Study design and patients

In this retrospective research, adult septic patients hospitalized in the intensive care unit (ICU) between January 2022 and January 2024 were screened, as depicted in Supplementary Fig. 1. The criteria for inclusion were outlined below: (1) The clinical diagnosis of sepsis based on the Sepsis 3.0 criteria [26]; (2) age ≥ 18 years; (3) ICU stay ≥ 24 h. The criteria for exclusion were as follows: (1) acute brain injury, encompassing epilepsy, traumatic brain injury, intracranial infection, ischemic stroke, or hemorrhagic stroke; (2) psychiatric disorders or neurological diseases; (3) chronic alcoholism or substance abuse; (4) hepatic encephalopathy, hypoglycemic coma, metabolic encephalopathy, or additional severe liver or kidney dysfunction impacting consciousness; (5) severe electrolyte disturbances; (6) pulmonary encephalopathy or PaCO2 ≥ 80 mmHg; and (7) acute or chronic liver dysfunction, including hepatic encephalopathy and cirrhosis.

Data within the first 24 h upon admission were collected, including BGLs and information on insulin treatment. Patients experiencing severe hyperglycemia (≥ 300 mg/dL) at least twice were administered insulin at a rate of 0.1–0.5 IU/kg/min and included in the HG group; patients experiencing moderate hyperglycemia (200–300 mg/dL) at least twice were included in the MG group; and patients with BGLs below 200 mg/dL and not requiring insulin treatment were included in the NG group. During hospitalization, delirium occurrence was assessed twice daily by two senior nurses, who were blinded to the study protocol, employing the Confusion Assessment Method for the Intensive Care Unit (CAM–ICU). This work was approved by the Ethics Committee of the First Affiliated Hospital of Nanchang University.

Baseline data, encompassing age, gender, diabetes history, site of infection, renal replacement therapy, APACHE II score, SOFA score, mechanical ventilation, vasopressors, ICU stay, and occurrence of delirium, were collected for all enrolled patients.

Rats and sepsis models

We obtained a cohort of healthy male Sprague–Dawley (SD) rats weighing between 200 and 240 g, between 12 and 14 weeks, from Vital River Lab Rotary in Zhejiang, China. The subjects were subjected to controlled environmental settings, which included a temperature range of 23–25 °C, humidity levels maintained between 50 and 60%, a 12-h light/dark cycle, and unrestricted availability of food and water. The handling and experimentation of animals were performed in complete compliance with the rules established by the Animal Welfare Ethics Committee of Nanchang University.

Following a 1-week acclimatization period, the rats underwent right jugular vein catheterization. A PE-50 catheter, having an inner diameter of 0.58 mm and an outside diameter of 0.99 mm, was placed into the jugular vein to assist in delivering the specified solution. Subsequently, the catheter was tunneled through a posterior neck incision and connected to a swivel harness, enabling the rats to move freely within their cages while maintaining access to the administered solution.

After catheterization, the SAE model was created by performing cecal ligation and puncture (CLP), as explained in prior research. In summary, the rats were administered 50 mg/kg of pentobarbital to induce anesthesia and a 1.5 cm incision was done along the midline to subject the cecum. Next, the cecal ileocecal valve was double-ligated at the distal one-third of the cecum. Subsequently, an 18G needle was utilized to puncture the cecum, causing a small quantity of intestinal detritus to be expelled into the abdominal cavity. Following the surgical procedure, the rats were put on a heating blanket set at a temperature of 37 °C and observed until they regained consciousness from the effects of the anesthetic. The rats in the SG were subjected to cecal exposure without CLP. Following the CLP operation, blood samples were obtained every 3 h from the right jugular vein catheter for blood glucose analysis.

