Abstract Sleep deprivation (SD) causes learning memory and cognitive impairment. Salvia miltiorrhiza Bunge (Danshen, DS), a medicinal plant in the family Labiatae, has been traditionally used for sleep-related disorders. Previous studies have shown that DS can ameliorate SD-induced cognitive impairment. However, the underlying mechanisms for its pharmacological effects remain unclear. This study aimed to evaluate the protective effects and mechanisms of DS extract against cognitive impairment in SD rats. UPLC-QTOF/MS was used to analyze DS extracts. The SD model was constructed utilizing a modified multi-platform aquatic sleep deprivation procedure that lasted 21 days. The Morris water maze test (MWM), hematoxylin and eosin (H&E) staining, and enzyme-linked immunosorbent assay (ELISA) were used to assess learning and memory ability, hippocampus injury, and serum inflammation, respectively. An integrated strategy of serum metabolomics combined with network analysis was used to explore the potential mechanisms by which DS exerts pharmacological effects. Molecular docking and experiments were used for further validation. UPLC-QTOF-MS/MS identified 32 diterpenoids in DS extract. The results showed that DS (1.35 and 2.70 g/kg) significantly improved spatial learning and memory abilities in SD rats while also reducing hippocampus pathological damage and serum inflammation. Serum metabolomics showed that DS modulated 26 differential metabolites, mainly involved in Glycerophospholipid metabolism, Glycerolipid metabolism, Phosphatidylinositol signaling system, and One carbon pool by folate. Network analysis screened 145 putative targets for DS to alleviate SD-induced cognitive impairment, involved in inflammation regulation and metabolic modulation. Integrated analyses of metabolomics and network analysis indicated that PIK3CA was a key target for DS’s regulatory effects, primarily engaged in the regulation of phosphatidylinositol phosphate metabolism. Validation experiments revealed that all eight components of DS extracts had a higher binding ability with PIK3CA, and DS restored the SD-induced abnormal expression of PIK3CA. Our study provides new insights into the development of DS as a dietary supplement for treating SD-induced cognitive impairment. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-14303-6. Keywords: Salvia miltiorrhiza bunge, Metabolomics, Network analysis, Sleep deprivation, Cognitive impairment, PIK3CA Subject terms: Metabolic disorders, Diseases, Neurological disorders Introduction Sleep deprivation (SD) is a disorder in which patients have difficulty falling asleep or staying asleep^[34]1 which leads to tiredness or sleepiness in patients and can even cause cognitive dysfunction such as memory loss or attention loss^[35]2. Sleep-wake cycle disturbances have grown frequent in all age groups as a result of the hurried pace of life^[36]3 having a significant detrimental influence on people’s life, learning, and work. Many experimental and clinical investigations have demonstrated that SD is harmful to the body in many ways, and in addition to producing cognitive impairment^[37]4,[38]5 it has also been linked to metabolic diseases such as obesity and diabetes^[39]6,[40]7. As a chronic disease prone to relapse, current conventional therapeutic drugs, such as acetylcholinesterase inhibitors^[41]8 caffeine, and donepezil^[42]2 can alleviate the symptoms to varying degrees but struggle to cure it. Furthermore, due to their various adverse effects or restrictions, these medications are ineffective in therapy, and their use is disputed and limited. As a result, there is an urgent need to discover a safe and effective therapy to address cognitive impairment caused by SD. Salvia miltiorrhiza Bunge (Danshen, DS), a medicinal plant in the family Labiatae, has been traditionally used for sleep-related disorders^[43]9 and is also customarily used as a functional tea in some regions of China^[44]10. In addition to alleviating cardiovascular and cerebrovascular diseases^[45]11studies on the neuroprotective effects of the fat-soluble components of DS have steadily expanded in recent years. Studies have shown that DS can exert cardioprotective effects by enhancing angiogenesis^[46]12. In vivo and in vitro experiments demonstrated that DS combats Alzheimer’s disease(AD) by promoting neurogenesis in neural progenitor cells^[47]13. In addition, DS inhibits the RAGE/NF-κB signaling pathway and ameliorates cognitive decline and neuroinflammation^[48]14. Our previous study found that DS significantly improved cognitive function in SD rats^[49]9 which was inextricably linked to the LPS-TLR4 pathway mediated by gut flora^[50]8. However, the mechanism of metabolic regulation by DS in SD rats is unclear. It is well known that off-target metabolomics analyses various metabolites at a specific period, a feature that fits with the holistic concept of Chinese medicine, and endogenous metabolites play a crucial role in the development of diseases^[51]15. Network pharmacology is an effective tool for elucidating the synergistic efficacy of multi-components^[52]16 and has great potential for uncovering drug targets and pharmacodynamic mechanisms^[53]17. Taken together, the combined analysis of metabolomics and network analysis help to reveal the potential mechanisms by which DS ameliorates SD-induced cognitive impairment. In the present study, the SD model was constructed by a modified multi-platform aqueous sleep deprivation method, and cognitive functions, hippocampal lesions, as well as serum inflammation in rats were observed to evaluate the therapeutic effects of DS. Following that, Ultra-high performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (UPLC-QTOF/MS) was used to analyze changes in serum metabolites in rats, and multivariate statistical analyses and metabolic networks were used to reveal the mechanisms of DS modulation in the SD model. Network pharmacology was utilized to predict the targets and mechanisms of DS to improve cognitive functions. Metabolomics and network analysis were then analyzed together using MetScape’s Compound-Reaction-Enzyme-Gene network. Finally, the key targets and metabolic networks in which DS plays a regulatory role were experimentally identified. Our research gives fresh insights into the therapeutic value of DS in the functional food and pharmaceutical sectors. Materials and methods Drugs and reagents As stated in our earlier work^[54]8 DS was obtained from HBUCM Affiliated Hospital. After two rounds of reflux extraction with 90% ethanol solution and freeze-drying, the DS extract was produced. Melatonin(purity ≥ 99%) was purchased from Shanghai