The protocols received institutional review and got approval from the Institutional Animal Care and Use Committee, Fuwai Hospital, Chinese Academy of Medical Sciences (FW-2021-0005). All experimental procedures complied with the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health. Sprague-Dawley rats were kept under standard laboratory conditions, within free access to food and water (provided by the HFK Bioscience, China). Rats (age, 12–14 weeks; weight, 450–550 g) were randomly allocated into two groups: sham group, DHCA group (n = 5, each group).

DHCA procedures were established as previously described (Yan et al., 2022; Yan and Ji, 2022). In the DHCA group, rats were first anesthetized with 2% sevoflurane and then intubated with a 16-G endotracheal tube. Mechanical ventilation was then initiated at a rate of 80 breaths per minute with a tidal volume of 10 mL/kg. Mean arterial blood pressure (MAP) was continuously monitored via the left femoral artery. The tail artery and the right external jugular vein were cannulated and connected to a DHCA circuit, which included a reservoir, a membrane oxygenator, and a heat exchanger. The DHCA circuit was primed with 12 mL of 6% hydroxyethyl starch and 2 mL saline with 150 IU heparin. Cardiopulmonary bypass was then initiated at a flow rate of 160–180 mL/kg/min for 10 min after heparinization (500 IU/kg). Blood flow was directed from the jugular vein through silicon tubes to the membrane oxygenator and then returned to the rat via the tail artery. The target deep temperature was set at 18°C. After 30 min of systemic cooling, DHCA was induced by draining the blood into the reservoir and lasted for approximately 45 min, which was confirmed by MAP = 0. During the rewarming phase, the temperature was gradually increased by rewarming the blood in the DHCA circuits. The rewarming phase lasted for more than 60 min. The rats were then subjected to 40 min for reperfusion to recovery. Afterward, the DHCA circuit was weaned off within another 20 min. Finally, rats were ventilated for 30 min without cardiopulmonary bypass support. Throughout the entire procedure, MAP was maintained above 50 mm Hg. Rats in the sham group were only anesthetized, cannulated, and heparinized.

Before conducting the open field test, animals should be habituated to the testing environment for a brief period to minimize novelty-induced stress. The testing environment consists of a large, square, enclosed area (40 cm × 140 cm) with high walls to prevent escape. On the test day, each animal is placed in the center of the open field arena and allowed to explore the area for 10 min. Video recordings are utilized to collect data on the animals’ movement patterns. DeepLabCut is employed for 2D markerless pose estimation using transfer learning with deep neural networks (Lauer et al., 2022). Motion Sequencing (MoSeq) is an unsupervised machine learning method for analyzing animal behavior (Weinreb et al., 2024). Given behavioral recordings, MoSeq identifies a set of stereotyped movement patterns and their occurrence over time. After extracting movement information with DeepLabCut, a probabilistic time-series model, specifically an autoregressive hidden Markov model (AR-HMM), parses behavior into a set of reusable sub-second motifs known as syllables. This segmentation naturally delineates boundaries between syllables, thereby revealing the structure that governs the interconnections between syllables over time, which we refer to as behavioral grammar.

Hippocampal tissues were immersion fixed in 4% paraformaldehyde and blocked in paraffin. Then they were sectioned at 4 μm. The sections were stained with hematoxylin-eosin (HE) staining (G1004, Servicebio) and Nissl staining (Cresyl Violet; GP2087, Servicebio), respectively.

The hippocampal sections were permeabilized with 0.1% Triton X-100 (ZSGB-Bio, Inc.) and incubated with 10% goat serum (ZSGB-Bio, Inc.). Next, the sections were incubated with the primary antibodies in a humidified chamber at 4 °C overnight. The primary antibodies used for immunohistochemistry in this study were NLRP3 (1:200; GB114320, Servicebio) and TLR4 (1:500; GB11519, Servicebio). After rewarming, the sections were incubated with enzyme-conjugated goat anti-rabbit IgG polymer using a Rabbit Two-step kit (ZSGB-Bio, Inc.). Between each step, a phosphate buffered saline (PBS) buffer wash was applied for 5 min each three times. Finally, sections were screened using Pannoramic SCAN (3D HISTECH, Inc.).

