Cholinergic basal forebrain modulation of glymphatic function: implications for depression

Abstract

Emerging evidence links glymphatic dysfunction to depression, yet its regulatory mechanisms remain unclear. Through multimodal neuroimaging of two independent cohorts, we tested the hypothesis that the cholinergic basal forebrain (ChBF) modulates glymphatic function and thereby contributes to depression. Glymphatic function was indexed using two indirect MRI markers: the diffusion tensor image analysis along the perivascular space (DTI-ALPS) index and coupling between global blood-oxygen-level-dependent signals and cerebrospinal fluid signals (gBOLD-CSF coupling). In a stroke cohort, we explored ChBF-glymphatic-depression interactions through dual perspectives: acute lesion effects and longitudinal recovery. Baseline lesion network mapping analysis (n = 189) revealed that functional connectivity (FC) between ChBF and lesion predicted glymphatic impairment. Longitudinal resting-state fMRI analysis (n = 71) demonstrated that recovery of ChBF whole-brain FC correlated with glymphatic restoration. Glymphatic dysfunction mediated the association between ChBF damage and post-stroke depression severity. Subsequently, we validated these findings in non-stroke cohort (n = 38) receiving transcranial magnetic stimulation. FC between the stimulation site and ChBF predicted glymphatic improvement, with enhanced ChBF whole-brain FC mediating this association. Furthermore, increased ChBF whole-brain FC was correlated with antidepressant efficacy and glymphatic enhancement mediated this relationship. Our findings identify that ChBF play a regulatory role in glymphatic function, which in turn influences depression. This study provides a theoretical foundation for the development of glymphatic-based therapeutic strategies for depression.

Introduction

Depression, the leading cause of global disability, imposes profound individual and societal consequences [1]. Despite advances in treatment, a substantial proportion of patients fail to achieve remission, underscoring the urgent need to refine pathophysiological models and identify novel therapeutic targets [2]. Recent breakthroughs have identified the glymphatic system, a brain-wide fluid transport network essential for metabolic waste clearance via perivascular cerebrospinal fluid dynamics [3], as a key player in neuropsychiatric disorders [4–9]. Growing evidence links its dysfunction to depression pathogenesis through disrupted brain homeostasis maintenance [10–12], suggesting it as a potential therapeutic target. However, the mechanisms regulating glymphatic function remain incompletely understood, limiting translational applications.

Emerging evidence suggests that cholinergic basal forebrain (ChBF), the primary source of cortical acetylcholine [13], may be a key regulator of glymphatic function. ChBF release acetylcholine, which can exert direct vasodilatory effects via endothelial muscarinic receptors [14]. These vascular changes have been shown to enhance arterial pulsatility and periarterial CSF flow [15]. Furthermore, early studies found that ChBF stimulation can induce gamma oscillations [16], which recent studies suggest may amplify glymphatic flow [17, 18]. Crucially, both preclinical and clinical investigations consistently link the ChBF dysfunction to depression-related behaviors [19, 20]. These findings collectively motivated our hypothesis that ChBF regulates glymphatic function, thereby playing a critical role in depression pathophysiology, and that ChBF maybe a potential glymphatic-based therapeutic target for depression.

Stroke patients provide an excellent experimental platform for testing this hypothesis. First, the high post-stroke depression (PSD) prevalence (~30%) provides robust statistical power [21]. Secondly, the causal association between acute lesions and ensuing functional deficits provides a powerful framework for investigating the neuroanatomical basis of glymphatic dysfunction. In this context, the recently developed lesion network mapping (LNM) [22, 23] technique links lesions in distinct brain areas that produce similar deficits to a common network. This technique allows for the exploration of causal relationships between ChBF damage and glymphatic dysfunction from network perspective. Third, longitudinal resting-state fMRI (rsfMRI) enables tracking of ChBF functional change and its association with glymphatic recovery. Importantly, findings from stroke patient models can identify neuromodulation targets, which may inspire glymphatic-based therapeutic strategies [24].

Neuroimaging innovations enable indirect evaluation of glymphatic function in vivo. For instance, the diffusion tensor imaging – analysis along the perivascular space (DTI-ALPS) index was used to assess subcortical glymphatic function near the lateral ventricles [25, 26], while another complementary measure—the coupling between global blood-oxygen-level-dependent signals and CSF (gBOLD-CSF coupling)—was used to evaluate cortical glymphatic function [27]. These imaging biomarkers have been widely applied in research on stroke [12], neurodegenerative diseases [28], and psychiatric disorders [12, 29, 30].

