Glymphatic Dysfunction and Related Brain Structure Changes in Major Depressive Disorder: Effects of Glymphatic Function in Mediating Neuroinflammation
Article information
Abstract
Objective
The glymphatic system, responsible for cerebrospinal fluid flow and waste clearance, is increasingly implicated in the pathophysiology of major depressive disorder (MDD) through its influence on neuroinflammation. This study investigated the association between glymphatic dysfunction, systemic inflammation, and brain volume changes in patients with MDD.
Methods
Glymphatic function was assessed using the diffusion tensor image analysis along the perivascular space (DTI-ALPS) index in 176 patients with MDD and 178 controls. Inflammatory cytokine levels, including plasma C-reactive protein (CRP) levels, were measured in 68 patients with MDD and 54 controls. Depressive symptoms were evaluated using the Hamilton Depression Rating Scale. Statistical analyses included a multivariate analysis of covariance and Pearson’s partial correlations adjusted for covariates. Mediation analysis examined the relationships between CRP, glymphatic function, and brain volume.
Results
Patients with MDD showed reduced glymphatic function compared to healthy controls. Reduced DTI-ALPS indices were correlated with higher CRP levels and increased ventricular volumes, including the choroid plexus. CRP levels were negatively correlated with DTI-ALPS indices and right choroid plexus volumes. Mediation analysis indicated that glymphatic dysfunction partially mediated the relationship between elevated CRP levels and decreased choroid plexus volume in patients with MDD.
Conclusion
This study found that in MDD, glymphatic dysfunction is associated with higher CRP and mediates the link between systemic inflammation and right choroid plexus volume. Given sample size and limited covariate control, these results are preliminary and need confirmation in larger longitudinal cohorts. Even so, impaired glymphatic function may represent a therapeutic target.
INTRODUCTION
Major depressive disorder (MDD) is a prevalent and debilitating psychiatric condition that affects millions of individuals worldwide [1-3]. Despite extensive research, the etiology of MDD remains complex and multifaceted, involving genetic, environmental, and neurobiological factors [4]. Recent studies have highlighted the potential role of glymphatic system dysfunction in the pathophysiology of MDD, suggesting a novel avenue for understanding its etiology [5-7].
The glymphatic system is a brain-wide network of perivascular channels that facilitates the flow of cerebrospinal fluid (CSF) and the clearance of metabolic waste products from the brain parenchyma. This system plays a crucial role in maintaining brain homeostasis by removing potentially toxic proteins and metabolites [8]. Glymphatic system dysfunction has been implicated in various neurological conditions, such as Alzheimer’s disease, traumatic brain injury, and even normal aging [9-11]. Depression is a common comorbidity in Alzheimer’s disease, and glymphatic dysfunction plays a role in this association [7].
Gu et al. [5] showed that glymphatic dysfunction induces oxidative stress and neuroinflammation in patients with MDD, suggesting a direct contribution of these pathological processes to depressive symptoms. Further supporting this notion, the regulation of stress-induced depression by ependymal cells and CSF flow emphasizes the critical role of CSF dynamics in mood regulation [12]. Loureiro-Campos et al. [13] discussed how disruptions in CSF circulation can influence depression, reinforcing the idea that glymphatic dysfunction may be intricately related to MDD pathogenesis. Additionally, polyunsaturated fatty acid supplementation alleviates depression-induced cognitive dysfunction by protecting the cerebrovascular and glymphatic systems, indicating that enhancing glymphatic function may have therapeutic benefits in depression [6]. Collectively, these findings suggest that the impaired glymphatic system may play a central role in the neurobiological foundations of MDD.
Several structural MRI findings suggest a possible connection between MDD and impaired glymphatic system. Studies have consistently reported lateral ventricular enlargement and increased CSF volumes in patients with MDD compared to healthy controls (HCs) [14,15]. Patients with MDD also show reduced volumes of several brain regions, including the hippocampus and basal ganglia, which are influenced by or contribute to impaired waste clearance in the brain [16,17]. Furthermore, recent studies have supported these findings by showing that enlargement of the brain ventricles and choroid plexus size can predict poor treatment response in patients with MDD [18]. Despite these observations, direct evidence linking glymphatic dysfunction to MDD in structural MRI studies remains limited. Diffusion tensor image analysis along the perivascular space (DTI-ALPS) index, calculated from diffusion tensor imaging (DTI) data, has been proposed as a measure of perivascular clearance activity in the human brain [19]. Importantly, as an MRI-derived surrogate, the DTI-ALPS index can be influenced by white-matter microstructure, crossing-fiber geometry, and scanner variability, and should not be interpreted as a direct measure of glymphatic clearance. Although not specifically studied in MDD, this index has shown associations with cognitive impairment in other neurological disorders and could potentially be applied to investigate glymphatic function in depression [20].