The PC-12 cells were purchased from Procell Life Science & Technology Co., Ltd. (CL-0481). PC-12 cells were maintained in a PC-12 cell culture medium (CM-0481, Procell) at 37°C in a 95% humidified 5% CO₂ cell culture incubator. Confluent cultures were passaged by trypsinization. The RNA sequences for transfection were synthesized by GenePharma (Shanghai, China) and listed in Supplementary Table 1. All cell transfections were performed using Lipofectamine 3,000 (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA) according to the manufacturer’s protocol. After 24 h transfection, the cells were treated with hypothermia oxygen-glucose deprivation/reperfusion (H-OGD/R) treatment as described before. Briefly, PC-12 cells were placed in a 16°C environment with glucose-free Dulbecco’s modified Eagle’s medium (DMEM) without FBS in an incubator containing 95% N2 and 5% CO₂ for 2 h. Then the medium was replaced with PC-12 cell culture medium (CM-0481, Procell) and cells were transferred to the incubator set at 37°C with reoxygenation for 24 h. Finally, the cells were harvested for further experiment. CCK-8 assay (G4103, Servicebio) was used for cell proliferation detection.

For luciferase reporter experiments, the specific segments of circFRRS1 and TLR4 predicted to interact with miR-27a-3p were amplified by PCR and inserted into GV272 luciferase reporter vectors (GenePharma Co., Ltd.). The 293T cells were cotransfected in 24-well plates with 0.1 μg of the firefly luciferase report vector and 0.02 μg of the control vector containing Renilla luciferase, pRL-TK (Promega), as well as with 0.4 μg miR-27a-3p mimics, inhibitor or control miRNA. At 48 h post-transfection, luciferase activity was analyzed using the Dual-Luciferase Reporter Assay System (Promega, United States) according to the manufacturer’s instructions. Data were normalized to the Renilla Luminescence and presented relative to control miRNA transfected group.

The shRNA lentiviral vectors targeting circFRRS1 (sh-circFRRS1) and the negative control (sh-NC) were obtained from Genechem (Shanghai, China). The detailed sequences are listed in Supplementary Table 2. The lentiviruses were administered to the hippocampus of rats using stereotactic injection as previously described (McSweeney and Mao, 2015). Briefly, rats were anesthetized via intraperitoneal injection of sodium pentobarbital (30 mg/kg) and then placed in a stereotaxic apparatus. A hole was drilled at the following coordinates relative to the bregma, according to the rat brain atlas (Swanson, 2018): 3.3 mm posterior, ± 2.0 mm lateral, and 3.0 mm ventral to the skull surface. A 10 μL Hamilton microliter syringe was used to deliver 2 μL of the lentiviral solution (sh-NC or sh-circFRRS1; titer: 1 × 108 TU). Three days after the injection, the rats underwent DHCA surgery. To assess the in vivo knockdown efficiency, hippocampal tissues were collected after DHCA intervention, and circFRRS1 expression was detected by RT-qPCR. Rats that did not receive lentivirus treatment were injected with 2 μL of normal saline to control for potential damage caused by the injection procedure itself.

The candidate hippocampal tissues or PC-12 cells used for WB analysis were homogenized and lysed in cold RIPA lysis buffer (Beyotime, Inc.). The protein concentration was measured with BCA assay (Beyotime, Inc.), and an equal amount of protein from all samples was subjected to gel electrophoresis using an 12% sodium dodecyl sulfate-polyacrylamide gel (SDS-PAGE) and then transferred to the nitrocellulose (NC) membranes. NC membranes were blocked with in TBS-T saline (Tris-buffered saline with 0.1% Tween 20) containing 5% non-fat dry milk for 1 h at room temperature. Then, primary antibodies were added and incubated overnight at 4°C. After the primary antibody protocol was completed, mMembranes were then washed in TBST and incubated with secondary antibody at room temperature for 1 h. The primary and secondary antibodies used in this experiment were as follows: primary antibodies: TLR4 (1:1,000; GB11519, Servicebio), Myd88 (1:1,000; GB111554, Servicebio), p-IKK (1:1,000; CST#2697, Cell Signaling Technology), IKK (1:1,000; WL01900, Wanleibio), p-IκB (1:1,000; CST #2859, Cell Signaling Technology), IκB (1:1,000; CST#4812, Cell Signaling Technology), p-P65(1:500; CST#3033, Cell Signaling Technology), P65 (1:1,000; CST#3034, Cell Signaling Technology) and NLRP3 (1:1,000; GB114320, Servicebio); secondary antibodies: HRP conjugated Goat Anti-Rabbit IgG (H+L) (GB23303, Servicebio). Bound proteins were detected using Amersham Imager 800 (Cytiva, Inc.) and quantified by ImageJ software (v1.54).¹ β-Actin was used as an internal control for protein inputs.

