The Untargeted Metabolomics Reveals Differences in Energy Metabolism in Patients with Different Subtypes of Ischemic Stroke

doi:10.21203/rs.3.rs-2851009/v1

Download PDF

Research Article

The Untargeted Metabolomics Reveals Differences in Energy Metabolism in Patients with Different Subtypes of Ischemic Stroke

https://doi.org/10.21203/rs.3.rs-2851009/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 06 Jan, 2024

Read the published version in Molecular Neurobiology →

Version 1

posted

You are reading this latest preprint version

Objective

Stroke has the characteristics of high morbidity, high mortality, high disability rate, and high recurrence rate, which brings great burden of disease. Ischemic stroke (IS) is the most common subtype of stroke, accounting for 71% of all stroke types. The risk factors and pathogenesis of IS are complex and varied due to different IS subtypes. The development of metabolomics technology provides new ideas for the study of the biomarkers and potential pathophysiological mechanisms of IS.

Methods

We included IS patients who visited Xiangya Hospital of Central South University from October 2017 to September 2018 and divided them into two groups based on the TOAST classification: large-artery atherosclerosis (LAA) subtype group (n = 87) and small-vessel occlusion (SVO) subtype group (n = 39). Plasma metabolomics analysis was performed using liquid chromatography-high-resolution mass spectrometry (LC-HRMS) to identify metabolic profiles in LAA and SVO subtype IS patients and to determine metabolic differences between patients with the two subtypes of IS.

Results

We identified 26 differential metabolites and 3 differential metabolic pathways between LAA and SVO subtype IS. A multiple prediction model based on the differential metabolites had good predictive ability for IS subtyping (AUC = 0.822, accuracy = 77.8%), with 12,13-DHOME being the most important differential metabolite in the model. The differential metabolic pathways between the two subtypes IS patients were related to the tricarboxylic acid cycle, alanine, aspartate and glutamate metabolism, and pyruvate metabolism, mainly focused on energy metabolism.

Conclusion

The energy metabolism level of SVO subtype IS patients is more active than that of LAA subtype IS patients.

stroke

metabolomics

large-artery atherosclerosis

small-vessel occlusion

13-DHOME

Stroke is a global public health problem with high morbidity, high disability, high mortality and high recurrence rate. The global burden of disease study showed that the overall lifetime risk of stroke in the Chinese population ranked first, which was 39.9%. From 1990 to 2016, the number of stroke deaths in China increased from 396,000 to 729,000, with a relative increase of 83.8%(1). In 2017, China has the largest number of stroke patients in the world(2). Ischemic stroke (IS) is the most common subtype of stroke, accounting for 71% of all strokes (3). In 2019, stroke ranked the third among the causes of death from cancer and cardiovascular diseases in China (4), among which patients with IS accounted for 81.9% of hospitalized patients with stroke. In recent years, with the extension of life expectancy and the acceleration of population aging, the prevalence rate of IS continues to rise (5, 6). Therefore, the establishment of effective prevention and treatment strategy IS an important problem faced by the society.

The TOAST classification method is the most widely used etiological classification method for IS internationally at present (7), which divided IS into 5 subtypes: large-artery atherosclerosis (LAA), cardio embolism (CE), small-vessel occlusion (SVO), IS of other determined etiology (SOE) and IS of undetermined etiology (SUE). TOAST classification of IS patients is conducive to early treatment of the disease and can reduce the risk of recurrence in patients (8). Patients with different subtypes of IS have different prognosis and outcomes (9). Previous studies have found that patients with LAA subtype IS have poor prognosis and significantly higher mortality than those with SVO subtype IS(10). The distribution proportion of patients with different subtypes of IS varies in different races, countries and regions. A number of previous retrospective analyses showed that the proportion of CE subtype IS patients in the United States and several European countries was higher than that in other regions, while the incidence of LAA subtype IS was the highest in Asia (11–16).

Most human diseases are associated with changes in metabolic processes. By capturing metabolites in blood or urine and conducting systematic biology analysis, disease-specific metabolic profiles can be identified (17). Metabolomics, which involves qualitative and quantitative analysis of all small molecule metabolites in the human body, can be used to study changes in metabolic profiles associated with different diseases or different stages of the same disease, in order to find relationships between metabolites and physiological or pathological changes. In the field of cerebrovascular diseases, several metabolic pathways related to IS have been identified. In 2011, Jung et al. used untargeted metabolomics based on nuclear magnetic resonance (NMR) spectroscopy to demonstrate that the one-carbon metabolism pathway and its metabolite homocysteine were closely related to IS (18). Subsequently, untargeted metabolomics techniques have confirmed the relationship between several metabolic pathways, such as amino acid metabolism, brain energy metabolism, and lipid metabolism, and IS (19). In 2021, a cohort study in Taiwan used ultra high-performance liquid chromatography with tandem mass spectrometry (UHPLC-MS/MS) to detect 220 serum metabolites and found that serum metabolite profiles of patients with IS within 7 days of acute onset could predict modified Rankin Scale (mRS) scores 3 months later(20). In these metabolomics studies, the detection limit of NMR technology ( > ~ 1 nmol) is higher than that of mass spectrometry, while liquid chromatography-mass spectrometry (LC-MS/MS) technology and its derivative UHPLC-MS/MS technology are not ideal for detecting unknown metabolites (21, 22). In recent years, emerging liquid chromatography-high-resolution mass spectrometry (LC-HRMS) technology has higher resolution and mass detection accuracy, not only capable of detecting known target metabolites, but also significantly improving the detection efficiency of unknown metabolites, and is currently one of the most ideal non-targeted metabolomics research methods (22–24).

The pathogenesis of IS is complex, and different subtypes of IS have distinct pathological and physiological mechanisms. Biomarkers can provide important evidence for the classification, early diagnosis, disease assessment, and prognostic evaluation of IS. Due to the presence of the blood-brain barrier (BBB), large molecules in cerebrospinal fluid are difficult to enter the peripheral blood, and so far, there is no definitive biomarker that can guide the classification and disease assessment of IS in clinical practice(25). Metabolomics technology can identify small molecule metabolic biomarkers that cross the BBB and has broad application prospects. In this study, we used LC-HRMS technology to perform untargeted metabolomics analysis on LAA and SVO subtypes of IS patients, identified the differential metabolic markers and metabolic pathway characteristics of different subtypes of IS. Our study would provide a theoretical basis and new treatment targets for the rapid classification and clinical evaluation of IS.

