Metabolomic Characterization of Acute Ischemic Stroke Facilitates Metabolomic Biomarker Discovery

Acute ischemic stroke (AIS) is characterized by a sudden blockage of one of the main arteries supplying blood to the brain, leading to insufficient oxygen and nutrients for brain cells to function properly. Unfortunately, metabolic alterations in the biofluids with AIS are still not well understood. In this study, we performed high-throughput target metabolic analysis on 44 serum samples, including 22 from AIS patients and 22 from healthy controls. Multiple-reaction monitoring analysis of 180 common metabolites revealed a total of 29 metabolites that changed significantly (VIP > 1, p < 0.05). Multivariate statistical analysis unraveled a striking separation between AIS patients and healthy controls. Comparing the AIS group with the control group, the contents of argininosuccinic acid, beta-D-glucosamine, glycerophosphocholine, L-abrine, and L-pipecolic acid were remarkably downregulated in AIS patients. Twenty-nine out of 112 detected metabolites enriched in disturbed metabolic pathways, including aminoacyl-tRNA biosynthesis, glycerophospholipid metabolism, lysine degradation, phenylalanine, tyrosine, and tryptophan biosynthesis metabolic pathways. Collectively, these results will provide a sensitive, feasible diagnostic prospect for AIS patients.


Introduction
Stroke, a rapid process of focal neurological deficits caused by a disruption of blood supply to the corresponding region of the brain, is the second-leading cause of death worldwide and a major contributor to the global disease burden [1,2]. Strokes are classified as either ischemic or hemorrhagic. Ischemic strokes account for 87% of all stroke-related incidents, caused by interruption of the blood supply to brain tissue, and result in infarction, thromboembolic strokes, or atherosclerotic strokes. Hemorrhagic strokes are defined as nontrauma-induced subarachnoid or intracerebral hemorrhages [3,4]. These two-stroke types showed an overlapped trajectory of functional recovery, with age and initial stroke severity as the main prognostic factors. Besides, it is now widely recognized that targeted interventions including reducing blood pressure and smoking and promoting physical activity and a healthy diet, could markedly reduce the burden of stroke [5,6].
AIS is a medical emergency and a major public health priority, occurred by an abrupt and sustained reduction in regional cerebral blood flow, and consequently led to acute loss of neurons, astroglia, and oligodendroglia as well as synaptic damage [7]. Due to the blood vessel blockage by a thrombus or emboli, oxygen, and glucose supply are limited to the affected brain region, clinical symptoms including aphasia and hemiparesis occurred [8]. The cornerstone of effective stroke care continues to be timely reperfusion treatment. Whereas, AIS treatment is racing against time-the therapeutic window for intravenous thrombolysis within 4.5 h or for mechanical thrombectomy within 8 h after symptom onset [7,9,10].
Patients with AIS now can be triaged for appropriate management with improved imaging techniques beyond a simple computed tomographic (CT) scan. Generally, CT, CT angiography (CTA), magnetic resonance imaging (MRI), and MR angiography (MRA) are used as an auxiliary for AIS; moreover, the advent of advanced imaging modalities, including multimodal CT and MR perfusion is helpful to delineate areas of infarcted tissue, areas of potentially salvageable tissue given timely reperfusion, and areas of benign oligemia [11][12][13]. Many stroke centers have been established nowadays in China to eliminate delays in neuroimaging. Collectively, speed of treatment is a critical factor in determining the outcome for patients with disabling AIS, and exploring early diagnostic biomarkers will save the necrotic tissue in the brain.
Because the pathogenesis of AIS is heterogeneous and complicated, it is crucial to establish high-throughput technologies such as metabolomics to facilitate the identification of novel diagnostic biomarkers and potential therapeutic targets for AIS. Metabolomics, which is the profiling of metabolites in biofluids, cells, and tissues, combines comprehensive analysis techniques with bioinformatics and detects subtle alterations of metabolites underline various physiological and pathological conditions [14][15][16]. Detailly, metabolites are typically recognized as small molecules, which are involved in cellular reactions, provide a functional signature of phenotype, and reveal biological relevant perturbations resulting from disease exposure. Metabolomics studies in the cohort of stroke patients primarily focused on biomarker discovery in biofluids [17][18][19], whereas studies in animals were largely designed to decipher molecular pathways and targets altered in brain tissue after stroke [20][21][22].
Metabolomics is broadly acknowledged to be the omics discipline that is closest to the phenotype. The metabolome reveals the products of the genome and its protein output, as well as metabolites from diet, drugs, and toxic compounds [23,24]. Although the nature of metabolites, particularly their diversity in both chemical structure and dynamic range of abundance, remains a major challenge in high-throughput abundance profiling in biological samples, enormous advances have been made with regard to the number of analytes about which accurate quantitative information and important biological information can be acquired [25,26]. In this study, we applied multiple reaction monitoring (MRM)-based targeted metabolomics approach incorporating liquid chromatography with mass spectrometry detection (LC-MRM/MS) to explore novel biomarkers and related pathways underlying AIS progression.
Currently, the MRM experiment can be used to selectively detect and quantify metabolites based on the screening of thousands of transitions (pairs of precursor and product m/z values) on a triple quadrupole (QqQ) MS instrument coupled to an LC, by welldesigned scheduling and selection of m/z windows [27,28]. Due to its high sensitivity, broad dynamic range, and good reproducibility, the MRM approach is promising to provide insights into the course of diverse diseases, discover novel biomarkers, and shed light on the impact of drug metabolism in vivo. LC-MRM/MS-based targeted metabolomics has been applied in newborn mass screening and antenatal diagnosis [29], metabolic disorders like serum 25-hydroxyvitamin D status [30], metabolic reprogramming in cancer metastasis [31], cardiovascular disease risk assessment [32], development of insulin resistance in obesity [33], and metabolic perturbations caused by drug addiction [34].
LC-MS/MS with MRM mode serves as the foundation for accurate simultaneous metabolites quantitation across a clinical cohort of sample sets to provide high-quality information on target molecular profiles. AIS metabolite panels were established here via combining standardized methods for extracting metabolites from clinical samples and highreproducible MRM-based targeted metabolomics technology. In summary, we develop a quantitative metabolomics workflow with an antibody-independent method to monitor AIS progression and assess stroke occurrence risk.

