Depression severity and sleep efficiency's associations with amyloid burden are additive but dependent on underlying genetic risk in late life depression

Abstract

Introduction

Evidence linking late‐life depression (LLD) to amyloid beta (Aβ) burden remains mixed, possibly in part due to limited examinations of the confounding role of comorbid sleep disturbance and the moderating role of the apolipoprotein E (APOE) ɛ4 gene.

Methods

We analyzed 110 older adults (ages 65–88, 66% female) with LLD to jointly model depression severity and sleep disturbance's relationship with Aβ, and to examine the moderating role of APOE ɛ4 genotype.

Results

Depression severity, sleep efficiency, and sleep duration did not demonstrate a direct relationship with Aβ; however, a significant APOE ɛ4 moderation effect occurred for depression severity and sleep efficiency. In APOE ɛ4 carriers only, higher depression severity and lower sleep efficiency were associated with greater total Aβ burden and indicated greater likelihood of reaching clinical cutoffs for amyloid positivity. The effects of depressive symptoms and sleep efficiency were statistically unique and additive. Exploratory analyses demonstrated APOE ɛ4 also moderated the effect of lifetime duration of untreated depression on increased Aβ.

Discussion

Depression severity and sleep efficiency's relationship with Aβ may be dependent on the presence of the APOE ɛ4 allele. This, among other individual differences, may partially explain mixed findings examining LLD as a risk factor for Alzheimer's disease. Furthermore, these findings do not suggest that the relationship between depression severity and amyloid is explained by the confounding role of sleep duration or sleep efficiency.

Highlights

• Depression and sleep's associations with amyloid burden were dependent on the apolipoprotein E (APOE) ɛ4 allele.

• Even mild depressive and sleep symptoms may relate to higher amyloid burden for APOE ɛ4 carriers.

• Depression and sleep efficiency appear to have additive relationships to amyloid burden.

• Lifetime untreated depression burden may be a risk factor for higher amyloid burden in late life.

• The APOE ɛ4 allele may explain mixed findings relating late‐life depression to amyloid.

1. BACKGROUND

Approximately 150 million people worldwide will meet criteria for a dementia diagnosis by 2050, ¹ the largest proportion of which is likely to be Alzheimer's disease (AD). ² The amyloid hypothesis revolutionized understanding of the pathogenesis of AD, proposing that the accumulation of amyloid beta (Aβ) pathology in the brain catalyzes a cascade of phosphorylated tau, neuroinflammation, and neurodegeneration, impairing cognition and functional independence. ³ However, recent updates to this framework propose greater examination of AD subtypes based on underlying genetic risk. ⁴ In the more prevalent sporadic AD, risk depends to a greater degree on complex gene and environment interactions and lifestyle risk factors. ³ , ⁴ , ⁵

The apolipoprotein E (APOE) ɛ4 allele has the largest known genetic effect in sporadic AD, conferring an estimated 3‐ to 4‐fold increased risk of AD by promoting the deposition of Aβ, impairing its clearance, and altering the progression of tau‐mediated neurodegeneration. ⁶ , ⁷ The APOE ɛ4 allele has also been linked to AD pathogenesis via more indirect pathways, including changes to neuroinflammatory processes, dysregulation of lipid metabolism, vascular dysfunction, and altering integrity of the blood–brain barrier. ⁶ Furthermore, emerging evidence suggests the APOE ɛ4 allele may have downstream effects on sleep and mood regulation that may confer further vulnerability to increased amyloid burden. ⁸ , ⁹

Major depressive disorder (MDD) and depressive symptoms have both been identified as risk factors for dementia, ¹⁰ but have a contested relationship with amyloid accumulation. While major late‐life depression (LLD) may reflect a prodrome of dementia, ¹¹ , ¹² and individuals with depression have an increased risk of a future AD diagnosis, ¹⁰ several recent findings have failed to find a link between LLD symptom severity and amyloid burden, ¹³ , ¹⁴ and have even found that individuals with LLD may have lower amyloid burden than non‐depressed controls. ¹³ However, while many studies control for APOE as a covariate, few have examined it as a moderating factor. Importantly, research has linked APOE ɛ4 to depression‐like symptoms in human and animal studies, ⁸ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ suggesting a possible genetic pathway for the emergence of depressive symptoms that may be altered in carriers compared to non‐carriers.

The APOE ɛ4 allele has also been linked to increased risk of disturbed sleep both subsequent and prior to the emergence of cognitive symptoms in AD. ¹⁹ Sleep disturbance has a well‐established bidirectional relationship with amyloid pathology. ²⁰ , ²¹ Sleep mechanistically clears amyloid pathology from the brain ²² and sleep deprivation studies have demonstrated that restricted sleep duration increases amyloid burden. ²³ Additionally, amyloid accumulation decreases sleep efficiency, sleep quality, sleep duration, and increases sleep latency and fragmentation. ²⁰ , ²¹ However, these findings have been equivocal, including when aggregating across studies. ²⁴ Recently, Feng et al. found that only APOE ɛ4 carriers demonstrated a positive relationship between sleep disturbance and amyloid burden. ²⁵ However, these findings did not assess multiple sleep metrics, nor did they account for the presence of significant depressive symptoms.