Total RNA was extracted from hippocampal tissues or PC-12 cells with Trizol (Invitrogen, Inc.) following the manufacturer’s protocol, and then converted into cDNA using PrimeScript™ RT reagent kit with gDNA Eraser (Takara Bio, Inc.). RT-qPCR was conducted with SYBR Premix Ex Taq™ II (Takara Bio, Inc.) and the protocol included: pre-denaturation (95 °C, 5 min, 1 cycle) and PCR reaction (95 °C, 10 s, 60 °C, 30 s, 40 cycles in total) followed by a dissolution curve. β-actin and U6 were chosen for the reference genes. The 2^–ΔΔCT algorithm was used to calculate the relative quantification expression levels. RT-qPCR primer sequences were listed in Supplementary Table 2.

R (v4.2.1; R Core Team, 2022) was used for data processing and statistical analysis. The experiment and RT-qPCR results were presented as mean ± SD. Student’s t-tests were used for comparisons between the two groups. One-way ANOVA with post hoc analysis was applied for all multi-group comparison. P < 0.05 were considered to be statistically significant.

Postoperative cognitive dysfunction encompasses delirium. However, in rats, cognitive status can only be indirectly assessed through their physical behaviors. Therefore, to validate the behavioral changes in DHCA rats, we assessed their performance in the open field test and marked their poses and movement trajectories using DeepLabCut (Figures 2A,B). Notably, rats subjected to DHCA exhibited a significant decrease in total traveling distance compared to the control group (Sham rats) (Figure 2C), suggesting impairment in exploratory behavior indicative of delirium-like symptoms (Cao et al., 2023). Further motion sequencing (Moseq) analysis revealed distinct patterns in the median trajectory of poses associated with specific movement syllables. The behaviors identified included working (syllables 0, 4, 8), left or right turning (syllables 6, 7), freezing (syllables 1 and 5), right looking (syllable 3), and running (syllable 2) (Figure 2D). Keypoint analysis provided aligned and centered coordinates during motion sequencing, unveiling a clear pattern of behavioral transitions (Figure 2E). The comparison of syllable transitions between Sham and DHCA rats illustrated distinct differences (Figure 2F). Importantly, we quantified the average time spent in each syllable per minute during the open field test, revealing that DHCA rats spent significantly more time in freezing behaviors and less time in exploration compared to the Sham group (Figure 2G), indicative of heightened anxiety or confusion. These findings collectively provide compelling evidence that DHCA results in significant behavioral changes in rats, reflecting symptoms consistent with delirium.

The HE staining revealed remarkable neuronal abnormalities in the hippocampus of rats in the DHCA group, evidenced by high numbers of necrotic neurons (Figures 3A,C). The Nissl staining also showed marked neuronal changes in the hippocampus of rats in the DHCA group, including the shrunken and deeply stained cell bodies and the disappeared Nissl body (Figures 3B,C). ELISA results showed that DHCA induced increasing levels of inflammatory factor (Figure 3D). Further WB analysis showed that the expression of NLRP3 was elevated in the DHCA group (Figure 3E). Immunohistochemistry staining of NLRP3 also displayed a similar trend to the WB results (Figure 3F). These results revealed that DHCA cause severe inflammation damage in hippocampal tissue.

In the previous study we have analyzed the expression profiles of circRNAs in rat hippocampus after DCHA. Among all dysregulated circRNAs, rno_circ_0028462 (rename as rno-circFRRS1, due to the host gene was FRRS1) confirmed highly expressed in DHCA group with both transcripts analysis and experiment validation (Wang et al., 2023). The detail sequence of circFRRS1 was shown in Figure 4A. In this study, we further explored the downstream mechanism of circFRRS1. In vitro experiments, PC-12 cells were treated with H-OGD/R to simulate DHCA process and transfected with si-circFRRS1. As shown in Figure 4B, the transfection of si-crcFRRS1 markedly reduced the high level of circFRRS1 induced by H-OGD/R in PC-12 cells. CCK-8 assays revealed that suppressing circFRRS1 raised the survival rate of PC-12 cells treated with H-OGD/R (Figure 4C). In addition, silencing circFRRS1 inhibit the inflammatory activation by H-OGDR, evidenced by decreased gene expression of TNF-α, IL-6, and IL-1β and less NLRP3 protein (Figures 4D–G).

