Sleep duration and amyloid status moderate the association between mood symptoms and amygdalar tau in preclinical Alzheimer's disease

Abstract

INTRODUCTION

Anxiety and depressive symptoms are common in Alzheimer's disease (AD), yet their relationships with amyloid, tau, and sleep remain unclear. We examined whether amyloid status and sleep duration moderate the relationships between anxiety and depressive symptoms and amygdalar tau burden in cognitively unimpaired older adults at risk for AD.

METHODS

Participants (n = 393) from the Anti‐Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4) and the Longitudinal Evaluation of Amyloid Risk and Neurodegeneration (LEARN) studies underwent tau and amyloid positron emission tomography imaging. Anxiety and depressive symptoms were evaluated using the State‐Trait Anxiety Inventory and Geriatric Depression Scale. Sleep duration was self‐reported.

RESULTS

Positive amyloid status moderated the relationship between depressive symptoms and amygdalar tau. Sleep duration moderated the relationship between anxiety and amygdalar tau, such that greater anxiety symptoms were associated with higher tau levels at shorter sleep durations.

DISCUSSION

Findings suggest biological and behavioral factors jointly influence neuropsychiatric symptom–tau relationships in preclinical AD, supporting an interactive model of early disease vulnerability.

Highlights

• Amyloid positivity strengthened the association between depressive symptoms and amygdalar tau burden, such that greater depressive symptoms were associated with higher tau in amyloid‐positive individuals.

• Shorter sleep duration amplified the association between anxiety and amygdalar tau.

• Findings suggest distinct biological (amyloid–tau) and behavioral (sleep–anxiety) pathways influencing early Alzheimer's disease (AD) vulnerability.

• Neuropsychiatric symptoms may serve as early indicators of underlying AD pathology or as contributing risk factors.

1. BACKGROUND

Alzheimer's disease (AD) develops along a biological continuum beginning years before cognitive symptoms appear. Early pathology includes amyloid beta (Aβ) and neurofibrillary tau accumulation. ¹ , ² Aβ aggregation occurs diffusely, disrupting synaptic function and promoting neuroinflammation, while hyperphosphorylated tau impairs neuronal stability and spreads through cortical networks, closely tracking cognitive symptoms. ³ , ⁴ In preclinical AD, pathological changes such as amyloid deposition, and, in some individuals, early tau abnormalities, emerge in the absence of clinically significant symptoms. Advances in biomarker detection have enabled identification of Aβ and tau abnormalities in asymptomatic individuals, though these markers do not fully predict when or how symptoms emerge. ⁵

Growing evidence highlights neuropsychiatric symptoms (NPS) as early indicators of AD processes. Anxiety and depression are common across the AD continuum, from the preclinical stage through dementia, with prevalence rising to nearly 40% in symptomatic stages. ⁶ , ⁷ , ⁸ Anxiety often presents as excessive worry, restlessness, or social withdrawal, whereas depression manifests as sadness, hopelessness, and loss of motivation. ⁷ , ⁹ These symptoms frequently co‐occur and are associated with poorer functioning, faster decline, and greater caregiver burden. ⁶ , ⁷ , ⁹ , ¹⁰ , ¹¹ Longitudinal and cross‐sectional studies show that older adults exhibiting these symptoms are at increased risk of developing mild cognitive impairment (MCI) and dementia. ^{1 2 − 2 1} Two hypotheses have been proposed: that anxiety and depression increase vulnerability to neurodegeneration, or that they reflect early AD‐related brain changes. ⁹ , ²² , ²³ , ²⁴ , ²⁵ A bidirectional relationship is also possible, emphasizing the need to examine associations between NPS and AD biomarkers such as Aβ and tau, particularly during the preclinical stage of the disease. ⁹ , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸

Sleep disturbances are common in AD and may both reflect and exacerbate underlying pathology. Poor sleep quality and shorter duration have been linked to greater Aβ and tau accumulation, cognitive decline, and heightened anxiety and depression symptoms across preclinical and symptomatic stages, suggesting that sleep may impact the relationship between NPS and AD biomarkers. ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵²

