Converging evidence from multiple meta-analyses (Jorm, 2001; Ownby et al., 2006; Diniz et al., 2013) suggests that depression approximately doubles an individual’s risk of developing dementia later in life. Brain amyloid-β (Aβ) deposition serves as a gold-standard hallmark of pathogenesis in Alzheimer’s disease (AD). Early autopsy studies (Rapp et al., 2006) have shown more pronounced Aβ plaque in AD patients who have lifetime major depressive disorder (MDD) as compared with those without MDD. Depression in old age has been increasingly investigated in terms of the relationship with in vivo cerebral Aβ via validated amyloid imaging in the past decade. However, the results regarding cerebral Aβ amounts in depression have been inconsistent and controversial. Butters et al. (2008) and Wu et al. (2014) found that MDD patients have increased cortical Aβ in comparison with non-depressed healthy controls; however, Madsen et al. (2012) found no difference between midlife MDD and cognitively normal individuals. Unexpectedly, a recent study by Mackin et al. (2021) showed an even more reduced cortical Aβ level in depressed elderly patients as compared with non-depressed cognitively normal subjects. Therefore, to date, related studies have yielded variable and conflicting results.

Depression in old age, not surprisingly, might represent an etiological entity with both clinical and pathophysiological heterogeneity. Our previous study (Wu et al., 2018) provided evidence of the diversity of involved neurodegenerative processes in elderly depressed individuals. We found that some depressed elderly patients entered the AD prodrome; others might be subject to a neurodegenerative pathway completely distinct from AD. In the past, we have always paid more attention to amyloid positivity for an accurate diagnosis of AD; however, the impact of subthreshold Aβ is gradually being explored (Bischof and Jacobs, 2019). Recent studies reported subthreshold Aβ and Aβ accumulation rate could predict early tau deposition in those who were nominally amyloid negative (Leal et al., 2018). Additionally, non-amyloid-mediated neurodegeneration could be associated with subthreshold Aβ changes (Jack et al., 2013). Subthreshold Aβ might provide clinically meaningful and useful information that may reflect various individual neurodegenerative processes. To date, a subthreshold Aβ condition among amyloid-negative MDD patients remains largely unclear. In order to make group comparisons on the same basis of amyloid negativity status, both samples of amyloid-negative MDD and control subjects were included in this study.

Among the amyloid-negative individuals, a suspected non-Alzheimer disease pathophysiology (SNAP) can be indicated by a negative test for β-amyloid (Aβ−) but a positive test for neurodegeneration (ND+). Hippocampal atrophy and glucose hypometabolism within AD-vulnerable regions have been widely adopted as ND biomarkers (Jack et al., 2012; Caroli et al., 2015; Mormino et al., 2016). In this study, amyloid-negative MDD patients were further classified into SNAP (Aβ−/ND+) MDD and Aβ−/ND− MDD groups. This study aimed to investigate the subthreshold Aβ characteristics in amyloid-negative MDD subjects, particularly in SNAP MDD patients, utilizing ¹⁸F-florbetapir PET imaging.

Unexpectedly, the SANP MDD group not only met the predefined criteria of an amyloid-negative status, but even exhibited a significantly reduced cerebral Aβ burden relative to the control and Aβ−/ND− MDD subjects in several brain regions. The most prominent decrease emerged in the bilateral mesial temporal cortex. The lack of differences of Aβ burden in the parietal and precuneus cortex between the SNAP MDD and control groups supported that the SNAP MDD patients had no preexisting early AD pathophysiology into the amyloid pathway. HVa atrophy, but not Aβ burden, had significant correlations with cognitive deficits in the total MDD and Aβ−/ND− MDD samples. However, in the SNAP MDD patients, neither HVa nor Aβ deposition were correlated with cognitive functions.

In the context of depression increasing the risk of incident dementia, early evidence connecting Aβ to depression came from postmortem studies of AD dementia patients, revealing an association between Aβ plaque and a lifetime history of MDD (Rapp et al., 2006). The advances of validated amyloid PET imaging enabled study of Aβ in vivo. In our previous studies (Wu et al., 2014, 2016), elderly patients with lifetime MDD, especially those with amnestic mild cognitive impairment, carried a greater Aβ burden in the parietal and precuneus cortex as compared with the controls. However, among subjects with midlife MDD, Madsen et al. (2012) demonstrated no global Aβ difference between cognitively normal MDD patients and control subjects. A population-based longitudinal study performed in Rotterdam (Mirza et al., 2016) provided some insight into the inconsistent results regarding Aβ in depression in old age, as that study uncovered that a substantially higher risk of dementia appeared in elderly depressed groups with an increasing depression trajectory, suggesting that depression might be a prodrome of dementia. In response to the postulation regarding depression as an AD prodrome, longitudinal studies have identified incident depressive symptoms corresponding to the baseline Aβ burden in cognitively normal older adults (Harrington et al., 2017; Donovan et al., 2018; Gatchel et al., 2019). Taken together, current evidence indicates that depression, especially late-onset depression, appears more likely to be a marker of incipient dementia than a true risk factor.

