A Histopathologic Correlation Study Evaluating Glymphatic Function in Brain Tumors by Multiparametric MRI

The glymphatic system plays a crucial role in maintaining brain extracellular homeostasis and facilitating the removal of metabolic waste products (1). Arterial pulsations drive cerebrospinal fluid (CSF) through the periarterial spaces, which are surrounded by astrocytic endfeet containing aquaporin-4 (AQP4) water channels that allow CSF to enter the brain parenchyma (2, 3). Within the brain parenchyma, CSF undergoes fluid exchange with interstitial fluid, leading to the removal of waste products and intracranial fluids through the perivascular spaces surrounding the veins (4). Therefore, dysfunctions in the glymphatic pathway have been linked to various neurological disorders (5, 6), such as Alzheimer’s disease, in which inefficiencies in glymphatic function are associated with reduced clearance of amyloid-beta and tau proteins (7, 8).

Brain tumors, whether primary or metastatic, can lead to obstructive hydrocephalus by exerting pressure and compromising the blood-brain barrier (BBB), ultimately affecting fluid dynamics in the perivascular region and the intracranial glymphatic system (9–11). Preclinical investigations have demonstrated that glioma-bearing mice exhibit a marked reduction in CSF outflow as a result of impaired circulation within the intracranial cavity, potentially exacerbating brain edema, altering the brain tumor microenvironment, and diminishing the efficacy of intracranial drug delivery (12). Hence, investigating the function of the glymphatic system in maintaining brain fluid equilibrium is essential for elucidating the pathophysiology of brain tumor-induced disruption of brain homeostasis and establishing a foundation for the advancement of therapeutic interventions aimed at restoring CSF outflow and normalizing CSF circulation. In line with the approach of utilizing intrathecal tracer injections in rodent models, human research has similarly documented CSF drainage through magnetic resonance imaging (MRI) scans conducted at various intervals following intrathecal gadolinium contrast administration (8, 13–15). Nevertheless, the invasive nature and extended scan time of this method render it unsuitable for frail patients with brain tumors.

Advanced MRI techniques offer a promising avenue for the noninvasive, quantitative, and repetitive evaluation of glymphatic function. Diffusion tensor imaging (DTI) analysis along the perivascular space (ALPS) approach enables the quantification of glymphatic activity in the perivascular space utilizing multidirectional diffusion coefficient maps derived from DTI data (16). A higher DTI-ALPS index signifies increased diffusivity of perivascular water, indicating enhanced glymphatic function (17). Multiparametric MRI (MTP) represents an innovative approach in which multiple MRI tissue properties are efficiently mapped through a single acquisition (18). Some studies have correlated various quantitative parameters with corresponding neuropathological changes. They demonstrated that proton density (PD) can be used to assess free tissue water content (19), while the T1 value exhibits positive correlations with tissue water content and gliosis, and negative correlations with iron and myelin content (20). Additionally, T2^∗ serves as an indicator of iron content (20, 21), whereas T1 values ranging from 1,850 to 3,200 ms have been predictive of higher tumor cell density (22). These quantitative parameters serve as indicators of subtle tissue characteristics and aid in the identification of variations in the local fluid environment within the tumor microenvironment and glymphatic system.

This study utilized quantitative maps derived from DTI-ALPS and MTP MRI data to assess alterations in glymphatic function related to brain tumors and investigate potential correlations with peritumoral brain edema (PTBE) and AQP4 expression levels. The primary objective was to identify neuroimaging markers that could provide insights into the interplay between the glymphatic system and tumor-induced edema.

This prospective study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of the Second Xiangya Hospital of Central South University (Project Number: 32200747). Written informed consent was obtained from all participants.

In contemplation of the divergent impact that tumors of distinct histopathological classifications may exert upon the glymphatic system, our study incorporated three comparative cohorts, namely glioblastoma, brain metastases, and aggressive meningiomas. Unlike generally encapsulated meningiomas, we define invasive meningiomas as those pathologically classified as nonmalignant but exhibiting growth that breaches the tumor capsule, resulting in infiltration into the brain parenchyma. Achieving Simpson Grade I resection is challenging as, during the excision process, it becomes necessary to remove contiguous portions of the invaded brain tissue. This circumstance served as the source of our meningioma and related specimens from the peritumor area. All subjects encompassed in this investigation were admitted to the Department of Neurosurgery at the Second Xiangya Hospital of Central South University during the period spanning from March 2022 to May 2023. The following patients were excluded if they (i) had other concomitant organic or mental disorders of the central nervous system; (ii) had severe sleep disorders; (iii) had a history of hypersensitivity reactions to contrast agents, a history of severe allergic reactions in general, and evidence of renal dysfunction and were pregnant or breastfeeding women; (iv) had contraindications for MRI; or (v) had images exhibiting significant artifacts.