The amygdala may represent a key region connecting these biological and behavioral features. Neurofibrillary changes in parts of this structure emerge by Braak stage II, with tau positron emission tomography (PET) uptake detectable up to a decade before clinical onset. ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ However, tau accumulation in medial temporal regions, including the amygdala, can be observed independent of amyloid, raising the question of whether associations between amygdalar tau and NPS reflect AD‐related pathology or more general age‐related tau processes. The amygdala regulates emotion and sleep, and its dysfunction has been implicated in anxiety and depression. Our prior work and that of others have shown that greater amygdalar tau burden is associated with depressive symptoms, ¹ , ³³ whereas other studies have demonstrated that sleep deprivation increases amygdala reactivity to negative stimuli. ⁵⁶ , ⁵⁷

Despite increasing evidence linking NPS, sleep, and AD pathology, relationships among these factors and early tau accumulation in the amygdala remain unclear. Specifically, it is unknown whether associations between anxiety and depressive symptoms and amygdalar tau are contingent on the presence of underlying amyloid pathology, suggesting AD‐specific tau effects, or whether they are evident independent of amyloid, potentially reflecting age‐related tau processes. In addition, sleep patterns may moderate these relationships. Studying cognitively unimpaired adults at risk for AD may provide insight into associations among anxiety, depression, and AD pathology and help contextualize these symptoms for efforts aimed at early identification. Such work may also clarify the conditions under which these symptoms are observed across the preclinical disease continuum and help disentangle whether NPS are associated with early AD‐related pathological changes, reflect broader age‐related processes, or both.

Thus, our study aims to: (1) examine whether amyloid status moderates the relationship between anxiety and depressive symptoms and amygdalar tau burden in preclinical AD; and (2) determine whether sleep duration moderates this relationship and whether effects differ by amyloid status.

2. METHODS

2.1. Participants

This study included participants from the Anti‐Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4) and the Longitudinal Evaluation of Amyloid Risk and Neurodegeneration (LEARN) studies. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ The A4 study was a 3‐year, randomized, placebo‐controlled clinical trial testing whether solanezumab could slow cognitive decline in cognitively unimpaired older adults with elevated brain amyloid, while LEARN enrolled individuals without elevated amyloid as a comparison cohort. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ Study design, screening procedures, and measures have been described previously. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰

The current analysis focused on a subset of participants with available tau PET data for the amygdala (n = 393). Participants were cognitively normal adults aged 65 to 85 years, living independently, and meeting standard criteria for normal cognition. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ Individuals with cognitive impairment, dementia, unstable medical conditions, or significant psychiatric disorders were excluded. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ Amyloid status was determined using [^{1 8}F]florbetapir PET, with standardized uptake value ratio (SUVR) ≥ 1.15 classified as amyloid positive. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ All study sites obtained institutional review board approval, and participants provided written informed consent prior to enrollment. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰

RESEARCH IN CONTEXT

• Systematic review: We reviewed prior literature using PubMed and Google Scholar to investigate neuropsychiatric symptoms (NPS), amyloid and tau accumulation, sleep, the amygdala, and Alzheimer's disease (AD). Although anxiety and depressive symptoms are recognized as early features across the AD continuum, their associations with tau pathology, particularly within the amygdala, and the moderating roles of amyloid and sleep remain incompletely understood, with limited work in preclinical AD.

• Interpretation: Our findings demonstrate that greater depressive symptoms were associated with higher tau burden in amyloid‐positive individuals, whereas greater anxiety symptoms were associated with higher tau levels at shorter sleep durations, particularly among amyloid‐positive participants. These findings highlight distinct behavioral and biological pathways through which NPS relate to early tau accumulation and support an interactive model in which NPS, amyloid, and sleep intersect to influence vulnerability to AD‐related change in the preclinical stage.

• Future directions: Future work should examine these relationships in individuals with clinically significant anxiety or depression, incorporate validated subjective and objective sleep measures, and use longitudinal and mechanistic approaches to clarify how stress, inflammatory, glymphatic, and network‐level processes link NPS and tau in preclinical AD.