Patients and Clinical Specimens
The study was conducted in accordance with the guidelines of the Declaration of Helsinki. AIS patients and healthy individuals were recruited from a single center, Xiamen Branch, Zhongshan Hospital of Fudan University, between March 2018 and February 2020. This study obtained the approval of the Research Ethics Committee from this hospital of Fudan University.
Inclusion criteria were as follows: an initial National Institutes of Health Stroke Scale (NIHSS) score from 6 to 22, aged 46 to 75 years, presentation < 24 h after stroke onset, and stroke localization in the area of the middle cerebral artery. Patients with diabetes, cardiovascular diseases, or other diseases that would affect the metabolic profiles were excluded from the study. Healthy donors with a stroke history or showing any sign of stroke based on CT or MRI evaluation were excluded from the control group. Metabolomics analyses began with an unbiased search for serum metabolites. Cases were randomly selected from 22 AIS patients, while an age-and gender-matched control group was randomly selected from 22 healthy individuals. Detailed information on these subjects is summarized in Supplementary Table S1. Written informed consent was provided by all participants. In order to ensure sample quality and result interpretation, the serum sample collection process was controlled carefully, including fasting patients with AIS and healthy volunteers overnight and preventing freeze-thaw cycles before analysis. Immediately after collection via vein blood sampling, blood samples were centrifuged at 3000 rpm 4 °C for 15 min, and the supernatants were collected and stored at − 80 °C for further analysis.

