An integrated multi-omics study to identify dynamic molecular alterations associated with acute brain injury

Neuroimmune cells are rapidly transition from a quiescent into an activated state in response to 26 acute brain injury (ABI) threats, but the dynamic molecular alterations are partially understood. Until recently, brain scientists were ineffectual to explore the molecular alterations in human, 28 owing to the obstacles for ABI related brain sample acquisition. Here, we integrated the dynamics 29 of multi-omics datasets in four ABI mice models. Transcriptomics revealed diversification of 30 thermogenesis, synaptic, and neuroinflammatory genes for ABI at the early phase (12H). 31 Transcriptomics and proteomics combined analysis singled out 15 co-variation risk genes for ABIs. 32 Besides, lipid metabolite alteration reflected a discrepancy between permanent ischemic brain 33 injuries and transient ischemic brain injuries at the middle phase (24H). Together, our data 34 elucidate a potential therapeutic resource for ABIs.

generated as described in method. The detected proteins, metabolites, and lipids were summarized 117 in Supplementary Table 1. All biological replicates for different omics assays in this study have a 118 nice reproducibility (Extended Data Fig.2A,C). 119 120 Transcriptome Profile of ABI mice models 121 Previous studies have reported the dysregulation of gene transcription in ABI mice models [28][29][30] . 122 However, transcriptome characterization of ABI-derived brain tissue has been limited. To 123 systematically illustrate the transcription dynamics underlying ABIs, which were mainly caused 124 by the secondary injury progresses, we analyzed transcriptome data we generated from ABI mice 125 models at three stages (12H, 24H, and 72H) and identified the DEGs (Extended Data Fig.2B). The 126 normalized counts (see method) of DEGs could reflect the dynamics of gene transcription along 127 with the progression of ABIs. 128 By comparing gene transcription between ABI subtypes and control group at each stage, we 129 successfully identified 132 (Control vs SAH of 72H) to 3969 (Control vs TM of 72H) DEGs 130 (Extended Data Fig.1D), while all of them had significant up-regulation (log2FC > 1; P < 0.05) 131 (Extended Data Fig.1E). Together, for the number of DEGs, PM and TM were the most in any 132 stage, and SAH was the least (Extended Data Fig.1D-E). This result revealed that ischemic brain 133 injury (PM and TM) at each stage of the ABI process caused more gene transcription alterations 134 than hemorrhagic brain injury (SAH and TBI). In turn, hemorrhagic brain injury (SAH and TBI) 135 or ischemic brain injury (PM and TM) have similar gene transcription patterns respectively. If 136 cross-compared, SAH and ischemic brain injury (PM and TM) both have a large difference in 137 transcription level, but TBI and TM have very similar transcription patterns at the stages (12H and 138 24H), while TBI' transcription pattern was closer to PM at a late stage (72H) (Extended Data 139 Fig.2B).