2.2. Clinical measures

Anxiety and depressive symptoms were assessed through self‐report measures. Anxiety symptoms were evaluated using a six‐item short form of the State‐Trait Anxiety Inventory (STAI). ⁶¹ Depressive symptoms were measured using the 15‐item Geriatric Depression Scale (GDS). ⁶² Sleep duration was evaluated using a lifestyle questionnaire in which participants reported their average hours of sleep per night. ⁵⁰ , ⁵⁸ , ⁵⁹

2.3. Amyloid and tau PET imaging

Amyloid burden was quantified using [¹⁸F]florbetapir PET, with SUVRs computed using the 50 to 70 minute postinjection window and the whole cerebellum as the reference region. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ Amyloid status was determined using both quantitative and qualitative methods. A SUVR of ≥ 1.15 was classified as amyloid positive (Aβ+), while SUVR values between 1.10 and 1.15 were considered Aβ+ only if confirmed by a visual read. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ The most recent PET classification was used for analysis.

A subset of participants underwent [¹⁸F]flortaucipir (FTP) PET to measure tau accumulation. Scans were acquired 90 to 110 minutes postinjection and reconstructed into six 5‐minute frames. ¹¹ , ⁵⁹ , ⁶⁰ PET data were processed using PETSurfer, with SUVRs calculated relative to the whole cerebellum. Partial volume correction was applied using the geometric transfer matrix method implemented in FreeSurfer. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰ Regional SUVRs were extracted from early tau‐accumulating regions in AD, including the entorhinal cortex, amygdala, and inferior temporal cortex.

2.4. Statistical analyses

All statistical analyses were conducted using R, with a significance threshold set at P < 0.05. Models controlled for age, sex, and education as covariates. To assess whether amyloid status and sleep duration moderate the relationship between anxiety and depressive symptoms and amygdalar tau burden, linear regression models with interaction terms were used.

For Aim 1, we examined whether amyloid status (positive vs. negative) moderates the relationship between anxiety and depressive symptoms and tau burden in the bilateral amygdala. A linear regression model was fitted with anxiety or depressive symptoms (STAI or GDS scores), amyloid status, and their interaction term as predictors of amygdalar tau burden:

$$\begin{array}{c}\hfill AmygdalarTauBurden={\beta }_0+{\beta }_1\left(STAI/GDS\right)\\ \hfill +{\beta }_2\left(AmyloidStatus\right)+{\beta }_3(STAI/GDS\end{array}$$

$$\times AmyloidStatus)+\epsilon$$

A significant interaction term (β₃) would indicate that the relationship between anxiety or depressive symptoms and tau burden differs based on amyloid status. If the interaction was significant, follow‐up stratified regression analyses were conducted separately within amyloid‐positive and amyloid‐negative groups to determine the direction and strength of the effects. In this model, amyloid‐negative individuals served as the reference group, meaning that the main effect of anxiety or depressive symptoms represents its association with tau burden in amyloid‐negative participants.

For Aim 2, we investigated whether sleep duration moderates the relationship between anxiety and depressive symptoms and tau burden, both in the overall sample and specifically within amyloid‐positive individuals. First, a linear regression model was conducted across the full sample, testing whether sleep duration modifies the association between anxiety or depressive symptoms and amygdalar tau burden:

$$\begin{array}{c}\hfill AmygdalarTauBurden={\beta }_0+{\beta }_1\left(STAI/GDS\right)\\ \hfill +{\beta }_2\left(SleepDuration\right)+{\beta }_3(STAI/GDS\end{array}$$

$$\times SleepDuration)+\epsilon$$

Here, the interaction term was tested to determine whether the effect of anxiety and depression on tau burden depends on sleep duration. Sleep duration was treated as a continuous variable, with higher values reflecting longer sleep duration.