Yuanye Biotechnology Co., LTD(Shanghai, China.). Interleukin-1β (IL-1β) (NO.RX302869R), interleukin-6 (IL-6) (NO.RX302856R), and tumor necrosis factor-α(TNF-α) (NO.RX302058R) assay kits were acquired from Ruixin Biotechnology(Quanzhou, China.). Chemical profiling of DS extract An ACQUITY UPLC liquid chromatography system coupled with a Waters Xevo G2-XSQTOF high-resolution time-of-flight mass spectrometer was used for the compositional characterization of DS extracts^[55]8. The chromatographic column was Waters ACQUITY UPLC BEH C18 (inner diameter 1.7 μm, 2.1 mm×100 mm). The chromatographic conditions were as follows: methanol as mobile phase A, 0.1% formic acid aqueous solution (v/v) as mobile phase B, column temperature at 40 ℃, detection wavelength at 280 nm, injection volume at 2µL, and flow rate at 0.3 mL/min. The chromatographic conditions are shown in Table [56]1. Mass spectrometry (MS): ESI ion source was used, and the data were collected in positive ion modes, and the primary and secondary MS ions were collected in the MSE mode in the range from 50 to 1200 Da. The secondary collision energy was 30 eV-40 eV, and the enkephalin (m/z 556.2771) tuning solution was used as the real-time correction solution. Table 1. Gradient conditions for compositional analysis. Time(min) A(methanol) B(0.1% formic acid aqueous solution) 0.01 45% 55% 8 70% 30% 11 75% 25% 15 75% 25% 20 80% 20% 25 95% 5% 30 95% 5% 31 45% 55% 35 45% 55% [57]Open in a new tab Animals Liaoning Changsheng Biotechnology Co. Ltd provided healthy adult male Sprague-Dawley rats (6–8 weeks, 180–220 g). The Ethics Committee of HBUCM approved the experiment. All rats were kept in a 12 h light/dark cycle with a room temperature of 24 ± 2 °C and a relative air moisture of 55 ± 15%. Protocol of animal experiments in vivo All rats were randomly allocated into five groups (n = 8): the control group, the model group, the melatonin group(100 mg/kg, MT), the low-dose DS group (1.35 g/kg, DSL), and the high-dose DS group(2.70 g/kg, DSH). The dosage of DS was calculated and expressed based on the weight of the crude drug material rather than the extracted components. The human and animal body surface area conversion method was utilized to calculate the administered dosage of DS and has been validated in our previous studies^[58]8. Following 7 days of acclimation feeding, all groups of rats, except the control group, were subjected to 21 days of modeling using a modified multi-platform aqueous sleep deprivation method. During the modeling period, the animals in each group were administered the respective medications via gavage once a day. The behavioral test was carried out in the last week. At the end of the experiment, rats were starved overnight but hydrated ad libitum, then anesthetized by intraperitoneal injection of sodium pentobarbital (40 mg/kg) and subsequently executed(Fig. [59]2A). All animal experiments were approved by the Ethics Committee of HBUCM (Permit code: HUCMS210726227631713255) and performed according to the relevant guidelines and regulations. All methods are reported in accordance with ARRIVE guidelines ([60]https://arriveguidelines.org). Fig. 2. [61]Fig. 2 [62]Open in a new tab DS prevents cognitive impairment, weight gain, hippocampus nerve degeneration, and serum inflammation in sleep deprivation-induced rats. (A) Flowchart of animal experiments. (B) Change from initial body weight. (C) Escape latency. (D) Time of crossing the platform. (E) Typical swimming trajectory. (F) Typical H&E staining of the hippocampus in the DG region. (G-I)The level of IL-1β, IL-6, and TNF-α in serum. The data are expressed as the means ± SEM. ^# p < 0.05, ^### p < 0.001 vs. control group; ^* p < 0.05, ^** p < 0.01, ^*** p < 0.001 vs. model group. Behavioral tests The spatial learning and memory abilities of rats were evaluated using the Morris water maze (MWM) experiment. As in our previous study^[63]4 the MWM consists of two phases, the navigation test and the spatial probe test, in which the animals’ swimming trajectories, escape latency, and the time of crossing the platform were recorded. Hematoxylin and Eosin staining Brain samples from three randomly selected animals in each group were fixed with 4% paraformaldehyde and then embedded in paraffin. Afterward, the wax blocks were cut into thin pieces about 4 μm thick, and stained with hematoxylin and eosin (H&E). Subsequently, the stained sections were placed under a light microscope for observation. Histopathological Scoring Criteria^[64]18,[65]19: Score 0: Lamina structure intact with distinct layering. Cells are densely packed. Neurons exhibit large somata, regular morphology, well-organized structure, and optically lucent cytoplasm. Score 1: Lamina structure partially disrupted or incomplete, with indistinct layering. Cellular packing demonstrates wider intercellular spacing. Some neurons display smaller somata, slightly irregular morphology, and darker cytoplasmic staining. Score 2: Lamina structure markedly disrupted or absent, lacking distinct layering. Cellular packing shows widest intercellular spacing. Numerous neurons exhibit small somata, irregular morphology, and dark cytoplasmic staining. Enzyme-linked immunosorbent assay(ELISA) Cytokines (IL-1β, IL-6, and TNF-α) are detected according to the corresponding kit instructions, and standard curves are used to determine cytokine concentrations. Serum metabolomics analysis Preparation of internal standard solution Precisely weigh about 2 mg of sofonadine standard and dilute to 500ng·mL^−1 by dissolving in methanol solution. Preparation of serum samples Pipette 40 µL of serum sample, then add 160 µL of acetonitrile solution and 50 µL of internal standard solution. After vortex mixing and centrifugation, take roughly 150 µL of supernatant in the injection bottle and store at 4 ℃. Quality Control(QC) sample Preparation Combine 10 µL of each serum sample, vortex mix, and store at 4 ℃. Chromatographic settings The gradient conditions are presented in Table [66]2. Waters ACQUITY BEH C18 column (100 mm×2.1 mm i.d., 1.7 μm), temperature: 35 °C, injection volume: 2 µL, flow rate: 0.3mL·min^−1. Table 2. Gradient conditions. Time(min) A(0.1% formic acid aqueous solution) B(methanol) 0 90% 10% 15 5% 95% 20 5% 95% 21 90% 10% 25 90% 10% [67]Open in a new tab Mass spectrometry conditions Mass spectrometric examinations were performed using a Waters Xevo G2-XS QTOF system with an electrospray ionization source (Waters, Massachusetts, USA). Positive ion electrospray was used to collect data during the sensitivity analysis mode. Metabolomics analysis was performed using positive ion mode. The structure of metabolites was established