Lipidomics Reveals Glycerophospholipid Metabolism Dysregulation in the Corpus Striatum of Mice Treated with Cefepime

Cefepime exhibits a broad spectrum of antimicrobial activity and thus is widely used for severe bacterial infection. Adverse effects on the central nervous system (CNS) have been reported in the patients treated with cefepime. Current explanation for the neurobehavioral effect of cefepime mainly attributes to its ability to across blood-brain barrier and to competitively bind to GABAergic receptor; however, the underlying mechanism is largely unknown. In this study, mice were intraperitoneally administered 80 mg/kg cefepime for different time period, including 1 day, 3 days, 5 days, 7 days and 10 days, and then LC/MS-MS-based metabolomics was used to investigate the effect of cefepime on the brain lipidomic proling and metabolic pathway. Repeated cefepime treatment time-dependently caused anxiety-like behavior accompanied with the reduced locomotor activity in the open eld test. Cefepime profoundly altered the lipid prole in the corpus striatum, and glycerophospholipid contributed to a large proportion among those signicantly modied lipids. Cefepime also signicantly modied acyl chain length and unsaturation of fatty acids. In addition, cefepime obviously altered the morphology of neurite, mitochondria and synaptic vesicles of striatal neuron in vitro. Collectively, our results show that cefepime reprograms glycerophospholipid metabolism of corpus striatum, which may underly its neurobehavioral effect.


Introduction
Cefepime is the fourth-generation cephalosporin antibiotic and a broad-spectrum antibiotic applying to hospitalized patients with a range of infections. However, about 10 ~ 23% of patients have been discovered aberrant nervous system-related behavior on average four to ve days after starting cefepime therapy. The most frequently encountered symptoms are the reduced consciousness, myoclonus, confusion, aphasia, seizures, agitation (Honore and Spapen. 2015; Kamboj et al. 2020;Khasani, 2015;Payne et al. 2017). The proposed pathophysiology for these adverse effects of cefepime has been attributed to its ability to across the blood-brain barrier (BBB) and competitively bind to the γ-amino butyric acid (GABA) receptors, eventually leading to the suppression of inhibitory neurotransmission (Amakhin et al. 2018; Bhattacharyya et al. 2016; Payne et al. 2017). Risks factors of cefepime include excessive dosing, enhanced serum level, renal insu ciency, preexisting brain injury, patients with hematological malignancy or underwent intensive care during hospitalization. Occasionally, the adverse side effects still occur despite appropriate cefepime application in clinic. The mechanism underlying the adverse event of cefepime is still largely unknown.
Brain is the second highest lipid-enriched organ behind adipose tissue, and total lipids mass contribute to the half of brain dry weight. Diverse lipids play pleiotropic crucial roles in the brain, ranging from the backbone for biological membrane, to cell signaling molecules, and regulation of immune responses (Beaulieu et  Phospholipids are amphipathic molecules with so-called polar head group and nonpolar tail ends. The non-polar tails contain hydrocarbon chains that vary in the chain length of fatty acid, double bond number and position. These characters determine the structure and geometry of fatty acyl chains, eventually governing the shape of lipids, the degree of lipid packing within bilayer, and membrane uidity. Membrane uidity is a key parameter for membrane fusion, in uencing lipid mobility and protein function. Disturbance of glycerophospholipids metabolism in the neural membrane remarkably alters the activities of membrane-bound proteins, such as enzymes, receptors and ion channels (Hanada,  In the present study, we performed unbiased lipidomic analysis to explore the effect of cefepime on neurobehaviors in mice. We found that several classes of lipids in the corpus striatum was profoundly altered after cefepime exposure for 5 continuous days, and glycerophospholipids took a large proportion among those modi ed lipids. In combination with the ndings from the cultured striatal neurons exposed to cefepime, we propose that dysregulated glycerophospholipid metabolism may contribute to CNS adverse effect of cefepime in mice.