Here, we used neuroimaging data acquired from two cohorts to achieve three aims. Firstly, to establish ChBF-glymphatic relationships, in stroke patients, we tested whether ChBF damage reduces glymphatic function in baseline LNM analysis and ChBF function variations coincide with glymphatic function changes in longitudinal rsfMRI analysis. Secondly, to determine ChBF-glymphatic-depression tripartite relationships, we examined the relationship of PSD with glymphatic function and ChBF function and tested whether glymphatic function can mediate the association between PSD and ChBF. Finally, to probe translational potential, we analyzed transcranial magnetic stimulation (TMS) data from a non-stroke population to assess whether ChBF stimulation enhances glymphatic function and whether this enhancement mediates antidepressant effects (Fig. 1 describes the detailed workflow).

Fig. 1. Flow Chart.

A Glymphatic Function Parameters: Based on each patient’s DTI and rsfMRI data, we compute DTI-ALPS and gBOLD-CSF coupling to assess subcortical and cortical glymphatic function. B Baseline Analysis: We computed FC between each lesion and ChBF using a normative connectome (N = 1000). Subsequently, we correlate these FC with DTI-ALPS and gBOLD-CSF coupling to explore the relationship ChBF damage and glymphatic function. C Longitudinal Analysis: We measure the number of voxels displaying significant FC with the ChBF over the whole brain at different time points, quantifying the dynamic projection strength of ChBF. By correlating changes in this projection strength with glymphatic function measures, we demonstrate the ChBF glymphatic relationship from a longitudinal perspective. D Depression, cholinergic function, and glymphatic Function: First, cross-sectional and longitudinal associations between depression, ChBF function, and glymphatic function were tested. Mediation analyses evaluated whether glymphatic function mediates the relationship between ChBF function and depression. E Therapeutic potential of modulating ChBF: Based on TMS data from 38 non-stroke participants, we validated the conclusions drawn from the stroke population. First, we quantified the FC between the stimulation site and ChBF to evaluate TMS-induced effects on ChBF. Then, correlations between ChBF-StimFC and post-TMS glymphatic changes were tested to explore the cholinergic-glymphatic association. Next, we examined the relationship of changes in depression scores with the whole-brain projection strength of ChBF, and improvements in glymphatic function. Finally, mediation models were used to test whether glymphatic improvement underlies the antidepressant effects of cholinergic modulation. ChBF, Cholinergic Basal Forebrain; DTI-ALPS, diffusion tensor image analysis along the perivascular space; gBOLD-CSF, blood-oxygen-level-dependent signals and cerebrospinal fluid signals; FC, functional connectivity; TMS, Transcranial Magnetic Stimulation; ChBF-StimFC, The stimulation site’s FC with ChBF.

Materials and methods

Stroke patient cohort

This study was approved by the Ethics Committee of the First Affiliated Hospital of the University of Science and Technology of China (USTC) (approval number: 2021-ky-012). All methods involving human participants were performed in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki. Written informed consent for study participation was obtained from all participants before enrolment.

We analyzed data from the CONNECT cohort [31], collected between March 2021 and November 2024, which met the predefined inclusion and exclusion criteria. To further meet the study’s requirements, at baseline, we excluded patients without imaging evidence of infarction and those whose diffusion tensor imaging (DTI) and resting-state functional MRI (rsfMRI) did not satisfy the requirements for calculating glymphatic function indices. For the longitudinal analysis, we further included patients who had imaging assessments at both baseline and follow-up. Ultimately, 189 patients were included in the baseline analysis, and 71 patients were included in the longitudinal analysis. All patients provided written informed consent. Supplemental Fig. 1 presents a flow chart of the study population.

Neuropsychiatric assessments and MRI scans were conducted at baseline and follow-up. Demographic and clinical data were recorded. Depression severity was evaluated using the Hamilton Depression Rating Scale (HAMD). All scale assessments were performed by a neurologist using a semi-structured interview approach.

Lesion segmentation

Lesions were manually segmented on MRI scans by an experienced neurologist using ITK-SNAP version 3.8.0 (www.itksnap.org). The resulting lesion masks were co-registered to the MNI152 T1 template utilizing both linear and nonlinear registration techniques available in SPM12. The normalized lesion maps were visually inspected and manually corrected as necessary to ensure accurate alignment. Supplemental Fig. 2 depicts the infarct distribution maps.