Metabolite Extraction for MRM Analysis
MS grade methanol (MeOH), acetonitrile (ACN), ammonium acetate (CH 3 COONH 4 ), and ammonium hydroxide (NH 4 OH) were purchased from ANPEL Laboratory Technologies Inc (Shanghai, China). Total metabolites were extracted from 100 μL serum using a MeOH:ACN:H 2 O (2:2:1, v/v) solvent mixture. To precipitate protein, serum was incubated at − 20 °C for 1 h, then centrifuged at 13,000 rpm at 4 °C for 15 min. The supernatant was removed and evaporated to dryness at a gentle nitrogen flow. The dry extracts were reconstituted in a 100-μL solvent mixture of ACN:H 2 O (1:1, v/v), vortexed for 30 s and sonicated for 10 min, then centrifuged at 12,000 rpm 4 °C for 15 min to remove the insoluble debris. Finally, 60 μL supernatant was transferred into a new LC-MS glass vial for the UHPLC-QQQ-MS analysis, and a pool of 15 μL from each sample and mixed as quality control (QC) samples.
To monitor the data quality and process variation, QC samples containing aliquots from serum samples of all participating subjects were parallel processed. Additionally, the orders of sample injection were randomized to avoid systematic biases.
For MRM-based targeted metabolomic analyses, 180 MRM transitions representing the 180 metabolites were simultaneously monitored, and the positive/negative polarity switching was applied. The dwell time for each MRM transition is 3 ms, and the total cycle time is 1.26 s. MRM transition parameters, including ionization polarity, precursor ion, product ion, collision energy, and fragmentor voltage, were optimized in-house. The retention time of each metabolite was determined by measuring the corresponding MRM transition individually. The metabolite list consisted of 180 metabolites with metabolite name, formula, HMDB ID, KEGG ID, METLINE ID, MRM transition parameters, and expected retention time (Supplementary Table S2).

Data Processing
A total of 180 metabolites were analyzed accurately and quantitatively via LC-MS/MSbased MRM approach. The raw data were assessed for peak detection and alignment using Profiling Solution software. Multivariate analysis, including principal component analysis (PCA) and orthogonal projections to latent structures-discriminant analysis (OPLS-DA), was used to visualize general clustering, trends and outliers of LC-MS/MS data via the software SIMCA-P + 14.1 (Umetrics, Sweden). The variable importance in projection (VIP) method was applied to select the most relevant variables for the interpretation of AIS changes. A volcano plot was used to filter important features that exhibited large variable significance in VIP > 1 and statistically adjusted p < 0.05 between AIS and control groups.
Potential biomarkers were assessed by receiver operating characteristic (ROC) analysis. The area under the ROC curve (AUC) of the proposed metabolite panel was used as a metric to evaluate the sensitivity and specificity of the biomarker performance. A list of these differential metabolites was imported into MetaboAnalyst 5.0 (http:// www. metab oanal yst. ca/) for pathway enrichment analysis [35].

Quantification of Metabolites in Serum Samples by LC-MRM/MS Approach from AIS Patients and Healthy Controls
A general outline of the experimental procedure was shown (Fig. 1). Fasting serum samples were collected from 22 patients with AIS and 22 healthy controls, and a total of 180 metabolites were analyzed via LC-MS/MS-based MRM approach. A diagnostic model for identifying AIS patients was then built up based on the ROC analysis.
To test the reproducibility of the sample preparation procedures and to assess the reliability of the LC-MS system, the QC samples were prepared by mixing aliquots of all the biological samples and analyzed between every six clinical samples. PCA analysis was performed to get an overview of the difference in metabolites profiling between both sample dataset and QC injections (Supplementary Fig. S1). The first principal component (PC1) accounted for 53.8% of the total variance. Seven QC samples were clustered tightly in the PCA score plot, which indicated that the precision and repeatability of the experiments were excellent.
A total of 112 metabolites were acquired from serum samples including 22 AIS patients and 22 healthy donors. Systematic metabolomic changes occurring in different groups were then assessed by two widely used multivariate methods-PCA and OPLS-DA. The PC1 accounted for 54.5% of the total variance and separates the AIS group from the control group (Fig. 2A). To achieve the maximum distinction and identify differential metabolites that accounted for the separation between groups, OPLS-DA analysis was further conducted (Fig. 2B). Samples were within a 95% confidence interval at Hotelling's t-test. The permutation test for OPLS-DA showed that the Q2 regression line had a negative intercept and all R2γ and Q2 values on the left were lower than the original points on the right (Supplementary Fig. S2), which demonstrated that the OPLS-DA model in the present study was valid.
Relative quantifications were applied to all the identified metabolites in these two groups, and VIP was applied to determine the most relevant differential metabolites. Totally, 29 significantly changed metabolites were presented in Supplementary Table S3. Significant differences in the variables between the AIS group and the control group were depicted as volcano plots, including 13 upregulated metabolites and 16 downregulated ones (Fig. 2C). Collectively, our metabolomic data unraveled a strikingly consistent separation between AIS patients and healthy donors.

