Background
Alzheimer’s disease (AD) may be described as a biological continuum that includes the hallmark pathological processes of amyloid-beta (Aβ) dysmetabolism, formation of amyloid deposits (A), neurofibrillary tangles (T), neurodegeneration (N), determined by measuring cerebrospinal fluid (CSF) levels of Aβ42, phosphorylated tau (P-tau), and total tau (T-tau) respectively. The presence or absence of pathological markers can be summarized as an A/T/N score, an unbiased classification of pathology and severity along the AD continuum [
1,
2]. In contrast, the clinical classification of the AD continuum is based on subjective accounts of cognitive deficits, performance on cognitive tests, and functioning in daily life [
3‐
6]. Patients who report experience of decline in cognitive function while performing within the normal range on cognitive tests, may be categorized as having subjective cognitive decline (SCD) [
3]. In contrast, mild cognitive impairment (MCI) requires the presence of subjective cognitive decline in combination with impaired cognitive performance yet retaining preserved independence in functional ability [
4‐
6]. We and others have made large efforts towards standardization of criteria for these stages, e.g., as part of the EU JPND-funded BIOMARKAPD study, and Norwegian national efforts [
7,
8].
While genetic evidence indicates that Aβ dysmetabolism is causal in familial AD, the initial sequence of events and causality in sporadic AD is still not determined. However, reduced Aβ clearance and deficient innate immune activity related to the triggering receptor expressed on myeloid cells 2 (TREM2) and clusterin (Apo J) function may play a role [
9‐
11]. While central nervous system (CNS) interstitial Aβ is released from neurons dependent on activity, clearance is a result of neuronal, astro-, and microglial uptake and degradation as well as transport to the glymphatic system, blood, and (CSF) [
12‐
14]. Microglia normally subserve synaptic homeostasis and synapse elimination [
15,
16]. They are CNS myeloid-derived innate immune effector cells, which together with reactive astrocytes also may acquire inflammatory properties. Genetic evidence supports a role for loss of balanced TREM2 activation, innate immunity, and microglial activity in AD pathogenesis [
9,
17‐
19]. Further, experimental studies support neuroinflammatory responses as drivers of AD pathogenesis, and there is evidence for associations to neuroinflammation and deficient microglia Aβ function in MCI due to AD and more advanced AD [
14,
20‐
22], though the initial microglial activation might be compensatory and advantageous. Aβ clearance decreases with age and could in combination with genetic liabilities for compromised innate immune clearance capacity contribute to age-related disease inception [
23,
24]. Notably, a recent translocator protein (TSPO) ligand positron-emission tomography (PET) study detecting activated microglia showed higher binding in AD “slow decliners” [
25]. Moreover, a longitudinal TSPO-PET study demonstrated reduced microglia activation over time in patients at the MCI stage, but increased activation in patients at the AD stage of dementia [
26]. These findings may be interpreted as an early beneficial role of microglial activation and a later inflammatory peak. Experimental evidence suggests that TREM2 increases in parallel with amyloid deposition, possibly limiting Aβ plaque-associated pathology [
27,
28]. Thus, initial microglial activation might induce phagocytosis of Aβ, stalling formation of oligomers, and restricting neurotoxicity from deposited Aβ in plaques, while further inflammatory activation might accelerate neurodegeneration. If supported, this distinction could aid patient stratification and guide intervention trials that include immune modification components.
Glial activation occurs as part of altered immune cytokine activities, which also change towards increased inflammatory activity during AD progression. However, micro- and astroglial activation are interlinked, and genetic evidence suggests that innate immunity could be a prime mover in the AD cascade [
9]. Based on the described findings of early microglial activation, our starting point was to investigate these events in CSF samples via soluble TREM2 (sTREM2) as a microglial activation marker, and clusterin and chitinase-3-like protein 1 (YKL-40) which both are suggestive of astroglial activation, a marker for neuron-microglia communication (chemokine ligand 1; CX3CL1; fractalkine) and a well-established marker for microglial mobilization and inflammatory reaction (monocyte chemoattractant protein 1, MCP-1).