Synthesizing this literature, prior work has prioritized structures directly linked to glymphatic function, including the ventricles and choroid plexus. Accordingly, the present study aimed to investigate glymphatic system function in patients with MDD using advanced structural MRI techniques. Furthermore, we explored the relationships between glymphatic function impairment, depressive symptoms, glymphatic-related brain structural changes, and inflammatory cytokine levels. The first hypothesis was that patients with MDD show reduced glymphatic function compared to HCs. The second hypothesis is that glymphatic dysfunction is associated with enlargement of the ventricle and choroid plexus and levels of inflammatory cytokines. Based on the results of our second hypothesis, if factors are correlated with glymphatic function, we intend to verify through mediation analysis whether neuroinflammation accompanying the onset of depression affects glymphatic function and, consequently, can induce changes in brain volume. Understanding the role of glymphatic dysfunction in MDD could provide valuable insight into the neurobiology of the disorder and potentially lead to the development of novel therapeutic approaches targeting the glymphatic system. Identifying biomarkers associated with glymphatic function can also improve diagnosis and inform personalized treatment strategies for patients with MDD.
METHODS
Participants
This study included 189 patients diagnosed with MDD and 181 HCs from June 2018 to August 2022. Board-certified psychiatrists (Ham B.J., Han K.M.) at Korea University Anam Hospital in Seoul, Republic of Korea, diagnosed MDD according to the guidelines of the Structured Clinical Interview and based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Axis I Disorders. The exclusion criteria included any additional major psychiatric disorders, psychotic symptoms such as delusions or hallucinations, a history of severe or uncontrolled medical conditions, underlying neurological disorders, or any factors that would prevent brain scanning due to physiological (e.g., metal implants) or psychological (e.g., claustrophobia) reasons.
Written informed consent was obtained from all participants and their right to withdraw from the study was acknowledged. All research procedures were conducted in accordance with the ethical principles outlined in the Declaration of Helsinki and approved by the Institutional Review Board (IRB) of Korea University Anam Hospital (IRB Nos. 2017AN0185, 2019AN0174, 2020AN0335, and 2202AN0190).
Clinical assessments
Demographic information from participants was collected, including age, sex, and educational level, and the Hamilton Depression Rating Scale (HDRS) was used to assess depressive symptoms in both the patient and control groups [21]. Widely used in empirical research on psychiatric patients, HDRS is a well-validated translated version known for its discriminant validity and reliable consistency in Korea [22]. The scale consists of 17 items rated on 3-point or 5-point scales. In this study, the total scores from all 17 questions were categorized as follows: 0–7, no depressive symptoms; 8–16, mild depression; 17–23, moderate depression; and ≥24, severe depressive symptoms [23]. Considering these cutoff scores, three participants in the HC group with an HDRS score of ≥8 were considered to have mild symptoms but had depressive symptoms and were excluded. Similarly, 13 patients in the depressed group with an HDRS score of ≤7 were considered in remission and excluded. Therefore, 176 patients in the MDD group and 178 in the HC group were included in the final analysis. The three insomnia items on the HDRS—early insomnia (item 4), middle insomnia (item 5), and late insomnia (item 6)—were rated on a scale of 0 (none), 1 (mild), or 2 (severe). We also calculated the global insomnia score by summing these items, resulting in a total possible score ranging from 0 to 6.24,25 State anxiety was quantified with the clinician rated Hamilton Anxiety Rating Scale (HARS) [26]. The scale comprises 14 items, each scored 0=“not present” to 4=“severe,” yielding a total score of 0–56. Consistent with conventional cutoffs, severity was classified as none/mild (0–13), mild (14–17), moderate (18–24), and severe (≥25). Current suicidal thoughts were assessed with the self reported Beck Scale for Suicide Ideation (SSI) [20], Korean version [14]. The SSI contains 19 items rating intensity of suicidal ideation and intent on a 0–2 scale, producing a total score of 0–38. Following established practice, a total score ≥6 was interpreted as clinically significant suicidal ideation [27]. To calculate the illness duration of depression, we assessed the period from the onset of symptoms to the time of evaluation, excluding euthymic periods whenever possible, based on comprehensive interviews conducted by psychiatrists.
Inflammatory cytokines
During the initial visit, all participants were assessed for depressive symptoms using HDRS and underwent an MRI. However, only those who consented to blood sampling had their neuroinflammatory cytokine levels measured using blood tests. Consequently, cytokine levels were determined in 68 participants in the MDD group and 54 participants in the HC group who agreed to undergo blood sampling. Briefly, blood plasma was separated from blood within 30 min, and the supernatant was stored at -80°C for up to 6 months before being thawed on the day of the assay. The plasma samples were assayed in duplicate using the Milliplex Human High Sensitivity T Cell Magnetic Bead Panel (Merck, 21-plex), a highly sensitive bead-based multiplex immunosorbent assay designed to simultaneously and quantitatively measure cytokines, including C-reactive protein (CRP), interleukin (IL)-1β, IL-6, IL-8, and tumor necrosis factor α (TNF-α). Further detailed analytical methods are provided in the Methods section of our previous publication [28].
MRI data and DTI-ALPS index
Participants underwent T1-weighted imaging at the Korea University MRI Center using a 3.0-Tesla TrioTM whole-body imaging system (Siemens Healthcare GmbH). Images were acquired parallel to the anterior commissure–posterior commissure line using a 3D T1-weighted magnetization-prepared rapid gradient-echo sequence with the following parameters: repetition time (TR) of 1,900 ms; echo time (TE) of 2.6 ms; field of view of 220 mm; matrix size of 256×256; slice thickness of 1 mm; 176 coronal slices without gaps; voxel size of 0.86×0.86×1 mm3; flip angle of 16°; and one excitation.