using positive ion modes. The mass number scanning range was: m/z 100–1200 Da; enkephalin (m/z 556.2771) tuning solution was used as the real-time correction solution; electrospray voltage: 3000 V; ion source temperature and desolvation temperature were 100 °C and 500 °C, respectively; air curtain gas and desolvation gas flow rates were 50 L/h and 500 L/h, respectively; the dissociation energies of the first-order and second-order collisions were 20 eV Ho 30 eV; collision energy fluctuation ± 10 eV. Identification of differential metabolites and metabolic pathway enrichment analysis The raw chromatographic data was analyzed using Masslynx, while SIMCA-p 14.1 (Umetrics, Umea, Sweden) was utilized for multivariate statistical analyses such as Principal Component Analysis (PCA), Partial Least Squares Discrimination Analysis (PLS-DA), and Orthogonal Partial Least Squares Discrimination Analysis (OPLS-DA). VIP > 1.0 and p < 0.05 were used as screening criteria to discover possible biomarkers. The human Metabolome Database (HMDB, [68]https://hmdb.ca/), Metlin (https://metlin.scripps.edu/landing_page.php? pgcontent=mainPage), and local databases were utilized to identify various metabolites. Metabolic pathway enrichment analysis was performed using the Kyoto Encyclopedia of Genes and Genomes Database (KEGG, [69]http://www.genome.jp/kegg/) and Metaboanalyst 6.0 ([70]https://www.metaboanalyst.ca/). Network analysis Acquisition of drug targets Based on the chemical composition of DS determined by UPLC-QTOF/MS, TCMSP ([71]https://old.tcmsp-e.com/tcmsp.php), SEA ([72]https://sea.bkslab.org/), and Swiss ([73]http://www.swisstargetprediction.ch/) were utilized to obtain the therapeutic targets of DS extracts. Acquisition of disease targets Drugbank ([74]https://go.drugbank.com), Genecards ([75]https://www.genecards.org), and DisGeNET ([76]https://disgenet.com) were used to obtain disease targets with the keywords ‘Sleep disorders’ and ‘Cognitive impairment’ respectively. The construction of the protein-protein interaction (PPI) network The intersection of the aforementioned two types of targets was taken to obtain the therapeutic targets of DS for SD-induced cognitive impairment. Following that, STRING ([77]https://cn.string-db.org) was used to analyze the inter-network between targets. Parameters were adjusted to High confidence (0.7), free nodes were removed and subsequently Cytoscape (version 3.10.2) ([78]https://cytoscape.org/) was used for further visualization adjustments. GO and KEGG enrichment analysis Metascape([79]https://metascape.org/)was used for Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes Genomes (KEGG) pathway analysis. Molecular Docking The AutoDock Vina-based CB-dock2 platform ([80]https://cadd.labshare.cn/) was used to perform molecular docking^[81]20,[82]21. Discovery Studio 2024 was used to visualize the molecular docking results in terms of docking details and 2D display. Smaller Vina scores indicate better binding results^[83]22. The 2D sdf structure files of the active compounds were obtained from PubChem ([84]https://pubchem.ncbi.nlm.nih.gov/), and the 3D structure of the target protein (PIK3CA:7PG5) was obtained from the PDB database ([85]https://www.rcsb.org/). Quantitative real-time polymerase chain reaction (qPCR) As previously published^[86]8 qPCR was performed according to the manufacturer’s instructions. mRNA relative expression was normalized with β-actin for further analysis. The primer sequences are described in Table [87]3. Table 3. Primer sequences for qPCR. Gene Forward primer Reverse primer PIK3CA GTGTGTGGCTGTGACGAGTA AATGCGCCTGGAGTAGGATG β-actin GAAGATCAAGATCATTGCTCC TACTCCTGCTTGCTGATCCA [88]Open in a new tab Statistical analysis The raw chromatographic data was analyzed using Masslynx, while the multivariate statistical analysis was performed by SIMCA-p. All results were expressed as mean ± SEM. The data was analyzed using Graphpad Pism 9.5. The nonparametric Kruskal-Wallis test and Mann-Whitney U post hoc test were used to compare count data between groups, whereas one-way analysis of variance (ANOVA) and the least significant difference (LSD) post hoc test were applied to metrological data. p < 0.05 were deemed significant. Results Chemical profiling of DS extract The main components of DS extract were characterized by UPLC-QTOF-MS/MS, and 32 compounds were identified^[89]8,[90]23,[91]24. The results are shown in Fig. [92]1, Fig. [93]S1-S2 and Table [94]S1. Fig. 1. [95]Fig. 1 [96]Open in a new tab Structures of 32 DS compounds. DS improves Spatial recognition and learning memory and mitigates weight gain in sleep deprivation-induced rats At the beginning of the experiment, there was no significant difference in the body weight of rats in each group. At the end of the experiment, the rats in the control group exhibited better symptoms, while the rats in the model group were in worse shape, characterized by disturbed mental states, lusterless fur, and higher food consumption. It is worth mentioning that the body weight of rats in the model group was significantly higher (p < 0.05) compared with the control group, while DS therapy helped SD rats return to near-normal levels (Fig. [97]2B). Impairment of spatial learning and memory functions is one of the serious complications associated with SD^[98]8,[99]25. As a classic experimental method to assess cognitive function in rodents^[100]4 the MWM was used as a behavioral test in this study. Among them, the navigation test was used to evaluate the learning ability of rats (Fig. [101]2C), and the spatial probe test was employed to detect the memory ability of animals (Fig. [102]2D). Figure [103]2E demonstrated the representative trajectories of animals in each group. After three days of training, rats in the model group had significantly longer escape latency compared to the control group (p < 0.05). On day 5, the escape latency of the DSL, DSH, and MT groups differed substantially from that of the model group (p < 0.05, p < 0.01, respectively). In addition, the times of crossing the platform was significantly lower in the model group than in the control group, suggesting poor memory capacity (p < 0.001), whereas the times of crossing the platform was significantly greater in the DSL, DSH, and MT groups than in the model group (ns, p < 0.05, p < 0.05). These results suggest that DS improves SD-induced cognitive deficits and that a dose of 2.70 g/kg is more effective than 1.35 g/kg. DS prevents hippocampus nerve degeneration and serum inflammation in sleep deprivation-induced rats Behavioral tests have shown that DS helps to reduce cognitive impairment caused by SD. The hippocampus, a critical brain