Category: 3. Business

Introduction

Background

Acute pancreatitis (AP) is characterized by the dysfunction of pancreatic cellular pathways and organelles due to various etiologies, with gallstones and alcohol abuse being the most prevalent.¹ This condition ultimately results in the death of pancreatic acinar cells.² In severe instances, AP may arise as a consequence of significant complications, including local and systemic inflammatory response syndrome (SIRS) and multiple organ failure (MOF).³ AP has acute onset, severe condition and many complications, and its incidence is increasing year by year.⁴ A study indicated that the annual global incidence of AP is 34 cases per 100,000 individuals,⁵ with an overall mortality rate of approximately 5%. In cases of severe acute pancreatitis (SAP), the mortality rate may approach 20%. Currently, there is no established clinical treatment for AP. The primary pathological response in AP is the premature activation of trypsinogen, resulting in damage and death of acinar cells. Recent research has identified necroptosis as a form of regulated cell death (RCD).⁶ It is crucial in the mechanism of acinar cell death and the premature activation of trypsinogen.⁷

Necroptosis is mediated by receptor interacting protein kinase 1 (RIPK1) and receptor interacting protein kinase 3 (RIPK3). The activation of the RIPK3/mixed lineage kinase like (MLKL) pathway exhibits characteristics of both necrosis and apoptosis. Specifically, necroptosis is actively regulated by various genes and occurs in an orderly manner through the activation of specific death pathways.^8–10 Salidroside (Sal) has a wide range of pharmacological activities, including Anti-inflammatory, anti-aging, antioxidant, and anti-tumor, etc.^11–14 Additionally, it has been shown to decrease the activity of pancreatic enzymes during the initial stages of SAP.^15,16

Aim of the Study

Based on the aforementioned evidence, we hypothesized that Sal might alleviate AP by modulating necroptosis. To test this hypothesis, the present study was designed to achieve the following specific aims: First, to investigate the therapeutic effects of Sal on pancreatic injury and systemic inflammation in a rat model of sodium taurocholate-induced AP; Second, to determine whether the protective effects of Sal are associated with the inhibition of the RIPK1/RIPK3/MLKL necroptosis pathway in both in vivo and in vitro (cerulein-stimulated AR42J cells) AP models; Finally, to explore the functional interaction between Sal and Nec-1 (a specific RIPK1 inhibitor) in order to elucidate the potential mechanism of action of Sal.

Materials and Methods

Experimental Animals and Cells

The Medical Laboratory Animal Center of Lanzhou University supplied 18 male Wistar rats with weights ranging from 250 to 280 grams. The animals were maintained in an environment characterized by alternating 12-hour light and dark cycles, with unrestricted access to food and water provided. All the rat experiments were conducted in accordance with the “Regulations on the Administration of Experimental Animals” (Lanzhou University) and were approved by the Ethics Committee of Lanzhou University Second Hospital (Approval No.D2024-410). At the same time,all methods are reported in accordance with ARRIVE guidelines.

Rat pancreatic exocrine cells-AR42J (Cellverse, iCell-r002) were cultured in AR42J specialized medium containing 20% Foetal Bovine Serum (FBS) (Procell, CM-0025) at 37°C, 5% CO₂.

AP Rat Model

Firstly, the AP rat model was established by retrograde injection of sodium taurocholate solution. All rats were fasted and water-deprived overnight before the operation, and then anesthetized by intraperitoneal injection of 0.3% pentobarbital sodium (0.2 mL/10 g). Next, the rats were randomly divided into three groups (n=6): (1) Sham surgery group: only underwent laparotomy and closure sham surgery; (2) AP group: Induction of pancreatic injury by retrograde injection of 3.5% sodium taurocholate (1mL/kg, injection rate 0.1 mL/min) into the pancreaticobiliary duct. Observation after approximately 10 minutes showed congestion, edema, local bleeding, and necrosis in the pancreatic tissue, confirming the establishment of the model^17,18 (3) Sal group: Based on our previous experimental results, Sal (60 mg/kg) (Medchemexpress, HY-N0109) was injected intraperitoneally 2 hours after AP modeling.¹⁵ And all animals were euthanized for sample collection at 24h post-modeling (Pentobarbital sodium, 200 mg/kg), followed by blood sample collection and serum separation through centrifugation. The pancreatic tissue was divided into three parts for pathological analysis, Western blot, and TEM detection. Finally, the animal carcasses were uniformly sent to the Gansu Province Hazardous Waste Disposal Center for processing. The final analysis included only those Wistar rats that successfully underwent the AP model induction surgery and survived the intended postoperative observation period. Rats that died during anesthesia or surgery (prior to model completion) were excluded from the analysis. Furthermore, any rat that did not exhibit a significant elevation in serum amylase levels at the predetermined time point post-modeling was also excluded, as it was deemed an induction failure. No animals were excluded from the final analysis based on these criteria.

Histopathology and Molecular Analysis

After fixation with 4% paraformaldehyde, pancreatic tissue was embedded in paraffin and sectioned (5 μm). Hematoxylin-eosin (H&E) staining was used to assess tissue pathological damage (edema, inflammatory infiltration, acinar necrosis), with a semi-quantitative scoring system on a 0–3 scale.¹⁹ The anti-p-MLKL antibody (Affinity, AF7420, diluted 1:200) was used in immunohistochemistry (IHC) to locate p-MLKL expression, followed by DAB staining and microscopic observation. Transmission electron microscopy samples were fixed with 2.5% glutaraldehyde and ultrathin sections were observed under an electron microscope to examine the ultrastructure of mitochondria.

Biochemical and Inflammatory Factor Testing

Serum and cell culture supernatant levels of AMY were quantified using a commercial kit (Yuanye, R22037) strictly according to the manufacturer’s instructions. Briefly, the assay involves the enzymatic hydrolysis of a defined substrate, and the resulting product is measured spectrophotometrically at a wavelength of 660 nm using a microplate reader. AMY activity is expressed in U/L. All samples were measured in duplicate. The concentrations of the inflammatory cytokines IL-6, IL-1β, and TNF-α were quantified using commercially available ELISA kits (Jonlnbio, catalog numbers JL20896, JL20884, and JL13202, respectively). The sensitivities of the assays were 1.11 pg/mL, 1.51 pg/mL, and 1.75 pg/mL for IL-6, IL-1β, and TNF-α, respectively. The detection ranges for all three cytokines were 3.12–200 pg/mL. Both intra- and inter-assay coefficients of variation were less than 10% for all assays, indicating high reproducibility and precision.

Western Blot Analysis

In Western blot analysis, tissues or cells are lysed with RIPA lysis buffer (containing protease/phosphatase inhibitors), and protein concentration is determined using the BCA method. 30 μg of protein was loaded into each well, separated by SDS-PAGE, and transferred to a PVDF membrane. After blocking with 5% BSA, the membrane was incubated with primary antibodies (RIPK1: Proteintech, 17519-1-AP, 1:1000; RIPK3: Bioss, bs-3551R, 1:1000; MLKL: Proteintech, 66675-1-Ig, 1:10000; p-MLKL: Affinity, AF7420, 1:1000; GAPDH: Selleck, F0003, 1:10000) and secondary antibodies. Protein expression levels were quantified by measuring the integrated density of the immunoreactive bands. Blots were imaged using a ECL chemiluminescence detection and analyzed with ImageJ software. The intensity of each target band was measured. To correct for potential variations in sample loading, the intensity of each target protein was normalized to that of the internal control (eg, GAPDH or MLKL) from the same sample. Data are presented in the bar graphs as the Mean ± SD of 4 independent experiments (due to tissue allocation constraints for protein extraction, Western blot analysis was performed on a subset of n=4 randomly selected samples per group).

AR42J Cell Experiment

First, the Cell Counting Kit-8 (CCK-8) assay was used to detect cell viability to determine the optimal drug concentrations of Sal and the RIPK1 inhibitor Necrostatin-1 (Nec-1): Cells were seeded in a 96-well plate (2×10⁴/well), allowed to adhere overnight, and then incubated with the corresponding drugs at gradient concentrations for 8 hours. CCK-8 reagent (10 μL/well) was added, and the incubation continued at 37°C for 4 hours. The optical density (OD) at 450 nm was measured using a microplate reader to calculate cell viability. Next, AR42J cells were pre-treated with Salidroside (50 μM) or Nec-1 (10 μM) for 4 hours prior to a 4-hour stimulation with cerulein (100 nM), resulting in a total intervention period of 8 hours.²⁰ Mitochondrial membrane potential was detected using the JC-1 dye (Beyotime, C2003S) and observed under a confocal microscope by measuring the red/green fluorescence ratio (excitation wavelength 490/525 nm, emission wavelength 530/590 nm). A decrease in the ratio indicates mitochondrial membrane potential depolarization. In the immunofluorescence experiment, after fixation and permeabilization of the cells, the anti-p-MLKL antibody (1:200) was co-stained with DAPI, and the subcellular localization of p-MLKL was observed using a confocal microscope. Only AR42J cell cultures with >95% viability were used. Any culture wells showing signs of bacterial or fungal contamination at the start of the experiment were excluded from the analysis.

Instrumentation and Equipment

The following key instruments were used in this study: Tissue FAXS PLUS microscope (Tissue FAXS PLUS, AUT), Electron microscope (Hitachi HT7800, Japan), Microplate reader (BioTek Synergy H1, USA), Confocal microscope (Zeiss LSM880, GER); ECL chemiluminescence detection (Bio-Rad ChemiDoc, SG).

Statistical Analysis

Data are expressed as mean ± standard deviation (Mean ± SD). Comparisons among multiple groups were performed using one-way analysis of variance (one-way ANOVA) followed by Tukey’s multiple comparison test for post-hoc analysis. All analyses were conducted using GraphPad Prism software (Version 10). A P-value of less than 0.05 (p < 0.05) was considered statistically significant. In the figures, significance levels are denoted as follows: ns (not significant, p > 0.05), *p < 0.05, **p < 0.01, and **p < 0.001.

Results

Sal Alleviates Pancreatic Injury in Rats with AP

The changes in serum AMY and inflammatory cytokine levels indicate that Sal exerts significant anti-inflammatory effects in AP. One-way ANOVA indicated a significant difference in serum AMY levels among the groups (Figure 1A, F (2, 15) = 27.13, p < 0.001). Post-hoc Tukey’s test revealed that compared to the Sham group, AMY levels were significantly elevated in the AP model group, whereas Sal treatment reduced AMY levels by 37.4% (p < 0.01). Similarly, serum levels of the pro-inflammatory cytokines TNF-α, IL-6, and IL-1β were markedly increased in the AP group (Figure 1B; one-way ANOVA, TNF-α: F (2, 15) = 97.37, p < 0.001; IL-6: F (2, 15) = 239.6, p < 0.001; IL-1β: F (2, 15) = 99.17, p < 0.001). Sal intervention significantly reduced these cytokines by 27.6%, 23.9%, and 45.3%, respectively (all p < 0.001). H&E showed extensive pancreatic edema, inflammatory cell infiltration, and fat necrosis in the AP model group (Figure 1C). In contrast, the Sal-treated group exhibited a marked reduction in pathological damage, with a 40% decrease in histopathological scores compared to the model group (Figure 1D, p < 0.001).

Figure 1 Sal alleviates pancreatic injury in rats with AP. (A) Serum AMY level detection (n=6). (B) Serum TNF-α, IL-6, and IL-1β levels (ELISA detection, n=6). (C) Representative images of pancreatic tissue H&E staining (scale bar: 100 μm) (black arrows indicate the areas of acinar cell necrosis; white arrows indicate the areas of fat necrosis; red arrows indicate the areas of inflammatory cell infiltration). (D) Pancreatic pathology score (0–3 points). **p < 0.01, ***p < 0.001 compared with AP group.

Sal Inhibits the Activation of the Necroptosis Pathway in Pancreatic Tissue

Since Sal alleviated pancreatic damage and inflammation in AP, we further investigated whether its protective effects depend on the necroptosis pathway. Western blot analysis revealed a significant upregulation of RIPK1, RIPK3, and p-MLKL protein expression in the pancreatic tissue of the AP group (Figure 2A–C). One-way ANOVA indicated statistically significant differences among the groups for all three proteins (RIPK1: F (2, 9) = 21.70, p < 0.001; RIPK3: F (2, 9) = 29.85, p < 0.001; p-MLKL: F (2, 9) = 39.33, p < 0.001). Post hoc analysis showed that Sal treatment significantly reduced their expression levels by 21.2%, 26.1%, and 18.7%, respectively (all p < 0.05). Consistent with these results, IHC staining demonstrated strong p-MLKL immunopositivity (brown) in the perimembranous region of pancreatic cells in the AP group, which was markedly attenuated in the Sal-treated group (Figure 2D and E). TEM was employed to evaluate the ultrastructural consequences of necroptotic signaling, with a focus on mitochondrial integrity. Pancreatic acinar cells from AP model rats exhibited severe organellar damage, most notably in the form of swollen mitochondria with vacuolation and disrupted cristae (Figure 2F, middle panel), which is a characteristic manifestation of ongoing necroptosis. Treatment with Sal markedly preserved mitochondrial morphology, presenting with only mild swelling and intact cristae (Figure 2F, right panel).

Figure 2 Sal inhibits the activation of the necroptosis pathway in pancreatic tissue. (A) Representative Western blot images showing protein levels of RIPK1, RIPK3, MLKL, and p-MLKL in pancreatic tissues from different experimental groups. (B) Quantitative analysis of RIPK1 and RIPK3 protein expression normalized to GAPDH (n=4). (C) Quantitative analysis of p-MLKL protein expression normalized to total MLKL (n=4). (D) Representative immunohistochemical staining of p-MLKL in pancreatic tissues (scale bar: 50 μm). Brown granules indicate positive signals. The AP group shows extensive p-MLKL expression localized around pancreatic acinar cell membranes. (E) Quantitative analysis of p-MLKL immunohistochemical staining expressed as mean optical density (OD) values (n=4). (F) Transmission electron microscopy observation of mitochondrial ultrastructure (scale bar: 1 μm). The mitochondria in the AP group showed swelling and cristae rupture, while the mitochondria in the Sal group appeared nearly normal (The black arrows indicate the mitochondrial structure of each group). *p < 0.5, **p < 0.01, ***p < 0.001 compared with AP group.

Sal Reduces Inflammatory Damage in the AR42J Cell AP Model by Mitigating Necroptosis

To further validate the anti-necroptotic and anti-inflammatory effects of Sal and elucidate its underlying mechanism, we established an in vitro AP model using cerulein-stimulated AR42J cells. First, the optimal non-cytotoxic concentrations of Sal (50 μM) and Nec-1 (10 μM) were determined via CCK-8 assay (Figure 3A and B). Cerulein (100 nM) stimulation significantly increased AMY release into the supernatant by 1.4-fold (Figure 3C; one-way ANOVA, F(3, 20) = 36.74, p < 0.001), indicating successful induction of acinar cell injury. Sal treatment alone reduced AMY levels by 15.3%, while co-treatment with Sal and Nec-1 did not yield a significant additive effect compared to Sal alone. Similarly, the release of pro-inflammatory cytokines (TNF-α, IL-6, IL-1β) was markedly elevated in the AP group (Figure 3D, one-way ANOVA, TNF-α: F (3, 20) = 37.14, p < 0.001; IL-6: F (3, 20) = 48.49, p < 0.001; IL-1β: F (3, 20) = 51.72, p < 0.001), with the most potent suppression observed in the Sal + Nec-1 co-treatment group (reductions of 29.8%, 13.2%, and 14.9%, respectively). Furthermore, cerulein induction resulted in a severe loss of mitochondrial membrane potential (indicated by a 94% decrease in red/green fluorescence ratio, p < 0.001), which was significantly restored by Sal pretreatment (Figure 3E and F, one-way ANOVA, F(3, 16) = 25.08, p < 0.001). This loss was significantly attenuated by Sal pretreatment (p < 0.05). These results confirm that Sal mitigates inflammatory damage and mitochondrial dysfunction in acinar cells under AP-like conditions.

Figure 3 Sal reduces inflammatory damage in the AR42J cell AP model. (A) The effect of Sal (1–200 μM) on the viability of AR42J cells (CCK-8 assay, n=4). Choose 50 μM (where cell viability significantly increased and there was no significant toxicity) for subsequent experiments. (B) The effect of Nec-1 (1–200 μM) on cell viability (n=4). Choose 10 μM (cell viability significantly increased and no significant toxicity) for subsequent experiments. (C) AMY levels in cell supernatants (n=6). (D) TNF-α, IL-6, and IL-1β levels in cell supernatants (n=6). (E) Quantitative analysis of mitochondrial membrane potential assessed by JC-1 staining (n=5). Data are presented as the red/green fluorescence ratio. (F) Representative fluorescence images of JC-1 staining in different experimental groups (scale bar: 20 μm). Red fluorescence indicates JC-1 aggregates (high membrane potential), while green fluorescence indicates JC-1 monomers (low membrane potential). A decrease in the red/green fluorescence ratio indicates a decline in membrane potential, and Sal pretreatment partially restores the ratio. *p < 0.5, **p < 0.01, ***p < 0.001 compared with AP group.

Sal Inhibits the RIPK1/RIPK3/MLKL Pathway to Suppress Necroptosis in AR42J Cells

We further investigated whether Sal confers protection through specific inhibition of the necroptotic pathway. Western blot analysis revealed that cerulein significantly upregulated key mediators of necroptosis, increasing RIPK1, RIPK3, and p-MLKL levels by 1.34-fold, 1.77-fold, and 1.61-fold, respectively (Figure 4A–D, one-way ANOVA, RIPK1: F(3, 12) = 10.25, p = 0.001; RIPK3: F(3, 12) = 10.47, p = 0.001; p-MLKL: F(3, 12) = 7.813, p = 0.004). Sal treatment alone significantly reduced the expression of these proteins, with p-MLKL decreasing by 34.7%. The absence of an additive effect between Sal and Nec-1 suggests that both compounds likely target the same node within the pathway. Immunofluorescence staining further demonstrated robust membrane translocation of p-MLKL in AP group cells, whereas both Sal and Nec-1 treatments markedly reduced p-MLKL membrane aggregation (Figure 4E and F, other representative fields of view for each group are shown in Supplementary Figure 1). Collectively, these data strongly support our hypothesis that Sal alleviates cerulein-induced AP injury in AR42J cells by inhibiting the RIPK1/RIPK3/MLKL-mediated necroptosis pathway.

Figure 4 Sal reduces necroptosis in AR42J cells by inhibiting the RIPK1/RIPK3/MLKL pathway. (A) Representative Western blot images showing protein levels of RIPK1, RIPK3, MLKL and p-MLKL in different experimental groups. (B) Quantitative analysis of RIPK1 protein expression normalized to GAPDH (n=4). (C) Quantitative analysis of RIPK3 protein expression normalized to GAPDH (n=4). (D) Quantitative analysis of p-MLKL protein expression normalized to total MLKL (n=4). (E) Quantitative analysis of p-MLKL fluorescence intensity from immunofluorescence staining (n=4). (F) Representative immunofluorescence images of p-MLKL subcellular localization (scale bar: 50 μm). Red fluorescence represents p-MLKL signal, and blue represents DAPI nuclear staining. *p < 0.5, **p < 0.01, ***p < 0.001 compared with AP group.