Cortical reconstruction and volumetric segmentation were performed using the FreeSurfer image analysis suite (https://surfer.nmr.mgh.harvard.edu/). Detailed procedures are described in earlier publications [29-32]. This process involves motion correction, removal of non-brain tissue, automated Talairach transformation, and segmentation of subcortical structures (e.g., hippocampi, amygdalae, ventricles, and choroid plexuses) [33,34]. It also includes intensity normalization [35], tessellation of the gray-white matter boundary, topology correction [30], and surface deformation based on intensity gradients [29]. Once cortical models are generated, additional procedures such as surface inflation,36 registration to a spherical atlas [37], and parcellation based on cortical folding [38] can be performed. Cortical thickness is measured as the distance from the gray/white to the gray/CSF boundary at each vertex, utilizing spatial intensity gradients [39]. These maps, which are not limited by voxel resolution, can detect submillimeter differences. The methods have been validated against histology [40] and manual measurements [41,42] and show good test-retest reliability across scanners [43,44].
DTIs were acquired using an echo-planar imaging sequence with the following parameters: TR, 6,300 ms; TE, 84 ms; field of view, 230 mm; matrix size of 128×128; 3-mm slice thickness with no gap; transversal orientation; voxel size, 1.8×1.8×3.0 mm; diffusion directions, 20; number of B0 images, 1; number of slices, 50; b-values, 0 and 600 s/mm2; flip angle, 90°; acceleration factor (iPAT- GRAPPA), 2 with 38 reference lines for phase encoding direction and 6/8-phase partial Fourier. Correction of artifacts in DTI was facilitated using MRtrix3 (https://www.mrtrix.org) software tools [45]. The command dwidenoise was utilized for denoising based on the Marchenko–Pastur Principal Component analysis approach, and the removal of Gibbs ringing artifacts was achieved using the mrdegibbs command. Additionally, corrections for distortions induced by eddy currents and adjustments for subjects’ movements were made using the eddy command provided by Functional MRI of the Brain Software Library (FSL) [46].
The generation of fractional anisotropy (FA) maps, as well as diffusivity maps across the x-, y-, and z-axes, was accomplished through the dtifit command in the FSL. The calculation of segmentation- based total intracranial volume (ICV) was performed using the run_samseg command from FreeSurfer software (version 7.4.1; https://surfer.nmr.mgh.harvard.edu/), which included the analysis of individuals’ T1 and T2 MRI, which were co-registered to the same subject’s T1 MRI [47]. The alignment of each subject’s FA maps to the Johns Hopkins University International Consortium of Brain Mapping (JHU-ICBM)-FA template was performed using the flirt command in FSL. The identification and labeling of the superior corona radiata (SCR) and superior longitudinal fasciculus (SLF) were based on the JHU-ICBM-DTI 81 White-Matter Labeled Atlas, which categorizes these structures as projections and associated fibers. Spherical regions of interest, each with a diameter of 5 mm, were automatically defined in the bilateral SCR and SLF areas, from which the diffusivity values were extracted: Dxxproj and Dyyproj for the projection fibers, and Dxxassoc and Dzzassoc for the associated fibers [48].
DTI-ALPS method exploits the unique anatomical arrangement at the level of the lateral ventricle body, where projection fibers (SCR) run perpendicular to association fibers (SLF), and both are perpendicular to the perivascular spaces coursing along medullary veins [49]. Because water diffusion along these perivascular channels is relatively unrestricted when glymphatic flow is intact, the ratio of diffusivity along the perivascular direction (x-axis) to that along other axes provides an indirect estimate of glymphatic activity [50]. Importantly, as an MRI-derived surrogate, the DTI-ALPS index can be influenced by white-matter microstructure, crossing-fiber geometry, and scanner variability, and should not be interpreted as a direct measure of glymphatic clearance. The overall mean DTI-ALPS index was calculated as the average of these indices (Figure 1).
Visualization of ROI localization via DTI-ALPS processing. This figure illustrates the automated identification of ROIs associated with the SLF and SCR using the DTI-ALPS pipeline. The color-coded ROIs are as follows: sky blue, right superior longitudinal fasciculus; yellow, right superior corona radiata; red, left superior corona radiata; and light green, left superior longitudinal fasciculus. The MRI scans are shown in radiological orientation, with the left side of the image representing the right cerebral hemisphere. DTI-ALPS, diffusion tensor image analysis along the perivascular space; ROIs, regions of interest; SLF, superior longitudinal fasciculus; SCR, superior corona radiata.
In addition to ventricular and choroid plexus morphometry, we included the DTI-ALPS index to evaluate glymphatic function. Whereas ventricular and choroid-plexus volumes represent macrostructural indices of CSF production and turnover that may reflect downstream consequences of impaired clearance [9,11,18,51], the DTI-ALPS index provides a microstructural functional surrogate of perivascular water-transport efficiency [19,52]. These two measures interrogate distinct yet complementary levels of the glymphatic–CSF system, enabling a systemslevel assessment of glymphatic dysfunction in MDD.
