Alzheimer’s disease is pathologically characterized by the extracellular accumulation of amyloid beta (Aβ) plaques and intracellular tau neurofibrillary tangles. Neuroinflammation has been considered as the third core pathological feature of AD (Kinney et al., 2018; McGeer and Rogers, 1992). Neuroinflammation, triggered by microglial and astrocytic activation, occurs decades before the clinical manifestation of AD and is strongly linked to the misfolded and aggregated proteins (Steardo et al., 2015). Evidence from clinical studies demonstrated that inflammatory biomarkers were elevated in AD patients (Shen et al., 2019), and utilizing anti-inflammatory drugs might reduce the risk of AD (Wang et al., 2015). Astrocytes, as a key component of the neurovascular unit (Iadecola, 2017), are crucial for maintaining the integrity of the blood-brain barrier (BBB) (Luissint et al., 2012). Notably, astrocytes are important regulators of central nervous system inflammatory response and actively contribute to neuroinflammatory cascades in AD pathogenesis (Ferrari-Souza et al., 2022). Elucidating the astrocyte-neuroinflammation interplay may therefore provide potential clinical value for diagnosis and treatments for AD.

Chitinase 3-like protein 1 (CHI3L1/YKL-40) is almost exclusively expressed by astrocytes throughout neurodevelopment in human brain tissue (Connolly et al., 2023), and strongly colocalizes with the astrocyte marker Glial fibrillary acidic protein (GFAP) (Llorens et al., 2017; Querol-Vilaseca et al., 2017). It is being increasingly considered as a biomarker of reactive astrocytes in AD (Carter et al., 2019; Pelkmans et al., 2024). Previous research has shown that high levels of CSF YKL-40 are significantly associated with the accumulation of AD pathologies and increase the risk of AD dementia (Ferrari-Souza et al., 2022; Janelidze et al., 2018; Pelkmans et al., 2024; Vergallo et al., 2020; Warmenhoven et al., 2024). However, there are studies reporting somewhat contradictory findings that higher CSF YKL-40 predicts slower cognitive decline in AD dementia (Kester et al., 2015). Together, the contribution of astrocytic activity, as reflected by the elevated levels of CSF YKL-40, to the pathogenesis of AD and its value in monitoring the progression of AD may be complex and require further elucidation.

Additionally, numerous studies have pointed out that vascular risk factors, especially hypertension, were strongly associated with AD (Kivipelto et al., 2001; McGrath et al., 2017). In vitro and in vivo studies indicate that hypertension exacerbates amyloid and tau pathology, activates neuroinflammation, and accelerates cognitive decline (Bajwa and Klegeris, 2022; Denver et al., 2019). In a mouse APP/PS1 model of AD, decreases in Chi3l1/YKL-40 promote astrocytic Aβ phagocytosis, and mitigate the formation of misfolded proteins (Lananna et al., 2020). It is noteworthy that YKL-40 levels are elevated in patients with hypertension (Xu et al., 2016), and other chronic inflammatory conditions, such as diabetes, atherosclerosis (Rathcke and Vestergaard, 2006). Hypertension has the potential to cause damage to the cerebral microcirculation, decrease cerebral blood flow, and subsequently lead to ischemia, initiating a neuroinflammatory cascade (Snyder et al., 2017) that involves the activation of astrocytes and microglia, which release YKL-40 (Bonneh-Barkay et al., 2010) and sTREM2 (Brendel et al., 2017) as part of the tissue repair process. Taken together, the complex interrelationships between hypertension, neuroinflammation, and YKL-40, and their implications in the progression of AD, remain to be thoroughly explored.

The present study utilized the cross-sectional and longitudinal data with the aims to (1) ascertain the relationships of CSF levels of YKL-40 with neuroinflammation, AD core biomarkers (Aβ₄₂ and p-tau [phosphorylated tau]), neurodegeneration (t-tau [total tau], MRI brain structures) and cognitive function; (2) explore whether AD modifiable risk factors (vascular risk factors, particularly hypertension) or non-modifiable risk factors (age, sex, apolipoprotein E4 [APOEε4] status) moderate these associations; (3) determine whether neuroinflammation serves as a mediating factor in the correlation between CSF levels of YKL-40 and AD-related pathologies; and (4) evaluate the predictive value of CSF level of YKL-40 for longitudinal changes in cognitive function and brain structures from the AD Neuroimaging Initiative (ADNI) database.