A key function of circRNA is to sequester miRNA by sponging, resulting in alteration of mRNA targets. In our previous study, bioinformatics analysis showed that miR-27a-3p was a candidate for circFRRS1 (Wang et al., 2023). Using miRTarbase, which display the largest amount of validated miRNA/mRNA pairs (Huang et al., 2020), the rno-miR-27a-3p/TLR4 pair was shown as strong evidence (Supplementary Figure 1), and miRWalk database showed that the interaction between miR-27a-3p and TLR4 was predicted across species (Supplementary Figure 2). Recent studies have shown that miR-27a-3p/TLR4 pair was a novel miRNA-related target that plays a major role in inflammatory activation and I/R injury, thus miR-27a-3p/TLR4 pair was selected as one candidate (Wang et al., 2019; Luo et al., 2022). In order to confirm the bioinformatics result, the dual luciferase reporter assay was performed and the regulatory relationships of circFRRS1/rno-miR-27a-3p and rno-miR-27a-3p/TLR4 pairs were confirmed (Figure 5).

Next, the expression tendency of miR-27a-3p/TLR4 pair was validated in the rat hippocampus tissues. The RT-PCR analysis results showed that the lower expression level of miR-27a-3p and higher level of TLR4 in the DHCA hippocampus tissues (Figure 6A). Immunohistochemistry showed that TLR4 was mainly expressed on the cell membrane, and confirmed that its expression was significantly increased in the DHCA group compared to the sham rats at the protein level (Figure 6B). TLR4/NF-κB signaling is an important pathway that triggers the formation of NLRP3 inflammasome, thus we detected the expression levels of key proteins in this pathway. As shown in Figures 6C,D, levels of TLR4, MyD88, and the ratios of p-IKK/IKK, p-IκB/IκB, and p-P65/P65 were markedly elevated in DHCA rat hippocampus relative to the Sham group.

We further used WB analysis to evaluate whether inhibition of circFRRS1 suppressed NLRP3 inflammasome activation by regulating the TLR4/NF-κB axis in H-OGD/R-treated PC-12 cells. As shown in Figure 7, levels of TLR4, MyD88, NLRP3, and the ratios of p-IKK/IKK, p-IκB/IκB, and p-P65/P65 were markedly elevated after H-OGD/R treatment relative to the controls. Levels of these proteins were markedly reduced in H-OGD/R + si-circFRRS1-treated cells compared with those treated with H-OGD/R + si-NC cells.

To confirm that the regulation of TLR4 by circFRRS1 in H-OGD/R-treated PC-12 cells is mediated by miR-27a-3p, we performed additional cell experiments. As shown in Figure 8A, administration of miR-27a-3p inhibitor could reverse the downregulation of TLR4 expression by inhibiting circFRRS1. Further protein-level detection gave the same results (Figure 8B).

Furthermore, we administered shRNA lentiviral vectors targeting circFRRS1 (sh-circFRRS1) to rats subjected to DHCA in advance (Supplementary Figure 3A). The vivo knockdown efficiency of circFRRS1 was assessed through RT-qPCR (Supplementary Figure 3B). The level of miR-27a-3p was significantly increased in the DHCA +sh-circFRRS1 group compared to the DHCA + sh-NC group. In DHCA rats, sh-circFRRS1 intervention restored the expression of miR-27a-3p and suppressed TLR4 protein levels in hippocampal tissue (Supplementary Figure 3C). Additionally, the DHCA + sh-circFRRS1 group exhibited lower NLRP3 protein expression compared to the DHCA + sh-NC group, indicating reduced inflammatory activation (Supplementary Figure 3C). Histopathological examination, including HE and PAS staining, revealed less damage in the hippocampal tissue of the DHCA + sh-circFRRS1 group relative to the DHCA + sh-NC group (Supplementary Figures 3D,E). In conclusion, the above results demonstrated that inhibition of circFRRS1 effectively alleviated inflammatory damage in the hippocampal tissue, further supporting the potential of circFRRS1 as an important therapeutic target for DHCA-associated neurological injury.