Animal treatment

The rats were first allocated into two primary groups: the CLP group and the SG. Preliminary experiments were conducted to establish the optimal amount of glucose required to attain certain goal BGLs in the past. The rats were allocated into three subgroups based on their target levels. The high glucose group (HG, n = 15) obtained a continuous intravenous infusion of 0.9% saline with 27% glucose (w/v) at a rate of roughly 39 mg/(kg·min) to preserve BGLs over 300 mg/dL. The moderate glucose group (MG, n = 15) acquired a continuous intravenous infusion of 0.9% saline that included 18% glucose (w/v) at a rate of about 24 mg/(kg·min) to preserve BGLs between 200 and 300 mg/dL; the none glucose group (NG, n = 15), wherein rats received an intravenous infusion of an equivalent volume of 0.9% saline, maintaining BGLs below 200 mg/dL; and the HG group treated with recombinant human insulin (HI, n = 15) at a dosage of 0.1 IU/kg/min to maintain BGLs between 200 and 300 mg/dL. In contrast, rats in the SG (n = 8) received an infusion of 0.9% saline at a volume equivalent to the other groups. Following the CLP procedure, the designated solution was continuously administered through a central vein catheter for 9 h.

Y-maze

48 h after CLP, the spatial working memory of rats was evaluated utilizing the Y-maze (Shanghai XinRuan, China), which consisted of three black polypropylene arms labeled A, B, and C, with each arm measuring 40 × 10 × 16 cm and placed at an angle of 120° from each other. During the 8-min experimental phase, the rats were given unrestricted access to explore the labyrinth. They were placed at the far end of any arm, and their entries and exits into each arm were recorded. For an entry or exit to be counted, all four limbs of the rat had to enter or exit an arm. A sequence of three different arm entries (such as ABC, CAB) was considered as spontaneous alternation performance (SAP), which reflects good working memory. The movement of a rat transitioning from one arm to the central region and then returning to the same arm was recorded as a same-arm return (SAR) each time. The alternation percentage was computed as [number of alternations/(total number of arm entries-2)] × 100. Following each experiment, the box was cleaned with 75% alcohol.

Novel object recognition (NOR) test

The NOR test was employed to assess alterations in short-term memory. In a plastic box with a side length of 40 cm, two objects of identical size and material (Object A) were placed. Prior to the test day, the rats were acclimated to the test box for 10 min. During the testing day, one of the objects, A (familiar object, F), was replaced with a similar-sized material object, B (novel object, N). Subsequently, the rats were individually introduced inside the enclosure and given unrestricted access to examine the items for 5 min. Subsequently, the box was cleaned using a solution of 75% alcohol. Object exploration was stated as orienting the nose toward an object at a distance ≤ 2 cm. The exploration times for the familiar and novel objects were recorded, and the discrimination index (DI) was computed as (t[N-F])/(t[F + N]) × 100%.

Evans blue (EB) staining

2 h prior to measurement, a 2% solution of EB was intravenously injected at a dose of 4 mL/kg. Following anesthesia with 5% pentobarbital sodium (50 mg/kg), rats underwent cardiac perfusion with saline. Tissue samples were homogenized in 5 mL of 0.1 M PBS, mixed with 5 mL of 60% trichloroacetic acid, underwent centrifugation at 4000 rpm for 15 min, and the supernatant was gathered. A spectrophotometer was employed to determine the supernatant's absorbance at 620 nm. Finally, the EB content in the brain tissue was calculated based on the unit weight.

Histological examination

Rats were euthanized using sodium pentobarbital after the behavioral tests, and blood and whole brain specimens were gathered. The brain specimens were preserved in a solution of 4% paraformaldehyde for 24 h. After that, they were embedded with paraffin and cut into coronal slices that were 5-μm thick. In order to perform a histological investigation, the slices were first treated to remove the paraffin and then dehydrated. Hematoxylin and eosin staining were then conducted to assess morphological changes in the hippocampal region.