To further explore whether the moderating effect of sleep is stronger in amyloid‐positive individuals, a separate interaction analysis was conducted exclusively within the amyloid‐positive subgroup:

$$\begin{array}{c}\hfill AmygdalarTauBurden={\beta }_0+{\beta }_1\left(STAI/GDS\right)\\ \hfill +{\beta }_2\left(SleepDuration\right)+{\beta }_3(STAI/GDS\end{array}$$

$$\times SleepDuration)+\epsilon$$

This analysis allowed us to determine whether the relationship between anxiety and depression symptoms and tau burden differs by sleep duration in individuals with greater amyloid accumulation, as prior research suggests that sleep disturbances are linked to both amyloid and tau pathology. ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³

If the interaction was significant, post hoc analyses were conducted using simple slopes analysis to examine the relationship between anxiety and depressive symptoms and tau burden at different sleep levels (e.g., low, medium, and high sleep durations defined as one standard deviation [SD] above or below the mean). Additionally, Johnson–Neyman analysis was performed to identify the specific range of sleep durations in which the interaction effect was significant. Parallel analyses were conducted for bilateral entorhinal and inferior temporal cortex tau SUVR using identical modeling approaches and covariates to assess regional specificity.

Sensitivity and exploratory analyses were conducted to evaluate potential sources of biological and methodological heterogeneity. Apolipoprotein E (APOE) ε4 carrier status and PET scanner class were included as additional covariates in sensitivity analyses of all primary models. Exploratory moderation analyses examined APOE ε4 carrier status as a moderator of the relationships between anxiety and depressive symptoms and amygdalar tau burden using interaction terms. Exploratory linear regression analyses were also conducted to examine associations among bilateral amygdalar tau burden, NPS, and global cognitive performance as measured by the Preclinical Alzheimer Cognitive Composite (PACC), adjusting for age, sex, and education.

3. RESULTS

Analyses included 393 participants from the A4 and LEARN studies with available tau PET data for the amygdala. Participant demographics and clinical characteristics by amyloid status are presented in Table 1. The amyloid‐positive and amyloid‐negative groups did not significantly differ in years of education, sex distribution, anxiety symptoms, depressive symptoms, or average sleep duration, indicating that the groups were generally well matched on these variables. However, the amyloid‐positive group was significantly older than the amyloid‐negative group (P < 0.001).

TABLE 1.

Demographics of A4 tau PET cohort.

	Amyloid negative	Amyloid positive	Total	P value
n	51	342	393
Age, years	69.6(4.4)	72.2(4.9)	71.9 (4.8)	<0.001
Education, years	16.7 (2.7)	16.1 (2.7)	16.2 (2.7)	0.200
Sex, female (%)	30 (58.8%)	200 (58.5%)	0 230 (58.5%)	1.000
STAI total	10.45 (3.17)	10.15 (3.02)	10.19 (3.03)	0.521
GDS total	1.24 (2.05)	1.03 (1.28)	1.06 (1.40)	0.487
Sleep duration, hours	6.94 (1.03)	7.04 (1.05)	7.03 (1.04)	0.533

Note: Mean (SD) or % values are reported.

Abbreviations: A4, Anti‐Amyloid Treatment in Asymptomatic Alzheimer's Disease; GDS, Geriatric Depression Scale; PET, positron emission tomography; SD, standard deviation; STAI, State Trait Anxiety Inventory.

3.1. Depressive symptoms, anxiety symptoms, amyloid status, and amygdalar tau

Amyloid status significantly moderated the relationship between depressive symptoms and amygdalar tau burden. This effect remained significant after adjusting for age, sex, and education (β = 0.06, standard error [SE] = 0.03, t[386] = 2.19, P = 0.029). There was no main effect of GDS on tau burden. Stratified analyses revealed that among amyloid‐positive participants, higher GDS scores were associated with greater amygdalar tau burden (β = 0.04, SE = 0.01, t[340] = 2.44, P = 0.015; Figure 1). No association was observed among amyloid‐negative individuals. Amyloid status did not moderate the relationship between anxiety symptoms (STAI total score) and amygdalar tau burden. There was no main effect of STAI on tau burden. Given the restricted range of depressive symptoms in this preclinical cohort, an additional sensitivity analysis was conducted and revealed a consistent positive GDS × amyloid interaction that was not driven by a small number of individuals with elevated depressive symptoms (see Tables S1–S3 in supporting information).

FIGURE 1.

Interaction between depressive symptoms (GDS) and amyloid status on amygdalar tau burden.