2.1. Participants

Participants were recruited from the Department of Neurology at Xiangya Hospital of Central South University between October 2017 and September 2018. A total of 87 patients with LAA subtype IS and 39 patients with SVO subtype IS were included. Inclusion criteria were listed as follows: (1) admission within 7 days of onset, (2) age ≥ 18 years, (3) confirmed infarction assessed diffusion-weighted imaging (DWI). The exclusion criteria are as follows: (1) patients received vascular recanalization therapy, such as intravenous thrombolysis and mechanical thrombolysis; (2) pre-onset mRS Score > 2; (3) patients with history of stroke, autoimmune disease, malignant tumors, severe infection, severe hepatic and renal dysfunction; (4) patients without complete clinical examination and imaging data. The study was approved by the Ethics Committee of Xiangya Hospital, Central South University. All enrolled participants signed an informed consent.

2.2. Neuroimaging

Neuroimaging was performed in all enrolled IS patients, including at least one of the 1.5T head magnetic resonance imaging (MRI), 3.0T head MRI or head plain computed tomography (CT) scan and corresponding vascular examination. Imaging assessments were conducted independently by two experienced neurologists.

2.3. Clinical Examination

We collected all participants’ demographic and clinical data from the medical record. Systolic blood pressure (SBP), diastolic blood pressure (DBP), blood count, overnight fasting blood glucose (FBG), blood urea nitrogen (BUN), uric acid (UA), serum creatinine (Scr), estimate glomerular filtration rate (eGFR), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), homocysteine (Hcy), and glycated hemoglobin A1c (HbA1c) were detected in department of clinical laboratory, Xiangya Hospital. Lifestyle factors (smoking, drinking) and vascular risk factors (hypertension, type2 diabetes, dyslipidemia) were defined as described in our previous articles(26).

2.4. Sample Preparation

Following an overnight fast, 3 mL of blood samples were collected and centrifuged at 3,000 rpm for 15 min at 4◦C. The resulting plasma samples were stored at − 80◦C to prevent multiple freeze-thaw cycles until analysis. To extract metabolites, we used the following procedure: 100 µL of each plasma sample was added to a 1.5 mL centrifuge tube and treated with 400 µL of a mixture of acetonitrile and methanol (1:1, v/v) for deproteinization. The mixture was vortexed for 30s and sonicated for 10 min, followed by centrifugation at 12,000 rpm for 15 min at 4℃. The supernatant was then transferred to a new 1.5 mL centrifuge tube and dried with nitrogen. The sample was reconstituted with 100 µL of a mixture of acetonitrile and water (1:1, v/v) and vortexed for 30s, followed by sonication for 5 min and centrifugation at 12,000 rpm for 15 min at 4℃. The resulting supernatant was transferred into autosampler vials for liquid chromatography-high-resolution mass spectrometry (LC-HRMS) analysis.

2.5 LC-HRMS Analysis

LC-HRMS analysis was carried out using an AB SCIEX TripleTOF 5600 system (AB SCIEX, Foster City, CA, US). Positive and negative ion modes were employed to detect the metabolites, and the ACQUITY UHPLC T3 column (2.1 × 100 mm; 1.8 µm) was utilized for both modes at a temperature of 4◦C. The mobile phase A consisted of 0.1% formic acid aqueous solution, while mobile phase B contained acetonitrile with 0.1% formic acid. The flow rate was set at 300 µl/min, and the injection volume was 2 µl.

A gradient elution program was implemented, as follows: 5% B (0– 2 min), 5–70% B (2–5 min), 70–90% B (5–14 min), 90–100% B (14–16 min), 100% B (16–22 min), 100–5% B (22–22.1 min), and 5% B (22.1–25 min). The eluent was then introduced into the TripleTOF 5600 electrospray tandem mass spectrometer and analyzed using the Information Dependent Acquisition (IDA) mode. The mass spectrometer conditions were as follows: ion source gas1 (GS1), 55 psi; ion source gas2 (GS2), 55 psi; curtain gas (CUR), 35 psi; ion spray voltage floating (ISVF), 5500/-4500 (+/-) V; declustering potential (DP), 80 V; collision energy (CE), 40 ± 20 V; temperature (TEM), 550 ℃. Dynamic background subtraction (DBS) was applied to screen the profile to meet the IDA criteria and trace all potential minor metabolites.

2.6. Data Processing and Statistical Analysis

2.6.1. Untargeted Metabolomics Data Processing

The raw LC-HRMS data were transformed into mzML format using ProteoWizard software and subsequently analyzed with XCMSplus software to perform mass-to-charge ratio (m/z) matching, mass spectrometry peak de-noising, baseline calibration, peak signal picking and peak area calculation (27). The identified metabolites were statistically analyzed to construct a data matrix containing m/z, retention time (RT) and peak area. Differential metabolites were identified using the HRMS/MS library and the MetDNA database(http://metdna.zhulab.cn/) (28).

The web-based software MetaboAnalyst 5.0 (https://www.metaboanalyst.ca) was used for further metabolite data analysis. The metabolite intensity values were first log-transformed and then autoscaled (mean-centered and divided by the standard deviation of each variable). Orthogonal projections to latent structures-discriminant analysis (OPLS-DA) multivariate statistical methods were employed to detect group differences, and variable influence on projection (VIP) values were obtained for each variable. The differences of normalized peak intensities of metabolites between two groups were analyzed by Student’s t test, followed by false discovery rate (FDR) adjustment for multiple comparisons. Metabolites with VIP > 1 and P < 0.05 after FDR adjustment were considered as differential biomarkers. The metabolic pathway analysis was performed using the Pathway Analysis module of MetaboAnalyst 5.0 and the pathway library of Kyoto Encyclopedia of Genes and Genomes (KEGG) (https://www.kegg.jp). Multivariable ROC (Receiver operating characteristic curve) analysis was used to evaluate the ability of candidate metabolites in discriminating IS types, and the area under the ROC curves (AUC) was calculated (Biomarker Analysis module of the MetaboAnalyst 5.0). Monte-Carlo cross validation was used to generate multivariable ROC curves.