Discussion

The pathological process of AP involves complex mechanisms, among which necroptosis has been widely recognized as a critical contributor to cellular damage and inflammatory responses. However, its regulatory mechanisms and potential as a therapeutic target in AP remain incompletely elucidated. This study suggests that Sal significantly alleviates pancreatic injury in both rat AP models and AR42J cells by suppressing RIPK1/RIPK3/MLKL-mediated necroptosis, providing novel experimental evidence to support the potential clinical application of Sal in the management of AP (Figure 5).

Figure 5 Sal inhibits the RIPK1/RIPK3/MLKL necroptosis pathway, prevents mitochondrial damage, and reduces inflammation in AP.

The present study indicate that Sal treatment markedly reduced serum levels of amylase (AMY) and pro-inflammatory cytokines (TNF-α, IL-6, and IL-1β) in AP rats (Figure 1A and B), consistent with previous studies reporting the protective effects of Sal against multi-organ injury through its anti-inflammatory and antioxidant properties.^16,21–24 For instance, Wang et al²⁵ showed that Sal inhibited furan-induced barrier damage and intestinal inflammation by suppressing TLR4/MyD88/NF-κB signaling, while Wang et al²⁶ reported Sal alleviated mitochondrial dysfunction by activating the PGC-1α/Mfn2 signaling pathway, and restrained the endoplasmic reticulum stress. However, unlike these earlier studies that focused primarily on general anti-inflammatory effects, our study provides novel mechanistic insights by specifically establishing the inhibitory effect of Sal on necroptosis. Histopathological evaluation further confirmed that Sal ameliorated pancreatic tissue edema, necrosis, and inflammatory infiltration (Figure 1C and D).

The most significant finding of this study is the identification of Sal as a regulator of necroptosis. Western blot and immunohistochemical analyses revealed that Sal significantly inhibited the expression of RIPK1, RIPK3, and p-MLKL in pancreatic tissues (Figure 2A–C). TEM further demonstrated that Sal preserved mitochondrial structural integrity (Figure 2F), suggesting that its protective effects may be closely associated with the mitigation of mitochondrial dysfunction during necroptotic stress.

The RIPK1/RIPK3/MLKL axis is a well-established pathway mediating necroptosis.^6,27 Our findings not only confirm the pivotal role of this pathway in AP but also provide new evidence for its pharmacological modulation. Sal significantly inhibited the activation of this pathway in both in vivo and in vitro AP models (Figures 2A and 4A). Notably, Nec-1 is a well-characterized and highly specific allosteric inhibitor of RIPK1. It functions by stabilizing RIPK1 in an inactive conformational state, thereby preventing its kinase activity and subsequent recruitment and phosphorylation of RIPK3. This specific inhibition blocks the initiation of the necroptotic cascade upstream of MLKL activation. In our study, the use of Nec-1 served a dual purpose: firstly, as a positive control to confirm the involvement of RIPK1-dependent necroptosis in our AP model, and secondly, as a pharmacological tool for pathway interrogation. The observation that the combination of Sal and Nec-1 did not produce a significant additive effect is a critical pharmacological clue. It suggests that Sal likely intersects with the necroptosis pathway at the level of RIPK1 or its upstream regulators, rather than acting on a parallel or downstream node.

Furthermore, Sal was found to restore mitochondrial membrane potential (Figure 3F), supporting the notion that it mitigates AP progression through a dual mechanism—inhibiting necroptosis and preserving mitochondrial function. This aligns with emerging studies emphasizing the crosstalk between necroptosis and mitochondrial dysfunction,^28–30 although the precise mechanisms require further investigation. While the inhibition of this pathway is a major mechanism, we cannot rule out contributions from other pathways. This is now framed as a key mechanism rather than the exclusive mechanism.

Despite these advances, several limitations should be acknowledged. First, this study did not directly determine the pharmacokinetic parameters of Sal in an ascites environment. The absorption and distribution characteristics of Sal need to be further clarified in future research. Second, a limitation of our in vitro study is the lack of a Nec-1 alone treatment group, which would have served as a crucial positive control to benchmark the efficacy of Sal against a known RIPK1 inhibitor. Future studies will include Nec-1 as a standalone treatment to fully validate the model and provide a direct comparison for the potency of novel inhibitors like Sal. Third, it is important to note that although the pharmacological data we obtained using Nec-1 strongly suggest that Sal acts on the RIPK1 pathway, the lack of in vivo inhibitor experiments or a rescue experiment (eg, through RIPK1 overexpression) remains a limitation. Future studies employing genetic approaches both in vitro and in vivo will be essential to conclusively validate RIPK1 as the direct target. To further solidify our conclusions, future work will involve administering Nec-1 in a rat AP model to directly compare its efficacy with that of Sal and to investigate potential synergistic effects, or transfecting AR42J cells with RIPK1 plasmids to examine whether forced RIPK1 expression can counteract the protective effects of Sal, which would provide definitive mechanistic validation. Additionally, while our TEM analysis provided clear evidence of mitochondrial damage and its prevention by Sal, future studies could aim to capture more panoramic views to document the full spectrum of necroptotic ultrastructural features, such as plasma membrane rupture. Most importantly, while pharmacological evidence strongly suggests that Sal targets the necroptosis pathway, direct binding assays and validation using genetic knockout animal models are needed to identify its precise molecular target.

From a clinical translation perspective, the pharmacokinetic profile, bioavailability, and long-term safety of Sal require systematic evaluation. Moreover, crosstalk between necroptosis and other cell death modalities (eg, apoptosis, pyroptosis^9,31,32) may influence its therapeutic efficacy. However, potential compensatory survival mechanisms resulting from excessive inhibition of cell death should be carefully monitored.

Conclusions

In conclusion, our study provides evidence that Sal alleviates AP by inhibiting necroptosis, likely through targeting the RIPK1/RIPK3/MLKL pathway. Our pharmacological data are consistent with RIPK1 being a potential target, although this requires direct genetic validation in future studies.

Abbreviations

AP, Acute pancreatitis; SAP, Severe acute pancreatitis; Sal, Salidroside; Nec-1, Necrostatin-1; AMY, Amylase; RCD, Regulated cell death; RIPK, Receptor interacting protein kinase 3; MLKL, Mixed lineage kinase like; TEM, Transmission electron microscope; H&E, Hematoxylin-eosin; IF, Immunofluorescence; IHC, Immunohistochemistry; ELISA, Enzyme-linked immunosorbent assay; IL, Interleukin; TNF, Tumor necrosis factor; CCK-8, Cell Counting Kit-8; OD, Optical density.

Data Sharing Statement

The data are available from the corresponding author upon reasonable request.

Ethics Approval Statement

All the rat experiments were conducted in accordance with the “Regulations on the Administration of Experimental Animals” (Lanzhou University) and were approved by the Ethics Committee of Lanzhou University Second Hospital (Approval No. D2024-410). At the same time,all methods are reported in accordance with ARRIVE guidelines.

Acknowledgments

We acknowledge the Cuiying Biomedical Research Center of the Lanzhou University Second Hospital for providing all the research equipments for this experiment.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Funding

This work was supported by the National Natural Science Foundation of China (82260135) and Science and Technology Program of Gansu Province (22YF7WA087).

Disclosure

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References

1. Lankisch PG, Apte M, Banks PA. Acute pancreatitis. Lancet. 2015;386(9988):85–96. doi:10.1016/S0140-6736(14)60649-8

2. Mederos MA, Reber HA, Girgis MD. Acute Pancreatitis: a Review. JAMA. 2021;325(4):382–390. doi:10.1001/jama.2020.20317

3. Gardner TB. Acute Pancreatitis. Ann Intern Med. 2021;174(2):Itc17–itc32. doi:10.7326/AITC202102160

4. Wang GJ, Gao CF, Wei D, Wang C, Ding SQ. Acute pancreatitis: etiology and common pathogenesis. World J Gastroenterol. 2009;15(12):1427–1430. doi:10.3748/wjg.15.1427

5. Petrov MS, Yadav D. Global epidemiology and holistic prevention of pancreatitis. Nat Rev Gastroenterol Hepatol. 2019;16(3):175–184. doi:10.1038/s41575-018-0087-5

6. Bertheloot D, Latz E, Franklin BS. Necroptosis, pyroptosis and apoptosis: an intricate game of cell death. Cell Mol Immunol. 2021;18(5):1106–1121.

7. Lee PJ, Papachristou GI. New insights into acute pancreatitis. Nat Rev Gastroenterol Hepatol. 2019;16(8):479–496. doi:10.1038/s41575-019-0158-2

8. Tong X, Tang R, Xiao M, et al. Targeting cell death pathways for cancer therapy: recent developments in necroptosis, pyroptosis, ferroptosis, and cuproptosis research. J Hematol Oncol. 2022;15(1):174. doi:10.1186/s13045-022-01392-3

9. Ketelut-Carneiro N, Fitzgerald KA. Apoptosis, Pyroptosis, and Necroptosis-Oh My! The Many Ways a Cell Can Die. J Mol Biol. 2022;434(4):167378. doi:10.1016/j.jmb.2021.167378

10. Khoury MK, Gupta K, Franco SR, Liu B. Necroptosis in the Pathophysiology of Disease. Am J Pathol. 2020;190(2):272–285. doi:10.1016/j.ajpath.2019.10.012

11. Liu X, Zhou M, Dai Z, et al. Salidroside alleviates ulcerative colitis via inhibiting macrophage pyroptosis and repairing the dysbacteriosis-associated Th17/Treg imbalance. Phytother Res. 2023;37(2):367–382. doi:10.1002/ptr.7636

12. Yang S, Wang L, Zeng Y, et al. Salidroside alleviates cognitive impairment by inhibiting ferroptosis via activation of the Nrf2/GPX4 axis in SAMP8 mice. Phytomedicine. 2023;114:154762. doi:10.1016/j.phymed.2023.154762

13. Huang G, Cai Y, Ren M, et al. Salidroside sensitizes Triple-negative breast cancer to ferroptosis by SCD1-mediated lipogenesis and NCOA4-mediated ferritinophagy. J Adv Res. 2024;2024:1.

14. Lei W, Chen MH, Huang ZF, et al. Salidroside protects pulmonary artery endothelial cells against hypoxia-induced apoptosis via the AhR/NF-κB and Nrf2/HO-1 pathways. Phytomedicine. 2024;128:155376. doi:10.1016/j.phymed.2024.155376

15. Ling Z, Peiwu L. Effect of salidroside on autophagy of pancreatic cells in rats with severe acute pancreatitis [J]Chin. J Emerg Med. 2021;30(1):53–58.

16. Wang X, Qian J, Meng Y, et al. Salidroside ameliorates severe acute pancreatitis-induced cell injury and pyroptosis by inactivating Akt/NF-κB and caspase-3/GSDME pathways. Heliyon. 2023;9(2):e13225. doi:10.1016/j.heliyon.2023.e13225

17. Wang X, Zhou G, Liu C, et al. Acanthopanax versus 3-Methyladenine Ameliorates Sodium Taurocholate-Induced Severe Acute Pancreatitis by Inhibiting the Autophagic Pathway in Rats. Mediators Inflamm. 2016;2016:8369704. doi:10.1155/2016/8369704

18. Wang X, Chu L, Liu C, et al. Therapeutic effects of Saussurea involucrata injection against severe acute pancreatitis- induced brain injury in rats. Biomed Pharmacother. 2018;100:564–574. doi:10.1016/j.biopha.2018.02.044

19. Wildi S, Kleeff J, Mayerle J, et al. Suppression of transforming growth factor beta signalling aborts caerulein induced pancreatitis and eliminates restricted stimulation at high caerulein concentrations. Gut. 2007;56(5):685–692. doi:10.1136/gut.2006.105833

20. Fu X, Xiu Z, Xu Q, Yue R, Xu H. Interleukin-22 Alleviates Caerulein-Induced Acute Pancreatitis by Activating AKT/mTOR Pathway. Dig Dis Sci. 2024;69(5):1691–1700. doi:10.1007/s10620-024-08360-6

21. Wang X, Qian J, Meng Y, et al. Salidroside alleviates severe acute pancreatitis-triggered pancreatic injury and inflammation by regulating miR-217-5p/YAF2 axis. Int Immunopharmacol. 2022;111:109123. doi:10.1016/j.intimp.2022.109123

22. Qian J, Wang X, Weng W, Zhou G, Zhu S, Liu C. Salidroside alleviates taurolithocholic acid 3-sulfate-induced AR42J cell injury. Biomed Pharmacother. 2021;142:112062. doi:10.1016/j.biopha.2021.112062

23. Fan H, Su BJ, Le JW, Zhu JH. Salidroside Protects Acute Kidney Injury in Septic Rats by Inhibiting Inflammation and Apoptosis. Drug Des Devel Ther. 2022;16:899–907. doi:10.2147/DDDT.S361972

24. Wu Q, Shan X, Li X, et al. Salidroside ameliorates neuroinflammation in autistic rats by inhibiting NLRP3/Caspase-1/GSDMD signal pathway. Brain Res Bull. 2025;220:111132. doi:10.1016/j.brainresbull.2024.111132

25. Wang Z, Li L, Li W, Yan H, Yuan Y. Salidroside Alleviates Furan-Induced Impaired Gut Barrier and Inflammation via Gut Microbiota-SCFA-TLR4 Signaling. J Agric Food Chem. 2024;72(29):16484–16495. doi:10.1021/acs.jafc.4c02433

26. Wang N, Gao Z, Zhan H, Jing L, Meng F, Chen M. Salidroside alleviates doxorubicin-induced hepatotoxicity via Sestrin2/AMPK-mediated pyroptotic inhibition. Food Chem Toxicol. 2025;199:115335. doi:10.1016/j.fct.2025.115335

27. He R, Wang Z, Dong S, Chen Z, Zhou W. Understanding Necroptosis in Pancreatic Diseases. Biomolecules. 2022;12(6):828. doi:10.3390/biom12060828

28. Ju IJ, Tsai BC, Kuo WW, et al. Rhodiola and Salidroside Attenuate Oxidative Stress-Triggered H9c2 Cardiomyoblast Apoptosis Through IGF1R-Induced ERK1/2 Activation. Environ Toxicol: Int J. 2024;39(11):5150–5161. doi:10.1002/tox.24372

29. Kang JS, Cho NJ, Lee SW, et al. RIPK3 causes mitochondrial dysfunction and albuminuria in diabetic podocytopathy through PGAM5-Drp1 signaling. Metabolism. 2024;159:155982. doi:10.1016/j.metabol.2024.155982

30. Ding Z, Wang R, Li Y, Wang X. MLKL activates the cGAS-STING pathway by releasing mitochondrial DNA upon necroptosis induction. Mol Cell. 2025;85(13):2610–25.e5. doi:10.1016/j.molcel.2025.06.005

31. Zheng M, Kanneganti TD. The regulation of the ZBP1-NLRP3 inflammasome and its implications in pyroptosis, apoptosis, and necroptosis (PANoptosis). Immunol Rev. 2020;297(1):26–38. doi:10.1111/imr.12909

32. Samir P, Malireddi RKS, Kanneganti TD. The PANoptosome: a Deadly Protein Complex Driving Pyroptosis, Apoptosis, and Necroptosis (PANoptosis). Front Cell Infect Microbiol. 2020;10:238. doi:10.3389/fcimb.2020.00238

Introduction

Aortic dissection (AD) is a life-threatening cardiovascular disorder with worldwide clinical significance, pathologically defined by an intimal rupture permitting blood entry into the medial layer, consequently creating separate true and false luminal spaces within the aortic wall.¹ The disease progression entails substantial architectural changes in the aorta, featuring VSMC loss, ECM breakdown, and leukocyte accumulation.² Crucially, inflammatory responses play a central mechanistic role in AD development. Invading immune cells (notably lymphocytes and macrophages) enhance protease and adhesion molecule production while releasing reactive oxygen species. These cells additionally trigger VSMC apoptosis, ultimately causing intimal deterioration – the principal pathological process driving AD formation.³ Although current management relying on surgical and endovascular techniques has advanced, the persistent lack of precise molecular targets and effective drug therapies emphasizes the critical necessity for discovering robust biomarkers to enable timely diagnosis and targeted treatment development.

Lactylation, an essential post-translational modification, significantly influences diverse biological pathways. Conventionally, lactate was viewed solely as the end-product of glycolysis.⁴ However, a paradigm shift occurred in 2019 when Professor Yingming Zhao’s research team identified lactylation as a novel protein modification occurring on lysine residues, thereby revolutionizing research in this field. This enzymatic process, catalyzed by specific enzymes, facilitates the attachment of lactate groups to lysine residues, subsequently modifying protein charge, structure, and function, and ultimately influencing diverse biological processes.⁵ The implications of lactylation extend across multiple cellular physiological activities. During tumor development, lactylation regulates essential oncogenic processes such as cancer cell metabolism, proliferation, invasion and metastatic spread via functional modulation of proteins.⁶ Within the cardiovascular domain, lactylation significantly contributes to various inflammatory responses and impacts cardiovascular system functionality.^7–9 Moreover, this modification critically governs immune regulation, profoundly altering immune cell functionality and response dynamics.

Given lactylation’s substantial involvement in cardiovascular diseases and inflammation,^10–14 coupled with the absence of reported expression and functions of lactylation-associated genes in AD, this study aims to explore potential connections between lactylation-associated genes, AD, and immunological characteristics. This investigation seeks to provide novel directions and insights for early AD diagnosis. Integrative analysis of bulk and single-cell transcriptomic data enabled identification of lactylation-associated genes in AD and uncovered their correlation with immune microenvironment characteristics.This research is anticipated to yield promising biomarkers for AD diagnosis and establish a novel approach for AD detection, thereby contributing to the theoretical foundation and technical support for early and precise diagnosis of aortic dissection.

Methods

Data Processing

Transcriptomic profiles from two independent GEO cohorts (GSE52093, n=12; GSE98770, n=11) were integrated for analysis. Corresponding clinical information and normalization files were also acquired from the GEO database. To ensure data consistency, batch effects were addressed through principal component analysis (PCA) and the application of the “ComBat” function from the “SVA” R package. This processing yielded a comprehensive AD cohort comprising 23 samples and 18,249 genes, suitable for subsequent analysis.

Identification of Differentially Expressed Genes Between AD and Healthy Individuals

Utilizing the integrated dataset, we performed differential expression analysis employing the limma package (version 4.2.1) in RStudio. DEGs were screened using cutoff criteria of |log2FC| > 0.58 (1.5-fold) and adjusted p-value < 0.05, with analytical results presented in volcano and heatmap visualizations.¹⁵ Subsequently, systematic functional annotation of significant DEGs was performed via Gene Ontology (GO) analysis and KEGG pathway enrichment.

Screening of Differentially Expressed Lactylation-Related Genes in AD

Based on comprehensive literature review, we identified 323 lactylation-related genes(Supplementary Data Sheet 1). Integration of differential expression results with lactylation-related genes enabled identification of lactylation-associated DEGs (LDEGs) through intersection analysis. Functional annotation of LDEGs was subsequently conducted using R-based clusterProfiler, incorporating GO enrichment and KEGG pathway analyses to delineate their biological roles. The functional enrichment results were visualized with R’s ggplot2 package to facilitate data interpretation.

Construction of Protein-Protein Interaction Networks

To elucidate the functional characteristics and metabolic pathways of differentially expressed lactylation-associated genes, we utilized the GeneMANIA database (http://genemania.org), a comprehensive online resource for gene function prediction and interaction network construction. This platform enables rapid generation of gene interaction networks by analyzing input gene lists, thereby elucidating relationships between genes and their interaction partners. GeneMANIA integrates diverse data sources and employs sophisticated network weighting models to predict functions of unknown genes based on their interactions with functionally characterized genes. In our study, we imported the differentially expressed lactylation-related genes into this platform to construct and visualize a protein-protein interaction (PPI) network, facilitating comprehensive analysis of gene interactions and functional associations.

