Peripheral inflammation impairs glymphatic function, contributing to neurodegeneration in Alzheimer’s disease

Summary

Alzheimer’s disease (AD) is associated with systemic immune alterations and glymphatic dysfunction, both of which are linked to brain structural and network changes that contribute to cognitive decline. In 570 participants with AD, mild cognitive impairment, or normal cognition, we combined peripheral immune profiling with multimodal magnetic resonance imaging (MRI) to evaluate glymphatic function, brain structure, and network organization. AD was characterized by reduced analysis along the perivascular space index, enlarged choroid plexus (CP) volume, increased white matter free water, reduced lymphocyte count, and elevated neutrophil-to-lymphocyte ratio (NLR). Immune indices, including NLR, platelet-to-lymphocyte ratio, systemic immune-inflammation index, and lymphocyte count, were associated with cognition and glymphatic-related MRI measures. Mediation analyses indicated that NLR influenced cognition indirectly through CP volume and downstream brain structural and network features. These findings link peripheral immune imbalance to cognitive decline through glymphatic and brain network alterations, supporting biomarker development and mechanism-guided therapeutic strategies.

Graphical abstract

Highlights

• •Peripheral immune dysfunction in AD links inflammation to neurodegeneration
• •Glymphatic dysfunction emerges as key imaging biomarkers in AD
• •Immune markers affect cognition via glymphatic, structural, and network pathways
• •Findings indicate biomarkers and therapies for the immune-glymphatic-brain axis

Health sciences; Neurology; Clinical stage

Introduction

With global population aging, the burden of Alzheimer’s disease (AD) continues to increase, imposing substantial pressure on patients, families, and society. The World Health Organization estimates that approximately 55 million people are living with dementia globally,¹ with nearly 10 million new cases occurring each year,² and that AD accounts for 60%–70% of dementia cases.³ The characteristic pathological features of AD include extracellular β-amyloid (Aβ) plaque deposition, intraneuronal neurofibrillary tangles formed by aggregated tau protein,⁴ synaptic and neuronal loss,⁵ and neuroinflammation,⁶ ultimately leading to brain atrophy and progressive cognitive impairment. However, its pathogenesis remains incompletely understood. Although multiple studies have focused on classical mechanisms such as the Aβ cascade hypothesis and abnormal tau phosphorylation, systematic insight into the associations and interaction mechanisms among peripheral immune alterations, glymphatic system dysfunction, brain structural and functional abnormalities, and cognitive decline across the AD spectrum is still lacking.

AD is increasingly recognized as a systemic disease involving dynamic interactions between peripheral and central immune responses.⁷ Growing evidence suggests that peripheral immunity dysregulation may exacerbate neurodegeneration in AD by altering the composition and function of circulating immune cells and promoting their infiltration into the brain,⁸ thereby playing a pivotal role in disease progression. An informative approach for evaluating peripheral immunity is the analysis of white blood cell-derived indices in peripheral blood.⁹ Measurement indicators such as the neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), platelet-to-lymphocyte ratio (PLR), and systemic immune-inflammation index (SII) provide quantifiable markers of immune activation and reflect the balance between innate and adaptive immunity.¹⁰ However, the specific mechanisms by which peripheral immune alterations influence brain structural atrophy, functional network disturbances, and cognitive decline in AD remain unclear.

The glymphatic system, a unique waste clearance pathway in the central nervous system (CNS), was first systematically described by Iliff et al. in 2012.¹¹ This system plays a crucial role in maintaining brain homeostasis,¹² by facilitating the efficient clearance of neurotoxic substances such as Aβ and tau protein through a dynamic perivascular exchange between cerebrospinal fluid (CSF) and interstitial fluid (ISF).¹³ This exchange is primarily mediated by densely expressed polarized aquaporin-4 (AQP4) water channels located on astrocyte endfeet and is facilitated by the perivascular pathway.¹⁴ Increasing attention has been directed toward the contribution of glymphatic dysfunction to AD pathogenesis, with accumulating evidence demonstrating impaired clearance efficiency and abnormal parenchymal accumulation of Aβ in AD.^15,16 Advances in noninvasive magnetic resonance imaging (MRI) have enabled indirect assessment of glymphatic function. For instance, diffusion tensor imaging (DTI)-based models can quantify the free water fraction in white matter (FWf-WM) and analyze the perivascular space (ALPS) index,¹⁷ while three-dimensional T1-weighted imaging (3D-T1WI) combined with automated segmentation allows for accurate extraction of choroid plexus (CP) volume.¹⁸ Collectively, these metrics (FWf-WM, ALPS index, and CP volume) serve as valuable indirect biomarkers for evaluating glymphatic function. Substantial evidence also indicates that progressive atrophy of key brain structures and disruption of functional networks occur in AD, both of which closely correlate with cognitive decline and contribute to disease progression.^19,20 These observations suggest that concurrent alterations in glymphatic function, brain structure, and brain networks may collectively underlie cognitive dysfunction in AD. Notably, peripheral immune dysregulation (e.g., altered immune cell migration) may further exacerbate AQP4 dysfunction and impair glymphatic clearance.²¹ Together, these findings support a multidimensional pathophysiological model of AD that involves dynamic interactions among peripheral immune disturbances, glymphatic dysfunction, brain atrophy, aberrant brain network activity, and cognitive decline.

In summary, the pathogenesis of AD extends beyond the classical amyloid and tau pathways, involving a complex interplay of systems. Growing evidence suggests that dysregulation of the peripheral immune system may impair the brain’s glymphatic clearance, a deficit thought to facilitate the accumulation of pathogenic proteins like Aβ and p-tau and sustain chronic neuroinflammation. This cascade is believed to drive progressive brain atrophy, functional network disruption, and the characteristic cognitive decline. However, it remains unclear whether peripheral immune dysregulation directly impairs glymphatic function and, if so, whether this impairment contributes to cognitive decline by mediating brain atrophy and network disruption. To address these questions, this study employed a multimodal approach that integrates MRI, peripheral immune profiling, and neuropsychological assessment. The aim of this study was to elucidate the “peripheral immunity-glymphatic system-brain structure/network-cognition” cascade pathway, thereby providing a scientific foundation for deeper insights into the pathophysiology of AD and the identification of precise therapeutic targets. Figure 1 presents the schematic overview of the study design.

