Research progress on the interaction between multiple organ-brain axes and perioperative neurocognitive disorders: a narrative review

Multiple factors (Mc Loughlin and Hinchion, 2023) during the perioperative period can significantly affect the diversity, function, and homeostasis of the gut microbiota. First, the type and duration of surgery play important roles. Compared to minor and shorter surgeries, major surgeries such as cardiac and abdominal procedures have a higher probability of causing dysbiosis. Additionally, the longer the duration of CPB, the higher the level of plasma lipopolysaccharide (LPS), which exacerbates systemic inflammation. Stein-Thoeringer CK (Stein-Thoeringer et al., 2025) and colleagues found that gastrointestinal surgeries, such as pancreatoduodenectomy, can lead to excessive proliferation of enterococci, thereby increasing the risk of postoperative infections. Furthermore, anesthesia and medications also influence the gut microbiota. General anesthetics, such as sevoflurane (Liu et al., 2022), can increase the likelihood of dysbiosis, whereas regional anesthesia, such as epidural anesthesia, has a less disruptive effect on the microbiota and may be more protective. The use of antibiotics, such as cephalosporins (Yin et al., 2025), during the perioperative period can significantly reduce the diversity of gut microbiota and promote the colonization of drug-resistant bacteria. Luo et al. (2021) found in studies on elderly mice that cefazolin sodium can alter the gut microbiome and increase the permeability of the BBB, thereby affecting postoperative cognitive function. Patient factors, including age and nutritional status, also influence the homeostasis of the gut microbiota. Research by Liufu et al. (2020) indicates that the alterations in gut microbiota caused by surgical anesthesia exhibit age-dependent changes, characterized by a significant decrease in both the abundance and diversity of the microbiota with increasing age. All of the aforementioned factors can contribute to a decline in cognitive function by affecting the composition and function of the gut microbiota.

Neuroinflammation is considered a major hallmark of PND (Subramaniyan and Terrando, 2019; Yang et al., 2024). The pathophysiological changes during the perioperative period and the stress response induced by surgical anesthesia may lead to dysbiosis of the gut microbiota, which in turn causes central inflammation through neural, immune, and endocrine pathways, ultimately resulting in the occurrence of PND. (1) Immune pathways. Perioperative stress, anesthesia, and antibiotic use can lead to dysbiosis of the gut microbiota, characterized by an increased ratio of enterococci to bacilli and elevated plasma LPS levels. A prospective study showed that intravenous injection of LPS is sufficient to directly induce significant systemic inflammatory responses in healthy volunteers, subsequently leading to decreased attention, memory, and a delirious state (van den Boogaard et al., 2010). This demonstrates that the translocation of gut-derived LPS is one of the key initiating factors of perioperative neuroinflammation and delirium. Mechanistically, LPS and the subsequent increase in pro-inflammatory factors can disrupt the integrity of the BBB. A critical experimental study showed that lateral ventricle injection of LPS significantly reduced the expression level of the tight junction core protein claudin-5 in the microvessels of the mouse cortex by about 40% within 24 h (Banks et al., 2015). It also increased the permeability of the blood–brain barrier to sucrose by approximately 75% (Banks et al., 2015). Interventional studies have confirmed that supplementation with probiotics such as Lactobacillus and Bifidobacterium not only restores gut homeostasis but also reduces the levels of inflammatory factors in the hippocampus by about 35–45%, significantly improving performance in cognitive behavioral tests—exploratory behavior in the elevated plus maze and open field test is enhanced by approximately 50–60% (Pei et al., 2024). (2) In terms of neural pathways, the metabolic products of gut microbiota, such as short-chain fatty acids (SCFAs), and the vagus nerve are key mediators. Brain-derived neurotrophic factor (BDNF) (Albini et al., 2023) maintains the growth and development of brain neurons. However, under stress conditions, perioperative dysbiosis of gut microbiota can be transmitted to the brain via the vagus nerve, inhibiting the expression of BDNF in hippocampal neurons. This makes neurons more susceptible to oxidative stress damage, thereby leading to changes in cognitive function. At the same time, gut microbial metabolites (Dalile et al., 2019), such as SCFAs, directly activate vagal afferent fibers, transmitting signals to the nucleus of the solitary tract. This, in turn, affects emotion- and cognition-related brain regions such as the amygdala and hippocampus. SCFAs (Que et al., 2024; Wu et al., 2020) improve cognitive function by inhibiting neuroinflammation and promoting neurogenesis. Animal experiments (Bravo et al., 2011) have shown that vagotomy can completely block the anxiolytic effects produced by probiotics (such as Lactobacillus rhamnosus), demonstrating the necessity of this pathway in gut-brain communication. (3) Endocrine pathways. Certain intestinal epithelial cells have endocrine or paracrine functions, such as enterochromaffin cells that produce 5-hydroxytryptamine, a precursor of the central neurotransmitter 5-HT (Shin and Kim, 2023). Surgical anesthesia disrupts gut microbiota and impairs the synthesis and metabolic homeostasis of neurotransmitters involving intestinal endocrine cells and gut microbiota (Qu et al., 2024). This disruption can lead to fluctuations in 5-HT and GABA levels, thereby affecting perioperative mood, behavior, and learning and memory abilities. The gut microbiota can also regulate the secretion of glucagon-like peptide-1 (GLP-1) and peptide YY (PYY), and receptors for these peptides are widely distributed in the hippocampus and prefrontal cortex, affecting appetite and cognitive function (Tolhurst et al., 2012). GLP-1 enhances synaptic plasticity through the cAMP-PKA pathway. Perioperative gut dysfunction can lead to a decrease in postoperative GLP-1 levels, which is significantly related to memory impairment (Ren et al., 2024).