140
Transcription patterns of all DEGs (combined DEGs for each ABI subtype and disease stages; 141 N=5632; Supplementary Table 2) were characterized into 6 clusters through the K-means 142 algorithm (see method) (Fig.1A). The consensus clustering results illustrated that the ABI subtypes 143 and secondary brain injury stages could be distinguished (Fig.1A). SAH and TBI shared the 144 highest transcription level at 12H compared with other ischemic brain injuries, which was also in 145 line with the feature that this group of genes was mainly enriched in the regulation of the blood 146 circulation pathway (cluster 1). 147 Note that the genes in cluster 2 were mostly characterized by synaptic and neuron functions, 148 and decreased transcription level in all ABI mice models, especially in PM and TM at 72H. This 149 indicated that after the occurrence of ABIs, the transcription of genes related to synapse and neuron 150 function was decreased, which may be related to neuronal dysfunction. Among them, PM and TM 151 may have the most serious neuronal dysfunction at this time point. Interestingly, in cluster 4 and 152 cluster 5, PM and TM have higher transcription levels and more DEGs (47.71% of all DEGs) than 153 other injuries at 72H. These genes were mainly enriched in pathways related to neuroinflammation, 154 such as cytokine, angiogenesis, cell adhesion, IL-6 production, innate immune response, ERK1/2 155 cascade, and reactive oxygen species metabolic pathway (P<0.05, Fisher's Exact Test). This 156 suggested that ischemic ABIs have a more serious neuroinflammatory response at the late stage of 157 the ABIs, which may be related to more severe damage to synapses and neurons in the brain tissues 158 (see cluster 2). It is worth noting that cluster4,5-related DEGs were also enriched in the "regulation 159 of vasculature development" pathway, suggesting that the progress of the ischemic ABIs have a 160 higher transcription of genes that regulate vascular development at the late stage, indicating the 161 angiogenesis and tissue repair activates which was also consistent with previous findings 31 . 162 Although most of DEGs of PM and TM shared similar transcription patterns in the late stage, there 163 were DEGs with opposed transcription patterns (cluster3 and cluster6), which were mainly 164 movement-related (such as cilium movement and microtubule bundle formation/movement, etc.) 165 and temperature adaptation and fat cell growth-related (such as regulation of fat cell differentiation,  166 adaptive thermogenesis, and cold-induced thermogenesis, etc.) genes (Fig.1A). Given that, genes 167 of the two clusters have the potential to be the markers to distinguish PM and TM, because of the 168 transcriptional differences only in PM and TM at 12H, but not in other ABIs.

169
We also investigated the transcription of 390 genes (cluster 6 in Fig.1A) in PM and TM, which 170 were further classified into two categories (Fig.1B). Based on this finding, these genes have shown 171 a significantly different transcription pattern in PM and TM at 12H (class 1: P=3.15e-07; class 2: 172 P=0.01; Fig.1D). Compared with the control group, DEGs, which are mainly related to 173 temperature regulation, were only highly transcribed in TM and PM at 12H respectively (Fig.1C). 174 To match the findings in ABI mice models with the characteristics of clinical cases, we further 175 performed a correlation analysis between NIH Stroke Scale/Score (NIHSS) and highest body 176 temperature and of ischemic stroke patients during the first 72H after admission. There was a linear 177 growth of body temperature with increasing NIHSS scores in both permanent acute ischemic 178 stroke (PAIS) and transient acute ischemic stroke (TAIS) patients (TAIS: R = 0.554, P < 0.001; 179 PAIS: R = 0.304, P < 0.001, Fig.1E). This finding revealed that thrombolysis or endovascular 180 thrombectomy for ischemic stroke has differences in neurological outcomes, and according to 181 previous literature, because of the higher body temperatures during both the intra-ischemic and 182 post-ischemic phases were associated with poorer clinical outcomes 32 , so that our results further 183 provide potential molecular evidence for thrombolytic therapy/mechanical thrombectomy 184 combined with mild hypothermia therapy/maintenance of body temperature for ischemic stroke.

186
The proteomic analysis reveals an increased level of neurological related proteins at 24H 187 In the previous study of proteomics in ABI mice models, only the disease subtypes or progression 188 stages were considered 19,33 . In our dataset, the Partial Least Squares Discrimination Analysis (PLS-189 DA) of protein abundance of all ABI mice models illustrated the formation of distinct clusters of 190 the ABI subtypes and the excellent data repeatability of replicates (Extended Data Fig.2D; 191 Supplementary Table 3; see method). All ABI subtypes were distinguished well from each other 192 and shared dynamic protein abundance across each disease progression stage (Extended Data 193 Fig.2C-D).

194
The comparison between the control group and ABI stages could lead to identifying the 195 differential expressed proteins (DEPs) at each disease progression stage. The DEPs were 196 determined by the log2 ratio of each protein abundance level to the abundance of the control sample.

197
We applied t-test analysis on all candidates to screen for DEPs with statistical significance. We 198 finally identified, in total, 1201, 1520, and 1514 DEPs at 12H, 24H, and 72H of ABIs respectively 199 (Supplementary Table 1), and heatmaps were used to display them graphically to show the 200 difference in protein abundance between ABI subtypes at the same stage of disease progression 201 (Extended Data Fig.2 C).