Figure 1.

Overview of the study

Abbreviations: ROI, region of interest; ALPS, analysis along the perivascular space; FWf-WM, free water fraction in white matter; CP, choroid plexus; WM, white matter; FCS, functional connectivity strength; RSNs, resting-state networks; AD, Alzheimer’s disease; MCI, mild cognitive impairment; NC, normal control; LY, lymphocyte; NLR, neutrophil-to-lymphocyte ratio; MMSE, Mini-Mental State Examination.

Results

Cohort characteristics

Table 1 summarizes the clinical and demographic characteristics of the participants. Compared with the mild cognitive impairment (MCI) and normal control (NC) groups, patients in the AD group exhibited significantly older age, lower educational level, decreased cognitive performance, and a higher proportion of APOE4 allele carriers. Furthermore, with respect to glymphatic system function, the results indicated significant inter-group differences in three evaluation indicators, ALPS, FWf-WM, and CP volume, among the three groups, showing a pattern consistent with progressive impairment across the AD spectrum. In terms of peripheral immunity, only lymphocyte (LY) count (AD vs. NC) and NLR (AD vs. NC and MCI vs. NC) showed significant inter-group differences.

Table 1.

Characteristics of overall participants

Characteristics	AD	MCI	NC	p value	F or χ2 value
Demographic data	n= 154	n= 244	n= 172
Age, years	67 (61–75)	63 (58–68)	59 (53.25–65)	<0.001	70.07
Gender, female	99 (64.3%)	155 (63.5%)	112 (65.1%)	0.946	0.11
Education, years	10 (9–15)	12 (9–15)	15 (12–15)	<0.001	50.82
Cognitive assessments	n= 154	n= 244	n= 172
MMSE	19 (13–22)	27 (26–29)	29 (28–30)	<0.001	372.10
MoCA	13 (8–16)	23 (20–25)	27 (26–28)	<0.001	394.55
MES	42.5 (26–59)	84 (77–91)	92 (89–96)	<0.001	371.29
MES-M	18 (10–24)	40 (36–45)	46 (42–48)	<0.001	334.53
MES-E	24 (14–32.25)	44 (40–48)	48.5 (45–50)	<0.001	303.58
ADL	28 (25–35)	21 (20–22)	20 (20–20)	<0.001	362.59
CDR	2 (1–2)	0.5 (0.5–0.5)	0 (0–0)	<0.001	532.54
NPI	2 (0–8)	0 (0–0.75)	0 (0–0)	<0.001	100.08
APOE4	57 (n = 106, 53.8%)	35 (n = 112, 31.3%)	7 (n = 34, 20.6%)	<0.001	17.34
Glymphatic system	n= 97	n= 124	n= 47
ALPS	1.31 ± 0.15	1.42 ± 0.15	1.47 ± 0.20	<0.001	21.78
FWf-WM	0.24 (0.21–0.26)	0.20 (0.18–0.21)	0.18 (0.17–0.20)	<0.001	89.55
CP volume	1.19 × 10−3 ± 3.04 × 10−4	1.01 × 10−3 ± 2.93 × 10−4	8.41 ± 1.95 × 10−4	<0.001	33.83
Peripheral immunity	n= 114	n= 152	n= 113
NE (×109/L)	3.54 (2.88–4.27)	3.60 (2.97–4.32)	3.41 (2.71–4.17)	0.291	2.47
LY (×109/L)	1.79 (1.47–2.17)	1.91 (1.51–2.31)	2.00 (1.64–2.51)	0.023	7.51
MO (×109/L)	0.40 (0.30–0.47)	0.39 (0.31–0.50)	0.38 (0.31–0.49)	0.715	0.67
NLR	1.98 (1.50–2.66)	1.85 (1.41–2.51)	1.75 (1.23–2.33)	0.021	7.73
LMR	4.43 (3.59–6.66)	4.64 (3.82–6.00)	5.19 (4.09–6.62)	0.056	5.75
PLR	120.97 (93.08–161.50)	115.18 (90.55–144.13)	107.07 (87.88–139.83)	0.078	5.11
SII	435.05 (293.48–609.33)	407.69 (288.39–531.45)	365.34 (269.70–506.74)	0.086	4.91

Data are presented as mean ± SD, median (interquartile ranges [IQRs]), or n (%), as appropriate. Comparisons among the three groups were performed using one-way ANOVA, the Kruskal-Wallis test, or the χ² test, as appropriate. Exact p values are reported where applicable; p values smaller than 0.001 are reported as p < 0.001. Abbreviations: AD, Alzheimer’s disease; MCI, mild cognitive impairment; NC, normal control; MoCA, Montreal Cognitive Assessment; MES, Memory and Executive Screening; MES-M, MES-Memory; MES-E, MES-Execution; ADL, Activity of Daily Living Scale; CDR, Clinical Dementia Rating; NPI, Neuropsychiatric Inventory; ALPS, analysis along the perivascular space; FWf-WM, free water fraction in white matter; CP, choroid plexus; NE, neutrophils; LY, lymphocytes; MO, monocytes; NLR, neutrophil-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio; PLR, platelet-to-lymphocyte ratio; SII, systemic immune-inflammation index (neutrophils × platelets/lymphocytes). Bold italic entries indicate category headings and sample sizes for each group.