2.7 Statistical Analysis

Statistical analysis was conducted using SPSS 22.0 (IBM SPSS, Chicago, IL, USA) and MetaboAnalyst 5.0. Continuous variables were described as median (interquartile range, IQR: 25th-75th percentile). Mann-Whitney U test was used to compare continuous variables between two groups. Categorical variables were expressed as numbers (N) and percentages (%) and were evaluated using the chi-square test. Correlation analysis was performed using Pearson's correlation analysis for continuous variables and a two-tailed test was used for the statistical analysis. Results were considered statistically significant when P < 0.05 for clinical data or when P < 0.05 after FDR adjustment during the mass spectrometry data analysis. Graphs and charts were generated using GraphPad Prism 8.0. ROC curves were generated using MetaboAnalyst 5.0 for differential metabolites and multivariate prediction models, and the area under the curve (AUC) was calculated.

3.1 Demographic and Clinical Characteristics of the Study Population

A total of 126 patients with IS were included in this study, including 87 patients with LAA- subtype and 39 patients with SVO subtype. There were no significant differences between the two groups in terms of gender, age, medical history (hypertension, type 2 diabetes, coronary heart disease, dyslipidemia), smoking and drinking history, systolic and diastolic blood pressure, blood cell count, renal function, blood lipids, and homocysteine. However, the FBG and HbA1c levels of LAA subtype IS patients on admission were significantly lower than those of SVO subtype IS patients (FBG, LAA group: 5.9 mmol/L vs. SVO group: 6.9 mmol/L, p = 0.045; HbA1c, LAA group: 5.8% vs. SVO group: 6.6%, p = 0.017) (Table 1).

Table 1

Baseline characteristics of patients with LAA and SVO subtype IS included in untargeted metabolomics analysis
	LAA (n = 87)	SVO (n = 39)	p-value
Age, years	61 (40–82)	62 (38–85)	0.992
Gender, male, N (%)	63 (72.4%)	26 (66.7%)	0.514
Medical history
Hypertension, N (%)	62 (71.3%)	26 (66.7%)	0.605
Type 2 diabetes, N (%)	29 (33.3%)	11 (28.2%)	0.569
Dyslipidemia, N (%)	15 (17.2%)	6 (15.4%)	0.797
Coronary heart disease, N (%)	12 (13.8%)	5 (12.8%)	0.883
Smoking, N (%)	48 (55.2%)	19 (48.7%)	0.504
Drinking, N (%)	39 (44.8%)	12 (30.8%)	0.139
SBP, mmHg	143 (128–156)	142 (127–149)	0.238
DBP,mmHg	83 (73–90)	83 (75–91)	0.496
WBC, ×10⁹/L	6.2 (5.1–8.1)	6.9 (5.5–9.3)	0.079
Plt, ×10⁹/L	189.0 (149.0–221.0)	184.0 (132.0–223.0)	0.650
BUN, mmol/L	4.8 (3.8–6.0)	5.3 (4.5-6.0)	0.100
UA, µmol/L	332.3 (279.5–339.0)	344.3 (276.1–384.3)	0.796
Scr, µmol/L	81.0 (69.0–89.0)	78.7 (64.1–95.7)	0.763
eGFR, ml/min per1.73 m²	84.2 (70.1–95.2)	86.4 (65.2–96.8)	0.633
TG, mmol/L	1.6 (1.2–2.1)	1.42 (1.1–2.0)	0.298
TC, mmol/L	4.1 (3.2-5.0)	4.3 (3.7–5.3)	0.208
HDL-C, mmol/L	1.0 (0.9–1.2)	1.0 (0.8–1.2)	0.739
LDL-C, mmol/L	2.5 (1.9–3.2)	2.7 (2.3–3.4)	0.249
FBG, mmol/L	5.9 (5.0–7.8)	6.9 (5.3–9.2)	0.045
HbA1c, %	5.8 (5.5–6.5)	6.6 (5.7–7.6)	0.017
Hcy, µmol/L	13.4 (11.3–17.6)	13.9 (10.6–17.9)	0.768
Continuous variables were presented as median (interquartile spacing 25 th-75 th); Categorical variables were presented as quantity (N) and percentage (%). LAA, large-artery atherosclerosis; SVO: small-vessel occlusion; SBP: systolic blood pressure; DBP, diastolic blood pressure; WBC, white blood cell count; Plt: platelet count; BUN: blood urea nitrogen; UA: uric acid; Scr, serum creatinine; eGFR: estimate glomerular filtration rate; TG: triglyceride; TC: total cholesterol; HDL-C, high density lipoprotein- cholesterol; LDL-C: low density lipoprotein-cholesterol; FBG, fasting blood glucose; HbA1c, glycosylated hemoglobin; Hcy: homocysteine.

3.2 Quality control of untargeted metabolomics mass spectrometry data

After data preprocessing and standardization using MetaboAnalyst 5.0, the mass spectrum data presented a good normal distribution (Supplementary Fig. 1–2).

3.3 Plasma metabolomics multivariate statistical analysis of LAA and SVO subtype IS patients

3.3.1 Principal component analysis (PCA)

Unsupervised pattern recognition was performed using principal component analysis (PCA) on the metabolites that were preprocessed in the previous steps. The PCA score plot was shown in Fig. 1A. LAA subtype IS patients clustered well, while SVO subtype IS patients showed less distinct clustering, indicating higher heterogeneity among the SVO subtype IS patients included in this study.

3.3.2 Partial least squares-discriminant analysis (PLS-DA)

Supervised multivariate statistical analysis was performed using partial least squares-discriminant analysis (PLS-DA) to analyze the differences between the two groups. The R2Y and Q2 values were 0.6902 and 0.30749, respectively. The score plot and cross-validation plot were shown in Fig. 1B and Fig. 1D, respectively.