Detection of Lactylation-Associated Hub Genes in Aortic Dissection Cases

Three machine learning approaches were employed to analyze pre-filtered lactylation-associated DEGs for hub gene identification in AD: (1) LASSO regression via R’s “glmNETs” package; (2) RF and SVM-REF analyses using “randomForest” and “kernlab” packages respectively. Integration of feature genes from all algorithms identified lactylation-related hub genes, followed by ROC curve analysis with AUC calculation (using “pROC”) to assess diagnostic performance for AD.

Expression, Correlation, and Gene Enrichment Analysis of Hub Genes

Conducting a correlation analysis and visualization between the selected hub genes and other differentially expressed genes using the “corrplot” package in R. Subsequently, analyze the differential expression of each hub gene utilizing the Wilcoxon rank-sum test. Finally, perform Gene Set Enrichment Analysis (GSEA) on the Hallmark gene sets (http://software.broadinstitute.org/gsea/msigdb/) for each hub gene using the R package “clusterProfiler” to elucidate the enriched pathways and functions associated with the hub genes.

Immune Infiltration Analysis

Existing evidence confirms immune cell involvement in AD pathophysiology. Using CIBERSORT, we profiled infiltration patterns of 22 immune cell subtypes in AD cohorts to examine hub gene-immune cell interactions.¹⁶ Immune cell correlations and infiltration variations were displayed via R’s “corrplot” and “ggplot2”, while Spearman analysis (“ggstatsplot”) revealed hub gene-immune cell associations, providing mechanistic insights into AD immunopathology.

External Dataset Validation

External validation using the GSE153434 dataset confirmed hub gene expression profiles, with ROC analysis evaluating their diagnostic potential.

scRNA-Seq Analysis

The scRNA-seq dataset (GSE213740: 6 AD cases vs 3 controls) was processed with Seurat in R. Quality control excluded cells with: >15% mitochondrial genes, >3% ribosomal genes, >0.1% erythrocyte genes, or gene counts <200/>7500. Post-QC, PCA-based dimensionality reduction preceded cell clustering (resolution=0.05) and visualization. Cluster-specific DEGs were identified using FindMarkers (log2FC>0.5, p<0.05), with top 5 markers visualized. Cellular lactylation levels were quantified via singscore for AD-normal comparisons.

Specimen Collection and RT-qPCR

The inclusion criteria for aortic dissection cases comprised patients who were: (1) diagnosed with aortic dissection through aortic CTA angiography, (2) underwent aortic artificial vessel replacement surgery, and (3) were aged over 18 years. Normal aortic tissues were collected from individuals undergoing coronary artery bypass grafting procedures. This study incorporated a total of nine clinical samples for differential gene expression validation, including six aortic dissection cases and three control cases (with aortic wall tissue obtained from the perforation site of the ascending aorta during coronary artery bypass grafting). The study protocol was conducted in strict compliance with the ethical principles outlined in the Helsinki Declaration. All experimental procedures and study protocols were reviewed and approved by the Ethics Committee of the Affiliated Hospital of Nantong University (Ethical approval number:2024-K247-01), and written informed consent was obtained from all participants prior to their inclusion in the study.

Total RNA extraction from aortic tissues was conducted with TRIzol reagent (ACCURATE BIOTECHNOLOGY). cDNA synthesis utilized HiScript II Q RT SuperMix (+gDNA wiper, Vazyme R223-01). qPCR amplification was performed with ChamQ SYBR Master Mix (Vazyme Q311-02), normalized to β-Actin expression (primers in Table 1). The 2–ΔΔCT method calculated relative expression, with p<0.05 considered statistically significant.

Table 1 Primer Sequences of Hub Genes

Drug Prediction and Molecular Docking Validation

To advance the identification of prospective therapeutic agents for aortic dissection, this research leveraged the Drug-Gene Interaction Database (DGIdb) to pinpoint targeted drugs corresponding to critical biological targets. The drugs that received the highest scores from this prediction were subsequently subjected to validation through molecular docking experiments with those key targets. To facilitate this process, the two-dimensional structures of small molecule ligands were sourced from the PubChem database (http://pubchem.ncbi.nlm.nih.gov/). These 2D structures were then converted into three-dimensional representations using ChemOffice software, and the resulting 3D structures were saved in mol2 file format for further analysis. For the molecular docking studies, high-resolution crystal structures of the relevant protein targets were obtained from the RCSB Protein Data Bank (RCSB PDB) (http://www.rcsb.org/). These structures were utilized as the docking receptors in our simulations. To prepare the proteins for the docking process, PyMOL software was employed to eliminate water molecules and phosphate groups, resulting in cleaned protein structures that were saved as PDB files. The molecular docking itself was conducted using AutoDock Vina 1.5.6 software, which facilitated the investigation of interactions between the proteins and the ligands. Throughout this process, the structures of both the proteins and the small molecule ligands underwent several modifications; hydrogen atoms were added to the proteins, water molecules were removed, and hydrogen atoms, along with specific torsional degrees of freedom for the small molecule ligands, were carefully managed. Following these preparations, the docking box coordinates were established.Finally, by analyzing and comparing the docking scores, the most favorable conformation from the molecular simulations was ultimately identified. Visualization and analysis of the interaction patterns between the candidate compounds and critical amino acids were conducted using PyMOL and Discovery Studio 2019 software, elucidating the 2D and 3D interaction diagrams that are crucial for understanding the binding characteristics of the predicted therapeutic agents.

Result

The Overall Expression Profile of AD Patients

After integrating and removing batch effects from two AD-related datasets (GSE52093 and GSE98770), we achieved expression profiles that included 13 patients diagnosed with AD and 10 healthy individuals (Figure 1A–D). The analysis for differential expression found 423 genes that were upregulated and 470 that were downregulated in patients with AD (Figure 2A). The 50 genes with the most significant differential expression were depicted using hierarchical clustering in a heatmap (Figure 2B). Following this, GO enrichment analysis of the differentially expressed genes (Figure 3A) indicated notable enrichment in biological processes linked to “mitotic cell cycle phase transition”, cellular components related to “collagen-containing extracellular matrix”, and molecular functions primarily associated with “actin binding.” Furthermore, an analysis of pathways using the KEGG revealed significant connections between AD and both the “cell cycle” as well as the “p53 signaling pathway” (Figure 3B).

Figure 1 Analysis of the merged and corrected two GEO datasets. (A and B) Expression profiles before and after raw data correction. (C and D) Differential gene expression profiles before and after data normalization.

Figure 2 Differential expression analysis of the AD patient dataset. (A) Volcano plot of differentially expressed genes. (B) Heatmap of the top 50 differentially expressed genes.

Figure 3 Functional enrichment analysis of differentially expressed genes. (A) GO enrichment analysis. (B) KEGG enrichment analysis.

Expression Characteristics of Lactylation-Related Genes

Following the acquisition of expression profiles, we examined 332 lactylation-associated genes and their differential expression patterns. The analysis revealed that, in contrast to three downregulated genes, ten lactylation-related genes were significantly upregulated in Alzheimer’s disease (AD) patients (Figure 4A and B). In order to substantiate these findings further, we conducted a functional analysis on 13 genes associated with lactylation that were expressed differentially. The GO analysis revealed notable enrichment in biological processes concerning “protein localization within organelles”, molecular functions related to “binding specific to protein domains”, and cellular components tied to “histone deacetylase binding” (Figure 4C–E). Furthermore, the KEGG pathway analysis uncovered significant changes in the pathways of “vascular smooth muscle contraction” and “cellular senescence” (Figure 4F).

Figure 4 Functional Analysis of Lactylation-Related Genes. (A and B) Differentially expressed genes related to lactylation. (C–E). GO enrichment analysis of lactylation-related genes. (F) KEGG enrichment analysis of lactylation-related genes.

Identification of Lactylation-Related Central Genes

To systematically identify hub genes associated with lactylation, we employed a multi-method approach. First, LASSO regression analysis was performed on the 13 differentially expressed genes, yielding 5 feature genes (Figure 5A and B). Subsequently, random forest ranking was applied to prioritize lactylation-related genes, identifying the top 10 candidates (Figure 5C). Further refinement using the SVM-REF algorithm revealed 4 feature genes (Figure 5D). Through integration of results from these three complementary methods, we identified three hub genes for lactylation: CALM1, PTBP1, and PARP1 (Figure 5E). The diagnostic potential of these hub genes was evaluated using ROC analysis, demonstrating their robust ability to distinguish AD patients from healthy controls (Figure 5F). Finally, we constructed a protein-protein interaction network for these three differentially expressed lactylation-related hub genes using the GeneMANIA database (Supplementary Figure 1).

Figure 5 Identification of Lactylation Hub Genes. (A–B) LASSO regression identified 5 feature genes. (C) Random Forest ranked the importance of 10 feature genes. (D) SVM support vector machine algorithm selected 4 feature genes. (E) The intersection of feature genes obtained from the three methods yielded 3 hub genes. (F) ROC curves of the 3 genes for predicting disease occurrence.

Expression, Correlation, and Gene Enrichment Analysis of Hub Genes

After identifying the hub genes, we performed comprehensive correlation analyses between these three hub genes and all other genes, visualizing the top three genes showing significant positive and negative correlations. Our findings revealed that RILPL1, PLS3, and GABARAPL2 exhibited significant positive correlations with CALM1, while TRIP12, SMARCC1, and CYB561D2 showed significant negative correlations with CALM1 (Supplementary Figure 2). Similarly, FES, TFDP1, and ELF3 demonstrated significant positive correlations with PARP1, whereas DNAJC24, DIXDC1, and AFTPH displayed significant negative correlations with PARP1 (Supplementary Figure 3). Furthermore, CANT1, ILF3, and XPO6 were significantly positively correlated with PTBP1, while CRBN, CD2AP, and TMEM106B showed significant negative correlations with PTBP1 (Supplementary Figure 4).

Subsequently, we investigated the roles of CALM1, PARP1, and PTBP1 in AD by comparing their expression profiles between normal and AD groups. The findings indicated that CALM1 levels were significantly elevated in the normal group in comparison to the AD group. Conversely, expression levels of PARP1 and PTBP1 were significantly increased in the AD group when compared to the normal group (Figure 6A). Additionally, we performed Gene Set Enrichment Analysis (GSEA) on these central genes (Figure 6B). This analysis uncovered a positive correlation between CALM1 levels and the myogenesis pathway, alongside a negative correlation with the glycolysis and oxidative phosphorylation pathways. Meanwhile, PARP1 expression demonstrated positive links to the NF-κB signaling pathway and the inflammatory response pathway, while showing a negative relationship with the myogenesis pathway. In a similar manner, PTBP1 expression was positively associated with both the NF-κB signaling and inflammatory response pathways, but negatively associated with the myogenesis pathway.

Figure 6 Characterization of Hub Gene Expression and GSEA Enrichment Analysis. (A) Expression of hub genes, with red indicating the normal group and blue representing the AD group. (B) GSEA enrichment results of hub genes.

The Correlation Between Key Target Expression and Immunological Characteristics in AD Patients

We conducted a comprehensive analysis of immune cell infiltration levels in patients with aortic dissection (AD) and healthy controls. The results revealed that healthy individuals exhibited significantly higher infiltration levels of γδT cells and resting mast cells compared to the AD group, while showing significantly lower levels of resting natural killer cell infiltration (Figure 7A and C). To further investigate the interrelationships among immune cells, we performed correlation analysis to elucidate potential interactions and their implications for immune-inflammatory mechanisms in AD (Figure 7B).

Figure 7 Immunoinfiltration Analysis. (A) The relative proportions of 22 immune cell subsets in all samples from the AD dataset. (B) The correlations of the 22 immune cell subsets. (C) Differences in the levels of 22 immune cell types between the normal group and the AD group.

Our analysis demonstrated a positive correlation between γδT cells and both M1 macrophages and resting mast cells, suggesting potential synergistic interactions and functional interdependence among these cell populations in the immune response to aortic dissection. Conversely, we observed negative correlations between resting natural killer cells and both resting mast cells and γδT cells, potentially reflecting disease-induced imbalances in the immune system’s inflammatory response, immune regulation, and immune surveillance capabilities. These findings highlight the crucial role of immune cell interactions in the pathogenesis and progression of aortic dissection.

Furthermore, we identified significant correlations between the three hub genes and specific immune cell populations (Figure 8). CALM1 expression showed a significant negative correlation with plasma cell infiltration, while demonstrating positive correlations with CD8-positive T cells, resting mast cells, and γδT cells. PARP1 expression exhibited a negative correlation with γδT cell infiltration and positive correlations with CD4-positive naive T cells and resting NK cells. PTBP1 expression was negatively correlated with γδT cell infiltration and positively correlated with resting NK cells, activated dendritic cells, monocytes, and naive B cells. These findings suggest complex interactions between lactylation-related genes and immune cell populations in the context of aortic dissection.

Figure 8 Correlation between Hub Genes and Immune Cells.

Validation of Hub Genes in External Datasets

To assess the reliability of hub genes related to lactylation that are differentially expressed in AD, we analyzed the expression profiles of three lactylation-associated genes—CALM1, PTBP1, and PARP1—using the independent dataset GSE153434. Our comparative analysis indicated that the expression of CALM1 was significantly reduced in the AD group when compared to the normal control group. In contrast, we found that PTBP1 expression was significantly increased in the AD group relative to the normal controls (Figure 9A and B). Interestingly, PARP1 expression did not demonstrate any notable differences between the two groups (Figure 9C). The receiver operating characteristic (ROC) curve analysis revealed that both CALM1 and PTBP1 had strong diagnostic capabilities, with area under the curve (AUC) values of 0.93 and 0.83, respectively. On the other hand, PARP1 exhibited minimal diagnostic value, yielding an AUC of 0.50 (Figure 9D).

Figure 9 External Dataset Validation of Hub Genes. (A–C) Expression characteristics of hub genes in the external dataset, with red representing the normal group and blue representing the AD group. (D) ROC curves of hub genes for predicting disease occurrence in the external dataset.

Identification of Cell Clusters in AD Patients

The single-cell dataset GSE213740, comprising six aortic dissection samples and three normal controls, was obtained from the GEO database. Initial quality control measures were implemented, with Figure 10A and B illustrating the distribution of gene counts (nFeature), sequencing depth (nCount), and mitochondrial gene percentage (percent.mt). For subsequent analysis, the top 2000 highly variable genes were selected, with the top 10 most variable genes displayed in Figure 10C. Based on the analysis presented in Figure 11A, we determined the optimal number of principal components (PCs) to be 9. Cluster analysis results suggested an appropriate resolution of 0.1 (Figure 11B), with marker genes for each cell population visualized through heatmap analysis (Figure 12A). Utilizing the UMAP algorithm, we classified the single-cell sequencing samples into 11 distinct clusters (Figure 12B), which were subsequently identified as 11 immune cell populations (Figure 12C). These populations included endothelial cells, smooth muscle cells, mesenchymal cells, macrophages, T cells, B cells, monocytes, mast cells, plasma cells, fibroblasts, M1 macrophages, and M2 macrophages. Differential expression analysis across all 11 cell populations identified the top five highly and lowly expressed genes in each population (Figure 13A).

Figure 10 Data Quality Control. (A and B) Changes in single-cell dataset quality control before and after. (C) Top 10 highly variable gene markers.

Figure 11 Selection of Principal Component Numbers and Resolution. (A) Heatmap is used to display the expression levels of marker genes for the top 20 PCs. (B) Clustering tree is used to select the appropriate resolution.

Figure 12 Cell Population Clustering. (A) Heatmap of marker genes for each cell population. (B) UMAP algorithm divides cells into 11 subgroups. (C) 11 cell subgroups are annotated as 11 immune cell types.

Figure 13 Differential Expression Analysis. (A) Differentially expressed genes (top 5) in each cell cluster. (B) Expression profiles of hub genes across different cell clusters.

Further investigation of lactylation hub gene expression across these 11 immune cell types revealed distinct expression patterns (Figure 13B). CALM1 exhibited high expression in smooth muscle cells, B cells, and fibroblasts, while showing low expression in endothelial cells, mesenchymal cells, mast cells, and M2 macrophages. PTBP1 demonstrated elevated expression in endothelial cells, T cells, B cells, mast cells, fibroblasts, and M1 macrophages, but was minimally expressed in smooth muscle cells, mesenchymal cells, monocytes, and M2 macrophages. Similarly, PARP1 showed high expression levels in T cells, B cells, fibroblasts, and M1 macrophages, with reduced expression observed in mesenchymal cells, monocytes, and mast cells.

Single-Cell Analysis Validates Lactylation and Immune Interactions in AD

To further investigate the immune-related interactions between lactylation and AD, we performed single-cell scoring analysis of the lactylation-related gene set. Our analysis revealed elevated lactylation levels in fibroblasts, smooth muscle cells, monocytes, and T cells, while other immune cell types showed no significant differences in lactylation levels (Supplementary Figure 5A). Subsequent comparative analysis of lactylation gene set scores between AD and normal samples demonstrated significantly higher lactylation scores in the AD group. This finding suggests that aortic tissues from AD patients exhibit elevated lactylation levels compared to normal controls. Furthermore, we observed significant differences in lactylation levels of immune cells between the two groups (Supplementary Figure 5B).

Clinical Sample Validation

RT-qPCR was utilized to assess the expression profiles of CALM1, PTBP1, and PARP1 in clinical samples, with comparisons made between patients with aortic dissection (AD) (n = 6) and normal controls (n = 3). The results indicated a notable decrease in CALM1 expression within the AD cohort relative to the normal controls. Conversely, the expression levels of PTBP1 and PARP1 did not exhibit any statistically significant variations between the AD and control groups (Figure 14).

Figure 14 RT-qPCR detection of hub gene mRNA expression levels. ns, no significant difference, *p < 0.05.

Drug Prediction and Molecular Docking Validation

Utilizing the DGIdb database, we forecasted potential drugs aimed at the key biomarker CALM1 and ultimately pinpointed the top five highest-scoring candidate drugs (Figure 15A). Among these candidates, PRENYLAMINE, which showed the highest targeting score, was chosen for molecular docking validation against CALM1. Generally, a binding energy that falls below −5.0 kcal/mol is indicative of favorable binding activity between the ligand and its target protein, where lower binding energy values signify greater binding affinity, increased stability, and more advantageous conformational interactions. Our findings revealed that PRENYLAMINE reached a binding energy of −6.6 kcal/mol when interacting with CALM1 (Figure 15B), indicating its potential as a therapeutic option for aortic dissection.

Figure 15 Drug Prediction and Molecular Docking Validation.(A) Potential targeted drugs for CALM1.(B) Molecular docking results of PRENYLAMINE with CALM1.

Discussion

The current understanding of aortic dissection (AD) pathogenesis recognizes inflammation, apoptosis, endothelial cell dysfunction, phenotypic transformation of smooth muscle cells, and extracellular matrix degradation as critical pathological components, with emerging evidence implicating epigenetic modifications in these processes.^17–21 While metabolic reprogramming and its epigenetic consequences are increasingly recognized in cardiovascular pathologies, the specific role of lactate accumulation, and its associated lactylation modifications in AD development remains unexplored. Our study bridges this knowledge gap by identifying three lactylation-associated hub genes (CALM1, PARP1, and PTBP1) through integrated RNA sequencing analysis and investigated their relationship with immune cell infiltration patterns.

