = Mink1	27	27	1.033506063	0.003627109
Q60738	zinc transporter 1, GN = Slc30a1	12	5	1.032821845	0.042890271
Q91YM4	FAST kinase domain-containing protein 4, GN = Tbrg4	5	4	1.032820741	0.014491126
Q921W4	quinone oxidoreductase-like protein 1, GN = Cryzl1	30	9	1.032033008	0.034037391
Q8JZN7	mitochondrial Rho GTPase 2, GN = Rhot2	9	5	1.030800469	0.029075355
Q9Z2I8	succinate–CoA ligase [GDP-forming] subunit beta, mitochondrial, GN = Suclg2	39	18	1.027400076	0.035306881
Q99MR0	actin-like protein 6B, GN = Actl6b	21	6	1.024814164	0.036668562
Q78PG9	coiled-coil domain-containing protein 25, GN = Ccdc25	19	4	1.023158245	0.035135783
E9PUL5	proline-rich transmembrane protein 2, GN = Prrt2	37	8	0.982084039	0.018897725
Q63829	COMM domain-containing protein 3, GN = Commd3	23	5	0.980662277	0.034438143
Q9CQF4	mitochondrial transcription rescue factor 1, GN = Mtres1	13	3	0.97898691	0.040461715
Q9DCT2	NADH dehydrogenase iron–sulfur protein 3, GN = Ndufs3	48	14	0.968398505	0.022386004
Q8CH72	E3 ubiquitin-protein ligase TRIM32, GN = Trim32	22	13	0.965597426	0.03594526
Q8R4F1	netrin-G2, GN = Ntng2	4	3	0.965217544	0.000994181
P11103	poly [ADP-ribose] polymerase 1, GN = Parp1	9	8	0.964471006	0.040301374
Q91VK4	integral membrane protein 2C, GN = Itm2c	34	6	0.962686811	0.0424762
Q91ZH7	phospholipase ABHD3, GN = Abhd3	10	4	0.960783265	0.008175226
Q91X97	neurocalcin-delta, GN = Ncald	77	7	0.958409806	0.033373857
O08644	ephrin type-B receptor 6, GN = Ephb6	6	5	0.94969615	0.038122544
P70408	cadherin-10, GN = Cdh10	16	10	0.946552755	0.030902986
Q9QUN9	dickkopf-related protein 3, GN = Dkk3	18	6	0.945815891	0.014993276
Q9WV76	AP-4 complex subunit beta-1, GN = Ap4b1	2	1	0.939870283	0.026232922
Q3TQF0	F-box only protein 31, GN = Fbxo31	2	1	0.938756477	0.023650774
O55033	cytoplasmic protein NCK2, GN = Nck2	23	8	0.934122038	0.006039627
P23953	carboxylesterase 1C, GN = Ces1c	11	6	0.933377651	0.022455845
Q3UFY7	7-methylguanosine phosphate-specific 5′-nucleotidase, GN = Nt5c3b	15	4	0.932564335	0.042808153
O08677	kininogen-1, GN = Kng1	11	6	0.931376613	0.046635589
Q8R2U6	diphosphoinositol polyphosphate phosphohydrolase 2, GN = Nudt4	25	1	0.928412788	0.00365918
P32848	parvalbumin alpha, GN = Pvalb	73	12	0.927436716	0.036932494
O35089	protein cornichon homolog 2, GN = Cnih2	6	1	0.921967153	0.030133405
Q8VE96	solute carrier family 35 member F6, GN = Slc35f6	3	1	0.920103006	0.011340532
Q9ESB3	histidine-rich glycoprotein, GN = Hrg	3	2	0.914787183	0.042555487
Q9CQ25	mitotic-spindle organizing protein 2, GN = Mzt2	9	1	0.914477964	0.046210189
Q9CQM0	nicolin-1, GN = Nicn1	4	1	0.90052069	0.001735753
O70250	phosphoglycerate mutase 2, GN = Pgam2	34	5	0.899310792	0.04662173
Q61838	pregnancy zone protein, GN = Pzp	20	30	0.893129596	0.008922305
Q8K013	GTP-binding protein 10, GN = Gtpbp10	2	1	0.87500987	0.046151603
P02798	metallothionein-2, GN = Mt2	13	1	0.859901481	0.019934253
P01864	Ig gamma-2A chain C region secreted form	15	3	0.859231943	0.026535001
Q8BGZ7	keratin, type II cytoskeletal 75, GN = Krt75	13	2	0.843522803	0.03382069
Q3V016	homeodomain-interacting protein kinase 4, GN = Hipk4	1	1	0.733170586	0.002472431
Q2WF71	leucine-rich repeat and fibronectin type III domain-containing protein 1, GN = Lrfn1	12	1	0.732267503	0.035409077

Note: Cov, covered by identified peptides; Uniq, unique peptide number; M, PER group; C, Con group.