During the study period, to verify the effects of different types of brain tumors on brain glymphatic drainage, a healthy control (HC) group matched by 1:1 age (based on age at the time of imaging) and gender were included. Inclusion criteria of HCs encompassed the following: (i) no history of tumors, (ii) no neurological or psychiatric disorders, (iii) no sleep disorders, and (iv) no use of any medication affecting the central nervous system.

Previous studies have reported the influence of circadian rhythms on the functionality of the glymphatic system; the glymphatic system tends to be more active during nocturnal sleep (23). To minimize the potential impact of scanning time differences on the outcomes, all patients in this study were scheduled for scanning between 5:00 pm and 7:00 pm. MRI data were acquired on a United Imaging uMR790 3T scanner equipped with a 32-channel head coil.

A 3D whole-brain MTP scan was acquired for all patients with brain tumors. MTP sequence features a dual-time-to-repetition (TR), dual-flip angle, and multiecho design. One MTP scan generates 2 (N1 + N2) sets of echo images. The acquisition parameters were as follows: α1/α2 = 4°/16°, TR1/TR2= 8.26/26.84 ms, five echoes with time to echo (TE) = 3.46–20.6 ms and ΔTE = 2.24 ms (bipolar readouts), BW = 200 Hz/px, matrix = 320 × 190 × 64, and voxel size = 1 × 0.7 × 2 mm³. The head scans used acquisition acceleration with 2× parallel imaging, 83% PE resolution, and elliptical k- space filling, leading to an acquisition time of 7 minutes 19 seconds. MTP enables the simultaneous generation of T1 maps, T2^∗ maps, PD maps, and quantitative susceptibility mapping (QSM). The imaging parameters of DTI: maximum b-value = 1,000 seconds/mm², 64 noncollinear directions, TR = 5,180 ms, TE = 71.7 ms, slice thickness = 3 mm with no gap between slices, matrix = 68 × 68, FOV = 204 × 204 mm². The scanning protocol further included 3D T1-weighted (T1) imaging pre-gadolinium and post-gadolinium injection, T2-weighted (T2) imaging, and 3D fluid-attenuated inversion recovery (FLAIR). Contrast-enhanced T1-weighted MRI scans were conducted using the contrast agent gadobutrol (Gadavist, German Bayer). The administration of the contrast agent was facilitated through a high-pressure injector, introduced via a venous bolus at a flow rate of 1.5 mmol/seconds at a dosage of 0.1 mmol/kg. Detailed MRI scan parameters are presented in Supplementary Material S1.

The tumor location, morphology, cystic changes, necrosis, lesion margins, tumor enhancement characteristics, intratumoral hemorrhage, and PTBE were collectively assessed by two neuroradiologists. All 3D MRI sequences, including postcontrast T1 (T1+C), FLAIR, and MTP images, were registered to the 3D precontrast T1 images using SPM12 software (RRID:SCR_007037). The T1 images were segmented into three tissue classes (CSF, gray matter and white matter) using the segmentation tool in SPM12. Subsequently, the tumor volume of interest (VOI), comprising both enhancing and nonenhancing areas (including central necrotic regions), was delineated based on the 3D T1+C images, while the PTBE VOI was delineated based on the 3D FLAIR images. The segmentation of lesions was performed by two neuroradiologists using a 3D Slicer (RRID:SCR_005619) through manual delineation. In case of discrepancies, a senior neuroimaging physician provided the final adjudication. Then, all segmentation labels were imported into the ITK-SNAP (RRID:SCR_002010) software, and the built-in statistical functionality was employed to compute the volumes of the respective labels.