area that regulates higher neural functions, has a unique role in memory formation and consolidation^[104]26. HE staining was utilized to evaluate hippocampal neurological impairment in rats (Fig. [105]2F). Compared to the control group, the model group’s hippocampus displayed apparent pathological damage, such as deeper staining and local nuclear consolidation. Neuronal damage was decreased following DS or MT therapy, as were neuronal staining and nuclear deformation. Pathological scores were statistically significantly higher in the model group compared to the control group (p < 0.001). The DSH and MT groups had significantly lower scores compared to the model group (p < 0.01, p < 0.05). DSL scores were reduced but not significantly different compared to the model group (ns)(Fig. S3). ELISA experiments were used to determine inflammation in rats. Serum IL-1β (Fig. [106]2G), IL-6 (Fig. [107]2H), and TNF-α (Fig. [108]2I) levels were significantly higher in the model group compared to the control group (p < 0.001, p < 0.001, p < 0.001). DSL, DSH, and MT significantly reduced serum IL-1β, IL-6, and TNF-α levels compared to the model group (p < 0.01 or 0.001). The above results indicate that DS inhibits hippocampal neuron degeneration and serum inflammation in SD-induced rats in a dose-dependent manner. DS regulates serum metabolic disturbances in sleep deprivation-induced rats To further explore the mechanisms by which DS improves cognitive impairment, UPLC-QTOF/MS was used to evaluate serum metabolic profiles in rats. To visually illustrate the dynamic changes in metabolites, this study employed volcano plots for visualization analysis, with the model group serving as the reference for intergroup comparisons. The results revealed that, compared to the model group, the control group exhibited significant upregulation of 270 metabolites and significant downregulation of 33 metabolites (Fig. [109]3A); similarly, the DSH group showed significant upregulation of 122 metabolites and significant downregulation of 38 metabolites compared to the model group (Fig. [110]3B). Fig. 3. [111]Fig. 3 [112]Open in a new tab Multivariate statistical analysis of non-target metabolomics. (A)Volcano plot between the control and model groups. (B) Volcano plot between the DSH and model groups. (C) Principal component analysis(The PCA score plots). (D)The PLS-DA score plots. (E) The OPLS-DA score plots between the control and model groups. (F) The OPLS-DA score plots between the DSH and model groups. (G)The validation plots from the 999-permutation tests of the PLS-DA model. (H) The validation plots from the 999-permutation tests of the OPLS-DA model between the control and model group. (I) The validation plots from the 999-permutation tests of the OPLS-DA model the DSH and model groups. Multivariate statistical analysis and identification of potential biomarkers Further unsupervised PCA score plots showed that serum QC samples were tightly clustered (Fig. [113]3C), which indicates that the instrumental method is stable and reliable. The PLS-DA model was constructed to evaluate intra-group and inter-group variations in overall metabolites in each group of rats (Fig. [114]3D). The results showed that the samples of each group were well aggregated, and the control and model groups were completely separated, while the DSH group tended to move closer to the control group, suggesting that SD caused metabolic disorders and DS therapy was able to restore aberrant metabolites. The 999 permutation test was used to evaluate the reliability of the PLS-DA model (R^2 = 0.989, Q^2 = 0.816), and all the predicted values were less than the target values, indicating that the PLS-DA model was stable and reliable(Fig. [115]3G). The OPLS-DA model was suitable for the comparison between the two groups for the screening of differential metabolites. The control group showed significant separation from the model group (Fig. [116]3E) and the model group from the DSH group (Fig. [117]3F). Similarly, the 999 permutation test was used to evaluate the reliability of the two OPLS-DA models, (R^2 = 0.997, Q^2 = 0.916; R^2 = 0.999, Q^2 = 0.850), and the intercept of the Q^2 regression line was less than 0, indicating that the OPLS-DA models were not overfitted and could be used to screen for differential metabolites(Fig. [118]3H-I). Metabolites that satisfied screening criteria of VIP > 1.0 and p < 0.05 were identified as possible biomarkers. Matching and comparing online databases (HMDB and METLIN) revealed 26 distinct metabolites (Table [119]4). The peak areas of these various metabolites were measured (Fig. [120]4). In the model group, 5 metabolites were up-regulated and 21 metabolites were down-regulated when compared to the control group. In the DSH group, 22 metabolites were up-regulated and 4 metabolites were down-regulated when compared with the model group. This indicates that DS therapy corrected SD-induced metabolic abnormalities. Next, a hierarchical clustering heat map analysis was utilized to show the content of each differential metabolite (Fig. [121]5A), with various colors representing different concentration levels. The results indicated that the control and DSH groups were more grouped together, while the model group differed more from them. The foregoing results indicate that the modeling of SD-induced cognitive impairment was successful, and the model had serum metabolic problems, whereas DS therapy can restore aberrant metabolite levels. Table 4. Differential metabolites. NO. Metabolites VIP RT (min) Formula Detected Mass (m/z) △ppm Change trend C/M D/M C1 gamma-Glutamylglutamic acid 1.17 3.3166 C[10]H[16]N[2]O[7] 277.1053 6.1349 ↑*** ↑*** C2 Phenylacetylglycine 1.21 4.5602 C[10]H[11]NO[3] 194.0813 −2.0610 ↑** ↑* C3 Tyrosine methylester 1.27 5.7277 C[10]H[13]NO[3] 196.0940 −17.3386 ↑*** ↑* C4 Suberoyl-L-carnitine 1.44 7.4230 C[15]H[27]NO[6] 318.1960 13.5137 ↑*** ↑*** C5 Tetrahydrofolic acid 1.19 7.7629 C[19]H[23]N[7]O[6] 446.1810 4.9307 ↑*** ↑*** C6 3-Sulfodeoxycholic acid 1.08 8.3605 C[23]H[38]O[7]S 459.2480 13.9358 ↑*** ↑*** C7 PIP(38:7-OH) 1.53 8.5202 C[47]H[78]O[17]P[2] 977.4927 13.8108 ↑*** ↑*** C8 7-Sulfocholic acid 1.28 8.5267 C[24]H[40]O[8]S 489.2488 −6.9494 ↑*** ↑*** C9 PI(40:9-2OH) 1.56 8.6675 C[49]H[77]O[15]P 937.4980 −10.4534 ↑*** ↑*** C10 PI(42:9-2OH) 1.42 8.8304 C[48]H[86]O[15]P[2] 965.5502 −1.8642 ↑*** ↑*** C11 LysoPA(20:5/0:0) 1.08 9.1133 C[23]H[37]O[7]P 457.2345 −2.1871 ↑*** ↑*** C12 (2R,3Z)-Phycocyanobilin 1.39 9.8158 C[33]H[38]N[4]O[6] 587.2885 2.5541 ↑*** ↑*** C13 PA(2:0/18:1-O) 1.22 10.2022 C[23]H[41]O[9]P 493.2610 8.9202 ↑*** ↑** C14 Maltotriose 1.12 12.2756 C[18]H[32]O[16] 505.1846 15.2420 ↑*** ↑*** C15 Oleoyl glycine 1.41 12.5998 C[20]H[37]NO[3] 340.2825 −7.9346 ↓** ↓* C16 Phosphatidylserine 1.13 12.7494 C[13]H[24]NO[10]P 386.1208 −2.0719 ↑*** ↑*** C17 L-Fucose 1.07 12.7512 C[6]H[12]O[5] 165.0762 −0.6058 ↑*** ↑*** C18 Disialyllactose 1.44 12.7530 C[34]H[56]N[2]O[27] 925.3256 11.5635 ↑*** ↑*** C19 Cortisol 21-sulfate 1.33 12.7546 C[21]H[30]O[8]S 443.1714 −5.8668 ↓** ↑ C20 Enkephalin L 1.11 12.9995 C[28]H[37]N[5]O[7] 556.2735 −6.4716 ↑*** ↑ C21 Glutamylphenylalanine 1.03 13.0097 C[14]H[18]N[2]O[5] 295.1322 9.4873 ↓* ↓** C22 Glycerol tripropanoate 1.21 13.7769 C[12]H[20]O[6] 261.1318 −7.6590 ↓* ↓** C23 Dodecanamide 1.30 14.3954 C[12]H[25]NO 200.2016 0.9990 ↓** ↓* C24 12-Oxo-20-trihydroxy-leukotriene B4 1.57 14.4014 C[20]H[30]O[7] 383.2052 −4.6972 ↑*** ↑*** C25 LysoPA(13:0/0:0) 1.40 14.6505 C[16]H[33]O[7]P 369.2040 −0.5417 ↑** ↑** C26 SM(18:2/20:5-3OH) 1.35 14.7958 C[43]H[75]N[2]O[9]P 795.5440 19.1064 ↑*** ↑*** [122]Open in a new tab C, Control group; M, Model group; D, DSH group. Fig. 4. [123]Fig. 4 [124]Open in a new tab Peak areas of 26 serum metabolites. Fig. 5. [125]Fig. 5 [126]Open in a new tab Serum metabolic pathway analysis. (A) Heatmap and cluster analysis of 26 metabolites in serum samples from three groups of rats. (B) Analysis of potential metabolic pathways using MetaboAnalyst 6.0. (a: Glycerophospholipid metabolism; b: Glycerolipid metabolism; c: Phosphatidylinositol signaling system; d: Inositol phosphate metabolism; e: One carbon pool by folate; f: Amino sugar and nucleotide sugar metabolism; g: Arachidonic acid metabolism) (C) Metabolic networks of DS in regulating SD-induced rats. Metabolites in red and blue reflected levels that rose and reduced in serum regulated by DS, respectively. Metabolic pathway enrichment analysis To further explore the metabolic pathways in which DS plays a regulatory role, we imported 26 identified differential metabolites into Metaboanalyst 6.0 and performed metabolic pathway analysis with the KEGG database. The results revealed that the metabolic pathways mainly included Glycerophospholipid metabolism, Glycerolipid metabolism, Phosphatidylinositol signaling system, Inositol phosphate metabolism, One carbon pool by folate, Amino sugar, and nucleotide sugar metabolism, Arachidonic acid metabolism (Fig. [127]5B). Among them, the metabolic pathways with Impact > 0.1 were considered the most influential^[128]17 mainly involving Glycerophospholipid metabolism, Glycerolipid metabolism, Phosphatidylinositol signaling system, One carbon pool by folate. The metabolic network was constructed according to the KEGG database (Fig. [129]5C). Network analysis of DS on cognitive impairment due to sleep disorders Network pharmacology, a method for elucidating potential drug targets and pathway analysis, was used in this study to explore the mechanism of DS to ameliorate cognitive impairment in SD rats. Our previous study analyzed the components of DS extracts (Table [130]S1), and as a basis, we obtained 748 targets of DS for the treatment of the disease (Fig. [131]6A). Moreover, we obtained 145 potential targets for DS to ameliorate SD-induced cognitive impairment (Fig. [132]6B). To further explore the mechanism of DS therapy, we constructed the PPI network of core targets by Cytoscape software (Fig. [133]6D), with targets such as TNF, IL6, NFKB1, and PIK3CA occupying important positions in the network and potentially being significant targets for DS function. In addition, enrichment analysis was performed on the top 20 targets in the Degree rankings of the PPI network.GO analysis included Biological Processes (BP), Cellular Components (CC) and Molecular Functions (MF) (Fig. [134]6C). The results showed that these potential targets were mainly located in the cell-substrate junction, RNA polymerase II transcription regulator complex and transcription regulator complex, regulating kinase binding, transcription coregulator binding and cytokine activity, and are involved in cellular response to cytokine stimulus, cellular response to lipid, positive regulation of phosphorus metabolic process and other biological processes. KEGG Sanger bubble map showed that DS may have a crucial role in modulating the MAPK signaling pathway, and the Toll-like receptor signaling pathway(Fig. [135]6E)^[136]27. These pathways are involved in inflammatory regulation and metabolic regulation^[137]28which are closely associated with the development of AD, lipid, and atherosclerosis, implying that DS may exert its efficacy on SD-induced cognitive impairment through these pathways. Fig. 6. [138]Fig. 6 [139]Open in a new tab Network pharmacological analysis of DS on cognitive impairment caused by sleep disturbances^[140]29. (A) Drug-compound-target network. (B) Venn diagram of DS, sleep disorders, and cognitive impairment intersection targets. (C) GO enrichment analysis histograms for BPs, CCs, and MFs. (D) PPI network of potential core targets. (E) Sanger bubble map for KEGG enrichment analysis. Integrating network analysis and serum metabolomics to reveal DS modulation in SD rats To fully comprehend the mechanism of DS to improve cognitive impairment, we conducted a combined analysis of metabolomics and network analysis, as well as the construction of a core action network. First, The identified differential metabolites were imported into Cytoscape’s MetScape plug-in, which was then used to create the Compound-Reaction-Enzyme-Gene network (Fig. [141]7A). We used the network to match 145 metabolite targets. Subsequently, We matched these targets to the 145 targets identified by network analysis to determine the primary target of PIK3CA (Fig. [142]7C). Further investigation revealed that PIK3CA was mainly associated with phosphatidylinositol phosphate metabolism(Fig. [143]7B). Fig. 7. [144]Fig. 7 [145]Open in a new tab Combined analysis of network analysis and serum metabolomics. (A) Potential Compound-Reaction-Enzyme-Gene networks for DS-regulated metabolism. (B) Network of Phosphatidylinositol phosphate metabolism. (C) Venn diagram for the intersection of key targets in network analysis and serum metabolomics. (D) Vina score. (E) Relative mRNA expression of PIK3CA. Molecular Docking and validation experiments for key targets of DS regulation Moreover, we validated the core targets obtained from the co-analysis by molecular docking and qPCR experiments. First, we queried the network analysis results (Fig. [146]6A) and discovered that eight DS components—dihydronortanshinone, przewaquinone C, przewaquinone A, tetrahydrotanshinone, Cryptotanshinone, Tanshinone IIA, Miltirone, and Salviolone—might have an effect on PIK3CA. Following that, molecular docking was performed using the CB-dock2 platform based on AutoDock Vina^[147]20,[148]21. The molecular docking results, including docking details and 2D display, were visualized using Discovery Studio 2024 (Fig. [149]8). The Vina scores of the above eight components binding to PIK3CA targets were all less than − 8.8 (Fig. [150]7D), indicating that the core components of DS had good binding effect to PIK3CA^[151]22. qPCR results showed that PIK3CA mRNA was significantly higher in the model group compared with the control group (p < 0.001). Following DS therapy, PIK3CA mRNA was significantly reduced in the DSH group compared with the model group (p < 0.001)(Fig. [152]7E). These results suggest that PIK3CA plays a key role in regulating the metabolic abnormalities in the SD-induced cognitive impairment model with DS. Fig. 8. [153]Fig. 8 [154]Open in a new tab Diagram of the molecular docking patterns of the 8 components of DS with PIK3CA.