Plasma proteins related to inflammatory diet predict future cognitive impairment

Abstract

Dysregulation of the immune system and dietary patterns that increase inflammation can increase the risk for cognitive decline, but the mechanisms by which inflammatory nutritional habits may affect the development of cognitive impairment in aging are not well understood. To determine whether plasma proteins linked to inflammatory diet predict future cognitive impairment, we applied high-throughput proteomic assays to plasma samples from a subset (n = 1528) of Women's Health Initiative Memory Study (WHIMS) participants (mean [SD] baseline age, 71.3 [SD 3.8] years). Results provide insights into how inflammatory nutritional patterns are associated with an immune-related proteome and identify a group of proteins (CXCL10, CCL3, HGF, OPG, CDCP1, NFATC3, ITGA11) related to future cognitive impairment over a 14-year follow-up period. Several of these inflammatory diet proteins were also associated with dementia risk across two external cohorts (ARIC, ESTHER), correlated with plasma biomarkers of Alzheimer's disease (AD) pathology (Aβ_42/40) and/or neurodegeneration (NfL), and related to an MRI-defined index of neurodegenerative brain atrophy in a separate cohort (BLSA). In addition to evaluating their biological relevance, assessing their potential role in AD, and characterizing their immune-tissue/cell-specific expression, we leveraged published RNA-seq results to examine how the in vitro regulation of genes encoding these candidate proteins might be altered in response to an immune challenge. Our findings indicate how dietary patterns with higher inflammatory potential relate to plasma levels of immunologically relevant proteins and highlight the molecular mediators which predict subsequent risk for age-related cognitive impairment.

Fig. 1. Study design and participant selection.

Olink® Inflammation and Immune Response Panels were applied to blood samples collected at baseline for WHIMS participants. The study sample was recruited from the larger Women’s Health Initiative (WHI) study, where enrollment began in 1996–1999. Participants continued annual cognitive assessments in-person until 2007–08, when the study was extended and cognitive assessments were continued via telephone. Cognitive status was assessed using standard neuropsychological testing coupled with central adjudication by a panel of expert clinicians. DII Dietary Inflammatory Index, WHIMS Women’s Health Initiative Memory Study.

Fig. 2. Association of inflammatory/immune proteins and pathways with Energy-Adjusted Dietary Inflammatory Index (EDII).

A Volcano plot showing correlations between EDII and 151 individual protein levels. Proteins above the dashed horizontal line were statistically significant (p < 0.05). B Heatmap showing 55 individual proteins significantly correlated with EDII, as well as corresponding associations with Age and unadjusted DII scores. C Lines of best fit illustrating significant correlations between EDII and immune/inflammatory pathway composite scores. Results derived from partial Spearman correlations adjusted for age. DII Dietary Inflammatory Index, EDII Energy adjusted Dietary Inflammatory Index.

Fig. 3. Inflammatory diet proteins and pathways related to incident cognitive impairment, validation of proteins in external cohorts and associations with neurodegenerative disease biomarkers.

Volcano plots showing the associations of (A) inflammatory diet proteins and (B) pathways with odds of incident cognitive impairment derived from multivariable logistic regression models. Proteins and pathways above the dashed horizontal line were statistically significant (p < 0.05). Models adjusted for age, APOE ε4 status, education, obesity, diabetes, hypertension, eGFR-creatinine, geographical region, hormone treatment. C Forest plot showing significant associations (shaded shapes, p < 0.05) of inflammatory diet proteins with risk for cognitive impairment (WHIMS) and dementia (ESTHER), and meta-analyzed results (WHIMS + ESTHER). Results derived from multivariable logistic regression models adjusted for covariates used in WHIMS (see above) and ESTHER (age, APOE ε4 status, education, BMI, diabetes, cardiovascular disease, depression, physical activity, sex). D Forest plot showing associations of inflammatory diet proteins with time to dementia onset in the ARIC Study. Results derived from Cox proportional hazards regression models adjusted for covariates in ARIC (age, center-race, sex, education, APOE ε4 status, diabetes, BMI, smoking status, hypertension, and eGFR-creatinine). Two plasma proteins associated with cognitive impairment in WHIMS (NFATC3, ITGA11) were not measured in ESTHER or ARIC. A-D: Hazards and Odds ratios represent differences in risk per each log2 increase in protein level. E Heatmap showing correlations and corresponding coefficients between candidate proteins and plasma biomarkers of brain amyloid-β (Aβ_42/40) and neurodegeneration (NfL). *Correlation statistically significant (p < 0.05). Results derived from partial Spearman correlations adjusted for age. ARIC Atherosclerosis Risk in Communities, EDII Energy adjusted Dietary Inflammatory Index, ESTHER Epidemiological Study on the Chances of Cure, Early Detection and Optimized Therapy of Chronic Diseases in the Elderly Population.

Fig. 4. Associations of inflammatory diet proteins with the SPARE-AD, an MRI-defined index of neurodegenerative brain atrophy.

Scatterplots and lines of best fit show levels of inflammatory diet proteins identified in WHIMS, A OPG, B CCL3, C CDCP1, significantly related to SPARE-AD scores in BLSA participants, as evidenced by multivariable linear regression models adjusted for age, sex, race, education, APOEε4, eGFR-creatinine and a comorbidity index. Density plots along x-and y-axes display distributions of inflammatory diet proteins and SPARE-AD scores, respectively. CCL3 C-C motif chemokine 3, CDCP1 CUB Domain Containing Protein 1, OPG Osteoprotegerin, RFU Relative Fluorescence Intensities, SPARE-AD Spatial Pattern of Atrophy for Recognition of Alzheimer’s Disease.

Fig. 5. Characterization of inflammatory diet proteins linked to incident cognitive impairment.

A Consensus transcript expression levels (normalized Transcripts per Million; nTPM) across immune tissues of interest based on transcriptomics data from the Human Protein Atlas and Genotype-Tissue Expression project. B Results from nine different RNA-seq analyses across four different immune cell types which reported differential expression of genes encoding candidate proteins in response to treatment with lipopolysaccharide (LPS), an immune challenge. C Protein-protein interaction network analyses were generated using STRING^©; predicted protein conformations are depicted in each protein’s circular node. GO enrichment analysis for biological processes and molecular functions utilized PANTHER™ functional classification platform. D Summary of evidence, including results linking candidate proteins to data obtained from AMP®-AD. Accelerating Medicines Partnership for Alzheimer’s Disease AMP®-AD, eQTL expression quantitative trait loci, DEG differentially expressed gene, LPS lipopolysaccharide; normalized Transcripts per Million, nTPM Protein Analysis Through Evolutionary Relationships Classification System, PANTHER™ Search Tool for the Retrieval of Interacting Genes/Proteins, STRING^©.