The number of intravascular CD45+ cells increase in reperfused brains with concurrent liver ischemia

Damage to the BBB triggers neuroinflammation (Johnson, 2023). Studies have shown increased immune cell infiltration in the brain tissue during ischemia-reperfusion injury (Dal-Bianco et al, 2016; Gelderblom et al, 2009). To determine whether concurrent liver ischemia can induce an increase in the number of immune cells infiltrating the brain, we performed HE and immunohistochemical staining. In HE-stained sections of the five brain regions, more nucleated blood cells were found in the vascular lumen in the brain, particularly in the frontal lobe of the BLWI-30 group compared to the BWI-30 group (Fig. 2A–G). Further characterization of nucleated hematopoietic cells through immunohistochemical staining of CD45 revealed more CD45+ cells in the brain, particularly in the frontal and temporal lobes of the BLWI-30 group compared to the BWI-30 group (Fig. 2H–N). These results suggest that the ischemic liver might increase the local immune response during brain ischemia-reperfusion injury.

Open in a separate window

Figure 2

A greater number of intravascular CD45-positive cells were observed in the BLWI-30 group compared to the BWI-30 group in vivo.

(A) Hematoxylin-eosin staining showed vascular and nucleated blood cells in the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and hippocampus of pigs (200×). The arrowhead pointed to nucleated blood cells. (B–G) The number of nucleated blood cells per field in the brain (average number in the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and hippocampus). (H) Immunofluorescence staining for CD45 in the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and the hippocampus of pigs (200×). The arrowhead pointed to brown-yellow CD45+ cells. (I–N) The mean number of CD45+ cells of two fields in the frontal lobe, parietal lobe, temporal lobe, occipital lobe, hippocampus or brain (average number of the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and hippocampus). (B–G), and (I–N) Sham, no ischemia, n=3; BWI-30, brain with 30-min warm ischemia, n=6; BLWI-30, brain and liver with 30-min warm ischemia, n=7; all replicates shown were biological replicates; Mean±SEM, two-tailed ratio unpaired t-test. Source data are available online for this figure.

Brain circulation and metabolic activity post-CA is restored during ex vivo brain NMP

In the in vivo global cerebral ischemia model, clamping of the portal vein transiently affected the hemodynamic stability (Fig. EV1A) and serum lactate levels (Fig. EV1B,C), which might affect the extent of brain injury. Recently, an ex vivo brain NMP model has been developed to investigate the pathogenesis of brain diseases (Vrselja et al, 2019). To comprehensively evaluate how the liver impacts the recovery of post-CA brain injury, we constructed ex vivo models: one focusing solely on the brain and the other incorporating liver-assisted brain NMP. In these models, the brain suffered complete, rather than global, ischemia before reperfusion, without affecting the hemodynamic stability by surgical manipulation of the liver. The pressure-controlled NMP provided normothermic, oxygenated perfusion to the brain with whole blood-based perfusate to mimic the return of spontaneous circulation post-CA. The perfusion flow, metabolic activity and electrocortical activity were monitored during ex vivo NMP. Figure 3A, Appendix Fig. S3A, and Appendix Table S2 demonstrate the technology for the liver-assisted brain NMP. There were six experimental groups in the current study: a brain-only control with rapid NMP (BOR group), a liver-assisted brain with rapid NMP (LABR group), and four liver-assisted brain groups in which brain NMP was preceded by 30–240min of warm ischemia time (WIT) (LABWI groups) (Appendix Fig. S3A). Brains isolated under identical conditions but without undergoing NMP were used as the Sham controls in our study.

Open in a separate window

Figure 3

Liver-assisted NMP reduces cell damage and hypoxic injury in pig brains.

(A) Technologies employed in the liver-assisted brain normothermic machine perfusion (NMP) model. (B) Whole brain structure in the three groups. The surface of the brain with cerebral oedema and the narrowed sulci (lower arrowhead) and widened gyri (upper arrowhead), in the BOR group. The characteristics were minimal in the LABR group. BOR, brain-only control with rapid NMP; LABR, liver-assisted brain with rapid NMP. (C) Hematoxylin-eosin staining (400×) of the cerebral cortex, thalamus, hippocampus, cerebellum, and brainstem after 6h of NMP (arrowheads pointing to the structures that indicate the neuronal shrinkage after suffering ischemia-reperfusion injury) in the five regions. (D–I) The mean injury scores of two fields in the cerebral cortex, thalamus, hippocampus, cerebellum, brainstem or brain (average score of the cerebral cortex, thalamus, hippocampus, cerebellum, and brainstem). (J) The perfusate levels of S100-β. (K) Immunohistochemistry analysis of hypoxia inducible factor-1α (HIF-1α) indicating the degree of hypoxic injury in the hippocampus and cortex. The black arrowhead pointed to HIF-1α positive cells. DG, dentate gyrus. (L, M) Number of cells with upregulated expression of HIF-1α were counted in the cerebral cortex (L) and hippocampus (M). (D–I, J, L, M) n=5, Mean±SEM; (D–I, L, M) two-tailed ratio unpaired t-test; (J) 2-way ANOVA. All replicates shown were biological replicates. Source data are available online for this figure.