Diagnosis of Patients with AIS Based on ROC Analysis and Abnormal Metabolism Pathways
The metabolite panel was developed based on a logistic regression model. ROC analyses were performed for 29 significantly changed metabolites. With the criterion of ROC area > 0.8, 5 metabolites were screened as potential biomarkers for AIS diagnosis, including argininosuccinic acid (AUC = 0.897), beta-D-glucosamine (AUC = 0.909), glycerophosphocholine (AUC = 0.816), L-abrine (AUC = 0.841), and L-pipecolic acid (AUC = 0.804) (Fig. 3A). ROC curve analysis further showed that these 5 metabolites Fig. 3 Diagnostic panel of AIS based on ROC analysis. A ROC curve analysis for argininosuccinic acid, beta-D-glucosamine, glycerophosphocholine, L-abrine, and L-pipecolic acid. B ROC curve analysis showed a potential biomarker panel combined these 5 metabolites for the diagnosis of AIS. C The bar charts presented intensities of these 5 metabolites in the AIS group and control group formed a potential diagnostic panel, presenting a good diagnostic accuracy for AIS (AUC = 0.919, Fig. 3B). Surprisingly, all these 5 metabolites including argininosuccinic acid, beta-D-glucosamine, glycerophosphocholine, L-abrine, and L-pipecolic acid were downregulated in AIS patients compared with healthy controls (Fig. 3C).
The KEGG pathway enrichment analysis was performed using 29 significantly changed metabolites compared to AIS patients with healthy controls. 29 out of 112 detected metabolites, enriched in four metabolic pathways, were found significantly affected during AIS progression, including aminoacyl-tRNA biosynthesis, glycerophospholipid metabolism, lysine degradation, as well as phenylalanine, tyrosine, and tryptophan biosynthesis (Fig. 4A). The color of each bubble reflected the significance of metabolism pathways (red indicated low p-values), while the size of the bubble indicated the number of differential metabolites. The detailed information of the top 8 enriched pathways were presented in Fig. 4B. Glycerophospholipid metabolism and inositol phosphate metabolism were significantly affected in the AIS metabolome via integrating pathway enrichment analysis and pathway topology analysis (Fig. 4C-D), which indicated lipid metabolism was remarkably disturbed in patients with AIS.