202
The DEPs between ABI subtypes could be potentially useful for distinguishing the subtypes of 203 ABIs. We performed K-Means Cluster analysis (see method) based on the protein abundance of 204 ABI subtypes in the same secondary brain injury progression stage and finally obtained 4, 2, and 205 3 categories at 12H, 24H, and 72H respectively ( Fig As shown in Fig.2A, we could distinguish TM and TBI from SAH and PM by protein abundance 208 at 24H and found that these proteins mainly related to synapse structure and other related protein 209 location (category 2), while in SAH and PM, the proteins related to hydrolase activity and 210 exocytosis were highly translated (category 1) (Fig.2B). This result was significantly different 211 from the other two stages (12H and 72H) and was also different from the previous transcriptome 212 results in the decline of gene transcription levels related to synapse function in TM and PM at 72H 213 (Extended Data Fig.4D), indicating that these genes related to synapse function might share the 214 other regulatory mechanism. 215 Given symptoms appear suddenly or worsen over time following an ABI, especially within the 216 first 24H after the injury 34 , we focused on the biological process pathways of subtype-specific 217 DEPs at 24H for further analysis. The functions of neurons and synapses were closely related to 218 the prognosis of ABIs. Based on the result that the synapse-related functional pathways in category 219 2 of Fig.2B were highly enriched, we further extracted synapse-related proteins, and analyzed the 220 correlation between their transcription levels and protein abundance (see method), we found that 221 transcription and protein level of all ABI subtypes have a very obvious positive correlation on 222 disease progression at 24H (Fig.2C,). This result was also applicable to the enriched exocytosis-223 related functional pathways in category 1 of Extended Data Fig.3A.

224
To further investigate the key molecules related to synaptic function in each ABI subtype, we 225 analyzed the dynamic profile of synapse-related proteins (Fig.2D) and exocytosis-related proteins 226 (Extended Data Fig.3B) of all synapse-related molecules in each ABI subtype (compared with the 227 control group). The Ephb3, mediating developmental processes in the nervous system, was down-228 regulated in all ABI subtypes at 72H, indicating that it was a key molecule that controls synaptic 229 function in the late stage of ABIs, and its down-regulation may be closely related to impaired 230 synaptic function. Besides, Nrp1 was also down-regulated almost at 72H of ABIs (except TM), 231 indicating that it might also be related to the damage of synaptic function. Interestingly, the Dgkb, 232 which regulates neuron-specific morphological changes including neurite branching and neurite 233 spine formation, was mainly up-regulated in ABI subtypes at 72H (except PM), which could be 234 related to the repairment of synaptic function. It was worth noting that only Lrrtm2 was highly 235 expressed at 24H of TBI, which could be used as a marker gene of TBI. For SAH and TM, Slc8a3 236 was specifically up-regulated at 12H in SAH and down-regulated at 72H in TM, so it could be 237 used as a marker gene to distinguish these two ABI subtypes. Similarly, Adnp has significantly 238 different protein levels in hemorrhagic ABI (SAH and TBI) and ischemic ABI (PM and TM), 239 indicating that it could be used as a marker for these two types of ABI subtypes.

241
Transcriptomic and proteomic analysis identified risk factors for ABI mice models 242 Transcription and translation events should be dynamic during the progression of secondary 243 damage to ABIs. We considered the correlation analysis gene clusters and protein categories and 244 reveal key risk factors that played a major role in ABIs and subtype-specific molecules (Fig.3A). 245 Based on the gene clusters in Fig.1A, we found that in cluster1 and cluster2, there was no 246 obvious correlation aggregation in SAH; However, in TBI, it has a significant negative correlation 247 aggregation (median: -0.36). Also, PM and TM hold a similar positive correlation aggregation 248 (median: 0.23 and 0.26). This indicates that the disorder of transcription level and translation 249 abundance in the genes of cluster 1 reflected their dynamic function of blood circulation regulation. 250 We further found that TBI also has negative correlation aggregation in cilium movement, 251 microtubule bundle formation, and microtubule-based movement (cluster3). More interesting, 252 except the cluster6, the correlations of PM and TM were almost aggregated to the positive side. 253 PM has a positive correlation aggregation in cluster1, 2, 3, and 4, especially in cluster3 and cluster4, 254 but there was no obvious correlation aggregation in cluster5.

