Plasma ADRD biomarkers predict longitudinal declines in intra‐network functional brain connectivity, and baseline functional connectivity predicts longitudinal cognition

Heather E. Dark, Andrea T. Shafer, Jenifer Cordon, Yang An, Alexandria Lewis, Abhay Moghekar, Bennett A. Landman, Susan M. Resnick, Keenan A. WalkerAlzheimer 2024 · DOI

TL;DR

This study reveals that specific blood-based biomarkers for Alzheimer's disease pathology, astrogliosis, and neuronal injury can predict future declines in functional brain connectivity within key networks in cognitively normal individuals. Furthermore, the baseline strength of this functional connectivity itself predicts subsequent cognitive changes, highlighting a potential pathway from molecular pathology to network dysfunction and cognitive decline.

What Problem Does This Paper Address?

Alzheimer's Disease (AD) is characterized by the accumulation of amyloid-beta (Aβ) plaques and tau neurofibrillary tangles, leading to neurodegeneration and cognitive decline. While these pathological hallmarks are well-established, understanding how they impact the brain's large-scale functional networks over time, especially in the preclinical stages, remains a critical area of research. Specifically, researchers want to know if easily measurable blood biomarkers can serve as early indicators of these network changes and if these network changes, in turn, predict future cognitive function. This paper addresses the gap in understanding the longitudinal relationship between peripheral AD biomarkers, intra-network functional brain connectivity, and subsequent cognitive trajectories in individuals who are not yet experiencing cognitive impairment.

Methods

The researchers utilized data from the Baltimore Longitudinal Study of Aging (BLSA). They measured baseline plasma biomarkers for AD pathology (Aβ42/40 ratio and phosphorylated tau [pTau-181]), astrogliosis (glial fibrillary acidic protein [GFAP]), and neuronal injury (neurofilament light chain [NfL]) using Quanterix SIMOA assays. Functional brain connectivity was assessed using 3T resting-state functional magnetic resonance imaging (rsfMRI). They defined seven intrinsic functional networks based on a predefined cortical parcellation.

To analyze the data, they employed linear mixed-effects (LME) models. These models were crucial for examining longitudinal changes and accounted for covariates such as age, sex, race, education, and gray matter volume. The LME models were used to:

  1. Determine if baseline plasma biomarkers predicted longitudinal changes in resting-state functional connectivity (rsFC).
  2. Investigate if the association between biomarkers and rsFC changes differed based on amyloid status (determined by Aβ42/40 levels).
  3. Assess whether baseline rsFC predicted longitudinal changes in cognitive function.

Datasets

The study utilized data from the Baltimore Longitudinal Study of Aging (BLSA). This dataset includes longitudinal measurements of plasma biomarkers, resting-state fMRI, and cognitive assessments from participants. The paper does not introduce a new benchmark dataset.

Key Insights and Ideas

  • Blood Biomarkers as Early Predictors: The study demonstrates that peripheral blood biomarkers, which are less invasive than PET scans or CSF analysis, can predict future changes in brain network function. This is a significant step towards earlier and more accessible AD detection and monitoring.
  • Amyloid Status Matters: The relationship between plasma biomarkers and rsFC changes is not uniform across all individuals. It is particularly pronounced in those with evidence of amyloid pathology, suggesting that amyloid accumulation might prime the brain for these network disruptions.
  • Network Dysfunction Precedes Cognitive Decline: The findings suggest a cascade: AD-related molecular changes (reflected in blood biomarkers) lead to alterations in functional brain connectivity, which in turn, precede measurable declines in cognitive abilities. This provides a potential mechanistic link between pathology and cognitive symptoms.
  • Multiple Biomarkers are Informative: The study highlights that not just AD pathology markers (Aβ, pTau) but also markers of neuroinflammation (GFAP) and neuronal injury (NfL) are associated with changes in brain connectivity, painting a more comprehensive picture of the pathological processes involved.

Key Results

  • Biomarkers and rsFC Decline: Higher levels of Aβ42/40, GFAP, and NfL were associated with faster declines in rsFC within several brain networks (P-range = 0.01-0.04).
  • Amyloid-Specific Associations: These biomarker-rsFC associations were significantly different based on amyloid status (P-range = 0.01-0.045).
  • Amyloid-Positive Individuals: In amyloid-positive participants, lower Aβ42/40 and higher GFAP and NfL levels were linked to faster rsFC declines in visual, attention (dorsal and ventral), limbic, and frontoparietal networks (P range ≤ 0.002-0.04).
  • No Associations in Amyloid-Negative: No statistically significant associations were found between plasma biomarkers and rsFC change in amyloid-negative participants.
  • rsFC Predicts Cognition: For a larger cohort (760 participants), baseline rsFC in several networks predicted longitudinal changes in cognition, including working memory, verbal fluency, and visuospatial abilities (P range = 0.02-0.049).