The DTI-ALPS index was employed to evaluate the water diffusion rate in the perivascular spaces surrounding the medullary veins at the level of the lateral ventricles (Fig. 1). It primarily reflects the brain's capacity to transport fluid from subcortical regions to the lateral ventricles (16) and has been used to indirectly reflect whole-brain glymphatic function. The analysis steps are outlined as follows: (i) DTI preprocessing was performed on a Linux workstation using the FSL (RRID:SCR_002823). This preprocessing included correction of distortions caused by eddy currents, participant motion, susceptibility-induced distortions, and bias fields. (ii) The diffusion tensor was calculated using Diffusion Toolkit (RRID:SCR_017345), which includes color-coded fractional anisotropy (FA) maps and diffusion coefficients along the x-, y-, and z-axes. (iii) Two 3-mm cubic VOIs were delineated by two neuroradiologists at the level of the lateral ventricle in the cerebral hemisphere contralateral to the tumor. These VOIs were located in the projection and association fibers. (iv) The diffusion coefficient D of the voxel levels within the VOIs was subjected to statistical analysis by two radiologists in the x-, y-, and z-axis directions to derive Dxxproj, Dxxassoc, Dyyproj, and Dzzassoc, respectively. Subsequently, the mean values of these coefficients were computed. The DTI-ALPS index was then determined using the following formula: DTI-ALPS index = mean (Dxxproj, Dxxassoc)/mean (Dyyproj, Dzzassoc).

Quantitative parameter maps, including T1, T2^∗, PD maps, and QSM, were automatically generated on the UIH workstation at the end of the MTP acquisition. To measure the quantitative parameter values in the tumor and peritumor areas, we aligned the segmented tumor and PTBE masks, as described in the previous section, with the preprocessed MTP quantitative maps and extracted the mean values based on voxels within the predefined VOI using FSL. We compared the mean quantitative values within the VOI between tumor types and analyzed the correlation between the quantitative parameters, DTI-ALPS, and AQP4 expression levels.

In adherence to the established principles of surgical neuro-oncology, every patient afflicted with malignant brain tumors or aggressive meningiomas underwent comprehensive tumor excision, accompanied by a prophylactic extended resection of cerebral tissue potentially infiltrated within a 2-mm radius from the tumor periphery. The task of histopathology analysis was undertaken by a seasoned, board-certified pathologist possessing 14 years of experience in the field. The pathologist conducting the analysis was unaware of the results of the MRI assessments. For each patient, a single paraffin-embedded block was made available. To prepare the histological samples, paraffin blocks were subjected to the processes of deparaffinization and rehydration, following which they were sectioned into 5-μm slices. The evaluation commenced with the application of standard hematoxylin and eosin (H&E) staining. Subsequently, the histological sections were subjected to staining procedures involving AQP4 (Proteintech Cat# 16473-1-AP, RRID:AB_2827426, dilution 1:150) and GFAP (Proteintech Cat# 60190-1-Ig, RRID:AB_10838694, dilution 1:100) antibodies. GFAP is a commonly used marker for astrocytes in central nervous system research. As an immunofluorescence stain, GFAP serves as a background indicator showcasing the arrangement of astrocytes in different local microenvironments surrounding tumors and their corresponding relationship with AQP4 expression levels. Because of its lack of specificity, GFAP is presented solely as a staining background indicator akin to DAPI (BD Biosciences Cat# 564907, RRID:AB_2869624). An appropriate representative region was chosen for subsequent analysis. It is imperative to deliberately exclude areas categorized as “necrotic zones,” characterized by the absence of viable tissue cells, “calcification zones,” in which viable tissue cells are absent, and “hemorrhage zones,” marked by the presence of a significant number of blood cells because of abnormal bleeding. Cellular quantification pertaining to the stained samples was executed using a magnification factor of 20×, with the assistance of the ImageJ (RRID:SCR_003070). The ensuing statistical analyses were conducted to prevent the pathologist from having access to the experimental conditions, ensuring a blinded approach. All data manipulation and statistical procedures were performed using GraphPad Prism (RRID:SCR_002798).

Descriptive statistics characterized participant features, presenting sample means, standard deviations, or appropriate measures for demographic and clinical variables. Categorical variable comparisons employed Pearson's χ² test. The normality of continuous variables was assessed using the Shapiro–Wilk test. For normally distributed data with homogeneity of variances, one-way ANOVA was used; if violated, Welch's ANOVA was applied. Nonnormally distributed data underwent Kruskal–Wallis and Mann–Whitney tests. Pearson's and Spearman's correlation cofficients evaluated relationships between MTP parameters, AQP4 expression levels, and the ALPS index. The Kruskal–Wallis test and Dunn–Bonferroni post hoc analysis compared MTP parameters across histological grade groups. SPSS (RRID:SCR_002865) version 28.0 facilitated all statistical analyses.