In summary, physiological changes during the perioperative period and stress responses triggered by surgical procedures may lead to dysbiosis of the gut microbiota, thereby affecting brain function (Figure 2). Therefore, regulating gut microbiota may help alleviate perioperative neurocognitive disorders and improve patients' cognitive function and quality of life.

Recent studies have shown that the types, concentrations, and dynamic balance of liver metabolites are closely related to neurocognitive function. The disruption of key metabolites such as SCFAs, bile acids, ammonia, and ketone bodies has been confirmed to have a significant association with the occurrence and development of cognitive dysfunction, neurodegenerative diseases, and mental disorders.

SCFAs, primarily produced by gut microbiota fermentation of dietary fibers (such as acetate, propionate, and butyrate), can cross the blood–brain barrier and directly act on microglia and neurons. They regulate the activity of histone deacetylase enzymes (HDACs) and promote the expression of BDNF in the brain, thereby improving synaptic plasticity as well as learning and memory functions (Silva et al., 2020). In chronic liver diseases such as cirrhosis, dysbiosis of the gut microbiota often leads to a significant reduction in SCFA levels. SCFAs are neuroprotective, and their deficiency is considered one of the important factors contributing to related cognitive impairment. Therefore, supplementation with butyrate is regarded as a promising therapeutic strategy (Dalile et al., 2019). Animal experiments have shown that a reduction in SCFAs can damage the blood–brain barrier and promote glial cell activation; however, supplementation with exogenous butyrate can significantly alleviate blood–brain barrier damage in model mice (Xiao et al., 2022). Furthermore, clinical studies have indicated that in patients with cirrhosis, circulating butyrate levels have a moderate to strong significant negative correlation with key inflammatory mediators, LPS and IL-6 (LPS: RS = −0.58; IL-6: RS = −0.49; both p < 0.001) (Juanola et al., 2019). Thus, the deficiency of butyrate associated with liver disease constitutes an important upstream risk factor for perioperative cognitive impairment. In addition, SCFAs may influence mood and decision-making functions by activating free fatty acid receptors (FFAR2/3) to regulate the activity of serotonergic neurons (Yano et al., 2015).

Bile acids are the end products of cholesterol metabolism in the liver. Recent studies have found that they can regulate neuroinflammation and mitochondrial function by activating the Farnesoid X receptor (FXR) and the TGR5 receptor (Xie et al., 2018). When liver function is impaired, the accumulation of secondary bile acids, such as deoxycholic acid, can induce excessive activation of microglia. This activation promotes the release of IL-1β through the NF-κB pathway, thereby exacerbating neuroinflammation (MahmoudianDehkordi et al., 2019). In contrast, ursodeoxycholic acid (UDCA) has neuroprotective effects by inhibiting α-synuclein aggregation. Moreover, clinical trials have suggested that UDCA can slow cognitive decline in early Parkinson's disease (PD) patients (Qi et al., 2021).