CALM1: A Calcium Signaling Nexus in AD Pathogenesis

Firstly, through comprehensive transcriptomic analysis, we identified CALM1 as a potential lactylation-related biomarkers distinguishing AD patients from healthy individuals. As the most evolutionarily conserved calmodulin isoform,^22,23 CALM1 regulates fundamental cellular process including motility, differentiation, and proliferation,²⁴ with established roles in oncogenesis,²⁵ neurodegeneration,²⁶ and fibrotic disorders.²⁷ Mechanistically, CALM1 overexpression activates PI3K-Akt pathway, subsequently suppressing hepatic gluconeogenesis – a pathway critically involved in apoptosis regulation, oxidative stress responses, and inflammatory signaling.²⁸

Our network analysis extends these finding by revealing novel association between CALM1 expression and mTORC1 signaling pathway, glycolysis – oxidative phosphorylation coupling, and the infiltration patterns of nearly all immune cell types. These associations suggest CALM1 as a potential orchestrator of the pro-inflammatory microenvironment characteristic of AD. However, further experimental investigations are necessary to fully elucidate the specific functions and molecular mechanisms of CALM1 in the pathogenesis of AD.

PARP1: Metabolic-Inflammatory Crosstalk

PARP1, also known as Poly(ADP-ribose) polymerase 1, is an enzyme that is highly conserved and dependent on NAD⁺. It is crucial for the regulation of various intracellular physiological processes, such as the repair of DNA damage, the regulation of gene transcription, chromatin remodeling, and the modulation of apoptosis.^29–31 The relationship between PARP1 and lactate/lactylation is both intricate and significant. Lactate exerts substantial influence on PARP1, not only by directly inhibiting its activity but also by indirectly modulating its function through alterations in cellular energy metabolism and redox balance.³² In chronic inflammation, excessive PARP1 activation establishes a pathogenic feedback loop by amplifying NF-κB-mediated TNFα/IL-6 production.^33,34 Our findings demonstrate elevated PARP1 expression levels in aortic tissues of AD patients compared to normal controls. This correlations between PARP1 elevation and CD4T cell/mast cell infiltration align with the previous paradigm, suggesting PARP1-mediated immune dysregulation contributes to AD progression.

PTBP1: Metabolic Reprogramming Driver

The RNA-binding protein PTBP1 emerged as a third lactylation-associated factor elevated in AD tissues. As one member of heterogeneous nuclear ribonucleoprotein (hnRNP) subfamily, PTBP1is encoded on human chromosome 19p13.3.³⁵ In the metabolism of cancer cells, PTBP1 enhances the production of the M2 isoform of pyruvate kinase (PKM2) and concurrently inhibits the expression of PKM1, resulting in a metabolic transition from oxidative phosphorylation to glycolysis, which enhances glycolytic flux and lactate accumulation, significantly influencing tumor initiation and progression.^36,37 Meanwhile, recent studies identifies lactylation-modified PTBP1 as a glycolysis amplifier in glioma stem cells via PFKFB4 mRNA stabilization.³⁸

Our single-cell analysis reveals broad PTBP1 upregulation across AD aortic cell types (endothelial cells, T cells, B cells, mast cells, fibroblasts, and M1 macrophages), suggesting it may similarly drive glycolytic flux and lactate accumulation in aortic tissues. The resultant lactylation surge could destabilize immune homeostasis through post-translational modification of regulatory proteins – a hypothesis requiring experimental verification. These findings demonstrate that PTBP1 likely functions as a promoting factor in the pathogenesis and progression of aortic dissection.

Lactylation Landscape in AD Microenvironment

Recent advancements in epigenetic research have demonstrated that lactate-induced histone lactylation plays a significant role in modulating cellular physiological functions and immune cell modifications.³⁹ To investigate this phenomenon in AD, we conducted a comprehensive analysis of cellular lactylation levels using single-cell sequencing data, comparing AD patients with healthy individuals. The results revealed significant lactylation elevation in AD lesions, particularly within fibroblasts, smooth muscle cells, and myeloid populations. While B/T lymphocytes exhibited relatively lower lactylation levels, their functional sensitivity to lactylation-mediated epigenetic changes warrants further investigation. The spatial correlation between lactate accumulation (from enhanced glycolysis) and lactylation patterns suggests a self-reinforcing metabolic-epigenetic circuit in AD pathogenesis.

Clinical Validation and Limitations

Finally, we performed experimental validation of the three hub genes using clinical samples, which demonstrated statistically significant differences in CALM1 expression between AD and control groups, whereas PARP1 and PTBP1 showed no significant differential expression. This divergence highlights the complex translation from bioinformatic prediction to biomarker validation, potentially reflecting post-transcriptional regulation or tissue-specific expression patterns. These results indicate that CALM1, rather than PARP1 or PTBP1, may serve as a promising biomarker for AD diagnosis.

While our multi-modal analysis implicates lactylation in AD-associated immune dysregulation, several limitations must be acknowledged: 1) The observational nature of human tissue studies precludes causal inference; 2) Lactylation’s cell-type-specific effects require functional validation in experimental models; 3) The diagnostic utility of lactylation markers needs prospective clinical evaluation.

Conclusion

Through a multidisciplinary investigation combining bioinformatic analyses and experimental validation, we conducted a systematic exploration of lactylation-mediated regulatory mechanisms in AD pathogenesis. Our findings demonstrate that CALM1 emerges as a novel and promising diagnostic biomarker, showing robust discriminative capacity between AD patients and healthy controls through comprehensive validation across multiple analytical platforms. Complementing these discoveries, high-resolution single-cell RNA sequencing revealed cell type-specific lactylation patterns within the AD aortic microenvironment, delineating previously unrecognized epigenetic regulatory networks across diverse cellular subpopulations. These results substantially advance our understanding of metabolic-epigenetic crosstalk in vascular remodeling disorders, providing mechanistic insights that bridge lactylation dynamics with AD progression. The identification of CALM1 as a lactylation-associated diagnostic indicator, coupled with our comprehensive cellular atlas of lactylation modifications, establishes a critical foundation for future investigations into AD’s molecular pathology. This work not only elucidates new pathophysiological dimensions of aortic dissection but also presents tangible translational applications, potentially enabling the development of lactylation-targeted diagnostic panels and therapeutic interventions for this life-threatening cardiovascular condition.

Data Sharing Statement

The datasets supporting the conclusions of this article are available in the GEO database, https://www.ncbi.nlm.nih.gov/geo/.

Ethics Approval and Consent to Participate

All experimental procedures and study protocols were reviewed and approved by the Ethics Committee of the Affiliated Hospital of Nantong University.All studies were conducted in accordance.

Funding

This work was supported by grants from Jiangsu Provincial Research Hospital (YJXYY202204-YSB58).

Disclosure

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Zhiming Yu, Yue Pan and Xiaoyu Qian contribute equally to this article as co-first author.

References

1. Vilacosta I, San Román JA, Di Bartolomeo R, et al. Acute aortic syndrome revisited: jacc state-of-the-art review. J Am Coll Cardiol. 2021;78(21):2106–2125. doi:10.1016/j.jacc.2021.09.022

2. Chen Y, Zhang T, Yao F, et al. Dysregulation of interaction between LOXhigh fibroblast and smooth muscle cells contributes to the pathogenesis of aortic dissection. Theranostics. 2022;12(2):910–928. doi:10.7150/thno.66059

3. Luo F, Zhou XL, Li JJ, et al. Inflammatory response is associated with aortic dissection. Ageing Res Rev. 2009;8(1):31–35. doi:10.1016/j.arr.2008.08.001

4. Chen AN, Luo Y, Yang YH, et al. Lactylation, a novel metabolic reprogramming code: current status and prospects. Front Immunol. 2021;12:688910. doi:10.3389/fimmu.2021.688910

5. Zhang D, Tang Z, Huang H, et al. Metabolic regulation of gene expression by histone lactylation. Nature. 2019;574(7779):575–580. doi:10.1038/s41586-019-1678-1

6. Malsi K, Xie S, Cai Y, et al. The role of lactylation in tumor growth and cancer progression. Front Oncol. 2025;15. doi:10.3389/fonc.2025.1516785

7. She H, Hu Y, Zhao G, et al. Dexmedetomidine ameliorates myocardial ischemia-reperfusion injury by inhibiting MDH2 lactylation via regulating metabolic reprogramming. Adv Sci. 2024;11(48):e2409499. doi:10.1002/advs.202409499

8. Wang J, Yang P, Yu T, et al. Lactylation of PKM2 suppresses inflammatory metabolic adaptation in pro-inflammatory macrophages. Int J Biol Sci. 2022;18(16):6210–6225. doi:10.7150/ijbs.75434

9. Wu D, Spencer CB, Ortoga L, et al. Histone lactylation-regulated METTL3 promotes ferroptosis via m6A-modification on ACSL4 in sepsis-associated lung injury. Redox Biol. 2024;74:103194. doi:10.1016/j.redox.2024.103194

10. Zhang N, Zhang Y, Xu J, et al. α-myosin heavy chain lactylation maintains sarcomeric structure and function and alleviates the development of heart failure. Cell Res. 2023;33(9):679–698. doi:10.1038/s41422-023-00844-w

11. Li X, Chen M, Chen X, et al. TRAP1 drives smooth muscle cell senescence and promotes atherosclerosis via HDAC3-primed histone H4 lysine 12 lactylation. Eur Heart J. 2024;45(39):4219–4235. doi:10.1093/eurheartj/ehae379

12. Xu X, Zhang DD, Kong P, et al. Sox10 escalates vascular inflammation by mediating vascular smooth muscle cell transdifferentiation and pyroptosis in neointimal hyperplasia. Cell Rep. 2023;42(8):112869. doi:10.1016/j.celrep.2023.112869

13. An S, Yao Y, Hu H, et al. PDHA1 hyperacetylation-mediated lactate overproduction promotes sepsis-induced acute kidney injury via Fis1 lactylation. Cell Death Dis. 2023;14(7):457. doi:10.1038/s41419-023-05952-4

14. Lin J, Ren J. Lactate-induced lactylation and cardiometabolic diseases: from epigenetic regulation to therapeutics. Biochim Biophys Acta Mol Basis Dis. 2024;1870(6):167247. doi:10.1016/j.bbadis.2024.167247

15. Phipson B, Wu D, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. doi:10.1093/nar/gkv007

16. Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12(5):453–457. doi:10.1038/nmeth.3337

17. Cui H, Chen Y, Li K, et al. Untargeted metabolomics identifies succinate as a biomarker and therapeutic target in aortic aneurysm and dissection. Eur Heart J. 2021;42(42):4373–4385. doi:10.1093/eurheartj/ehab605

18. Luo S, Kong C, Zhao S, et al. Endothelial HDAC1-ZEB2-NuRD complex drives aortic aneurysm and dissection through regulation of protein S-Sulfhydration. Circulation. 2023;147(18):1382–1403. doi:10.1161/CIRCULATIONAHA.122.062743

19. Yin Z, Zhang J, Zhao M, et al. EDIL3/Del-1 prevents aortic dissection through enhancing internalization and degradation of apoptotic vascular smooth muscle cells. Autophagy. 2024;20(11):2405–2425. doi:10.1080/15548627.2024.2367191

20. Chakraborty A, Li Y, Zhang C, et al. Epigenetic induction of smooth muscle cell phenotypic alterations in aortic aneurysms and dissections. Circulation. 2023;148(12):959–977. doi:10.1161/CIRCULATIONAHA.123.063332

21. Rombouts KB, van Merrienboer TAR, Ket JCF, et al. The role of vascular smooth muscle cells in the development of aortic aneurysms and dissections. Eur J Clin Invest. 2022;52(4):e13697. doi:10.1111/eci.13697

22. Ikura M, Ames JB. Genetic polymorphism and protein conformational plasticity in the calmodulin superfamily: two ways to promote multifunctionality. Proceedings National Academy Sci United States Am. 2006;103(5):1159–1164. doi:10.1073/pnas.0508640103

23. Zhang M, Abrams C, Wang L, et al. Structural basis for calmodulin as a dynamic calcium sensor. structure. Structure. 2012;20(5):911–923. doi:10.1016/j.str.2012.03.019

24. Chin D, Means AR. Calmodulin: a prototypical calcium sensor. Trends Cell Biol. 2000;10(8):322–328. doi:10.1016/S0962-8924(00)01800-6

25. Yao M, Fu L, Liu X, et al. In-Silico multi-omics analysis of the functional significance of calmodulin 1 in multiple cancers. Front Genet. 2022;12:793508. doi:10.3389/fgene.2021.793508

26. Esteras N, Alquézar C, de la Encarnación A, et al. Calmodulin levels in blood cells as a potential biomarker of Alzheimer’s disease. Alzheimers Res Ther. 2013;5(6):55. doi:10.1186/alzrt219

27. Ji D, Chen GF, Wang JC, et al. Identification of TAF1, HNF4A, and CALM2 as potential therapeutic target genes for liver fibrosis. J Cell Physiol. 2019;234(6):9045–9051. doi:10.1002/jcp.27579

28. Ko HL, Ren EC. Functional aspects of PARP1 in DNA repair and transcription. Biomolecules. 2012;2(4):524–548. doi:10.3390/biom2040524

29. Hussain M, Khadka P, Pekhale K, et al. RECQL4 requires PARP1 for recruitment to DNA damage, and PARG dePARylation facilitates its associated role in end joining. Exp Mol Med. 2025;57(1):264–280. doi:10.1038/s12276-024-01383-z

30. Swindall AF, Stanley JA, Yang ES. PARP-1: friend or Foe of DNA damage and repair in tumorigenesis? Cancers. 2013;5(3):943–958. doi:10.3390/cancers5030943

31. Conceição CJF, Moe E, Ribeiro PA, et al. PARP1: a comprehensive review of its mechanisms, therapeutic implications and emerging cancer treatments. Biochim Biophys Acta Rev Cancer. 2025;11(2):189282. doi:10.1016/j.bbcan.2025.189282

32. Wang L, Liang C, Li F, et al. PARP1 in carcinomas and PARP1 inhibitors as antineoplastic drugs. Int J Mol Sci. 2017;18(10):2111. doi:10.3390/ijms18102111

33. Demény MA, Virág L. The PARP enzyme family and the hallmarks of cancer part 2: hallmarks related to cancer host interactions. Cancers. 2021;13(9):2057. doi:10.3390/cancers13092057

34. Won SJ, Jang BG, Yoo BH, et al. Prevention of acute/severe hypoglycemia-induced neuron death by lactate administration. J Cereb Blood Flow Metab. 2012;32(6):1086–1096. doi:10.1038/jcbfm.2012.30

35. Pina JM, Hernandez LA, Keppetipola NM. Polypyrimidine tract binding proteins PTBP1 and PTBP2 interact with distinct proteins under splicing conditions. PLoS One. 2022;17(2):e0263287. doi:10.1371/journal.pone.0263287

36. Shinohara H, Kumazaki M, Minami Y, et al. Perturbation of energy metabolism by fatty-acid derivative AIC-47 and imatinib in BCR-ABL-harboring leukemic cells. Cancer Lett. 2016;371(1):1–11. doi:10.1016/j.canlet.2015.11.020

37. He X, Arslan AD, Ho TT, et al. Involvement of polypyrimidine tract-binding protein (PTBP1) in maintaining breast cancer cell growth and malignant properties. Oncogenesis. 2014;3(1):e84. doi:10.1038/oncsis.2013.47

38. Zhou Z, Yin X, Sun H, et al. PTBP1 lactylation promotes glioma stem cell maintenance through PFKFB4-Driven glycolysis. Cancer Res. 2025;85(4):739–757. doi:10.1158/0008-5472.CAN-24-1412

39. Gu J, Zhou J, Chen Q, et al. Tumor metabolite lactate promotes tumorigenesis by modulating MOESIN lactylation and enhancing TGF-β signaling in regulatory T cells. Cell Rep. Cell Reports. 2022;39(12):110986. doi:10.1016/j.celrep.2022.110986

The Community Studio, located on the first and second floors, welcomes guests with an expansive and interactive Auto Library featuring over 2,500 curated books and 500 mobility-related lifestyle and exhibition items. As a collaboration between Hyundai Motor, CCC (Culture Convenience Club), the lifestyle company behind TSUTAYA BOOKSTORE, French luxury publisher ‘Assouline’, and Korean mobility culture company ‘EREVO’, the Auto Library is a must for all those who want a deeper insight into automotive milestones. Visitors can browse through vintage catalogs, admire collectible model cars, read contemporary automotive lifestyle articles, and view racing goods, along with a variety of Hyundai memorabilia.

Rare items such as the Camel Trophy archive—one of only five limited sets worldwide are displayed among other, carefully sourced volumes that narrate the backstories, personal histories, people, and diverse mobility brands shaping modern car culture.

Introduction

Teicoplanin (TEC), a glycopeptide antibiotic, is widely used for treating invasive infections caused by Gram-positive bacteria, including methicillin-resistant Staphylococcus aureus (MRSA), methicillin-sensitive Staphylococcus aureus (MSSA), Streptococcus pneumoniae, and Enterococcus species.¹ Its mechanism of action involves binding to the D-alanyl-D-alanine termini of peptidoglycan precursors, inhibiting bacterial cell wall synthesis.² Due to its broad-spectrum activity and favorable safety profile, TEC is commonly administered for pediatric infections such as bloodstream infections, endocarditis, and osteomyelitis.³

Given its poor oral bioavailability, TEC is typically administered intravenously or intramuscularly in pediatric patients. The drug exhibits a long half-life, enabling less frequent dosing; however, its high protein-binding capacity (90–95%) may influence distribution and elimination.⁴ The weight-normalized clearance (CL) of TEC in children (0.21–1.99 L/h/70 kg) is generally higher than that in adults (0.45–6.33 L/h/70 kg, though mostly concentrated at 0.5–1.0 L/h/70 kg). Additionally, the CL of TEC in children is more significantly influenced by age (eg, differences between infants and older children), body weight, and renal function, whereas in adults, it is mainly associated with renal function. Although TEC exhibits linear pharmacokinetics (PK) within the therapeutic concentration range, pediatric patients face unique PK challenges due to developmental changes in renal/hepatic function, body composition, and drug metabolism.⁵ These factors contribute to significant interindividual variability in drug exposure, necessitating tailored dosing strategies to avoid subtherapeutic concentrations or toxicity.

The clinical efficacy of TEC is closely associated with pharmacokinetics/pharmacodynamics (PK/PD) indices, particularly the ratio of the area under the concentration–time curve to the minimum inhibitory concentration (AUC₂₄/MIC).⁶ The trough concentration (C_min) represents another critical efficacy parameter, with C_min >10 mg/L demonstrating improved clinical outcomes in invasive infections.^7–9 However, achieving optimal PK/PD targets in pediatric populations remains challenging due to limited exposure-response data and the impact of covariates such as age, weight, and renal function.¹⁰

Treatment failure in pediatric patients may result from bacterial resistance, inadequate dosing, or variability in drug exposure.¹¹ Population pharmacokinetic (PPK) modeling provides a robust approach to identify covariates influencing TEC pharmacokinetics and optimize dosing regimens.¹² Despite its clinical significance, no systematic review had comprehensively evaluated PPK studies of TEC in pediatric populations. Therefore, this review aims to: (1) summarize significant predictors of teicoplanin PK parameters; (2) analyze the probability of target attainment (PTA) for primary PK/PD indices across pediatric subgroups using model-based simulations; and (3) identify knowledge gaps requiring further investigation.

