Graphical abstract graphic file with name ga1.jpg [49]Open in a new tab Keywords: Traumatic brain injury, Hydroxysafflor yellow A, Metabolomics, Network pharmacology, Mechanisms Abstract Traumatic brain injury (TBI) has become a leading cause of mortality, morbidity and disability worldwide. Hydroxysafflor yellow A (HSYA) is effective in treating TBI, but the potential mechanisms require further exploration. We aimed to reveal the mechanisms of HSYA against acute TBI by an integrated strategy combining metabolomics with network pharmacology. A controlled cortical impact (CCI) rat model was established, and neurological functions were evaluated. Metabolomics of brain tissues was used to identify differential metabolites, and the metabolic pathways were enriched by MetaboAnalyst. Then, network pharmacology was applied to dig out the potential targets against TBI induced by HSYA. The integrated network of metabolomics and network pharmacology was constructed based on Cytoscape. Finally, the obtained key targets were verified by molecular docking. HSYA alleviated the neurological deficits of TBI. Fifteen potentially significant metabolites were found to be involved in the therapeutic effects of HSYA against acute TBI. Most of these metabolites were regulated to recover after HSYA treatment. We found 10 hub genes according to network pharmacology, which was partly consistent with the metabolomics findings. Further integrated analysis focused on 4 key targets, including NOS1, ACHE, PTGS2 and XDH, as well as their related core metabolites and pathways. Molecular docking showed high affinities between key targets and HSYA. Region-specific metabolic alterations in the cortex and hippocampus were illuminated. This study reveals the complicated mechanisms of HSYA against acute TBI. Our work provides a novel paradigm to identify the potential mechanisms of pharmacological effects derived from a natural compound. 1. Introduction Traumatic brain injury (TBI) is a leading cause of mortality, morbidity and disability worldwide [50][1]. The disease triggers a cascade of pathophysiological events, such as disrupting biochemical, metabolic, and molecular functions, disturbing brain cell homeostasis and impairing cognitive, motor, or neuropsychological health [51][2]. Despite large efforts to develop neuroprotective therapies for TBI, no drugs have been approved by the Food and Drug Administration (FDA). Natural bioactive compounds tend to be promising agents against brain injury [52][3], [53][4]. Safflower (Carthamus tinctorius L.), a well-known traditional Chinese medicine, is widely used to treat cerebrovascular diseases. Hydroxysafflor yellow A (HSYA, C[27]H[32]O[16], 612.500 g/mol, [54]Fig. S1) is the main active ingredient of safflower and exerts antiinflammatory, antiapoptotic, antioxidative and neuroprotective effects [55][5], [56][6]. Our previous research demonstrated that HSYA could across the injured blood–brain barrier of TBI patients to exert a neuroprotective effect [57][7]. Further investigation suggested that HSYA prevents oxidative stress post TBI by increasing the activity of antioxidant enzymes [58][8]. However, the mechanisms and targets of HSYA in treating TBI have not been fully elucidated. Given that TBI reflects perturbations in complex metabolic physiologies, metabolomics is powerful for monitoring the dynamic changes in pathological metabolites [59][9]. However, traditional metabolomics could only reflect the terminal variation of disease and treatment [60][10], [61][11], [62][12]. It is unclear about the endogenous mechanisms of metabolites’ changes, including how these metabolites are produced, what their upstream pathways and proteins are, and which proteins HSYA exerts effects through. Thus, metabolomics alone may limit the application of HSYA. Currently, the paradigm of developing single target-based drugs as therapeutics has been challenged mainly due to lack of efficacy and emerging resistance [63][13]. Thus, natural compounds that selectively act on two or more targets of interest in theory should be more efficacious than single-target agents [64][14]. Network pharmacology appears in this setting to construct an alternative systems-level approach to find new drug candidates. Instead of looking for a single disease-causing gene and drugs which act solely on an individual target, the whole drug-disease