The data generated in this study are available within the article and its Supplemental Data Files. A structured repository of data does not exist. Additional data (e.g., deidentified raw data) are available upon request to the corresponding author, Jun Liu, subject to necessary authorization from the Data Review Committee of the author’s institution. All data requests must comply with the Personal Information Protection Act to ensure the privacy and confidentiality of patient data. The corresponding author, Jun Liu, will provide guidance on the privacy-preserving procedures required for data access.

Of the 143 individuals recruited for this study, 84 were diagnosed with brain tumors (average age ± SD: 55 ± 12 years, ranging from 23 to 80 years) and 59 were healthy controls (HCs; average age ± SD: 54 ± 8 years, ranging from 24 to 69 years). The clinical and histopathological traits of the participants are tabulated in Table 1. The representativeness of the study participants is displayed in Supplementary Table S1. Figure 2 details the flowchart of patient enrollment and classification. All patients underwent MTP scanning and surgical tumor resection. DTI scanning and tumor specimen AQP4 fluorescence staining were limited to 61 patients. The mean duration between MRI scans and surgery was approximately 3 days.

Neuroimaging biomarkers such as tumor volume, PTBE volume, CSF volumes, and DTI-ALPS index were used as surrogate markers of tumor burden, edema magnitude, intracranial CSF balance, and glymphatic function, respectively. The segmentation showed excellent interobserver agreement [intraclass correlation coefficient, 0.91 (95% confidence interval, 0.86–0.97)]. Intergroup analyses across three distinct levels revealed the following findings: (i) compared with HCs, patients with brain tumors demonstrated increased CSF volumes (201.376 cm³ vs. 115.957 cm³, P < 0.001) and diminished DTI-ALPS indices (2.315 vs. 2.879, P = 0.001) (Supplementary Fig. S1A). (ii) Cross-examination of various tumor categories disclosed no statistically significant dissimilarities in neuroimaging biomarkers, with the exception of alterations in tumor volume (P < 0.001). (iii) Exploration of various regional VOIs underscored a markedly elevated AQP4 expression level in PTBE at 22.225 compared with the tumor itself at 4.593 (P < 0.001). The comprehensive between-group comparative findings are showcased in Table 1. The results of pairwise comparisons among different tumor subgroups are shown in Supplementary Figure S1B.

In the tumor mass, the DTI-ALPS index significantly decreased with increasing tumor volume (r = −0.715, P < 0.001), concomitant with a corresponding increase in CSF volume (r = −0.747, P < 0.001). The AQP4 expression levels showed no significant correlation. Conversely, within the PTBE region, the AQP4 expression level was significantly positively correlated with PTBE volume (r = 0.989, P < 0.001). However, the DTI-ALPS index was negatively correlated with PTBE volume only in meningiomas (r = −0.508, P = 0.016), with no significant correlation observed in gliomas and brain metastases. Comprehensive statistical findings are visually represented in Figure 3.

Figure 4 presents the H&E-stained images (Fig. 4A), AQP4 immunofluorescence images (Fig. 4B), and quantitative parameter maps of T1, T2^∗, PD, and QSM (Fig. 4C) for representative cases of meningioma, glioblastoma, and metastatic tumors. Significant statistical differences in the mean values of T1, T2^∗, PD, and QSM were observed among the different tumor subgroups within the tumor regions (P = 0.004, P = 0.001, and P = 0.018), whereas only the mean value of T2^∗ showed significant differences in the PTBE region (P = 0.006). Detailed comparisons can be found in the Supplementary Figure S2.

The heatmaps in Figure 5 display the correlation analysis results between these four quantitative parameters and the DTI-ALPS index, as well as AQP4 expression levels. The PD map exhibited the strongest correlation, showing a significant positive correlation between the mean PD and the DTI-ALPS index within the tumor region (ρ = 0.503, P < 0.001). In the PTBE region, mean PD showed a significant negative correlation with AQP4 expression levels in the edema area (r = −0.506, P < 0.001).