Materials and Methods

Study Identifications

A systematic search was conducted for all PPK studies of TEC in the PubMed, Web of Science, and EMBASE databases from their inception until 25 March 2025. The search terms used were: “Teicoplanin”, “Targocid”, “Teichomycin”, “Teichomycin A2.” Additionally, we included terms related to pharmacokinetics such as “population pharmacokinetic*” “pharmacokinetic model*” “nonlinear mixed-effects model”, “NONMEM”, “WINNONMIX”, “P-PHARM”, “MWPHARM”, “nlmixed”, “NLME”, or “MONOLIX.” Finally, we searched for terms related to age groups: “Children”, “Child”, “Neonate*” “Neonatal”, “Infant*” “Newborn”, and “Pediatric*.” The reference lists of all included studies were manually checked to retrieve potentially relevant studies. Detailed search strategies were listed in Supplementary Table S1.

A study was deemed eligible if it met the following criteria: (1) the study population consisted of humans, (2) TEC was the study drug, (3) a population-based analysis was employed, and (4) a nonlinear mixed-effect modeling approach was adopted. A study was excluded if (1) it was a review, conference abstract, or focused on methodology/algorithm/software, (2) it was published in a non-English language, (3) its data overlapped with an article published later, and (4) the information on modeling was insufficient to reproduce.

Moreover, two investigators independently conducted the literature search and study selection process. For discrepancies, consensus was achieved through consultation with a senior researcher. The systematic review methodology adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.¹³

Data Extraction

The following information was extracted from the included articles: (1) the characteristics of the target population (eg age, weight range, and sex); (2) the study design (eg type of study, number of participants and collected samples, sampling design, and TEC formulations); and (3) the information on PPK analyses such as data analysis software, structural models, between-subject variability (BSV), residual unexplained variability (RUV), parameter estimates, covariates, and model evaluation approaches. The study characteristics and PPK analyses were summarized in a tabular format.

Reporting Quality

The reporting quality of the included studies was assessed based on the reporting guidelines for PK studies.^14,15 A checklist containing 35 items was created to summarize the key points of the PK report (Supplementary Table S2). All items were categorized into five sections: title/abstract, background, methods, results, and discussion/conclusion. If an item in the checklist was reported in the study, one point was assigned; otherwise, no point was counted. The compliance rate was used to evaluate the quality of each study, and the calculation formula was as follows (Equation 1):

Compliance rate (%) = sum of items reported / sum of all items × 100% (1)

Study Comparison

Assessment of Visual Predictive Distributions

Monte Carlo simulations of PK profiles were performed to provide visual predictive distributions (VPDs) based on the established model and the study cohort in each included study.¹⁶ According to the classifications of patient characteristics in the retrieved studies,^17–24 we adapted the following 5 groups of virtual populations for Monte Carlo simulation: preterm infants, neonates, infants, children, and adolescents. Differences in each profile were compared by visual inspection because it was assumed that the predictive distribution of the simulated PK profile of TEC sufficiently represented the features of each model and its original data.¹⁶ A total of 1000 virtual patients were simulated for each scenario. All simulations were performed using NONMEM (version 7.4.1; ICON Development Solutions, Ellicott City, MD, U.S.A). R software (version 4.3.1; www.r-project.org, accessed on 16 June 2023) was employed for data exploration and visualization.

Assessment of the Covariates’ Impact

Clearance (CL) is the most critical PK parameter in long-term pharmacotherapy and has been comprehensively investigated in previous studies. Therefore, the covariate effect on CL was explored. The effects of the included covariates on PK parameters were assessed using forest plots. For continuous covariates, the maximum and minimum values based on the demographic information in the included studies were extracted and scaled to the same range. For binary covariates such as sex, 0 and 1 were used.

Based on the range of the identified covariate in each study, the minimum and maximum CL values were calculated. The CL value, normalized to median covariate values in each study, was selected as the reference. Thus, the effect of the identified covariate on CL in each study was displayed by the following equation (Equation 2):

Effect of covariate_j in study_i = The range of calculated CL / The CL reference in study_i × 100% (2)

where study_i means the ith study and covariate_j means the jth identified covariates in study_i.

Moreover, the change in CL within 80%–125%, a threshold widely used to assess bioequivalence,²⁵ was not considered clinically significant in our study. All data were analyzed and plotted using R software (version 4.3.1; www.r-project.org, accessed on 16 June 2023).

Monte Carlo Simulation for the Probability of Target Attainment

AUC₂₄/MIC ≥ 400 and C_min > 10 mg/L are associated with effective bacterial suppression and reduced development of antimicrobial resistance in the treatment of MRSA infections.^19–21,26 Based on the EUCAST standard, the MIC breakpoint for susceptibility against MRSA is defined as 2 mg/L.²⁷ These values are crucial for guiding the rational use of TEC in clinical practice, facilitating the development of appropriate dosing regimens and the accurate assessment of bacterial resistance.

Monte Carlo simulations were conducted to evaluate the rationality of a consistent dosing regimen across five age groups: preterm infants, neonates, infants, children, and adolescents. The dosing protocol included a loading dose of 10 mg/kg every 12 hours for three doses, followed by a maintenance dose of 10 mg/kg once daily. For each age group, 1,000 virtual patients were simulated. Monte Carlo modeling provided the PK profiles used to determine AUC₂₄ on day 4 by trapezoidal rule integration and C_min as trough concentrations.

Separate PTA calculations for AUC₂₄/MIC ≥400 and C_min >10 mg/L were performed across pediatric age groups, with comparative analyses conducted for each target. We set the target that the PTA should be ≥ 90% for determining the optimal dosage for MRSA.

Results

Study Identification

Figure 1 presents the PRISMA flowchart of study selection. A total of 148 records were initially identified through database searches (PubMed: n=39; Embase: n=47; Web of Science: n=62). After duplicate removal (n=64) and automated screening exclusion, 84 records underwent title/abstract screening. Following exclusion of irrelevant studies (n=44), in vitro/animal studies (n=2), non-PPK studies (n=18), reviews (n=2), and conference abstracts (n=1), 17 full-text articles were assessed. Six studies employing adult population models and three utilizing nonparametric approaches were subsequently excluded, yielding eight eligible studies for final inclusion.

Figure 1 PRISMA flow diagram used to identify teicoplanin (TEC) population pharmacokinetics studies.

Reporting Quality

The reporting quality assessment of included studies is summarized in Table 1. The median compliance rate was 84.3% (range: 82.9–94.3%), demonstrating generally satisfactory reporting quality. However, several reporting deficiencies were identified: only 37.5% of studies adequately described co-administration or food-related aspects, and 62.5% failed to report methods for handling missing data. Notably, base model evaluation methods were universally unreported, while only 62.5% and 37.5% of studies provided schematic representations of the final model and summaries of model-building processes, respectively. Despite these limitations, all studies maintained compliance rates exceeding 80%, indicating overall methodological robustness.

Table 1 Reporting Quality of Included Studies

Study Comparison

Study Characteristics

All included studies were single-center investigations, with characteristics detailed in Table 2. Only Yamada et al²³ and Zhang et al²⁰ employed retrospective designs, while the remaining studies^{17–19,21,22,24} were prospective. The studies encompassed diverse pediatric populations: Lukas et al¹⁷ stratified participants by a 12-month age threshold; Zhao et al¹⁸ Zhang et al,²⁰ and Aulin et al²² examined broader age ranges; Gao et al¹⁹ focused on younger subjects (median age: 1.25 years); Kontou et al²¹ specifically studied preterm infants and neonates (age range: 6–17.7 days); Yamada et al²³ differentiated neonates, infants, and children; and Laiseca et al²⁴ categorized patients by continuous kidney replacement therapy status.

Table 2 Characteristics of Included Studies

Quantification methods included high-performance liquid chromatography (HPLC), liquid chromatography-tandem mass spectrometry (LC-MS/MS), ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS), reverse-phase HPLC (RP-HPLC), fluorescence polarization immunoassay (FPIA), quantitative microsphere system (QMS), and latex turbidimetric immunoassay (LTIA). The lower limit of quantification (LLOQ) varied widely, ranging from 25 ng/mL to 3 mg/L.

A total of eight PPK models were identified across the included studies. The majority (7/8)^{17,18,20–24} employed NONMEM software, with one study using NLME.¹⁹ Structural models (7/8) predominantly followed a two-compartment disposition, while one study²⁰ adopted a one-compartment approach. Parameter estimation primarily utilized first-order conditional estimation (FOCE), with or without η-ε interaction. Sampling strategies varied, with two studies^22,24 employing intensive sampling and the remaining^17–21,23 relying on sparse data from therapeutic drug monitoring (TDM). Study sample sizes ranged from 20 to 214 participants, with total TEC observations per study spanning 143–399. Detailed modeling approaches and final PK parameters are summarized in Table 3.

Table 3 Modeling Strategies and Final Parameters of Included Studies

Population Pharmacokinetic Studies

Virtual patients spanning from preterm infants to adolescents were stratified into five demographic groups for VPDs. All simulated cohorts comprised male subjects receiving TEC at steady-state conditions. Population characteristics for five representative groups were detailed in Supplementary Table S3. Monte Carlo simulations generated 1,000 virtual patients per cohort, with results visualized in Figure 2.

Figure 2 Concentration–time profiles of TEC in (A) preterm infants, (B) neonates, (C) infants, (D) children, (E) adolescents in retrieved studies. The solid line represents the median of the simulated concentration–time profile, and the light shadows represent the 10th–90th percentiles of those profiles. All patients were assumed to be male and received TEC at a 10 mg/kg weight-normalized dosage.

At therapeutic doses, adolescents exhibited lower median weight-normalized CL values compared to other populations, while neonates showed higher median CL values than infants and children. The estimated median CL of TEC varied by age group: 0.0131 L/h/kg for preterm infants (1.5 kg, 10 days), 0.0150 L/h/kg (range: 0.0114–0.0185) for neonates (3.5 kg, 10 days), 0.0136 L/h/kg (range: 0.0110–0.0488) for infants (7 kg, 0.5 years), 0.0138 L/h/kg (range: 0.00946–0.0370) for children (21 kg, 6 years), and 0.0116 L/h/kg (range: 0.00855–0.0285) for adolescents (60 kg, 14 years). Additionally, no significant population differences were found in the median volume of distribution (V) values of TEC. The median values of V per kilogram (L/kg) were similar for neonates, infants, and children, though higher than those for preterm infants. Specifically, the median values were: 0.159 L/kg for preterm infants, 0.208 L/kg (range: 0.159–0.257) for neonates, 0.248 L/kg (range: 0.0871–0.476) for infants, 0.186 L/kg (range: 0.0871–0.476) for children, and 0.171 L/kg (range: 0.0871–0.476) for adolescents.

The steady-state AUC₂₄ at day 4 (72–96 h post-dose) across age groups is presented in Figure 3. Preterm infants demonstrated the highest median AUC₂₄ (740 mg·h/L), exceeding the 540–690 mg·h/L range observed in neonates. Infants showed substantial between-study variability (median AUC₂₄ range: 206–715 mg·h/L), while children exhibited similar heterogeneity (273–835 mg·h/L). Adolescents displayed the most pronounced variability (350–1110 mg·h/L).

Figure 3 TEC main pharmacokinetic parameter (AUC) at steady state for various typical populations. All simulated patients were assumed to be male. They received TEC with an initial regimen of 10 mg/kg administered every 12 hours for three doses, followed by a maintenance dose of 10 mg/kg once every 24 hours. Notes: The text in different colors in the figure corresponds to each population pharmacokinetic (PPK) study included in this research, with each color uniquely identifying one of the eight studies.

All tested covariates that affected CL, and the distribution volume of the central compartment (V_c) are summarized in Table 4. The stepwise covariate screening included forward inclusion and backward elimination. A comparison of identified and investigated covariates is presented in Figure 4.

Table 4 List of Tested and Significant Covariates in the Models

Figure 4 A histogram of the amount of investigated and identified covariates in included studies.

All included PPK studies of TEC aimed to identify potential covariates explaining between-subject variability (BSV) in drug exposure. The most frequently screened covariates were: weight (investigated in 7 studies), age (6 studies), serum creatinine (Scr; 6 studies), estimated glomerular filtration rate (eGFR; 4 studies), creatinine clearance (CLcr; 3 studies), albumin (ALB; 3 studies), body surface area (BSA; 3 studies), and sex (3 studies). Among these, weight was consistently identified as a significant covariate in all 7 studies that examined it. Other covariates showing statistical significance in final models included: CLcr (2/3 studies), eGFR (2/4 studies), Scr (2/6 studies), postmenstrual age (PMA; 1/1 study), postnatal age (PNA; 1/2 studies), filter surface area (FSA; 1/1 study), age (1/6 studies), ALB (1/3 studies), and continuous kidney replacement therapy (CKRT; 1/1 study). Notably, despite being frequently investigated, covariates such as sex and BSA were not retained in any final models, possibly due to collinearity with stronger predictors or limited sample sizes in individual studies.

The effect of identified covariates on CL was presented in the forest plot (Figure 5). Seven of the 8 studies indicated that body weight was associated with the CL of TEC.^18–24 In contrast, one study¹⁷ found no relationship between body weight and CL, but it did find an association between weight and CL when conducting population-wide modeling. Weight significantly affected CL changes across all age groups. One study²³ revealed that PMA significantly influenced the CL of TEC in a threshold-dependent manner: for patients with PMA ≥48 weeks, eGFR was the primary covariate affecting CL; whereas in those with PMA <48 weeks, CL was determined by PMA itself, PNA, Scr, and ALB. Moreover, the eGFR could explain the BSV of CL in two studies,^19,23 all of which showed a clinically significant influence on CL, with absolute changes exceeding 20%.

Figure 5 Covariate effect on the clearance (CL) of TEC. The horizontal bars represent the effect of covariates on the CL according to each study. The typical CL in each study was considered to be 1. The x-axis refers to the effect of each identified covariate on CL, expressed as the ratio of the value of CL in the range of each covariate to the typical value of CL. The shaded area ranges from 0.8 to 1.25. N, no; Y, yes.

Two studies^18,21 reported CLcr significantly influenced TEC clearance. One study¹⁸ demonstrated that when CLcr ranged from 25 to 464.1 mL/min, CL varied by 0.30-to 1.78-fold relative to baseline. The other study,²¹ for the same range of CLcr values, showed a much larger variation, from 0.99 to 7.04 times. Two independent studies^20,23 identified Scr as a significant CL determinant. ALB showed particular clinical relevance for CL in neonates and infants with PMA < 48 weeks.²³ One study²⁴ revealed CKRT reduced the CL of TEC by approximately 30% compared to non-CKRT patients. Notably, filter surface area significantly affected CL in CKRT patients: medium and large filters increased CL by 2.58-fold and 4.04-fold, respectively.

All included studies characterized between-subject variability (BSV) using exponential models. The median BSV was 33.1% (range: 14.1%–65.9%; n=8) for CL and 45.7% (range: 22.2%–105.4%; n=7) for V. Residual unexplained variability (RUV) was predominantly modeled using proportional error models, demonstrating a range of 4.15–34% (n=4). Notably, Zhang et al²⁰ reported significantly higher BSV for CL (65.9%) in pediatric populations compared to other studies.

Internal evaluations were conducted for all pharmacokinetic (PK) models. Goodness-of-fit (GOF) plots, demonstrating predicted versus observed value correlations, were the most frequently employed validation method. All studies^17–24 incorporated model-based simulations to characterize the PK profiles of TEC. Furthermore, six studies^{17–20,23,24} extended simulations to quantify covariate effects and derive optimized dosing regimens.

Analysis of Probability of Target Attainment

The simulations in the included studies were based on two PK/PD targets: AUC₂₄/MIC and C_min. PTA simulations focused solely on MRSA, with each PPK model evaluated using strain-specific MIC data (Figure 6). This study found that MIC and C_min values significantly affected drug efficacy in different pediatric populations. When the MIC was 2 mg/L, the PTA of all populations did not exceed 90%, indicating that existing standard treatment regimens were ineffective and medication strategies needed adjustment. When the C_min was 10 mg/L, there were differences in attainment rates among populations. The attainment rates for preterm infants, neonates, infants, children, and adolescents were 100% (with a sample size of 1 study), 50% (from 2 studies), 42.86% (from 7 studies), 28.57% (from 7 studies), and 33.33% (from 3 studies), respectively. The PTA results suggested that for pediatric patients, drug dosages should be determined by combining the target C_min and MIC values.

Figure 6 The probability target achievement of TEC in included studies. The gray dashed lines represent a 90% target attainment rate, while the green vertically oriented dashed lines correspond to the respective clinical breakpoints for MIC specific to MRSA and Cmin of TEC.

Discussion

Teicoplanin, a glycopeptide antibiotic, is widely used for methicillin-resistant Staphylococcus aureus (MRSA) infections. Although numerous PK studies on TEC existed, few had systematically evaluated the physiological and pathological factors contributing to its exposure variability. To our knowledge, this study represented the first systematic evaluation of PPK models for TEC, specifically addressing the need for dose adjustment in pediatric populations.

The PK of TEC exhibited significant differences among pediatric patients at different age stages. The median CL values for children were similar to those of preterm infants and infants. However, the median CL in neonates was significantly higher than in adolescents. The difference in CL between neonates and adolescents could be explained by several key factors. Studies^28–30 had shown that neonates exhibited nearly twice the glomerular filtration rate per unit body weight compared to adolescents. This was accompanied by significantly increased expression of renal tubular transporters, including organic anion transporters (OATs) and organic anion transporting polypeptides (OATPs), which markedly enhanced drug excretion. Moreover, neonates have only about 50% of the α1-acid glycoprotein (AAG) levels observed in adolescents. This lower AAG concentration leads to higher free drug fractions, which are more readily cleared by the kidneys.^31–33 Of equal importance, neonates exhibit a larger volume of distribution and higher extracellular fluid ratio, which enhances initial drug clearance. In contrast, adolescents show lower clearance as metabolic enzymes mature toward hepatic dominance.^34–36

The elimination of TEC was influenced by the maturity of renal function and body composition. Studies^23,36 show that CL increased significantly with PMA growth. In neonates (PMA: 23.7–42.4 weeks), the CL increased 4.7-fold, rising from 0.00473 to 0.0223 L/h. In infants (PMA: 30.6–136.4 weeks), the CL escalated 5.4-fold, from 0.0299 to 0.161 L/h. While in children (PMA: 134–878 weeks), it stabilized at a level ranging from 0.520 to 0.536 L/h. This PMA-related pattern correlated with glomerular filtration rate and muscle mass development.^37,38 Thus, critically ill neonates with lower PMA require dosage adjustment to avoid accumulation, while infants nearing 1 year may need higher weight-normalized doses to maintain therapeutic levels.

The study by Yamada et al²³ demonstrated that PNA significantly affected the CL of TEC in pediatrics with PMA < 48 weeks. During the first 27 days of life, the CL of TEC remained at a low level of 0.0126–0.0141 L/h, only 1.3–1.4% of the reference value (0.990 L/h), confirming immature renal elimination pathways in the first month of life. In contrast, during infancy (36.5–657 days), CL increased markedly to 0.134 L/h, representing 13.5% of the values. This 10.6-fold rise from neonatal levels reflected rapid renal maturation in this developmental stage. These results indicated that PNA was a critical determinant of TEC dosing, particularly during the transition from neonates to infants, where clearance capacity undergoes a dramatic surge.²¹ Consequently, dose adjustments were essential for neonates due to their substantially lower clearance capacity. The nonlinear relationship between CL and PNA necessitates age-stratified dosing regimens for this special population.