(A-H) Molecular docking pattern of PIK3CA with dihydronortanshinone (A), przewaquinone C (B), przewaquinone A (C), tetrahydrotanshinone (D), Cryptotanshinone (E), Tanshinone IIA (F), Miltirone (G), and Salviolone (H), respectively, including local docking details and 2D model. Discussion UPLC-QTOF-MS/MS identified 32 diterpenoids in DS extract, all belonging to tanshinones, with pharmacological activities such as neuroprotective, anti-inflammatory and cardiovascular protection^[155]8,[156]23,[157]24. Our research showed that DS corrected spatial learning and memory deficits caused by SD. In addition, DS attenuated hippocampal nerve damage, serum inflammation, and metabolic disturbances in SD rats. In this study, 2 different doses of DS were used, and 2.70 g/kg had a better-moderating effect on behavioral assessment experiments and serum inflammation than 1.35 g/kg, suggesting that 2.70 g/kg of DS had a superior protective effect on cognitive performance. This is the first study to find that DS improves cognitive function through metabolic modulation, in which PIK3CA is one of the key targets for its action. There is consensus that SD has several harmful impacts on the body, especially its severe impairment of cognitive function^[158]30,[159]31. The hippocampus, a major brain area in the regulation of cognition and emotion, is essential for the consolidation of learning memory and cognitive function^[160]4,[161]26. Disruptions in the sleep-wake cycle might cause systemic chronic inflammation, which increases hippocampus neuronal death^[162]32. Microglia in the hippocampus respond to inflammation by producing additional inflammatory substances, and this vicious cycle further causes cognitive dysfunction^[163]33. DS is a medicinal plant that is widely consumed as a tea beverage or supplement^[164]10. Not only has DS been historically used to treat insomnia-related disorders^[165]9 but current pharmacological research has also shown that the lipid-soluble ingredients have neuroprotective properties^[166]13. The ingredients of the DS extracts in this experiment were identified, including tanshinones such as Tanshinone IIA, which are natural medicinal components with great activity^[167]14,[168]34,[169]35. The multi-platform aquatic environment method is classic for developing SD models that is simple, successful, and has been widely used in studies related to rodent SD models^[170]3,[171]36–[172]38. Although our previous studies have shown that DS ameliorates SD-induced systemic inflammation and cognitive deficits by modulating the TLR4/MyD88/NF-κB signaling pathway mediated by intestinal flora^[173]8 and some clinical studies have pointed to a possible causal relationship between insomnia and coronary artery disease^[174]39 however, the SD negative effects on metabolic homeostasis and the metabolic regulation of serum by DS are poorly understood. For these reasons, the present study adopted a combined UPLC-QTOF/MS metabolomics and network analysis strategy to explore the mechanisms by which DS exerts metabolic modulation to improve cognitive function in SD rats. Metabolomics directly focuses on metabolite changes in organisms, and this analysis integrates information from gene, transcript and protein level changes, as well as post-translational modifications, which is a distinct advantage not found in other “genomics technologies”^[175]40. Metabolomics reveals the differential expression of metabolites in the body and is used to characterize the relationship between physiological and pathological processes, as well as to predict the body’s response to drugs^[176]15. Clinical studies have shown that sleep restriction negatively affects the rhythm of plasma metabolites in patients with insomnia, with 27 metabolites being significantly elevated during SD, including 3 sphingolipids and 13 glycerophospholipids, which is a clear link to metabolically disturbed obesity and cardiovascular disease^[177]40. Furthermore, a study in mice reported that 6 h of SD affected glycolysis and lipid metabolism in the body^[178]41. Serum metabolic disorders are present in animal models of SD or patients with insomnia, and modulation of serum metabolic profiles may be a potential mechanism by which DS improves cognitive function. The present study identified 26 metabolites that can be significantly modulated by DS, and further enrichment analysis revealed that eight of them may be key, mainly involved in Glycerophospholipid metabolism, Glycerolipid metabolism, Phosphatidylinositol signaling system, and One carbon pool by folate that are closely related to phospholipid metabolism^[179]42^,^[180]43 It has been reported that dietary phospholipid supplementation protects against phospholipid compositional abnormalities in several brain regions in a mouse model of n-3 polyunsaturated fatty acid deficiency, which is associated with improved learning and memory functions^[181]44. Previous studies have shown a significant increase in hepatic triglyceride content in C57BL/6J male mice subjected to SD, which is mediated by upregulation of hepatic lipogenic enzymes^[182]45. The results on body weight in our study are consistent with the report, which partly explains that DS can reverse SD-induced weight gain in rats. 12-Oxo-20-trihydroxy-leukotriene B4 is a metabolite generated upon lipid oxidation of leukotriene B4, a pro-inflammatory lipid mediator that exerts multiple physiological activities during inflammation and host defense^[183]46that can induce leukocyte adhesion. Inflammation-induced reduction in leukotriene B4 omega-oxidation^[184]47 may be responsible for the reduced serum 12-Oxo-20-trihydroxy-leukotriene B4 levels in SD rats in the present study. Phosphatidylinositol (PI) is a major cell