3.3.3 Orthogonal partial least squares-discriminant analysis (OPLS-DA)

An OPLS-DA model was established, with R2Y and Q2 values of 0.690 and 0.331, respectively. The score plot was shown in Fig. 1C, indicating that the current model can differentiate between LAA and SVO subtypes of IS, but with limited discrimination ability. Figure 1C showed the score plot, and Fig. 1E showed the permutation test plot.

3.4 Univariate statistical analysis of plasma metabolic profiles of LAA and SVO subtype IS patients

We performed univariate statistical analysis and identified 98 differential metabolites based on VIP values obtained from PLS-DA and p-values in t-test(VIP > 1, p < 0.05). After FDR correction, 26 of these metabolites remained significantly different between the two groups (VIP > 1, p < 0.05 after FDR correction) (Table 2).

Table 2

Untargeted metabolomic differential metabolites in patients with LAA and SVO subtype IS
Metabolites	KEGG	mz	RT	P value	FDR	VIP	FC	AUC
12,13-DHOME	C14829	313.24	761.91	< 0.01	< 0.01	2.90	2.32	0.830
9,10-DHOME	C14828	313.24	761.91	< 0.01	< 0.01	2.90	2.32	0.830
Dehydroepiandrosterone sulfate	C04555	289.22	469.46	< 0.01	< 0.01	2.19	0.42	0.773
Pyruvic acid	C00022	87.01	72.00	< 0.01	< 0.01	1.80	0.68	0.750
P1, P4-Bis(5'-guanosyl) tetraphosphate	C01261	869.05	496.07	< 0.01	< 0.01	1.81	1.83	0.757
L-Glutamine	C00064	145.06	50.90	< 0.01	< 0.01	1.54	0.82	0.715
(R)-3-((R)-3-Hydroxybutanoyloxy) butanoate	C04546	191.09	410.39	< 0.01	0.01	1.74	0.84	0.757
3-Hydroxybutyric acid	C01089	103.04	96.00	< 0.01	0.01	1.78	2.04	0.727
Feruloyl-CoA	C00406	966.16	496.15	< 0.01	0.01	1.70	1.71	0.752
L-Malic acid	C00149	133.01	72.00	< 0.01	0.01	1.66	0.71	0.715
Citric acid	C00158	191.02	72.00	< 0.01	0.01	1.73	0.78	0.720
Glutaryl-CoA	C00527	882.15	495.40	< 0.01	0.01	1.66	1.95	0.721
p-Aminobenzoate	C00568	138.06	60.00	< 0.01	0.01	1.88	0.49	0.716
Trigonelline	C01004	138.06	60.00	< 0.01	0.01	1.88	0.49	0.716
Fumaric acid	C00122	115.00	72.00	< 0.01	0.01	1.49	0.80	0.691
3',5'-Cyclic AMP	C00575	330.06	371.37	< 0.01	0.01	1.34	0.82	0.672
Ornithine	C00077	131.08	53.53	< 0.01	0.01	1.51	0.72	0.697
2-Furoic acid	C01546	111.01	72.00	< 0.01	0.02	1.58	0.78	0.701
5'-Oxoinosine	C01993	301.04	487.71	< 0.01	0.02	1.29	0.81	0.702
N-Formylmethionine	C03145	176.04	385.11	< 0.01	0.02	1.62	0.81	0.690
Catechol	C00090	128.07	378.58	< 0.01	0.03	1.37	1.37	0.671
Nicotine	C16150	163.13	469.23	< 0.01	0.04	1.57	0.87	0.688
Deoxyadenosine	C00559	252.11	371.79	< 0.01	0.04	1.63	0.85	0.667
Isocitrate	C00311	191.02	134.56	< 0.01	0.04	1.52	0.66	0.684
Guanine	C00242	152.06	372.34	< 0.01	0.04	1.43	0.84	0.735
Lupinine	C10773	192.14	377.37	< 0.01	0.04	1.53	0.84	0.667
Notes: KEGG, Kyoto Encyclopedia of Genes and Genomes Metabolite ID; mz, mass/charge ratio; RT, retention time; Multiples of FC, LAA subtype IS group/SVO subtype IS group; LAA, large-artery atherosclerosis; SVO: small-vessel occlusion

3.5 Differential metabolic pathway analysis of LAA and SVO subtype IS patients

The KEGG ID list of 26 differential metabolites was input into MetaboAnalyst 5.0 for metabolic pathway analysis and pathway bubble plots were drawn (Table 3 and Fig. 1F). Our results showed significant differences in the tricarboxylic acid cycle (TCA cycle), alanine, aspartate, and glutamate metabolism, and pyruvate metabolism pathways between LAA and SVO type IS patients. Among them, the TCA cycle pathway had the strongest discriminative ability between LAA and SVO subtype IS (p < 0.0001), and was also the most significantly altered metabolic pathway between the two groups (impact value was 0.256).

Table 3

Results of KEGG differential pathway analysis between patients with LAA and SVO subtype IS
Pathway	p-value	Impact
Tricarboxylic acid cycle (TCA cycle)	< 0.0001	0.256
Alanine, aspartate and glutamate metabolism	0.0007	0.116
Pyruvate metabolism	0.0041	0.238

3.6 ROC analysis of differential metabolites in LAA and SVO type IS patients

We performed ROC analysis on all 26 differential metabolites in IS patients to explore the effectiveness of their relative abundance changes in distinguishing between LAA and SVO subtype IS. The results showed that only one metabolite, 12, 13-DHOME (KEGG ID: C14829) with 0.9 > AUC > 0.8 in plasma of patients with different subtypes of IS (its isomer, 9, 10-DHOME, KEGG ID: C14828), its AUC = 0.837 (0.744–0.910) (Supplementary Fig. 3A). Fifteen metabolites with 0.8 > AUC > 0.7 were identified in plasma of patients with different subtypes of IS (Table 2). In all the differential metabolites, isocitrate (KEGG ID: C00311), L-malic acid (KEGG ID: C00149), citric acid (KEGG ID: C00158), pyruvic acid (KEGG ID: C00022), and fumaric acid (KEGG ID: C00122) belonged to the TCA cycle pathway; L-glutamine (KEGG ID: C00064), citric acid, fumaric acid, and pyruvic acid belonged to alanine, aspartate, and glutamate metabolism pathway; L-glutamine, cyclic AMP (3', 5'-Cyclic AMP, KEGG ID: C00575), deoxyadenosine (KEGG ID: C00559), guanine (KEGG ID: C00242), and P1, P4-bis(5'-guanosyl) tetraphosphate (KEGG ID: C01261) belonged to pyruvate metabolism pathway. The relative abundance and ROC curves of these 10 differential metabolites were shown in Supplementary Fig. 3A-K.