Functional changes of the glymphatic system in the spectrum of AD

To evaluate functional alterations in the glymphatic system at different stages of the AD spectrum, this study assessed the glymphatic system in the AD, MCI, and NC groups. The results revealed that the ALPS index was significantly reduced in the AD group (1.31 ± 0.15) compared with the NC (1.47 ± 0.20, p < 0.001) and MCI (1.42 ± 0.15, p < 0.001) groups, while FWf-WM was significantly increased in the AD group (0.24, interquartile range [IQR]: 0.21–0.26) relative to the NC (0.18, IQR: 0.17–0.20, p < 0.001) and MCI (0.20, IQR: 0.18–0.21, p < 0.001) groups. CP volume also showed significant inter-group differences (NC: 8.41 ± 1.95 × 10⁻⁴, MCI: 1.01 × 10⁻³ ± 2.93 × 10⁻⁴, AD: 1.19 × 10⁻³ ± 3.04 × 10⁻⁴), progressively increasing with disease severity (all pairwise comparisons p < 0.001, Bonferroni post hoc test), as shown in Table 1 and Figures 2A–2C. Subsequently, correlation analyses were conducted between these indicators and cognitive scales to further examine the relationship between glymphatic system function and cognition. The results, as shown in Figure 2F, indicate that glymphatic system function was significantly associated with global cognition, memory, and executive function after adjusting for age, sex, and years of education.

Figure 2.

The functional alterations of the glymphatic system and peripheral immune changes in the AD spectrum, encompassing inter-group differences and their correlation with cognition

(A–C) Box-and-whisker plots showing individual data points for the ALPS index, FWf-WM, and CP volume, respectively, among the AD, MCI, and NC groups.

(D and E) Box-and-whisker plots showing individual data points for LY counts and NLR, respectively, among the AD, MCI, and NC groups.

(F) Forest plot showing the adjusted associations of the ALPS index, FWf-WM, and CP volume with cognitive scale scores, including MMSE, MES-M, and MES-E. The models were adjusted for age, gender, and years of education. In all boxplots, center lines indicate the median, boxes indicate the interquartile range (IQR; 25th–75th percentiles), whiskers extend to the most extreme data points within 1.5× IQR from the lower and upper quartiles, and each diamond represents an individual sample. In the forest plots, squares indicate β coefficients, and horizontal lines indicate 95% confidence intervals (CIs). The vertical reference line indicates β = 0, representing no association. Group differences were assessed using ANOVA with Bonferroni post hoc tests or Kruskal-Wallis H tests with Bonferroni-corrected Dunn’s post hoc tests, as appropriate. Forest plots show adjusted β coefficients and 95% CIs from general linear regression models. Exact p values are shown where applicable; values < 0.001 are reported as p < 0.001. Abbreviations: AD, Alzheimer’s disease; MCI, mild cognitive impairment; NC, normal control; ALPS, analysis along the perivascular space; FWf-WM, free water fraction in white matter; CP, choroid plexus; LY, lymphocytes; NLR, neutrophil-to-lymphocyte ratio.

Peripheral immune alterations across the AD spectrum

Regarding peripheral immune function across the AD spectrum, LY count was significantly reduced in the AD group compared with the NC group (1.79 [1.47–2.17] vs. 2.00 [1.64–2.51], p = 0.018). NLR was significantly elevated in the AD group relative to the NC group (1.98 [1.50–2.66] vs. 1.75 [1.23–2.33], p = 0.022). All pairwise comparisons were performed using Dunn’s post hoc test with Bonferroni correction. After adjusting for age, sex, and education level, global cognition (Mini-Mental State Examination [MMSE]) was negatively correlated with NLR (β = −0.194, 95% confidence interval [CI]: −0.291 to −0.097), PLR (β = −0.150, 95% CI: −0.247 to −0.054), and SII (β = −0.166, 95% CI: −0.262 to −0.069). These associations were also observed for memory and executive function. LY count was significantly positively correlated with MMSE (β = 0.109, 95% CI: 0.012–0.206), as shown in Table 1 and Figures 2D and 2E.

Associations between peripheral immunity and glymphatic system function

Regarding whether peripheral immunity affects glymphatic system function across the AD spectrum, this study found that neutrophil (NE) count (β = 0.151, 95% CI: 0.019–0.283), NLR (β = 0.155, 95% CI: 0.045–0.265), PLR (β = 0.147, 95% CI: 0.011–0.284), and SII (β = 0.159, 95% CI: 0.039–0.279) were positively correlated with CP volume, indicating that a stronger peripheral innate immune response was associated with greater impairment in the glymphatic system. In contrast, LY count was negatively correlated with CP volume (β = −0.136, 95% CI: −0.264 to −0.007), suggesting that higher peripheral LY counts, a marker of adaptive immunity, were associated with relatively preserved glymphatic function. PLR (β = 0.184, 95% CI: 0.036–0.332) and SII (β = 0.172, 95% CI: 0.047–0.297) also show a positive correlation with FWf-WM, while LMR (β = −0.167, 95% CI: −0.317 to −0.017) was negatively correlated with FWf-WM. The results are shown in Figure 3.

Figure 3.

Associations between peripheral immunity and glymphatic function

Squares indicate β coefficients, and horizontal lines indicate 95% confidence intervals (CIs). The vertical reference line indicates β = 0, representing no association. β coefficients were estimated using general linear regression models adjusted for age, sex, and years of education. Exact p values are shown where applicable; values < 0.001 are reported as p < 0.001. Abbreviations: NE, neutrophils; LY, lymphocytes; MO, monocytes; NLR, neutrophil-to-lymphocyte ratio; LMR, lymphocyte-to-monocyte ratio; PLR, platelet-to-lymphocyte ratio; SII, systemic immune-inflammation index; MMSE, Mini-Mental State Examination; MES-M, Memory and Executive Screening-Memory; MES-E, Memory and Executive Screening-Execution; ALPS, analysis along the perivascular space; FWf-WM, free water fraction in white matter; CP, choroid plexus.