network is considered with the aim to find multi-targets drugs to reduce side effects [65][15]. Nonetheless, network pharmacology is limited by the single computational methods that rely on public databases. Network pharmacology alone could only predict the possibility of compound-target combination and pathway analysis [66][16]. It is uncertain whether HSYA binds to targets in vivo and which effect HSYA exerts on targets: inhibition, activation or ineffective combination. Therefore, we integrated metabolomics with network pharmacology. Untargeted metabolomics was applied to determine the influences of HSYA on TBI and to identify the essential metabolites. Subsequently, network pharmacology was performed to analyze the proteins and reactions that modulated the metabolites, as well as the targets that HSYA acted on. Collectively, this strategy compensates network pharmacology for lacking experimental validation and metabolomics for lacking upstream molecular mechanisms and drug-binding targets. This strategy will hopefully contribute to a better understanding of the therapeutic principle of natural compounds for TBI treatment. In the present study, we first developed a novel integrated strategy to explore the key targets and mechanisms of HSYA in treating acute TBI based on metabolomics and network pharmacology. Furthermore, we identified region-specific metabolic responses (cortex and hippocampus) in the HSYA treated rat model of TBI. This study provides new insight into the neuroprotective effects of HSYA in treating TBI. The research flowchart is shown in [67]Fig. 1. Fig. 1. [68]Fig. 1 [69]Open in a new tab The schematic flowchart of the integrated strategy. The mechanisms of HSYA against TBI were analyzed by metabolomics of brain tissues (Part 1). Hub genes were extracted by network pharmacology (Part 2). Key metabolites and targets were identified and linked based on Part 1 and 2. These key targets were further verified by molecular docking (Part 3). 2. Material and methods 2.1. Reagents and materials Hydroxysafflor yellow A (HSYA) was purchased from Shanghai Yuanye Bio-Technology Co., Ltd (Shanghai, China; purity: 90%, lot number: S26799). Ammonium acetate, ammonium hydroxide and methanol (grade: for HPLC) were provided by Sigma-Aldrich (St. Louis, MO, USA), and acetonitrile and H[2]O (grade: for HPLC) were obtained from J.T.Baker (PA, USA). All of the remaining reagents were of analytical grade. 2.2. Animals and the controlled cortical impact (CCI) model 7-week-old male specific-pathogen-free Sprague Dawley rats were obtained from the Laboratory Animal Centre of Central South University (Changsha, China). All rats were housed in a well-ventilated room at 25°C, with a 12 h dark-light cycle and free access to food and water. Animal care was performed under the guidelines of Central South University for the care and use of animals and the protocol was approved by the Medical Ethics Committee of Central South University. Rats were randomly assigned to 3 groups (n = 10 per day per group): sham group, CCI group and HSYA group. Rats in the HSYA group were orally administrated HSYA (0.87 mg/ml, dissolved in 0.9% saline) once a day at a dose of 13.88 mg/kg. Rats in the sham and CCI groups were treated with an equal volume of saline solution. Replication of the CCI rat model was performed according to a previous study [70][17]. The parameters were set as follows: impact depth, 5.0 mm; striking speed, 6.0 m/s; dwell time, 50 ms. Rats in the sham group were operated identically to those in the CCI and HSYA groups, except for cortical impact. Rats were anaesthetized by intraperitoneal injection of sodium pentobarbital (60 mg/kg) and sacrificed at day 1 (n = 10 per group) and day 3 (n = 10 per group) after CCI. The ipsilateral cortex and hippocampus were collected from all of the rats after perfusion with ice-cold saline and stored at −80 °C for further use. 2.3. Neurological function testing All animals were assessed by the modified neurologic severity score (mNSS) test and weight change. The 18-point mNSS comprises motor, sensory, reflex abilities and balance tests [71][18]. Higher scores indicate more serious damage. Rats were evaluated before and after injury to verify the neuroprotective effect of HSYA in the CCI model. The body weight of rats was recorded before and after injury, and the percentage of weight change was calculated. 