All experimental procedures were conducted in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Fujian Medical University. Nine-week-old male Sprague‒Dawley rats (weight range: 300–400 g) were housed under standard laboratory conditions (12 h light/dark cycle, 22 ± 1 °C, and 55 ± 5% humidity) with ad libitum access to food and water.

A rat model of cardiopulmonary bypass (CPB) with deep hypothermic circulatory arrest (DHCA) was established based on an adapted methodology (19). The detailed surgical and perfusion protocol proceeded was as follows:

Preoperative preparation and anaesthesia: After a 6-hour fast, the rats were weighed and anaesthetized. Surgical anaesthesia was then maintained by continuous inhalation of 3% sevoflurane in 100% oxygen throughout the surgical procedures, except for the circulatory arrest period itself.

Twenty-one rats were subjected to the CPB/DHCA procedure. Twelve animals completed the full experimental protocol and were included in the final analysis. Nine rats were excluded: four because of technical failure during CPB establishment, two because of acute complications (e.g., air embolism), and three because of death within 24 h post-operation. At 24 h postmodelling, the rats were euthanized via pentobarbital sodium (50 mg/kg) for tissue harvesting. Upon harvest, target tissues were immediately snap-frozen in liquid nitrogen and stored at −80 °C for subsequent analysis. The hippocampus was then processed for transcriptomic profiling using RNA sequencing. Owing to its recognized susceptibility to ischemia, the hippocampus was chosen for molecular analysis as it is a sensitive indicator of global cerebral injury after DHCA (13, 19).

Total RNA was extracted using TRIzol (Invitrogen). The RNA concentration and purity were determined by measuring the A260/A280 ratio using a NanoDrop ND-1000 (Thermo Fisher Scientific, Waltham, MA, USA). RNA samples were immediately frozen and stored at −80 °C until use. The miRNA was sequenced by CloudSeq, Inc. (Shanghai, China). Small RNA libraries were constructed using a GenSeq® Small RNA Library Prep Kit (GenSeq, Inc.) following the manufacturer's instructions. Briefly, 3′ and 5′ adapters were subsequently ligated to the RNA samples. Subsequently, the adapter-ligated RNA was reverse transcribed into cDNA, followed by polymerase chain reaction (PCR) amplification. After amplification, miRNA fractions were selected from the cDNA libraries based on size and then sequenced on a platform.

After sequencing, the raw data were subjected to image analysis, base calling, and initial quality filtering. Quality was first controlled using the Q30 standard. Adapter sequences were removed using cutadapt (v1.9.3), retaining reads ≥15 nt after trimming. These reads were subsequently aligned to a combined pre-miRNA reference (including known miRBase entries and newly predicted pre-miRNAs) via Novoalign (v3.02.12), permitting up to one mismatch. Mature miRNA-mapped reads were counted as raw expression values and normalized using TPM (tags per million aligned miRNAs). Novel miRNAs were predicted using miRDeep2 (v2.0.0.5) on pooled trimmed reads from all the samples. miRNAs with an absolute fold change ≥1.5 and a P value ≤0.05 were considered as differentially expressed. Experimentally supported or predicted miRNA targets were identified using established tools, and miRNA‒target networks were visualized with Cytoscape (v2.8.0). A functional enrichment analysis was performed on the target genes of the top 20 differentially expressed miRNAs.

One hundred forty consecutive patients undergoing open surgical repair for AAD who were admitted to cardiac surgery center from January 2024 to June 2024 were enrolled. This retrospective study was reviewed and approved by the Fujian Medical University Union Hospital Ethics Committee and strictly complied with the Declaration of Helsinki. Informed consent was waived by the ethics committee because of the retrospective and anonymized nature of the data. The inclusion criteria were as follows: (1) age ≥18 years and (2) underwent open surgical repair. The exclusion criteria were as follows: (1) pre-existing neurological or psychiatric disorders (e.g., dementia, stroke, schizophrenia, or depression); (2) concurrent chronic hepatic or renal dysfunction; (3) preoperative shock or haemodynamic instability secondary to cardiac tamponade; and (4) death within 24 h after surgery. Eligible patients were divided into two groups: patients with PONC and patients without PONC. The diagnosis of stroke was made clinically and confirmed by computed tomography or magnetic resonance imaging. Neurological deficit severity was assessed using the National Institutes of Health Stroke Scale at admission. Postoperative delirium was assessed using the Richmond Agitation–Sedation Scale and Confusion Assessment Method for the ICU, while the Glasgow Coma Scale provided objective criteria for defining postsurgical coma.