Open in a separate window

Figure EV1

Systolic blood pressure and lactic acid levels in the global cerebral ischemia model.

(A) Systolic blood pressure of the pigs in the BWI-30 and BLWI-30 groups. (B, C) Lactic acid levels of the right mammary artery (B) and right external jugular vein (C) of the pigs in the BWI-30 and BLWI-30 groups. (A–C) Sham, no ischemia, n=3; BWI-30, brain with 30-min warm ischemia, n=6; BLWI-30, brain and liver with 30-min warm ischemia, n=7. All replicates shown were biological replicates. I, ischemia; R, reperfusion. Mean±SEM, 2-way ANOVA analysis; BLWI-30 versus BWI-30.

Due to the preparation of whole blood-based perfusate, the brains in the BOR group underwent a WIT of 10.1±1.8min (Appendix Fig. S3A). Because of the additional procedures, which included procuring the liver and connecting it to the system, the mean WIT of the brains was about 4min longer in the LABR group (14.2±0.7min) than in the BOR group (P=0.0670) (Appendix Fig. S3A). The livers continued to produce bile during perfusion with a portal flow exceeding 500mL/min. They exhibited a homogeneous appearance with parenchyma tissue of soft consistency throughout the procedure, indicating that the livers were functioning well. Ex vivo circulation and oxygen consumption were successfully restored in both the BOR and LABR groups. There was no significant difference in the perfusion pressure, pO₂, pCO₂, pH, as well as lactate and glucose levels between the BOR and LABR groups during ex vivo NMP (Appendix Fig. S3B–G). However, the left middle cerebral artery (MCA) flow declined in the BOR group, while it maintained stable in the LABR group (Appendix Fig. S3H). Consistently, the arterial resistance declined during the first 2.5h but increased thereafter in the BOR group. In contrast, the arterial resistance remained stable in the LABR group (Appendix Fig. S3I).

Collectively, we have established both ex vivo brain-only NMP and liver-assisted brain NMP models. These models enable the restoration of circulation and metabolic activity in the brains ex vivo after CA.

Liver-assisted brain NMP reduces post-CA brain injury

Next, we tested how the inclusion of the liver in the ex vivo NMP circuit could impact the recovery of post-CA brain injury.

In line with the increased arterial resistance and reduced perfusion flow, obvious edema of the reperfused brains was observed after 6-h NMP in the BOR group, while edema was not obvious in the LABR group (Fig. 3B). To assess ischemic brain injuries, HE staining of different brain areas was conducted and tissue injury scores were calculated. In the BOR group, notable pathological changes were observed, including shrunken cells, pyknotic nuclei, and expanded perinuclear space in cortical and hippocampal cells, compared to those in the Sham group. In contrast, in the LABR group, pyramidal neurons exhibited plump cell bodies with large, central nuclei and abundant nerve fibers, similar to those observed in the Sham group (Fig. 3C). The tissue injury scores in the brains, particularly in the cortex and hippocampus, were lower in the LABR group than in the BOR group (Fig. 3D–I).

In addition, the liver-assisted perfused brains had a substantially decreased perfusate level of S100-β, a biomarker of neural injury (Michetti et al, 2019), compared to the brains in the BOR group (Fig. 3J). Moreover, immunohistochemistry (IHC) was performed on the cerebral cortex to evaluate ischemia-reperfusion injury using markers such as hypoxia inducible factor-1α (HIF-1α), heat shock protein 70 (HSP70), antioxygen nuclear factor erythroid 2 like 2 (NRF2) and DNA repairing 8-oxoguanine DNA glycosylase (OGG). The results demonstrated that the brains in the BOR group had a higher density of these markers compared to both the Sham and LABR groups (Fig. 3K–M; Appendix Fig. S4), suggesting that ischemia-reperfusion injury of the brain is aggravated when the liver is absent from the NMP circuit.

Taken together, these results suggest that the liver can protect the brain from post-CA ischemia-reperfusion injury.

Liver-assisted brain NMP improves neuronal viability and protects the cytoarchitecture

Considering that the concurrent ischemia of the liver increased the infarct area of the frontal lobe in the in vivo global cerebral ischemia model, we then assessed the density of live neurons in the ex vivo model using Nissl staining. The staining revealed comparable densities of live neurons between the Sham and LABR groups in both the CA1 region and dentate gyrus of the hippocampus. In contrast, the density of live neurons in the CA1 region was lower in the BOR group than in the LABR group (Fig. 4A–C). Immunofluorescence analysis of a microglial marker ionized calcium binding adapter molecule 1 (IBA1) produced fragmented signals, with signs of cellular destruction in the CA1 region of the BOR brains but preserved density in the dentate gyrus of all perfused brains in the LABR group (Appendix Fig. S5A–C). Staining for the astrocytic marker glial fibrillary acidic protein (GFAP) revealed preserved astrocyte density in all three groups (Appendix Fig. S5D–F). Collectively, incorporating the liver into the ex vivo brain NMP setup can protect the reperfused brains from cell death in the neurons and microglia within the CA1 region.

Open in a separate window

Figure 4

Improved neuronal viability and electrocortical activity with ex vivo liver-assisted brain NMP compared to brain-only NMP.