Glymphatic function index

We employed the DTI-ALPS index to assess subcortical glymphatic function as introduced in prior studies [12, 25]. We placed four spherical ROIs in white matter regions where deep medullary veins were perpendicular to the lateral ventricles (Fig. 1A and Supplemental Fig. 3). Diffusivity within the ROIs on the projection fibers and association fibers was defined as Dxproj, Dyproj, Dzproj, Dxassoc, Dyassoc and Dzassoc, respectively. DTI-ALPS index = (mean (Dxproj, Dxassoc)) / (mean (Dyproj, Dzassoc)). We excluded DTI-ALPS indices where lesions overlapped with the ROIs. Although this method may introduce some bias, it ensures the accuracy of glymphatic assessment in stroke patients.

Additionally, the large cortical BOLD signal has been reported to be associated with strong CSF movement, potentially indicating cortical glymphatic function [27]. We examined cortical glymphatic function by assessing the coupling between the global BOLD signal and CSF flow (Fig. 1A and Supplemental Fig. 4). Following standard rsfMRI preprocessing, we extracted time-series signals from cortical gray matter and CSF in the upper spinal canal, then calculated their cross-correlation. The maximal negative correlation (Supplemental Fig. 4C) was identified as the gBOLD-CSF coupling strength, with more negative values indicating tighter synchronization (i.e., stronger glymphatic pumping). gBOLD-CSF coupling statistical significance was evaluated by permutation testing.

A detailed description of DTI-ALPS and gBOLD-CSF coupling is provided in the Supplemental Method.

Lesion network mapping

In the baseline analysis, to test our hypothesis that “ChBF damage reduces glymphatic function”, we employed previously validated methods for LNM [32] (Fig. 1B).

Initially, we computed functional connectivity (FC) between each lesion location and all other brain voxels using rsfMRI data from 1000 healthy subjects (https://doi.org/10.7910/DVN/ILXIKS). This process generated an unthresholded lesion connectivity map for each patient, with T-values representing the strength of brain regions connected to the lesion. This unthresholded method can avoid bias associated with a choice of threshold.

Subsequently, we conducted hypothesis-driven analyses. Firstly, the ChBF was defined as region of interest (ROI) based on a prior atlas [33], which has been widely used in multiple studies to assess cholinergic function in the brain [34, 35]. Next, we computed the connectivity (sum of T-values within the ChBF mask) for each lesion following previous methodologies [32]. Then, average FC was regressed against lesion volume and the residuals were used as ChBF network damage score to control for lesion volume. Simultaneously, to control for confounding factors of glymphatic function indexes, we regressed these indexes against onset time, age, gender, and lesion volume and extracted the residuals to create an adjusted glymphatic indexes, which were used in further analyses. Finally, we performed spearman correlation analyses between ChBF network damage and glymphatic indexes in order to the relationship of ChBF damage and glymphatic function. Statistical significance was determined using non-parametric permutation testing with 5000 iterations and a p-threshold of 0.05.

Additionally, we conducted data-driven analyses across the whole brain. Following previous methodologies [32], lesion connectivity maps were compared to adjusted glymphatic function indexes using general linear model with 5000 permutations, including lesion volume as a covariate. Regions significantly associated with glymphatic indices were identified using a voxelwise family-wise error (FWE) control threshold of p < 0.05.

Longitudinal functional changes

To validate our hypothesis that “ChBF function variations coincide with glymphatic function changes”, we analyzed the longitudinal data of rsfMRI, DTI-ALPS, and gBOLD-CSF coupling (Fig. 1C).

Originally, fMRI data preprocessing was conducted using fMRIPrep and the CONN Toolbox. Detailed preprocessing procedures are provided in the Supplemental Method. Subsequently, considering ChBF’s primary physiological function is cholinergic projection in our study, we quantified its whole-brain projection intensity to assess ChBF function. Specifically, using the ChBF as ROI, we calculated the number of FC between ROI and every voxel in the brain, applying a threshold of r > 0.25. The number of voxels exhibiting suprathreshold FC (r > 0.25) was considered the whole-brain projection intensity of ChBF. Then, we calculated the changes in ChBF projection intensity, DTI-ALPS, and gBOLD-CSF coupling between follow-up and baseline. Finally, we performed Spearman correlation analyses between the changes in whole-brain projection intensity of ChBF and the two glymphatic indices to explore the association between ChBF functional changes and glymphatic function improvement.

We repeated the aforementioned analyses using alternative threshold values of r > 0.2 and r > 0.3 to conduct sensitivity analyses.

Depression, ChBF function and glymphatic function

To investigate the relationship among ChBF, glymphatic function and depression, we utilized stroke patients’ baseline and longitudinal data to explore these associations (Fig. 1D).