Venous blood samples (5 mL) were collected from all participants using EDTA-coated tubes at two time points: pre-anaesthesia induction and 24 h postoperatively. Plasma was isolated by low-speed centrifugation (1,500×g, 4 °C, 15 min), after which the supernatant was aliquoted into cryovials and stored at −80 °C for subsequent biochemical assays.

The operation was performed under general anaesthesia via a sternal incision. General anaesthesia was induced with intravenous midazolam, sufentanil, and etomidate and maintained with sevoflurane and a continuous infusion of sufentanil and propofol. Muscle relaxation was achieved with rocuronium. During CPB, anaesthesia was maintained with propofol and midazolam. Haemodynamic monitoring included intra-arterial pressure and transesophageal echocardiography. After heparinization, CPB was established by venous and arterial cannulation. When the nasopharyngeal temperature had decreased to 32 °C, the ascending aorta was clamped, and 4 °C cold blood cardioplegia was infused for myocardial protection. Subsequently, the repair operation was performed. After cooling to the required temperature (nasopharyngeal temperature, 20–23 °C; rectal temperature, 23–26 °C), circulatory arrest began, and bilateral selective antegrade cerebral perfusion was performed. The flow rate was approximately 8–10 mL/kg/min. For arch repair, total arch replacement using a 4-branched Dacron graft with stented elephant trunk implantation or open triple-branched stent graft placement was performed, as previously reported (20, 21). The triple-branched stent graft technique was selected when preoperative CT measurements revealed that the diameters of the native aortic arch and arch vessels were 10%–20% smaller than those of the graft, and the distances between adjacent arch vessels matched the corresponding interbranch distances of the stent graft. Otherwise, total arch replacement with a 4-branched Dacron graft and a stented elephant trunk was performed.

For miRNA expression profiling, total RNA was isolated using the miRcute miRNA Isolation Kit (TIANGEN, Beijing, China; DP503), followed by first-strand cDNA synthesis using the miRcute Plus miRNA First-Strand cDNA Synthesis Kit (TIANGEN Biotech, Beijing, China; KR211-02), in strict accordance with the manufacturer's protocols. Quantitative PCR analysis was performed using the miRcute Plus miRNA SYBR Green qPCR Kit (TIANGEN Biotech, Cat# FP411-02) in strict compliance with standardized protocols. Serum miR-29 expression was quantified through by real-time PCR, where threshold cycle (Ct) values derived from RT‒PCR amplification kinetics were computationally transformed using the 2^−ΔΔCt algorithm. The relative expression metric was defined as ΔΔCt = [Ct (target miRNA) − Ct (U6 snRNA)] − [Ct (calibrator sample) − Ct (U6 snRNA)], with U6 snRNA serving as the endogenous normalization control. The miRNA primers used (Sunya Biotech) were as follows: miR-29a-5p forward, 5′-CCGCGACTGATTTCTTTTGGTGTTCAG-3′, and miR-29b-3p forward, 5′-CCGCTAGCACCATTTGAAATCAGTGTT-3′. The sequences were designed by Sunya Biotech (ISO-certified) following stem‒loop RT‒qPCR design principles.

Statistical analyses were performed using SPSS (version 20.0; IBM, USA). Normally distributed measurement data are expressed as the mean ± standard deviation, whereas non-normally distributed are expressed as the median and interquartile range. Two independent groups were compared using t tests or Mann‒Whitney U tests (for continuous variables) and chi‒square tests or Fisher's exact tests (for categorical variables). Significant univariate predictors (p < 0.05) were advanced to multivariate logistic regression for identifying independent PONC risk factors, and a nomogram was constructed. The prediction model was evaluated through internal validation using receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and confusion matrices. Survival probabilities over the 30-day observation period were estimated using the Kaplan‒Meier method, with between-group comparisons assessed via log-rank test. A two-tailed p-value < 0.05 was considered statistically significant.

To investigate miRNA dynamics under hypothermic circulatory arrest conditions, high-throughput sequencing coupled with comparative transcriptomic analysis of hippocampal tissues from the CPB/DHCA group and control group was used to characterize endogenous miRNA profiles. More than six hundred miRNAs were identified by high-throughput sequencing. Differentially expressed miRNAs were defined by a selection criterion of |FC|>1.5 with statistical significance (P < 0.05). Through this analytical pipeline, 31 miRNAs met the defined thresholds, comprising 10 upregulated and 21 downregulated miRNAs (Figure 2A). Notably, the changes in the expression of miR-29a-5p and miR-29b-3p were the most pronounced (Figure 2B).