255
To further identify the key regulatory molecules in the progression of ABIs, we analyzed the 6 256 clusters. Taken the intersection for top 10 candidates of 5 screening algorithms by cytohubba 257 plugin in Cytoscape (Closeness, EPC, Degree, Radiality, and MNC), we got 5, 8, 3, 1, 7, and 6 258 hub genes in cluster1 to cluster6 respectively (Extended Data Fig.5A). The summary of these 30 259 hub genes was listed in Supplementary Table 4. To further explore the interaction of robust hub 260 genes, we constructed the PPI network by STRING database (Fig.3B). The hub genes in cluster 1 261 were mainly enriched in behavior and cytosolic calcium ion concentration pathways. The Oxt, 262 encodes a precursor protein that is processed to produce oxytocin and neurophysin I, involved in 263 cognition, adaptation, and regulation of water excretion and cardiovascular functions. The increase 264 in transcription level expression of Oxt could positively regulate the translation level of protein in 265 SAH and PM (Extended Data Fig.5B), but there might be post-translational in TM. The 266 transcription and translation level of Oxt at three stages of four ABI models had significant 267 characteristics, so this gene may be used as a marker gene in different progression stages of ABIs. 268 In cluster 2, these hub genes were highly related to synaptic transmission and regulation of 269 neuronal synaptic plasticity, which was also consistent with the results in Fig involved in the formation of the brain and multi-synaptic globular rings 35,36 . MKI67, the hub gene 278 of cluster 4, encodes a nuclear protein associated with cellular proliferation. In cluster 5 and cluster 279 6, hub genes were mainly enriched in pathways related to abiotic stimulus, cytokine, apoptotic and 280 inflammatory response. 281 Overall, the identification of risk factors (hub genes), and particularly of ABI subtype-specific 282 molecules underscores their potential roles in ABI progression.

284
Metabolic Characteristics of ABI 285 These dynamic transcriptomic and proteomic changes in ABIs promoted us to study alterations in 286 the metabolic process. For ABIs, probably together with astrocytic and microglial activation and 287 induction of a metabolic disorder state that can induce secondary injury events, offering several 288 potential clinical biomarkers and therapeutic targets 37 . We explored whether the alterations in 289 transcription and translation levels could modulate the metabolites. Because changes of genes in 290 transcription or translation levels can lead to broad metabolites alterations due to modulation of 291 metabolic pathways. 292 To manipulate the dynamic changes in metabolites at different stages of ABIs, we utilized LC-293 MS to perform metabolomic and lipidomic assays (Extended Data Fig.1A, Supplementary Table  294 5). Further abundance profile indicated high repeatability of lipidomic and metabolomic datasets 295 (Extended Data Fig.2C). Given our focus on the metabolites alterations, we characterized the 296 detected metabolites into four categories according to the public metabolic molecules database 297 ( others. The organic compounds accounted for the vast majority in the three stages (early stage: 299 52.63%; mid-stage: 52.67%; and late-stage: 52.63%) in the four ABIs. Among them, we found 300 that lipids (9%) hold dynamic changes. The nucleotides category (9.33%) was significantly higher 301 in TM compared to the other ABI subtypes at 24H. Generally, the metabolome profile exhibited 302 more altered patterns between PM and TM at 24H (Fig.4A). Besides, this finding was also 303 consistent with the results in 12H and 72H (Extended Data Fig.6B); and could catch out attention 304 to the difference in metabolites caused by the progression stages of ABIs. 305 Inspection of differential expressed metabolites (DEMs) (Fig.4A (Fig.4C, S8). Interestingly, in TM, we found DELs in PE accounted for 325 27.27% (12H) and 18.18% (72H) proportion at the early and late stages respectively; however, we 326 did not detect PE expression at 24H. This suggests that cephalin-related lipids mainly changed in 327 the early and late stages of TM, but at 24H, it preferred the regulation of lecithin-related lipids 328 changes. Also, we found 12H and 72H had more overlapped DELs (Fig.4D) in four ABI subtypes. 329 This analysis revealed that there might be more metabolites alterations in the mid-stage of ABIs 330 (24H) compared with the early (12H) and late stages (72H). 331 These metabolic analyses identified the dynamic alterations in metabolites, especially in lipids, 332 associated with the progression of ABIs, supporting the roles for metabolic disorders in ABIs.