Firstly, we utilized a general linear model to explore the relationship of PSD with ChBF and glymphatic function. For baseline analyses, general linear models were used with baseline HAMD scores as the dependent variable and either the ChBF network damage score, baseline DTI-ALPS, or baseline gBOLD-CSF coupling as independent variables. In longitudinal analyses, changes in HAMD scores were modeled as changes in ChBF projection intensity, DTI-ALPS, or gBOLD-CSF coupling. These models included age, gender, onset time and lesion volume as covariates. Additionally, mediation analyses were conducted using the mediation package in R (with 5000 bootstrap iterations) to test whether glymphatic function mediates the relationship between cholinergic function and depressive symptoms. ChBF function (i.e., network damage score or changes in ChBF projection intensity) was treated as the independent variable, depression measures (i.e., baseline HAMD scores or changes in HAMD scores) as the dependent variable, and DTI-ALPS or gBOLD-CSF coupling (baseline or changes) as the mediating variables. The above analyses were all performed on normalizing data.

TMS validation

To explore the therapeutic potential of modulating ChBF for improving glymphatic function and alleviating depressive symptoms, we conducted a retrospective analysis of TMS data from a non-stroke population (Fig. 1E). Data were drawn from a double-blind, randomized, sham-controlled trial investigating the effect of high-frequency rTMS targeting the left dorsolateral prefrontal cortex in patients with cocaine use disorder [36, 37]. To maximize statistical power, patients in the active treatment arm were combined with those who, after completing the sham phase, received open-label active TMS. A total of 38 patients with complete imaging data were included.

Initially, multimodal imaging data and Hamilton Depression Rating Scale (HDRS) scores obtained pre- and post-stimulation allowed us to analyze the following parameters: (1) The stimulation site’s FC with ChBF (ChBF-StimFC), serving as a measure of the stimulation effect on ChBF, similarly previous methods [23, 38]. (2) Changes in DTI-ALPS and gBOLD-CSF coupling, representing improvements in glymphatic function. (3) Changes in whole-brain ChBF projection intensity as an index of ChBF functional alteration. (4) Changes in HDRS scores to quantify depressive symptom improvement. To control for confounding factors (medication, treatment group, gender and age), each parameter was regressed against these variables, and the residuals were used in further analyses. The detailed processing steps for this section are described in the supplementary material.

Next, spearman correlation analyses were performed to determine whether ChBF stimulation could modulate glymphatic function by correlating ChBF-StimFC with changes in glymphatic parameters (DTI-ALPS and gBOLD-CSF coupling). Subsequently, mediation analyses were executed to assess whether improvements in ChBF function mediated the relationship between ChBF stimulation and changes in glymphatic function. In this model, the independent variable was ChBF-StimFC, the mediator was the change in ChBF projection intensity, and the dependent variables were changes in glymphatic function.

Finally, to elucidate the interrelationship among ChBF modulation, glymphatic function and depressive symptom alleviation, we first examined spearman correlations of changes in HDRS scores with alterations of glymphatic parameters and ChBF projection intensity. We then performed mediation analyses using the changes in ChBF projection intensity as the independent variable, changes in glymphatic function as the mediator and the changes in HDRS scores as the dependent variable.

The TMS validation cohort used publicly available, de-identified SUDMEX-TMS data from OpenNeuro (ds003037, version 2.1.0). The original trial was approved by the Ethics Committee of the Instituto Nacional de Psiquiatría “Ramón de la Fuente Muñiz” (CEI/C/070/2016), registered at ClinicalTrials.gov (NCT02986438), and obtained written informed consent from all participants.

Results

Demographic and clinical features

Clinical and demographic characteristics for the baseline (n = 189; 62 females; age: 57.9 ± 11.7 years) and the follow-up data (n = 71; 23 females; age: 56.1 ± 11.8 years) are summarized in Table 1. Although a substantial number of participants were lost to follow-up, there were no statistically significant differences in demographic variables or glymphatic parameters between the baseline and follow-up patients. The lesion incidence maps for both baseline and follow-up patients are presented in Supplemental Fig. 2.

Table 1.

Clinical and demographic information on the cohort of patients.