2.4. Sample preparation 50 mg of tissue was homogenized with 400 µL of H[2]O. A BCA protein assay was performed to measure the total protein concentration on each of the individual homogenates. 100 µL aliquots of homogenates were precipitated by adding methanol and acetonitrile as a ratio of 1:1 (v/v). After vortexing for 30 s and sonicating for 10 min, the samples were incubated for 1 h at 20 °C and centrifuged at 20,000 g at 4 °C. Then, the supernatants were collected and dried in a vacuum concentrator. Finally, the dry extracts were reconstituted with 40 µL/mg acetonitrile and H[2]O (1:1, v/v) for HPLC/MS analysis. The pooled quality control (QC) samples were made by mixing 10 µL aliquots from each sample (one per six samples). 2.5. HPLC-MS/MS analysis Metabolomics was applied using 1260 infinity high-performance liquid chromatography (Agilent, CA, USA) coupled with Q-Exactive MS/MS (Thermo, MA, USA). Chromatographic separations were performed on an amide column at 25 °C. The mobile phase consisted of water mixing with 25 mM ammonium acetate, 25 mM ammonium hydroxide (solvent A) and acetonitrile (solvent B). The gradient program was as follows: 90% B (0–1.0 min), 90 to 87% B (1.0–11.0 min), 87–80% B (11.0–14.0 min), 80–70% B (14.0–16.5 min), 70–50% B (16.5–18.5 min), 50–20% B (18.5–20.5 min), 20% B (20.5–25.0 min), 20–90% (25.0–25.1 min) and maintained at 90% B until 34 min. The injection volume was 4 µL and the flow rate was 0.4 mL/min. MS analysis was carried out on the Q-Exactive MS/MS in both positive and negative ion modes. Setting the relevant tuning parameters for the probe: aux gas heater temperature, 400 °C; spray voltage, 3.5 kV; sheath gas, 40 psi; auxiliary gas, 13 psi; capillary temperature, 350 °C. Building a DDA method as follows: full scan range was 60–900 m/z; maximum injection time for MS1 and ddMS2: 100 ms and 45 ms; resolution for MS1 and ddMS2: 70,000 and 17,500 respectively; automatic gain control for MS1 and ddMS2: 3e^6 and 2e^5; isolation window: 1.6 m/z; normalized collision energies: 10, 17, 25 or 30, 40, 50. Building a full scan method as follows: full scan range: 60 to 900 m/z; resolution: 140,000; maximum injection time: 100 ms; automatic gain control: 3e^6 ions. 2.6. Data processing and analysis The acquired raw files were preprocessed using Thermo Compound Discover 2.1 (Thermo, MA, USA) software. Intensities were corrected for signal drift and batch effect by fitting a locally quadratic (loess) regression model to the median intensity of pooled QC samples. The alpha parameter controlling the smoothing was set to 2 to avoid overfitting. After correction, the median area of all pooled QC samples was the same. The data were pretreated using the “80% rule” [72][19] to reduce the missing value input. HPLC-MS/MS analysis and data processing were conducted by KangChen Bio-tech (China). The features with relative standard deviations (RSDs) > 30% were removed from all the QC samples. The pretreated data were calibrated with median, transformed with log, and scaled with Pareto, then analyzed using principal component analysis (PCA), supervised partial least squares discrimination analysis (PLS-DA), and orthogonal partial least squared discriminant analysis (OPLS-DA) in R software (version 3.6.0) by the ropls R package. 7-round cross-validation and 200 permutation test were performed to evaluate the accuracy of the models. Features were further subjected to one-way analysis of variance (ANOVA) with a false discovery rate (FDR) at the univariate level to measure the significance of each metabolite (q-value). The features with variable importance in the projection (VIP) > 1 and q-value < 0.05 were considered to be differential compounds. These features were identified by performing retention time alignment, unknown compound detection, and compound grouping across all samples. For retention time alignment, the max time shift was 2 mins, and a tolerance of 0.5 min was used for grouping unknown compounds. Mass tolerance for feature detection and compound annotation was set as 10 ppm and 5 ppm respectively. The formula and accurate mass of each feature were submitted to ChemSpider ([73]http://www.chemspider.com/) with 4 databases selected (BioCyc; Human Metabolome Database; Kyoto Encyclopedia of Genes and Genomes [KEGG]; LipidMAPS). The metabolite with the most references was