Figure 3

Characteristics of the identified proteins in PER and CON groups. (A) Hierarchical cluster analysis of 93 differentially expressed proteins. Whole hippocampal samples are represented in the columns, and the IDs of differentially expressed proteins are delineated in rows. The color bar located below the figure shows that red indicates significant up-regulation and blue represents down-regulation. M1, M2, and M3, three replicates of perampanel-treated mice; C1, C2, and C3, three replicates of control mice. (B) Volcano plot for the DEPs in the comparison group according to fold change (M/C) and p value (t-test). Red and blue dots indicate significantly up-regulated and down-regulated proteins, respectively, and proteins with no difference are gray.

Enrichment Analysis of Differentially Expressed Proteins

To better understand how the functional role of DEPs was disrupted by PER exposure, GO annotation was used to categorize according to molecular function (MF), cellular component (CC), and biological process (BP). To detect significantly enriched biological function types, we performed GO enrichment analysis and ranked the terms by enrichment score (−log10(p value)). According to the GO enrichment results, the top 20 enriched items were displayed by the enrichment factors (Figure ).

Figure 4

GO functional enrichment analysis of the DEPs in three categories, biological process, molecular function, and cellular component. The p-value represents the enriched degree.

The biological process of the identified proteins mainly involved regulation of neuronal synaptic plasticity, regulation of AMPA receptor activity, postsynaptic density organization, regulation of synapse structure or activity, vocalization behavior, regulation of vesicle-mediated transport, and anatomical structure homeostasis. The results for molecular function showed the strong enrichments of SH3 domain binding, neurexin family protein binding, structural constituent of postsynaptic density, nerve growth factor binding, and neuroligin family protein binding. Cellular component annotation displayed the intrinsic component of postsynaptic density membrane, synaptic membrane, dendritic spine membrane, dendrite membrane, anchored component of synaptic membrane, and neuron projection membrane. KEGG pathway analysis (p < 0.05) revealed that the statistically significant enriched terms included the Ras signaling pathway, IL-17 signaling pathway, JAK–STAT signaling pathway, Rap1 signaling pathway, focal adhesion, axon guidance, glycine, serine, and threonine metabolism, cell adhesion molecules, and neuroactive ligand–receptor interaction (Figure ). The Ras signaling pathway and axon guidance are the most significant changes in up-regulation and down-regulation pathways.

Figure 5

KEGG pathway enrichment analysis of the DEPs. The p-value represents the enriched degree. Orange, significant increase; blue, significant decrease.

STRING Analysis of Protein–Protein Interaction Networks

To find out the correlation between protein interactions and related biological processes, a protein–protein interaction (PPI) network of the 93 DEPs was constructed by the STRING database (Figure ).

Figure 6

Protein–protein interaction network analysis of the DEPs using STRING. The line thickness represents the confidence scores, and thicker connection lines indicate the higher confidence of protein–protein interaction. Nodes are labeled with gene names.

Proteins regarding synaptic function accounted for about half of all regulated proteins. It was obviously revealed that at least four closely coexpressed clusters were displayed in the complex PPI network, including Vasp, Nck2, DLg4, and Shisa6. These clusters were mainly associated with neural development and dendritic outgrowth, Ras pathway, postsynaptic density protein, and AMPA-type glutamate receptor. The neural development and dendritic outgrowth-related cluster included Vasp, Actl6b, Ptpn6, Insr, and Mapk9. Proteins in the Ras pathway-related cluster included Nck2, Insr, Ephb6, Mink1, Pdgfrb, and Syngap1. Proteins in the postsynaptic density protein cluster included Dlg4, Syngap1, Lrrtm2, Shank3, Pvalb, Lrfn1, Prrt2, and Ntng2. In addition, the AMPA-type glutamate receptor-related cluster included Shisa6, Shisa9, Cnih2, and Dlg4. Taken together, the bioinformatic analysis of our study revealed that synaptic plasticity and the Ras signaling pathway may play an important role in the pathogenesis of aggressive behavior.

Validation of Differentially Expressed Proteins

To confirm the potential bioinformatic analysis significance in aggressive behavior induced by PER, further verification of Syn, PSD95, p-ERK1/2, and GluA1 was performed using western blot. The following reasons were used for choosing the four proteins: (a) most significant proteins for functional enrichment and KEGG analysis; (b) signal proteins of synaptic plasticity and the Ras signaling pathway; (c) relevance to synaptic function and commercially available antibodies; (d) perampanel is a relatively safe drug, and the only drug target is the AMPA receptor. The results of western blot showed that expression of Syn and PSD95 was down-regulated and that of p-ERK and GluA1 proteins was up-regulated in the PER-treated mice compared to the control groups (p < 0.05 and p < 0.001; Figure ).