This study was the first to systematically explore the interrelationships of CSF YKL-40 with AD biomarkers, neuroinflammation, cognitive function and brain structures, and elucidate the additive effects of neuroinflammation and hypertension on the associations between astrocyte reactivity and AD-related pathologies. Overall, these results suggest that specific neuroinflammatory biomarkers might mediate the associations between CSF YKL-40 and p-tau and t-tau with hypertensive dependency, particularly in the preclinical and prodromal stages of AD (Figure 5). Specifically, the key findings are that: (1) higher baseline levels of CSF YKL-40 are positively related to p-tau and t-tau; (2) hypertensive status modifies the significant associations between CSF YKL-40 and p-tau and t-tau; (3) the mediation of VCAM1, sTNFR1 and sTNFR2 on abovementioned significant associations persist in the whole sample and individuals with hypertension; (4) higher baseline levels of CSF YKL-40 interact with hypertension to promote a faster decline in cognitive function and atrophy in hippocampal volume, entorhinal volume, middle temporal lobe volume, and whole-brain volume. Our results indicate that higher levels of CSF YKL-40 may promote AD progression, and constitute a reasonable pathway to connect the effects of hypertension and neuroinflammation with astrocyte reactivity in contributing to AD pathogenesis.

CSF Chitinase 3-like protein 1 (CHI3L1/YKL-40) is identified as a significant indicator of astrocyte reactivity (Carter et al., 2019). An increasing amount of evidence indicates that the positive correlations between CSF YKL-40 levels and tau pathologies, but not Aβ (Ferrari-Souza et al., 2022; Janelidze et al., 2018; Pelkmans et al., 2024). Pelkmans et al. (2024) reported that YKL-40 was released into the CSF slightly later in the AD cascade, specifically mediating the link between the downstream p-tau induced by Aβ and the neuronal damage induced by tau protein, whereas it did not mediate the association between soluble and insoluble Aβ aggregates. GFAP, another promising biomarker of astrocytic activation, was considered to be predominantly a response to Aβ pathology (Pelkmans et al., 2024). Our results were consistent with these findings that CSF YKL-40 was significantly associated with p-tau and t-tau. Similarly, the positive associations between CSF YKL-40 and p-tau and t-tau were particularly robust in the A + MCI and the T + subgroups in our exploratory analyses, which favored the hypothesis that CSF YKL-40 might contribute to tau accumulation secondary to Aβ plaques in the downstream AD cascade. In contrast, the EMIF-AD Multimodal Biomarker Discovery study revealed an association between YKL-40 and Aβ only in the non-demented population (Bos et al., 2019). A postmortem study found that YKL-40 was independent of tau in other neurodegenerative diseases (Querol-Vilaseca et al., 2017), which indicated that the astrocyte activity as measured by the CSF levels of YKL-40 might not be exclusively influenced by amyloid-tau dichotomy (Pelkmans et al., 2024; Salvadó et al., 2023). Besides, the marginally negative correlation between baseline CSF YKL-40 and the entorhinal cortex volume did not survive FDR correction in our results, warranting replication and validation in a larger sample. However, an animal experiment revealed that calcium-binding protein centrin-2-positive astrocytes were distributed in clusters within the entorhinal cortex and co-localized with YKL-40, providing a possible explanation for our results (Degl’Innocenti et al., 2024). Nevertheless, it should be noted that the choice of brain regions in our mediation analyses was to some extent guided by statistical considerations, which may lead to insufficient exploration of the mechanisms in crucial biological regions such as the hippocampus.

This study further proves that astrocytic reactivity with increased neuroinflammation play a prominent role in AD pathologies (Sofroniew and Vinters, 2010). We observe a robust correlation between CSF YKL-40 and ICAM1, VCAM1, sTNFR1, and sTNFR2. Simultaneously, these neuroinflammatory biomarkers partially mediate the effects of YKL-40 on tau pathology. These results reinforce the value of neuroinflammatory biomarkers to establish a more precise AD diagnostic framework in the clinic and provide a new perspective on understanding how astrocyte affects tau pathology. Of importance, the mediation analysis in this study is based on observational data and these CSF biomarkers (YKL-40, sTNFR1, sTNFR2, ICAM1, and VCAM1) may reflect overlapping aspects of a shared neuroinflammatory process, with the results reflecting correlation rather than causation. Further additional in vitro and in vivo studies are needed to verify the sequential and mechanistic associations between these molecules. Nevertheless, our findings suggest that interventions targeting astrocyte activation and neuroinflammation may benefit the prevention of tau pathology.