Changes in brain structure volume and brain network function in the spectrum of AD

Measurements of brain structural volumes and brain network function were obtained, and the results are presented in Table S1. These findings indicate that, compared with the NC group, key brain structures in the MCI and AD groups exhibited progressive atrophy, and brain network function showed a stepwise decline as the disease progressed.

Simple and multiple mediation analyses between peripheral immunity and cognition

The aforementioned results indicate significant associations between peripheral immunity and cognitive function, as well as between the glymphatic system and cognition. Previous studies have confirmed that changes in brain structure and function are associated with cognitive impairment. Therefore, we further analyzed peripheral immune indicators to determine whether they affect cognitive function by mediating pathways involving the glymphatic system, brain structure, and brain function. We employed simple and multiple mediation models to conduct three mediation path analyses: (1) peripheral immunity → glymphatic system function → global cognition, (2) peripheral immunity → brain structure and brain network function → global cognition, and (3) peripheral immunity → glymphatic system function → brain structure and brain network function → global cognition.

First, the results of simple mediation analyses indicated that peripheral immune indicators, including NLR and SII, indirectly affect overall cognitive function through CP volume, while also exerting direct effects on cognition. PLR, in contrast, is fully mediated by CP volume and subsequently influenced cognitive function. PLR and SII also indirectly affected cognitive function through FWf-WM. When brain structural volume was used as a mediating variable, NLR and SII indirectly affected MMSE through the volumes of regions including the hippocampus, entorhinal cortex, prefrontal lobe, precuneus, parietal lobe, and posterior cingulate cortex. However, no significant indirect pathway was identified between NLR and cognition through brain network function alone. These results are illustrated in Figure 4 and Table S2.

Figure 4.

The glymphatic system function and volume of key brain structures mediate the association between peripheral immunity and cognition

(A) Mediation model showing CP volume as a mediator of the association between NLR and MMSE.

(B) Mediation model showing FWf-WM as a mediator of the association between SII and MMSE.

(C) Mediation model showing hippocampal volume as a mediator of the association between NLR and MMSE.

(D) Mediation model showing precuneus volume as a mediator of the association between NLR and MMSE. Values on the arrows indicate standardized path coefficients with corresponding p values. Positive and negative values indicate positive and inverse associations, respectively. c denotes the total effect, and c’ denotes the direct effect after including the mediator in the model. Detailed standardized estimates of the indirect, direct, and total effects are provided in Table S2. Mediation analyses were performed using models adjusted for age, sex, and years of education. Exact p values are shown where applicable; values < 0.001 are reported as p < 0.001. Abbreviations: CP, choroid plexus; NLR, neutrophil-to-lymphocyte ratio; MMSE, Mini-Mental State Examination; FWf-WM, free water fraction in white matter; SII, systemic immune-inflammation index.

Subsequently, in the multiple mediation model, NLR, PLR, and SII indirectly affected overall cognitive function through indicators of the glymphatic system (CP volume), brain structural volume, and brain network function. Under the influence of peripheral innate immune activity on cognition, the glymphatic system emerged as a central mediating pathway linking peripheral immunity to alterations in brain structure and function, which in turn affected cognitive performance. These findings identify the glymphatic system as an important loop and a key intermediary mechanism in the immunity-brain-cognition axis. The results are presented in Figure 5 and Table S3.

Figure 5.

Mediation analysis based on multiple mediation modeling

(A) Serial multiple mediation model evaluating the pathway NLR → CP volume → hippocampal volume → MMSE.

(B) Serial multiple mediation model evaluating the pathway NLR → CP volume → precuneus volume → MMSE.

(C) Serial multiple mediation model evaluating the pathway NLR → FWf-WM → hippocampal volume → MMSE.

(D) Serial multiple mediation model evaluating the pathway NLR → FWf-WM → precuneus volume → MMSE.

(E) Serial multiple mediation model evaluating the pathway NLR → CP volume → mean FCS in aDMN → MMSE.

(F) Serial multiple mediation model evaluating the pathway NLR → CP volume → mean FCS in LEN → MMSE. Values on the arrows indicate standardized path coefficients with corresponding p values. Positive and negative values indicate positive and inverse associations, respectively. c denotes the total effect, and c’ denotes the direct effect after including the mediators. Solid and dashed lines indicate significant and nonsignificant paths, respectively. Detailed standardized estimates are provided in Table S3. Serial multiple mediation analyses were adjusted for age, sex, and years of education. Exact p values are shown where applicable; values < 0.001 are reported as p < 0.001. Abbreviations: CP, choroid plexus; NLR, neutrophil-to-lymphocyte ratio; MMSE, Mini-Mental State Examination; FCS, functional connectivity strength; aDMN, anterior default mode network; LEN, left executive network.

Discussion

This study revealed a dynamic multisystem cascade underlying the progression of AD by integrating peripheral immune markers, glymphatic system function, brain structural and network imaging, and cognitive assessments. The findings of the present study suggest that peripheral immune imbalance represents an early alteration in AD, whereas glymphatic system dysfunction may emerge and progressively worsen during the early stages of the disease, thereby mediating the effect of peripheral immunity on the CNS. Structural atrophy and functional degradation of brain networks constitute the direct neurobiological substrates of cognitive decline. In summary, this study elucidates a sequential pathway of “peripheral immunity → glymphatic system → brain structure/network → cognition,” in which glymphatic system dysfunction appears to serve as a central mechanism linking peripheral inflammation to hippocampal atrophy and cognitive decline across the AD spectrum. Accordingly, these results suggest that early modulation of inflammatory responses and preservation of glymphatic function may help limit the progression from inflammation to neurodegeneration, thereby providing potential targets for the early treatment and prevention of AD.