The plot depicts the relationship between GDS scores and bilateral amygdalar tau SUVR separately for amyloid‐positive (blue) and amyloid‐negative (red) participants. Among amyloid‐positive individuals, higher GDS scores were associated with greater tau burden, whereas no relationship was observed among amyloid‐negative individuals. Shaded regions represent 95% confidence intervals. GDS, Geriatric Depression Scale; SUVR, standardized uptake value ratio.

3.2. Depressive symptoms, anxiety symptoms, sleep duration, and amygdalar tau

Sleep duration did not significantly moderate the association between depressive symptoms and amygdalar tau burden in the overall sample or the amyloid‐positive subgroup. No significant main effects were observed. Sleep duration significantly moderated the relationship between anxiety symptoms and amygdalar tau burden. The raw distribution of STAI scores and bilateral amygdalar tau SUVR is shown in Figure S1 in supporting information. In the full sample, the STAI × sleep interaction was significant after adjusting for covariates (β = −0.01, SE = 0.00, t[386] = –2.28, P = 0.023). Main effects for both STAI (β = 0.08, SE = 0.04, P = 0.021) and sleep duration (β = 0.11, SE = 0.05, P = 0.041) were also significant. Among amyloid‐positive participants, a significant STAI × sleep interaction was observed (β = –0.01, SE = 0.01, t[335] = –2.50, P = 0.013). There were also significant main effects of STAI (β = 0.11, SE = 0.04, P = 0.009) and sleep duration (β = 0.13, SE = 0.06, P = 0.026).

In the overall sample, simple slopes analyses demonstrated that at a low sleep duration (−1 SD ≈ 6.0 hours; n = 112, 28.5% of the sample), the relationship approached significance (β = 0.02, SE = 0.01, t = 1.89, P = 0.060). At the mean (≈ 7.0 hours) and +1 SD (≈ 8.1 hours), slopes were non‐significant. Johnson–Neyman analysis revealed that the effect of STAI on tau burden was significant when sleep duration was outside the range of 5.82 to 9.93 hours. Given the observed sleep range (4.0–9.0 hours), only the lower bound was meaningful. Within amyloid‐positive individuals, at a low sleep duration (−1 SD ≈ 6.0 hours; n = 95, 27.8% of the sample), STAI scores were significantly associated with greater amygdalar tau burden (β = 0.02, SE = 0.01, t = 2.48, P = 0.010; Figure 2). Slopes were not significant at the mean or higher levels of sleep duration. Johnson–Neyman analysis revealed that the effect of STAI on tau burden was significant when sleep duration was outside the range of 6.66 to 10.37 hours (Figure 3).

FIGURE 2.

Interaction between anxiety symptoms STAI and sleep duration on amygdalar tau burden among amyloid‐positive participants. The lines represent the relationship between STAI and bilateral amygdalar tau burden at low (−1 SD), average (mean), and high (+1 SD) levels of sleep duration; shaded bands indicate 95% confidence intervals. At low sleep, higher anxiety was associated with increased tau burden, while this relationship was attenuated at average and high sleep durations. SD, standard deviation; STAI, State Trait Anxiety Inventory; SUVR, standardized uptake value ratio.

FIGURE 3.

Johnson–Neyman plot depicting the moderating effect of sleep duration on the association between anxiety symptoms (State Trait Anxiety Inventory [STAI]) and amygdalar tau burden among amyloid‐positive participants. The plot shows the slope of STAI predicting tau as a function of sleep duration. The shaded blue region indicates values of sleep where the relationship is statistically significant (P < 0.05), while the pink region indicates non‐significance. The gray shading denotes the observed range of sleep duration in the sample. Results show that the effect of anxiety on tau burden is significant only at lower levels of sleep (< 6.66 hours), suggesting that short sleep amplifies the association between anxiety symptoms and tau accumulation.