Statistical analysis
For categorical variables such as sex, the chi-square test was used for data analysis. Additionally, continuous variables, including age and years of education, were evaluated using the Student’s t-test. Before parametric testing, homogeneity of variance was assessed with Levene’s test and was not violated for the reported comparisons. When comparing DTI-ALPS and serum cytokine levels between the two groups, we used one-way multivariate analysis of covariance, adjusting for age, sex, years of education, and ICV as covariates. Furthermore, to investigate the relationships between DTI-ALPS, insomnia severity, and serum cytokine levels, two-tailed Pearson’s partial correlations were conducted, controlling for age, sex, years of education, and ICV. Additionally, when analyzing the 17 brain volumes, we applied Bonferroni correction to reduce Type I errors. The p-value threshold was set at 0.00294, calculated by dividing 0.05 by 17 regions of the brain [53]. Before applying statistical tests (e.g., t-tests, analysis of variance), Levene’s test was used to assess the equality of variances across comparison groups. In contrast, the cytokine correlation analyses were exploratory, aimed at identifying potential associations between inflammatory markers and glymphatic function for hypothesis generation in future studies.
Following the guidelines established by Hayes, we used SPSS PROCESS macro version 3.5 to examine the hypotheses regarding the mediating role of glymphatic function between inflammation and brain volume [54,55]. In our mediation models, we adjusted for demographic variables, including age, sex, years of education, presence or absence of medication, and ICV. Using Model 4, we investigated whether inflammation affects choroid plexus volume by mediation of glymphatic function. Furthermore, using Model 6, we explored the impact of chronic disease on glymphatic function. All statistical analyses were performed using SPSS software version 21.0 (IBM Corp.). The number of bootstrap was 5,000 for all mediation analyses.
RESULTS
Demographics, glymphatic function, and related brain volume between MDD and HC
No significant differences in age, sex, years of education, or relative ICV were observed between the MDD and HC groups. Patients with MDD exhibited significantly higher HDRS scores (mean, 16.71) than HCs (mean, 0.82; p<0.001), indicating greater depressive symptoms and sleep disturbance. There were also markedly higher anxiety and suicidal ideation levels in the MDD group: mean scores on the HARS and SSI were significantly elevated compared with HCs (both p<0.001). In the MDD group, 29.5% of the patients were not medicated. Among those receiving treatment, 26.1% were prescribed selective serotonin reuptake inhibitors, 38.6% used benzodiazepines, 20.5% received antipsychotics (mean olanzapine equivalent dose, 2.87 mg), and 13.1% received serotonin-norepinephrine reuptake inhibitors (Table 1 and full data in Supplementary Table 1). And there were no significant group differences by medication status in age, sex, years of education, ICV, total HDRS score, illness duration, or DTI-ALPS indices (right/left), whereas the medication-free group showed higher HDRS sleep scores (p=0.021) and the medicated group had higher CRP levels (p=0.038). When comparing participants with versus without available laboratory data, age, sex, education, ICV, and total HDRS did not differ significantly; however, the lab-available group had longer illness duration and higher HARS scores (both p<0.001), while SSI was higher in the lab-unavailable group (p<0.001) (Supplementary Tables 2 and 3).
Demographics of patients with MDD and HC
The right DTI-ALPS index decreased in the MDD group (1.574±0.207) versus HCs (1.645±0.168; p=1.476×10-3), and a similar reduction was observed in the left DTI-ALPS index (MDD: 1.531±0.186; HC: 1.600±0.158; p=6.822×10-4). The left inferior lateral ventricle was enlarged in patients with MDD (337.565 mm3±193.569 mm3) compared to HCs (282.196 mm3±137.025 mm3; p=2.606×10-3), and this difference remained significant after Bonferroni correction (Table 2 and Figure 2).
Differences in DTI-ALPS and glymphatic function-related brain structure between MDD and HC groups
Three-dimensional view of the choroid plexus. This image provides a 3D rendering of the choroid plexus (shown in sky blue), which is segmented and visualized to highlight its detailed structure. Segmentation was performed using the FreeSurfer software package, which calculates the volume of each labeled region. This image helps to understand the spatial distribution and volumetric data of the choroid plexus within the ventricular system of the brain. A, anterior; P, posterior; L, left; R, right.
Correlation between glymphatic function, depressive symptoms, and brain volume in the MDD group
In the analysis of the correlations between DTI-ALPS indices, depressive symptoms, illness duration, and brain volumes within the MDD group, no significant correlations were found between the total HDRS scores and the DTI-ALPS index on the right or left side. However, a significant negative correlation was observed between illness duration and the right and left DTI-ALPS indices, but only the correlation with the right DTI-ALPS survived the Bonferroni correction (ALPS_R: r=-0.265, p<0.001; ALPS_L: r=-0.201, p=0.008).
The DTI-ALPS indices showed significant negative correlations with ventricular volume in patients with MDD. Specifically, the right DTI-ALPS index (ALPS_R) showed significant negative correlations with the left lateral ventricle (r=-0.274, p<0.001), right lateral ventricle (r=-0.314, p<0.001), and right inferior lateral ventricle (r=-0.263, p<0.001). Similarly, the left DTI-ALPS index (ALPS_L) was negatively correlated with the left lateral ventricle (r=-0.232, p=0.002), right lateral ventricle (r=-0.222, p=0.004), and right inferior lateral ventricle (r=-0.211, p=0.006). In addition to ventricular volumes, significant negative correlations were observed between the DTI-ALPS indices and choroid plexus volume. The right DTI-ALPS index showed negative correlations with the left choroid plexus (r=-0.195, p=0.011) and right choroid plexus (r=-0.208, p=0.006). The left DTI-ALPS index was also negatively correlated with the left choroid plexus (r=-0.241, p=0.002) and right choroid plexus (r=-0.177, p=0.021). Although the correlation with third ventricle volume did not reach statistical significance after Bonferroni correction, there was a trend toward a negative correlation with the right DTI-ALPS index (r=-0.149, p=0.051) and left DTI-ALPS index (r=-0.181, p=0.018).