Ammonia is a product of protein metabolism that involves the liver. When liver function is impaired, accumulated ammonia can interfere with the activity of glutamine synthetase in astrocytes, disrupt neurotransmitter cycling, induce brain edema and bioenergetic dysfunction, directly damage the hippocampus and prefrontal cortex, and affect memory and executive function (Hadjihambi et al., 2017). In addition, hyperammonemia may also lead to calcium overload by activating NMDA receptors, promote pathological tau protein hyperphosphorylation, and contribute to Alzheimer's disease-like changes (Rodrigo et al., 2010), ultimately resulting in neurocognitive dysfunction. Although hyperammonemia is considered one of the core mechanisms underlying delirium in patients with liver disease, its general applicability as a predictive indicator should be viewed with caution. A systematic review and meta-analysis focused on liver transplant patients clearly pointed out that preoperative ammonia levels have no significant correlation with the occurrence of postoperative delirium. In contrast, the patient's baseline liver function status, indicated by high Child-Pugh or MELD scores, advanced age, history of preoperative hepatic encephalopathy, and duration of surgery are more definitive risk factors (Zhou et al., 2021).

In addition, abnormal liver function causes metabolic disorders involving metal ions such as manganese and copper, leading to excessive generation of free radicals, inhibiting the activity of superoxide dismutase (SOD) and inducing lipid peroxidation in neurons (Jiang et al., 2009), directly affecting the function of the CNS.

Liver dysfunction can amplify systemic inflammatory responses. Patients with liver dysfunction during the perioperative period exhibit higher levels of inflammation compared to healthy individuals, accompanied by intestinal barrier damage and gut microbiota dysbiosis. LPS entering the liver activates Kupffer cells, which then release increased amounts of pro-inflammatory cytokines (Yang et al., 2024; Liu et al., 2022; Fukui, 2015). These pro-inflammatory cytokines signal through the vagus nerve or directly cross the BBB, causing central inflammation and impairing synaptic plasticity (Llansola et al., 2024). Animal experiments have confirmed that mice with chronic liver conditions, such as cirrhosis, show significantly elevated levels of IL-1β in the hippocampus after surgical trauma, which is associated with spatial memory impairment (Cao et al., 2010). Clinical data indicate that liver disease patients with elevated preoperative IL-6 levels have more than double the risk of developing postoperative delirium (OR > 2.0) (Capri et al., 2014). Based on this mechanism, preoperative anti-inflammatory interventions in liver disease patients have shown promising potential. However, a randomized controlled trial in liver resection patients demonstrated that although preoperative administration of high-dose corticosteroids significantly suppressed the intraoperative increase of IL-6 levels by over 50%, it did not significantly reduce the overall incidence of postoperative delirium (Awada et al., 2022). This suggests that the pathophysiology of delirium during liver resection may involve more complex inflammatory mechanisms beyond IL-6 that warrant further investigation.

In section 3.2.2, it is mentioned that SCFAs derived from gut microbiota, such as butyrate (Silva et al., 2020; Dalile et al., 2019), are absorbed into the liver via the portal vein and then distributed to the brain through hepatic blood circulation. These SCFAs exhibit neuroprotective effects by inhibiting HDAC activity and promoting BDNF expression. Dysbiosis of gut microbiota leads to a metabolic imbalance and reduced levels of SCFAs in patients with liver disease, which may weaken their inhibitory effects on neuroinflammation (Xu et al., 2024). The role of the gut-liver-brain axis—comprising gut barrier disruption, endotoxin translocation, impaired hepatic detoxification, and the resulting neuroinflammation cascade—in central inflammation and neurocognition is becoming clearer. Recent clinical trials have shown that perioperative supplementation with probiotics can reduce the incidence of delirium by 48%, suggesting that regulating the gut-liver-brain axis is a potential intervention strategy (Fukui, 2015).

The risk of cognitive changes significantly increases in patients with hepatic dysfunction after receiving anesthesia. In hepatic dysfunction, the sensitivity of GABA receptors is enhanced, while the metabolism of dopamine and serotonin is abnormal, which may lead to a postoperative imbalance between excitation and inhibition (Hadjihambi et al., 2017). The metabolism of anesthetic drugs, such as benzodiazepines and propofol, depends on hepatic enzymes, such as CYP450. In patients with impaired liver function, the clearance rate decreases, prolonging the duration of drug action and further exacerbating cognitive suppression (Elvir-Lazo et al., 2021; Tsai et al., 2015). These findings demonstrate that liver function affects cognitive function in patients after anesthesia. Retrospective analyses also show that the incidence of cognitive impairment after surgery is significantly higher in patients with Child-Pugh grade B or C cirrhosis compared to those with grade A, and this is significantly related to the dosage of anesthetic drugs used during surgery (Zhou et al., 2023).

In summary, hepatic dysfunction affects the CNS through metabolic, inflammatory, and pharmacokinetic pathways, significantly increasing the risk of PND (Figure 3).