This study identified significant peripheral immune dysregulation in patients with AD, characterized by decreased LY levels and increased NLR. Furthermore, LY, NLR, PLR, and SII were significantly associated with cognitive function, suggesting that peripheral immune status plays a key regulatory role in AD pathophysiology. The observed reduction in peripheral LY in patients with AD is consistent with previous research findings^6,22,23 and reflects a relative suppression of peripheral adaptive immunity.¹⁰ During systemic inflammatory responses, the integrity of the blood-brain barrier in AD is often compromised,^10,24 facilitating LY trafficking into the CNS. This can lead to increased LY infiltration into the brain parenchyma, particularly in the hippocampus and temporal cortex, thereby contributing to reduced LY counts in peripheral circulation.²⁵ Notably, increased numbers of differentiated CD3⁺ T cells in the hippocampus of AD patients²⁶ have been shown to promote Aβ deposition, cognitive decline, and disease progression through the secretion of proinflammatory cytokines such as interferon-γ.²⁷

Furthermore, composite inflammatory markers derived from relative counts of different leukocyte populations and platelets, including NLR, PLR, and SII, provide integrated indices of systemic inflammatory status.²⁸ The NLR reflects the balance between innate and adaptive immune responses, with elevated values indicating a shift toward innate immune activation or suppression of adaptive immunity.²⁹ Multiple studies have demonstrated that peripheral NLR levels are significantly increased in patients with AD^30,31 and are positively associated with dementia risk,^9,10,32 further supporting the role of systemic inflammation in AD pathogenesis. Collectively, these results suggest that an imbalanced peripheral immune state, characterized by heightened innate immune activity and relative suppression of adaptive immune function, represents an important pathological mechanism contributing to AD-related cognitive impairment.

The glymphatic system is a major fluid clearance pathway in the brain,¹¹ and accumulating evidence suggests that its dysfunction plays a significant role in the development and progression of AD. Several imaging-derived indicators, including ALPS index, FWf-WM,³³ and CP volume,³⁴ have been proposed as indirect markers of glymphatic function. ALPS, calculated using DTI, is commonly used to assess glymphatic system integrity and has been suggested as a biomarker of disease progression in AD.^16,35 In this study, ALPS values were significantly reduced in patients with MCI and AD compared with NCs and were negatively correlated with cognitive performance. This finding corroborates previous reports^17,36 and directly links glymphatic dysfunction, as indexed by ALPS, to cognitive decline. Given that glymphatic impairment precedes substantial Aβ deposition and predicts amyloid accumulation and neurodegeneration,^15,36 the results suggest that reduced ALPS may represent an early biomarker of glymphatic failure that is clinically relevant even before extensive pathological burden. Free water (FW) imaging uses DTI to estimate the fraction of freely diffusing water molecules within each voxel,³⁷ reflecting extracellular water content and processes related to inflammation, atrophy, and extracellular edema.³⁸ In the present study, FWf-WM was significantly increased in patients with AD and MCI and showed a significant association with cognitive impairment, consistent with prior reports.^17,39,40 Increased FWf-WM likely reflects impaired interstitial fluid clearance,⁴¹ a hallmark of glymphatic dysfunction. The CP is a highly vascularized ventricular structure responsible for CSF production and constitutes an important component of the glymphatic system involved in metabolite clearance. Increasing attention has been directed toward its role in neurodegenerative disorders.³⁴ In this study, we found that CP volume was significantly increased in patients with AD and MCI and was negatively correlated with cognitive performance. These findings align with previous imaging and autopsy studies^42,43,44 and indicate that CP enlargement is evident across the AD continuum and is linked to cognitive impairment. In line with prior evidence showing negative associations between CP volume and CSF levels of total tau and p-tau in AD,⁴⁵ these results provide further support for the involvement of CP alterations in impaired protein clearance and AD-related pathology. It should be acknowledged that ALPS, FWf-WM, and CP volume are indirect imaging markers. Given the indirect nature of these measures, future studies incorporating direct assessments of CSF dynamics, inflammatory biomarkers, and longitudinal follow-up are warranted to further clarify the underlying mechanisms.

A key finding of this study was the identification of the glymphatic system as a central mediating component in the pathway linking systemic inflammation to cognitive impairment. The glymphatic system, a unique waste clearance mechanism in the brain, critically depends on the polarized distribution of AQP4 water channels on astrocytic endfeet.⁴⁶ Accumulating evidence suggests that systemic inflammation may impair glymphatic function through several mechanisms: (1) proinflammatory cytokines, such as interleukin-6 (IL-6), interleukin-1β (IL-1β), and tumor necrosis factor alpha (TNF-α), directly disrupt the expression and polarization of AQP4 in astrocytes, thereby weakening CSF-interstitial fluid exchange⁴⁷; (2) inflammation-induced vascular responses can alter CSF dynamics, further impeding glymphatic flow⁴⁷; and (3) systemic inflammation is often accompanied by disturbances in sleep architecture, and because glymphatic clearance is most active during deep sleep, such disruption may indirectly compromise clearance efficiency.⁴⁸ Together, these findings point to impaired glymphatic function as a potential convergence point and amplifier of systemic inflammatory effects within the brain, thereby providing a plausible mechanistic link between peripheral inflammation and cognitive decline.

