Plasma proteomic evidence for increased β-amyloid pathology after SARS-CoV-2 infection

Abstract

Previous studies have suggested that systemic viral infections may increase risks of dementia. Whether this holds true for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus infections is unknown. Determining this is important for anticipating the potential future incidence of dementia. To begin to do this, we measured plasma biomarkers linked to Alzheimer's disease pathology in the UK Biobank before and after serology-confirmed SARS-CoV-2 infections. SARS-CoV-2 infection was associated with biomarkers associated with β-amyloid pathology: reduced plasma Aβ42:Aβ40 ratio and, in more vulnerable participants, lower plasma Aβ42 and higher plasma pTau-181. The plasma biomarker changes were greater in participants who had been hospitalized with COVID-19 or had reported hypertension previously. We showed that the changes in biomarkers were linked to brain structural imaging patterns associated with Alzheimer's disease, lower cognitive test scores and poorer overall health evaluations. Our data from this post hoc case-control matched study thus provide observational biomarker evidence that SARS-CoV-2 infection can be associated with greater brain β-amyloid pathology in older adults. While these results do not establish causality, they suggest that SARS-CoV-2 (and possibly other systemic inflammatory diseases) may increase the risk of future Alzheimer's disease.

Fig. 1. Study overview.

a, Experimental design. Protein concentrations were assayed from plasma samples acquired from the UK Biobank imaging assessment visits, the second of which was specifically recruited for the study of COVID-19. b, Distribution of participant ages at the pandemic assessment. c, Sources of evidence for case selection. Antibody, home-based lateral-flow SARS-CoV-2 antibody test; Antigen, PCR antigen (swab) test; Health records, GP and/or hospital records. d, Distribution of pre-pandemic assessment visit dates. e, Distribution of pandemic assessment visit dates. f, Distribution of intervals between assessments. g, Estimated dates of COVID symptoms (from participants with antigen test results). Figure created with BioRender.com.

Fig. 2. Standardized regression model parameter estimates for the case–control term in each of the AD protein change models.

The parameters represent the SARS-CoV-2-infection-associated effect on change in protein levels between pre-pandemic and pandemic assessment sessions. Aβ42:Aβ40 ratios showed significant relative reductions in individuals with SARS-CoV-2. P values correspond to one-sided t-tests for differences corresponding to previously reported associations with AD. Full model fits are in Supplementary Tables 2 and 3 (basic and extended models). **FDR significant.

Fig. 3. Standardized regression model parameter estimates for models including age-dependent vulnerability terms in each of the protein biomarker change models.

a, Parameters for the age-dependent vulnerability term. b, Parameters for the age-dependent vulnerability term interaction with case–control status. c, Age-dependent vulnerability term weightings. P values correspond to one-sided t-tests for differences corresponding to previously reported associations with AD. Full model fits are in Supplementary Tables 4 and 5 (basic and extended models). *P < 0.05; **FDR significant.

Fig. 4. Time plots of plasma protein levels (pg ml⁻¹) of SARS-CoV-2 cases.

a, Aβ40. b, Aβ42. c, Aβ42:Aβ40 ratio. d, pTau-181. Plots show the 7 year rolling means for control (blue) and case (orange) groups with 90% confidence intervals on the mean shown (shaded areas).

Fig. 5. Associations between plasma protein biomarker levels and comorbidities and other factors as determined from regression models across all cases and controls.

The shadings show the magnitudes of standardized betas for particular factors, estimated in models with additional variates controlling for age, sex and time interval between assessments. The P values correspond to two-sided t-tests. All maps masked at |P uncorrected| = 0.20. *P uncorrected < 0.05; **FDR significant. a, Associations within the baseline assessment data only. The red colors represent positive associations. b, Associations of factors with the change in plasma protein biomarker levels across assessments. The red colors represent a positive association of factors with higher levels of protein biomarker in the pandemic assessment session. c, Interaction of association from b with case–control status (case = +1). The red colors represent a greater positive association in cases.

Extended Data Fig. 1. CONSORT Diagram.

Schematic summary of data availability for the study. Primary analyses focus on 624 individually matched cases and controls from the COVID19 imaging repeat study with SIMOA plasma proteomic data. Full details of participant selection and control matching for the COVID study can be found at https://biobank.ndph.ox.ac.uk/ukb/ukb/docs/casecontrol_covidimaging.pdf.