	Baseline N = 189	Follow-up N = 71	p values
Age, mean (SD)	57.9 (11.7)	56.1 (11.8)	0.285
Female, n (%)	62 (32.8%)	23 (32.4%)	1.000
Infarct volume (mm3), mean (SD)	9670 (13680)	10553 (15195)	0.669
Interval between onset and scanning, mean (SD)	7.58 (5.29)	220.57 (90.83)	-
Baseline HAMD	5.86 (4.21)	5.83 (3.91)	0.963
Follow-up HAMD	-	5.03 (4.15)	-
Baseline DTI-ALPS	1.27 (0.15)	1.29 (0.12)	0.257
Baseline gBOLD-CSF	−0.18 (0.15)	−0.18 (0.15)	0.933
Follow-up DTI-ALPS	-	1.28 (0.13)	-
Follow-up gBOLD-CSF	-	−0.20 (0.14)	-

HAMD hamilton depression scale, gBOLD-CSF coupling, the coupling between

blood-oxygen-level-dependent signals from global gray matter and cerebrospinal fluid signals; DTI-ALPS, diffusion tensor image analysis along the perivascular space index.

Association between glymphatic function and ChBF damage

Using lesion network mapping, we first tested the hypothesis that ChBF damage reduces glymphatic function. For each participant, we computed FC between their lesion mask and the whole brain using a normative connectome. In our hypothesis-driven analysis, the ChBF network damage score was determined by the average FC within ChBF. Spearman correlation analyses revealed that network damage score was significantly associated with glymphatic function parameters (DTI-ALPS: r = –0.302, p = 0.002; gBOLD-CSF coupling: r = 0.408, p < 0.001; Fig. 2A). A data-driven whole-brain FC analysis confirmed that significant clusters indeed overlapped within the ChBF mask (p < 0.05, FWE-corrected; Fig. 2B).

Fig. 2. Association between the ChBF and glymphatic function in stroke cohort.

A Lesion-Induced ChBF Network Damage Correlates with Glymphatic Dysfunction: ChBF network damage scores, derived from FC between lesion and the ChBF using a normative connectome, were negatively associated with subcortical glymphatic function (DTI-ALPS index, left) and positively correlated with cortical glymphatic dysfunction (gBOLD-CSF coupling, right). Spearman correlations (p-values) are overlaid on permutation-derived null distributions (histograms). B Whole-Brain Lesion Network Mapping Confirms ChBF-Glymphatic Associations: Voxelwise general linear models identified brain regions where lesion connectivity patterns significantly predicted glymphatic parameters (FWE-corrected p < 0.05). Top: Regions associated with DTI-ALPS. Bottom: Regions associated with gBOLD-CSF coupling. T-maps reflect correlation strength, with darker colors indicating more significant statistics. C Longitudinal ChBF Projection Recovery Parallels Glymphatic Restoration: Changes in ChBF whole-brain projection intensity (voxels with FC > 0.25) between baseline and follow-up correlated with improvements in DTI-ALPS (left) and gBOLD-CSF coupling (right). Permutation-derived p-values confirm robustness. FC functional connectivity, ChBF Cholinergic Basal Forebrain, DTI-ALPS diffusion tensor image analysis along the perivascular space, gBOLD-CSF blood-oxygen-level-dependent signals and cerebrospinal fluid signals.

Correlation between changes in ChBF projection intensity and glymphatic function

Next, to test the hypothesis that functional alterations in the ChBF occur in parallel with changes in the glymphatic function, we analyzed the longitudinal data of ChBF function and glymphatic function parameters. As shown in Fig. 2C, changes in ChBF projection intensity were significantly correlated with alterations in both DTI-ALPS (r = 0.295, p = 0.023) and gBOLD-CSF coupling (r = −0.428, p < 0.001). These associations remained robust in sensitivity analyses using correlation thresholds of r ＞ 0.2 (DTI-ALPS: r = 0.367, p = 0.007; gBOLD-CSF: r = –0.293, p = 0.013) and r ＞ 0.3 (DTI-ALPS: r = 0.329, p = 0.017; gBOLD-CSF: r = –0.237, p = 0.048) (Supplemental Fig. 5).

PSD, ChBF function and glymphatic function

In this section, we explored the interplay among ChBF function, glymphatic function, and PSD.

As summarized in Table 2, general linear modeling demonstrated that both baseline and longitudinal changes in PSD severity were significantly linked to glymphatic function and ChBF function. At baseline, higher depression scores were associated with lower DTI-ALPS indices (β = –0.381, p < 0.001) and higher gBOLD–CSF coupling (β = 0.274, p < 0.001). Similar patterns emerged for symptom trajectories over time, with worsening depression correlating with decreased DTI-ALPS (β = –0.382, p = 0.0033) and increased gBOLD–CSF coupling (β = 0.275, p = 0.0267). Additionally, higher baseline ChBF network damage predicted more severe depressive symptoms (β = 0.373, p < 0.001); similarly, reductions in ChBF projection intensity during follow-up were linked to symptom exacerbation (β = −0.283, p = 0.0196).

Table 2.