Conversely, significant positive correlations were found between DTI-ALPS indices and thalamic volumes. The right DTI-ALPS index was positively correlated with the left (r=0.241, p=0.002) and right thalamus (r=0.257, p<0.001). Similarly, the left DTI-ALPS index showed positive correlations with the right thalamus (r=0.227, p=0.003) (Figure 3 and Supplementary Table 4 for MDD, Supplementary Table 5 for HC).
Correlation matrix heatmap of depressive symptoms and glymphatic function-related features in the MDD group. **p<0.00294 (Bonferroni 0.05/17=0.00294); ***p<0.001. MDD, major depressive disorder; R, right; L, left; HDRS, Hamilton Depression Rating Scale; ALPS, analysis along the perivascular space.
Inflammatory cytokines between the MDD and HC groups
When comparing inflammatory cytokine levels between the MDD and HC groups, IL-6 and IL-8 levels were significantly elevated in the MDD group, while no significant differences in CRP levels were observed between the two groups (Table 3 and full data in Supplementary Table 6 for MDD, Supplementary Table 7 for HC).
Inflammatory cytokines levels between MDD and HC groups
Correlation between inflammatory cytokines and glymphatic function in the MDD group
After controlling for age, sex, educational level, medication status, and ICV, a significant correlation was found between IL-6 and IL-8 levels. However, these inflammatory markers were not significantly associated with depression severity, sleep symptoms, illness duration, or brain volume. In contrast, the CRP level showed a significant negative correlation with both bilateral DTI-ALPS indices and right choroid plexus volume, and these associations remained statistically significant even after Bonferroni correction (Table 4 and Supplementary Table 6). However, in HCs (n=48), only left DTI-ALPS showed a negative association with CRP (r=-0.341, p<0.05), with an effect size attenuated relative to MDD. Right DTI-ALPS did not correlate with CRP or any other inflammatory cytokines (for CRP r=-0.277, p>0.05; all other cytokines p>0.05) (Supplementary Table 7).
Correlation between inflammatory cytokines and glymphatic function in the MDD group (N=61)
The illness duration of patients with MDD showed a wide distribution, with a positively skewed pattern (Supplementary Figure 1). To verify the stability of this association, a sensitivity analysis was performed after excluding two extreme outliers of illness duration. And the positive correlation between illness duration and CRP remained significant (r=0.368, p<0.05), confirming the robustness of the relationship between chronicity and systemic inflammation (Supplementary Table 8).
Mediation analysis between CRP, DTI-ALPS, and choroid plexus volume in the MDD group
Mediation analysis highlights that systemic inflammation, measured by CRP levels, significantly affects choroid plexus volume in patients with MDD and, importantly, part of this effect appears to be mediated by reduced DTI-ALPS indices, suggesting a role for glymphatic dysfunction. Specifically, the total effect of CRP levels on choroid plexus volume was significant (effect=-88.305, standard error [SE]=24.915, lower-level confidence interval [LLCI]=-138.940, upper-level confidence interval [ULCI]=-37.671), as was the direct effect (effect=-126.951, SE=26.527, LLCI=-180.921, ULCI=-72.981), suggesting that higher levels of CRP are associated with lower choroid plexus volume. The indirect effect of CRP levels on choroid plexus volume through the DTI-ALPS indices was also significant (effect=38.646, SE=20.006, LLCI=6.613, ULCI=85.209), indicating that reduced glymphatic function (represented by lower DTI-ALPS values) partially mediates the relationship between elevated inflammation and decreased choroid plexus volume. Conversely, the indirect effect of CRP on choroid plexus volume through the duration of illness was not statistically significant (effect=4.388, SE=16.652, LLCI=-15.791, ULCI=49.384), nor was the indirect effect involving both the duration of illness and DTI-ALPS (effect=9.420, SE=8.306, LLCI=-2.776, ULCI=27.894) (Table 5 and Figure 4).
Mediation analysis between CRP, ALPS, and choroid plexus volume in MDD (N=41)
Mediation effect of ALPS on CRP and choroid plexus. A: Right ALPS index mediates the association between serum CRP and right choroid plexus volume. B: Reciprocal model with CRP tested as the mediator between right ALPS and right choroid plexus volume. C: Serial mediation model incorporating illness duration, CRP, right ALPS, and right choroid plexus volume. Red solid lines indicate statistically significant negative associations; blue solid lines indicate statistically significant positive associations; blue dashed lines indicate non-significant paths. Covariates age, sex, years of education, ICV, taking medication. *p<0.05; **p<0.00294 (Bonferroni 0.05/17=0.00294); ***p<0.001. CRP, C-reactive protein; ALPS, analysis along the perivascular space; ICV, intracranial volume.