Figure 7

Western blot validations of Syn, PSD95, p-ERK1/2, and GluA1 proteins in the hippocampus. (A) Western blot images. (B) The protein expression levels were quantitatively analyzed with the tubulin level. Data are presented as mean ± SD. Student’s t-test, *p < 0.05, ***p < 0.001 versus the control group. n = 6 for each group.

Discussion

There is increasing amount of clinical research revealing that the long-term use of perampanel for refractory partial-onset epilepsy has caused an abundance of concern about the adverse events and health risks of aggressive behavior.[26−28] However, the underlying pathogenesis and neurobiological significance remain poorly understood. In our current study, we have carried out quantitative proteomic methods to identify changes in protein profile alternations induced by PER in the hippocampus of an aggressive behavior mice model. Half of DEPs were mainly directed to neurite growth, postsynaptic density, AMPA receptor, and Ras signaling pathway. Furthermore, western blot assay further demonstrated the alteration of Syn, PSD95, GluA1, and p-ERK in PER-treated mice. Gene function enrichment and pathway visualization analysis objectively displayed that dysregulation of synaptic plasticity and the Ras signaling pathway might mainly mediate the pathological mechanism.

Our study successfully established aggressive behavior in mice after oral administration of PER at a dose of 2 mg/kg per day for 2 months. Resident–intruder test results showed that aggressive behavior was significantly enhanced in the PER group compared to the control group. To our knowledge, no previous animal studies have formalized the representation of PER-induced aggressive behavior in male mice. Nonetheless, these results are comparable with the already published controlled randomized clinical studies. Irritability and aggressive behavior were listed in perampanel’s instructions since it was first marketed.[29] There was evidence shown that around 25% patients with severe drug-resistant focal epilepsy revealed PER-related aggressiveness.[23] In a recent prospective study, Goji and Kanemoto indicated that PER increases assessment scores indicative of aggression.[30] Collectively, these observations seen in experimental animal models are qualitatively consistent with previous cases of aggressive behavior.

Proteomics has recently been exploited to analyze protein expression profiling and contributes to revealing the signal pathways of external environmental stimuli and pathological processes of drug action as well as the biological regulatory mechanisms associated with disease pathology.[31,32] However, there are no reports on studies of PER-induced aggression using proteomic methods. Our results showed that PER-induced changes in hippocampal proteins were manifested as a disturbance of synaptic plasticity and the Ras signaling pathway.

We found that PER treatment induced the decrease in Syn and PSD95 proteins in the hippocampus. Synaptic plasticity-related proteins mainly include synaptophysin and postsynaptic density 95,[33] which have received increasing attention as a means of modifying innate and adaptive behavioral responses.[34] Changes in synaptic plasticity-related proteins reflect their roles in axonal growth, synaptic reconstruction, and synaptic connection.[35] The most significantly enriched pathway associated with aggressive behavior in humans is axon guidance.[36] Previous work has shown that perinatal fluoxetine exposure resulted in significantly decreased rates of hippocampal synaptophysin density, which was correlated with increased social aggression.[37] In another study, maternal separation mice showed reduced synaptic function at the hippocampal synapses, which was significantly associated with more aggressive behavior.[38] In additional animal work, Stxbp1+/– mice that exhibited impaired synaptic function and excitatory neurons showed a pronounced greater number and prolonged duration of aggression than the wild-type mice.[39] In our opinion, synaptic protein and neurite growth seem to play an important regulatory role in aggressive behavior.

Postsynaptic density 95 (PSD95) is highly gathered at the synapses and accounted for the assembly of the abundant protein complexes indispensable for synaptic plasticity and neural development.[40] PSD95 plays a crucial role in capturing and immobilizing surface AMPA receptors,[41] which interacted with the stargazin auxiliary subunit to anchor and increase the number of AMPA receptors at the synapse.[42] Synaptic morphological plasticity and constructional plasticity are the material foundation of functional plasticity.[43] PSD95 has been extensively related to the modulation of pro-social and emotional behavior.[44] A previous study showed that male PSD95 mice demonstrated higher levels of aggression.[45] In addition, Dlgap2 mice exerted a shorter and thinner PSD with respect to wild-type mice in cortical synapses; thus, Dlgap2 mice displayed heightened aggressive behavior.[46] The DLG5 gene related to neuritogenesis was associated with aggressive behavior in candidate gene association studies (CGAS) mentioned above.[36] Overall, the down-regulation of PSD95 protein was correlated with aggressive behavior, supporting a common mechanism of aggressive behavior.