membrane component, an acyl derivative resulting from esterification of the phosphate portion of phosphatidic acid with inositol. Early studies have shown that neurotransmitter receptor activity can be regulated by phosphatidylinositol, while phosphatidylinositol is a phospholipid derivative of PI^[185]44. To be precise, acetylcholine can be synthesized from phosphatidylcholine, phosphatidylethanolamine, or sphingomyelin. Given the role of acetylcholine and cholinergic neurotransmission in brain development and cognition^[186]48dietary supplements containing phospholipids may provide additional substrates for acetylcholine synthesis. In addition, phosphatidylinositol is essential for synaptic vesicle transport in the presynaptic compartment, as well as receptor endocytosis and cytosolization on the postsynaptic side^[187]49. Abnormal phosphatidylinositol metabolism has been found to be involved in the adipose tissue-brain axis intervening in the cognitive decline in obese mice^[188]50. The reduced serum levels of PI(42:9-2OH) and PI(40:9-2OH) in the SD rats in the present study somewhat explains the close association between SD-induced weight gain and impaired cognitive function. Research indicates that Aβ-induced PIP depletion causes slower release at hippocampal excitatory synapses, reduces the likelihood of release at Schaffer collateral-hippocampal CA1 pyramidal neurons in AD model mice, and improves spatial learning and memory in APP/PS1 mice^[189]51. Since Aβ-induced reduction in synaptic release rate precedes synaptic and neuronal loss, controlling supplemental PIP levels may become an effective strategy to prevent AD development. These reports are consistent with the trend of PIP (38:7-OH) changes after DS treatment in the present study, which further validates our conclusions. Phosphatidic acid (PA) and Lyso-phosphatidic acid (LysoPA) are both glycerophospholipids, and they all have a glycerol backbone and a phosphoryl group, with the structural difference in the number of acyl groups.PA can be hydrolyzed to form LysoPA catalyzed by phospholipase A2. LysoPA is an active class of phospholipid signaling molecules that not only induces macrophage differentiation and polarization, regulates macrophage migration and infiltration^[190]52 but also participates in the homeostatic regulation of glutamatergic transmission and the excitability of cortical networks, thus playing a role in the regulation of feeding behavior and feeding habits, and thus exerts a modulatory role on feeding behavior and psychiatric disorders^[191]43. In addition, LysoPA induces pathogenic factors such as IL-1 and reactive oxygen species that directly promote the release of pro-inflammatory mediators from macrophages and regulate the development of neuroinflammatory diseases such as ischemic brain injury^[192]53. In our study, DS administration improved lowered serum PA (2:0/18:1-O) and LysoPA (20:5/0:0) levels in SD rats, indicating that enhanced cognitive function by DS is strongly associated to inflammatory management, which is consistent with our earlier results^[193]8. Phosphatidylserine accounts for 13–15% of phospholipids in the human cerebral cortex^[194]54 and its primary role is to increase nerve cell function, control nerve impulse transmission, and boost brain memory performance. For example, aged rats orally administered krill-derived phosphatidylserine for 7 days showed improved cognitive performance in the MWM test^[195]55. Similarly, human subjects receiving oral phosphatidylserine supplementation showed improved cognitive performance. EEG measurements showed that treatment with 200 mg of soy-based phosphatidylserine for 42 days resulted in greater relaxation in a clinical sample with significant mental stress^[196]56. Additionally, early clinical studies suggest that phosphatidylserine intake may have a compensatory effect on cognitive deficits associated with older adults and AD patients^[197]42 which is the first report to confirm that phospholipids may be an important candidate for improving cognitive performance in pathological and neurodegenerative diseases. This result was further validated in our study, where the reduction of phosphatidylserine in SD rats was reversed by DS treatment, suggesting that regression of phosphatidylserine improves cognitive performance. Glycerol tripropanoate is a triglyceride in which all three hydroxyl groups of glycerol undergo esterification with the propionic acid molecule Glycerol lipids. Elevated blood triglyceride levels release hunger signals to the brain, inducing harmful and persistent eating behaviors^[198]57. Moreover, hypertriglyceridemia, a typical dyslipidemia symptom of obesity, is closely associated with diseases such as metabolic syndrome and type II diabetes, and is a direct risk factor for inducing cognitive impairment^[199]58. Metabolomics results showed that DS significantly reduced SD-induced high serum levels of Glycerol tripropanoate. Metabolite disorders were normalized in SD rats treated with DS in the present study. Therefore, improving the metabolic microenvironment in SD individuals may be an effective strategy to ameliorate the decline in learning and memory abilities in insomnia patients. In conclusion, serum metabolomics results explain the mechanisms by which DS exerts metabolic regulation. However, it is clear that metabolomics cannot adequately explain the regulatory mechanisms, and therefore, network analysis is used for complement. Network pharmacology emphasizes the mechanism of interaction between multiple components of traditional Chinese medicine(TCM)and multiple targets of disease, which is consistent with the holistic concept of TCM^[200]15. Network analysis has unearthed 145 potential targets for DS to exert therapeutic effects, suggesting that DS may play an important role in regulating the MAPK signaling pathway and Toll-like receptor signaling pathway (Fig. [201]5E)^[202]27. MAPK signaling pathway is a classical pathway of lipid metabolism, which is relevant to the metabolomics results of this study. In addition, MAPK signaling is involved in the regulation of CNS myelin formation, which is associated with synapse formation and neuronal circuits essential for normal brain function^[203]59. It has also been reported that neuronal defects in p38α-MAPK reduce Aβ and phosphorylated tau proteins in the brains of AD mice, slowing down the development of cognitive impairment in APP/PS1 double-transgenic mice^[204]60. Toll-like receptors (TLRs) are an important class of pattern-recognition receptors, which are involved in the regulation of myelin formation in the CNS through the recognition of pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs) to activate immune responses^[205]4. The activation of TLRs initiates downstream signaling pathways mediated by the adapter MyD88, which in turn activates transcription factors such as NF-κB, MAPK, and IRF, which in turn regulate the expression of a variety of inflammation-associated genes including cytokines, chemokines, and adhesion molecules, thereby triggering an inflammatory response. Our earlier study has demonstrated that DS modulates the LPS-mediated TLR4/MyD88/NF-κB signaling pathway to ameliorate the development of cognitive impairment in SD rats^[206]8. Moreover, the network analysis predicted the KEGG pathway involved in inflammatory regulation and metabolic modulation^[207]28which is closely related to the development of AD, lipid and atherosclerosis, which is consistent with the cognitive impairment, serum inflammation, and metabolic disorders in SD rats in the present study. The above results suggest that the network analysis prediction results are informative. To further explore the underlying mechanism of DS, MetScape was used to synthesize the results of metabolomics and network analysis. Molecular docking is a common tool used to predict the binding sites of active compounds to target proteins^[208]61 and Wang combined network pharmacology and molecular docking to reveal that Daqinjiao decoction ameliorated neuronal cell damage and memory deficits in cerebral small vessel disease^[209]62. Our study revealed that all eight components of DS extracts had good binding capacity to PIK3CA, which is mainly involved in the regulation of Phosphatidylinositol phosphate metabolism. Current studies have reported the neuroprotective effects of these compounds against neuroinflammation. Among them, dihydronortanshinone is a potent anti-inflammatory agent that exerts protective effects in LPS-induced RAW264.7 macrophages, which is closely associated with the inhibition of ROS generation and preservation of mitochondrial function^[210]63. Przewaquinone A has been shown to induce protective autophagy, thereby contributing to its neuroprotective properties^[211]64. In vitro studies have confirmed that Cryptotanshinone suppresses AD-related tau hyperphosphorylation by modulating the PI3K/Akt/GSK3β signaling pathway^[212]65. Additionally, Cryptotanshinone alleviates the effects of Aβ on PSD95 and synaptophysin expression levels in HT22 cells, demonstrating neuroprotective activity. Tanshinone IIA, a widely studied natural compound, exhibits significant neuroprotective effects^[213]66. Research indicates that Tanshinone IIA improves spatial learning and memory deficits in APP/PS1 mice by inhibiting the RAGE/NF-κB signaling pathway^[214]14. Our previous study confirmed that Tanshinone IIA ameliorates cognitive impairment in SD rats by regulating the CNR1/PI3K/AKT pathway^[215]9. Miltirone, a phenanthraquinone, modulates the PI3K/Akt pathway and mitigates ROS-dependent neuronal apoptosis in Parkinson’s disease cell models^[216]67. Salviolone has been reported to suppress key malignant features of melanoma cells^[217]68demonstrating both anti-inflammatory and anti-tumor activities^[218]69. Currently, there are limited reports on the neuroprotective effects of przewaquinone C and tetrahydrotanshinone. Therefore, our future research will focus on investigating these potential key bioactive compounds for their protective roles in cognitive function. The protein encoded by the PIK3CA gene is the catalytic subunit α of phosphatidylinositol-3-kinase (PI3K). It has been reported that PIP2 is converted to PIP3 under the regulation of PIK3CA, which induces phosphorylation of AKT1 and subsequent activation of the NF-κB pathway, thereby regulating protein synthesis, cell proliferation, differentiation, autophagy and apoptosis^[219]15. Clinical studies have found that mutations in the PIK3CA gene cause a rare syndrome known as PIK3CA-Related Overgrowth Spectrum (PROS), which causes symptoms including fibrofatty hyperplasia or overgrowth, megalencephaly, cortical dysplasia, and epilepsy, which negatively affects an individual’s cognitive function^[220]70. In addition, the PI3K/Akt signaling pathway negatively affects synaptic plasticity by regulating the phosphorylation pattern of tau proteins, affecting Aβ clearance, and causing oxidative stress and overproduction of ROS^[221]71. Several researchers have also discovered hyperactivation of the PI3K/Akt signaling pathway in patients with AD or cognitive impairment^[222]72. The validation experiments in this study also found that the mRNA changes of PIK3CA showed a consistent trend with the above reports. Phosphatidylinositol phosphate metabolism plays an important role in cell signaling, metabolic regulation, and energy homeostasis. As mentioned earlier, PI and its derivative PIP are involved in the regulation of neurotransmitter receptor activity including acetylcholine^[223]44^,^[224]48, synaptic signaling^[225]49 and neuronal energy supply^[226]51 they are involved in regulating neuronal survival and function, thereby directly affecting cognitive function. In the present study, an untargeted serum metabolomics study of integrated network analysis focused the regulatory mechanisms of DS on a large number of phospholipid metabolites, necessitating more research using targeted lipid metabolomics techniques. Likewise, reports have indicated that the gut microbiota also produces phospholipid metabolites. For example, some commensal bacteria from the phylum Mycobacterium are capable of synthesizing sphingolipids, which are critical for gut homeostasis, and their deficiency leads to increased intestinal inflammation^[227]44^,^[228]73. Our previous study also pointed out that DS ameliorates SD-induced cognitive deficits by modulating the LPS-TLR4 signaling pathway mediated by intestinal flora^[229]8 implying that gut-derived phospholipid metabolites may be a key mechanism by which DS exerts its regulatory effects and warrants further investigation. Conclusion In summary, UPLC-QTOF-MS/MS identified 32 diterpenoids in DS extract. The results indicated that DS medication for 21 days significantly improved SD -induced learning memory deficits in rats. Furthermore, DS was involved in serum metabolic regulation to improve homeostasis (Fig. [230]9). Our study provides new insights into the development of DS as a dietary supplement for treating SD with cognitive impairment. Fig. 9. [231]Fig. 9 [232]Open in a new tab The mechanism of DS against SD-induced cognitive impairment. Supplementary Information Below is the link to the electronic supplementary material. [233]Supplementary Material 1^ (716KB, docx) Acknowledgements