Introduction
Alzheimer’s disease (AD) is a debilitating neurodegenerative condition characterized by the presence of extracellular amyloid-β (Aβ) plaques, intracellular tau-containing neurofibrillary tangles (NFTs), neurodegeneration, and cognitive impairment [
1,
2]. Traditionally, hallmarks of AD have been measured in vivo using positron emission tomography (PET) imaging or cerebral spinal fluid (CSF) measurement. Biomarkers of AD pathology as well as non-specific markers of neurodegeneration are now measurable in blood using ultrasensitive assays, providing a cost-effective, scalable, and minimally invasive alternative to PET and CSF measurement. Several studies demonstrate that plasma biomarkers of AD pathology (e.g., Aβ
42/40, phosphorylated tau [pTau-181]), astrogliosis (glial fibrillary acidic protein [GFAP]), and to a lesser extent, neuronal injury (neurofilament light chain [NfL]) relate to – and can be used to predict – cortical Aβ and tau burden [
3‐
6]. Further, plasma pTau-181, GFAP and NfL differentiate between AD, mild cognitive impairment, and control participants [
7‐
9], and plasma pTau-181 and GFAP increase over time in preclinical AD [
10]. Although plasma Aβ
42/40 predicts Aβ-PET positivity [
11,
12], its utility for differentiating clinically-defined disease stages is comparatively limited [
13,
14]. These blood biomarkers are poised to play an important role in dementia research and clinical practice and are already being used to screen participants and as secondary endpoints in clinical trials for AD [
15]. While these blood biomarkers have been extensively characterized in recent years, the prognostic significance of these measures among cognitively normal older adults remains largely unknown.
Prior cross-sectional studies have shown that that higher NfL [
7,
16], GFAP [
5,
17], pTau-181 [
18], and lower Aβ
42/40 [
5] in blood were associated with smaller brain volume in regions vulnerable to AD (e.g., hippocampus, entorhinal cortex), as well as white matter volume [
19]. Additionally, higher blood levels of NfL and GFAP [
7,
17] and longitudinal increases blood levels of NfL [
16,
20,
21] and pTau-181 [
18] have been associated with decreased volume in AD-vulnerable brain regions.
Although previous studies have shown that lower Aβ
42/40 and higher ptau-181 [
18,
22], NfL [
7,
17], and GFAP [
17] in blood are associated with increased rate of cognitive decline [
23], the vast majority of studies connecting these blood biomarkers to cognitive function have been cross-sectional [
7,
17,
19,
24]. Accordingly, the extent to which these measures relate to future brain volume loss and cognitive decline in a cognitively normal community sample is not well understood. Additionally, whether demographic, genetic, and disease staging factors modify these associations is unknown. To address these questions, we used data from the Baltimore Longitudinal Study of Aging (BLSA) to examine the association of Aβ
42/40, pTau-181, GFAP, and NfL with baseline and longitudinal measures of regional brain volume and domain-specific cognition. We further sought to determine whether sex,
APOEε4 carrier status, and plasma amyloid status modified the association between plasma biomarkers and longitudinal brain volume and cognition.
Discussion
The present study examined whether plasma markers of AD pathology (Aβ42/40, pTau-181), astrogliosis (GFAP), and neuronal injury (NfL) predict longitudinal changes in brain volume and cognitive performance in a sample of cognitively unimpaired adults over a median follow-up of five to six years. We found that higher plasma pTau-181 was a robust predictor of brain volume loss, particularly in gray matter structures, while higher plasma GFAP showed a strong association with longitudinal ventricular enlargement and accelerated declines in verbal fluency. Although plasma Aβ42/40 and NfL were not associated with brain volumetric changes, lower plasma Aβ42/40 (indicative of greater Aβ burden) was associated with faster declines in verbal memory and visuospatial performance. Overall, our results suggest that higher pTau-181 and GFAP are indicative of future brain volume loss in cognitively normal older adults, while Aβ42/40 and GFAP levels are associated with faster subsequent cognitive declines in specific cognitive domains.
Given that declines in brain volume typically precede cognitive impairment [
50], early predictors of brain atrophy may aid in identifying individuals at risk for age-related cognitive decline and cognitive impairment, including dementia. The finding that plasma pTau-181 was associated with steeper longitudinal declines in brain volume aligns with a previous study that found baseline pTau-181 to be associated with greater brain atrophy in cognitively unimpaired individuals [
18]. In contrast to other studies [
7,
17] which have examined both participants with and without cognitive impairment, we found that higher abundance of plasma NfL was not associated with faster subsequent declines in brain volume. One reason for this discrepancy may be disease stage. Given our focus on cognitively unimpaired older adults, participants may not be advanced enough from a neuropathological perspective to detect significant associations with a marker of neuronal injury. Previous work has demonstrated that plasma NfL does not increase in those with AD compared to healthy controls until approximately 10 years before dementia diagnosis [
51]. In the present study, only 8.5% of participants went on to develop any type of dementia or mild cognitive impairment during the follow-up period. Increases in plasma NfL are thought to coincide with brain atrophy and occur after elevations in pTau-181, Aβ
42, and GFAP [
50,
52,
53]. The lack of NfL associations with brain volume change in the context of associations of pTau-181 and GFAP with brain volume change suggests that, among cognitively unimpaired older adults, NfL is comparatively less informative as a prognostic indicator of neurodegeneration.