In this prospective study, we leveraged state-of-the-art neuroimaging techniques, including MTP and the DTI-ALPS index, to elucidate the alterations in glymphatic function precipitated by brain tumors and the ensuing modulations in CSF dynamics. Our investigation unveiled two pivotal insights: First, our results are the first in humans to observe an inverse relationship between tumor size and the efficacy of the glymphatic system's clearance capabilities, manifesting in augmented retention of CSF. This finding is consistent with the hypothesis that physical obstruction or pressure exerted by larger tumors can impair the physiological mechanisms responsible for fluid clearance. Second, regions adjacent to tumors demonstrated not only an increase in edema but also an elevated AQP4 expression level, which was significantly associated with a reduction in PD within these edematous areas. These results underscore the potential of neuroimaging biomarkers to quantitatively delineate alterations in water molecular composition and flux, providing a novel approach for assessing the integrity and function of the glymphatic system in the aftermath of brain tumor–related injuries.

Our data suggest that decreased glymphatic function and increased CSF volume in patients with brain tumors may be attributable to an increased degree of physical compression of brain tissue by tumor volume, which reduces the periarterial space and impedes the entry of CSF, thus affecting intracranial fluid circulation. This finding is consistent with earlier animal-centered findings (12). Our research identified a novel and significant association in human subjects. This discovery has potential clinical implications as a subgroup of individuals with brain tumors may manifest hydrocephalus and necessitate ventriculoperitoneal CSF shunting for symptom management.

Several studies have reported that increased PTBE volume can lead to a decrease in the DTI-ALPS index, suggesting impaired glymphatic function (9, 10, 24). In our research, this correlation demonstrated a distinctive feature of aggressive meningiomas and was not statistically significant for malignant brain tumors such as glioblastomas and brain metastases. This difference may be related to the unique nature of extracranial occupancy by meningiomas in which PTBE primarily arises from the compression of the mass on adjacent vessels (25). The biological behavior of tumor compression results in a narrowing of the periarterial space, diminished glymphatic drainage, and expansion of the volume of cerebral edema (26). In contrast, glymphatic function in malignant brain tumors may be influenced by a combination of factors beyond mere physical compression. These factors incorporate the disruption of the BBB and altered interstitial fluid dynamics, which results from the infiltration of tumor and inflammatory cells (27) and eventually complicate the inner link between PTBE volume and glymphatic efficiency. Our hypothesis was confirmed in a separate study utilizing DTI. De Belder and colleagues demonstrated a discernible difference in FA and apparent diffusion coefficient values between the vasogenic edema surrounding meningiomas and tumor-infiltrating edema surrounding high-grade gliomas (28). The FA value was observed to be lower in tumor-infiltrating edema compared with vasogenic edema, which is potentially attributable to the disruption of typical white matter tracts in tumor-infiltrating edema. This distinction underscores the complexity of tumor-induced alterations in brain fluid dynamics and highlights the importance of considering tumor type and associated pathophysiological mechanisms when evaluating glymphatic function.

One significant finding of our study was the robust positive association between the volume of PTBE and the AQP4 expression levels. This finding is consistent with previous studies that demonstrated a positive correlation between AQP4 expression levels in the peritumoral region and edema indices (29), irrespective of tumor grade, tumor volume, Ki-67 expression, and cell count (30, 31). Within the context of brain tumors, disruption of the BBB is indicative of the onset of peritumoral vasogenic edema, prompting a compensatory increase in AQP4 expression levels to regulate brain water homeostasis (32, 33). Simultaneously, the tumor's physical compression impacts cerebral glymphatic drainage, leading to an alteration in interstitial hydrostatic pressure and subsequent development of interstitial brain edema. This condition, in conjunction with vascular and cellular cerebral edema, forms a detrimental cycle. It is hypothesized that the absence of a clear association between PTBE expansion and diminished glymphatic function in malignant tumors in our findings may be attributed to the enhanced movement of water molecules facilitated by the compensatory upregulation of AQP4 expression levels. This finding is consistent with the role of aquaporins in reducing parenchymal resistance and facilitating the movement of water and solutes (34). Furthermore, Alghanimy and colleagues confirmed through various MRI techniques that, compared with the vehicle group, rats treated with the AQP4 facilitator TGN-073 showed more extensive distribution and higher parenchymal uptake of Gd-DTPA. The TGN-073 group demonstrated an increased rate of water diffusion within the brain, indicating greater intracerebral water flux. This suggests that the AQP4 facilitator can improve glymphatic function (35). Our study extends these observations to encompass a range of brain tumors, such as glioblastomas with PTBE zones and brain metastases characterized by “small lesion with extensive edema.” The presence of significant edema in these tumors has implications for surgical timing, surgical outcomes, and the initiation of radiotherapy and chemotherapy. Future research aimed at modulating AQP4 expression levels to enhance glymphatic system function offers a promising therapeutic approach for addressing diseases associated with glymphatic system dysfunction.