Once glymphatic clearance is compromised, two major downstream consequences may arise, collectively contributing to structural and functional brain abnormalities.^11,49,50 First, the clearance efficiency of neurotoxic proteins, including Aβ and tau, is markedly reduced. The abnormal accumulation of these proteins represents a core pathology feature of AD, as they exert direct neurotoxic effects that promote synaptic loss and ultimately lead to hippocampal atrophy, one of the earliest and most critical pathological hallmarks of AD.⁵¹ Second, the accumulation of metabolic waste generates a microenvironment unfavorable for neuronal homeostasis and directly triggers neuroinflammatory responses, thereby disrupting the functional integration of large-scale brain networks.⁵² Consistent with this view, the present findings showed that glymphatic dysfunction was associated with impaired functional connectivity between the default mode network (DMN) and executive network (EN), two networks involved in memory integration and executive control and particularly vulnerable in early AD.^53,54

The peripheral immunity → glymphatic system → brain structure/network → cognition model proposed in this study holds substantial theoretical value, as it effectively bridges the gap between the “peripheral inflammation hypothesis” and the “glymphatic system hypothesis.” Although these two mechanisms have largely been investigated independently, these results suggest that they represent interconnected components of a shared pathological cascade. Considering this integration, a multitarget therapeutic strategy that combines modulation of systemic inflammation with enhancement of glymphatic clearance of pathological proteins may be particularly effective. Such an approach could open new avenues for the treatment and prevention of AD. While direct clinical evidence is lacking, lifestyle interventions with anti-inflammatory properties—such as anti-inflammatory diets, regular physical activity, and improved sleep⁵⁵—and selected anti-inflammatory pharmacological approaches⁵⁶ may represent promising strategies to preserve glymphatic function, pending future mechanistic studies. Furthermore, the NLR, an inexpensive and readily accessible blood-based biomarker, when combined with imaging-based assessments of glymphatic function (e.g., the DTI-ALPS index or CP volume), shows promise as a composite biomarker for identifying individuals at high risk for AD and for disease staging.

Conclusions

This study identified an association pattern across the AD spectrum, where peripheral immune dysregulation, glymphatic dysfunction, brain structural/network impairment, and cognitive decline co-varied along the AD continuum. Glymphatic system dysfunction emerged as a mediator linking peripheral inflammation to brain atrophy, network dysfunction, and cognitive decline. These findings generate a hypothesis that systemic inflammation may contribute to cognitive decline through glymphatic impairment and highlight glymphatic dysfunction as a promising target for future mechanistic studies. Although the specific mechanisms underlying these interrelationships remain to be further investigated, these findings provide a foundation for future mechanistic studies and may offer new insights into potential diagnostic markers and therapeutic strategies.

Limitations of the study

This study has several limitations that should be acknowledged. First, the cross-sectional and observational design precludes causal or mechanistic conclusions. Thus, the proposed immune-glymphatic-brain-cognition framework should be interpreted as a biologically plausible, association-based, and hypothesis-generating model requiring validation in future longitudinal studies. Second, although AD biomarker data were available for some participants, these data were not obtained using a uniform modality across the entire cohort because of the retrospective nature of the study. Future prospective studies with more complete and standardized biomarker assessments in all participants are needed to further verify these findings. Third, the imaging markers used to reflect glymphatic-related function are indirect and remain methodologically evolving; their sensitivity, specificity, and biological validity need further confirmation. Fourth, the peripheral inflammatory indices examined in this study, including LY, NLR, PLR, and SII, are nonspecific and may be influenced by various comorbidities and physiological conditions. These indices were interpreted as pragmatic indicators of peripheral immune status rather than markers of specific inflammatory pathways. Fifth, APOE4 status was not included as a covariate in the adjusted analyses because APOE genotyping data were not available for all participants; therefore, potential residual confounding related to genetic background cannot be completely excluded. Finally, the mediation analyses were conducted in complete-case subsamples rather than in the full cohort, and the reduced sample size may have limited statistical power and the stability of mediation estimates.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact and corresponding author, Yu Yang (yang_yu@jlu.edu.cn).

Materials availability

This study did not generate new unique reagents or materials.

Data and code availability

De-identified subject-level data supporting the findings of this study, including key hematological measures, imaging-derived summary metrics, cognitive scores, and relevant demographic and covariate information, may be made available from the corresponding author upon reasonable request and, where required, subject to institutional approval. Raw clinical records and raw MRI data are not publicly available because of privacy and ethics restrictions, as well as the scope of the original informed consent. This paper does not report original code.

Acknowledgments

The authors want to thank all the participants enrolled in this study. This work was supported by the Brain Science and Brain-like Intelligence Technology-National Science and Technology Major Project (2022ZD0211600), the Doctor of Excellence Program (DEP) of the First Hospital of Jilin University (JDYY-DEP-2024015), the National Natural Science Foundation of China (no. 81600923), and Jilin Province Natural Science Foundation of China (JLSRCZX2026-16).

Author contributions

B.X. designed and conceptualized the study, analyzed the data, interpreted the findings, and drafted and revised the manuscript. Y.Y. designed and conceptualized the study, supervised the work, and critically reviewed and revised the manuscript. X.F. and C.G. performed data analysis. J.L., Y.H., and S.Y. participated in data collection. All authors reviewed the manuscript.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
De-identified participant-level clinical, cognitive, hematological, and MRI-derived dataset	The First Hospital of Jilin University/This paper	Restricted access; see Availability of data and materials
Software and algorithms
IBM SPSS Statistics	IBM	RRID:SCR_016479
MATLAB	MathWorks	Version R2021b; RRID:SCR_001622
SPM12	Wellcome Center for Human Neuroimaging	Version 12, revision 7771; RRID:SCR_007037
FSL	FMRIB Software Library	Version 6.0.7; RRID:SCR_002823
MRtrix3	MRtrix3 developers	Version 3.0.1; RRID:SCR_024123
Computational Anatomy Toolbox 12 (CAT12)	Structural Brain Mapping Group, University of Jena	Version 12.8.2, revision 2170; RRID:SCR_019184
FreeSurfer	FreeSurfer developers	Version 7.4.1; RRID:SCR_001847
Other
Cognitive assessment instruments	Original publications and scale manuals	MMSE, MoCA, etc.; see method details

Experimental model and study participant details

Human participants

We continuously recruited 733 patients, their caregivers, and volunteers from the Cognitive Disorders Clinic of First Hospital of Jilin University, China. After excluding participants who did not meet the inclusion criteria, 570 individuals were enrolled in this study, including 154 with AD, 244 with MCI, and 172 NC (115 caregivers and 57 volunteers). Demographic characteristics, including age, biological sex, and education level, are summarized by group in Table 1. Biological sex was determined based on medical records and was reported for all participants, whereas gender identity was not collected in the original clinical records.

