Neutrophil activation may trigger tau burden contributing to cognitive progression of chronic sleep disturbance in elderly individuals not living with dementia

The data on 2272 adults were retrieved from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (http://​adni.​loni.​usc.​edu). ADNI is a multi-site dataset launched in 2003 and designed to test the clinical symptoms, imaging, genetic, and biochemical biomarkers of AD. Data collection and sharing in ADNI were approved by the institutional review boards of all participating institutions. Written informed consent was obtained from all participants or their guardians in accordance with the Declaration of Helsinki. The participants are older adults aged 55–90. Each participant had an in-person neuropsychological assessment interview at baseline and annual follow-up [23, 24]. The inclusion and exclusion criteria are shown in the flow chart (Additional file 1). The subjects in this study received an NPI-sleep assessment [25, 26] at least twice in follow-up visits, with an interval of 6 months. The participants reporting normal sleep at each follow-up were selected for the normal sleep group (n = 528), and the subjects reporting sleep difficulties at baseline and follow-up (≥ 2 times in total) were selected for the CSD group (n = 256). To avoid information bias, patients who reported sleep disturbance only once were excluded. Finally, a total of 784 non-dementia elderly were included in this study. Detailed NPI-sleep assessments are provided in Additional file 2.

Cognitive assessments included the Mini-Mental State Examination (MMSE), the Alzheimer’s Disease Assessment Scale-cognitive (ADAS-cog), the Functional Activities Questionnaire (FAQ), the ADNI-MEM for memory, and the Montreal Cognitive Assessment (MoCA). In the ADNI, cognitively normal subjects had MMSE scores of 24–30, a Clinical Dementia Rating (CDR) of 0, and no memory complaints. Subjects with mild cognitive impairment (MCI) had MMSE scores of 24–30, a CDR score of 0.5, an informant-reported memory complaint, and objective evidence of memory loss. Through follow-up, dementia was diagnosed with MMSE scores of 20–26, a CDR score of 0.5–1, and subjective memory concern as reported by the subject, study partner, or clinician. The detailed criterion can be found in Additional file 2.

Individuals had in-person interviews at baseline and follow-up per 6 months, and the follow-up time was up to 168 months (28 times). A total of 784 non-dementia elderly were involved in this study at baseline, which included normal cognitive and MCI subjects. Through follow-up, cognitively stable individuals were defined as (1) stable normal cognitive, (2) stable MCI, or (3) MCI reversing into cognitively normal. Meanwhile, cognitively progressive individuals were defined as (1) MCI progressing into dementia or (2) cognitively normal progressing into MCI or dementia.

After fasting overnight, the participants’ blood was collected in the morning using vacuum tubes with ethylene diamine tetraacetic acid (EDTA). The blood samples were then sent for analysis on the same day of collection for routine blood test, including blood neutrophil percentage.

Annual change rates of hippocampal volumes were measured longitudinally in 198 subjects who had ≥ 3 available MRI assessments provided by ADNI imaging data. The methodology details of measuring hippocampal volumes are in the Additional file 3 [27].

Before analysis, all concentrations were normalized into Z scores, and outliers beyond ± 3δ were excluded. The CSF from 771 subjects had typical AD biomarkers, including Aβ₄₂, p-tau, and t-tau proteins. These biomarkers were detected using a fully automated and highly standardized Roche Elecsys immunoassay. Meanwhile, the CSF from 189 subjects had inflammatory factors, including tumor necrosis factor receptor 1,2 (TNFR1,2); transforming growth factor 1,2,3 (TGFβ1,2,3); interleukin 6,7,21 (IL6,7,21); intercellular adhesion molecule 1 (ICAM1); and VCAM1. These inflammatory markers were detected using commercially available multiplex immunoassays (Millipore Sigma, Burlington, MA) modified for CSF analyte levels.

The PAXgene Blood RNA Kit (Qiagen Inc., Valencia, CA, USA) was used to purify total RNA from the whole blood collected in a PAXgene Blood RNA Tube. The Affymetrix Human Genome U219 Array (Affymetrix, Santa Clara, CA, USA) was used for expression profiling in ADNI. The quality of gene expression data, including sample quality, hybridization, and overall signal quality, was analyzed using Affymetrix Expression Console software and Partek Genomic Suite 6.6. Raw expression values were pre-processed using the robust multichip average (RMA) normalization method.