Soluble TREM2 is released upon microglial activation, leading to increased levels of CSF sTREM2 in AD [
29,
30]. This receptor might subserve Aβ uptake by peptides being bound to its ligands APOE and clusterin [
31‐
34]. Clusterin is abundantly expressed by astrocytes and select neuronal populations, e.g., within the hippocampus, and may modulate Aβ metabolism as a chaperone protein [
35]. In binding Aβ, clusterin may increase clearance and inhibit plaque formation in processes that are coupled to immune responses [
35‐
37]. YKL-40 is produced mainly by astrocytes, but also microglia, often surrounding amyloid plaques. While early expression levels vary, increased expression has been reported at the MCI stage associated with neuroinflammation [
38,
39]. Experimental data suggest a role for YKL-40 in microglia-astroglia crosstalk [
38,
40]. Fractalkine is a CXC chemokine (CX3CL1) that is highly expressed by neurons in the hippocampus and cortex, while its receptors (CX3CR1) are found on microglia [
41]. Fractalkine neuron-to-microglia communication strengthens the neuroprotective role of microglia, by inhibiting TNFα secretion [
42], reducing neurotoxicity, and reducing microglial activation [
43,
44]. The expression level of fractalkine has been reported to reflect progression of AD [
45]. MCP-1 is a CC chemokine produced by micro- and astroglia and endothelial cells with receptors (CCR2) largely restricted to immune cells but also found on neurons. In the brain, MCP-1 attracts microglial and peripheral immune cells to sites of inflammation. It may stimulate microglia to change from resting to activated morphology, and the level of CSF MCP-1 increases with advancing pathology in AD [
46].
These individual markers have been studied in predementia and in AD dementia stages with variable reported findings (see Additional file
1: Table S1). To our knowledge, none of the included CSF immune markers (sTREM2, MCP-1, YKL-40, fractalkine, and clusterin) have been studied in a defined SCD group; however, both CSF sTREM2 and YKL-40 have been studied in preclinical AD (SCD cases and asymptomatic subjects) with pathological (low) CSF Aβ. Neither sTREM2 nor YKL-40 was reportedly increased in this mixed group [
29,
30,
38,
40]. Clusterin and fractalkine have been little studied in MCI [
47,
48], but sTREM2, MCP-1, and YKL-40 have all shown contradictory results, either unchanged [
49‐
52] or increased [
29,
30,
38,
40,
53,
54] compared to controls. Except for YKL-40 [
38,
51] and fractalkine [
47] which respectively have been found unchanged or reduced compared to controls, all the other immune markers have shown contradictory results in AD dementia compared to controls, either unchanged [
29,
49,
51,
55‐
59], reduced [
60,
61], or increased CSF values [
38,
50,
53,
54,
62‐
65].
Intrathecal levels of glial- and inflammation markers may reflect both CNS AD pathogenic processes and responsivity, as well as inflammatory reactivity upon stimulation such as therapeutic interventions and infectious agents. To our knowledge, CSF sTREM2, MCP-1, YKL-40, clusterin, and fractalkine have never been analyzed in the same cohorts across predementia AD stages. Thus, we currently lack information on putative disparate or concerted micro- and astroglial patterns of activation and inflammation related to clinical and neuropathological changes in predementia AD. As activation may be bi- or multiphasic along the AD continuum, highly standardized protocols and measurements on standardized platforms, tightly controlled clinical staging, and biomarker-based stratification may be necessary to detect relevant differences.
Microglial activation per se does not need to be inflammatory, but may be a compensatory response at the synapse. Following Fan et al. [
26], we hypothesize that the earliest stage of demonstrable microglial activation occurs at the pre-clinical stage, only coincident with other inflammatory and astroglial activation markers at later stages. We also explore relations between CSF biomarker-derived A/T/N stages and glial activation markers.