Associations between post-stroke depression severity, glymphatic parameters, and ChBF function.

	Baseline HAMD				Change HAMD
	β value	Std. Error	P values		β value	Std. Error	P values
DTI-ALPS	−0.381	0.081	<0.001	DTI-ALPS change	−0.382	0.123	0.0033
gBOLD-CSF coupling	0.274	0.070	<0.001	gBOLD-CSF coupling change	0.275	0.122	0.0267
Network damage score	0.373	0.067	<0.001	Changes in ChBF projection intensity	−0.283	0.118	0.0196

HAMD hamilton depression scale; gBOLD-CSF coupling, the coupling between

blood-oxygen-level-dependent signals from global gray matter and cerebrospinal fluid signals; DTI-ALPS, diffusion along perivascular space index; ChBF, cholinergic basal forebrain.

In order to test whether glymphatic function mediates the association between ChBF damage and PSD, we performed mediation analyses in both the baseline and longitudinal data. At baseline (Fig. 3A), ChBF network damage score was entered as the independent variable, baseline HAMD score as the dependent variable, and DTI-ALPS and gBOLD–CSF coupling as mediators. Results showed significant indirect paths through both glymphatic indices: DTI-ALPS mediated the effect of ChBF damage on depression (indirect β = 0.069, p < 0.001), and gBOLD–CSF coupling likewise conveyed a significant indirect effect (indirect β = 0.059, p = 0.048). These findings indicate that impaired glymphatic function partially explains how ChBF function disruption contributes to depressive symptom soon after stroke.

Fig. 3. Glymphatic Function Mediates Cholinergic-Depression Relationships in Stroke.

A Baseline Mediation: Glymphatic dysfunction (DTI-ALPS, β = 0.069, p < 0.001, top; gBOLD-CSF coupling, β = 0.059, p = 0.048, bottom) mediated the association between ChBF network damage and depression severity (HAMD scores). B Longitudinal Mediation: Improvements in ChBF projection intensity predicted HAMD score reductions, with trends toward mediation by DTI-ALPS (β = −0.095, p = 0.073) and gBOLD-CSF coupling (β = −0.055, p = 0.106). * represent p < 0.05.

By contrast, in the longitudinal analysis (Fig. 3B), we substituted change in ChBF projection intensity for the network damage score, and change in HAMD for the outcome, with corresponding changes in DTI-ALPS and gBOLD–CSF as mediators. Neither glymphatic index showed a significant indirect effect on the relationship between ChBF alterations and depression change (DTI-ALPS: indirect β = −0.095, p = 0.073; gBOLD–CSF: indirect β = −0.055, p = 0.106). Thus, although baseline ChBF damage appears to drive depressive severity via glymphatic dysfunction, dynamic changes in ChBF and glymphatic over time do not significantly mediate PSD trajectories.

ChBF stimulation and glymphatic function

In order to validate our findings from stroke patients and explore the therapeutic potential of modulating cholinergic function, we analyzed TMS data from a non-stroke population. Firstly, to explore the causal link between ChBF stimulation and glymphatic function enhancement, we tested the relationship of ChBF Stimulation and glymphatic function using spearman correlation. Results showed that stronger ChBF-StimFC was associated with more glymphatic function enhancement (DTI-ALPS: r = 0.377, p = 0.012; gBOLD-CSF coupling: r = −0.320, p = 0.040, Fig. 4A).

Fig. 4. Validation of ChBF-Glymphatic-Depression Relationships in Non-Stroke Cohort Using Transcranial Magnetic Stimulation.

A ChBF-Targeted Stimulation Modulates Glymphatic Function: Functional connectivity between TMS stimulation sites and the ChBF predicted post-intervention increases in DTI-ALPS (left) and normalized gBOLD-CSF coupling (right). Permutation testing validated associations. B Cholinergic Function Engagement Mediates Glymphatic Effects: Changes in ChBF projection intensity mediated the relationship between ChBF-StimFC and glymphatic improvements (DTI-ALPS: β = 0.125, p = 0.026, left; gBOLD-CSF: β = −0.150, p = 0.020, right). C Antidepressant Response Correlates With Glymphatic and Cholinergic Changes: Depression improvements were associated with glymphatic restoration (DTI-ALPS, left; gBOLD-CSF, middle) and ChBF projection enhancement (right). D Glymphatic Function Mediates Antidepressant Effects of ChBF Modulation: Improvements in glymphatic function significantly mediated the association between enhanced ChBF projections and depression alleviation (DTI-ALPS: β = −0.162, p = 0.040, left; gBOLD-CSF: β = −0.133, p = 0.020). * represent p < 0.05; † represent p < 0.1.