3.7 Correlation analysis between differential metabolites and clinical indicators

We performed correlation analysis between the 26 differential metabolites and clinical indicators in all IS patients, including SBP, DBP, WBC, TG, TC, HDL-C, LDL-C, FBG, HbA1C, and Hcy (Supplementary Table 1). The results showed that the relative abundances of pyruvic acid, P1, P4-bis(5'-guanosyl) tetraphosphate, isocitrate, fumaric acid, 3',5'-cyclic AMP, and deoxyadenosine had no significant correlation with common IS risk factors. The relative abundance of L-glutamine was negatively correlated with WBC and FBG levels (p < 0.001 and p = 0.038, respectively). The relative abundance of L-malic acid was negatively correlated with LDL-C (p = 0.048), while the relative abundance of citric acid was negatively correlated with WBC and LDL-C levels (p = 0.027 and p = 0.018, respectively). The relative abundance of guanine was positively correlated with SBP, WBC, and FBG levels (p = 0.030, p < 0.001, and p = 0.006, respectively).

3.8 Performance evaluation of multiple differential metabolites in distinguishing between two subtypes of IS

We employed an exploratory analysis method based on the multivariate ROC curve, and used a metabolite classification method based on the PLS-DA and a feature ranking method based on the built-in PLS-DA to establish a multivariate algorithm. All models were subjected to Monte-Carlo cross-validation using MetaboAnalyst 5.0, and were subjected to multiple rounds of feature selection, model building, and diagnostic performance evaluation (Fig. 2A-B). Model 2 had the highest AUC among the 6 models (AUC = 0.822, 0.741–0.906), with a predictive accuracy of 77.8%, indicating the best diagnostic performance among the 6 models. The feature metabolites for constructing Model 2 included 9,10-DHOME (KEGG ID: C14828), 12,13-DHOME (KEGG ID: C14829), pregnenolone sulfate (KEGG ID: C18044), pyruvic acid (KEGG ID: C00022), citric acid (KEGG ID: C00158), (R)-3-((R)-3-Hydroxybutanoyloxy) butanoate (KEGG ID: C04546), p-aminobenzoate (KEGG ID: C00568), trigonelline (KEGG ID: C01004), P1,P4-bis(5'-guanosyl) tetraphosphate (KEGG ID: C01261), and L-ornithine (KEGG ID: C00077). During the model building and cross-validation process, 12,13-DHOME (the isomer of 9,10-DHOME) had the highest probability and average importance on the predictive biomarker panel (Fig. 2C-D), and was the most important differential metabolite for distinguishing between LAA and SVO subtype of IS.

This study conducted an untargeted plasma metabolomics analysis on different subtypes of IS patients and found 26 differential metabolites and three differential metabolic pathways between the two groups. Ten differential metabolites were included in the differential metabolic pathways, including isocitrate, L-malic acid, citric acid, pyruvic acid, fumaric acid, L-glutamine, cyclic AMP, deoxyadenosine, guanine, and P1, P4-bis(5'-guanosyl) tetraphosphate. We also found that 12,13-DHOME had the highest VIP value and an AUC of 0.830 among the differential metabolites between LAA and SVO subtype IS patients. In the constructed multivariate ROC model, 12,13-DHOME was also the metabolite with the highest probability of appearing on the predicted biomarker panel and the highest average importance. Compared with SVO subtype IS, the relative abundance of 12,13-DHOME in the plasma of LAA subtype IS patients was significantly higher. Previous studies have shown that 12,13-DHOME can regulate multiple genes related to energy metabolism and lipid metabolism by binding to PPAR receptors (29, 30). Other studies have shown that 12,13-DHOME can inhibit mitochondrial function and increase oxidative stress in mammals(31, 32). We speculate that the different pathophysiological mechanisms between LAA-subtype and SVO-subtype IS may involve mitochondrial function, oxidative stress, and energy metabolism, which require further investigation.

The brain is the most energy-demanding organ in the human body, consuming approximately 20% of the total human energy production. Neurons in the brain are highly sensitive to ischemia and hypoxia, and mitochondrial respiratory chain dysfunction and disruptions in the TCA cycle can lead to decreased brain energy metabolism, which can in turn affect brain function and cause brain injury. Studies have shown that hindrance of the citric acid metabolism process in ischemic brain tissue of middle cerebral artery occlusion (MCAO) mice leads to decreased TCA cycle flux and adenosine triphosphate (ATP) production, ultimately resulting in neuronal death (33, 34). Our study found that the pyruvate metabolism and TCA cycle pathways in LAA subtype IS patients were less active compared to SVO subtype IS patients, consistent with in vitro study results. Additionally, combining the relative abundance ratio (FC value) of several metabolites that belong to the energy metabolism pathway showed that the energy metabolism level was more active in SVO subtype IS patients than in LAA subtype IS patients. Furthermore, several differential metabolites involved in the pyruvate metabolism and TCA cycle pathways (pyruvic acid, P1, P4-bis(5'-adenosyl) tetraphosphate, isocitrate, fumaric acid, cyclic AMP, deoxyadenosine) had no significant correlation with IS risk factors, confirming the existence of energy metabolism differences between LAA and SVO-subtype IS patients. Over 95% of the energy required by the brain is supplied by glucose, and almost all substrates of the mitochondrial respiratory chain come from cytoplasmic glycolysis products; therefore, if glycolysis is disrupted, even with adequate oxygen supply, the brain cannot maintain normal energy metabolism levels (35). Although oxygen plays a crucial role in mitochondrial ATP production, neurons and glial cells in the brain can survive for several hours under hypoxic conditions by relying solely on anaerobic glycolysis occurring in the cytoplasm (36, 37). However, if both glucose and oxygen supply are simultaneously halted, brain cells will completely deplete ATP within three minutes(38). In this study, FBG and HbA1C levels at admission were significantly lower in LAA subtype IS patients than in SVO subtype IS patients. Th lower blood glucose levels may also be one of the causes of energy metabolism dysfunction in LAA subtype IS patients.

