Serum IL-6, sAXL, and YKL-40 as systemic correlates of reduced brain structure and function in Alzheimer's disease: results from the DELCODE study

Abstract

Background: Neuroinflammation constitutes a pathological hallmark of Alzheimer's disease (AD). Still, it remains unresolved if peripheral inflammatory markers can be utilized for research purposes similar to blood-based beta-amyloid and neurodegeneration measures. We investigated experimental inflammation markers in serum and analyzed interrelations towards AD pathology features in a cohort with a focus on at-risk stages of AD.

Methods: Data of 74 healthy controls (HC), 99 subjective cognitive decline (SCD), 75 mild cognitive impairment (MCI), 23 AD relatives, and 38 AD subjects were obtained from the DELCODE cohort. A panel of 20 serum biomarkers was determined using immunoassays. Analyses were adjusted for age, sex, APOE status, and body mass index and included correlations between serum and CSF marker levels and AD biomarker levels. Group-wise comparisons were based on screening diagnosis and routine AD biomarker-based schematics. Structural imaging data were combined into composite scores representing Braak stage regions and related to serum biomarker levels. The Preclinical Alzheimer's Cognitive Composite (PACC5) score was used to test for associations between the biomarkers and cognitive performance.

Results: Each experimental marker displayed an individual profile of interrelations to AD biomarkers, imaging, or cognition features. Serum-soluble AXL (sAXL), IL-6, and YKL-40 showed the most striking associations. Soluble AXL was significantly elevated in AD subjects with pathological CSF beta-amyloid/tau profile and negatively related to structural imaging and cognitive function. Serum IL-6 was negatively correlated to structural measures of Braak regions, without associations to corresponding IL-6 CSF levels or other AD features. Serum YKL-40 correlated most consistently to CSF AD biomarker profiles and showed the strongest negative relations to structure, but none to cognitive outcomes.

Conclusions: Serum sAXL, IL-6, and YKL-40 relate to different AD features, including the degree of neuropathology and cognitive functioning. This may suggest that peripheral blood signatures correspond to specific stages of the disease. As serum markers did not reflect the corresponding CSF protein levels, our data highlight the need to interpret serum inflammatory markers depending on the respective protein's specific biology and cellular origin. These marker-specific differences will have to be considered to further define and interpret blood-based inflammatory profiles for AD research.

Keywords: Alzheimer’s disease; Biomarker; Blood-based; Cognition; Inflammation; Structural MRI.

Fig. 1

Serum biomarker correlation matrix. Partial Spearman correlation matrix adjusted for covariates age, sex, BMI, and APOE status. Serum biomarker levels (yellow) were correlated against their CSF counterparts (CSF: serum), to CSF AD biomarkers (gray, Aβ42/40 ratio, p-tau-181, t-tau, neurogranin, and FABP-3), as well as to plasma levels of Nf-L determined in DELCODE (red). The upper panel depicts the strength of correlation (Spearman r), and the lower panel, significance (Spearman p). Between CSF and serum, several proteins showed weak to modest correlations, whereas others were not correlated at all. The strongest correlation was observed between serum and CSF CRP protein. FABP-3, ApoE, sAXL, and complement C4 showed weak correlations to amyloid. FABP-3 and YKL-40 were consistently correlated to CSF levels of tau isoforms and neurodegeneration markers. Plasma levels of Nf-L were correlated to several proteins from the serum panel, even if the respective proteins did not relate to their CSF counterparts or to CSF AD markers

Fig. 2

Elevation of serum sAXL in biomarker-positive DAT subjects. Differences of sAXL serum levels between subject groups based on screening diagnosis or CSF biomarker profile. Violin plots with median and interquartile range. Groups colored red were elevated against groups colored blue. Gray color indicates indifferent groups. Further statistical details are described in Additional file 1: Tables S2-S4. Serum sAXL levels were elevated in DAT versus SCD or HC subjects (A), as well as in either p-tau-based A+T+ subjects (B) or t-tau-based A+N+ subjects (C) against A−T− or A−N− subjects, respectively. There was no significant elevation in either MCI subjects or those with only amyloid or tau isoform-positive CSF biomarker profile

Fig. 3

Relations of serum biomarkers to structural MRI. A Bivariate Spearman partial correlation matrix calculated for serum markers versus Braak stage composite scores of brain structure, adjusted for age, sex, BMI, and APOE status (yellow box). Correlation matrices (left panel) show the strength (rho) of all correlations that were significant at uncorrected p value < 0.05 (right panel). Results for CSF AD markers (Aβ42/40 ratio, tau) are depicted for comparison purposes (gray box). The ratio Aβ42/40 was inverted to visualize how more pathological amyloid levels relate to lower structural scores. All significant serum markers showed negative relations to structure, of which FABP-3, YKL-40, and IL-6 were most consistent. Aside from these, only sAXL and C1q were related to more than one Braak region. Several correlations were also observed for single markers in Braak stage I, consistent with results of a multiple regression, using serum inflammation markers and predicting structure, significant only for the Braak stage I ROI, explaining 12% of the variance in this region. p values of “0” in the right panel denote p < 0.001. B Freesurfer-region specific, bivariate Spearman relations of FABP-3, sAXL, YKL-40, and IL-6. Correlation strength (rho), at uncorrected p > 0.05 for illustration purposes, is plotted rendered on the brain surface. All correlations were adjusted for age, sex, BMI, and APOE status

Fig. 4

Serum biomarker relations to the longitudinal outcome. Serum biomarker levels were correlated against longitudinal PACC score values and plotted as one standard deviation (SD) above and below the mean standardized biomarker level. Available subject numbers throughout follow-ups were as follows: baseline, N = 289; Y1, 245; Y2, 213; Y3, 164; Y4, 89; Y5, 20. Correlations are displayed with confidence intervals. Results are adjusted for age, sex, BMI, and APOE status. Serum FABP-3 did not differ at baseline, but higher baseline levels predicted cognitive decline at follow-up (A, p = 0.003). In contrast, sAXL (B) did not predict change over time, but differed at baseline, with higher levels in subjects with stronger cognitive impairment (p = 0.003)