(A) Nissl staining (400×) shows the structural integrity of neuronal somas and axons in the hippocampus of pigs. The white arrowhead pointed to active neurons. BOR, brain-only control with rapid NMP; LABR, liver-assisted brain with rapid NMP. (B, C) Neurons with intact cell bodies were counted in the CA1 region (B) and dentate gyrus (C). (D, E) Transmission electron micrographs of the three groups in the pig cerebral cortex (D) and hippocampus (E). The arrowhead pointed to microvessels, neurites, myelin sheaths, mitochondria or synapses. RBC, red blood cell; GC, granulocyte. (F) Electroencephalography (EEG) results from a representative pig brain in the BOR and LABR groups. (G, H) The EEG scores (G) and spectroscopic entropy (H). (B, G, H) n=5, Mean±SEM; (B, H) Two-tailed ratio unpaired t-test; (G) 2-way ANOVA. All replicates shown were biological replicates. Source data are available online for this figure.

To further evaluate the ultrastructural characteristics of reperfused brains, we conducted the transmission electron microscopy analysis on the brain tissues. In the cerebral cortex and hippocampus of the BOR group, we observed severe tissue vacuolization, mild microvascular edema, slightly loosened myelin sheath, extensive mitochondrial swelling or membrane rupture, rupture and degradation of mitochondrial cristae, and significant synaptic degradation. In contrast, the microvessels, neurites, myelin sheaths, mitochondria, synapses, and neurotransmitters were structurally intact in the majority of cells in the LABR group. Erythrocytes were observed in the microvascular lumen of the LABR and Sham groups, whereas granulocytes were observed in the BOR group (Fig. 4D,E), which is in line with in vivo studies that observed more CD45⁺ cells trapped in the vessels of the brains with concurrent liver ischemia.

Collectively, integrating a liver into the ex vivo brain NMP circuit can enhance neuronal viability and maintain cytoarchitecture.

Liver-assisted brain NMP preserves global electrophysiological activity

It is crucial to determine if the reduction in brain injury translates to improved global electrophysiological activity in the LABR group. Global electroencephalograph (EEG) monitoring revealed the presence of α and/or β waves, both of which are considered to represent conscious activity (Koch et al, 2016; Mashour and Hudetz, 2018), within 30min after the start of NMP. However, these waves disappeared after 3–5h of NMP in four of five brains in the BOR group (Fig. 4F; Appendix Table S3). In contrast, frequent α and β waves appeared within 1h of NMP and persisted throughout the entire 6-h NMP period in all the perfused brains in the LABR group (Fig. 4F; Appendix Table S3). At the end of NMP, the EEG scores and spectroscopic entropy were much higher in the LABR group than in the BOR group (Fig. 4G,H). Collectively, these data show that electrocortical activity can be maintained ex vivo by liver-assisted brain NMP instead of brain-only NMP.

To determine the maximum duration of warm ischemia after which global electrophysiological activity can be restored and maintained ex vivo in pigs, we extended the WIT of the brains to 30min (LABWI-30), 50min (LABWI-50), 60min (LABWI-60), or 240min (LABWI-240) with liver support (Fig. EV2). Global electrophysiological activity was restored in all five brains, and the presence of α or β waves was maintained until the end of NMP in four of five brains in the LABWI-30 group and in all five brains in the LABWI-50 group (Appendix Table S3). In the LABWI-60 group, all brains exhibited α or β waves during 3–4h of NMP, although the waves disappeared shortly thereafter (Appendix Table S3). In the LABWI-240 group, no α or β waves were observed (Appendix Table S3). These results indicate that global electrophysiological activity of the brain can be restored and maintained ex vivo following a WIT of over 50min when a functioning liver is incorporated into the NMP circuit.

Open in a separate window

Figure EV2

Electrophysiological activity during ex vivo liver-assisted brain NMP.

(A–D) Electroencephalography results of a representative pig brain in the liver-assisted brain groups, where normothermic machine perfusion (NMP) followed intervals of 30-, 50-, 60-, or 240min. LABWI, liver-assisted brain groups in which brain NMP was preceded by 30–240min of warm ischemia time; Pre-op, Pre-operation.

Transcriptome analysis indicates decreased cell death and immune response in brain tissues without simultaneous hepatic ischemia compared to those with the condition

To elucidate the molecular mechanisms underlying the protective effects of the liver on brain injury, RNA-seq was employed to profile the frontal and temporal lobes of the brain with or without concurrent hepatic ischemia. In the frontal lobe, we observed a significant upregulation and downregulation of genes in the presence of simultaneous hepatic ischemia. We identified 245 upregulated genes and 350 downregulated genes in the frontal lobe with simultaneous hepatic ischemia (BLWI-30 group) (Fig. 5A). Functional annotation of the differentially expressed genes highlighted distinct functional terms. The upregulated genes in the BLWI-30 group were enriched in GO terms related to the regulation of programmed cell death and immune response, such as positive regulation of apoptotic process, regulation of T cell chemotaxis, and B cell activation. Conversely, genes downregulated in the BLWI-30 group were enriched in GO terms associated with neuronal functions, including neuron projection development and regulation of neurotransmitter levels (Fig. 5B). Similar transcriptome differences were observed in the temporal lobe, with 269 upregulated genes and 337 downregulated genes in the BLWI-30 group (Fig. 5C). These differentially expressed genes exhibited comparable enrichment in functional gene ontology terms (Fig. 5D). These molecular alterations corroborate our prior findings on brain function and pathological analyses, indicating that the support of a functioning liver can reduce post-CA brain injury.