The brain tissue is the organ with the highest oxygen consumption in the human body (accounting for 20% of total oxygen consumption), and its energy supply almost entirely relies on aerobic metabolism (Amin et al., 2024; Li C. et al., 2022). Surgery-related respiratory suppression, the effects of anesthetic drugs, or pre-existing pulmonary diseases can lead to perioperative hypoxia in patients. When the arterial partial pressure of oxygen (PaO₂) falls below 60 mmHg or chronic hypoxia (SpO₂ < 90% sustained for more than 5 min) occurs (Ding et al., 2018; Zhang K. et al., 2024; Jing et al., 2022), mitochondrial function in the hippocampus is impaired, resulting in a 40–60% reduction in ATP production. Insufficient ATP leads to the inhibition of presynaptic glutamate transporter function, causing extracellular glutamate concentration to increase several dozen-fold (Jing et al., 2022), which in turn triggers the degradation of neuronal cytoskeletal proteins. At the same time, during hypoxia, brain tissue initiates glycolysis, and lactic acid production increases by 3–5 times. This metabolic acidosis further inhibits GABA aminotransferase activity, affecting the metabolism of inhibitory neurotransmitters (Mácha et al., 2024). Clinical studies have found that serum levels of hypoxia-inducible factor HIF-1α are significantly higher in patients with obstructive sleep apnea (OSA) who develop postoperative delirium (p < 0.01); elevated HIF-1α is an independent risk factor for PD (OR = 4.82) (Breitkopf et al., 2025). Additionally, the duration of mild hypoxemia (55 mmHg ≤ PaO₂ < 60 mmHg) accumulated during surgery is significantly associated with an increased risk of postoperative delirium. For every additional 10 min of cumulative exposure time, the risk of developing postoperative delirium increases by approximately 18% (OR = 1.18) (Ahrens et al., 2023). These findings indicate that preoperative chronic hypoxia load (reflected by hypoxia biomarkers) lays a pathological basis for the occurrence of delirium. Moreover, intraoperative hypoxia greatly enhances the brain's vulnerability to perioperative stress. Together, these factors lead to the occurrence of PND. Therefore, does monitoring regional cerebral oxygen saturation (rSO₂) during the perioperative period have value in predicting and preventing postoperative delirium? Some studies suggest that a decrease in brain oxygen is a risk factor; however, there are also high-quality prospective observational studies that have yielded negative results, finding no association between perioperative rSO₂ parameters and delirium risk (Fischer-Kumbruch et al., 2022). This inconsistency may be related to the heterogeneity of the study population, types of surgery, and methods of delirium assessment.

Perioperative respiratory dysfunction leads to abnormal levels of carbon dioxide in the blood, either hypercapnia or hypocapnia, each of which is considered an important modifiable factor that induces and promotes PND. The core mechanism lies in the fact that carbon dioxide (Meng and Gelb, 2015), a potent vasodilator and neuromodulator that can freely penetrate the BBB, significantly affects neuronal excitability through direct and indirect pathways. This interference disrupts the normal function of cognitive-related brain regions.

Hypocapnia has been supported by multi-level evidence as a clear risk factor for PND. Early studies found that intraoperative end-tidal hypocapnia (EtCO₂ < 30 mmHg) is an independent predictor of postoperative delirium (OR = 3.2, 95% CI 1.2–8.9) (Mutch et al., 2018). This finding was subsequently confirmed in more precise studies: the latest evidence shows a clear dose-dependent relationship between intraoperative hypocapnia (PaCO₂ < 30 mmHg) and the risk of delirium, with each additional 10 min of intraoperative hypocapnia exposure increasing the risk of delirium by 30% (OR = 1.30) (Ahrens et al., 2023). Mechanistically, hypocapnia induces cerebral vasoconstriction, leading to reduced cerebral blood flow, which may cause covert ischemia in brain tissue and interfere with neuronal energy metabolism and excitability balance.