For immunofluorescence analysis, the sections were subjected to antigen retrieval by immersion in sodium citrate buffer at 100 °C for 15 min. Afterward, the sections were blocked utilizing 0.01 mM PBS with 10% goat serum for 15 min. The sections were then incubated overnight at 4 °C with primary antibodies, including rabbit anti-IBA1 and CD68 (Abcam, USA), to target specific cellular markers. On the next day, the samples were rinsed three times with PBS, followed by incubation in the dark with fluorescently labeled secondary antibodies (FITC or CY3) at ambient temperature for 2 h. Subsequently, the samples were counterstained with DAPI for 10 min and sealing the tissue slices. The average number of IBA1 + CD68-positive cells in three non-overlapping areas was observed and calculated using a fluorescence microscope (ZEISS, Germany).

TUNEL staining

TUNEL staining was conducted utilizing the TUNEL Assay Kit (Beyotime Biotech, China) in accordance with the company’s protocol. The sections were treated with 4% paraformaldehyde to fix them, made permeable with 0.1% Triton X-100 for 10 min, and then subjected to the TUNEL reaction mixture for 1 h at 37 °C in a lack of light. DAPI was employed to stain the cell nuclei. Imaging was performed using an optical microscope (Olympus, USA).

Cell culture and treatment

BV2 cells (obtained from the Pricella Life science & Technology, China) were cultivated in Dulbecco's Modified Eagle Medium (DMEM, Thermo Fisher Scientific, USA) enriched with 10% fetal bovine serum and 1% penicillin/streptomycin and maintained in a 37 °C humidified incubator with a constant oxygen supply (5% CO2, 21% O2, and 74% N2). To simulate sepsis, BV2 cells underwent treatment with 1 μg/mL LPS and then incubation in DMEM supplemented with 5.6 mM (LG group), 16.7 mM (MG group), and 25.0 mM (HG group) glucose. BV2 cells were pretreated with 10 μM KC7F2, followed by 1 μg/mL LPS and/or 25.0 mM glucose for the cellular signaling pathway study.

shRNA-mediated ChREBP knockdown

The lentiviral vectors for the knockdown of ChREBP (sh-ChREBP) or the non-silencing control shRNA (NC) were synthesized by Santa Cruz Biotechnology (cat. no. sc-3861, USA). BV2 cells were infected with lentivirus-mediated sh-ChREBP or NC for 48 h. To serve as a negative control, non-silencing shRNA was used (NC). The sh-ChREBP or non-targeting shRNA lentiviral vectors were added to DMEM at a final multiplicity of infection of 50. Stable transformed cell lines were subsequently screened using puromycin (5 μg/mL). Silencing efficiency was assessed by Western blot.

Cell viability

BV2 cells were placed in a 96-well plate and subjected to 1 μg/mL LPS for a duration of 12 h. Subsequently, a predetermined amount of glucose was introduced into the wells. Afterward, 10 μL of CCK-8 solution was introduced according to the manufacturer's recommendations (Beyotime, China). Afterward, the plate was put in a humidified incubator that was adjusted at 37 °C for 1 h. Following incubation, the plate's absorbance was measured at 450 nm utilizing a microplate reader.

ELISA

The IL-1β, TNF-α, HMGB1, NSE, and S100β levels were ascertained by employing a commercial ELISA kit following the manufacturer's recommendations (Elabscience, China). After collecting blood samples or hippocampal tissue homogenate from animal experiments, the specimens immediately centrifugated at 3000 rpm for 15 min at 4 °C, and the supernatant was gathered. For in vitro experiments, BV2 cells were initially stimulated with LPS for 12 h and subsequently exposed to 25.0 mM glucose. After collecting the supernatant, the IL-1β, TNF-α, and HMGB1 levels were measured.