Subsequent functional annotation was performed through integrated Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses to systematically characterize miRNA-mediated regulatory networks. Functional annotation analysis revealed the GO landscape across biological processes (BP), molecular functions (MF), and cellular components (CC), as depicted in Figures 2C,D. KEGG pathway profiling revealed predominant enrichment of differentially expressed genes (DEGs) in three critical pathways: MAPK signalling, tight junction, and gap junction assembly (Figures 2E,F). These pathways are functionally implicated in BBB homeostasis through coordinated regulation of endothelial cell–cell adhesion complexes and paracellular permeability (22, 23).

Between January and June 2024, 140 consecutive patients with AAD were initially enrolled in this study. Following rigorous screening, 45 patients were excluded based on the following predefined criteria: incomplete clinical records (n = 23), preexisting neurological/psychiatric comorbidities (n = 7), chronic hepatorenal insufficiency (n = 8), and mortality events (aortic rupture-related deaths, n = 4; death within 24 h after surgery, n = 3). Ultimately, 95 eligible patients composed the final analytical cohort. The baseline demographic and clinical characteristics of the study cohort are summarized in Table 1.

A comparative analysis of the perioperative parameters revealed that compared with the non-PONC cohort, the PONC cohort had significantly longer CPB times, extended aortic cross-clamp durations, longer mechanical ventilation requirements, and longer ICU stays (Table 2). Furthermore, in-hospital mortality rates were significantly higher in the PONC cohort (28% vs. 4.29%, p = 0.003). Preoperative biomarker analysis revealed significant intergroup disparities (Figure 3). Compared with non-PONC controls, the PONC cohort demonstrated marked increases in the expression of inflammatory mediators, such as CRP (p = 0.019), procalcitonin (PCT) (p = 0.027), IL-6 (p < 0.001), and miR-29b-3p (p < 0.001). Postoperative inflammatory marker levels also demonstrated similar trends (Figure 3). CRP (p = 0.001), PCT (p < 0.001), and IL-6 (p = 0.001) levels significantly increased in the PONC group. Conversely, miR-29a-5p expression was markedly decreased in PONC patients (p < 0.001).

Univariate logistic regression revealed multiple perioperative predictors of PONC, including preoperative inflammatory biomarkers, preoperative miR-29b-3p expression, postoperative miR-29a-5p expression, and CPB time. Notably, multivariate adjustment confirmed that preoperative miR-29b-3p expression (OR = 2.53, 95%CI 1.17–5.47, p = 0.018) and postoperative miR-29a-5p expression (OR = 0.21, 95%CI 0.05–0.96, p = 0.044) were independent risk factors for PONC development, whereas IL-6, CRP, and CPB time lost statistical significance in the multivariable model (Table 3).

A clinically oriented nomogram incorporating the identified independent predictors was developed (Figure 4A), followed by rigorous multidimensional validation. The nomogram exhibited strong discriminative ability, with an area under the ROC curve (AUC) of 0.867 (Figure 4B). At the optimal probability threshold, the model achieved high overall accuracy, sensitivity, and specificity, along with favorable predictive values and likelihood ratios. The diagnostic odds ratio and Youden index further supported its excellent diagnostic performance (Table 4). The model exhibited optimal calibration fidelity, as evidenced by the nonsignificant Hosmer–Lemeshow test results (p = 0.441) (Figure 4C). Decision curve analysis revealed substantial clinical utility across the 3%–87% risk threshold spectrum, achieving peak net benefit (Figure 4D).

Thirty-day survival analyses stratified by miRNA expression thresholds (median premiR-29b-3p and postmiR-29a-5p levels) revealed distinct prognostic patterns. Preoperative miR-29b-3p expression was significantly associated with survival (log-rank P = 0.041; Figure 5A), with an elevated hazard ratio (HR = 4.372; 95%CI 0.928–20.596). In contrast, postoperative miR-29a-5p expression was not significantly associated with prognosis, indicating only a marginal increase in risk (Figure 5B).

While this study established a significant correlation between miR-29 dysregulation and PONC, several limitations warranted consideration. First, this study was unable to establish a direct causal relationship between miR-29 levels and the pathogenesis of PONC. Second, our mechanistic exploration was confined to transcriptomic analysis of the rat hippocampus. Third, the study lacked corroborative RNA-seq data from patient blood or tissue samples, which limited our ability to directly translate the rodent transcriptomic findings to the human pathophysiology. Finally, the absence of functional validation through in vitro neuronal models or in vivo behavioral assays in animals meant the precise neuroprotective or detrimental roles of miR-29 in the context of DHCA remains to be experimentally confirmed.