334
Dynamic lipid alterations are associated with PM and TM 335 Metabolomics analysis revealed that the abundance of a batch of lipid molecules changed 336 dynamically in PM and TM (Fig.4A, Extended Data Fig.6B). These lipid alterations may pinpoint 337 the genes involved in the lipids metabolic process. We considered the analysis of lipid-related 338 genes for PC, PE, Cer, CerG1, and PS (downloaded from Mouse Genome Informatics; see method) 339 ( Fig.5A and Supplementary Table 6). The abundance of genes and proteins were correlated well 340 regardless of PM and TM progression in PC-related genes. Compared with TM, the log2FC (versus 341 Ctrl) of PE-related protein Esyt1, CerG1-related proteins Gba and Gba2, and Cer-related protein 342 Agk in PM showed a gradual increase trend during the first 72 hours. For PS-related proteins, the 343 log2FC of Syt12 (versus Ctrl) was higher at 12H, decreased at 24H in PM compared with TM, 344 then back to a high level (log2FC of PM versus log2FC of TM) at 72H (Fig.5B, Supplementary 345 Table 7). This result was confirmed that in the late stage (72H) of ischemic brain injury (PM and 346 TM), lipid-related proteins involved in an important pathological process of ABI. However, PM 347 and TM always occupied an inconsistent pace in the abundance regardless of the transcription and 348 translation levels and ABIs progressions. 349 To further explore the lipids alterations in PM and TM, we analyzed in detail the abundance 350 changes of DELs in PM and TM with the progress of ABIs. It was remarkable that we found 351 similar lipids abundance patterns associated with PM and TM in 12H and 72H (Fig.5C). However, 352 just like the dynamic changes at the transcription and protein levels, the abundance of DELs also 353 showed dynamic changes in 24H. Among them, almost all Cer and PS were down-regulated, and 354 only CerG1 was up-regulated in PM. As for PE and PC, which are the largest portion of lipids in 355 brain tissues, DELs abundances of PM and TM hold a very obvious mutually exclusive pattern. 356 Based on the results above, we found that the Cer and PS of PM were greatly decreased compared 357 to TM, and these two types of lipid molecules appear to have a positive role in neurological 358 function post stroke 38,39 . These data provide molecular evidence of ischemic time-depended 359 secondary brain injury related to lipid metabolism.