Based on the exploratory analysis method of multivariate ROC curves, we constructed multiple prediction models using plasm metabolites. The prediction model (Model 2) that included 10 differential metabolites had the best predictive performance with an AUC of 0.822 and a prediction accuracy of 77.8%. Citric acid, pyruvic acid, and deoxyadenosine were among the top 10 metabolites with the highest selection probability in the predicted biomarker panel, while citric acid, pyruvic acid, and P1, P4-bis(5'-guanosyl) tetraphosphate were among the top 10 metabolites with the highest average importance in the predicted biomarker panel. These differential metabolites were all involved in the pyruvate metabolism and TCA cycle pathways. These results further confirmed the difference in energy metabolism between LAA and SVO subtype IS patients.

In summary, this study used untargeted metabolomics technology to discover 26 differential metabolites between LAA and SVO subtype IS patients. The multivariate ROC curve constructed with 10 differential metabolites showed good predictive ability for subtyping. 12, 13-DHOME was the most important differential metabolite in distinguishing different subtypes of IS. Metabolic pathway analysis revealed three differential metabolic pathways between the two groups, all of which were concentrated in energy metabolism. Differential metabolite analysis found that several metabolites belonging to the energy metabolism pathway had a higher relative abundance in SVO subtype IS patients than in LAA subtype IS patients and were not significantly correlated with IS risk factors. Overall, the findings suggest that the energy metabolism level of SVO subtype IS patients is more active than that of LAA subtype IS patients. This result was not been identified in previous metabolomic studies of different types of IS, which providing new potential targets for exploring the pathological and physiological mechanisms of different IS subtypes(39, 40).

However, there are some limitations to our study. First, this was a cross-sectional study, so we were unable to establish a causal relationship between differential metabolites and IS typing. Secondly, the participants in our study came from one center and were all Han, which may lead to patient selection bias and racial heterogeneity. Third, LC-HRMS cannot distinguish metabolites with different spatial conformations with the same mass-to-charge ratio and retention time, and the results need further targeted metabolomics verification.

Conflicts of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgement

This work was supported by the National Key Research and Development Projects (2022YFC3602400, 2022YFC3602401), the National Natural Science Foundation of China (82271369) and the Provincial Key Plan for Research and Development of Hunan (Grant No. 2020SK2067).

Conflicts of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Ethics statement

The studies involving human participants were reviewed and approved by Xiangya Hospital Ethics Committee. The patients/participants provided their written informed consent to participate in this study.

Data Availability Statement

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.

Authors’ contributions

Study design: Xi Li, Jiaxin Li; data collection and analysis: Xi Li, Jiaxin Li, Fang Yu, Xianjing Feng, Yunfang Luo, Weijia Xie, Yinghuan Pan; writing: Xi Li, Jiaxin Li; funding: Jian Xia; administration: Jian Xia