Discussion
Stroke, especially AIS, is a high mortality disease caused by blood vessel blockage. In this study, argininosuccinic acid, beta-D-glucosamine, glycerophosphocholine, L-abrine, and L-pipecolic acid were downregulated in AIS patients compared with healthy volunteers. Through metabolomic analysis, we found that Aminoacyl-tRNA biosynthesis, glycerophospholipid metabolism, and lysine degradation, phenylalanine, tyrosine, and tryptophan biosynthesis were associated with AIS.
Membranes of eukaryotic cells are composed principally of five distinct classes of phospholipids, which maintain the stability and fluidity of cell membrane, facilitate cellular regulation, and enable cell-to-cell communication [36]. Notably, phosphatidylcholine (PtdCho) is generally the most abundant phospholipid class in a membrane. Phospholipase B deacylates PtdCho, producing glycerophosphocholine (GroPCho) and two free fatty acids [37]. Inhibition of the activity of phospholipase B chemically or genetically resulted in slow neurodegeneration in mice and Drosophila [38,39]. Fernández-Murray JP and McMaster CR presented metabolic evidence indicating that intracellular GroPCho was further metabolized and that the Cho moiety was reused for PtdCho biosynthesis [37]. Accordingly, the alteration of GroPCho metabolism could be the reason for the cerebral damage from AIS.
The non-proteinogenic amino acid L-pipecolic acid (L-PA) is a precursor of immunosuppressants, peptide antibiotics, and a metabolic intermediate of L-lysine [40,41]. L-lysine is an essential charged amino acid transported into the central nervous system. The lysine catabolism pathway differs in the adult mammalian brain from that in extracerebral tissues. The saccharopine pathway is the predominant lysine degradative pathway in extracerebral tissues, whereas the pipecolic acid pathway predominates in the adult brain [42,43]. Our metabolic evidence showed downregulated level of L-pipecolic acid in the AIS group and perturbed lysine degradation, which caused brain dysfunction in patients with AIS.
Other abnormal metabolic pathways, including the phenylalanine pathway as well as tyrosine and tryptophan biosynthesis pathway, were related to the downregulation of L-abrine [44]. L-abrine is a small molecule found in the seed of Abrus precatorius. L-abrine also belongs to prenylated indole alkaloids, which represent a group of natural products with diverse chemical structures and are widely distributed [45]. It had been reported that various endogenous indoles might provide an antioxidant defense in the brain and participate in free radical scavenging [46]. In our study, the level of L-abrine was downregulated in patients with AIS. As a component of indoles, L-abrine deficiency might cause the accumulation of radicals. Consequently, the overproduction of radicals might trigger necrosis or apoptosis in the brain following AIS. Fig. 4 The KEGG pathway enrichment analysis of differential metabolites in clinical specimens compared AIS patients with healthy controls. A Metabolic pathways organized by pathway enrichment analysis (p-value) and pathway topology analysis (pathway impact). The tendency of red circles in the bubble plot presented the importance of metabolism pathways. B Top 8 enriched pathways were summarized, including p-value and pathway impact information. C-D Glycerophospholipid metabolism and inositol phosphate metabolism were enriched in the AIS metabolome. Compound colors within the pathway-light blue means those metabolites are not in the targeted metabolomic data of this study and are used as background for enrichment analysis, while red means the metabolites are in this dataset with different levels of significance In conclusion, this study developed a high-throughput metabolomic assay and manifested its utility for metabolite profiling of large-scale biological samples. By using a quantitative MS with an antibody-independent approach, a flux of 180 metabolites was established via the MRM detection method. Among which, 112 metabolites were successfully identified in serum samples, indicating that many serum metabolites associated with AIS are affected, including 29 metabolites changed markedly and 5 metabolites are downregulated. Finally, these metabolites are found to be mainly enriched in 4 metabolic pathways. Our work represented the significant distinct metabolites in plasma samples from Ischemic, demonstrating that metabolites profiling could possibly provide a sensitive, feasible diagnostic prospect for Ischemic patients.

Author Contribution
All authors contributed to the study's conception and design. The methodology was performed by Biao Qi and Yanyu Zhang. Data analysis and investigation were performed by Bing Xu and Guoqiang Fei. The original manuscript was written by Biao Qi, Yanyu Zhang, and Bing Xu, while the manuscript review and editing was done by Ling Lin and Qiuping Li. Funding acquisition was gained by Biao Qi, Guoqiang Fei, and Qiuping Li. Resources were prepared by Yuhao Zhang. All authors have read and approved the final manuscript. Data Availability All data generated or analyzed during this study are available from the corresponding authors on reasonable request.

Declarations
Ethics Approval This article contains a metabolomic study with human subjects. Ethical approval from the Research Ethics Committee from Xiamen Branch, Zhongshan Hospital of Fudan University was obtained.
Consent to Participate Patient written informed consent was obtained.

Consent for Publication Not applicable.
Competing Interests The authors declare no competing interests.