The impact of hypercapnia on PND presents a complex dual nature. Randomized controlled trials indicate that mild controlled hypercapnia (PaCO₂ 50–55 mmHg) can significantly increase intraoperative cerebral oxygen saturation by approximately 5–7%, demonstrating physiological benefits in improving brain oxygenation (Wong et al., 2020). However, this benefit has not been consistently validated in clinical outcomes. Recent studies have found that while controlled hypercapnia (EtCO₂ 40–50 mmHg) can improve brain oxygen saturation, it fails to reduce the incidence of postoperative delirium (Li et al., 2025). Meanwhile, basic research has clearly revealed hypercapnia's direct neurotoxicity. The decrease in cerebrospinal fluid pH caused by hypercapnia (Zhou et al., 2023) directly modulates the gating of voltage-gated sodium channels (Nav1.2) and TWIK-related acid-sensitive potassium channels (TASK-1/3), resulting in an overall reduction of neuronal excitability by 30–40% (O'Donohoe et al., 2019; Linden et al., 2007). Additionally, the release of presynaptic vesicles is significantly reduced (Tachibana et al., 2013), particularly affecting glutamatergic synapses, which enhance inhibitory neural impulse transmission (Sylvestre et al., 2022), exacerbating postoperative sedation and consciousness disorders. Maintaining moderate hypercapnia (PaCO₂ ≈ 60 mmHg) for 2 h can lead to a suppression of long-term potentiation (LTP) amplitude in the mouse hippocampus by approximately 45%, and cause a decline in spatial memory performance by about 26% (Tachibana et al., 2013; Tregub et al., 2023). This indicates that hypercapnia exceeding the physiological compensation range—that is, the body's ability to maintain acid–base balance—directly affects ion channel function and has clear neurotoxicity.

It is noteworthy that clinical observational studies have found that, compared to the clear risks of hypocapnia, mild to moderate hypercapnia during surgery (PaCO₂ > 45 mmHg) is not significantly associated with an increased risk of delirium (Ahrens et al., 2023). Hypocapnia poses a clear risk, while hypercapnia has both physiological benefits and potential neurotoxicity, suggesting that in clinical practice, avoiding hypocapnia caused by excessive ventilation may be prioritized over concerns regarding mild hypercapnia.

Pulmonary infection not only leads to a systemic inflammatory response but also exacerbates cognitive function impairment by affecting oxygenation status and increasing neuroinflammation (Gao et al., 2023). During pulmonary infection, pathogen-associated molecular patterns (PAMPs) activate alveolar macrophages (Gao et al., 2023; Ryu et al., 2023). These activated macrophages release pro-inflammatory factors such as IL-1β, IL-6, and TNF-α, which can enter the CNS through disrupted barriers, directly damaging neurons and synaptic function (Yang et al., 2024; Dalile et al., 2019; Llansola et al., 2024). Additionally, pulmonary infection causes hypoxia, and the abnormal or excessive activation of HIF-1α can disrupt the integrity of the BBB, promoting inflammatory cell infiltration and further exacerbating neural damage (Arias et al., 2023). Animal experiments have shown that 24 h after infection with Streptococcus pneumoniae, the activation of microglia in the hippocampal region of mice is significantly increased, and levels of pro-inflammatory cytokines such as TNF-α and IL-6 also rise significantly, which is accompanied by a decline in spatial memory ability (Zhang et al., 2021). Furthermore, studies targeting human populations have confirmed that patients infected with respiratory viruses such as SARS-CoV-2 (COVID-19) and SARS-CoV exhibit significant activation of inflammatory signaling pathways in their cerebrospinal fluid (Reinhold et al., 2023; Jarius et al., 2022), along with a marked increase in the incidence of acute cognitive dysfunction (with a pooled incidence rate of delirium as high as 21.7%) (Pranata et al., 2021). This highlights the significant impact of pulmonary infection on brain function through systemic inflammation and other mechanisms.

These mechanisms work synergistically to form a vicious cycle. Hypoxia exacerbates the inflammatory response, which further impairs lung function. Meanwhile, hypercapnia or hypocapnia resulting from reduced lung function aggravates cerebral metabolic disorders, further promoting the occurrence of PND (Figure 4). Clinical interventions should comprehensively improve oxygenation, optimize ventilation strategies, and control inflammation to interrupt the pathological signaling along the lung-brain axis.