361
Discussion 362 The key to developing effective brain injury treatments is to better understand and determine the 363 exact mechanism of the secondary pathology associated with ABI. Current guidelines are agreed 364 on the general principles of early management or medical care for ABI patients 40,41 . However, the 365 progression and survival prediction of patients with late-stage ABI are extremely challenging 366 obstacles to successful treatment selection, partly due to a lack of understanding of the complex 367 pathophysiological changes during secondary brain injury. 368 Here, we reported an integrated omics analysis of ABI mice models, which uncovered molecular 369 alterations associated with ABIs. A total of 15 co-variation risk factors were identified as key 370 regulators for secondary pathophysiological changes of ABIs. The relationship between 371 transcriptome and proteome patterns and prognosis may facilitate the precise treatment and 372 evaluation of late-stage ABI patients. 373 In particular, for the early stages of ABIs, widespread transcriptome abnormalities with 374 prominent signatures of body temperature in the prognosis of ischemic brain injuries. The 375 relationship between abnormal body temperature and ABI severity and clinical outcome has been 376 reported previously 20,42 , but the molecular mechanism behind it, especially the regulatory factors 377 of gene and protein network under the multi-omics model, has not been revealed yet. inflammatory cytokine (e.g. interleukin-6) is reported to be associated with elevated body 383 temperature after stroke 47 . In this study, inflammation-related pathways such as interleukin-6 384 production, cytokine-mediated signaling pathway, and regulation of innate immune response are 385 significantly changed in ABI brains. Our transcriptomics/proteomics combined analysis provides 386 potential molecule targets for the regulation of body temperature after ischemic ABIs. 387 Specifically, although the primary injury mechanism is considered different, traumatic and 388 hemorrhagic injury share dysregulation of cerebral blood flow, resulting in secondary ischemic 389 injury. For ischemic stroke, if reperfusion is not achieved in the subacute phase, a delayed phase 390 (days to weeks after symptom onset) may occur in which ischemic injury is further exacerbated 391 by secondary oxidative stress, brain edema, neuroinflammation, and other associated and relevant 392 deleterious molecular mechanisms. Thus, the multi-omics dynamic analysis would provide 393 evidence of the effect of ischemic time on brain injury from a molecular level. Between the two 394 subtypes of ischemic ABI models, PM and TM, these ischemic injuries harbored few differences 395 in their transcriptomic signatures. However, we found that lipidomics of PM and TM revealed 396 variations between them during the first 72H after injury, showing that lipid metabolism is 397 sensitive to ischemia, which is consistent with the previous investigation 48 . 398 Based on the integrated omics dataset of ABI mice models, we have established a resource 399 database of disease progression and molecular pathology characteristics of ABI to allow 400 researchers to understand and evaluate this disease more comprehensively. Through the results of 401 transcriptome and proteomics, we have identified the genes related to temperature disturbance in 402 the early stage of ischemic ABI and the key molecules in the regulatory network of the entire 403 disease progression. In addition to thermoregulation-related pathways, other pathways related to 404 early brain damage such as synaptic plasticity correlations and neurofilament-based process were 405 significantly changed in ABI models, which have been reported as neurobiological foundations 406 for biomarker applications in brain injury and neurodegeneration 49 . These results will bring about 407 substantial assistance for early intervention and diagnosis of ischemic ABIs, and will also provide 408 a molecular basis for the development of ABI therapeutic drugs.

409
In summary, our integrated omics dataset focused on ABI providing the global characteristics 410 of transcripts, proteins, and metabolites. Based on our observations of divergence changes in 411 multiple omics characteristics for distinct progression of ABIs, we found that correlation analysis 412 combined with transcriptomics and proteomics data can accurately reflect ABIs. Also, the changes 413 in metabolites, especially lipid metabolites, reflect the secondary brain injury progression stage of 414 ABIs. These results strongly suggest that the treatment and rehabilitation of ABI patients should 415 focus on changes in brain metabolites in both traumatic/hemorrhagic and ischemic ABIs for 416 effective, individualized treatment strategies.