Animals
Adult male C57BL/6 mice (6-8 weeks, body weight 18-22 g) were purchased from Vital River Laboratory Animal Technology Co. Ltd (Beijing, China). All mice were housed under standard condition (a 12-h light/12-h dark cycle with light on from 07:00-19:00) with a constant room temperature. All experimental procedures and use of the animals were conducted in accordance with the guidelines established by the Association for Assessment and Accreditation of Laboratory Animal Care and the Institutional Animal Care and Use Committee of Sichuan University.

Drug Injection
Mice were intraperitoneally administered 80 mg/kg cefepime for different time period. This dosage was based on the pharmacokinetic study of cefepime single regimen and mono-therapeutic dose in murine model (Bhagwat et al. 2019;Pechère and Vladoianu, 1992).

Open eld test (OFT)
All experiments were carried out in a dark cycle, started from 90 minutes turned off the lights, and arrived at the test room after at least 2 hours of adaptation time. This test schedule and method followed previously published research (Tomihara et al. 2009). Open led test was applied to check the movement activity and anxiety-like behavior. A 16-beam animal activity monitor was used to divide the black acrylic box (48 x 48 x 30 cm) into center and periphery. EthoVision version 7.0 software (Noldus Information Technology) was applied to analyze a series of parameters basing on the recorded activity, including total movement distance, time spent in the inner zone (24 x 24 x 30 cm) and the number of entries into the inner zone. The movement time only refers to the time consumed by the animal moving but exclude the time without moving. At the beginning of rst adaption, mouse was placed in the same position (rear left corner) of the arena and allowed to spontaneously move in the open eld arena for 5 minutes.

Blood collection and tissue isolation
Mice were immediately killed by rapid decapitation after the end point of open eld test. Blood was collected for the measurement of serum biochemistry. Brain tissue and other major organs were subsequently dissociated out for lipidomic analysis and histopathological evaluation. For lipidomic pro ling analysis, the isolated cerebral cortex, hippocampus and corpus striatum were quickly frozen via dipping into liquid nitrogen. The frozen tissues were stored at -80°C until lipid extraction. Histopathological assessment was performed through HE staining. Heart, liver, spleen, lung, kidney and whole brain of the mouse were dissociated and xed in a 4% paraformaldehyde solution, followed by tissue sectioning and HE staining.

Blood biochemical test
A series of blood chemical parameters were detected using the puri ed serum without any contamination of blood cells. Brie y, the collected fresh blood samples were puri ed through centrifuging at 1500 rpm for 10 min and the serum in the upper layer was aspirated for further test. The biochemical parameters were assessed using an automatic biochemical analyzer (Cobas c311, Switzerland). The following parameters, the serum alanine aminotransferase (ALT), serum albumin (ALB), creatine kinase isoenzyme (CK-MB), alkaline phosphatase (ALP), aspartate aminotransferase (AST), serum total protein (TP), cholesterol (CHOL), uric acid (UA), urea, lactate dehydrogenase (LDH), triglyceride (TG), direct bilirubin (DBiL) and serum creatinine (Cre) were measured.

HE staining
The major organs, including heart, liver, spleen, lung, kidney and brain, were exteriorized and xed in 4% paraformaldehyde solution. The xed tissue was embedded with para n, and was further sectioned with a thickness of 5 µm after dewaxing and rehydration. The sections were stained with hematoxylin solution for 5 minutes, and then immersed in 1% acidic ethanol for another 5 times. The sections were then counterstained with eosin-phloxine solution for 1 minutes, and dehydrated with a gradient alcohol and cleared with in xylene.

Lipid extraction
Lipids extracts from mouse brain were prepared basing on methyl tert-butyl ether (MTBE) lipid extraction methods. Brie y, pre-chilled methanol (150 µl) and MTBE (450 µl) were sequentially added into the frozen tissue (25-30 mg). The mixture was incubated at room temperature for 10 minutes and subsequently homogenized at 6,500 rpm in three cycles for 15 s with 20s intervals using a bead-based homogenizer (Precellys Evolution, Bertin Technologies, Montigny le Bretonneux, France). Then, aspirate 300 µl of 25% methanol diluted in sterile MiliQ water and add it into the homogeneous mixture in order to form a phase separation. Centrifuge at 14,000 g for 10 minutes after vigorous mixing. Aspirate the upper organic phase carefully without any disturbing the middle layer and transfer it into another tube. The extracted lipids were evaporated at room temperature under a gentle stream of nitrogen and stored at -80°C until use.