The CL of TEC was influenced by ALB levels, particularly in neonates and infants.³⁹ Yamada et al²³ demonstrated that higher ALB values were associated with reduced the CL of TEC in these populations. In neonates with ALB levels ranging from 1.1 to 5.5 g/dL, the CL decreased from 0.0370 to 0.00650 L/h, representing an approximately 5.7-fold reduction. Similarly, in infants (ALB: 2–5.1 g/dL), the CL declined from 0.245 to 0.0892 L/h, with a 2.8-fold difference. This inverse correlation suggested that albumin binding may limit the free fraction of TEC, thereby reducing its clearance.³⁹ Consequently, dose adjustment may be required for neonates and infants with elevated ALB levels to ensure the attainment of effective therapeutic concentrations.

The CL of TEC was significantly influenced by renal function parameters. In pediatric populations, both eGFR and CLcr had been independently incorporated into PPK models as key covariates affecting TEC elimination. Studies^18,19,21,23 demonstrated that the developmental stage of renal function substantially impacted TEC clearance. Results^19,23 indicated that variations in eGFR could lead to a 6.3-fold fluctuation in TEC clearance, ranging from 0.151 to 0.949 times the reference value. In comparison, CLcr variations exerted a more pronounced effect on clearance, showing up to a 7.1-fold difference, from 98.8% to 703.9% of the reference value.^18,21 This discrepancy mainly arose because CLcr more directly reflected glomerular filtration function and was independent of muscle metabolism interference, thus showing greater sensitivity to changes in renal function.⁴⁰

Multiple studies^20,23 had demonstrated a significant inverse correlation between Scr levels and TEC clearance. Specifically, when neonatal Scr concentrations increased from 17.68 to 159.12 μmol/L, TEC clearance decreased from 0.0250 to 0.00792 L/h, representing a 3.1-fold reduction. Notably, in infant populations, as Scr rose from 8.84 to 150.3 μmol/L, TEC clearance showed a more pronounced decline from 0.357 to 0.0491 L/h, reflecting a 7.3-fold difference. This Scr-dependent clearance pattern was particularly pronounced in early infancy, showing a close correlation with the rapid developmental changes in renal function that were characteristic of this period.⁴¹

One study²⁴ reported that during continuous kidney replacement therapy (CKRT), the 30% reduction in TEC clearance was mainly attributed to the drug being adsorbed onto the dialysis membrane and the convective loss of the unbound fraction during ultrafiltration.⁴² Notably, the study further revealed that the surface area of the filter had a significant impact on the CL of TEC in patients undergoing CKRT. Medium and large filters could notably enhance the CL of TEC by 2.58-fold and 4.04-fold, respectively. This effect was likely attributed to their larger surface areas, which facilitated more efficient drug clearance.⁴³

In this study, the notable influence of MIC and C_min on drug efficacy in various pediatric groups stemmed from multiple factors. Literature indicated that preterm infants and neonates had immature metabolism and clearance mechanisms, which resulted in unique PK profiles.⁴⁴ Their reduced drug-processing ability leads to diverse drug concentrations, affecting the achievement of PK/PD targets like C_min. When C_min was set at 10 mg/L, the significant variation in probability of target attainment (PTA) across pediatric age groups revealed the key role of age-related physiology.⁴⁵ Preterm infants, with less-developed drug-elimination pathways, more easily reached the C_min target in the single included study. In contrast, older children and adolescents have more active metabolisms, leading to lower attainment rates. When MIC was 2 mg/L, all age groups failed to achieve a PTA above 90%, which suggested that current standard treatments may be inadequate. The reduced efficacy across all pediatric groups likely stemmed from elevated MRSA resistance (MIC = 2 mg/L) to TEC, which compromised its growth-inhibitory capacity.⁴⁶

In summary, these results highlighted the need for personalized dosing in pediatric patients. Considering both target C_min and AUC₂₄/MIC, healthcare providers could optimize drug treatment, factoring in the distinct physiology and resistance patterns of different pediatric age groups.

The establishment of the 90% PTA threshold as a standard is based on several key considerations. First, populations such as children and patients with renal impairment exhibit significant pharmacokinetic variability. Monte Carlo simulations quantify this variability to assess the robustness of dosing regimens, with the >90% threshold ensuring coverage of the vast majority of patients and reducing treatment failures caused by individual differences. Second, although therapeutic drug monitoring (TDM) is the gold standard, its immediate implementation in medical institutions remains limited. A dosing regimen validated by Monte Carlo simulations with >90% PTA can provide the optimal choice for initial treatment before TDM results are available; if PTA is below 90%, it indicates the need for dosage adjustment or early initiation of TDM. Finally, relevant studies have also set PTA at ≥90%, demonstrating that this standard ensures most pediatric patients achieve drug exposures associated with effective bacterial suppression, thereby indirectly reducing the risk of drug resistance.^47,48

It is important to note that despite the consistent core findings across most included studies, certain conclusions regarding TEC pharmacokinetic covariates remain inconsistent. For instance, discrepancies exist regarding the impact of ALB on TEC clearance: Yamada et al²³ reported a 5.7-fold reduction in CL with increasing ALB in neonates, whereas Aulin et al²² did not identify ALB as a significant covariate—potentially attributed to differences in study populations (neonates vs broader pediatric cohorts) and ALB measurement ranges (1.1–5.5 g/dL vs 2–5.1 g/dL). Similarly, variations in the effect of PNA on CL were noted between Kontou et al²¹ (no significant association) and Yamada et al²³ (10.6-fold CL increase in infancy vs neonates), likely driven by differences in PNA stratification (6–17.7 days vs 1–657 days) and concurrent adjustment for PMA in the latter.

To date, PK/PD studies on TEC remain limited, particularly those focusing on pediatric patients. A critical gap exists in clarifying the quantitative relationship between pharmacodynamic indicators and adverse effects. Therefore, larger prospective PK/PD studies in pediatric populations are essential to better characterize the associations among dose, drug exposure, and therapeutic response.

This review is subject to a number of limitations. First, we only included English-language literature, which may have led to the omission of potentially valuable studies from non-English-speaking regions. Second, we did not incorporate nonparametric PPK models, a choice that may result in an incomplete interpretation of certain analytical approaches. Third, the absence of external validation using an independent dataset, which may limit the generalizability of the models to other clinical settings. Notably, plasma exposure is a surrogate marker for the infection site. This discrepancy may cause deviations between the predicted PTA and actual clinical outcomes. Treatment failure cannot be assumed based solely on PTA targets, which are derived from combining drug concentrations in the central compartment and microbiological MICs over time.

Conclusion

This systematic review of published PPK studies on TEC highlights significant PK variability of TEC across pediatric populations, with neonates exhibiting a higher weight-normalized CL than adolescents. Methodologically, this review incorporates a distinct innovative approach through the integrated application of VPD and Monte Carlo Simulations (MCS). This combined framework enables systematic comparison of PK parameters across pediatric age cohorts and rigorous evaluation of the PTA, thereby furnishing a robust technical underpinning for the key findings presented herein. The key covariates influencing TEC pharmacokinetics include body weight, PMA, renal function parameters (eGFR, CLcr, Scr), ALB levels, and CKRT status. Furthermore, dosage regimen adjustments should be considered based on pediatric age subgroups and pathogen-specific MICs, particularly for infections caused by MRSA with an MIC of 2 mg/L. Notably, this analysis establishes a practical framework for optimizing future clinical dosing regimens of TEC. For example, preterm infants may require lower maintenance doses to prevent excessive drug exposure, whereas neonates and children may need higher weight-normalized doses to achieve therapeutic targets. Additionally, further large-scale prospective population studies are essential to clarify the dose-exposure-response relationship of TEC in pediatric populations. Nevertheless, significant external validation will be required before clinical implementation.

Acknowledgments

The authors would like to thank Xiping Li from Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan for his critical comments and suggestions regarding the data visualization.

Funding

This research was financially supported by the Foundation of Health Commission of Anhui Province (Grant No. AHWJ2023BAa20159).

Disclosure

The authors declare no conflicts of interest in this work.

References

1. Behring AH, Mariani C, Gerbi M, et al. Appropriateness of loading dose and TDM in patients treated with teicoplanin: a single-centre observational retrospective cohort study comparing single-daily and thrice-weekly regimens. J Antimicrob Chemother. 2025;80(6):1535–1542. doi:10.1093/jac/dkaf097

2. Beltrametti F, Consolandi A, Carrano L, et al. Resistance to glycopeptide antibiotics in the teicoplanin producer is mediated by van gene homologue expression directing the synthesis of a modified cell wall peptidoglycan. Antimicrob Agents Chemother. 2007;51(4):1135–1141. doi:10.1128/aac.01071-06

3. David MZ, Daum RS. Treatment of staphylococcus aureus infections. Curr Top Microbiol Immunol. 2017;2017:325–383.

4. Lode H, Boeckh M, Höffken G, et al. Comparative pharmacokinetics of glycopeptide antibiotics, and the influence of teicoplanin on granulocyte function. Scand J Infect Dis Suppl. 1990;72:9–13.

5. Mouton JWA, De Clercq A, De Paepe P, et al. Pharmacokinetics and target attainment of teicoplanin: a systematic review. Clin Pharmacokinet. 2025;64:467–509. doi:10.1007/s40262-025-01483-7

6. Matsumoto K, Watanabe E, Kanazawa N, et al. Pharmacokinetic/pharmacodynamic analysis of teicoplanin in patients with MRSA infections. Clin Pharmacol. 2016;8:15–18. doi:10.2147/cpaa.S96143

7. Sato Y, Hiramatsu K, Suzuki Y, et al. Optimal trough concentration of teicoplanin in febrile neutropenic patients with hematological malignancy. Chemotherapy. 2018;63(1):29–34. doi:10.1159/000481725

8. Hanai Y, Takahashi Y, Niwa T, et al. Optimal trough concentration of teicoplanin for the treatment of methicillin-resistant Staphylococcus aureus infection: a systematic review and meta-analysis. J Clin Pharm Ther. 2021;46(3):622–632. doi:10.1111/jcpt.13366

9. Hanai Y, Takahashi Y, Niwa T, et al. Clinical practice guidelines for therapeutic drug monitoring of teicoplanin: a consensus review by the Japanese society of chemotherapy and the Japanese society of therapeutic drug monitoring. J Antimicrob Chemother. 2022;77(4):869–879. doi:10.1093/jac/dkab499

10. Tsuji Y. Hospital pharmacometrics for optimal individual administration of antimicrobial agents for anti-methicillin-resistant staphylococcus aureus infected patients. Biol Pharm Bull. 2021;44(9):1174–1183. doi:10.1248/bpb.b21-00002

11. Merino-Bohórquez V, Docobo-Pérez F, Sojo J, et al. Population pharmacokinetics and pharmacodynamics of fosfomycin in non-critically ill patients with bacteremic urinary infection caused by multidrug-resistant Escherichia coli. Clin Microbiol Infect. 2018;24(11):1177–1183. doi:10.1016/j.cmi.2018.02.005

12. Chen CY, Xie M, Gong J, et al. Population pharmacokinetic analysis and dosing regimen optimization of teicoplanin in critically ill patients with sepsis. Front Pharmacol. 2023;14:1132367. doi:10.3389/fphar.2023.1132367

13. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021:372:n71. doi:10.1136/bmj.n71

14. Kanji S, Hayes M, Ling A, et al. Reporting guidelines for clinical pharmacokinetic studies: the ClinPK statement. Clin Pharmacokinet. 2015;54(7):783–795. doi:10.1007/s40262-015-0236-8

15. Jamsen KM, McLeay SC, Barras MA, Green B. Reporting a population pharmacokinetic-pharmacodynamic study: a journal’s perspective. Clin Pharmacokinet. 2014;53(2):111–122. doi:10.1007/s40262-013-0114-1

16. Duffull SB, Wright DF. What do we learn from repeated population analyses? Br J Clin Pharmacol. 2015;79(1):40–47. doi:10.1111/bcp.12233

17. Lukas JC, Karikas G, Gazouli M, Kalabalikis P, Hatzis T, Macheras P. Pharmacokinetics of teicoplanin in an ICU population of children and infants. Pharm Res. 2004;21(11):2064–2071. doi:10.1023/b:pham.0000048198.56873.d8

18. Zhao W, Zhang D, Storme T, Baruchel A, Declèves X, Jacqz-Aigrain E. Population pharmacokinetics and dosing optimization of teicoplanin in children with malignant haematological disease. Br J Clin Pharmacol. 2015;80(5):1197–1207. doi:10.1111/bcp.12710

19. Gao L, Xu H, Ye Q, et al. Population pharmacokinetics and dosage optimization of teicoplanin in children with different renal Functions. Front Pharmacol. 2020;11:552. doi:10.3389/fphar.2020.00552

20. Zhang T, Sun D, Shu Z, et al. Population pharmacokinetics and model-based dosing optimization of teicoplanin in pediatric patients. Front Pharmacol. 2020;11:594562. doi:10.3389/fphar.2020.594562

21. Kontou A, Sarafidis K, Begou O, et al. Population pharmacokinetics of teicoplanin in preterm and term neonates: is it time for a new dosing regimen? Antimicrob Agents Chemother. 2020;64(4). doi:10.1128/aac.01971-19

22. Aulin LBS, De Paepe P, Dhont E, et al. Population Pharmacokinetics of unbound and total teicoplanin in critically Ill pediatric patients. Clin Pharmacokinet. 2021;60(3):353–363. doi:10.1007/s40262-020-00945-4

23. Yamada T, Emoto C, Fukuda T, et al. Optimal teicoplanin dosing regimen in neonates and children developed by leveraging real-world clinical information. Ther Drug Monit. 2022;44(3):404–413. doi:10.1097/ftd.0000000000000930

24. Butragueño-Laiseca L, García-Orueta G, Riva N, et al. Population pharmacokinetic analysis of teicoplanin in paediatric patients, including those receiving continuous kidney replacement therapy: a prospective cohort study. J Antimicrob Chemother. 2025;80(3):868–875. doi:10.1093/jac/dkaf012

25. Christians U, Klawitter J, Clavijo CF. Bioequivalence testing of immunosuppressants: concepts and misconceptions. Kidney Int Suppl. 2010;(115):S1–7. doi:10.1038/ki.2009.504

26. Pea F, Brollo L, Viale P, Pavan F, Furlanut M. Teicoplanin therapeutic drug monitoring in critically ill patients: a retrospective study emphasizing the importance of a loading dose. J Antimicrob Chemother. 2003;51(4):971–975. doi:10.1093/jac/dkg147

27. Chang HJ, Hsu PC, Yang CC, et al. Influence of teicoplanin MICs on treatment outcomes among patients with teicoplanin-treated methicillin-resistant staphylococcus aureus bacteraemia: a hospital-based retrospective study. J Antimicrob Chemother. 2012;67(3):736–741. doi:10.1093/jac/dkr531

28. Staub E, Bolisetty S, Allegaert K, Raaijmakers A. neonatal kidney function, injury and drug dosing: a contemporary review. Children. 2025;12(3). doi:10.3390/children12030339

29. Smeets NJL, IntHout J, van der Burgh MJP, Schwartz GJ, Schreuder MF, de Wildt SN. Maturation of GFR in term-born neonates: an individual participant data meta-analysis. J Am Soc Nephrol. 2022;33(7):1277–1292. doi:10.1681/asn.2021101326

30. Wu Y, Allegaert K, Flint RB, et al. Prediction of glomerular filtration rate maturation across preterm and term neonates and young infants using inulin as marker. Aaps j. 2022;24(2):38. doi:10.1208/s12248-022-00688-z

31. Thompson LE, Joy MS. Endogenous markers of kidney function and renal drug clearance processes of filtration, secretion, and reabsorption. Curr Opin Toxicol. 2022;31. doi:10.1016/j.cotox.2022.03.005

32. Ligi I, Boubred F, Grandvuillemin I, Simeoni U. The neonatal kidney: implications for drug metabolism and elimination. Curr Drug Metab. 2013;14(2):174–177.

33. Lu H, Rosenbaum S. Developmental pharmacokinetics in pediatric populations. J Pediatr Pharmacol Ther. 2014;19(4):262–276. doi:10.5863/1551-6776-19.4.262

34. Mahmood I. Prediction of total and renal clearance of renally secreted drugs in neonates and infants (≤3 months of age). J Clin Trans Res. 2022;8(6):445–452.

35. Alcorn J, McNamara PJ. Ontogeny of hepatic and renal systemic clearance pathways in infants: part I. Clin Pharmacokinet. 2002;41(12):959–998. doi:10.2165/00003088-200241120-00003

36. Rhodin MM, Anderson BJ, Peters AM, et al. Human renal function maturation: a quantitative description using weight and postmenstrual age. Pediatr Nephrol. 2009;24(1):67–76. doi:10.1007/s00467-008-0997-5

37. Mahmood I, Staschen CM. Prediction of human glomerular filtration rate from preterm neonates to adults: evaluation of predictive performance of several empirical models. Aaps j. 2016;18(2):445–454. doi:10.1208/s12248-016-9868-3

38. Groothof D, Post A, Polinder-Bos HA, et al. Muscle mass and estimates of renal function: a longitudinal cohort study. J Cachexia Sarcopenia Muscle. 2022;13(4):2031–2043. doi:10.1002/jcsm.12969

39. Yano R, Nakamura T, Tsukamoto H, et al. Variability in teicoplanin protein binding and its prediction using serum albumin concentrations. Ther Drug Monit. 2007;29(4):399–403. doi:10.1097/FTD.0b013e3180690755

40. Khader NA, Kamath VG, Kamath SU, Rao IR, Prabhu AR. Kidney function estimation equations: a narrative review. Ir J Med Sci. 2025;194(2):725–743. doi:10.1007/s11845-025-03874-y

41. Starr MC, Griffin RL, Harer MW, et al. Acute kidney injury defined by fluid-corrected creatinine in premature neonates: a secondary analysis of the PENUT randomized clinical trial. JAMA Network Open. 2023;6(8):e2328182. doi:10.1001/jamanetworkopen.2023.28182

42. Yoshimoto K, Matsuura R, Komaru Y, et al. Solute clearance evaluation and filter clotting prediction in continuous renal replacement therapy. J Clin Med. 2023;12(24). doi:10.3390/jcm12247703

43. Tsujimoto Y, Fujii T. How to prolong filter life during continuous renal replacement therapy? Crit Care. 2022;26(1):62. doi:10.1186/s13054-022-03910-8

44. Bardol M, Norman E, Lagercrantz H, Fellman V, Standing JF. Fentanyl dosage for preterm infants suggested by a pharmacokinetic, -dynamic, and -genetic model. Pediatr Res. 2025;97(1):239–245. doi:10.1038/s41390-024-03404-z

45. Strenger V, Hofer N, Rödl S, et al. Age- and gender-related differences in teicoplanin levels in paediatric patients. J Antimicrob Chemother. 2013;68(10):2318–2323. doi:10.1093/jac/dkt176

46. Mandell JB, Gish C, Cappellini AJ, et al. Methicillin resistant staphylococcus aureus mazEF expression promotes infections by influencing cellular growth, antibiotic sensitivity, and formation of biofilms. Sci Rep. 2024;14(1):21269. doi:10.1038/s41598-024-70829-1

47. Li XP, Liu XQ, Mao JH, Liu D, Jiao Z. Evaluation of population pharmacokinetic models of micafungin: implications for dosing regimen optimization in critically Ill patients. Pharmaceutics. 2024;16(9):1145. doi:10.3390/pharmaceutics16091145

48. Zhou YZ, Huang C, Ji JR, et al. Optimization of dosing regimens of vancomycin and teicoplanin in treatment of methicillin-resistant staphylococcal bloodstream infection based on monte carlo simulation. Herald Med. 2022;41(7):1026–1032. doi:10.3870/j.issn.1004-0781.2022.07.017

This article first appeared on GuruFocus.