418
Methods 419 Animals and ABI mice models. All experiments were approved by the Southwest Medical 420 University Animal Studies Committee (201903-105). We followed the Guide for the Care and Use 421 of Laboratory Animals of China. Male C57BL/6 wild-type mice were purchased at 10-12 weeks 422 of age, housed in a pathogen-free facility with access to food and water. Our methods also included 423 randomization, blinding, and statistical criteria consistent with ARRIVE guidelines (Animals in 424 Research The SAH mouse model was performed by endovascular perforation as previously described 429 with slight modifications 29 . Briefly, anesthesia was induced by inhalation of 3% isoflurane in a 430 nitrous oxide/oxygen mixture (70% oxide, 30% oxygen) and maintained by 1.5% isoflurane 431 administered through a facemask. After the right internal carotid artery (ICA) was exposed, a 5-0 432 prolene filament was introduced into the right external carotid artery (ECA) and advanced through 433 the ICA until the resistance was felt, further advanced 3 mm to induce arterial rupture. 434 Subsequently, the filament was immediately withdrawn. Body temperature was maintained at 37 435 ± 0.5 °C throughout the procedure using a heating pad. 436 The TBI mouse model was induced under isoflurane anesthesia using a controlled cortical 437 impact (CCI) technique 27 . Briefly, the animals were anesthetized with 1.5% isoflurane and checked 438 for pain reflexes. A 3 mm right lateral craniotomy centered at 2.7 mm lateral from the midline and 439 3 mm anterior to lambda was performed with a motorized drill. The skull was removed without 440 disrupting the dura. The CCI was produced with a pneumatic cylinder (Precision Systems and 441 Instrumentation) using a 3 mm diameter flat-tip impounder (velocity, 3 m/s; dwell time, 100 ms; 442 depth, 1.0 mm). A polyvinylidene fluoride skull cap was secured over the craniotomy and sealed. 443 Body temperature was maintained at 37 ± 0.5 °C throughout the procedure using a heating pad. 444 The anesthetized mice were wrapped in a blanket (37 °C) until recovered and were able to freely 445 ambulate.

446
For the ischemic stroke mice models, the transient focal cerebral ischemia was produced by the 447 right cerebral artery occlusion (MCAO) with slight modifications 50 . Briefly, under isoflurane 448 anesthesia (3% induction, 1.5% maintenance), mice were placed in a supine position, a midline 449 incision was made on the neck, and the right common carotid artery (CCA) was exposed. A 6-0 450 nylon monofilament was inserted through the stump of ECA into the ICA and advanced into the 451 middle cerebral artery until light resistance was felt (12 mm). After 90 min of MCAO, reperfusion 452 was initiated by withdrawing the nylon monofilament. Body temperature was maintained at 37 ± 453 0.5 °C throughout the procedure using a heating pad. The same procedure is performed with the 454 permanent MCAO model, but no reperfusion until euthanasia. Technologies, CA, USA) and the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system 476 (Agilent Technologies, CA, USA), respectively.