In addition, PER treatment induced the increasing levels of AMPA-type glutamate receptors in mice. AMPA receptors are especially intriguing since they have direct and lasting effects on aggression.[47] A growing body of experimental and clinical evidence suggests that AMPA receptors have gradually been valued for their role in aggressive behavior.[1,48] Previous studies have confirmed that genetic modification of the AMPA receptor leads to changes in aggressive behavior; blocking the AMPA receptor can reduce or increase aggressive behavior.[6,49,50] It has been suggested that mice lacking the AMPAR subunit and GluA1 mutations, resulting in a significantly reduced AMPAR function, are less aggressive than wild-type littermates.[36] Furthermore, several recent studies demonstrated that chronic ultrasonic exposure induces aggressive behavior in mice, resulting in decreased GluA1 expression and increased GluA2 expression in the hippocampus and prefrontal cortex.[51] The aggressiveness of intermale mice with decreased AMPAR function and AMPAR knockout mice is significantly lower than that of wild-type (WT) littermates.[52] AMPA receptors mainly mediate fast excitatory synaptic transmission, which plays an important role in synaptic transmission efficiency, neuron integration function, and synaptic plasticity.[53] Synaptic plasticity largely depends on the amount of AMPA receptors and the adjustment of gating characteristics.[54] Accumulating evidence suggests that dysfunctional synaptic rearrangement of AMPA receptors may play a critical role in aggression.[55] AMPA receptors are highly expressed in the CA1 and CA3 subregions and the dentate gyrus of the hippocampus, which is also believed to play a significant role in determining whether to attack or not. Compared with the non-aggressive control group, the levels of AMPA receptors in the isolation-induced aggressive mice were significantly higher in the amygdala and prefrontal cortex.[56] Studies have shown that AMPA receptors are expressed throughout the brain, and their normal activation is essential for modulation of synaptic plasticity (such as long-term potential enhancement (LTP) and long-term depression (LTD)).[57] LTP and LTD can occur on the hypothalamic synapses, which are important for attacking.[58] LTP is part of the causal mechanism and increases aggressiveness through repeated successful attacks.[59] Taken together, our results reveal that PER-induced aggression is involved in synaptic plasticity.

Proteomic data showed that PER treatment activated Ras pathway-related proteins in mice. The extracellular signal-regulated kinase (ERK) is a key member in the Ras intracellular signaling pathway and could regulate hippocampal synaptic transmission and plasticity.[60] In this study, it was clear that the phosphorylation of ERK1/2 expression was activated in PER-treated mice in contrast with the control group. Our results are consistent with the previous studies stating that PER could affect the phosphorylation of ERK1/2 expression in the hippocampus.[61] Enrichment analysis studies showed that aggression genes are involved in Ras–ERK signaling. ERK1/2 signaling forms intracellular signaling cascades and the phosphorylation of ERK1/2 coordinates cellular responses that are usually associated with aggression.[62] There is growing evidence that the ERK1/2 kinases trigger cascade phosphorylation events to regulate AMPAR subunit gene expression and synaptic plasticity.[63]Calotropis gigantea significantly increased the p-ERK1/2 expression in hippocampal neurons to promote neurite outgrowth and synaptogenesis.[64] In particular, Ras initiates extracellular signal-regulated kinase (ERK) signaling to drive synaptic delivery of AMPA receptors containing subunits with long cytoplasmic termini.[65] These results suggest that abnormal Ras/ERK signaling and phosphorylation of ERK1/2 may be involved in the aggressive behavior induced by PER treatment.

Limitations and Considerations

Some limitations related to the current study should be considered here. Although we were able to validate our findings from the proteomic comparison study, it is possible that some important synaptic protein level changes may have been missed due to the relatively small number of sample size in the LC–MS analyses; thus, these findings should be interpreted with caution and need to be confirmed in larger samples. Aggressive behavior is a complex phenomenon that may involve different emotional states in humans, such as increased anger, high emotional responses, or decreased control. Animal models only reflect the specific aggressive manifestations and cannot fully explain how perampanel induced aggressive behavior. Therefore, further proteomic research should be performed with cerebrospinal fluid or blood samples from aggressive behavior individuals. Neuroimaging studies help quantify the activity of specific brain regions or circuits associated with aggressive behavior. The study of proteomics combined with neuroimaging data can more effectively determine the cause of perampanel-induced aggressive behavior.

Conclusions

This is the first report of the whole proteomic changes in the hippocampus for perampanel-induced aggressive behavior. Chronic PER exposure induces aggressive behavior in mice, accompanied by changes in certain hippocampal proteins, particularly those involved in synaptic plasticity and the Ras–ERK signaling pathway. Changes in hippocampal synapses and AMPA receptor proteins may be involved in PER-induced aggression. Our results will contribute to the understanding of the roles of proteins in pathophysiological mechanisms of aggressive behavior induced by PER. The disturbance of these proteins in the hippocampus could be related to the aberrant aggressive behavior. The identified biological functions and pathways in the hippocampus are one of the mechanisms of PER-induced aggressive behavior in humans and thus must be further investigated.