Open in a separate window

Figure 5

Transcriptomic differences between brain tissues with and without simultaneous hepatic ischemia by RNA-seq.

(A) Volcano plot displaying the differential genes in the frontal lobe of the pig brains with (BLWI-30 group) and without (BWI-30 group) simultaneous hepatic ischemia. Genes significantly upregulated or downregulated in the BLWI-30 group (P value<0.05, log2(fold change) > 1) are represented in red and blue, respectively. Functional interest genes were annotated and labeled according to their functional terms, as shown in (B). (B) Bar plot illustrating GO enrichment of significant differential genes in the frontal lobe. Upregulated and downregulated genes were tested separately and represented in red and blue, respectively. GO terms were color-coded corresponding to (A). (C) Volcano plot depicting the differential genes in the temporal lobe of the brains of the two groups. Genes significantly upregulated or downregulated in the BLWI-30 group (P value<0.05, log2(fold change) > 1) are represented in red and blue, respectively. Functional interest genes were annotated and labeled according to their functional terms, as shown in (D). (D) Bar plot displaying GO enrichment of significant differential genes in the temporal lobe. Upregulated and downregulated genes were tested separately and represented in red and blue, respectively. GO terms were color-coded to correspond with (C). (E) Violin plot showing gene expression (RPKM) in each biological replicate. The vertical lines (whiskers) connecting the box represented the maximum and minimum values. The box signified the upper (75th percentiles) and lower quartiles (25th percentiles). The central band inside the box represents the median (50th percentiles). Outliers were shown. F, frontal lobe; T, temporal lobe. (A, C, E) P values and log2(fold change) were calculated by the Wald test using the DESeq2 R package. (B, D) P values were calculated by the hypergeometric test using the MetaCore database. (A–E) Frontal lobe: BWI-30, n=6, BLWI-30, n=4; temporal lobe: BWI-30, n=5, BLWI-30, n=4. All replicates shown were biological replicates. Source data are available online for this figure.

Furthermore, we identified differential genes involved in metabolic processes. For instance, genes such as AK7, AK9, TYMS, involved in nucleotide metabolism, may be influenced by cell death and immune responses (Fig. EV3). Intriguingly, we also identified differentially expressed genes enriched in terms related to the response to energy reserve metabolic processes, including genes involved in lipid, linoleic (PLA2G2D), and ketone (BMP5, PTGDR2) metabolism processes (Fig. 5E), as well as in glycolysis/gluconeogenesis (ALDOB/PCK1) (Fig. EV3). These results suggest that the liver might regulate the post-CA brain injury through altering metabolic activity of the brain.

Open in a separate window

Figure EV3

Differentially expressed genes involved in metabolic processes in the global cerebral ischemia model.

(A, B) Violin plot illustrating the expression (reads per kilobase per million, RPKM) of differential genes in each of the biological replicates in the frontal lobe (BWI-30, n=6; BLWI-30, n=4) (A) or temporal lobe (BWI-30, n=5; BLWI-30, n=4) (B) of pigs. All replicates shown were biological replicates. (A, B) The vertical lines (whiskers) connecting the box represented the maximum and minimum values. The box signified the upper (75th percentiles) and lower quartiles (25th percentiles). The central band inside the box represents the median (50th percentiles). Outliers were shown. P values were calculated by the Wald test using the DESeq2 R package.

Dramatic metabolome differences in brain tissues with and without simultaneous hepatic ischemia were revealed

To unveil the distinct metabolic dynamics during brain ischemic injury with or without liver ischemia, we conducted an ultra-high-performance liquid chromatography quadrupole time-of-flight mass spectrometry (UHPLC-QTOFMS) analysis of the frontal and temporal lobes. Supervised partial least squares analysis (PLS-DA) of the overall metabolic profiles from 5–7 biological replicates in both groups clearly demonstrated distinct metabolomes (Figs. 6A,B and EV4A,B). The general distribution of metabolism classes is summarized in Fig. EV4C,D.

Open in a separate window

Figure 6

Metabolic differences in pig brain tissues with and without simultaneous hepatic ischemia by UHPLC-QTOFMS.

(A, B) Volcano plot illustrating the differential metabolites with secondary mass spectrometry (MS2) names in the frontal lobe (A) and temporal lobe (B) of the pig brains with and without simultaneous hepatic ischemia. Significant metabolic alterations (P value<0.05, VIP>1) were denoted in red or blue, respectively. (C, D) Bubble plot displaying the enriched metabolic pathways for increased metabolites in the frontal lobe and temporal lobe, respectively. The value on the x-axis and the size of the circle indicates the impact of the difference. The value on the y-axis and the color indicates the significance of enrichment in −ln(P). (E, F) Matchstick plot illustrating the top metabolic alterations with lowest P values in the frontal lobe (E) and temporal lobe (F). The x-axis displays the log2(fold change) of BLWI to BWI. Each item’s VIP score is denoted by color. (A–F) Frontal lobe: BWI-30, n=5, BLWI-30, n=5; temporal lobe: BWI-30, n=6, BLWI-30, n=7. All replicates shown were biological replicates. P values were calculated by the t-test (A, B, E, F) and Fisher’s exact test (C, D) using the ggplot2 R package. Source data are available online for this figure.