477
A total amount of 3 μg RNA per sample was used as input material for the RNA sample 478 preparations. Sequencing libraries were generated using NEBNext® UltraTM RNA Library Prep 479 Kit for Illumina® (NEB, USA) following the manufacturer's recommendations and index codes 480 were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA 481 using poly-T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations 482 under elevated temperature in NEBNext First Strand Synthesis Reaction Buffer (5X). First strand 483 cDNA was synthesized using a random hexamer primer and M-MuLV Reverse Transcriptase 484 (RNase H-). Second strand cDNA synthesis was subsequently performed using DNA Polymerase 485 I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase 486 activities. After adenylation of 3' ends of DNA fragments, NEBNext Adaptor with hairpin loop 487 structure was ligated to prepare for hybridization. To select cDNA fragments of preferentially 488 250~300 bp in length, the library fragments were purified with the AMPure XP system (Beckman 489 Coulter, Beverly, USA). Then 3 μl USER Enzyme (NEB, USA) was used with size-selected, 490 adaptor-ligated cDNA at 37°C for 15 min followed by 5 min at 95 °C before PCR. Then PCR was 491 performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers, and Index (X) 492 Primer. At last, PCR products were purified (AMPure XP system) and library quality was assessed 493 on the Agilent Bioanalyzer 2100 system. and negative) Two modes); the secondary mass spectrum is obtained by information-dependent 562 acquisition (IDA), and the high sensitivity mode is adopted, Declustering potential (DP): ±60 V 563 (both positive and negative modes), Collision Energy: 35 ± 15 eV, IDA setting The following 564 Exclude isotopes within 4 Da, Candidate ions to monitor per cycle: 10. The data collection is 565 segmented according to the mass range, 50-300, 290-600, 590-900, 890-1200, thereby expanding 566 the collection rate of the secondary spectrum, each method collects four repetitions per segment. 567 The collected data were used to identify the structure of metabolites using self-built MetDDA and 568 LipDDA methods. After the sample is tested, use AB Triple TOF 6600 mass spectrometer to 569 collect the primary and secondary spectra of the sample. 570 The ESI source conditions after HILIC chromatographic separation are as follows: Ion Source 0.20 s/spectra, product ion scan accumulation time 0.05 s/ spectra; the secondary mass spectrum 575 is obtained by information-dependent acquisition (IDA), and the high sensitivity mode is adopted, 576 Declustering potential (DP): ±60 V (both positive and negative modes), Collision Energy: 35±15 577 eV, IDA The settings are as follows: Exclude isotopes within 4 Da, Candidate ions to monitor per 578 cycle: 6 (Shanghai Applied Protein Technology, Shanghai, China). 579 580 Lipid library preparation and sequencing. Lipid extraction and mass spectrometry-based lipid 581 detection were performed by Applied Protein Technology. And a separate sample in each group 582 and mix them equally together to create a pooled QC sample. QC samples were inserted into the 583 analysis queue to evaluate the system stability and data reliability during the whole experimental 584 process. Precisely weigh 30 mg of the sample and transfer it to a 2 mL centrifuge tube pre-installed 585 with an appropriate number of magnetic beads, add 200 μL of 4°C water, and put it in the solution. 586 Flash freezing in nitrogen for 5 seconds, and homogenize it with MP homogenizer (24×2, 6.0M/S, 587 the 60s, three times). Add 240μL of pre-cooled methanol, vortex to mix, add 800μL of MTBE, 588 vortex to mix, sonicate in a low-temperature water bath for 20 minutes, place at room temperature 589 for 30 minutes, and centrifuge at 14000 g at 10°C 15 min, take the upper organic phase, blow dry 590 with nitrogen, and store the sample at -80°C. The samples were separated using UHPLC Nexera 591 LC-30A ultra-high performance liquid chromatography system. The column temperature is 45°C; 592 the flow rate is 300 μL/min. Mobile phase composition A: 10 mM ammonium formate acetonitrile 593 aqueous solution (acetonitrile: water=6:4, v/v), B: 10 mM ammonium formate acetonitrile 594 isopropanol solution (acetonitrile: isopropanol=1:9, v/v ). The gradient elution procedure is as 595 follows: 0-2 min, B is maintained at 30%; 2-25 min, B changes linearly from 30% to 100%; 25-596 35min, B is maintained at 30%. The sample is placed in the 10°C autosamplers during the entire 597 analysis. IToavoid the influence caused by the fluctuation of the detection signal of the instrument, 598 a random sequence is adopted to carry out continuous analysis of the sample. 599 Electrospray ionization (ESI) positive ion and negative ion modes were used for detection. Mass 600 spectrometer (Thermo ScientificTM) performs mass spectrometry analysis. The mass-to-charge 601 ratios of lipid molecules and lipid fragments are collected according to the following method: 10 602 fragment patterns (MS2 scan, HCD) are collected after each full scan. MS1 has a resolution of 603 70,000 at M/Z 200, and MS2 has a resolution of 17,500 at M/Z 200. Lipid identification (secondary 604 identification), peak extraction, peak alignment, and quantification were assessed with 605 LipidSearch software version 4.1 (Thermo Scientific™). In the extracted ion features, only the 606 variables having more than 50% of the nonzero measurement values in at least one group were 607 kept (Shanghai, China). 608 609 Transcriptome analysis. The raw fastq files were trimmed using trim galore (version 1.18) to 610 remove adaptor sequences and low-quality reads. Then FastQC (version 0.11.9) was used for 611 quality control. The remaining reads were aligned to the GRCm38 mouse genome using HISAT2 612 (v2.2.0) with default parameters and further filtered with samtools (version 1.10, parameters used: 613 samtools view -F 1804 -f 2 -q 30). Gene counts were calculated from the mapped reads using 614 featureCounts (v2.0.1) with the Ensembl gene annotation (version mm10). Subsequently, TPM 615 (Transcripts Per Kilobase of exon model per Million mapped reads) in each gene was calculated 616 for subsequent analysis. 617 618