Mass spectrometry analysis of lipid metabolites
Lipid analysis was performed on an Acquity Ultra Performance Liquid Chromatography (UPLC system) coupled with hybrid quadrupole orthogonal time-of-ight lass Spectrometer (Waters, Milford, MA, USA). Electrospray positive and negative ionization modes were used. In detail, the extracted lipids powder was rstly re-dissolved in acetonitrile/isopropanol (v/v, 7:3). Then, the injected lipids were separated through Waters Acquity HSS T3 Column (Waters, Milford, MA, USA) with a constant temperature at 55°C in column oven. The average column pressures were maintained around ca. 10,000 psi. Elution buffer A consists of acetonitrile and water with 10 mM ammonium acetate (40:60, v/v), and elution buffer B consists of acetonitrile and isopropanol (10:90, v/v) with 10 mM ammonium acetate. The sample analysis was performed by a binary gradient system over 15 min total run time. At the start point of the gradient, it was held at 60% elution buffer A and 40% elution buffer B. In the following 10 min, the gradient was ramped in a linear fashion to 100% elution buffer B and held at this composition for another 2 min. The system was switched back to 60% elution buffer B and 40% elution buffer A, and equilibrated for an additional 3 min. The ow rate was set at 0.4 mL/min and the injection volume was 10 µl.
The capillary voltage and cone voltage were separately set at 3.

Data processing and analysis
Progenesis QI software (Newcastle, UK) was used to process UPLC-ESI-TOFMS data. Data were sequentially processed through comparison, peak selection and lipid identi cation. We identi ed the lipid metabolites by referring primarily to the lipid map database (www.lipidmaps.org) and the Human Metabolome Database (https://hmdb.ca/). Data tables were obtained from the Progenesis QI software and the absolute intensities of all identi ed compounds were recalculated based on the relative abundance of lipid molecules. The data was initially processed by unsupervised principal component analysis (PCA) to obtain group clusters. Then, the supervised orthogonal partial least-least squares discriminate analysis (OPLS-DA) model was utilized to screen those m/z that impacts on the group clustering. All m/z with variable importance (VIP) value above 1 were screened for further analysis of variance (ANOVA) using Progenesis QI software. The corresponding lipids of m/z with VIP > 1 and pvalues < 0.05 were identi ed as signi cant difference. Lastly, pathway enrichment analysis of signi cantly altered lipids was performed in MetaboAnalyst 4.0. For normality, log transformations and automatic scaling were applied to the data.

Primary neuron cultures
The primary striatal neurons were obtained from mouse embryos (E16 to E17) (Fath et al. 2009;Wu et al. 2016). In brief, corpus striatum was carefully cut into small pieces and subsequently digested with 0.25% trypsin at 37 ℃ for 20 min. The puri ed neuronal pellet was resuspended with culture medium containing neurobasal medium, 1% B27 supplement (Gibco), 2 mM Glutamine and 0.2% Primocin (InvivoGen). Neurons were cultured in 6-well plate with poly-D-ornithine (0.5mg/ml)-coated coverslip at 37°C and 5% CO 2 in a humidi ed incubator. Half of medium was changed with fresh culture medium every 3 days.