In patients with cardiac insufficiency during the perioperative period, the decrease in cardiac output inevitably leads to a reduction in CBF, leading to PND (Ju et al., 2023; Bowden et al., 2022). Cardiac-origin cerebral hypoperfusion damages neurons through dual pathways. Firstly, there is the direct energy crisis, which refers to the state of cerebral hypoperfusion leading to downregulation of neuronal glucose transporter GLUT3 expression (Kisler et al., 2017), reduced activity of mitochondrial complex IV (Xie et al., 2013), and decreased ATP production (Bhushan et al., 2021), ultimately resulting in a supply–demand imbalance of brain energy. Secondly, the decoupling of the neurovascular unit (NVU) occurs (Iadecola, 2017). The reduction in CBF velocity decreases the shear stress on the vessel wall. This leads to a reduction of up to 55% in the expression of the key tight junction protein claudin-5 of the BBB through the PECAM-1/Gαq/PLCβ pathway (Ma et al., 2023). This disrupts the structural and functional connections of the NVU, thereby increasing the permeability of the BBB to peripheral toxins, which can then enter the CNS and damage neurons. Moreover, imaging studies have confirmed that under hypotensive stress, delirium-related brain regions, such as the hippocampus and cingulate gyrus, exhibit significant hypoperfusion, as evidenced by typical hypoperfusion imaging signals (Mutch et al., 2020). This directly proves that hemodynamic disturbances can lead to hypoperfusion in delirium-related brain regions (Mutch et al., 2020). Importantly, clinical intervention data show that for patients with bradycardia, improving cardiac output can enhance cognitive function scores measured by standardized neurocognitive tests by approximately 21%, accompanied by a reduction in serum IL-1β concentration of about 25% (Martis et al., 2021).

The heart also affects the CNS through the "heart-brain inflammation axis." Cardiac vagal afferent fibers express IL-1R1 and TLR4 receptors, which can transmit peripheral inflammatory signals to the solitary nucleus (Lammers-Lietz et al., 2025). This process segmentally activates microglia, resulting in a 2-3-fold increase in Iba-1 positive cells in the hippocampus (Xia et al., 2024) and significantly elevates the level of CNS inflammation. Patients who experience cardiac arrest (CA) and achieve the return of spontaneous circulation (ROSC) often suffer from severe neuroinflammation due to global cerebral ischemia–reperfusion injury, with excessive activation of microglia in multiple brain regions (Chen et al., 2025). Furthermore, patients with heart failure generally exhibit a systemic inflammatory state (serum IL-6 is commonly elevated) (Gui and Rabkin, 2023; Tromp et al., 2018), and surgical trauma can further amplify this inflammatory response. Baseline data from a multicenter randomized controlled trial indicate that in the control group, patients who did not receive neuroprotective interventions showed that the serum level of the neuroinflammatory marker S100β can increase to approximately 2.8 times the preoperative baseline value within 24 h post-surgery (Liu et al., 2023); the magnitude of this increase is significantly correlated with the incidence of postoperative delirium (p < 0.01) (Dunne et al., 2021). Long-term patients with cardiac insufficiency experience a 1–2% annual reduction in hippocampal volume, with a faster decline in memory scores compared to peers, which may also be related to chronic inflammation (Traub et al., 2024).

Insufficient cerebral blood flow leads to neuronal energy metabolism disorders, and chronic inflammation activates microglia through cytokines, causing CNS inflammation. These two factors intertwine, forming a vicious cycle: low perfusion exacerbates inflammation and oxidative stress, which in turn further impairs vascular regulatory function, disrupts energy metabolism, and ultimately plays an important role in the occurrence and development of PND in the heart-brain axis (Figure 5).

The mechanism by which the spleen-brain axis regulates PND involves multiple pathways, the most important of which may be mediating the balance between peripheral inflammation and the CNS. Because the spleen is the largest lymphoid organ in the body, perioperative stress causes it to release inflammatory mediators that promote neuroinflammation, leading to cognitive decline (Feng et al., 2017). The spleen can influence the progression of neuroinflammation by regulating the polarization of peripheral immune cells into M1 and M2 macrophage phenotypes (Aiello et al., 2020); M1 macrophages promote pro-inflammatory responses, while M2 macrophages have anti-inflammatory and neuroprotective effects. The dual role of the splenic branch of the vagus nerve in anti-inflammatory and pro-inflammatory responses also plays an important role. On one hand, the spleen exerts neuroprotective effects through the cholinergic anti-inflammatory pathway via the vagus nerve. The efferent branches of the splenic vagus nerve release acetylcholine, which binds to α7nAChR on splenic macrophages, inhibiting NF-κB activation and reducing the release of pro-inflammatory factors from the spleen (Pavlov et al., 2018). Animal studies have shown that electrical stimulation of the vagus nerve to activate the cholinergic anti-inflammatory pathway in the spleen can significantly reduce TNF-α levels in the spleen and brain (by approximately 40%) (Mihaylova et al., 2012), and correspondingly improve postoperative cognitive function, with behavioral test scores increasing by about 35% (Mihaylova et al., 2012). Meta-analysis also confirmed that animal behavioral test performances showed highly significant improvement compared to the control group, with a large overall effect size (SMD = −2.87, p < 0.001) (Zhang H. et al., 2022). On the other hand, peripheral inflammatory mediators activate specific receptors on the afferent fibers of the splenic vagus nerve, transmitting peripheral inflammatory signals to brain regions associated with cognition such as the hippocampus and prefrontal cortex via the nucleus of the solitary tract. This leads to exacerbation of local neuroinflammation and thereby impairs synaptic plasticity (Ben-David et al., 2020). In animal experiments, aged rats significantly exhibited a postoperative decline in spatial learning and memory ability of about 30–40% and impaired executive function (Zhang et al., 2023b). The splenic vagus nerve pathway is a complex bidirectional signaling pathway. Its efferent branches exert anti-inflammatory and neuroprotective effects, while its afferent branches are responsible for transmitting pro-inflammatory signals, serving as a key bridge mediating peripheral inflammation leading to CNS inflammation and cognitive impairment (Zhang et al., 2023b). In summary, the spleen is a critical link in the amplification of neuroinflammatory responses.

