Dissection of blood-brain barrier dysfunction through CSF PDGFRβ and amyloid, tau, neuroinflammation, and synaptic CSF biomarkers in neurodegenerative disorders

Abstract

Background: Blood-brain barrier (BBB) dysfunction is an early event in neurodegenerative disorders. Pericytes are key cells for BBB maintenance. Upon pericyte injury, the platelet-derived growth factor receptor-β (PDGFRβ) is released in the cerebrospinal fluid (CSF). The relation of CSF PDGFRβ with markers of amyloid pathology, neuroinflammation, and axonal and synaptic damage across dementia remains unclear.

Methods: Retrospectively, we quantified CSF PDGFRβ and CSF core Alzheimer's disease (AD), astrocytic (GFAP), microglial (sTREM 2, YKL-40), axonal (NfL), and synaptic (GAP-43, neurogranin) biomarkers in 210 patients from the Cognitive Neurology Centre, Paris, France, including n = 23 neurological controls (NC), n = 84 patients with mild cognitive impairment (MCI) [AD, n = 41; non-AD, n = 43], and n = 103 patients with dementia (AD, n = 73; non-AD, n = 30).

Findings: Comparing clinical stages, CSF PDGFRβ levels were increased at the MCI stage (Cohen's d = 0.55 [CI_95% 0.066, 1.0], P = 0.025) compared with NC. Non-AD MCI displayed higher levels than controls (Cohen's d = 0.74 [CI_95% 0.22, 1.3], P = 0.042). No association was observed with CSF Aβ42/Aβ40 ratio but with p-tau 181 (β = 0.102 [CI_95% 0.027, 0.176], P = 0.0080) and t-tau levels (β = 0.133 [0.054, 0.213], P = 0.0010). CSF PDGFRβ levels were positively associated with CSF neuroinflammation and synaptic markers levels. Higher CSF PDGFRβ levels were associated with lower MMSE scores at MCI (β = -1.23 [CI_95% -2.33, -0.260], P = 0.015) and dementia stages (β = -2.24 [CI_95% -3.62, -0.85], P = 0.0020). CSF neuroinflammation biomarkers mediated the association of CSF PDGFRβ with neurodegeneration and synaptic integrity markers.

Interpretation: CSF PDGFRβ, a candidate biomarker of BBB dysfunction, is increased in the early stages of neurodegenerative disorders, in association with neuroinflammation and axonal and synaptic damage.

Funding: Association des Anciens Internes des Hôpitaux de Paris, Edmond de Rothschild Program, Fondation Vaincre Alzheimer, Demensförbundet, Gamla Tjänarinnor, Anna-Lisa och Bror Björnssons Stiftelse.

Keywords: Alzheimer's disease; Brain blood barrier; CSF biomarkers; Neurodegenerative disorders; Neuroinflammation; PDGFRβ.

Fig. 1

Association of CSF PDGFRβ and Q-Alb with clinical syndromes and diagnosis. a, CSF PDGFRβ levels across syndrome groups; b, CSF PDGFRβ levels across diagnosis groups; c, Q-Alb levels across syndrome groups; d, Q-Alb levels across diagnosis groups. P-values were obtained through one-way ANCOVA adjusting for age, sex, and levels of education, followed by post hoc Tukey's test, adjusting for multiple comparisons (n = 210). Significant differences (P < 0.05) are reported. Box and whiskers plots with the central line denoting the median value and the box containing the 25th to 75th percentile values. Association of CSF PDGFRβ with Q-Alb, e, in the whole sample; f, neurological controls; g, MCI; and h, dementia group. The association between biomarkers was studied using linear regression, adjusting for age and sex (n = 207). Individual points and the regression line are displayed with a shaded area above and below the line, representing the upper and lower bounds of the 95% confidence interval. Results are presented as unstandardised β (95% confidence interval) and P-values.

Fig. 2

Association of CSF PDGFRβ with CSF AD biomarkers. a, CSF PDGFRβ levels across AT(N) groups, compared using one-way ANCOVA adjusting for age and sex (n = 210). Box and whiskers plots with the central line denoting the median value and the box containing the 25th to 75th percentile values. b-e, Association of CSF PDGFRβ with core AD biomarkers, including b, CSF Aβ42; c, CSF Aβ40; d, CSF Aβ42/Aβ40; e, CSF p-tau 181; and f, CSF t-tau. The association between biomarkers was studied using linear regression, adjusting for age and sex (n = 210). Individual points and the regression line are displayed with a shaded area above and below the line, representing the upper and lower bounds of the 95% confidence interval. Results are presented as unstandardised β (95% confidence interval) and P-values.

Fig. 3

Association with axonal, synaptic, and neuroinflammation CSF markers. a-d Association of CSF PDGFRβ with CSF GFAP: a, in the whole cohort; b, in the NC group; c, in the MCI group; d, in the dementia group. e-h Association of CSF PDGFRβ with CSF sTREM 2: e, in the whole cohort; f, in the NC group; g, in the MCI group; h, in the dementia group. i-l Association of CSF PDGFRβ with CSF YKL-40: i, in the whole cohort; j, in the NC group; k, in the MCI group; l, in the dementia group. m-p Association of CSF PDGFRβ with CSF NfL: m, in the whole cohort; n, in the NC group; o, in the MCI group; p, in the dementia group. q-t Association of CSF PDGFRβ with CSF neurogranin: q, in the whole cohort; r, in the NC group; s, in the MCI group; t, in the dementia group. u-x Association of CSF PDGFRβ with CSF GAP-43: u, in the whole cohort; v, in the NC group; w, in the MCI group; x, in the dementia group. Linear regressions adjusting on age, sex, and CSF Aβ42/40 ratio (n = 210). Individual points and the regression line are displayed with a shaded area above and below the line, representing the upper and lower bounds of the 95% confidence interval. Results are shown as unstandardised β estimate (95% confidence interval), P-value.

Fig. 4

Mediation analysis of the effect of PDGFRβ on neurodegeneration and synaptic integrity. Mediation analysis exploring: a, CSF YKL-40 as a mediator of the relationship between CSF PDGFRβ and CSF t-tau; b, CSF sTREM 2 as a mediator of the relationship between CSF PDGFRβ and CSF t-tau; c, CSF YKL-40 as a mediator of the relationship between CSF PDGFRβ and CSF GAP-43; d, CSF sTREM2 as a mediator of the relationship between CSF PDGFRβ and CSF GAP-43. Mediation effects are reported as the unstandardised estimate β (95% CI), percentage of the total effect for n = 210. Age, sex and APOE4 status were added as covariates.