Next, we tested whether TMS-induced changes in ChBF projections mediate these glymphatic effects. In mediation models (Fig. 4B), ChBF-StimFC served as the independent variable, changes in glymphatic parameters as the dependent variable, and changes in ChBF projection intensity as the mediator. Significant indirect paths were observed for both glymphatic measures (DTI-ALPS: indirect β = 0.125, p = 0.026; gBOLD–CSF: indirect β = −0.150, p = 0.020). The above results suggest that ChBF stimulation may cause glymphatic function improvement by enhancing ChBF projection.

ChBF and glymphatic function in depression improvement

Then, we further explored the relationship among depressive symptom improvement, glymphatic function and cholinergic function. Firstly, correlation analysis showed significant associations between depression improvement and glymphatic function enhancement (DTI-ALPS: r = −0.369, p = 0.027; gBOLD-CSF coupling: r = 0.529, p = 0.001, Fig. 4C), as well as ChBF projection intensity strengthen (r = −0.340, p = 0.030, Fig. 4C).

Subsequently, in mediation analysis, changes in ChBF projection intensity were entered as the independent variable, changes in depression scores as the dependent variable, and differences in glymphatic function as the mediator. Results showed that glymphatic function enhancement significantly mediated the association between changes in ChBF projection intensity and depression score improvements (DTI-ALPS: β = −0.162, p = 0.040; gBOLD-CSF: β = −0.133, p = 0.020; Fig. 4D).

Discussion

Glymphatic systemin is increasingly recognized as the critical pathophysiology of depression, underscoring its promise as a therapeutic target. However, the mechanisms that govern glymphatic flow remain poorly defined, hindering the development of targeted interventions. Concurrently, recent research have pointed to the ChBF as a critical regulator of perivascular CSF dynamic, yet direct data have been lacking. Here, we present the critical human evidence that ChBF modulated glymphatic function and, in turn, influenced depressive symptoms. Specifically, we found that: (1) In stroke patients, damage and subsequent recovery of ChBF function align with corresponding impairments and improvements in glymphatic function. (2) PSD is associated with both glymphatic function and ChBF function, with glymphatic function mediating the relationship between ChBF damage and PSD. (3) In retrospective TMS analysis, enhancing cholinergic projections through ChBF stimulation modulates glymphatic function. Moreover, glymphatic function mediates the association between enhanced cholinergic projections and improvement in depressive symptoms. Together, these findings identify ChBF as regulators of glymphatic function, providing a theoretical foundation for future glymphatic-based therapies in depression.

Our first novel conclusion is the establishment of a causal link between ChBF function and glymphatic function. Several lines of evidence support this conclusion: (1) Utilizing LNM analysis techniques, the FC between ChBF and lesion exhibited significant associations with glymphatic function. (2) Longitudinal analyses showed that changes in ChBF projection intensity were significantly correlated with alterations in glymphatic function. (3) TMS data showed that patients with stronger ChBF-StimFC experienced greater improvements in glymphatic function, with enhancing ChBF projections mediating this effect. Mechanistically, cholinergic modulation may influence glymphatic function through direct vasodilatory effects [14, 15] or by indirectly inducing gamma oscillations that promote vascular motility [17, 18] —a mechanism that has been the subject of recent debate. Tsai et al. demonstrated that 40 Hz sensory stimulation can induce gamma oscillations within specific brain circuits, enhancing vascular motility and glymphatic flow [39]. However, this conclusion has been challenged by recent research [40, 41]. Interestingly, one study found that while 40 Hz stimulation did not enhance glymphatic flow, it did elevate cholinergic levels [41]. This phenomenon implies that 40 Hz sensory stimulation, as a non-specific stimulus, may be influenced by individual differences, and that modulation of cholinergic levels could represent a more precise downstream target. Our research established a connection between ChBF function and glymphatic function based on stroke model. More importantly, by leveraging neuromodulation data, we further confirmed this relationship from a therapeutic perspective, revealing the potential value of modulating ChBF to improve glymphatic function for therapeutic purposes. However, due to the absence of electroencephalographic data, we remain uncertain whether ChBF regulates glymphatic function through endogenous gamma oscillations. In fact, ChBF is also closely related to brain states and alertness [42], which are similarly associated with glymphatic function. Future studies should combine multimodal imaging to further analyze whether this regulatory pathway is mediated by gamma oscillations.