Release Date: October 24, 2025

For the complete transcript of the earnings call, please refer to the full earnings call transcript.

• Consolidated net sales increased by 4.8% to 149.3 billion pesos for the first nine months of 2025.

• Gross profit rose by 6.2% to 36.4 billion pesos, outpacing revenue growth.

• The food segment saw a sales increase of 4.5% in the third quarter, driven by same-store sales growth and new store contributions.

• Drugstores delivered strong performance with double-digit sales growth in the third quarter.

• Robinsons Retail Holdings Inc (PHS:RRHI) was recognized as one of the world’s best companies of 2025 by Forbes magazine.

• Net income attributable to the parent decreased by 60% to 3.1 billion pesos due to the absence of a one-time gain from the previous year.

• Department store sales declined by 11.7% in the third quarter due to a shift in the school year and increased competition.

• EBITDA for the DIY segment declined due to higher operating expenses on rent and process improvements.

• Higher interest expenses and equity losses from wholesale operations negatively impacted net income.

• The department store segment faced stiff competition from online marketplaces, affecting sales.

Q: Can you give us an update on the remaining balance of the debt used for the acquisition of the DPI shares? Also, what is the interest expense related to this debt? A: The outstanding balance on the DPI-related acquisition loan is 10.8 billion pesos, unchanged from June 2025. The interest expense is 500 million pesos. (Respondent: Unidentified_1)

Q: What is the SSSG in the supermarket and CVS segments in the third quarter of 2025, respectively? How are basket size and transaction trends in both segments? A: For supermarkets, SSSG is about 3%, and for Uncle John’s, it’s 1%. Basket sizes for supermarkets were up about 7 to 8% versus last year, and for Uncle John’s, they were up by about 1%. (Respondent: Unidentified_1)

Q: What is the latest update on the approvals for the premium bikes acquisition? When are you expecting it to close? A: The acquisition is still under review by the Philippine Competition Commission, and we expect it to close this year. (Respondent: Unidentified_1)

Q: How have the different segments performed so far for the month of October? Are we seeing sales momentum pick up for discretionary items? A: As of mid-October, food SSSG is holding up well, and the drugstore business is stable. However, we are still facing challenges in the discretionary items segment. (Respondent: Unidentified_1)

Q: What is the outlook for 2025 in terms of top-line SSSG and margins per segment? A: We are targeting a blended SSSG of 2 to 4% and a gross margin expansion of up to 30 basis points. The food segment is expected to be the main driver with a 3 to 5% SSSG and similar margin expansion. (Respondent: Unidentified_1)

Q: Can you comment on the overall demand scenarios across your various business formats? Any trends in consumption that you can share? A: Demand is healthy for our food and drugstore businesses, with basket sizes increasing. We are focusing on increasing the mix of privately branded items and imported products to drive margins. (Respondent: Unidentified_1)

Q: What led to the 6% year-on-year decline in royalty and other revenues in 2025? A: This decline is primarily due to timing differences in revenue recognition. We can provide more details later. (Respondent: Unidentified_1)

Q: How is SSSG trending so far in October, and what is your sense of consumer behavior? A: SSSG is trending slightly above the midpoint of our 2 to 4% blended guidance. Consumer behavior shows no significant trading down, with basket sizes increasing, indicating positive consumer sentiment. (Respondent: Unidentified_1)

For the complete transcript of the earnings call, please refer to the full earnings call transcript.

By Rhiannon Hoyle

  A Rio Tinto-controlled company has asked law enforcement to help with an internal investigation into allegations of corruption and unethical conduct at the giant Oyu Tolgoi copper operation in Mongolia. 

  The probe is the latest twist in a saga to develop and mine one of the world's biggest copper and gold deposits, found in the southern Gobi Desert, about 50 miles north of the border with China. 

  "We are aware of allegations involving procurement-related activities," Oyu Tolgoi LLC said in a statement. The company is "conducting a comprehensive internal investigation" and has "sought cooperation of law enforcement authorities." 

  It said it couldn't comment further given the probe is ongoing. 

  Oyu Tolgoi is 66% owned by Rio Tinto, while the government of Mongolia owns the rest. Rio Tinto manages the operation, which it expects to become the world's fourth-largest copper mine by 2030. 

  The operation is at the heart of Rio Tinto's plans to grow and diversify its portfolio away from steel ingredient iron ore, which it currently relies on for the bulk of its earnings. Increasing copper output is a priority for many of the world's biggest miners given the metal is used heavily to build electric cars, renewable energy and data centers. 

  Rio Tinto, the world's second-biggest miner by market value, has invested billions of dollars in an underground expansion of the Oyu Tolgoi mine, where it expects production will increase by more than 50% this year. More than 80% of Oyu Tolgoi's total value lies deep underground, according to the miner. 

  The development of Oyu Tolgoi, established as an open-pit mine in 2011, has been beset by delays, cost overruns and complicated negotiations with Mongolia's government, including a multiyear dispute over taxes. 

  In June, Rio Tinto agreed to pay $138.75 million to resolve a U.S. class-action lawsuit that alleged the company concealed problems during the expansion underground. The company agreed to the settlement to avoid the uncertainty of continued litigation, according to the court documents, and denied all allegations of wrongdoing. 

  Earlier this year, Oyu Tolgoi said it had faced "false and defamatory allegations" related to its procurement processes via some media and online platforms. In a statement published to its website in February, the company described its procurement processes as transparent and fully compliant with Mongolian laws and regulations. 

  "No single individual has unilateral decision-making authority over procurement and our operations are subject to [regular] audits by both national and international bodies," Oyu Tolgoi said in that statement. 

  Write to Rhiannon Hoyle at rhiannon.hoyle@wsj.com 

  (END) Dow Jones Newswires

  October 27, 2025 00:39 ET (04:39 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.

Introduction

In the past few decades, the global age-specific mortality rate of chronic kidney disease(CKD) has continued to rise, affecting approximately 9.1% of the global population and imposing a heavy burden on patients and healthcare.^1,2 CKD has always been an established risk of cardiovascular disease, and timely diagnosis can help prevent adverse outcomes such as end-stage renal disease and related cardiovascular diseases.^3,4 Specifically, in various stages of CKD, the extracellular matrix undergoes renal interstitial fibrosis and tubular atrophy (IF/TA), which is negatively correlated with renal function.^5–7 Therefore, timely assessment of CKD progression through IF/TA testing is necessary. Up to now, glomerular filtration rate is commonly used to evaluate renal function in CKD patients, but it cannot accurately reflect renal function and is easily influenced by the population. In addition, although percutaneous renal biopsy is the “gold standard” for diagnosing and understanding renal fibrosis changes, it is not suitable for long-term monitoring of disease progression and evaluation of treatment effectiveness.⁸ Given this, it is extremely urgent to non invasively and accurately detect and monitor the degree of TA/IF in the kidneys of CKD patients to assist clinicians in evaluating the progression of fibrosis caused by CKD.

Ultrasound imaging is a commonly used diagnostic method for evaluating CKD in clinical diagnosis and treatment.^9–11 Although there may be texture information in ultrasound images that cannot be observed with the naked eye, it can be obtained through radiomics. In recent years, radiomics has used computer technology to extract texture features and high-dimensional image features from ultrasound images to quantify image information, which are correlated with biological and pathological information.^12,13 At present, radiomics in the field of kidney has found that wavelet transform based features in ultrasound radiomics are of great significance for distinguishing CKD kidneys from healthy kidneys, and ultrasound radiomics can be used to evaluate the degree of renal function damage in CKD.^14–16 However, the feature extraction of ultrasound imaging omics and optimization of advanced machine learning(ML) algorithms are still ongoing, providing alternative diagnostic strategies for non-invasive diagnosis of renal fibrosis. Encouraged by this, we hope to use convolutional neural networks to extract ultrasound image parameters more widely in this study, and use artificial intelligence algorithms to construct a kidney fibrosis prediction model with better predictive performance.

As a biomarker for evaluating systemic inflammation, the Aggregate Index of Systemic Inflammation(AISI) was initially designed to assess the inflammatory status of patients with idiopathic pulmonary fibrosis (IPF), primarily by reflecting the ratio of immune cell subsets (such as neutrophils, lymphocytes, and monocytes) to platelet count to evaluate systemic inflammatory response.^17,18 The advantages of AISI include low cost, easy collection, and simple calculation, making it potentially valuable for clinical applications. At present, some scholars have found that AISI is a more comprehensive inflammatory indicator than SII and SIRI, which can more comprehensively evaluate the systemic inflammatory status, especially in evaluating proteinuria.^19–21 With the help of NHANES, the relationship between AISI and CKD or low eGFR has been confirmed.²² However, the potential role of AISI in kidney disease (especially renal fibrosis) still needs further clinical research to verify.

Therefore, this study collected a large sample size of clinical data and attempted to construct a clinical ML prediction model based on renal ultrasound radiomics and AISI to evaluate the degree of fibrosis in CKD, providing clinicians with a low-cost and efficient method for evaluating the degree of fibrosis.

Materials and Methods

Study Population

We retrospectively collected 758 patients diagnosed with CKD at the Second Hospital of Jingzhou and Jingzhou Hospital Affiliated to Yangtze University from January 2017 to July 2024. The inclusion criteria are as follows: (1) Diagnosis of CKD through percutaneous renal biopsy; (2) Optical microscope specimens>10 glomeruli; (3) Age>18 years old; (4) Ultrasound examination should be performed within 3 days before the patient’s renal puncture surgery. Exclusion criteria: (1) Acute kidney injury and heart valve disease; (2) Renal artery stenosis or urinary tract obstruction; (3) Cysts or tumors; (4) Ultrasound images are missing or of poor quality.

Percutaneous renal biopsy was performed by two experienced ultrasound physicians, both of whom selected the left kidney for biopsy. According to the distribution range of IF/TA in the glomerular cortex, patients are divided into three categories: TA/IF0,0–25%; TA/IF1, 26%~50%; TA/IF2 >50%. Due to the small number of TA/IF class 2 patients, they were not included in this study. The construction process of patient inclusion and prediction models was detailed in Figure 1.

Figure 1 Workflow diagram for patient inclusion and prediction model establishment.

Ethical Statement

This study was approved by the Ethics Committee of the Second Hospital of Jingzhou. Due to its retrospective nature, involving de-identified existing medical records, the Committee waived the requirement for individual patient consent, as the research posed minimal risk and obtaining consent was impracticable. Strict confidentiality measures were applied, all records were anonymized, access was restricted to the research team, and data handling complied with relevant privacy regulations to prevent patient identification. This study was conducted in accordance with the principles of the Declaration of Helsinki (World Medical Association, 2013 revision) for medical research involving human subjects.

Data Collection

We collected demographic data (age, gender, body mass index[BMI], etc)., laboratory tests (blood routine, liver and kidney function electrolytes, coagulation function, etc)., and ultrasound image information of patients from standardized sources such as electronic health records in hospitals. Other clinical data: gender, age, mean arterial pressure, hemoglobin, platelet count, creatinine eGFR1, Urea, uric acid, 24-hour urine protein, 24-hour urine volume. All data collection was independently entered and verified by two professionals. To ensure the accuracy and verifiability of the entered data, we used Epidata software for programming and input. Finally, a third-party professional performed final verification and data cleaning on the entered data.

Ultrasound Radiomics Feature Extraction

Two ultrasound doctors with over 5 years of experience in abdominal ultrasound examination use ITK 3.8.0 software (http://www.itksnap.org/) to outline the region of interest (ROI) of renal ultrasound images and perform manual segmentation (2×2 cm²), image resolution (512×512 pixels), and CNN architecture (12 layers with ReLU activation). Next, we randomly selected renal ultrasound images of patients and had two doctors independently perform ROI delineation. After 2 weeks, we repeated the same steps to evaluate the matching degree of feature extraction, while retaining features with good correlation for subsequent analysis. Finally, we used PyRadiomics 3.0.1 online analysis tool to automatically extract and quantify the features of ultrasound.

Construction of ML-Based Prediction Models

As shown in Figure 1, the features with intra group correlation coefficient (ICC)>0.75 in the training queue were retained. Then, single factor logistic regression analysis was used to screen out the features with significant differences between IF/TA class 0 and class 1 in the training group. Finally, the minimum absolute shrinkage and selection operator algorithm was used to select the optimal features, and the ML algorithm was used to establish an ultrasound radiomics prediction model.

Evaluation of ML-Based Predictive Model Performance

In this study, we used seven ML algorithm prediction models, namely extreme gradient boosting (EGE), support vector machine(SVM), random forest(RF), multilayer perceptron(MLP), artificial neural network(ANN), naive Bayes(NB), and generalized linear logistic regression(GLR), to construct prediction models. The 10-fold cross validation was used to ensure the stability of the model. We evaluated the performance of the prediction model based on the AUC, sensitivity, specificity, recall, F1 score, accuracy, and of the ROC curve. Additionally, we also plotted decision curve analysis (DCA) and calibration curves to demonstrate its true clinical use. To determine the optimal threshold probability of our model, we generated a Clinical Impact Curve (CIC) for rigorous evaluation and determination of the most effective clinical application decision threshold.

Statistical Analysis

We used SPSS 25.0 and R software (version 4.3.2). Metric data that follows a normal distribution are represented by mean ± standard deviation, metric data that does not follow a normal distribution are represented by M (Q1, Q3), and count data are represented by example (%). The t-test, Mann Whitney U-test, and chi square test were used for univariate analysis, while binary logistic regression analysis was used for multivariate analysis. We used Delong test to compare the differences in area under the curve (AUC) of each prediction model between the training group and the validation group. P<0.05 indicates a statistically significant difference.

Results

Baseline Characteristics

A total of 758 patients diagnosed with CKD were included in the final study. Among them, a total of 135 patients were diagnosed with renal fibrosis through renal biopsy, accounting for approximately 17.8%. 515 patients were randomly divided into the predictive model training set and the internal test set at a ratio of 7 to 3(as shown in Table 1). In addition, as shown in Supplementary Table 1, another 243 patients were considered as an external test set, of which 24.7% were diagnosed with renal fibrosis. In the comparison of baseline data between the renal fibrosis group and the non-renal fibrosis group, the results showed that there were no significant statistical differences in age, gender, BMI, etc. of the patients (P>0.05). However, among the relevant indicators of laboratory examinations, there were significant statistical differences between groups in terms of neutrophil count(NEU), platelet count(PLT), monocyte count(MONO), and lymphocyte count(LYM), etc. (P<0.05). The calculation formula of AISI is as follows: AISI=. The results showed that there was a significant statistical difference in AISI between the renal fibrosis group and the non-renal fibrosis group (P < 0.05). The ultrasound radiomics parameters are mainly obtained from ultrasound images, including texture features, gray-scale matrix parameters, and ultrasound hemodynamic features, etc. We used the convolution kernel parameters and bias parameters of CNN to adjust the output range of convolution. Ultimately, we extracted 167 generalized ultrasound radiomics parameters. Among them, the parameters based on the gray matrix showed a significant statistical difference between the renal fibrosis group and the non-renal fibrosis group (P<0.05). The ultrasound radiomics data were summarized in Supplementary Table 2.

Table 1 Analysis of Baseline Data for Training and Internal Validation Sets

Selection of Candidate Predictor Variables

As shown in Figure 2A, in the heat map of candidate predictor variables and outcomes (ie, renal fibrosis) constructed by Pearson correlation coefficient, the results indicated a significant positive correlation between AISI and renal fibrosis (P < 0.05), while based on the parameter characteristics extracted from ultrasound radiomics, For example, Angular Second Moment(ASM), Contrast, Correlation, Entropy, Inverse Difference Moment(IDM) were significantly correlated with renal fibrosis (P < 0.05). Then, as shown in Figures 2B and C, the LASSO regression analysis results suggest that the optimal number of variables corresponding to the equation λ-se is 5. Therefore, we finally included 5 variables into the equation. Including AISI, ASM, Contrast, Correlation, Entropy, and IDM. In addition, for the weight value distribution of the candidate variables, we also sorted the weight values based on the SHAP interpretability algorithm, as shown in Figure 2D. Based on the fact that the ultrasound radiomics parameters occupy a large weight proportion in all candidate variables, especially the ultrasound feature parameters extracted based on the co-occurrence gray matrix, it has shown extremely promising predictive power in becoming a candidate predictor variable.

Figure 2 Inclusion and weight analysis of predictive feature factors. (A) Pearson coefficient correlation analysis between candidate variables and outcomes. (B) Iterative analysis of candidate variables based on Lasso log lambda. (C) Ten-fold cross validation based on Lasso regression analysis. (D) Comparison of weight values of candidate variables interpretable based on SHAP.

Comparison of Predictive Performance Across Different Models

We constructed seven ML-based prediction models based on candidate predictor variables. As shown in Table 2, among the candidate variables screened based on logistic regression algorithm, both ultrasound radiomics parameters and AISI were independent risk factors for renal fibrosis (P<0.05). As shown in Table 3 and Figure 3, in the training set, the AUC values of various prediction models ranged from 0.72 (95% CI: 0.67~0.77) to 0.96 (95% CI: 0.92~0.99). In the internal test set, the AUC values of various prediction models ranged from 0.71 (95% CI: 0.66~0.76) to 0.93 (95% CI: 0.88~0.98). Among them, the AUC values of the prediction model based on RF algorithm in the training set and internal validation set were 0.96 (95% CI: 0.92~0.99) and 0.93 (95% CI: 0.88~0.98), respectively, while the prediction model GLR with the worst prediction performance had AUC values of 0.72 (95% CI: 0.67~0.77) and 0.71 (95% CI: 0.66~0.76) in the training set and internal validation set, respectively. In addition, the AUC value of RF in the external test set can still reach AUC values of 0.95 (95% CI: 0.89~0.99). Moreover, as shown in Figure 4, in the DCA evaluation of the net benefits of various prediction models, the prediction model constructed by the RF algorithm still has the best net benefit, indicating that among the seven ML constructed renal fibrosis prediction models, RF can become the best prediction model for predicting renal fibrosis.

Table 2 Univariate and Multivariate Logistic Regression Analysis for Independent Risk Factors

Table 3 Predictive Performance Comparison of the Seven Types of ML Algorithms

Figure 3 Evaluation of ROC for ML-based prediction model. The AUC values of 7 prediction models for the (A) Training set and (B) Internal validation set (C) External validation set.

Figure 4 Evaluation of DCA for ML-based prediction model. The DCA of 7 prediction models for the (A) Training set and (B) Internal validation set (C) External validation set.

Performance Evaluation of the RF Model on the External Cohort

After 1000 resampling times, as shown in Figure 5, RF demonstrated satisfactory robustness in all three datasets (the C-index is between 0.99 and 1.00). Then, as shown in Supplementary Table 3, since both the ultrasound radiomics parameters and AISI are assigned quantization values in the RF algorithm, the calculation formula of the RF prediction model is as follows:. Among them, n is the number of decision trees in the random forest, and k is regarded as the included predictor variable parameter. As shown in Figure 6 and Supplementary Figure 1, the discriminative efficacy of the renal fibrosis prediction model constructed based on the RF algorithm is excellent in the training set, internal and external test sets. Collectively, the renal fibrosis prediction model constructed based on RF has extremely significant clinical application value and generalization ability.