Open in a separate window

Figure EV4

Metabolome differences in pig brain tissues with and without simultaneous hepatic ischemia.

(A) Scatter plot depicting orthogonal projections to latent structures-discriminant analysis (OPLS-DA) results for the frontal lobe comparing the BLWI-30 and BWI-30 groups. (B) Scatter plot depicting the OPLS-DA data for the temporal lobe comparing the BLWI-30 and BWI-30 groups. (C, D) The bar plot displays the KEGG annotation of detected metabolites for the frontal lobe (C) and temporal lobe (D). The x-axis represents the percentage of identified metabolites in each KEGG class. (E, F) Depiction of overall changes in differential metabolites within a specific pathway for the frontal lobe (E) and temporal lobe (F). The Differential Abundance Score (DA Score) was calculated as the ratio of the difference between the upregulated metabolite count and the downregulated metabolite count on a specific pathway to the total count of metabolites on that pathway. (A–F) Frontal lobe: BWI-30, n=5, BLWI-30, n=5; temporal lobe: BWI-30, n=6, BLWI-30, n=7. All replicates shown were biological replicates. P values were calculated by Fisher’s exact test (E, F) using the ggplot2 R package.

Interestingly, we found that the downregulated metabolites in the BLWI-30 group, which were higher in the BWI-30 group, were significantly enriched in pathways such as pantothenate and CoA biosynthesis, and glycerophospholipid metabolism in both brain regions (Fig. 6C,D). Further analysis revealed that metabolites enriched in the linoleic acid, arachidonic acid, and glycerophospholipid pathways were decreased in the BLWI-30 group (Fig. EV4E,F). Both linoleic acid and arachidonic acid metabolism are crucial energy metabolism pathways and are important for brain function. Linoleic acid, arachidonic acid, and their derivatives are essential components in the modulation of synaptic transmission and neurotransmitter release, thereby influencing various neurological processes (Wei et al, 2012). Linoleic acid, arachidonic acid, and their derivatives may contribute to neuroprotection and support neuronal survival under various pathological conditions, including ischemic injuries (Nayeem et al, 2022). In contrast, the upregulated metabolites in the BLWI-30 group were fewer and enriched in divergent pathways, such as pantothenate and CoA biosynthesis, beta-alanine metabolism, as well as cholinergic synapses (Fig. EV4E,F). The most significantly differentially expressed metabolites are shown in Fig. 6E,​E,F,F, including variant compounds contributing to energy metabolism and neuronal functions. The distinct metabolome observed in brains experiencing ischemic injury in the presence or absence of concurrent hepatic ischemia highlights the critical influence of the liver on brain metabolic activities under ischemic conditions.

Animals

The feed given to Tibet minipigs (female and male) complied with the GB 14924.3-2010 Chinese standard for “Laboratory animals-Nutrients for formula feeds”. The pigs were fed in an ordinary environment. Animal facilities had an independent ventilation system, the number of air changes was greater than 8 times per hour, the temperature was controlled between 16 and 26°C, the daily temperature variation did not exceed 4°C, and the relative humidity was maintained between 40% and 70%. The living space complied with the GB 14925 Chinese standard for “Laboratory animals—Environment and housing facilities”.

In vivo global cerebral ischemia with or without liver ischemia

Seventeen Tibet minipigs (weight, 25.0±3.1kg; age, 6 months) purchased from Guangdong Pearl Biotechnology Co. LTD were included in this experiment. Seven pigs underwent 30min of global cerebral and hepatic ischemia followed by reperfusion (BLWI-30 group). Three pigs underwent sham operations (control group). Appendix Fig. S6 summarizes the number of pigs included and excluded in each experimental group for various analyses. Throughout the procedure, the pigs were maintained under general anesthesia via endotracheal intubation, using a combination of inhaled isoflurane (1–2%) and propofol (100mg/h). Prior to the procedure, pancuronium (0.1mg/kg) was administered for neuromuscular blockade, and sufentanil (10μg/kg) was given for perioperative analgesia. Post-operatively, the pigs were monitored for 24h, unless they died prematurely. Blood and tissue samples were obtained at scheduled time points. Antibiotics were administered every 8h perioperatively, and body temperature was maintained at 37°C using a heating pad.

A pig model of global cerebral ischemia and neurological severity scores has previously been established (Allen et al, 2012). Through a 7–8cm supra-sternal incision, the following arteries were isolated: the innominate artery, left subclavian artery, both internal mammary arteries, both vertebral arteries, both common carotid arteries. The right mammary artery and right external jugular vein were catheterized for the purpose of obtaining blood gas samples and administering drugs. The innominate artery, left subclavian artery just distal to the aortic arch, both internal mammary arteries, and both distal subclavian arteries were clamped for global brain ischemia. Heparin (250 IU/kg) was administered prior to the ischemia insult. All vascular clamps were removed after 30min of brain ischemia. In vivo Optical Spectroscopy (INVOS) was employed to monitor cerebral tissue oxygen saturation, while ambulatory blood pressure monitoring and neural examination confirmed complete global brain ischemia. Thirty minutes after reperfusion, protamine was injected, and the incision was closed. A 20 Fr drainage tube was placed for mediastinal drainage. Midazolam (0.1mg/kg) was administered in cases where postoperative seizures persisted for more than a few minutes.