This study was approved by the Ethics Committee of First Hospital of Jilin University (approval no. 2016-028), and written informed consent was obtained from all participants before study participation. Biological sex was determined based on medical records and was reported for all participants. Gender identity was not systematically collected in the original clinical records and therefore could not be analyzed. The association of biological sex with the study outcomes was considered by including sex as a covariate in the statistical models. The absence of gender identity information may limit the generalizability of the findings across populations with different gender identities.

Enrollment criteria for the AD group included: (1) CDR >0.5; (2) meeting the 2011 National Institute on Aging-Alzheimer’s Association (NIA-AA) core diagnostic criteria for AD.⁵⁷ Enrollment criteria for the MCI group included: (1) MMSE scores between 24 and 30; (2) CDR score of 0.5; (3) Presence of a complaint of memory loss (not a required condition); and (4) Does not meet the diagnostic criteria for dementia, shows only mild cognitive impairment, and retains the ability to work and live independently. Enrollment criteria for the NC group included: (1) MMSE scores between 24 and 30; (2) CDR score of 0. The exclusion criteria for all groups included: (1) Neurological or psychiatric disorders such as epilepsy, encephalitis, and severe depression; (2) other systemic diseases that may lead to dementia, cancer, anemia, or thyroid dysfunction; (3) inability to complete neuropsychological assessments or the presence of MRI contraindication; (4) severe white matter lesions, defined as a Fazekas score >2 on MRI; (5) T2-weighted image (T2WI) and T2-fluid attenuated inversion recovery (T2-FLAIR) MRI showing cerebral infarction, space-occupying lesions, or other pathological abnormalities. All clinical diagnoses were independently made by two neurologists with 12–15 years of clinical experience. Any disagreements were resolved through consensus discussions. Peripheral immune assessments, 3D-T1WI, DTI, and resting-state functional MRI (rs-fMRI) were performed in the subgroups. Figure S1 shows the grouping flowchart of the study participants. The demographic details of the 3D-T1WI, DTI, and rs-fMRI subgroups are presented in Tables S4, S5, and S6.

Method details

Study design

This study was designed to investigate the relationships among glymphatic function, peripheral immune markers, brain structural and functional imaging features, and cognitive performance in participants with Alzheimer’s disease, mild cognitive impairment, and normal cognition. The analyses were based on clinical and cognitive data together with available hematological and multimodal MRI-derived measures. Figure 1 presents the schematic overview of the study design. All participant-level clinical, cognitive, hematological, and imaging data were de-identified before analysis.

Cognitive function assessments

Cognitive function was evaluated using MMSE, MoCA, MES, and CDR scale. Daily functioning was assessed using ADL. Neuropsychiatric behavioral evaluations were conducted using NPI. MES is considered a practical and efficient cognitive screening tool for Chinese populations. MES comprises two domains, MES-E and MES-M, both of which demonstrate high sensitivity and specificity for cognitive functioning.⁵⁸

Head MRI data acquisition and preprocessing

MRI acquisition and preprocessing

Head MRI data were acquired using a Philips Ingenia 3.0T MR system in the Department of Radiology of the First Hospital of Jilin University. The scanning sequences for each participant included DTI, rs-fMRI, 3D-T1WI, T2WI, and T2-FLAIR. The DTI, rs-fMRI, and 3D-T1WI scanning parameters were as follows.

• (1)DTI scanning parameters: diffusion gradient b value of 1000s/mm2 in 64 non-collinear directions and a non-diffusion gradient b value of 0s/mm2; repetition time (TR) = 9640 ms; echo time (TE) = 91 ms; flip angle (FA) = 90°; slice thickness = 2 mm; field of view (FOV) = 112 × 112 mm; voxel size = 2 × 2 × 2 mm; (2) rs-fMRI scanning parameters: TR = 2000 ms; TE = 30 ms; FA = 90°; slice thickness = 3.5 mm; FOV = 64 × 64 mm; voxel size = 3.5 × 3.5 × 4.2 mm; total scanning time = 480 s; (3) 3D-T1WI scanning parameters: TR = 7.2 ms; TE = 3.3 ms; FA = 7°; slice thickness = 1 mm (no gap); FOV = 256 × 256 mm; voxel size = 1 × 1 × 1 mm; total scanning time = 360 s.

MRI data preprocessing

• (1)DTI data preprocessing was primarily performed using FSL and MRtrix3 software and included the following steps: file format conversion; artifact corrections using a denoising algorithm and Gibbs-unringing³³; eddy current and head motion correction to reduce image deformation caused by coil eddy current and movement; skull stripping to decrease computational complexity during tensor reconstruction; tensor fitting to obtain voxel-wise quantitative diffusion parameters; and spatial normalization using a nonlinear two-step registration method to achieve accurate alignment of native diffusion images with Montreal Neurological Institute (MNI) space.⁵⁹ (2) rs-fMRI data preprocessing was conducted mainly using-based toolboxes and included: File format conversion; removal of initial time points; slice timing correction; realignment with exclusion of images showing excessive head motion; spatial normalization to MNI space with resampling to 3 × 3 × 3 mm voxels; detrending and nuisances regression; and band-pass filtering (0.01–0.08 Hz) to eliminate the high-frequency noise. (3) 3D-T1WI data were used to obtain key brain structural volumes using the Computational Anatomy Toolbox 12 (CAT12) and FreeSurfer. The preprocessing steps included correction for head motion; spatial segmentation using voxel-based morphometry (VBM); spatial normalization; B1 field inhomogeneity correction; and skull stripping.