Imaging
Brie y, neurons grown on the coverslip were directly xed with warm 4% paraformaldehyde/ 4% sucrose solution for 10 min at room temperature after treatment with or without cefepime for 24 h. Neurons were permeabilized with 0.1% Triton X-100 in PBS for 10 minutes at room temperature, and subsequently blocked with 5% bovine serum albumin (BSA). The coverslip was incubated with primary antibody (1:50 − 1:1000) at 4°C overnight, followed by incubation with Alexa-conjugated secondary antibody (1:200) for 1 h. All IF images were obtained using a laser scanning confocal microscope (Leica SP8 X, Leica) with built-in LAS X software, using a 63x3 1.3 NA oil lens. In order to ensure the authenticity of the uorescence statistics, the laser intensity of each sample group was consistent. The following antibodies were used: TOM20 (11802-1-AP, Proteintech), Alexa uor 647 Phalloidin (A22287, Invitrogen), SV2 (AB2315387, Developmental Studies Hybridoma Bank).
For the quanti cation of mitochondrial morphology and density, as well as the density of synaptic vesicles, around 20 neurons were blindly selected and utilized for quanti cation. The quanti cation procedure was performed in ImageJ (Version 1.52P). The line tool was rstly used to trace the pro le of neurite, and the size of area or the uorescence intensity of TOM20 or SV2 was calculated. For obtaining the density of mitochondria, the uorescence intensity of mitochondria was divided into the corresponding neuritic area.

Statistical Analysis
All data were expressed as mean ± SD deviation, and measured by GraphPad Prism 8 software using multiple comparison, ordinary one-way Analysis of Variance (ANOVA). The profoundly modi ed lipids were measured using multiple comparison, two-way ANOVA. The p value < 0.05 was considered statistically signi cant (*p < 0.05, ** or & p < 0.01, *** or # p < 0.001).

Cefepime administration causes anxiety-like behaviors in mice
To investigate the effect of cefepime on neurobehaviors, the mice were intraperitoneally administered 80 mg/kg cefepime daily for 1, 3, 5, 7 or 10 days, respectively. Open eld test was applied to investigate the effect of cefepime on general activity, exploratory and anxious behavior of mice. Mice were placed into the at open eld box (48×48×30 cm) after continuous cefepime administration (80 mg/kg) for indicated time period, and movement path during 5 min of spontaneous movement was simultaneously tracked (Fig. 1a).
Overall, cefepime-treated mice preferred to move around the periphery area of box, especially after cefepime treatment for 5 days. The trajectory around the periphery area displayed an increasing trend in a time-dependent manner (Fig. 1b). In contrast to control mice, the total exploratory time in inner zone of the at arena gradually decreased with the extension of cefepime administration (Fig. 1b). The total exploratory time spent in inner zone was signi cantly reduced from the beginning of the fth day after cefepime treatment (Fig. 1c). The reduced exploratory activity in cefepime-exposed mice was positively correlated to the decreased frequency of entries into the inner zone (Fig. 1d), suggesting that cefepime may induce anxiety-like behavior. Moreover, the total distance traveled in the whole open area was markedly decreased along with the extension of cefepime administration, suggesting that cefepime exposure may impair locomotor activity of mice (Fig. 1e). Collectively, these results showed that treatment of cefepime causes anxiety-like behaviors in mice.
3.2 Cefepime shows no effect on serum chemistry and organ histology in mice Serum chemistry and histology of major organs were assessed to evaluate the toxicity of cefepime on blood pro le and the integrity of tissue, respectively. As shown in Fig. 2a, in comparison to the control mice, the levels of ALT, AST and ALP were slightly reduced in the mice treated with cefepime for 5 continuous days. There were no signi cant difference in the levels of urea, creatinine or uric acid between control and cefepime-treated mice (Fig. 2b). Cefepime showed no obvious effect on the levels of serum lactate dehydrogenase (LDH) and myocardial-associated isoenzyme creatine kinase-MB (CK-MB) (Fig. 2c). The levels of serum albumin (ALB) and serum total protein (TP) in cefepime-treated mice were similar to those in control mice (Fig. 2d). Moreover, the concentration of serum cholesterol (CHOL) and triglyceride (TG) as well as the direct bilirubin were not altered by cefepime treatment (Fig. 2e and 2f). Finally, alteration of various organs was microscopically assessed via the histopathology. Cefepime treatment showed no effect on the histology of the major organs, including brain (hippocampus and cerebral cortex), heart, liver, spleen, lung and kidney (Fig. 2g).