Challenging existing assumptions, Mackin et al. (2021) presented a shocking finding contrary to expectations, in that depressed elderly patients showed less Aβ accumulation than non-depressed subjects. Compared with the non-depressed group, which included individuals with a matched proportion of MCI, the depressed group exhibited decreased global Aβ accumulation and a lower proportion of Aβ positivity. They performed a second similar analysis restricting the sample to subsets of cognitively normal participants both in the depressed and non-depressed groups, which yielded results indicating a significant difference with regards to Aβ positivity, but not for Aβ burden.

In the present study, we observed that the amyloid-negative MDD patients had a significantly lesser Aβ burden than the control subjects. Moreover, the SNAP MDD population was the group that contributed most strongly to the result of reduced Aβ deposition. Corresponding to the reports by Mackin et al. (2021), our findings might provide a possibility as well as evidence to explain that other non-amyloid-mediated pathways may be associated with potential cortical Aβ reduction in depressed older adults. Nevertheless, we did not agree with the interpretation speculated by Mackin et al. that reduced cerebral blood flow or hypometabolism in depression may limit regional Aβ uptake (Mackin et al., 2021). Regional cerebral hypometabolism in SNAP by definition would mimic the metabolism pattern in AD; however, AD patients present with a typical abundant Aβ burden, and therefore decreased Aβ accumulation in the SNAP MDD group might be related to their own pathology-specific factors. To date, no study has investigated the characteristic Aβ patterns showing reduced Aβ depositions in SNAP patients with or without depression (Jack et al., 2012; Burnham et al., 2016).

Even though our findings demonstrated that other non-amyloid-mediated pathways were likely associated with reduced Aβ in the SNAP MDD patients, we did not refute the possibility that an increased Aβ burden is associated with the AD pathway in depression in old age. We very much agree with the comments of Christopher et al. (Van Dyck et al., 2021) that these diverse results should remind us of the tremendous heterogeneity regarding neurodegenerative pathophysiology in depression in old age. We should not expect most of depression in old age to have a uniform relationship with dementia pathogenesis. Some depressed individuals who enter the Aβ cascade in preclinical or prodromal AD stages may carry a greater Aβ burden; other depressed individuals have diverse Aβ depositions when they are on other non-AD pathways. Global cortical Aβ deposition was not found to be associated with any neuropsychological tests, whether in SNAP, Aβ−/ND− MDD, or all amyloid-negative MDD patients. However, HVa was positively associated with several neuropsychological tests in the amyloid-negative MDD sample, and when restricting the MDD sample to the Aβ−/ND− MDD patients. These correlations disappeared when further restricting the sample to the SNAP MDD subjects, the lack of a correlation in the SNAP MDD sample perhaps being associated with unrevealed factors other than Aβ and HVa.

Several limitations of this study must be acknowledged. First, this study was limited to small sample sizes; however, our results remained robust, even when using atrophy-corrected PVC data and a stringent threshold of FDR for multiple comparisons. Second, there are amyloid-positive individuals in the cognitively healthy general population, but we included control subjects with the Aβ−/ND− profile for group comparisons on the same amyloid-negative basis. However, the SNAP MDD patients exhibited much lower Aβ depositions, while amyloid-positive control subjects were also included in the study. Third, although atrophy and hypometabolism are widely recognized as ND biomarkers in relevant SNAP studies (Caroli et al., 2015; Jack et al., 2016; Mormino et al., 2016), an operational standardized approach to measurement remains lacking a consensus (Jack et al., 2012). Individuals may be misclassified due to arbitrary cut-off values. Moreover, the pathophysiology of SNAP should remain heterogeneous. Other non-amyloid neuropathological changes, such as Aβ-independent tauopathy, hippocampal sclerosis with TDP-43, α-synucleinopathy, or argyrophilic grain disease, might contribute to and coexist with each other (Wirth et al., 2013; Mormino et al., 2016; Villeneuve, 2016). Data regarding specific pathologies of the SNAP subjects were unavailable in this study. Finally, interpretation and generalization of the findings to non-depressive subjects must acknowledge the potential limitations of this study. Notably, the finding that the SNAP and Aβ-/ND-MDD groups differed in amyloid burden despite similar psychopathological features between these two depressed groups, supported that SNAP, but not depression, is associated with reduced amyloid burden. Future studies in non-depressed subjects with SNAP are critical to determine the role of isolated SNAP on brain Aβ changes.

In this study, the amyloid-negative MDD patients had a significantly lesser Aβ burden than the control subjects. The SNAP MDD patients were the group that contributed most strongly to the result of reduced Aβ deposition. SNAP MDD subjects represent a distinct study population with ND biomarkers mimicking AD, but pathological biomarkers not doing so. Our findings might provide a possible evidence that other non-amyloid-mediated pathways may be involved in underlying cortical Aβ reduction in depressed older adults. We envisage that the currently labeled entity “amyloid-negative individuals” might be further refined to discern individuals with substantially low amounts of Aβ. Meaningfully low amounts of subclinical Aβ might provide critical information on the pathogenesis of non-AD individuals.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The studies involving human participants were reviewed and approved by Chang Gung Medical Foundation Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.

K-YW, K-JL, and I-TH designed the study. K-YW, K-JL, I-TH, C-HC, Y-MW, and C-YL acquired the data. K-YW, K-JL, Y-MW, T-CY, C-SC, and I-TH analyzed the data. K-YW, K-JL, T-CY, and I-TH wrote the manuscript. All authors revised and approved the article for publication.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.