Glymphatic function metrics

ALPS index calculation

The extraction of the ALPS index was completed in template space. First, four 5-mm diameter spherical regions of interest (ROIs) were positioned within bilateral projection and association fibers at the level of the lateral ventricular body based on the JHU-ICBM white matter atlas. Second, the diffusion tensor images along the x axis, y axis, and z axis for all participants were registered to template space using the nonlinear “twice registration” method (first using the “flirt” command for contour-level alignment, followed by the “fnirt” command for internal detail registration). The ALPS index was calculated as the ratio of the average x axis diffusivity in both the projection (Dxx,proj) and association (Dxx,assoc) fibers to the mean y axis diffusivity in projection fibers (Dyy,proj) and z axis diffusivity in association fibers (Dzz,assoc) The mathematical expression is as follows:

$$ALPSindex=\frac{\text{Mean}\left(\text{Dxx},\text{proj}+\text{Dxx},\text{assoc}\right)}{\text{Mean}\left(\text{Dyy},\text{proj}+\text{Dzz},\text{assoc}\right)}$$

Calculation of FWf-WM

FW fraction maps in the native space were estimated from diffusion images to eliminate the influence of tissue water using a single-shell FW fraction estimation algorithm based on Python, and were then registered into the MNI space. The pure white matter mask was obtained using FreeSurfer commands and was likewise registered into the MNI space. Finally, the mean FW fraction within the pure white matter mask of each participant was extracted for subsequent statistical analysis.^38,60,61

Volume extraction of CP

FreeSurfer software was used for subcortical nuclei segmentation of the 3D-T1WI data to obtain CP volume. All final segmentation results were reviewed and approved by a neuroimaging researcher with years of experience, without any manual correction. The volume of the region was expressed as the proportion of its volume to the total intracranial volume.

Peripheral immunity markers

Blood samples were analyzed within 24 h after blood collection at the laboratory department of our center. Peripheral blood cell counts were analyzed using a Japanese SYSMEX XN-9000 hematology analyzer, and quality control was performed according to the manufacturer’s recommendations. Data on neutrophil (NE), monocyte (MO), platelet, and lymphocyte (LY) counts were extracted. Subsequently, we calculated four ratios based on the peripheral blood cell counts: NLR, LMR, PLR, and SII (SII = neutrophils × platelets/lymphocytes). Elevated neutrophil and monocyte counts, NLR, PLR, and SII indicated a stronger peripheral innate immune response, whereas elevated lymphocyte counts and LMR indicated a stronger peripheral adaptive immune response.¹⁰ A total of 379 participants completed the hematological testing in this study, and detailed demographic information for each subgroup is provided in Table S7.

Brain structures volume extraction

Using the BA⁶² and JulichBrain^63,64,65 atlases as anatomical references, the CAT12 toolbox was applied to automatically extract the volumes of key brain structures such as the hippocampus, insular gyrus, and precuneus. The analyzed brain regions were selected a priori based on prior studies, their relevance to AD-related neurodegeneration, and their presumed involvement in glymphatic and immune-related processes.

Functional connectivity strength (FCS) assessment in key brain network

FCS represents voxel-wise weighted degree centrality, defined as the sum of the functional connectivity values between a given voxel and all other voxels in the whole brain. To examine the strength of the functional connections between key brain networks and other gray matter regions, the mean FCS values within resting-state networks (RSNs) were extracted. In this study, the calculation of FCS was limited to the gray matter region to avoid unnecessary effects caused by the loss of signals of interest. The functional connectivity threshold was set to 0.6, following previous studies, to eliminate low-relevance functional connectivity or noise signals.^66,67

Quantification and statistical analysis

SPSS Statistics (IBM, New York, U.S.A.) was used to conduct statistical analyses, and intergroup differences in demographic, cognitive, glymphatic system function, and peripheral immune indices were assessed. Variables conforming to a normal distribution were described by their means and standard deviations, whereas those not conforming were described by medians and IQR. Categorical variables were described by counts and percentages. Univariate comparisons between groups were performed using ANOVA with Bonferroni post hoc test for normally distributed continuous variables, Kruskal-Wallis H test with Dunn’s post hoc test (Bonferroni-corrected) for non-normally distributed continuous variables, or χ2 tests according to the types and distributions of variables. Furthermore, general linear regression was employed to estimate relationships between the different variables.

We utilized PROCESS software (version 4.1) for the mediation analysis.⁶⁸ PROCESS Model 4 was initially used to investigate whether the relationship between peripheral immunity (X) and cognitive level (Y) was mediated by glymphatic system function, brain structural volume, or brain network function (M). Subsequently, additional serial multiple mediation analyses were conducted using PROCESS Model 6 to assess whether the glymphatic system (M1) and changes in brain structure or brain network function (M2) mediated the impact of peripheral immunity (X) on cognition (Y). The serial ordering in Model 6 was specified according to the conceptual framework of the present study and prior literature. All mediation analyses utilized 5,000 bias-corrected bootstrap replicates, and the significance of indirect effects was determined based on the 95% confidence interval (CI); an indirect effect was considered statistically significant when the CI did not include 0. The main mediation analyses were conducted in a complete-case subset of 187 participants with available blood, MRI, cognitive, and covariate data. For DTI- and fMRI-related mediation analyses, sample sizes were smaller and varied across models according to data availability and image quality. All continuous variables were z-standardized before regression and mediation analyses. Statistical significance was defined as p < 0.05. Exact p values are reported where applicable; values < 0.001 are reported as p < 0.001.

Age, biological sex, and years of education were included as covariates in all applicable regression and correlation analyses. All mediation models were adjusted for age, biological sex, and years of education. Biological sex was determined based on medical records and was included as a covariate to account for its potential association with the variables of interest. Gender identity was not collected in the original clinical records and therefore could not be included in the statistical models.

Published: June 30, 2026