The lack of an association between Aβ
42/40 levels and subsequent brain changes is surprising, particularly in the context of the robust pTau-181 associations. In a recent analysis using the same (BLSA) cohort, we found that Aβ
42/40 predicted amyloid PET status with good – but less than optimal – levels of accuracy (AUC = 0.72), with pTau-181 showing similar accuracy for prediction of amyloid status (AUC = 0.72) [
12]. The divergent associations with brain volume loss despite similar predictive accuracy for cortical amyloid suggest that plasma pTau-181, in addition to acting as a measure of amyloid status [
54], captures a broader range of neuropathological processes linked to neurodegeneration compared to plasma Aβ
42/40 at this point in the disease process [
18]. There are at least three additional explanations for the null association between plasma Aβ
42/40 and brain volume. First, accuracy of the Aβ
42/40 quantification may be limited by assay-specific factors, as the Simoa Aβ
42/40 assay has been shown to be inferior to mass spectrometry and other immune-assays for prediction of cortical Aβ [
55,
56]. Second, plasma Aβ
42/40 has a limited dynamic range with a group level difference between healthy control and AD patients of only 10–20%, making discrimination between AD and non-AD challenging, perhaps even more so among cognitively normal individuals [
50,
57]. Lastly, the accumulation of Aβ
42 peptides, which has been proposed to be an initiating factor in AD, may be too distal in the disease processes from the neurodegenerative processes underlying brain atrophy [
12].
In the current study, we found that lower plasma Aβ
42/40 and higher GFAP were associated with faster declines in verbal memory and fluency, respectively, over approximately 6 years. The association of Aβ
42/40 with accelerated decline in verbal memory supports the clinical utility and specificity of this biomarker in asymptomatic AD. The association of higher GFAP with faster declines in verbal fluency aligns with prior studies that suggest astrogliosis promotes synaptic dysfunction and subsequent cognitive impairment [
58]. While the association of Aβ
42/40 and GFAP with cognitive decline has been reported previously, the lack of association of NfL and pTau-181 with cognitive decline runs counter to previous findings [
23]. Despite finding that higher baseline pTau-181 was associated with accelerated declines in temporal lobe brain volume, pTau-181 was not associated with cognitive decline. Given that soluble pTau-181 levels are known to rise prior to significant increases in cortical tau [
52], and cortical tau is strongly associated with cognitive decline [
59], associations of plasma pTau-181 with cognitive decline may be observed only at a later disease stage or with a longer follow-up period. Notably, we did see nonsignificant trends in pTau-181 associations with cognitive change, suggesting that higher levels may relate to declines in several cognitive domains.
While only 8.5% of participants went on to develop mild cognitive impairment or dementia during the follow-up period of the current study, 40% of participants were estimated to be amyloid positive at the time plasma biomarkers were measured based on plasma Aβ42/40 levels. We found that amyloid-positive status – an indicator of the presence of AD pathology – modified the association of plasma pTau-181 with brain volume loss such that higher pTau-181 was more strongly associated with increased ventricular volume among those with higher Aβ burden. In contrast to these results, we also found that higher pTau-181 was more strongly associated with accelerated declines in total brain and hippocampal volume among participants with a lower Aβ burden. The reason for this seemingly contradictory finding is unknown. However, these results suggest that the association of pTau-181 with total brain and hippocampal volume loss may be magnified early in the course of AD pathogenesis (ahead of any major elevations in cortical amyloid), whereas pTau-181 and GFAP are associated with faster declines in verbal memory later in the course of AD pathogenesis, in the context of high Aβ burden.
This study has several strengths, including a large sample, the availability of longitudinal outcome data from a comprehensive neuropsychological examination and repeated neuroimaging assessments on a 3T MRI, and the focus on cognitively unimpaired (asymptomatic) older adults. Despite these strengths, the current study has several limitations. First, the BLSA is a relatively healthy sample and mean level of education in the present sample was 17 years. While the sample is relatively large and diverse, it is not sufficiently powered for stratified analyses to determine whether findings are consistent across included racial and ethnic groups. Therefore, the findings derived from the present study may not be generalizable to the broader population. Second, only a small percentage of participants progressed to cognitive impairment, which may have limited our ability to detect significant associations among some plasma biomarkers (e.g., NfL) and brain volume. Finally, the present study did not include measurement of the pTau-217 isoform, which compared to pTau-181, has demonstrated greater accuracy for prediction of cortical AD pathology [
60,
61]. Further, the addition of pTau-217 may have enabled us to detect differences in brain volume and cognition as a function of sex, as prior work has demonstrated that higher pTau-217 is related to declines in verbal memory and brain atrophy for cognitively unimpaired women but not men [
62].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.