In the BLWI-30 group, the portal vein and hepatic artery were thoroughly dissected. The liver underwent simultaneous ischemia and reperfusion with the brain through clamping and unclamping of these two vessels.

Neurological severity scores

Neurological behavior was independently assessed by two laboratory team members 6h after reperfusion, using a species-specific behavior scale. The neurological severity scores assess five components as follows (Allen et al, 2012): (I) Central nerve function (0–100 points): pupil size, eye position, light reflex, lid reflex, corneal reflex, ciliospinal reflex, oculocephalic reflex, auditory reflex, gag reflex, and carinal reflex (each 0–10); (II) Respiration (0–100 points): normal (0), hyperventilation (25), abnormal (50), to absent (100); (III) Motor sensory function (0–100 points): stretch reflex, motor response to pain, positioning, and muscle tonus (each 0–25); (IV) Level of consciousness (0–100 points): normal (0), cloudy (30), delirium (45), stupor (60), to coma (100); and (V) Behavior (0–100 points): drinking, chewing, sitting, standing (each 0–15), and walking (0–40). A total score of 0 indicated normal neurological function, while a score of 500 indicates brain death.

Detection of blood biochemical indicators

Venous blood samples were collected for the detection of biochemical indicators. The blood samples were centrifuged at 3000rpm for 10min at 4°C. Serum samples were collected for aminotransferase (AST) and lactate dehydrogenase (LDH) measurement using the Automatic Chemical Analyzer 7600-100 (Hitachi, Ltd, Tokyo, Japan). Lactic acid and glucose were measured from whole blood samples using a blood gas and chemistry analyzer (EDAN, I15 Vet).

Triphenyl-tetrazolium chloride (TTC) staining

Frontal coronal brain slices of the pigs were cut with a thickness of 2mm. The slices were put into 2% triphenyl-tetrazolium chloride solution (Asegene, G3005) in a dark and were bathed at 37°C for 30min. After being washed with PBS for 3min, brain slices were photographed immediately. The white area indicates infarct tissue.

Surgical procedure for ex vivo brain perfusion

Atropine at a dose of 0.05mg/kg was administered via intramuscular injection 15 to 30min before tracheal intubation. Zoletil 50, a combination of tiletamine hydrochloride and zolazepam hydrochloride (5mg/kg, Virbac), was administered via intramuscular injection for basic anesthesia. Anesthesia was sustained with inhalation of isoflurane (1–2vol% end expiratory). Intraoperative infusion of fentanyl (6μg/kg, repeated every 45min) was used to maintain analgesia.

After anesthesia, the skin, connective tissues, and musculature were cut around the connection between the C7 and T1 vertebrae with a high-frequency electric knife. The bilateral common carotid arteries and veins were separated, and the vertebral arteries from the subclavian arteries were ligated. Thoracotomy was then performed to expose the roots of the common carotid artery trunk and jugular vein stem. Next, the pigs were intravenously injected with 62500 U heparin. The pigs were euthanized by exsanguination (800–1000mL of whole blood) via the abdominal aorta, then injected with 1g of potassium chloride into the left cardiac ventricle to induce cardiac arrest. After cardiac arrest, the common carotid artery trunk and jugular vein stem were separated, and the esophagus, trachea and spinal cord were disconnected.

In the Sham group, immediately after the above surgery, the whole brain was isolated without normothermic machine perfusion (NMP) for pathological assessment. In the brain-only NMP group, a 16 Fr arterial cannula was inserted into the common carotid artery trunk to perfuse the brain. For the liver-assisted brain NMP group, the abdominal aorta (with the celiac artery) was connected to the common carotid artery trunk, and the 16 Fr arterial cannula was inserted into the abdominal aorta to simultaneously perfuse the liver and brain. A straight 24 Fr cannula was inserted into the portal vein to perfuse the liver.

Ex vivo brain NMP technology

Twenty Tibet minipigs (weight, 30–50kg) purchased from Southern Medical University, Guangzhou, China, were used in the development of the brain NMP technology. In these preliminary experiments, the technology, including the NMP conditions, perfusate components, methods for brain function assessment, and the appropriate ex vivo NMP duration, was optimized. After that, brains from 27 additional pigs were perfused in the formal experiments, and five more brains were harvested immediately after CA without NMP and used as controls (Sham group) in the pathological assessments. In our ex vivo brain NMP system (Fig. 3A), a perfusate based on whole blood (Appendix Table S2) circulated through a pulsatile arterial line and in a nonpulsatile portal vein line. It passed through an oxygenation unit, heater, and filtration unit, after which it perfused the brain and liver. The filtration unit filtered out small blood clots and impurities, and the heater maintained the temperature of the perfusate at 37°C. Over 10min, the arterial perfusion pressure was gradually increased from 60mmHg to 90–95mmHg to achieve an MCA flow close to the preoperative value. The portal vein perfusion pressure was set at 5–12mmHg to maintain a flow higher than 500mL·min⁻¹.