Enrichment analysis was performed on 88 cognitively stable and 54 cognitively progressive subjects with CSD to investigate the activated pathways for cognitive progression. Gene set enrichment analysis (GSEA) [28] in R was used to screen the biological process (BP) of the Gene Ontology (GO) term through the c5.go.bp.v7.5.symbols.gmt in the Molecular Signatures Database (MSigDB) [28], the GO-Biological-Process-2018.txt in the Enrichr database as a reference gene set, and the Disease-Perturbations-from-GEO-up.txt in the Enrichr database [29] for the pathway term between the groups. A value of adjusted p < 0.05 was considered significant.

The linear mixed-effects model depicted the effects of CSD on longitudinal clinical outcomes, including cognitive function and social activity function. The model included random slope and intercept terms for each participant. The effect of neutrophils on cognitive function was demonstrated using hierarchical regression. The risk factor of CSD for cognitive progression was predicted using a time-dependent Cox proportional hazards model. To control for confounding by multiple factors, such as tau burden including CSF t-tau and p-tau, and other psychiatric symptoms including depression, appetite, aberrance, irritability, disinhibition, elation, and agitation, were all included in the Cox regression model. To eliminate the difference caused by a range of cognitive function between normal and MCI, besides demographic factors, baseline global cognition score (MMSE) was incorporated in the Cox regression model to assess the cognitive progression risk of CSD or neutrophil. A restricted cubic spline (RCS) curve was used to explore the association between blood neutrophils and cognitive outcome. RCS allows threshold identification of neutrophils on cognitive progression risk. According to the threshold of 61.33, we then divided the cohort into two subgroups (low neutrophil group < 61.633% versus high neutrophil group > 61.633%). Additionally, a Kaplan–Meier curve was used to plot the risk of cognitive progression. To determine annual change rates in cognition and hippocampal volume, we used the fitted linear mixed models with MMSE, ADAS-cog, FAQ, MoCA, ADNI-MEM, and hippocampal volumes as dependent variables and time (years from baseline) as independent variables, controlling for random intercept and slope. Then, a slope for the annual change rate was created for each subject. Longitudinal analyses were restricted to subjects with at least 3 time points.

Routine blood test of neutrophil and CSF inflammatory factors including TNFR1,2; TGFβ1,2,3; IL-6,7,21; ICAM1; and VCAM1 were compared between the normal and CSD groups using the Mann–Whitney U test. The Benjamini–Hochberg correction was used to adjust the p-value.

We used analysis of covariance (ANCOVA) to test whether CSF t-tau or p-tau positivity aggravated the association between CSD and hippocampal atrophy, after controlling for the main effects of sleep, tau burden (CSF p-tau or t-tau), age, gender, education, APOE, and intracranial volume. p-tau positivity was defined as p-tau > 21.8 pg/mL, and t-tau positivity was defined as t-tau > 245 pg/ml. The p-value was corrected using a Bonferroni correction with an α-threshold of 0.025.

Mediation analysis was used to explore what pathways mediated the link between blood neutrophils and cognitive function. Through exploratory analysis, we assigned X as the blood neutrophil percentage, M (mediator) as brain tau burden (CSF t-tau or p-tau at baseline), and Y as the outcome (cognitive function and social activity function at 2-year follow-up). We then interpreted the total effect as the amplitude of neutrophils in cognitive progression, both directly and through tau pathology intermediates. Next, we estimated the effects of the tau burden on the cognitive outcome after observing the association for a potential mediator. Age, education, gender, and APOE genotype were included as covariates, and the total, direct, and indirect effects were estimated using model 4 (mediation) of the PROCESS macro by the bruceR package in R with a bootstrapping module of 1000 iterations.

A total of 784 elderly were included in the present study, and their characteristics from ADNI are summarized in Table 1. The participants were 72.91 ± 7.04 years old, had MMSE scores of 28.31 ± 1.72, and had a moderate education of 16.29 ± 2.64 years. Male participants accounted for 55.7% of the total participants. Subjects with CSD had less education, more impaired cognitive function, and lower social activity function (p < 0.05) (Table 1, Fig. 1A1–A5). Cognitively progressive elderly with CSD tended to be APOE ε4 carriers.

A total of 784 individuals had in-person interviews at baseline and annual follow-up, with a total follow-up time of up to 168 months (n = 4253 person-times in total). Subjects with CSD displayed more longitudinal cognitive impairment and lower social activity function compared to normal sleep. There is no significant difference in the slope of the lines between the two groups (Fig. 1B1–B5).