Methods
Subjects
For the purposes of the present study, we selected 121 participants from two Norwegian cohorts. Healthy controls with normal CSF (
n = 36), participants with SCD (
n = 18), and MCI (
n = 20) patients, both with CSF Aβ42 confirmed amyloid pathology, were selected from the Norwegian multicenter study, “Dementia Disease Initiation” (DDI) [
7]. A patient group meeting the National Institute on Aging–Alzheimer’s Association (NIA-AA) criteria for dementia due to AD [
6] (
n = 27) and an additional 19 MCI patients with CSF Aβ42 confirmed amyloid pathology were included from the Norwegian part of the Gothenburg-Oslo MCI (MCI-GO) cohort [
66]. Classification of A/T/N groups [
1] was done using CSF Aβ42 (A), phosphorylated tau (P-tau) (T), and total-Tau (T-tau) (N). All subjects were assigned binary scores for each category, rated positive when the CSF biomarker value was defined as pathological. The cut-off for CSF was Aβ42 < 708 pg/ml for amyloid plaque pathology, subsequently denoted Aβ-positive (Aβ+) and A+ cases. This Aβ optimal cut-off was determined by a PET [
18F]-Flutemetamol uptake study [
67]. Cases with Aβ42 values close to cutoff (± 30 ng/ml) were excluded from this study material. The abnormality cut-off values for CSF T-tau and P-tau were set in accordance with reference values from Sjögren et al. [
68]. For P-tau, the cut-off value was ≥ 80 pg/ml, and values above this threshold were classified as a T+ score. For T-tau, cut-off values were > 300 pg/ml for age < 50 years, > 450 pg/ml for age 50–69 years, and > 500 pg/ml for age ≥ 70 years. Subjects were denoted N+ cases when their T-tau value exceeded the respective thresholds.
Further criteria for inclusion were age between 40 and 80 years and a native language of Norwegian, Swedish, or Danish. Exclusion criteria were brain trauma or disorders, including clinical stroke, dementia, severe psychiatric disorder, and severe somatic disease that might influence the cognitive functions, intellectual disability, or other developmental disorders.
Both DDI and MCI-GO employ a standardized protocol for participant selection, assessment, and disease stage classification according to published criteria [
3,
4,
6]. All patients were interviewed and examined by a physician trained in diagnosing cognitive disorders. They all underwent cognitive testing, either cerebral MRI or CT, blood screening, and standard lumbar puncture as part of the clinical assessment.
Classification of SCD and MCI
Participants were classified as SCD according to the SCD-I framework, which requires normal objective cognitive performance on neuropsychological tests while experiencing a subjective decline in any cognitive domain [
3]. MCI was classified according to the NIA-AA criteria which require the presence of subjective cognitive impairment or decline in combination with lower performance than expected in one or more cognitive domains, yet preserved independence in functional ability and not fulfilling the criteria of dementia [
4,
6]. Performance was classified as normal or abnormal according to published norms (adjusted for age, sex, and educational effects) for the different tests [
69‐
71]. Due to mutually exclusive criteria, the cut-off values for SCD vs. MCI (defined as normal or abnormal cognition) were ≤ 1.5 standard deviation below normative mean on either Consortium to Establish a Registry for AD (CERAD) word list (delayed recall), Visual Object and Space Perception (VOSP) silhouettes, Trail Making Test part B (TMT-B), or Controlled Oral Word Association Test (COWAT). For the DDI cohort global cognitive status was also assessed by the Clinical Dementia Rating Scale (CDR), whereas the Global Deterioration Scale was used for MCI-GO [
72,
73].
CSF collection and handling
Lumbar punctures were performed similarly on four sites all following a detailed BIOMARKAPD SOPs as described previously [
8]. Briefly described, sampling was done before noon and CSF was collected in polypropylene tubes (Thermo Fisher Scientific, MA, USA) which were centrifuged within 4 h at 2000
g for 10 min at room temperature. The supernatant was subsequently transferred to new defined tubes, directly frozen at − 80 °C on site and kept at − 80 °C until thawed for analysis. All CSF samples were analyzed either at the Department of Interdisciplinary Laboratory Medicine or Section of Clinical Molecular Biology (EpiGen) at Akershus University Hospital. The exception was the sTREM2 analysis, which was assayed at the Department of Pharmacology at the University of Oslo.