The brains were rapidly reperfused ex vivo or after various intervals. WIT was defined as the interval between cardiac arrest and brain reperfusion. When a liver was perfused simultaneously, it was reperfused ex vivo immediately upon harvesting to ensure its viability.

Global electroencephalogram (EEG) monitoring of the brains

EEG was monitored during ex vivo brain NMP. After sedation but before tracheal intubation, the initial global EEG of the pig brain was recorded. During the total duration of ex vivo NMP, global EEG activity of the isolated pig heads was continuously recorded. Real-time detection and assessment of global EEG activity from the scalp were performed with a qEEG analysis system (NicoletOne software version 5.94, Natus Medical Inc., San Carlos, CA, USA). The spectral entropy of qEEG was used to assess the conscious states of the isolated pig heads, especially the awake state. Global EEG signals were recorded using 12 silver disc electrodes placed at standard locations: Fp1, Fp2, C3, C4, T3, T4, O1, O2, Fz, Cz, A1, and A2. Based on the presence of δ (1 point), θ (2 points), α (3 points) and β waves (4 points), the EEG score was calculated hourly. Each cerebral hemisphere was scored, and the scores on both sides were averaged.

Measurement of whole blood gas and biochemistry during the ex vivo brain NMP

Hourly venous perfusate samples were collected from the jugular vein during NMP. Perfusates were collected from the common carotid artery trunk every 20–30min during NMP. One hundred microliters of each sample were immediately analyzed using the i-STAT®1 Blood Analyzer system (Abbott, Flextronics, Singapore) with CG4+ and EC8+ test cartridges. The arteriovenous gradients of glucose, lactate, pH, and partial pressures of carbon dioxide (pCO₂) and oxygen (pO₂) were calculated.

Transcranial Doppler (TCD) cerebral blood flow chart

The initial cerebral blood flow of the pigs was measured after sedation but before tracheal intubation. During ex vivo NMP, cerebral blood flow measurements of the perfused brains were taken every hour. These measurements were conducted on the left middle cerebral artery (MCA) using TCD with a 1.6MHz probe (EMS-9PB, Delica, Shenzhen, China) through ocular windows at a depth of 4–4.5cm.

S100-β measurement

Perfusate samples from the arterial perfusion line were collected and shaken. Samples of 80mL were applied onto the chip of the S100-β Rapid Test Kit (Immunochromatography) (Wuhan Easy Diagnosis Biomedicine, China). The chip was incubated for 15min and analyzed in an immunoassay analyzer (QMT8000, Wuhan Easy Diagnosis Biomedicine Co., China).

Tissue processing and histology

RNA-seq sample preparation, library construction, and data analysis

Total RNA was extracted from the cortical tissues of the frontal and temporal lobes of 13 pigs (n=6 in the BWI group and n=7 in the BLWI-30 group). Each pig served as a biological replicate. Cerebral cortex tissue samples with an RNA Integrity Number (RIN)<5, indicating poor data quality, were excluded from the study. There were 10 remaining biological replicates in the frontal lobes and nine in the temporal lobes. The cDNA libraries were constructed using a strand-specific RNA-seq protocol. The libraries were subjected to high-throughput sequencing using DNBSeq T7 platforms.

RNA-seq reads were mapped to the reference genome (susScr11) using alignment algorithms (STAR version 2.7.3a), and gene expression levels were quantified using established bioinformatics pipelines (featureCounts Version 2.0.1). Differential gene expression analysis was performed to identify genes exhibiting significant changes in expression between the experimental conditions, using a threshold of log2(fold change) >1 and a P value<0.05. GO Processes analysis was analyzed using MetaCore.

Sample preparation for Liquid chromatography-tandem mass spectrometry (LC-MS/MS) from brain tissues

A brain tissue sample weighing 25mg was placed in the Eppendorf tube, to which 500μL of extract solution (methanol:acetonitrile:water = 2:2:1, with isotopically labeled internal standard mixture) was added. Then, the samples were homogenized and sonicated thrice in an ice-water bath and incubated for 1h at −40°C and centrifuged at 12,000rpm for 15min at 4°C. The resulting supernatant was transferred to a fresh glass vial for analysis. The quality control (QC) sample was prepared by mixing an equal aliquot of the supernatants from all the samples. LC-MS/MS analyses were performed using an UHPLC system (Vanquish, Thermo Fisher Scientific) equipped with a UPLC HSS T3 column (2.1mm × 100mm, 1.8 μm) coupled to an Orbitrap Exploris 120 mass spectrometer (Orbitrap MS, Thermo). The ESI source parameters included a capillary temperature of 320°C, spray voltage set at 3.8kV (positive) or 3.4kV (negative), and collision energies of 10/30/60 in NCE mode.