During the follow-up period, in the CSD group, there were 35 stable normal cognitive subjects, 125 stable MCI subjects, 22 normal subjects developing MCI, and 74 MCI subjects developing dementia. In the normal sleep group, there were 224 stable normal cognitive subjects, 175 stable MCI subjects, 33 normal subjects developing MCI, 84 MCI subjects developing dementia, 1 normal subject developing dementia, and 11 MCI subjects reverting to normal. The results of the Kaplan–Meier analysis showed a significant difference in the cumulative proportion of individuals with cognitive stability between the normal sleep and CSD groups (p = 0.009) (Fig. 1C). The CSD group showed an increased risk of cognitive progression compared with the normal sleep group, according to the Cox proportional hazards model, with age, education years, gender, APOE, brain tau burden, baseline MMSE, and multiple psychiatric symptoms as covariates (Fig. 1D).

Among the CSD elderly, there were 88 cognitively stable and 54 cognitively progressive subjects. A total of 19,456 expressed genes were used for GSEA. GO biological process (BP) analysis showed that pathways related to neutrophil activation were significantly enriched in cognitively progressive subjects from the Enrichr (Fig. 2A1, A2) and MSig (Fig. 2C1, C2) databases. Disease pathway analysis showed that the AD-relevant pathway was significantly enriched in cognitively progressive subjects from the Enrichr database (Fig. 2B1, B2).

A total of 720 participants had routine blood test of neutrophils, and 771 subjects had the CSF AD biomarker test. Cognitively progressive subjects with CSD had higher blood neutrophil percentage as compared with cognitively normal subjects with CSD; however, there was no difference in blood neutrophil percentage between cognitively progressive and cognitively normal subjects without CSD (Table 1) (Fig. 3A1) (adjusted p < 0.05). There were significant differences in Aβ, t-tau, or p-tau levels between cognitively normal and cognitively progressive subjects, in both normal sleep and CSD groups (Table 1) (adjusted p < 0.05) (Table 1). Hierarchical regression demonstrated the impact of neutrophil percentage on the annual change in cognitive function, including FAQ, MMSE, ADAS-cog, and ADNI-MEM in all the subjects (Table 2).

RCS analysis was performed to further investigate the relationship between blood neutrophil and cognitive outcome, and the risk of cognitive progression increased with increasing blood neutrophil levels in CSD subjects (Fig. 3B1), with a threshold of 61.633 for blood neutrophil percentage. This such relationship was not observed in normal sleep subjects (Fig. 3B2). Individuals with higher neutrophil levels showed an increased risk of cognitive progression through the Cox regression, with age, education, sex, APOE, and baseline MMSE as covariates in CSD subjects (Fig. 3C1), but not in normal sleep subjects (Fig. 3C2). Survival analysis using the Kaplan–Meier curve revealed that subjects with higher neutrophil levels had a significant tendency for cognitive progression in all subjects (Fig. 3D). Significant correlations were also observed between blood neutrophil and annual changes in cognitive function, including MMSE, MEM, FAQ, MoCA, and ADAS-cog (adjusted p < 0.05) (Fig. 3E).

We investigated whether higher t-tau and p-tau levels exacerbated the effect of CSD on future hippocampal volume. Our findings showed a significant interaction between CSD and p-tau positivity (F = 7.847, adjusted p = 0.021) and interaction between CSD and t-tau positivity (F = 8.459, adjusted p = 0.018) on the annual change rate of left hippocampal volume in samples with higher neutrophil (Fig. 4B1, B2), but no significant results in all the samples (Fig. 4A1, A2), after controlling for the main effects of sleep, tau burden (t-tau or p-tau levels), and demographics. Furthermore, no such interaction was observed in the annual change rate of the right hippocampal volume. We found that t-tau and p-tau both mediated the association between blood neutrophil and cognitive function (ADAS-cog, MMSE, and FAQ) at 2 years follow-up. The mediation effect was considered to be partial, with the proportion of mediation ranging from 21 to 36% (Fig. 4C1–C6).

We assessed CSF inflammatory factors in 189 subjects and found that higher levels of inflammatory factors, including ICAM1, VCAM1, TNFR2, and TGFβ1, were present in cognitively progressive subjects with CSD (adjusted p < 0.05) but not in cognitively progressive subjects with normal sleep (Table 1, Fig. 5A1–A4). Other six inflammatory factors, including TNFR1; TGFβ2,3; IL-6,7,21, were not observed significantly different between the groups (Table 1). Correlation analysis showed that these four inflammatory factors were significantly associated with CSF p-tau (Fig. 5B1–B4) and t-tau (Fig. 5C1–C4) levels, especially in individuals with CSD (p < 0.05).