Supplementary

Supplementary material 1: Supplementary Tables

Supplementary material 2: Supplementary Figures

Liu S, Li Y, Zeng X, Wang H, Yin P, Wang L, et al. Burden of Cardiovascular Diseases in China, 1990-2016: Findings from the 2016 Global Burden of Disease Study. JAMA cardiology (2019) 4(4):342-52. Epub 2019/03/14. doi: 10.1001/jamacardio.2019.0295.
Wu S, Wu B, Liu M, Chen Z, Wang W, Anderson CS, et al. Stroke in China: Advances and Challenges in Epidemiology, Prevention, and Management. The Lancet Neurology (2019) 18(4):394-405. Epub 2019/03/18. doi: 10.1016/s1474-4422(18)30500-3.
Campbell BCV, De Silva DA, Macleod MR, Coutts SB, Schwamm LH, Davis SM, et al. Ischaemic Stroke. Nature reviews Disease primers (2019) 5(1):70. Epub 2019/10/12. doi: 10.1038/s41572-019-0118-8.
Wang YJ, Li ZX, Gu HQ, Zhai Y, Jiang Y, Zhao XQ, et al. China Stroke Statistics 2019: A Report from the National Center for Healthcare Quality Management in Neurological Diseases, China National Clinical Research Center for Neurological Diseases, the Chinese Stroke Association, National Center for Chronic and Non-Communicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention and Institute for Global Neuroscience and Stroke Collaborations. Stroke and vascular neurology (2020) 5(3):211-39. Epub 2020/08/23. doi: 10.1136/svn-2020-000457.
Liu M, Wu B, Wang WZ, Lee LM, Zhang SH, Kong LZ. Stroke in China: Epidemiology, Prevention, and Management Strategies. The Lancet Neurology (2007) 6(5):456-64. Epub 2007/04/17. doi: 10.1016/s1474-4422(07)70004-2.
Guan T, Ma J, Li M, Xue T, Lan Z, Guo J, et al. Rapid Transitions in the Epidemiology of Stroke and Its Risk Factors in China from 2002 to 2013. Neurology (2017) 89(1):53-61. doi: 10.1212/WNL.0000000000004056.
Adams HP, Jr., Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, et al. Classification of Subtype of Acute Ischemic Stroke. Definitions for Use in a Multicenter Clinical Trial. Toast. Trial of Org 10172 in Acute Stroke Treatment. Stroke (1993) 24(1):35-41. Epub 1993/01/01. doi: 10.1161/01.str.24.1.35.
Montaner J, Perea-Gainza M, Delgado P, Ribó M, Chacón P, Rosell A, et al. Etiologic Diagnosis of Ischemic Stroke Subtypes with Plasma Biomarkers. Stroke (2008) 39(8):2280-7. Epub 2008/06/07. doi: 10.1161/strokeaha.107.505354.
Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. Classification and Natural History of Clinically Identifiable Subtypes of Cerebral Infarction. Lancet (London, England) (1991) 337(8756):1521-6. Epub 1991/06/22. doi: 10.1016/0140-6736(91)93206-o.
Sacco SE, Whisnant JP, Broderick JP, Phillips SJ, O'Fallon WM. Epidemiological Characteristics of Lacunar Infarcts in a Population. Stroke (1991) 22(10):1236-41. Epub 1991/10/01. doi: 10.1161/01.str.22.10.1236.
Petty GW, Brown RD, Jr., Whisnant JP, Sicks JD, O'Fallon WM, Wiebers DO. Ischemic Stroke Subtypes: A Population-Based Study of Incidence and Risk Factors. Stroke (1999) 30(12):2513-6. Epub 1999/12/03. doi: 10.1161/01.str.30.12.2513.
Boodt N, Compagne KCJ, Dutra BG, Samuels N, Tolhuisen ML, Alves H, et al. Stroke Etiology and Thrombus Computed Tomography Characteristics in Patients with Acute Ischemic Stroke: A Mr Clean Registry Substudy. Stroke (2020) 51(6):1727-35. Epub 2020/05/15. doi: 10.1161/strokeaha.119.027749.
Vemmos KN, Takis CE, Georgilis K, Zakopoulos NA, Lekakis JP, Papamichael CM, et al. The Athens Stroke Registry: Results of a Five-Year Hospital-Based Study. Cerebrovascular diseases (Basel, Switzerland) (2000) 10(2):133-41. Epub 2000/02/25. doi: 10.1159/000016042.
Kelly DM, Li L, Rothwell PM. Etiological Subtypes of Transient Ischemic Attack and Ischemic Stroke in Chronic Kidney Disease: Population-Based Study. Stroke (2020) 51(9):2786-94. Epub 2020/08/20. doi: 10.1161/strokeaha.120.030045.
Grau AJ, Weimar C, Buggle F, Heinrich A, Goertler M, Neumaier S, et al. Risk Factors, Outcome, and Treatment in Subtypes of Ischemic Stroke: The German Stroke Data Bank. Stroke (2001) 32(11):2559-66. Epub 2001/11/03. doi: 10.1161/hs1101.098524.
Hajat C, Heuschmann PU, Coshall C, Padayachee S, Chambers J, Rudd AG, et al. Incidence of Aetiological Subtypes of Stroke in a Multi-Ethnic Population Based Study: The South London Stroke Register. Journal of neurology, neurosurgery, and psychiatry (2011) 82(5):527-33. Epub 2010/10/27. doi: 10.1136/jnnp.2010.222919.
Nielsen J. Systems Biology of Metabolism. Annual review of biochemistry (2017) 86:245-75. Epub 2017/03/17. doi: 10.1146/annurev-biochem-061516-044757.
Jung JY, Lee HS, Kang DG, Kim NS, Cha MH, Bang OS, et al. 1h-Nmr-Based Metabolomics Study of Cerebral Infarction. Stroke (2011) 42(5):1282-8. Epub 2011/04/09. doi: 10.1161/strokeaha.110.598789.
Au A. Metabolomics and Lipidomics of Ischemic Stroke. Advances in clinical chemistry (2018) 85:31-69. Epub 2018/04/16. doi: 10.1016/bs.acc.2018.02.002.
Chi NF, Chang TH, Lee CY, Wu YW, Shen TA, Chan L, et al. Untargeted Metabolomics Predicts the Functional Outcome of Ischemic Stroke. Journal of the Formosan Medical Association = Taiwan yi zhi (2021) 120(1 Pt 1):234-41. Epub 2020/05/18. doi: 10.1016/j.jfma.2020.04.026.
Fan TW, Higashi RM, Lane AN. Integrating Metabolomics and Transcriptomics for Probing Se Anticancer Mechanisms. Drug metabolism reviews (2006) 38(4):707-32. Epub 2006/12/06. doi: 10.1080/03602530600959599.
Wissenbach DK, Meyer MR, Remane D, Philipp AA, Weber AA, Maurer HH. Drugs of Abuse Screening in Urine as Part of a Metabolite-Based Lc-Msn Screening Concept. Analytical and bioanalytical chemistry (2011) 400(10):3481-9. Epub 2011/05/03. doi: 10.1007/s00216-011-5032-1.
Rocha BA, Asimakopoulos AG, Barbosa F, Jr., Kannan K. Urinary Concentrations of 25 Phthalate Metabolites in Brazilian Children and Their Association with Oxidative DNA Damage. The Science of the total environment (2017) 586:152-62. Epub 2017/02/09. doi: 10.1016/j.scitotenv.2017.01.193.
Gago-Ferrero P, Schymanski EL, Bletsou AA, Aalizadeh R, Hollender J, Thomaidis NS. Extended Suspect and Non-Target Strategies to Characterize Emerging Polar Organic Contaminants in Raw Wastewater with Lc-Hrms/Ms. Environmental science & technology (2015) 49(20):12333-41. Epub 2015/09/30. doi: 10.1021/acs.est.5b03454.
Jickling GC, Sharp FR. Biomarker Panels in Ischemic Stroke. Stroke (2015) 46(3):915-20. Epub 2015/02/07. doi: 10.1161/strokeaha.114.005604.
Yu F, Zhou X, Li Z, Feng X, Liao D, Liu Z, et al. Diagnostic Significance of Plasma Levels of Novel Adipokines in Patients with Symptomatic Intra- and Extracranial Atherosclerotic Stenosis. Frontiers in neurology (2019) 10:1228. doi: 10.3389/fneur.2019.01228.
Smith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak G. Xcms: Processing Mass Spectrometry Data for Metabolite Profiling Using Nonlinear Peak Alignment, Matching, and Identification. Analytical chemistry (2006) 78(3):779-87. Epub 2006/02/02. doi: 10.1021/ac051437y.
Shen X, Wang R, Xiong X, Yin Y, Cai Y, Ma Z, et al. Metabolic Reaction Network-Based Recursive Metabolite Annotation for Untargeted Metabolomics. (2019) 10(1):1516. doi: 10.1038/s41467-019-09550-x.
Medina-Gomez G, Gray S, Vidal-Puig A. Adipogenesis and Lipotoxicity: Role of Peroxisome Proliferator-Activated Receptor Gamma (Ppargamma) and Ppargammacoactivator-1 (Pgc1). Public health nutrition (2007) 10(10a):1132-7. Epub 2007/11/21. doi: 10.1017/s1368980007000614.
Lecka-Czernik B, Moerman EJ, Grant DF, Lehmann JM, Manolagas SC, Jilka RL. Divergent Effects of Selective Peroxisome Proliferator-Activated Receptor-Gamma 2 Ligands on Adipocyte Versus Osteoblast Differentiation. Endocrinology (2002) 143(6):2376-84. Epub 2002/05/22. doi: 10.1210/endo.143.6.8834.
Edwards LM, Lawler NG, Nikolic SB, Peters JM, Horne J, Wilson R, et al. Metabolomics Reveals Increased Isoleukotoxin Diol (12,13-Dhome) in Human Plasma after Acute Intralipid Infusion. Journal of lipid research (2012) 53(9):1979-86. Epub 2012/06/21. doi: 10.1194/jlr.P027706.
Viswanathan S, Hammock BD, Newman JW, Meerarani P, Toborek M, Hennig B. Involvement of Cyp 2c9 in Mediating the Proinflammatory Effects of Linoleic Acid in Vascular Endothelial Cells. Journal of the American College of Nutrition (2003) 22(6):502-10. Epub 2003/12/20. doi: 10.1080/07315724.2003.10719328.
Chen L, Chao Y, Cheng P, Li N, Zheng H, Yang Y. Uplc-Qtof/Ms-Based Metabolomics Reveals the Protective Mechanism of Hydrogen on Mice with Ischemic Stroke. Neurochemical research (2019) 44(8):1950-63. Epub 2019/06/27. doi: 10.1007/s11064-019-02829-x.
Zhang SJ, Sandström ME, Lanner JT, Thorell A, Westerblad H, Katz A. Activation of Aconitase in Mouse Fast-Twitch Skeletal Muscle During Contraction-Mediated Oxidative Stress. American journal of physiology Cell physiology (2007) 293(3):C1154-9. Epub 2007/07/07. doi: 10.1152/ajpcell.00110.2007.
Baxter P, Chen Y, Xu Y, Swanson RA. Mitochondrial Dysfunction Induced by Nuclear Poly(Adp-Ribose) Polymerase-1: A Treatable Cause of Cell Death in Stroke. Translational stroke research (2014) 5(1):136-44. Epub 2013/12/11. doi: 10.1007/s12975-013-0283-0.
Longuemare MC, Hill MP, Swanson RA. Glycolysis Can Prevent Non-Synaptic Excitatory Amino Acid Release During Hypoxia. Neuroreport (1994) 5(14):1789-92. Epub 1994/09/08. doi: 10.1097/00001756-199409080-00026.
MacMillan VH. Cerebral Energy Metabolism in Cyanide Encephalopathy. Journal of cerebral blood flow and metabolism : official journal of the International Society of Cerebral Blood Flow and Metabolism (1989) 9(2):156-62. Epub 1989/04/01. doi: 10.1038/jcbfm.1989.23.
Lowry OH, Passonneau JV, Hasselberger FX, Schulz DW. Effect of Ischemia on Known Substrates and Cofactors of the Glycolytic Pathway in Brain. The Journal of biological chemistry (1964) 239:18-30. Epub 1964/01/01.
Lee EJ, Kim DJ, Kang DW, Yang W, Jeong HY, Kim JM, et al. Targeted Metabolomic Biomarkers for Stroke Subtyping. Transl Stroke Res (2023). Epub 2023/02/11. doi: 10.1007/s12975-023-01137-5.
Wang X, Zhang L, Sun W, Pei LL, Tian M, Liang J, et al. Changes of Metabolites in Acute Ischemic Stroke and Its Subtypes. Front Neurosci (2020) 14:580929. Epub 2021/01/29. doi: 10.3389/fnins.2020.580929.