DISCUSSION
In this study, we investigated glymphatic dysfunction and related brain structural changes in patients with MDD, with the aim of exploring its association with inflammatory cytokines. Our findings revealed that patients with MDD exhibited significantly reduced glymphatic function, as indirectly estimated by decreased DTI-ALPS indices on the right and left sides, compared to HCs. Additionally, patients with MDD showed significant volumetric differences in the brain regions associated with glymphatic function, including the enlarged left inferior lateral ventricle. In particular, although DTI-ALPS scores were not associated with the severity of depressive symptoms or specific symptoms, such as sleep disturbances and appetite changes, after Bonferroni correction, the right DTI-ALPS index showed a statistically significant association with the duration of depression. Furthermore, decreased glymphatic function was correlated with large ventricular and choroid plexus volumes, and the right DTI-ALPS index showed significant negative correlations with the bilateral lateral ventricle and right inferior lateral ventricle, while the left DTI-ALPS index showed similar negative correlations. Conversely, both DTI-ALPS indices were positively correlated with thalamic volume, suggesting that better glymphatic function is associated with increased thalamic volume in patients with MDD. Furthermore, the inflammatory marker CRP showed significant negative correlations with bilateral DTI-ALPS indices and right choroid plexus volume. The mediation analysis highlighted that higher levels of CRP were associated with reduced DTI-ALPS indices, which, in turn, were associated with a reduced choroid plexus volume. These findings highlighted the impact of inflammation on glymphatic function and its subsequent effects on brain structures involved in CSF dynamics. The absence of significant mediating effects of the duration of illness suggests that the chronicity of the depressive illness itself may not play a direct role in linking inflammation with changes in choroid plexus volume. Instead, the inflammatory response exerts its influence primarily by impairing glymphatic clearance, leading to structural changes.
Our findings are consistent with those of previous studies that reported a decrease in DTI-ALPS indices in patients with depression, suggesting potentially impaired glymphatic function in patients with MDD [56]. Although previous research has shown a proportional relationship between DTI-ALPS and the severity of depressive symptoms, our study did not confirm such an association. In ischemic stroke and cerebral small vessel disease, lower DTI-ALPS values persistently correlate with white-matter lesion burden even after acute recovery, indicating an enduring clearance-impairment trait [57]. Similarly, in MDD, DTI-ALPS was reduced relative to controls yet did not track concurrent HDRS scores, and in the blood-consenting subgroup DTI-ALPS mediated the link between CRP and choroid-plexus volume—findings consistent with a longer-timescale mechanism. Taken together, these observations support DTI-ALPS as a potential trait marker reflecting glymphatic dysfunction in MDD rather than a state marker of momentary symptom severity.
Instead, we found that DTI-ALPS data showed a negative correlation with the duration of illness. This result has not previously been reported in patients with MDD; however, our findings are consistent with those of patients with bipolar disorder, in which the DTI-ALPS index was associated with the duration of illness [58]. In addition, patients with MDD who responded to treatment had significantly smaller volumes of the lateral ventricles and choroid plexus than non-responders, suggesting that ventricular enlargement may be related to treatment resistance and disease severity [18]. Furthermore, one of the interesting findings was that the bilateral decrease in the DTI-ALPS index was associated with an increase in the volume of the thalamus on each side. Numerous neuroimaging studies have reported that patients with MDD often exhibit reduced thalamic volume compared to HCs [59,60]. This thalamic atrophy has been associated with symptoms such as impaired cognitive function, disrupted emotional processing, and difficulties in sensory perception [61-63]. An association between the bilateral analysis DTI-ALPS index and increased thalamic volumes on both sides suggests that reduced perivascular water transport, as estimated by DTI-ALPS, might correlate with decreased function of thalamic structures in patients with MDD, potentially mitigating some cognitive and emotional symptoms associated with thalamic atrophy. Integrating these findings, our study suggests that glymphatic dysfunction is associated with factors that prolong the duration of illness or reduce treatment response in MDD, while the mediation model including illness duration did not indicate a significant indirect effect of DTI-ALPS on right choroid plexus volume. Specifically, this dysfunction, manifested as ventricular enlargement and decreased thalamic volume, can lead to a reduced response to treatment, which may contribute to the chronicity of depressive symptoms. As the glymphatic system plays a crucial role in the removal of metabolic waste from the brain, its dysfunction can lead to the accumulation of neurotoxic substances, exacerbating neuronal damage, and promoting ventricular enlargement. Consequently, glymphatic impairment can not only contribute to the prolonged duration of illness, but can also decrease the efficacy of treatments, highlighting its role in the pathophysiology of MDD and its potential as a therapeutic target. The glymphatic system is most active during sleep, especially during slow-wave sleep [64]. Previous studies in animal models and patients with Alzheimer’s disease have reported that glymphatic dysfunction is associated with decreased sleep quality [65,66]. Contrary to these findings, our study of patients with MDD did not reveal a correlation between glymphatic function and the severity of sleep disturbances. However, since our study did not directly assess specific indicators of sleep quality, we cannot definitively conclude an absence of association between glymphatic function and sleep quality in patients with depression. Further studies are required to determine the association between insomnia symptoms and glymphatic function in patients with depression.