The spleen, as a complex immune organ, contains various immune cells, such as splenic macrophages, dendritic cells, mesenchymal stem cells (MSCs) and B lymphocytes. These different types of cells can remotely regulate gene expression in the brain by secreting exosomes carrying microRNAs (miRNAs) and other bioactive molecules. These exosomes affect synaptic plasticity and neuronal survival (Pitt et al., 2016; Muntasell et al., 2007), thereby playing a dual regulatory role in both the initiation and progression of PND. Under physiological conditions, the spleen can secrete exosomes with neuroprotective effects: macrophage-derived miR-146a-5p can downregulate TLR4 expression in microglia, alleviating neuroinflammation (Zhang et al., 2023a); MSC-derived miR-125b can counteract Aβ-induced synaptic damage and promote synaptic plasticity (Zhang J. et al., 2022). Moreover, Deng et al. (2024) and others found that exosomes derived from splenic MSCs have a positive effect on nerve regeneration and cognitive function recovery. However, under pathological conditions, surgical trauma can activate splenic immune cells (especially red pulp macrophages and marginal zone B cells), leading to a large release of pro-inflammatory exosomes. These exosomes carry pro-inflammatory miRNAs such as miR-155-5p and miR-193a-5p (Fang et al., 2024). They can be taken up by neurons and microglia, where they inhibit the expression of BDNF. This inhibition causes sustained amplification of neuroinflammation and mediates a decline in cognitive function levels. Clinical studies have found that the level of exosomal miR-193a-5p in the plasma of PND patients after surgery is several times higher (approximately 3–5 times) than that of non-PND patients, and its expression level is positively correlated with the duration of delirium (R = 0.51) (Xu et al., 2025), suggesting it could serve as a potential biomarker. This indicates that perioperative stress leads to a shift in the "pro-inflammatory/anti-inflammatory" balance of splenic exosomes toward a pathological direction, increasing central inflammation levels and promoting the occurrence of PND.

In summary, the spleen-brain axis participates in the occurrence and development of PND through various mechanisms, including regulating inflammatory balance, neuroendocrine functions, and extracellular vesicles (Figure 6).

Various uremic toxins accumulated during renal function decline (Li et al., 2021; Yamamoto, 2019), including small, water-soluble molecules, middle molecular toxins, and protein-bound toxins, can damage the integrity of the BBB and directly injure neurons, ultimately promoting the progression of PND. The protein-bound toxin p-cresol sulfate (PCS) activates microglia by promoting the release of TNF-α and IL-1β, inducing neuroinflammation (Azargoonjahromi et al., 2025). In contrast, indole-3-propionic acid can inhibit the expression of synaptic plasticity-related proteins in the hippocampus, such as postsynaptic density protein 95 (PSD-95) and BDNF, impairing learning and memory function (Sun et al., 2021). β2-microglobulin (β2-MG) (Yamamoto, 2019), a middle molecular toxin, leaks into the CNS through the BBB, promoting the hyperphosphorylation of tau protein and facilitating the deposition of amyloid proteins in the brain alongside amyloid-beta (Aβ). Studies have found that blood levels of β2-MG in patients with renal insufficiency are negatively correlated with MMSE cognitive scores (β ≈ −0.10, p < 0.001) (Ke et al., 2025). In addition, the decline in renal clearance function leads to the accumulation of small, water-soluble molecules such as homocysteine (Hcy), which exacerbates oxidative stress by activating the NMDAR / ROS pathway, increasing the apoptosis rate of neurons by approximately 20–40% (Luzzi et al., 2022).