Cefepime profoundly alters brain lipidomic pro le of corpus striatum
To address the mechanism contributing to behavioral abnormalities in cefepime-treated mice, a UPLC-QTOF-MS-based lipidomic approach was applied to characterize the in uence on globe lipids among interconnected brain regions, including hippocampus, corpus striatum and cerebral cortex. Firstly, to investigate the differences of lipidomic pro le among control group and multiple time-dependent cefepime-treated groups, all identi ed ion peaks (mass-to-charge ratio, m/z) under both positive ion mode (ESI+) and negative ion mode (ESI-) were subjected to the supervised orthogonal partial least squares discriminant analysis (OPLS-DA) model. The OPLS-DA score plots using one orthogonal component displayed clearly separation under both positive and negative modes among all groups in all three brain regions (Fig. 3a). Notably, with the prolongation of cefepime treatment, the clustering between control and cefepime groups exhibited more manifest in all detected brain regions (Fig. 3a), suggesting that brain lipids were signi cantly perturbated in cefepime-treated mice. Then, the ion peaks mediating clustering in score plots were investigated and exhibited via loading plot. The value of variable importance (VIP) above 1 (VIP≥1) was labeled with red color for great contribution to the separation among control and multiple cefepime groups (Fig. 3b). In the three brain regions, corpus striatum, hippocampus and cerebral cortex, 81, 111, and 74 positive ion peaks, and 89, 73 and 31 negative ion peaks, respectively, were remarkably modi ed by cefepime.
Among all detected m/z ratios in both modes, the profoundly up-regulated (p < 0.05, Fold-change > 2) and down-regulated (p < 0.05, Fold-change<-2) m/z ratios were subsequently identi ed. The results discovered much more m/z ratios were dysregulated by cefepime administration in corpus striatum than those in hippocampus and cerebral cortex (Fig. 3c). In combination with total numbers satisfying VIP≥1 under both ESI + and ESI-, corpus striatum was the brain region mostly affected by cefepime ( Fig. 3b and 3c).
To determine whether those dysregulated ion peaks in corpus striatum accumulated with the extension of cefepime application, the total number of profoundly altered ion peaks were further separately quanti ed using the acquired ion peaks from the different time period of cefepime treatment. The results showed that total numbers of m/z ratios indeed increased with the extension of cefepime treatment. Notably, the total number of dysregulated m/z ratio was signi cantly increased after cefepime exposure for 5 continuous days (149 of m/z ratios) than that on day 1 (16 of m/z ratios) and day 3 (56 of m/z ratios). Interestingly, there was no obvious changes of dysregulated lipid peaks after cefepime treatment for 7 or 10 days in comparison to cefepime treatment for 5 days continuously. The amount of disturbed lipids peaks boosted almost three times from 3 days to 5 days of cefepime treatment, indicating that lipidomic pro le of corpus striatum may be remarkably perturbed after cefepime treatment for 5 days (Fig. 3d).