Evaluation of glycolysis and metabolic pathway shift

The glycolysis/oxidative phosphorylation assay kit (DOJINDO laboratories, Japan) was employed to assess glycolysis levels and shifts in metabolic pathways. Briefly, following the methods outlined in a previous study [27], BV2 or HMGB1–shRNA/NC cells (2.0 × 104 cells/well) were seeded in a 96-well plate and subsequently treated with 1 μg/mL LPS and DMEM containing a pre-defined level of glucose for 12 h. To evaluate shifts in the metabolic pathway shift, cells were initially stimulated with LPS and subsequently treated with or without 1.25 μM oligomycin (an inhibitor of OXPHOS) for 3 h. Following this, 100 μL of ATP working solution was introduced to each well, and absorbance was quantified at 450 nm. The shift in metabolic pathways was evaluated by measuring mitochondrial ATP and glycolytic ATP production.

To examine glycolysis levels, after LPS stimulation, 20 μL of the supernatant was incubated with 80 μL of the lactate working solution for 30 min. The mixture absorbance was subsequently ascertained at 450 nm, employing a microplate reader. The glycolysis degree was assessed by measuring the lactate production.

Western blotting

The hippocampus tissues or BV2 cells were homogenized utilizing RIPA lysis buffer, aligning with the manufacturer's recommendations. The protein concentration was quantified employing a BCA assay kit (Boster, China). Then, 20 μg of protein specimens were separated employing a 10% SDS–PAGE gel and then deposited onto a PVDF membrane (MILLIPORE, USA). The membrane was subsequently obstructed with 5% skimmed milk and underwent incubation at 4 °C overnight with primary antibodies specifically targeting ZO-1, Claudin-5, IL-1β, TNF-α, HMGB1, iNOS, ChREBP, HK2, PKM2 (Abcam, USA), and β-actin (Jinqiao Biotechnology, China). It was then incubated with HRP-conjugated secondary antibodies at ambient temperature for 2 h. Finally, protein bands were observed and quantified utilizing the ChemiDoc Imaging System (Bio-Rad Laboratories, USA) and Image Lab software.

RNA extraction and real-time PCR

The TRIzol reagent (Invitrogen, Thermo Fisher Scientific, USA) was employed to extract the total RNA, while the cDNA was generated utilizing the PrimerScript reagent Kit (Takara) following the manufacturer's recommendations. The ABI StepOne real-time PCR machine (Life Technologies, USA) and the SYBR Green PCR Master Mix (TransGen Biotech, China) were deployed to conduct real-time PCR. The PCR products were amplified, measured, and normalized utilizing β-actin as a control. All the PCR primers are provided as follow. ChREBP: Forward: 5′-CACTCAGGGAATACACGCCTAC-3′, Reverse: 5′-GGACTTCTGGGATTCTGGTTCTA-3′. β-Actin: Forward: 5′-ACCCAGAAGACTGTGGATGG-3′, Reverse: 5′-CACATTGGGGGTAGGAACAC-3′. HIF-1α: Forward: 5′-CTGGGACTTTCTTTTACCATGC-3′, Reverse: 5′-AATGGATTCTTTGCCTCTGTGT-3′. HK2: Forward: 5′-TGATCGCCTGCTTATTCACGG-3′, Reverse: 5′-AACCGCCTAGAAATCTCCAGA-3′. PKM2: Forward: 5′-GCCGCCTGGACATTGACTC-3′, Reverse: 5′-CCATGAGAGAAATTCAGCCGAG-3′.

Data analysis

The Shapiro–Wilk test was applied to estimate the normality of continuous data. The Kruskal–Wallis test was applied to compare non-normally distributed continuous variables, which were expressed as the median (interquartile range [IQR]) or count (percentage). The Pearson rank correlation test was employed to analyze the link between BGLs and delirium. The study conducted logistic regression and receiver operating characteristic (ROC) analyses to assess the predictive significance of blood glucose levels in relation to the occurrence of delirium and death within 28 days. The continuous variables in both in vitro and in vivo experiments had a normal distribution and were expressed as the mean ± Standard Deviation (SD). Statistical significance was assessed utilizing either one- or two-way analysis of variance, depending on the situation. A log-rank test was implemented to evaluate the variances in survival rates among the groups.