3.3. Regional specificity analysis: entorhinal and inferior temporal tau

To evaluate the regional specificity of the observed associations with amygdalar tau, we conducted parallel moderation analyses in two additional early tau regions: the bilateral entorhinal cortex and bilateral inferior temporal cortex. Models mirrored the primary analyses and included the same covariates (age, sex, and education). Amyloid status did not moderate the relationship between depressive symptoms and tau burden in either the entorhinal cortex (GDS × amyloid: β = 0.003, SE = 0.03, P = 0.913) or the inferior temporal cortex (β = −0.01, SE = 0.02, P = 0.595), and no main effects of depressive symptoms on tau burden were observed in either region. Similarly, sleep duration did not moderate the association between anxiety symptoms and tau burden in the entorhinal cortex (STAI × sleep: β = −0.006, SE = 0.006, P = 0.233) or inferior temporal cortex (β = −0.004, SE = 0.004, P = 0.341), and no significant main effects of anxiety symptoms or sleep duration were observed.

3.4. Exploratory analyses

To assess the potential influence of genetic risk, APOE ε4 carrier status was included as an additional covariate in sensitivity analyses of all primary models. Adjustment for APOE ε4 modestly attenuated effect sizes but did not alter the overall direction or pattern of the amyloid × depressive symptom or sleep × anxiety interactions, though some effects were reduced to trend‐level significance. Across all models, APOE ε4 carrier status was independently associated with greater bilateral amygdalar tau burden (all P < 0.001).

In exploratory moderation analyses, APOE ε4 carrier status significantly moderated the relationships between NPS and bilateral amygdalar tau burden. Significant interactions were observed between APOE ε4 status and depressive symptoms (β = 0.06, P = 0.009) and between APOE ε4 status and anxiety symptoms (β = 0.03, P = 0.016), such that higher levels of depressive and anxiety symptoms were associated with greater amygdalar tau burden primarily among ε4 carriers. No significant NPS–tau associations were observed among non‐carriers.

To evaluate the robustness of primary findings with respect to imaging acquisition, all models additionally included PET scanner class as a fixed‐effect covariate. Inclusion of scanner did not alter the direction or statistical significance of any primary effects.

Finally, to provide clinical context, associations among bilateral amygdalar tau burden, NPS, and global cognitive performance were examined using the PACC. Higher bilateral amygdalar tau burden was associated with worse global cognitive performance after adjustment for age, sex, and education (β = −1.88, P < 0.001). In contrast, neither anxiety nor depressive symptoms were independently associated with cognitive performance. Full regression results are provided in Table S4 in supporting information.

4. DISCUSSION

In this study, we examined whether amyloid status and sleep duration moderate the relationships between NPS, specifically anxiety and depression, and bilateral amygdalar tau burden in individuals at risk for AD. We found that amyloid status moderated the association between depressive symptoms and bilateral amygdalar tau burden, such that increased depressive symptoms were related to higher amygdalar tau only in amyloid‐positive individuals. Conversely, sleep duration moderated the relationship between anxiety and tau burden in both the overall sample and amyloid‐positive individuals. Increased anxiety symptoms were associated with higher tau levels in individuals reporting shorter sleep duration, especially in the presence of amyloid. Observed relationships among depressive symptoms, anxiety symptoms, amyloid status, sleep duration, and tau burden were specific to the amygdala and were not evident in other early tau‐accumulating regions. These findings highlight potentially distinct mechanisms through which depression and anxiety symptoms may relate to AD pathology in the preclinical stage and suggest that sleep may be an important modifiable factor in the relationship between anxiety and tau accumulation. Although amygdalar tau was associated with cognitive performance, anxiety and depressive symptoms were not, consistent with the cognitively unimpaired, preclinical nature of the cohort, in which subtle tau‐related cognitive differences may be detectable on sensitive composites prior to the emergence of clinically meaningful symptom–cognition associations.

While our primary aims focused on amyloid status and sleep as moderators, APOE ε4 represents a major source of biological heterogeneity in preclinical AD. In exploratory analyses, APOE ε4 carrier status also moderated the relationships between anxiety and depressive symptoms and amygdalar tau burden, such that NPS–tau associations were evident primarily among ε4 carriers. These findings suggest that genetic vulnerability may heighten sensitivity to neuropsychiatric processes in the context of early tau accumulation and may help explain individual differences in symptom emergence during the preclinical stage. However, because APOE moderation was not a pre‐specified aim, these results should be interpreted as hypothesis generating and warrant further investigation.