Limitations and Future Work

  • Cognitively Normal Cohort: The study focused on cognitively unimpaired individuals. While this is crucial for understanding preclinical changes, the findings may not directly translate to individuals with established cognitive impairment.
  • Amyloid Status Definition: Amyloid status was defined using plasma Aβ42/40 and ROC analysis for optimal prediction of amyloid PET positivity. While this is a practical approach, direct PET or CSF confirmation of amyloid status for all participants would provide a more definitive classification.
  • Future Work: The authors suggest that hypo- and hyper-intra-network connectivity may drive changes in cognitive performance. Future research could explore interventions targeting these network dysfunctions to potentially slow or prevent cognitive decline. Further investigation into the specific mechanisms by which these biomarkers influence network connectivity is also warranted.

Summary updated

Abstract

Abstract Background AD is defined by cortical amyloid‐β (Aβ), tau neurofibrillary tangles, and neurodegeneration, pathological processes which may contribute to cognitive decline by altering large scale functional brain networks. To test this hypothesis, we examined whether plasma biomarkers of AD pathology (Aβ 42/40 , phosphorylated tau [pTau‐181]), astrogliosis (glial fibrillary acidic protein [GFAP]), and neuronal injury (neurofilament light chain [NfL]) related to longitudinal changes in resting‐state functional connectivity (rsFC) in cognitively unimpaired participants from the Baltimore Longitudinal Study of Aging. Method Baseline plasma biomarkers were measured with Quanterix SIMOA assays. Functional connectivity (3T resting‐state fMRI) was derived using a predefined cortical parcellation mask from which intra‐network connectivity from seven functional networks was extracted for each participant. Amyloid status (positive/negative) was defined using plasma Aβ 42/40 (Figure 1). Linear mixed effects models adjusted for age, sex, race, education, gray matter volume, and time‐covariate interactions were used to determine whether 1) baseline plasma biomarkers predicted longitudinal changes in rsFC, 2) the magnitude of the biomarker‐related rsFC changes differed by amyloid status, and 3) rsFC predicted longitudinal changes in cognition. Result Longitudinal connectivity analyses (mean age±SD=65.49±16.17) included 490 participants (1190 visits; mean follow‐up time=4.31±1.68 years). Higher Aβ 42/40 , GFAP, and NfL were associated with faster declines in rsFC within several networks (P‐range=0.01‐0.04; Figure 2). Overall, plasma biomarker‐rsFC associations differed by amyloid status (P‐range=0.01‐0.045). Among amyloid‐positive participants, lower levels of Aβ 42/40 , and higher levels of GFAP, and NfL (Figure 2) were associated with faster declines in rsFC in the visual, dorsal and ventral attention, limbic, and frontoparietal networks (P range=<0.002‐0.04). There were no statistically significant associations between plasma biomarkers and rsFC change among amyloid‐negative participants. Among 760 participants with at least one rsFC scan (mean age±SD=67.21±14.93; 1550 visits, follow‐up time=3.94±1.60 years), we found that baseline rsFC in several networks predicted changes in cognition, e.g., working memory, verbal fluency, and visuospatial abilities (P range=0.02‐0.049; Figure 3). Conclusion Among cognitively normal individuals, plasma biomarkers of Aβ 42/40 , astrogliosis, and neuronal injury are associated with future intra‐network functional brain changes, particularly in the context of elevated amyloid. Hypo‐ and hyper‐ intra‐network connectivity may drive changes in cognitive performance.
BibTeX
@inproceedings{dark2024plasma,
  author = {Heather E. Dark and Andrea T. Shafer and Jenifer Cordon and Yang An and Alexandria Lewis and Abhay Moghekar and Bennett A. Landman and Susan M. Resnick and Keenan A. Walker},
  title = {{Plasma ADRD biomarkers predict longitudinal declines in intra‐network functional brain connectivity, and baseline functional connectivity predicts longitudinal cognition}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.092515},
}
Figure