To the best of our knowledge, this is the first observational study to explore the interaction of hypertension and neuroinflammation on the associations between astrocyte reactivity and AD-related pathologies. We observed that hypertension reinforced the positive correlation between CSF YKL-40 and p-tau and t-tau in our interaction and subgroup analyses. Previous studies have shown that YKL-40 levels were elevated in hypertensive patients (Xu et al., 2016). Although our results revealed no difference in baseline levels of YKL-40 between the hypertensive and normotensive subgroups, further analyses indicated that the partial mediating effects of neuroinflammation (VCAM1, sTNFR1, and sTNFR2) on the associations between CSF YKL-40 and p-tau and t-tau were replicated in the HTN + subgroup rather than in the HTN- subgroup. An experiment investigating the additional or synergistic effects between AD and hypertension in rat models revealed that hypertension increased neuroinflammation (triggered by microglia and astrocytes) and tau pathology with minimal effect on Aβ pathology, which supported our results (Denver et al., 2019). Thus, these data seem to confirm this assumption that hypertension and astrocyte-driven neuroinflammation may interact to exacerbate AD pathologies (Bajwa and Klegeris, 2022; Nelson et al., 2014). However, the modification effects of other AD risk factors, such as age, sex, or APOEε4 status, are not observed in this study. A potential explanation for the lack of interaction may be the relatively limited sample size. Further studies are warranted to validate our findings.

Existing clinical evidence has suggested that an elevated levels of soluble cell adhesion molecules (CAMs), hallmarks of vascular endothelial dysfunction (Lee and Benveniste, 1999), are correlated with numerous vascular risk factors (Hwang et al., 1997), dementia (Drake et al., 2021), and immune-mediated diseases (Kong et al., 2018). Nevertheless, the underlying mechanism of its involvement in the pathogenesis of AD is not fully understood. Earlier studies on which CAMs were most relevant to cognition have yielded inconsistent results (Drake et al., 2021; Nielsen et al., 2007), with our present research suggesting a preference for VCAM1. Another cohort study from ADNI observed a synergistic effect of CAMs and AD pathology on the clinical severity of AD (Drake et al., 2021). Yousef et al. (2019) found that glial activation in the hippocampus of aging mice was associated with focal upregulation of VCAM1 on the luminal side adjacent to the BBB. Additionally, a study from postmortem brain tissue from AD patients found that perivascular astrocytes in the white matter exhibit high CHI3L1 expression, further suggesting the involvement of YKL-40 in BBB disruption (Moreno-Rodriguez et al., 2020). A well-established fact is that endothelial dysfunction and chronic vascular inflammation are critical pathophysiological changes in hypertension. In the present study, we observed that CAMs mediated the associations between YKL-40 and p-tau and t-tau, and the mediation effects of VCAM1 were particularly robust in hypertensive patients. Altogether, these findings support the hypothesis that hypertension may influence the neuroinflammatory pathological cycle of astrocyte activation, subsequently inducing endothelial dysfunction and further aggravating tau pathology in AD.

This work identified sTNFRs as another potential mediating factor in the association between CSF YKL-40 and tau pathology. TNF-α, a biomarker of proinflammatory cytokine and neurodegenerative diseases (McCoy and Tansey, 2008), is secreted by activated astrocytes and microglia. TNFR could be stimulated upon binding to soluble TNF and cleaved to yield sTNFR. The measurement of sTNFRs provides a more sensitive determination of TNF-signaling activity than the direct measurement of TNF-α (Kreuzer et al., 1996). In line with our results, a prior study had demonstrated positive associations between sTNFRs and Aβ, as well as tau pathology in CSF (Buchhave et al., 2010). Furthermore, we and other previous studies found that sTNFRs were elevated in hypertensive patients compared with normotensive patients (Verma et al., 2019). These findings were supported by AD rat models showing that administration of telmisartan (an AT1 receptor blocker) can decrease TNF-α mRNA expression in the hippocampus and ameliorate spatial memory (Shindo et al., 2012). By coincidence, our results showed that the mediation effects of sTNFRs were notably stronger pronounced in the hypertensive group. The exact mechanism by which hypertension influences the mediation effects of sTNFRs was not entirely elucidated. Considering sTNFRs as an important signaling mechanism leading to inflammation and endothelial dysfunction, we speculate that microvascular injury and BBB disruption may be the key underlying mechanisms. Additionally, Shen et al. (2019) found that the levels of plasma TNF-α were increased and vascular endothelial growth factor A were decreased in MCI patients compared to cognitively healthy controls, while plasma sTNFR-1 were inversely correlated with MoCA scores in MCI patients. These findings supported neuroinflammatory and vascular hypothesis for dementia. However, no significant correlation between CSF sTNFRs and cognitive scores was found in the present study. The reason for the inconsistent results may be the heterogeneity of the study population, sample size, differences in the measurement of the biomarker or other confounding factors.