Our findings contribute to a growing body of research linking depressive symptoms to AD‐related tau pathology and suggest that this relationship, particularly in the bilateral amygdala, differs by amyloid status. ⁹ , ¹¹ , ³² , ³³ These results are consistent with our prior work, which demonstrated that amyloid‐positive individuals demonstrated increased tau binding in the medial and lateral amygdala, and that this tau pathology was associated with depressive symptoms. In that study, we also observed alterations in functional connectivity from amygdalar subregions to temporal, insular, orbitofrontal, and retrosplenial regions, highlighting how disruptions in networks supporting emotional processing may contribute to mood‐related symptoms. ¹¹ Similarly, studies of PSEN1 mutation carriers have shown that tau accumulation in the lateral amygdala is associated with both depressive symptoms and memory decline. ³³ These findings align with evidence that the amygdala is an early site of tau deposition, with elevated PET signal observed in this region up to a decade before clinical diagnosis. ⁵³ Recent work also suggests that the amygdala may serve as a key node in the propagation of tau pathology within the medial temporal lobe due to its strong reciprocal connectivity with early neurofibrillary tangle–affected regions such as the entorhinal cortex and anterior hippocampus. ⁵³ , ⁶³ , ⁶⁴ This may help explain why tau accumulates in the amygdala early in the disease course, particularly in individuals with existing amyloid pathology. In this context, our finding that depressive symptoms were associated with amygdalar tau only in amyloid‐positive individuals suggests that depressive symptoms may reflect emerging tau‐related changes in emotion‐regulating circuits, but only once a certain threshold of underlying Aβ pathology is present. Consistent with this interpretation, sensitivity analyses modeling amyloid burden continuously suggested that this moderating effect reflects a threshold associated with amyloid positivity rather than a linear dose response across the full range of amyloid burden (Table S3). Furthermore, by examining tau burden across the bilateral amygdala, our study expands upon these unilateral or subregion‐specific findings to offer a more comprehensive view of how depressive symptoms may relate to tau accumulation within the amygdala, a key structure involved in emotion processing. Therefore, our study highlights a conditional relationship in which depressive symptoms and amygdalar tau burden co‐occur primarily in the presence of amyloid in preclinical AD.

Our finding that sleep moderated the relationship between anxiety symptoms and amygdalar tau burden—both in the overall study population and among amyloid‐positive individuals—offers a novel contribution to the literature. Anxiety has been linked most consistently to amyloid burden in preclinical and prodromal AD, whereas evidence for anxiety‐related tau effects has been mixed. ⁹ , ¹⁶ , ²² , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ These results help clarify the less understood relationship between anxiety and tau pathology and point to sleep as a key behavioral factor. Sleep disturbances are highly prevalent in individuals with AD, affecting up to 40% of those with dementia and often preceding cognitive decline. ³⁹ , ⁴⁰ Prior research suggests that poor sleep exacerbates anxiety symptoms and impairs glymphatic clearance of toxic proteins, including tau. ³⁹ , ⁴⁶ , ⁴⁷ , ⁴⁸ Experimental studies in animal models demonstrate that sleep deprivation accelerates tau spread, and human imaging research shows that poor sleep is associated with increased tau burden and functional disruptions in limbic regions. ⁴⁴ , ⁴⁵ , ⁵⁷ Additionally, sleep loss increases amygdala reactivity to emotional stimuli, which may compound anxiety‐related stress and neuroinflammatory responses in at‐risk individuals. ⁵⁶ It is possible that insufficient sleep co‐occurs with anxiety in contexts associated with greater amygdalar tau burden, rather than uniformly across individuals. In this sense, sleep duration may represent a contextual modifier of the association between anxiety symptoms and tau burden. Given that anxiety and sleep disturbances frequently co‐occur, ⁴⁶ our findings suggest that sleep may be a critical factor to consider when interpreting anxiety‐related differences in tau burden in preclinical AD.