No competing interests reported.

Download PDF

Journal Publication

published 06 Jan, 2024

Read the published version in Molecular Neurobiology →

Version 1

posted

You are reading this latest preprint version

The Untargeted Metabolomics Reveals Differences in Energy Metabolism in Patients with Different Subtypes of Ischemic Stroke

Status:

Journal Publication

Version 1

Abstract

Figures

1 Introduction

2 Materials And Methods

2.1. Participants

2.2. Neuroimaging

2.3. Clinical Examination

2.4. Sample Preparation

2.5 LC-HRMS Analysis

2.6. Data Processing and Statistical Analysis

2.6.1. Untargeted Metabolomics Data Processing

2.7 Statistical Analysis

3 Results

3.1 Demographic and Clinical Characteristics of the Study Population

3.2 Quality control of untargeted metabolomics mass spectrometry data

3.3 Plasma metabolomics multivariate statistical analysis of LAA and SVO subtype IS patients

3.3.1 Principal component analysis (PCA)

3.3.2 Partial least squares-discriminant analysis (PLS-DA)

3.3.3 Orthogonal partial least squares-discriminant analysis (OPLS-DA)

3.4 Univariate statistical analysis of plasma metabolic profiles of LAA and SVO subtype IS patients

3.5 Differential metabolic pathway analysis of LAA and SVO subtype IS patients

3.6 ROC analysis of differential metabolites in LAA and SVO type IS patients

3.7 Correlation analysis between differential metabolites and clinical indicators

3.8 Performance evaluation of multiple differential metabolites in distinguishing between two subtypes of IS

4 Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1