The most important finding of this study was the analysis of inflammatory cytokines, which revealed that inflammation affects brain volume, mediated by glymphatic function. Within the depression group, CRP levels showed a significant negative correlation with the DTI-ALPS index and right choroid plexus volume, and mediation analysis was applied. In HCs, however, only left DTI-ALPS showed a weaker negative association with CRP, whereas right DTI-ALPS did not correlate with any inflammatory cytokines. This pattern suggests that the bilateral CRP–DTI-ALPS coupling observed in MDD is amplified or disorder-specific relative to controls. Whether glymphatic dysfunction precedes neuroinflammation or vice versa in depression has not yet been determined. Therefore, in this study, we conducted mediation analyses for both models: one in which neuroinflammation precedes glymphatic dysfunction and another in which glymphatic dysfunction precedes neuroinflammation. In both cases, we confirmed an indirect effect on brain volume. To date, chronic neuroinflammatory states are known to activate glial cells such as microglia and astrocytes, promoting the secretion of various cytokines [67,68]. Since functional changes in astrocytes can affect the function of the glymphatic system, astrocyte alterations due to neuroinflammation are more likely to induce glymphatic dysfunction [8]. Therefore, it is more probable that Model 1 (CRP→DTI-ALPS→choroid plexus volume model) reflects the changes that occur in depression. Furthermore, the choroid plexus, which showed an association with CRP levels in our study, has been consistently reported in previous research to show an increase in the volume in depression [51]. Our group-level comparisons confirmed this pattern, with MDD patients showing larger choroid plexus volumes than HCs. However, within the MDD group, higher CRP levels were associated with reduced right choroid plexus volume, which appears to contrast with prior findings. This discrepancy may reflect the substantial illness chronicity of our sample: under prolonged systemic inflammation, the choroid plexus may initially enlarge as a compensatory response but subsequently undergo atrophy due to cumulative inflammatory damage. Unlike previous studies that did not find a correlation between choroid plexus volume and CRP levels, our study demonstrated a strong association, possibly because we adjusted for factors such as medication use and applied Bonferroni correction, whereas an acknowledged limitation of previous studies is the lack of clearly defined covariates. In addition, inflammatory markers other than CRP did not show significant correlations in our results. CRP is a downstream acute-phase protein synthesized in the liver in response to upstream cytokines, particularly IL-6, reflecting the integrated output of multiple inflammatory pathways [69]. With a biological half-life of approximately 19 hours and minimal diurnal variation, CRP captures sustained inflammatory burden more reliably than individual cytokines such as IL-6 and IL-8, which are inherently phasic and reflect acute inflammatory responses [70,71]. Because the DTI-ALPS index is a trait-level microstructural measure that likely reflects cumulative changes in perivascular water transport, it is biologically plausible that CRP demonstrates stronger coupling with DTI-ALPS and choroid plexus volume than more transient cytokines [72-74]. Consistent with this, in our MDD cohort characterized by substantial illness chronicity, CRP but not IL-6 or TNF-α correlated with DTI-ALPS, supporting the view that sustained inflammatory tone is most relevant to glymphatic impairment. Therefore, more extensive research is needed to determine whether consistent findings regarding the associations between inflammatory cytokines, such as CRP levels, glymphatic function, and the choroid plexus, will continue.
The first strength of our study is that we recruited a relatively large number of patients with depression and HC, allowing us to evaluate and compare glymphatic function using MRI. Second, although it involved a subgroup, we measured various levels of cytokines that represent neuroinflammation (e.g., CRP, IL-1, IL-6, and TNF-α) and confirmed changes in glymphatic function, as well as in the volumes of associated brain regions. Third, rather than merely identifying simple correlations, we applied a mediation analysis to verify the associations among neuroinflammation, glymphatic function, and brain structural changes, which had been theoretically proposed in previous hypotheses, in actual clinical patients. Finally, even within the patient group analyses, we adjusted for various covariates, including medication use, sex, age, and years of education. We also confirmed our results under rigorous statistical criteria by applying the Bonferroni correction to adjust for statistical errors due to multiple comparisons in the brain structure analyses.
Despite these strengths, our study has several limitations. The most significant limitation is that, although the DTI-ALPS index provides a noninvasive proxy for perivascular water transport, it should be regarded as an indirect surrogate of glymphatic function [19,52]. The metric is derived from directional diffusivity and can be influenced by crossing-fiber geometry, local tissue anisotropy, and scanner variability. Moreover, validation of DTI-ALPS in psychiatric populations remains limited [20,58]. Therefore, interpretations in MDD should remain cautious. Future multimodal studies combining DTI-ALPS with quantitative CSF flow or tracer-based MRI will be essential to confirm its physiological specificity. A second important limitation is insufficient control for potential confounders of inflammation. Although we adjusted for available covariates including medication status, important factors such as body mass index, smoking status, fasting duration, and comorbid metabolic conditions were not accounted for. Furthermore, medication status was treated as a binary variable (medicated vs. medication-free), which does not capture the heterogeneous effects of different psychotropic medication classes on inflammation and brain structure. Although no significant differences in ALPS indices were observed between medicated and medication-free patients, stratified analyses by individual medication class were not feasible due to insufficient subgroup sizes (Supplementary Table 2). Future studies should incorporate comprehensive covariate assessment, standardized protocols, and detailed medication data to enable class-specific stratified analyses. Until such designs are implemented, the generalizability of the current findings should be interpreted with