The endocrine disorders caused by renal insufficiency represent another important mechanism of the kidney-brain axis. Erythropoietin (EPO), secreted by the kidneys, has neuroprotective effects; however, its level decreases during renal insufficiency, which can lead to reduced neurogenesis in the hippocampus (Vittori et al., 2021). A phase III clinical trial showed that treatment with recombinant human EPO (rhEPO) in patients with chronic kidney disease increased their MoCA scores by an average of about 2.5 points (Ehrenreich et al., 2009). Additionally, secondary hyperparathyroidism (SHPT) is a significant complication of renal insufficiency. The hyperphosphatemia and elevated fibroblast growth factor 23 (FGF23) caused by SHPT may induce brain edema by disrupting cerebral vascular autoregulation and may also inhibit neuronal energy metabolism (Gong et al., 2020; Drew et al., 2020). Clinical studies have confirmed that preoperative vitamin D deficiency is an independent risk factor for PND in elderly patients, increasing the risk by 60 to 150% (OR = 1.6–2.5) (Hung et al., 2022). This association likely results from insufficient activity of renal 1α-hydroxylase, which leads to low levels of active vitamin D in the brain and subsequently reduces the synthesis of neurotrophic factors (Hung et al., 2022).

In recent years, exosome-mediated inter-organ communication has received widespread attention. Under stress conditions such as surgical trauma and ischemia–reperfusion injury, the kidneys can release a large number of exosomes (Lv et al., 2020) into the circulatory system. These exosomes carry specific miRNAs, such as miR-21-5p and miR-155, and express specific integrins on their surface, such as ITGα6β4 and ITGαvβ5. This enables the exosomes to selectively target brain endothelial cells and cross the BBB through transcytosis. Once in the brain, they regulate multiple inflammatory pathways and interfere with neuronal energy metabolism, which plays an important role in cognition and memory functions (Zang et al., 2019; Ianni et al., 2024). For example, pro-inflammatory miRNAs such as miR-155 carried by renal exosomes are key mediators in regulating neuroinflammation. Specifically, miR-155 (Zingale et al., 2021) targets the zinc finger protein Tristetraprolin (TTP), thereby enhancing the stability of IL-6 and IL-8 mRNA and further amplifying the neuroinflammatory cascade. In addition to pro-inflammatory effects, renal exosomes can also interfere with neuronal energy metabolism through molecules such as miR-668 (He and Zhang, 2020). In AKI animal models, exosomal miR-668 reduces the membrane localization of glucose transporters GLUT1 and GLUT3 by inhibiting the expression of insulin-like growth factor 1 receptor (IGF1R) in the hippocampus, leading to insufficient energy supply for neurons and impaired synaptic plasticity (Li S. et al., 2022). This metabolic disorder can further exacerbate Aβ deposition and tau protein hyperphosphorylation (Lang et al., 2023), forming pathological features similar to Alzheimer's disease. It is noteworthy that renal exosomes have bidirectional regulatory potential. Under physiological conditions, the kidneys secrete exosomes carrying anti-inflammatory miRNAs such as miR-146a-5p (Zhang et al., 2023a; Valiuliene et al., 2023), which alleviate neuroinflammation by inhibiting the IRAK1/TRAF6 signaling pathway. However, during perioperative stress conditions, this protective mechanism is often disrupted, resulting in a predominance of pro-inflammatory exosomes. Although there is currently insufficient clinical research directly targeting the association between renal-derived exosomes and PND, a large body of basic research evidence suggests that renal exosomes are a promising mediator connecting peripheral organ injury and CNS function, making them an important focus for future exploration as early warning biomarkers or intervention targets for PND.

In addition, the kidneys also regulate the brain's neuroinflammatory response through the renin-angiotensin system (RAS) (Chagas et al., 2025). Angiotensin II (Ang II) excessively activates the AT1 receptor (AT1R), promoting NADPH oxidase-dependent oxidative stress in the nervous system. Angiotensin receptor blockers (ARBs), such as losartan, can alleviate postoperative cognitive impairment (Althammer et al., 2023). These findings indicate that renal inflammatory responses are also key contributors to the development of PND.

In summary, the kidney–brain axis participates in the pathological process of PND through multiple mechanisms, including barriers involved in toxin clearance, endocrine disorders, and exosome signaling (Figure 7). The therapeutic potential of renal protection strategies in the prevention and treatment of PND may be a focus of future research.