Cefepime causes a remarkable dysregulation of glycerophospholipids in corpus striatum
To discover the differentially changed lipids in the corpus striatum of cefepime-treated mice, those ion peaks with both VIP≥1 basing on OPLS-DA model and p < 0.05 acquiring from the statistical analysis were subsequently de ned referring to the Lipid Maps Database (www. lipidmaps.org) and the Human Metabolome Database (http://www. hmdb.ca/). The results revealed that a total of 116 lipids of corpus striatum were signi cantly altered by cefepime (Supplemental Table 1). Basing on the current lipid category, besides a small proportion of lipids without classi cation, we found that cefepime widely modi ed seven groups of lipid classes among the total eight categories of lipids (Fig. 4a). The proportion of identi ed lipid classes glycerophospholipids (GP), glycerolipids (GL), prenol lipids (PR), fatty acyls (FA), sphingolipids (SP), sterol lipids (ST), saccharolipids (SL) and the other uncategorized lipids were 62.07%, 8.62%, 6.03%, 5.17%, 5.17%, 3.45%, 0.86% and 8.62%, respectively. Moreover, among all identi ed glycerophospholipids subclasses, phosphatidylethanolamine (PE) and phosphatidylcholine (PC) were two major subclasses accounting for the most proportion of modi ed glycerophospholipids caused by cefepime (Fig. 4b). Collectively, cefepime predominantly dysregulated glycerophospholipid pro le, especially PC and PE, in the corpus striatum.
Next, the level of change of those modi ed lipids along with the prolongation of cefepime treatment was quanti ed and presented in heatmap (Fig. 4c). Interestingly, the majority of PC and PE presented an upregulated trend. In consistent with the discovery in OPLS-DA model, a great number of lipids were remarkably boosted only after continuous treatment with cefepime for 5 days. Those lipids with acute dysregulation were subsequently maintained a relatively stable level after cefepime treatment for 7 or 10 days (Fig. 4c). Lysophosphatidylcholines (LysoPC) was declined at the beginning of cefepime treatment but recovered to normal level from the 5th day. Different from PC and PE, the pro le of several other glycerophospholipids subclasses phosphatidylglycerol (PG), phosphatidylinositol (PI) and diacylglycerol (DG) were declined in response to cefepime. Moreover, the lipids of phosphatidylserine (PS), a subclass of GP, were nonuniformly affected in response to cefepime (Fig. 4c). Similar to the GP, most of the lipids belonging to the FA, GL, ST, PR and SP groups were remarkably altered after continuous cefepime treatment for 5 days (Fig. 4d). These data further con rmed that cefepime signi cantly disturbed the lipid pro le after 5 days' treatment. Taken together, cefepime markedly alters lipid pro le of corpus striatum, particular PC and PE, which contributes to a large proportion of aberrant lipidomic pro le.

Cefepime remarkably modi es the composition and structure of glycerophospholipid in corpus striatum
The effect of cefepime on the composition and structure of glycerophospholipid was studied. The results showed the overall abundance of GP subclasses PE, PC, PS, PG and PI were gradually increased with the extension of cefepime treatment, and maintained relatively stable level after 5 days of cefepime treatment (Fig. 5a). Individually, except for 10 days of cefepime treatment, the gently upregulated abundance of PE, PG and PI showed no signi cance among the control and cefepime-treated groups. Different from these subclasses, cefepime dramatically drove up the abundance of PC and PS after cefepime treatment for 5 days (Fig. 5a).
The major and minor species of PE, PC, PS, PG and PI were separately analyzed. Although the total abundance of PE only remarkably changed after cefepime treatment for 10 days, the composition of three major PE species (20:5/P-18:1, 20:5/20:0 and 20:4/18:0) were remarkably up-regulated even after cefepime treatment for one day (Fig. 5b). In addition, the composition of another one major PE species (22:6/16:0) and three minor (14:1/24:0, 20:5/18:3 and O-14:0/18:0) PE species were dramatically increased from the beginning of cefepime treatment for 5 days (Fig. 5b and Fig. S1B). Compared to the other GP subclasses, almost all identi ed major PC species were remarkably up-regulated in response to cefepime treatment on day 5. As shown in the Fig. 5b, six of seventh major PC species were altered by cefepime. Moreover, all identi ed minor PC species and PS species were also increased by cefepime starting at day 5 (Fig. S1B). Moreover, the composition of two major (20:3/21:0, 22:6/18:0) PS species were also signi cantly up-regulated by cefepime starting from the 5th day of exposure (Fig. S1A). Although the major species of PG was less affected by cefepime, the minor species of PG (20:0/20:3) was greatly increased after cefepime treatment for 5 days (Fig. S1B). Collectively, cefepime administration largely affected the composition of PC in corpus striatum.
The length and unsaturation degree of acyl chain are two key factors affecting the lipid geometry and uidity of lipid bilayer, and the distribution of acyl chains differs from disease mutations or depletion of metabolic enzymes (Andresen et al. 1999;Nochi et al. 2017). We then analyzed the pool of fatty acyl chains related to GP. Cefepime treatment uniformly altered the composition of GP, especially from the 5th day of cefepime treatment (Fig. 5c). In addition, the intensity of long fatty acyl chain, such as C32, C34, C36, C37, C38 and C40, were signi cantly higher in cefepime-treated mice those in control mice (Fig. 5d). Except for the length of acyl chain, the unsaturation degree of acyl chain was also profoundly affected by cefepime (Fig. 5e). Collectively, cefepime signi cantly remodeled GP pro le, especially PC species, accounting for the majority of those modi ed lipids in response to cefepime treatment.