appropriate caution. The third limitation was the relatively small number of participants in the cytokine analysis and mediation analyses. Cytokine data were available only for a subgroup who consented to blood sampling (n=68 MDD, n=54 HC), and the final mediation analysis included only 41 MDD patients with complete data, which limits statistical power. To mitigate this, we employed bootstrap resampling with 5,000 iterations for robust confidence interval estimation. Additionally, this subsample exhibited longer illness duration and higher anxiety severity than the full sample, raising the possibility of selection bias, although key demographic and clinical variables (age, sex, education, ICV, and HDRS) did not differ significantly (Supplementary Tables 2 and 3). The longer illness duration may have enriched for chronic inflammatory profiles, potentially strengthening CRP–ALPS associations while limiting generalizability. Despite these constraints, the observed large effect sizes support the robustness of our findings; however, replication in larger samples with complete cytokine data is warranted. The fourth limitation is that in this study the measurements were limited to the lateral choroid plexus, which introduces several limitations. By excluding the choroid plexus in the third and fourth ventricles, this study may not have fully captured the variations and pathological changes specific to these regions. This restriction could lead to an incomplete understanding of the physiological functions of the choroid plexus and its role in neurological conditions. Consequently, the findings may not accurately reflect the overall dynamics and contributions of the choroid plexus system, potentially skewing interpretations related to brain health and disease. Future studies should consider incorporating all the choroid plexus regions for a more comprehensive analysis. Fifth, regarding statistical considerations, our multiple-comparison control targeted a pre-specified family of 17 glymphatic-relevant structural outcomes using Bonferroni correction, which aligns with common practice but may be less conservative than pooling all variables into a single family. Sensitivity false discovery rate analyses across a broader test set yielded qualitatively similar conclusions, although replication in larger samples remains essential. For the cytokine correlation analyses, we did not apply Bonferroni correction because these analyses were exploratory in nature, and CRP was designated a priori as the primary inflammatory marker based on its established relevance to systemic inflammation. Additionally, given the relatively small cytokine subsample size (n=68 for MDD, n=54 for HC), stringent correction would substantially increase the risk of Type II error. We therefore reported uncorrected p-values while emphasizing the large effect sizes observed for CRP associations to support the robustness of these findings. Finally, our examination of the glymphatic sleep interface was constrained to the sleep disturbance subscale of the HDRS. We did not administer standardized sleep questionnaires (e.g., Pittsburgh Sleep Quality Index) nor collect objective physiological indices such as actigraphy or overnight polysomnography. Future studies that incorporate both subjective and objective sleep metrics will be essential to delineate more precisely how specific aspects of sleep architecture and quality modulate glymphatic clearance in MDD.
This study identified a significant association between glymphatic dysfunction, neuroinflammation, and structural brain changes in patients with MDD. Patients with MDD exhibit reduced glymphatic function, as proven by decreased DTI-ALPS indices and volumetric alterations in the brain regions involved in CSF dynamics, such as enlarged ventricles and altered thalamic volume. In particular, CRP levels showed significant negative correlations with bilateral DTI-ALPS indices and right choroid plexus volume. Mediation analysis suggested that glymphatic function, as indirectly assessed by DTI-ALPS, mediates the relationship between neuroinflammation and brain volume. Although this mediating effect is believed to be associated with prolonged duration of illness or reduced treatment responsiveness in depression, these results should be regarded as preliminary and require replication in larger independent samples and longitudinal designs.
Supplementary Materials
The Supplement is available with this article at https://doi.org/10.30773/pi.2025.0456.
Differences in ALPS and glymphatic function related brain structure between MDD and HC (full table)
Demographics of MDD with medication and MDD without medication
Demographics of MDD with lab and MDD without lab
Correlation depressive symptoms and glymphatic function related features in MDD group
Correlation depressive symptoms and glymphatic function related features in HC group (N=178)
Correlation between inflammatory cytokines and glymphatic function in MDD (N=61, full table)
Correlation between inflammatory cytokines and glymphatic function in HC (N=48)
Sensitivity correlation analysis between inflammatory cytokines and illness duration excluding two extreme values
Histogram of illness duration in major depressive disorder.
Notes
Availability of Data and Material
The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.
Conflicts of Interest
Kyu-Man Han, a contributing editor of the Psychiatry Investigation, was not involved in the editorial evaluation or decision to publish this article. All remaining authors have declared no conflicts of interest.
Author Contributions
Conceptualization: Daun Shin, Kyu-Man Han, Youbin Kang, Dorothee P. Auer, Woo Suk Tae. Data curation: Kyu-Man Han, Youbin Kang, Dorothee P. Auer, Woo Suk Tae, Byung-Joo Ham. Formal analysis: Daun Shin, Youbin Kang, Woo Suk Tae. Funding acquisition: Woo Suk Tae, Byung-Joo Ham. Investigation: Daun Shin, Kyu-Man Han, Dorothee P. Auer, Woo Suk Tae, Byung-Joo Ham. Methodology: Daun Shin, Kyu-Man Han, Youbin Kang, Woo Suk Tae, Byung-Joo Ham. Supervision: Kyu-Man Han, Byung-Joo Ham. Writing—original draft: Daun Shin. Writing—review & editing: all authors.
Funding Statement
This work was supported by National Research Foundation of Korea (NRF) grants funded by the Korean government (MSIT) (NRF-2020M3E5D9080792; NRF-2022R1A2C2093009; Author Byung-Joo Ham). This work was supported by a grant from the National Research Foundation of Korea (NRF) funded by the Korean government (MSIT) (Grant No. RS-2023-00241730; Author Woo-Suk Tae).
Acknowledgments
We sincerely thank all the participants who voluntarily took part in this study. We also extend our gratitude to the staff and collaborators who assisted in the research.