Figure 5 Evaluation of calibration curve for RF prediction model. The calibration curve after 1000 bootstrap for the (A) Training set; (B) Internal validation set; (C) External validation set.

Figure 6 Evaluating the efficacy of RF in clinical applications based on CIC. (A) Training set; (B) Internal validation set; (C) External validation set.

Discussion

CKD, as a chronic progressive disease, is expected to pose a health and medical burden to 10% to 14% of the global population worldwide.^23–25 As CKD progresses, renal fibrosis, as a hallmark manifestation of different progressive CKD, is characterized by excessive deposition of extracellular matrix leading to scar formation.²⁶ Cautiously, as the degree of renal injury progresses, progressive loss of glomerular capillary structure, tubular atrophy and narrowing, and replacement of glomerular cellular components by expanded extracellular matrix and fibrous tissue lead to a decrease in glomerular effective filtration area and eGFR.^27,28 The degree of IF/TA indicates the loss of functional nephrons and the progression of renal injury.⁶ Normally, patients with CKD are often accompanied by endothelial damage, vascular calcification, hypertension, and peripheral arterial disease, which are closely related to the development of end-stage renal disease and the occurrence of cardiovascular disease.^29,30 Hypertension is the main cause of CKD progression and persistent vascular disease.³ In this study, significant differences were found in eGFR and mean arterial pressure among CKD patients with different levels of IF/TA, indicating that as the degree of renal fibrosis increases, the risk of renal dysfunction and cardiovascular disease also increases. In view of this, non-invasive and accurate assessment of the degree of IF/TA is of great clinical significance in the selection of treatment plans and prognosis evaluation for patients with CKD.

Non invasive assessment of the degree of fibrosis in CKD has always been regarded as an urgent need in current clinical diagnosis and treatment, and radiomics based on ultrasound images has the potential to meet this demand. For example, Huang et al extracted ultrasound feature parameters through multimodal ultrasound, particularly based on shear wave elastography, and angio planewave ultrasensitive imaging characteristics.¹⁴ The constructed prediction model has an AUC value of over 0.7 in predicting CKD related renal fibrosis, indicating that ultrasound parameters have certain advantages in predictive performance. In addition, Chen et al also constructed a renal fibrosis prediction model using Ultrasonic renal length. However, renal length presented limited discrimination ability in distinguishing degrees of renal fibrosis while controlling the key confounding factors, yielding an area under the ROC curve of only 0.58 (95% CI 0.45–0.70).¹⁵ Previous studies have shown that ultrasound affects the efficacy and robustness of omics in predicting renal fibrosis, which may be closely related to the parameters extracted by ultrasound and the algorithms used to construct prediction models.^9,15,31 In addition, due to the inability of small sample queues to generalize predictive models, continuous optimization is still needed for the exploration of ultrasound radiomics. In this study, we extracted a large number of ultrasound imaging parameters based on convolutional neural networks, and screened reliable ultrasound imaging omics prediction candidate parameters based on multiple iterations. The optimal AUC values of 0.96 and 0.95 were obtained in the training and validation sets, respectively. Therefore, the renal fibrosis prediction model constructed based on ultrasound parameters and ML algorithms has great clinical application value.

AISI, as an easily accessible indicator and a novel prognostic biomarker, AISI has been used to predict patients with idiopathic pulmonary fibrosis (IPF).²¹ Previous studies have shown that it can significantly distinguish between IPF patients and healthy subjects, and AISI levels are independently associated with poor prognosis.^21,32 In addition, research should also demonstrate a significant correlation between AISI and poor prognosis in patients with viral pneumonia.¹⁹ However, few studies have investigated the predictive value of AISI for renal fibrosis outcomes.

Previous studies have shown that AISI is significantly positively correlated with the incidence of CKD and has better predictive power compared to other inflammatory indicators.²² Consistent with previous research findings, our study found that the AUC for predicting renal fibrosis based on AISI independent indicators was 0.88 and 0.89 in the training and validation sets, respectively. We speculate that perivascular cells are now considered the main innate immune sentinels in the kidneys, producing pro-inflammatory cytokines and chemokines after injury. These mediators promote immune cell infiltration, leading to persistent inflammation and progression of renal fibrosis.³³ Therefore, the interactions between perivascular cells and renal tubular epithelial cells, immune cells, and endothelial cells are key processes in physiological and pathophysiological states.

Although ML models have been proven to achieve high accuracy in clinical applications, the impact of individual variables on these models is often still unknown.³⁴ This lack of transparency limits the application of ML in clinical practice. In this study, we performed SHAP interpretability weight ranking on ultrasound imaging parameters and AISI. By combining optimal credit allocation with local interpretation, we intuitively represented the importance of each variable in the model, providing more interpretable outputs. More importantly, in the RF prediction model constructed using ultrasound radiomics combined with AISI, the AUC values for predicting renal fibrosis in the training and internal validation sets were 0.96 and 0.95, respectively, indicating that the combined parameters and ML can achieve better prediction performance. The predictive model constructed by Wu et al using RF algorithm combined with serum creatinine based approach has a predictive performance AUC value of 0.89, which is consistent with the results of this study.³⁵ This indicates that RF algorithm is suitable for constructing renal fibrosis prediction models. However, the selection of candidate variables directly determines the predictive performance and superiority of RF algorithm.

This research also inevitably has the following limitations. Firstly, as a retrospective study, selection bias is inevitable in the collection of patients’ clinical data and ultrasound images. Notably, substandard-quality ultrasound images could not be utilized, directly leading to the loss of patient information. To address this, future prospective randomized controlled trials are needed for compensation, with stricter inclusion criteria to minimize data loss. Secondly, although internal and external tests of the prediction model were conducted using data from two tertiary hospitals, the generalization ability and generalizability of the model still require repeated verification through large-sample, multi-center cohort studies. Additionally, inter-operator variability in ultrasound examinations may affect radiomic feature extraction. Although our center implemented standardized training for sonographers and regularly assessed inter-observer agreement (with a kappa coefficient of 0.82, indicating good consistency), differences in operational proficiency across different institutions may still impact model performance. Thirdly, this study incorporated ultrasound imaging parameters and albumin-to-iron ratio to construct the renal fibrosis predictive model. However, for non-invasive diagnostic models, it is necessary to continue exploring candidate markers with potential value, such as uromics and MRI radiomics. In summary, the goal is to identify optimal, convenient, and cost-effective predictive factors for clinical application. Regarding clinical implications, to enhance the practicality of the model, we suggest integrating it with portable ultrasound devices, which can facilitate its use in resource-limited settings, enabling more widespread screening and monitoring of chronic kidney disease patients.

Conclusion

In summary, we developed a RF model based on ML algorithms that combines ultrasound radiomics parameters and AISI to evaluate renal fibrosis. Their integration is a key innovation, with synergistic effects enhancing diagnostic performance. The model achieved a superior AUC of 0.96 in the training set, outperforming conventional methods. The included ultrasound radiomics parameters and AISI are easy to obtain, with high diagnostic value, good reproducibility, economy and cost-effectiveness, and they do not increase patients’ medical burden. Specifically, for high-risk renal fibrosis patients, this RF model enables dynamic evaluation of renal fibrosis and assists in pre-biopsy decision-making. It provides a practical non-invasive alternative with potential for wide clinical application to improve the efficiency and accuracy of renal fibrosis assessment.

Disclosure

The authors report no conflicts of interest in this work.

References

1. KDIGO. 2024 clinical practice guideline for the evaluation and management of chronic kidney disease. Kidney Int. 2024;105(4s):S117–s314. doi:10.1016/j.kint.2023.10.018

2. Bello AK, Okpechi IG, Levin A, et al. An update on the global disparities in kidney disease burden and care across world countries and regions. Lancet Glob Health. 2024;12(3):e382–e95. doi:10.1016/S2214-109X(23)00570-3

3. Jankowski J, Floege J, Fliser D, Böhm M, Marx N. Cardiovascular disease in chronic kidney disease: pathophysiological insights and therapeutic options. Circulation. 2021;143(11):1157–1172. doi:10.1161/CIRCULATIONAHA.120.050686

4. Schuett K, Marx N, Lehrke M. The cardio-kidney patient: epidemiology, clinical characteristics and therapy. Circu Res. 2023;132(8):902–914. doi:10.1161/CIRCRESAHA.122.321748

5. Jia YY, Yu Y, Li HJ. The research status and prospect of periostin in chronic kidney disease. Renal Failure. 2020;42(1):1166–1172. doi:10.1080/0886022X.2020.1846562

6. Tampe D, Schridde L, Korsten P, et al. Different patterns of kidney fibrosis are indicative of injury to distinct renal compartments. Cells. 2021;10(8):2014. doi:10.3390/cells10082014

7. Bullen AL, Vaingankar S, Madero M, et al. Urine uromodulin, kidney tubulointerstitial fibrosis, and furosemide response. Nephron. 2024;148(7):443–447. doi:10.1159/000534578

8. Poggio ED, McClelland RL, Blank KN, et al. Systematic review and meta-analysis of native kidney biopsy complications. Clin J Am Society Nephrol. 2020;15(11):1595–1602. doi:10.2215/CJN.04710420

9. Cè M, Felisaz PF, Alì M, Sartò GV R, Cellina M. Ultrasound elastography in chronic kidney disease: a systematic review and meta-analysis. J Med Ultrasonics. 2023;50(3):381–415. doi:10.1007/s10396-023-01304-z

10. Ng KH, Wong JHD, Leong SS. Shear wave elastography in chronic kidney disease – the physics and clinical application. Phys Eng Sci Med. 2024;47(1):17–29. doi:10.1007/s13246-023-01358-w

11. Bob F, Grosu I, Spore I, et al. Is kidney stiffness measured using elastography influenced mainly by vascular factors in patients with diabetic kidney disease? Ultrasonic Imaging. 2018;40(5):300–309. doi:10.1177/0161734618779789

12. Sohn JH, Fields BKK. Radiomics and deep learning to predict pulmonary nodule metastasis at CT. Radiology. 2024;311(1):e233356. doi:10.1148/radiol.233356

13. Li Y, Lyu B, Wang R, et al. Machine learning-based radiomics to distinguish pulmonary nodules between lung adenocarcinoma and tuberculosis. Thoracic Cancer. 2024;15(6):466–476. doi:10.1111/1759-7714.15216

14. Huang X, Wei T, Li J, et al. Multimodal ultrasound for assessment of renal fibrosis in biopsy-proven patients with chronic kidney disease. Ultraschall in der Medizin. 2025. doi:10.1055/a-2559-7743

15. Chen Z, Jiang J, Gunda ST, et al. Ultrasonic renal length as an indicator of renal fibrosis severity in non-diabetic patients with chronic kidney disease. Clin Experi Nephrol. 2025;29(4):460–468. doi:10.1007/s10157-024-02598-0

16. Okda HI, Keshk RAE, El-Anwar N, El Behiry AM, Gamaleldin SM. The role of strain wave elastography in the evaluation of renal fibrosis in patients with kidney diseases. Nefrologia. 2025;45(3):259–270. doi:10.1016/j.nefro.2024.10.007

17. Tuzimek A, Dziedzic EA, Beck J, Kochman W. Correlations between acute coronary syndrome and novel inflammatory markers (Systemic immune-inflammation index, systemic inflammation response index, and aggregate index of systemic inflammation) in patients with and without diabetes or prediabetes. J Inflamm Res. 2024;17:2623–2632. doi:10.2147/JIR.S454117

18. Xiu J, Lin X, Chen Q, et al. The aggregate index of systemic inflammation (AISI): a novel predictor for hypertension. Front Cardiovasc Med. 2023;10:1163900. doi:10.3389/fcvm.2023.1163900

19. Hosseninia S, Ghobadi H, Garjani K, Hosseini SAH, Aslani MR. Aggregate index of systemic inflammation (AISI) in admission as a reliable predictor of mortality in COPD patients with COVID-19. BMC Pulm Med. 2023;23(1):107. doi:10.1186/s12890-023-02397-5

20. Ghobadi H, Mohammadshahi J, Javaheri N, Fouladi N, Mirzazadeh Y, Aslani MR. Role of leukocytes and systemic inflammation indexes (NLR, PLR, MLP, dNLR, NLPR, AISI, SIR-I, and SII) on admission predicts in-hospital mortality in non-elderly and elderly COVID-19 patients. Front Med. 2022;9:916453. doi:10.3389/fmed.2022.916453

21. Zinellu A, Collu C, Nasser M, et al. The Aggregate Index of Systemic Inflammation (AISI): a novel prognostic biomarker in idiopathic pulmonary fibrosis. J Clin Med. 2021;10(18):4134. doi:10.3390/jcm10184134

22. Huang D, Wu H. Association between the aggregate index of systemic inflammation and CKD: evidence from NHANES 1999-2018. Front Med. 2025;12:1506575. doi:10.3389/fmed.2025.1506575

23. Fletcher BR, Damery S, Aiyegbusi OL, et al. Symptom burden and health-related quality of life in chronic kidney disease: a global systematic review and meta-analysis. PLoS Med. 2022;19(4):e1003954. doi:10.1371/journal.pmed.1003954

24. Zhao WM, Li XL, Shi R, et al. Global, regional and national burden of CKD in children and adolescents from 1990 to 2019. Nephrol Dial Transplant. 2024;39(8):1268–1278. doi:10.1093/ndt/gfad269

25. Song J, Pan T, Xu Z, et al. A systematic analysis of chronic kidney disease burden attributable to lead exposure based on the global burden of disease study 2019. Sci Total Environ. 2024;908:168189. doi:10.1016/j.scitotenv.2023.168189

26. McDonnell T, Banks RE, Taal MW, Vuilleumier N, Kalra PA. Personalized care in CKD: moving beyond traditional biomarkers. Nephron. 2025;2025:1–19.

27. Huang R, Fu P, Ma L. Kidney fibrosis: from mechanisms to therapeutic medicines. Signal Transduct Target Ther. 2023;8(1):129. doi:10.1038/s41392-023-01379-7

28. Panizo S, Martínez-Arias L, Alonso-Montes C, et al. Fibrosis in chronic kidney disease: pathogenesis and consequences. Int J Mol Sci. 2021;22(1):408. doi:10.3390/ijms22010408

29. Banerjee D, Rosano G, Herzog CA. Management of heart failure patient with CKD. Clin J Am Society Nephrol. 2021;16(7):1131–1139. doi:10.2215/CJN.14180920

30. Szlagor M, Dybiec J, Młynarska E, Rysz J, Franczyk B. Chronic kidney disease as a comorbidity in heart failure. Int J Mol Sci. 2023;24(3):2988. doi:10.3390/ijms24032988

31. Qin X, Liu X, Xia L, Luo Q, Zhang C. Multimodal ultrasound deep learning to detect fibrosis in early chronic kidney disease. Renal Failure. 2024;46(2):2417740. doi:10.1080/0886022X.2024.2417740

32. Zinellu A, Paliogiannis P, Sotgiu E, et al. Blood cell count derived inflammation indexes in patients with idiopathic pulmonary fibrosis. Lung. 2020;198(5):821–827. doi:10.1007/s00408-020-00386-7

33. Tanaka S, Portilla D, Okusa MD. Role of perivascular cells in kidney homeostasis, inflammation, repair and fibrosis. Nat Rev Nephrol. 2023;19(11):721–732. doi:10.1038/s41581-023-00752-7

34. Cabitza F, Rasoini R, Gensini GF. Unintended consequences of machine learning in medicine. JAMA. 2017;318(6):517–518. doi:10.1001/jama.2017.7797

35. Wu LH, Zhao D, Niu JY, et al. Development and validation of multi-center serum creatinine-based models for noninvasive prediction of kidney fibrosis in chronic kidney disease. Renal Failure. 2025;47(1):2489715. doi:10.1080/0886022X.2025.2489715

Klinefelter syndrome is the most common genetic cause of azoospermia. Owing to limited awareness and phenotype variability, the disease has been historically diagnosed in men during mid-adulthood during workup for fertility issues. However, advances in prenatal testing and screening have altered the diagnostic paradigm, raising questions about whether surgical sperm retrieval should occur in adolescence or be delayed until a desired time in adulthood.

This was the basis for a new study led by Scott Lundy, MD, PhD, and colleagues, who conducted a meta-analysis to assess the relationship between age and the rate of retrieval in sperm extraction in patients with nonmosaic (47, XXY) Klinefelter syndrome. They published their findings in the prestigious journal, Fertility and Sterility.

Patients with Klinefelter syndrome are likely to experience germ cell apoptosis, seminiferous tubular hyalinization and testicular interstitial hyperplasia at puberty, which complicates the perceived window for intervention and has, historically, created urgency around sperm retrieval around the time of puberty.

“We don’t know why, but if there are no germ cells or sperm precursors, then there can’t be any sperm down the road. Some providers advocate for testicular sperm extraction surgery on children soon after puberty to freeze it for future fertility treatment, like in vitro fertilization or intracytoplasmic sperm injection,” explains Dr. Lundy, adding, “This has been somewhat controversial in our field.”

This approach raises several ethical concerns, including potentially unnecessary surgery in children, who may not fully understand its implications or ultimately want to have children. Further, it does not guarantee a successful outcome, and there is a psychological cost associated with knowledge of likely infertility so early in life.

A closer look at the study

“We wanted to understand this to guide patients and parents to a more nuanced degree,” says Dr. Lundy. A previous meta-analysis was conducted in 2017 but did not include retrieval rates for adolescents.

Using PubMed, Embase and Medline, the research team extracted data from 48 studies, with a total of 2,815 participants. The researchers included outcomes from both conventional sperm extraction methods and microdissection, the current standard of care, in those with nonmosaic Klinefelter syndrome. In addition to age and sperm retrieval rate, they analyzed live birth rate, total testicular volume, preprocedural testosterone, and blood follicle-stimulating hormone and luteinizing hormone levels.

Of the 48 studies, researchers found a median sperm retrieval rate of 44%. In total, 24 studies found groups of patients with positive sperm retrieval compared to negative sperm retrieval. Ages of the positive sperm retrieval cohorts were, on average, 2.8 years younger than those with negative sperm retrieval (95% confidence interval: − 3.62 to − 2.02 years; I² = 78%). The researchers also reported no difference in sperm retrieval rates between the adolescent and adult groups (45% vs. 42%) across all studies.

The authors also reported a nonsignificant quadratic relationship between age and sperm retrieval rates, suggesting that rates may decline after age 40.

No meaningful difference in sperm retrieval rates

The takeaway, Dr. Lundy emphasizes, is that there’s no meaningful difference in the surgery’s success rate when performed in puberty versus at the average age for desired family planning, which tends to be around 30. “We can now provide some reassurance when counseling parents and patients alike that there is no urgency to rush into surgery.”

Although he cautions, one of the studies indicates that sperm retrieval may become less successful in this patient population around 40, which is just another data point for consideration when counseling patients with Klinefelter syndrome.

Questions that remain

Still, questions involving testosterone therapy in pediatric patients remain. In some cases, Dr. Lundy says it’s “certainly necessary” to facilitate puberty in patients with Klinefelter’s syndrome. In others with lower-to-normal levels of testosterone, it could negatively affect testicular function and sperm production.

“There is a possibility that pediatric patients who have gone through puberty and are placed on testosterone might have a better fertility outcome if they weren’t on testosterone. We need more data to guide which patients receive testosterone and which ones don’t,” he cautions.