Here, we applied an unsupervised data-driven method in a broad range of CN and MCI individuals to evaluate the association between spatiotemporal variability in amyloid deposition and AD profiles. We identified two reproducible subtypes characterized by different regional progression patterns in two independent datasets. The spatiotemporal subtypes of amyloid deposition showed distinct cognitive and biomarker characteristics. Our findings suggest that spatiotemporal progression patterns in amyloid accumulation could provide insight into the disease monitoring and enrollment of therapeutic trials in CN and MCI.
In this study, the amyloid deposition was aptly described as two distinct progression patterns in a purely data-driven model, offering regional abnormality trajectories from early to late stages. In the cortex-priority subtype, the cingulate was the first region to show abnormality. In line with this, previous studies demonstrated that the cingulate has high-intensity values early on and may be the seed of amyloid deposition propagation [12, 24–26]. The following abnormal regions, the parietal lobe and frontal region, seem to be affected by the neighboring cingulate, and the association between them has been indicated in a previous study [27]. In general, the progression pattern of the cortex-priority subtype was similar to previous neuropathologic findings [8, 28, 29], except for the early appearance of the cingulate. In addition, some PET estimates showed that the medial frontal, medial parietal, and lateral temporo-parietal areas were the initial sites of amyloid deposition [30, 31]. Different ROI choices and analytic approaches may contribute to these inconsistencies. Data-driven methods may lead to unappreciated subtypes. The progression pattern in the subcortex-priority subtype was an intriguing result. In this subtype, the abnormalities in the subcortical regions occurred prior to those in the cerebral cortex. In particular, the thalamus and basal ganglia became abnormal earlier than the other regions. In line with this, previous studies have shown that the thalamus and basal ganglia are vulnerable to amyloid deposition and appear to have high SUVRs across preclinical AD phases [32, 33]. However, such a sequence contradicts the progression of amyloid from the neocortex to the subcortical regions in previous studies [14, 15, 34]. One possible explanation for the discrepancy is that the cutoffs we defined may make subcortical SUVRs less reliable and more sensitive to variation. Therefore, such subtype may fall short of explaining the ground truth neuropathology. Meanwhile, a distinct advantage of the present consideration is that it provided additional information of amyloid deposition and captured the early variation in the vast majority of individuals with few clinical symptoms.
The spatiotemporal subtypes enable linkage of amyloid accumulation trajectories with genetic and biomarker characteristics. APOE ε4 has consistently been found to be the strongest risk factor for AD, increasing the risk via oligomerization, aggregation, degradation, and clearance of amyloid [35]. Previous studies showed that individuals with positive amyloid-PET were more frequently APOE ε4 carriers than those with negative ones [36]. Similar findings have been found in the two subtypes we identified. CSF indicators usually provide highly concordant information with PET measures [4, 37]. In line with this, we found significant associations between the stages and the CSF Aβ concentrations in distinct subtypes, but the CSF level was lower in the subcortex-priority subtype. Previous studies showed that amyloid-PET is associated with tau pathology in the preclinical and prodromal stages of AD [38, 39]. Because of the extensive absence of tau-PET data from the two datasets, we used CSF t-tau and p-tau data as measures of tau pathology and found that higher stages in either one of the subtypes were correlated with increased levels of t-tau and p-tau. Our findings indicate that all the sequences of regional vulnerabilities in amyloid deposition are associated with tau accumulation, potentially providing insight into the monitoring of AD neuropathology.
Few studies have examined the association between spatiotemporal variations in amyloid deposition and clinical presentation in CN and MCI. We found that individuals with the cortex-priority subtype had lower cognitive performance and executive function compared with those in the subcortex-priority subtype. Notably, individuals in the cortex-priority subtype had a higher probability of conversion to dementia within 6 years. Moreover, the stage may be a useful marker for cognitive decline or conversion risk for individuals in the cortex-priority subtype. However, there was no apparent conversion until the presence of cortical abnormality for individuals in the subcortex-priority subtype, providing evidence that a significant proportion of elderly subjects remain cognitively normal with amyloid deposition [40, 41]. Overall, participants with different progression patterns of amyloid accumulation have distinct disease trajectories in individuals in CN and MCI. Incorporating these findings into clinical trials with differentiated populations holds great promise for improving the accuracy of individualized diagnosis and providing opportunities for future therapeutic intervention to prevent or slow the rate of disease progression [42]. In particular, our findings suggested that amyloid abnormality in the cortical regions are a key predictor in the progression of cognitive decline and subsequent conversion to AD dementia.
Although we identified consistent spatiotemporal subtypes of amyloid burden across two independent datasets in CN and MCI that differed in disease profiles, our results need to be replicated in other cohorts. The subcortical regions are primarily being used to characterize the global amyloid accumulation. Current findings lack clear cutoffs of the subcortical area, especially in individuals with few clinical symptoms. Although the two subtypes identified by cutoffs defined in the context of our study resulted in more differentiated population stratification, future work is needed to balance the sensitivity and specificity to account for the underlying neuropathology. Furthermore, AD is a complex neurodegenerative disorder that is associated with amyloid deposition, tau accumulation, hypometabolism, brain atrophy, and a variety of biological processes. Future work should investigate other AD-related biomarkers to further improve risk stratification.
In summary, we found two spatiotemporal subtypes with regional progression patterns of amyloid accumulation in a broad range of individuals with few clinical symptoms. The amyloid subtypes showed distinct regional progression patterns and AD profiles. Furthermore, the regional progression patterns were associated with clinical and biomarker characteristics. Our findings highlight the importance of uncovering the spatiotemporal variations of amyloid deposition in CN and MCI for clinical trials and precision medicine.