We also systematically reviewed the previous literature and found that previous studies frequently, but not consistently reported the relation of IL-6, IL-7, or IL-10 to AD biomarkers, neuroinflammation, cognitive performance, and brain structures (Cisbani et al., 2020; Hazen et al., 2020; Janelidze et al., 2018; Marsland et al., 2015). In line with our results, a prior study using ADNI database had also demonstrated that there was no significant associations of CSF IL-6, IL-7, or IL-10 with CSF AD biomarkers (Xu et al., 2021). But we observed suggestive inverse associations between CSF IL-7 and MMSE, MEM and ADAS13. The biological mechanism between these correlations is currently unknown, although it can be speculated that various physiologic, medical, environmental, or lifestyle factors could impact on inflammation and AD pathologies. The potential biological mechanism remains to be clarified in the future.

Furthermore, a better understanding of the predictive role of YKL-40 in AD progression will help to develop reliable prognostic and outcome indicators in clinical practice. We therefore evaluated the utility of baseline CSF YKL-40 levels in predicting longitudinal changes of cognitive function and brain atrophy. Consistent with previous studies, our longitudinal investigations show that higher CSF YKL-40 levels are associated with fast decline in cognitive performance (Warmenhoven et al., 2024). Other studies have also reported that higher levels of CSF YKL-40 increased risk of developing AD dementia in non-demented elderly (Janelidze et al., 2018). A Chinese community-based aging cohort demonstrated that higher cortical tau tangle load and lower temporal cortical thinning were more pronounced in the presence of higher levels of astrocyte activation, as measured by plasma GFAP (Guo et al., 2024). Likewise, our study observes significant associations between higher baseline CSF levels of YKL-40 and accelerated rates of atrophy in AD signature ROI volumes (whole brain, hippocampus, entorhinal cortex, and middle temporal lobe), which are vulnerable regions associated with dementia. Aligning with previous research that hypertension was related to global gray matter atrophy (Glodzik et al., 2012), our study found that hypertension had additional deleterious effects on the typical atrophy pattern linking to CSF YKL-40. These findings underscored that CSF YKL-40 and hypertension could contribute to cognitive decline and brain atrophy in an additive way. Interestingly, our longitudinal sensitivity analyses showed that higher baseline levels of CSF YKL-40 were associated with faster atrophy in the volumes of whole brain, hippocampus, and mid temporal lobe in the A + CN rather than the A-CN subgroup. As expected, combining blood pressure profile with T status, faster atrophy in the volumes of whole brain, hippocampus, and mid temporal lobe was replicated in the T + HTN + subgroup but not in other subgroups. This is theoretically plausible given that Aβ burden in the early phases of AD is strongly associated with brain atrophy, and as the disease progresses, downstream pathological events resulting from Aβ accumulation combined with hypertension may accelerate the sustained brain atrophy (Chételat et al., 2010). Our findings highlighted the clinical application of CSF YKL-40 in combination with hypertension for monitoring and predicting the progression of AD.

The study included participants with CN, MCI and AD, had long-term follow-up within repeated tests covering multiple domains of cognitive function and brain structures, and utilized multiple analysis models, which might strengthen the plausibility and robustness of our findings. However, there were some possible limitations in this study. First, the assessment of CSF YKL-40 concentration at a single time point limited our ability to ascertain the impact of changes in YKL-40 levels on the primary AD-related pathologies, highlighting a significant area for future research endeavors. Secondly, this study only employed a simple binary classification for hypertension. A more detailed differentiation of specific blood pressure parameters, including systolic, diastolic blood pressure, pulse pressure, as well as duration, could provide a better understanding of the additive impacts of hypertension. Thirdly, the scarcity of other biomarkers for astrocyte activation (i.e., GFAP) in the ADNI database constrained our capacity to perform a more comprehensive analysis. Last, despite the exploration of potential pathways through mediation analysis, the biological overlap among these inflammatory biomarkers could impact the directional inferences made.

In summary, we have observed that ICAM1, VCAM1, sTNFR1, and sTNFR2 mediate the associations between CSF YKL-40 and p-tau and t-tau. Baseline CSF YKL-40 levels adversely affected longitudinal changes in cognitive function and brain structures in the AD spectrum. Additionally, our findings highlight that hypertension has an additive effect on the mediation effects, as well as the longitudinal correlations. Taken together, this study provides convincing evidence for the development of therapeutic strategies targeting astrocyte reactivity and neuroinflammation while emphasizing the importance of controlling blood pressure to prevent AD progression.