Metabolites extraction for perfusate serum

Triplicate perfusate samples from the arterial perfusion line were collected after 2 and 3h of NMP in the BOR and LABR groups. 100μL of the sample was transferred to an EP tube, and 400μL extract solution (acetonitrile:methanol = 1:1) containing internal standard (L-2-Chlorophenylalanine, 2μg/mL) was added. After vortexing for 30s, the samples were sonicated for 5min in an ice-water bath. Then, the samples were incubated at −40°C for 1h and centrifuged at 10,000rpm for 15min at 4°C. 425μL of supernatant was transferred to a fresh tube and dried in a vacuum concentrator at 37°C. Then, the dried samples were reconstituted in 200μL of 50% acetonitrile by sonication on ice for 10min. The mixture was then centrifuged at 13,000rpm for 15min at 4°C, and 75μL of supernatant was transferred to a fresh glass vial for LC/MS analysis. The QC sample was prepared by mixing an equal aliquot of the supernatants from all the samples.

LC-MS/MS analyses to detect metabolites in the perfusate serum

UHPLC separation was carried out using an Agilent 1290 Infinity series UHPLC System (Agilent Technologies) equipped with a UPLC BEH Amide column (2.1 × 100mm, 1.7 μm, Waters). The mobile phase consisted of 25mmol/L ammonium acetate and 25mmol/L ammonia hydroxide in water (pH = 9.75) (solvent A) and acetonitrile (solvent B). The elution gradient was set as follows: 0–1.0min, 95% B; 1.0–14.0min, 95%–65% B; 14.0–16.0min, 65%–40% B; 16.0–18.0min, 40% B; 18.0–18.1min, 40%–95% B; 18.1.1–23.0min, 95% B. The flow rate was set at 0.5mL/min. An Agilent 6495 triple quadrupole mass spectrometer (Agilent Technologies) was applied for assay development. The capillary voltage settings were +3000V for the positive ion mode and −2500 V for negative ion mode. The gas (N2) temperature was maintained at 170°C with a flow rate of 16L/min. The sheath gas (N2) temperature was set at 350°C with a flow rate of 12L/min. Nebulizer pressure was maintained at 40psi, and the fragmentor voltage was set at 380V.

LC-MS/MS analysis was conducted to analyze metabolites in both in vivo brain tissues and ex vivo perfusate

The data were analyzed using SAS 9.0 (SAS Institute Inc.). The resultant three-dimensional data involving the peak number, sample name, and normalized peak area were fed into SIMCA 14.1 (Sartorius Stedim Data Analytics AB; Umea, Sweden) for principal component analysis (PCA) and OPLS-DA. PCA depicted the distribution of the original data. To obtain an enhanced level of group separation and to better understand the variables responsible for classification, supervised OPLS–DA was applied. This permutation test was conducted to validate the model. Based on OPLS–DA, a loading plot was constructed to illustrate the variables’ contributions to the differences between the two groups. To refine this analysis, the first principal component of VIP was obtained. If P<0.05, then the variable was identified as a significantly differential metabolite between the groups. Using the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg), we obtained the metabolic pathways associated with each differential metabolite.

Exclusion criteria

In the results regarding the infarct area ratio, data from one pig in the BWI-30 group was identified as an outlier according to GraphPad’s outlier calculator (https://www.graphpad.com/quickcalcs/grubbs1/). Therefore, the infarct area ratio and all histological data from this pig were excluded from the final analysis in this study. Cerebral cortex tissue with an RNA integrity number (RIN)<5, indicating poor data quality, was excluded. Outlier data based on PCA were excluded from the metabolomic profile of cerebral cortex tissue. No data were excluded in the ex vivo experiments (Appendix Fig. S6).

Statistical analysis

All data are presented as either the mean±SEM or as the median with interquartile range, as appropriate. T tests and Mann–Whitney test were conducted for statistical analyses using GraphPad Prism v9.4.0. For comparisons between multiple experimental groups, ordinary 2-way analysis of variance (ANOVA) followed by Sidak’s multiple comparison test was employed. P<0.05 was considered statistically significant. When differences were not statistically significant, no indication was reported in the figures. The number of repeats performed and the statistical tests used were showed in the figure legends. Before the experiment, the pigs were randomly grouped. Infarct area measurements, injury scores, and cell counting (active neurons, nucleated blood cells, CD45+ cells, HIF-1α+ cells, HSP70+ cells, NRF2+ cells, and OGG+ cells) were performed blindly.

Graphics

Figures 1A,B, ​A,B,3A,3A, S3A, and synopsis graphics were created with BioRender.com, Adobe Illustrator, and Keynote.

This study was supported by grants from the National Natural Science Foundation of China (81970564, 82170663, 82370664, to Dr Guo; 82070670, to Dr He; 82300744, to Dr Xu), Guangdong Basic and Applied Basic Research Foundation (2024B1515040011 to Dr Guo), the Guangdong Provincial Key Laboratory Construction Projection on Organ Donation and Transplant Immunology (2023B1212060020, to Dr He), Guangdong Provincial International Cooperation Base of Science and Technology (Organ Transplantation) (2020A0505020003, to Dr He), and the Science and Technology Program of Guangdong (2020B1111140003, to Dr He). This study was also supported by the China organ transplantation development foundation (to Dr Guo) and the MedicalScientificResearchFoundationofGuangdongProvinceofChina (A2020275 to Dr Yin).