There are several limitations to this study. First, our sample did not include individuals with clinically significant psychiatric disorders, which may limit generalizability to broader populations. Because we examined subclinical depressive and anxiety symptoms, future studies should explore whether similar relationships hold in individuals with major depressive disorder or generalized anxiety disorder. Second, our assessment of sleep relied on self‐reported duration as part of a lifestyle questionnaire, and future studies should incorporate validated instruments like the Pittsburgh Sleep Quality Index (PSQI) or objective measures such as actigraphy or polysomnography. Furthermore, we had limited representation of individuals reporting longer sleep durations (> 9 hours), which restricted our ability to assess non‐linear or U‐shaped associations between sleep and tau burden. Although very short sleep durations were also relatively uncommon (n = 6 reporting 4 hours; n = 23 reporting 5 hours), the distribution was considerably sparser at the upper end, which primarily limited our ability to characterize associations across the full sleep spectrum. Prior work has suggested that both short and long sleep durations are associated with increased depression and amyloid burden, indicating that our findings may underestimate effects at the extremes of sleep. ⁴⁹ , ⁵⁰ Third, although we tested interaction effects consistent with effect modification, such models are not designed to establish temporal directionality or underlying mechanisms linking sleep, affective symptoms, and tau burden. Nevertheless, the observed interaction patterns help identify key relationships and candidate pathways that can be examined in future work to better understand how NPS and sleep disturbances relate to early AD pathology. Finally, although we focused on the amygdala due to its relevance to emotional processing and early tau accumulation, future research could examine broader neural circuits in which the amygdala plays a key role, such as the amygdala‐prefrontal and salience network connectivity.

Future studies should build on these findings by using longitudinal designs to clarify the temporal sequence among NPS, sleep, and AD pathology. Chronic stress and hypothalamic–pituitary–adrenal (HPA) axis dysregulation, common in both depression and anxiety, have been implicated in Aβ production and tau accumulation via elevated glucocorticoid levels in animal models, while serotonergic and inflammatory pathways have been proposed as potential links between mood and anxiety symptoms to neurodegeneration. ²⁰ , ⁶¹ , ⁶⁵ Additionally, the glymphatic system plays a critical role in clearing toxic proteins such as tau and amyloid during sleep, and its dysfunction may contribute to their accumulation. ³⁹ , ⁶⁶ , ⁶⁷ Therefore, investigating biological pathways and processes such as HPA axis dysregulation, inflammatory markers (e.g., glial fibrillary acidic protein [GFAP], neurofilament light chain [NfL]), and glymphatic function may help elucidate the hypothesized biological pathways linking mood symptoms, sleep disturbances, and neurodegenerative processes. Overall, integrating these biological and network‐level approaches will be essential to understanding how mental health symptoms interact with AD pathology and sleep disturbances across the disease continuum.

Collectively, these findings provide insight into how biological and behavioral moderators, specifically amyloid status and sleep duration, influence the relationship between NPS and tau accumulation. Our moderation analyses suggest that these symptoms may be more strongly linked to tau burden in distinct ways, which has important clinical implications. Depressive symptoms may serve as early behavioral markers of tau pathology in amyloid‐positive individuals, underscoring the importance of mental health screening in at‐risk populations. In contrast, the relationship between anxiety symptoms and tau appears more context dependent, and our findings suggest that improving sleep may buffer the deleterious effects. Future trials should assess whether targeted behavioral interventions to enhance sleep quality and reduce anxiety might delay or mitigate AD progression during this critical window. Of note, our results were observed in a cognitively unimpaired population, highlighting the relevance of neuropsychiatric and lifestyle factors even in the preclinical stage of AD. Although NPS are well documented in the later stages of AD, relatively little is known about their role in the preclinical phase. Our findings suggest that depressive symptoms may function more as early indicators of underlying AD pathology, whereas anxiety symptoms, particularly in the context of shorter sleep duration, may reflect a potential risk pathway that contributes to or amplifies tau accumulation. Taken together, these results support an interactive model of disease progression, where NPS, early AD pathology, and sleep intersect to influence vulnerability to AD‐related changes, with implications for early detection and targeted behavioral interventions in preclinical AD.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All study sites obtained institutional review board approval, and participants provided written informed consent prior to enrollment. ¹¹ , ²³ , ⁵⁸ , ⁵⁹ , ⁶⁰

Supporting information

Supporting Information

Supporting Information