Cefepime causes aberrant neuronal morphology and function in vitro
To examine the effect of cefepime exposure on lipid disorder of neurons, the embryonic primary neuron isolated from mouse striatum was cultured and treated with cefepime in vitro (DIV). We observed that 50 µM or 500 µM cefepime exposure for 24 h caused a signi cant increase of fragmented neurotic structure ( Fig. 6a). High concentration of cefepime caused more neurite fragmentation; however, the integrity of soma was unaffected and no obvious alteration was noted in cefepime-exposed neurons (Fig. 6a), suggesting cefepime may mainly affect the dendrite of striatal neuron.
Studies have shown that abnormally high or low ratio of PC/PE is able to in uence energy metabolism, linking to disease progression (Green et al. 2009; Ledesma et al. 2012). In forementioned results, we quanti ed the abundance of all PC and PE species, and found that the ratio of PC/PE abundance was signi cantly increased in vivo from the 7th day of cefepime exposure as compared with control groups (Fig. 6b). To further determine whether the increased PC/PE ratio in cefepime-treated brain linked to aberrant energetic system or those abnormal neurites observed in vitro, the morphology of the mitochondria which is the predominate regulator of bioenergetic metabolism was subsequently assessed (Friedman and Nunnari, 2014;Liesa and Shirihai, 2013;Rolfe and Brown, 1997). Mitochondrial morphology of primary striatal neuron was visualized by co-immunostaining with the translocase of mitochondrial outer membrane 20 (TOM20). The results showed that mitochondrial morphology of neuron was obviously altered after cefepime treatment (50 µM and 500 µM) for 24h. In contrast to tubular mitochondrial structure in normal primary neurons, cefepime exposure resulted in swelled and fragmented mitochondria (Fig. 6c). We quanti ed the area parameter of individual mitochondria using Image J, and found that the mitochondrial size was markedly decreased by cefepime (500 µM). In spite that the mitochondrial size showed relatively normal under 50 µM cefepime treatment, the overall mitochondrial size was clearly reduced (Fig. 6d). In addition, we further analyzed the average mitochondrial intensity in the selected neuritic area, which was rstly drawn out based on the intensity of the uorescently-labeled phalloidin, a probe of actin cytoskeleton. Mitochondrial intensity per neuritic area was signi cantly decreased in cefepime-exposed neurons (Fig. 6e). Taken  We found that the density of synaptic vesicle was obviously reduced by cefepime treatment (Fig. 6g and 6f), implying that cefepime treatment impaired the formation of synaptic vesicles in neurites. Taken together, these results indicated that cefepime treatment may interfere with neurotic morphology, axonal mitochondria and synaptic vesicle, which may eventually lead to the de ciency of neuronal function.

