DOI: 10.1002/alz.079240 ISSN: 1552-5260

Braak‐stage II atrophy links lower locus coeruleus – medial temporal lobe connectivity to cognitive decline in preclinical AD

Christoph Schneider, Prokopis C. Prokopiou, Kathryn V. Papp, Stephanie Hsieh, Aaron P. Schultz, Dorene M. Rentz, Reisa A. Sperling, Keith A. Johnson, Heidi I.L. Jacobs
  • Psychiatry and Mental health
  • Cellular and Molecular Neuroscience
  • Geriatrics and Gerontology
  • Neurology (clinical)
  • Developmental Neuroscience
  • Health Policy
  • Epidemiology



In preclinical human Alzheimer’s disease (AD), novelty‐related functional connectivity of the locus coeruleus (LC) with the medial temporal lobe (MTL) (ConnLC‐MTL) predicted beta‐amyloid (Aβ)‐related longitudinal cognitive decline. In animals, LC lesions in AD models led to dwindling LC projections to the hippocampus, hippocampal atrophy, and cognitive impairment. Following these observations, we investigated the contribution of cross‐sectional and longitudinal atrophy in early Braak stages as a potential mechanism underlying this relationship.


We included 128 participants of the Harvard Aging Brain Study (HABS) with baseline Pittsburgh Compound‐B Positron Emission Tomography – Distribution Volume Ratio (PiB PET, DVR; 36 Aβ+), as well as baseline functional and longitudinal structural MRI (ntot = 372; ∼ every 3 years; follow‐up 5±3.4 years) and yearly cognitive assessments (PACC5 = Preclinical Alzheimer’s Cognitive Composite; ntot = 874; follow‐up 6.5±3.3 years) (Table 1). ConnLC‐MTL values were derived voxel‐wise and averaged across significant clusters. FreeSurfer‐estimated cortical thickness and hippocampus/amygdala volume (adjusted for intracranial volume) were z‐scored and averaged in two variables (THV2/THV3), encompassing regions associated with Braak stage II and III respectively (Table 1). Baseline associations were evaluated with linear models, longitudinal relationships between ConnLC‐MTL and THV2/THV3 with mixed‐effects models (random intercept and slope). Mediation of ConnLC‐MTL and PACC5 slopes (year 5‐13) by THV2/THV3 slopes (year 0‐5) was calculated (n = 79; bootstrapped n = 1000; Table 1). All models contained age, sex and education as covariates.


At baseline, ConnLC‐MTL positively associated with THV2 (p = 0.005) and with PACC5 (p = 0.045), but not with THV3 (p = 0.26; Figure 1A‐C). Longitudinally, ConnLC‐MTL did not predict changes in THV2 nor THV3. However, we observed faster cortical atrophy with lower ConnLC‐MTL at higher levels of neocortical PiB (THV2: p = 0.02; THV3: p<0.001; Figure 1D‐E). Finally, THV2, but not THV3, mediated the relationship between ConnLC‐MTL and PACC5 decline (p = 0.024) and this effect was stronger in PiB+ individuals compared to PiB‐ (p = 0.026; Figure 2).


Worse LC‐MTL connectivity was associated with downstream atrophy in the target Braak‐stage II areas, which impacted cognitive performance in preclinical AD. Given that in similar previous animal findings activation of the norepinephrine LC system reversed memory deficits, maintaining optimal ConnLC‐MTL may be a promising intervention target to delay disease progression.

More from our Archive