Our second major conclusion is that glymphatic function mediated the relationship between ChBF function and depression. This conclusion is supported by multiple findings: (1) In baseline analyses of stroke patients, glymphatic function mediated the association between FC of lesion-ChBF and depression scores. (2) In TMS data, changes in glymphatic function mediated the relationship between ChBF projection enhancement and depression improvements. These observations align with literature highlighting acetylcholine’s role in mood and cognitive regulation; dysregulated central cholinergic signaling has been implicated in depression, whereas cholinergic activation can enhance learning, memory, and affective responses [19]. The glymphatic system serves as a foundation for maintaining brain homeostasis, which is a critical step in the pathogenesis of various diseases, and may represent a therapeutically targetable final common pathway [10]. Our mediation analysis thus suggests that cholinergic-based antidepressant strategies may exert their effects by globally improving the brain’s internal milieu via enhanced glymphatic clearance. Combined with our findings in TMS data that ChBF stimulation can enhance ChBF projection intensity and subsequently improve glymphatic function, this may reveal a new glymphatic-based regulatory pathway for depression.

Notably, our longitudinal analysis revealed no significant mediating effects, which could be explained by several factors. First, cholinergic systems interact with various neurotransmitter systems to regulate glymphatic flow [43]. Adaptive changes in other neurotransmitter systems may correlate with glymphatic function, thereby diminishing the correlation between glymphatic function and cholinergic projections in longitudinal analyses. Second, our assessment of cholinergic projections relied solely on whole-brain FC of ChBF. This approach may fail to detect functional redistribution—for example, shifts to hippocampal regions following ChBF neuron damage—thereby reducing the correlation. Third, non-biological factors and the clinical heterogeneity of depression could further obscure the correlation. Finally, limited statistical power due to withdrawn may have influenced results. Future studies should address these limitations by incorporating multimodal assessments of neurotransmitter activity and accounting for compensatory mechanisms.

Additionally, there are other limitations to this study. First, methodologically, DTI-ALPS and gBOLD–CSF coupling are MRI-based indirect surrogates of neurofluid dynamics and do not directly measure glymphatic flow/clearance [44]. The ALPS index, in particular, depends on fiber geometry at the lateral ventricular body and periventricular white-matter integrity [45]; in stroke, periventricular injury and ventricular deformation may reduce its reliability and interpretive specificity, warranting cautious interpretation. Future studies using more direct or multimodal glymphatic assessments are needed to validate and generalize our conclusions regarding glymphatic function. Second, stroke and glymphatic function may have a bidirectional relationship [46]. However, the pre-stroke brain state of patients is mostly unknown, and individual differences in pre-lesion brain conditions could confound the relationship between brain injury and glymphatic function. Future studies could address this by investigating populations with available pre-lesion neuroimaging data or by combining large imaging databases for normative modeling. Third, we analyzed the associations among depression, glymphatic function, and cholinergic function in two heterogeneous cohorts, but all conclusions were based on depression scale scores rather than strict clinical diagnoses. Therefore, caution is warranted when generalizing our findings to primary mood disorders.

In conclusion, our study supports the conclusion that ChBF play a regulatory role in glymphatic function, which in turn influences depressive symptoms. This study provides a vital theoretical foundation for the development of future glymphatic-based therapeutic strategies for depression.

Author contributions

Yirong Fang: Conceptualization, Formal analysis, Data curation, Methodology, Writing – original draft. Zeyu Lu: Conceptualization, Writing – review and editing, Data curation. Xian Chao: Validation, Visualization. Ran Shi: Data curation, Investigation. Dawei Yin: Data curation, Investigation. Yanan Lu: Data curation, Investigation. Peng Wang: Resources, Supervision, Project administration, Data curation. Xinfeng Liu: Resources, Supervision, Project administration, Data curation, Funding acquisition. Wen Sun: Resources, Supervision, Project administration, Data curation, Funding acquisition, Writing – review and editing.

Funding

This work was supported by a collaborative research grant to National Natural Science Foundation of China (NSFC-U22A20341, NSFC-82471341 and NSFC-82571496), Natural Science Foundation of Anhui Province (No. 2108085MH271), Research Funds of Centre for Leading Medicine and Advanced Technologies of IHM (No. 2023IHM01051 and 2023IHM01054) and Joint Fund for Medical Artificial Intelligence (No. MAI2022Q018).

Data availability

Code for replicating all analyses uploaded to github (https://github.com/fyr12345/Cholinergic-and-Glymphatic-). Some of the data may be private and unpublished, the corresponding author can be contacted if necessary. TMS data were made publicly available by Eduardo A.Garza-Villarreal et al. to OpenNeuro (https://openneuro.org/datasets/ds003037/versions/2.1.0).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04284-z.