Discussion
The mechanism for the CNS adverse effect of cefepime has not been fully elucidated. Using lipidomic analysis, we aimed to investigate whether the disorder of lipids hemostasis participates such adverse effect. We observed that 5-day continuous intraperitoneal injection of cefepime dysregulated the hemostasis of glycerophospholipid pro le in the corpus striatum of mice. Cefepime not only dysregulated the abundance of GP but also altered biophysical properties of neuronal membrane through modifying the length and unsaturation of acyl chain. Morphological analysis of primary striatal neuron revealed that cefepime-induced GP disorder may contribute to the altered neuritic integrity and mitochondrial dynamic . Monotherapy with cefepime in clinic is usually administered intravenously 2 ~ 4 g/day (33 ~ 66 mg/kg/day for human dose), and the trough level of cefepime arrives to 11.5 ~ 44.6 mg/L. Previous studies showed that the trough level of cefepime above 20 mg/L has a 50% probability of developing neurological symptom (Lamoth et al. 2010). In this study, to avoid general toxicity of mice which may interfere with the lipidomic pro ling of brain, mice were injected daily with 80 mg/kg cefepime, which is half of equivalent dose in human based on body surface area (Nair and Jacob, 2016). Similar to patients in clinic, the anxiety-like behavior as well as the attenuated locomotor activity were observed in the mice treated with cefepime for 5 continuous days. Both serum chemistry and histopathology of the major organs showed no signi cant alteration.
The pathophysiology of cefepime neurotoxicity is thought to across BBB and modulate GABA receptor.
As the brain is the second most lipid-rich organ, we hypothesized that dysregulated lipid metabolism may participate in the initial or developing progress of cefepime neurotoxicity. To this end, through UPLC-MS/MS based lipidomic analysis, we found a widely alteration of lipid pro le in all three interconnected brain areas. In particular, the number of dysregulated lipids were remarkedly increased after 5-days of cefepime exposure, supporting the ndings in the neurobehaviors of cefepime-treated mice.
It is well known that homeostasis of lipids as well as fatty acids are the crucial determinants of neural

Cefepime predominately dysregulates glycerophospholipid pro le of corpus striatum
Through combined OPLS-DA model and statistical analysis, we discovered that the ratio of dysregulated ion peaks of corpus striatum was higher than that in hippocampus or prefrontal cortex, indicating that dysregulated lipids of corpus striatum may contribute to cefepime-induced CNS disorders. Cefepime is able to competes with γ-aminobutyric acid, thus weakening the function of GABAergic neuron and eventually leading to CNS disorder (Lam and Gomolin, 2006). In addition, the proportion and diversity of GABAergic neuron within the striatum are higher than that in the cortex or hippocampus (Huang and Paul, 2019; Murata and Colonnese, 2020; Shi et al. 2021). Thinking above, we speculate that GABAergic neuron-enriched striatum may be more sensitive to cefepime treatment than the other two brain areas. Indeed, the data from lipidomic pro le revealed that corpus striatum was the brain region that was mostly affected by cefepime.
Our results showed that GP accounted for the majority of cefepime-dysregulated lipids. GP mainly act as amphipathic molecules for forming the backbone of biological membrane as well as a reservoir for the Consistently, the mitochondrial structure and mitochondrial density in the cultured striatal neuron was profoundly reduced after cefepime treatment. We infer that the aberrant mitochondrial morphology and reduced mitochondrial content in neurite may participate neurite degeneration. Interestingly, we also observed a decreased density of synaptic vesicle in the neurite. Due to the essential role of synaptic vesicle in neurotransmission (Brunger et al. 2018;Südhof, 2013), the reduced synaptic vesicles may somehow represent the reduction of neuronal activity.
In summary, cefepime profoundly changes the GP pro le in the corpus striatum of mice. Cefepime alters the membrane structure of neurites, mitochondria and synaptic vesicle in the cultured striatal neurons.
Our study reveals that dysregulation of GP homeostasis may contribute to cefepime-induced CNS adverse effect.

Declarations
Ethical approval All applicable international, national and/or institutional guidelines for the care and use of animals were followed. All experimental procedures and use of the animals were conducted in accordance with the guidelines established by the Association for Assessment and Accreditation of Laboratory Animal Care and the Institutional Animal Care and Use Committee of Sichuan University. All efforts were made to minimize the suffering of the mice.

Con ict of Interest
The authors declare that they have no con icts of interest.    Color bar represents the value of log2-fold changed level of each lipid.