Utilizing the novel 3D MTP technique, our research produced comprehensive whole-brain quantitative parametric maps encompassing T1, T2^∗, PD, and QSM values, offering valuable insights into the evaluation of subtle tissue alterations in neurological disorders (36–39).

Our rigorous quantitative analyses of the parametric mappings revealed a notable positive association between PD in tumor regions and overall brain glymphatic function. Conversely, a distinct negative relationship was observed between PD and AQP4 expression levels within PTBE. The PD maps primarily represent the spatial distribution of water molecules, given that water is the predominant proton-containing entity (40). The decrease in PD within the tumor region may be explained by multiple factors, including elevated nuclear-to-cytoplasm ratio in tumor cells, hemorrhage, and necrosis, collectively leading to significant disruption of normal cellular and tissue structures (41–43). Hence, the observed reduction in PD within the tumor parenchyma in this study may indicate a heightened intricacy of the tumor microenvironment, potentially impacting intracranial fluid dynamics. In the realm of noninvasive imaging methodologies, the PD map could serve as an adjunctive modality for DTI-ALPS in forthcoming research endeavors, facilitating a connection between glymphatic system functionality and AQP4 expression levels. Although the findings based on MTP are theoretically sound and can be interpreted from a pathophysiological point of view, there remains a lack of histological validation in numerous instances. Consequently, it is imperative for researchers within the realm of MTP to strive for correlations between quantitative parameters and histological evidence of neurological diseases whenever feasible.

Nonetheless, it is crucial to underscore specific limitations inherent to our study, warranting consideration in subsequent investigations. First, our study cohort size was relatively constrained, especially when contrasting various brain tumor subtypes, potentially attenuating the statistical robustness of our outcomes. We aim to increase the cohort size in future studies. Second, because of ethical strictures, the procurement of HC brain tissues for AQP4 immunofluorescence staining has remained elusive. Incorporating such tissues in future studies could furnish pivotal insights into normative AQP4 expression levels. Third, our study demonstrated the intricate nature of the glymphatic system, with many pathological interpretations relying on imaging findings that await validation through future molecular imaging studies. Additionally, a promising direction for further investigation would involve longitudinal surveillance to track the restoration of the glymphatic system following surgery-induced reduction of mass effects. These studies have the potential to shed light on the possibility of reversing glymphatic dysfunction in individuals with brain tumors. Addressing these nuances would undoubtedly enrich the depth and breadth of prospective studies in this domain.

Our research indicates a potential relationship between tumor-induced mechanical compression, resulting in glymphatic dysfunction, and elevated peritumoral fluid accumulation through the integration of DTI-ALPS and MTP methodologies. This cascade may trigger the upregulation of AQP4 expression levels, potentially aiding in molecular efflux. These results contribute to the current knowledge of intracranial fluid dynamics and provide insights for the development of therapeutic interventions targeting tumor-related cerebral edema.

H. Lin and J. Yuan are experts in scientific research cooperation at the Central Research Institute, United Imaging Healthcare. No disclosures were reported by the other authors.

M. Gao: Conceptualization, resources, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft. Z. Liu: Writing–review and editing. H. Zang: Data curation. X. Wu: Data curation. Y. Yan: Data curation. H. Lin: Methodology. J. Yuan: Methodology. T. Liu: Writing–review and editing. Y. Zhou: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, project administration, writing–review and editing. J. Liu: Conceptualization, resources, supervision, funding acquisition, validation, investigation, project administration, writing–review and editing.

This work was supported by the National Natural Science Foundation of China (82102157); the National Natural Science Foundation of China Youth Fund (32200747); the Hunan Provincial Health Commission Natural Science Foundation (202204043379); and the Key Research and Development Project of the Science and Technology Department of Hunan Province (2022SK2047; 2023SK2033).