Protein biomarker measurements
Commercial enzyme-linked immunosorbent assays (ELISAs) based on monoclonal antibodies were used to measure CSF levels of the following protein biomarkers: Aβ42, T-tau, and P-tau. They were determined using Innotest β-Amyloid (1-42), Innotest T-tau Ag, and Innotest P-tau (181P)(Fujirebio, Ghent, Belgium), respectively.
CSF sTREM2 was also analyzed using a sandwich ELISA as described earlier [
49] with some modifications; the plates were coated over night with the capture antibody (0.25 μg/ml; AF1828, R&D Systems, MN, USA) and samples incubated for 2 h prior to TREM2 detection with a rabbit-monoclonal anti-human TREM2 antibody (0.5 μg/ml; SEK11084, Sino Biologics, Beijing, China).
The QuickPlex SQ 120 system from Meso Scale Discovery (MSD, MD, USA) was used to measure YKL-40, MCP-1, and fractalkine in a U-plex format and clusterin in an R-plex format, where YKL-40 and clusterin were in a singleplex setup and MCP-1 and fractalkine were in the same multiplex setup. The MSD analyses were carried out according to the manufacturers’ procedures, with the adjustments that CSF samples were diluted 200 times prior to YKL-40 and clusterin analyses, and the multiplex setup was used with 100 μl neat CSF and 25 μl buffer.
All the lower limits of quantifications (LLOQs) were defined as the lowest concentration at which the coefficient of variation (CV) of the calculated concentration was < 20% in > 75% of the analyses or the mean CV was < 20% in our test set. All biomarker values in all samples were well above LLOQ. All samples were analyzed in duplicates and reanalyzed if relative deviations (RDs) exceeded 20%. In addition, quality control samples with RD threshold of 15% assured inter-plate and inter-day variation.
Statistical analysis
Normality was assessed through the inspection of Q-Q plots, histograms, and the Shapiro-Wilks test of normality.
In order to explore and adjust for age and sex, and APOE-ɛ4 allelic effects on CSF inflammatory markers in healthy aging, simple and multiple regression analyses (controlling for several covariates) were performed between these variables and CSF immune markers within the healthy control group. If a significant relationship was observed between these covariates and an inflammatory marker in the healthy control group, the standardized residuals from the pertinent regression model was obtained for the entire sample and used in further analysis in order to adjust for these covariates in between-group comparisons. To assess differences in biomarker levels between groups, we performed one-way ANOVAs with planned comparisons for variables with normal distributions. We performed Kruskal-Wallis test with Dunn’s non-parametric pairwise post hoc test with Bonferroni corrections to assess group differences in variables with non-normal distributions (CSF Aβ42, CSF T-tau, CSF P-tau, MMSE, and A/T/N groups). Non-parametric pairwise comparisons and ANOVA contrasts were performed in a hierarchical manner. We compared Aβ + SCD, MCI, and AD dementia groups to healthy controls, and finally we compared the SCD with the MCI group and both SCD and MCI to the AD dementia group. The dichotomous variable “sex” was assessed using a chi-square test. For the A/T/N groups, A−T−N− and A+T−N− were compared to all other groups. Only one patient had A+T+N− classification, and this patient was excluded from both statistical analysis and figure.
To assess clinical stage dependent relationships between the innate immune response to AD pathology, correlational analyses between the inflammatory markers (sTREM2, YKL40, MCP-1, fractalkine, and clusterin) and CSF AD biomarkers (Aβ42, T-Tau, and P-tau) were performed using Pearson’s r within the pertinent symptomatic groups (SCD, MCI, and AD dementia).
All analyses were performed in the Statistical Package for Social Sciences (SPSS) version 25, and the significance level was defined as p < .05.
Ethics
The regional medical research ethics committee approved this study. Participants gave their written informed consent before taking part in the study. All further study conduct was in line with the guidelines provided by the Declaration of Helsinki (1964; revised 2013) and the Norwegian Health and Research Act (2009).