RESEARCH IN CONTEXT

1. Systematic Review: The authors reviewed the literature to assess possible explanations for the mixed literature relating late‐life depression to increased amyloid. Both depressive symptoms and sleep dysfunction are linked to the apolipoprotein E (APOE) ɛ4 allele and are highly comorbid, yet little research has examined nuanced relationships between these variables in their relation to amyloid burden.

2. Interpretation: Greater depression severity, lower sleep efficiency, and greater untreated lifetime depression burden were correlated with increased amyloid burden for APOE ɛ4 carriers, but not non‐carriers. For APOE ɛ4 carriers, sleep and depressive symptoms had additive relationships with amyloid, despite high comorbidity.

3. Future Directions: Whether late‐life depression and sleep dysfunction is accompanied by increased amyloid burden may be partially explained by the moderating role of APOE ɛ4. Future research would benefit from examining directionality of relationships between depression and sleep dysfunction with amyloid in APOE ɛ4 carriers, and whether mood and sleep interventions can reduce amyloid burden in high‐risk individuals.

To date, no research has examined whether symptoms of sleep disturbance and major depression, as commonly co‐occurring syndromes, are distinctly related to amyloid burden in APOE ɛ4 carriers and non‐carriers. Here, we hypothesize that mixed findings in the literature can be explained in part by a moderating role of APOE ɛ4 genotype, and whether sleep and depression are uniquely related to amyloid. We hypothesize that lower sleep duration, lower sleep efficiency, and greater depression severity will have a stronger relationship with amyloid burden in APOE ɛ4 carriers, and that jointly accounting for sleep and depression will yield better estimates of amyloid burden.

2. METHODS

2.1. Participants

Participants consisted of 110 older adults enrolled in the Alzheimer's Disease Neuroimaging Initiative – Depression Project (ADNI‐D; ¹³ NCT02434393), a study independent from the larger ADNI project but designed to collect similar data from individuals with depression. The study was conducted in accordance with the Declaration of Helsinki for protection of human subjects, with procedures approved by the institutional review boards of each study site. Participants provided written informed consent upon enrollment in the study. Participants completed neuropsychological assessment, magnetic resonance imaging (MRI), amyloid positron emission tomography (PET), blood draw APOE genotyping, and self‐reported measures of emotional and physical health. Data were collected across two visits at either the University of California, San Francisco, or the University of Pittsburgh Medical Center—an MRI/PET session and a behavioral/cognitive session, aimed to be completed within a 2‐week timeframe.

Participants were considered eligible if they were age ≥ 65, and met criteria for a current diagnosis of unipolar MDD, without psychotic features, with symptom severity of at least 15 on the Hamilton Depression Rating Scale ²⁶ and a current episode of depression lasting at least 6 weeks at the time of screening. Diagnosis of MDD was made by a licensed clinical psychologist using the Structured Clinical Interview (SCID) for the Diagnostic and Statistical Manual of Mental Disorders 4th Edition (DSM‐IV). As previously reported, ¹³ participants had an average of 2.5 (standard deviation [SD] = 1.8) depressive episodes in their lifetime, averaging 263.4 months (SD = 223.8), of which for 139.1 months (SD = 190.0) participants reported their symptoms were untreated. This sample consisted of individuals with both early‐onset depression (before the age of 65) and late‐onset depression (after the age of 65), and thus while all individuals met criteria for a current depressive episode, this sample does not exclusively examine individuals with depression occurring for the first time in late‐life.

Exclusion criteria consisted of: the presence of another Axis I disorder, including history of bipolar disorder; diagnosis of a neurological disorder such as epilepsy, Parkinson's disease, traumatic brain injury, or cortical stroke; diagnosis of dementia or a score of < 25 on the Mini‐Mental State Examination. ²⁷

2.2. Measures

To assess amyloid burden, PET imaging was completed in the same manner as the ADNI project. ²⁸ Data were collected in four, 5‐minute frames acquired between 50 and 70 minutes post‐injection with florbetapir. Images were co‐registered to structural MRI, processed with FreeSurfer version 5.3. ²⁹ The cortical summary standardized uptake value ratio (SUVR) was calculated by averaging amyloid across the frontal, cingulate, temporal, and parietal regions referenced to the whole cerebellum. For the primary analyses, amyloid burden was treated as a continuous variable, and for a supplementary analysis, participants were designated as amyloid positive or negative based on a 1.11 threshold.

For APOE genotyping, blood sample DNA extraction was performed using commercial reagents (FlexiGene, Qiagen). Two APOE single‐nucleotide polymorphisms, rs7412 and rs429358, were typed with allelic discrimination assays using TaqMan reagents (Applied Biosystems) and designated as alleles ɛ2, ɛ3, or ɛ4. For the current analysis, genotype was categorized as a dichotomous variable based on the presence or absence of at least one ɛ4 allele.

Sleep measures were taken from self‐reported data on the Pittsburgh Sleep Quality Index. ³⁰ Sleep efficiency was calculated as the amount of estimated time spent sleeping divided by the estimated amount of time in bed, averaged over the preceding 2 weeks. A sleep efficiency score of below 80% to 85% is generally considered a clinical target for sleep restriction in the treatment of insomnia. ³¹ Sleep duration was measured as the participant's average estimated nightly hours of sleep over the preceding 2 weeks.

Given the primary aim to assess independence of sleep and depressive symptomatology, depression symptom severity was measured using the 15‐item Geriatric Depression Scale (GDS‐15) ³² because it does not include items assessing sleep. Scores of 5 to 8 are typically considered indicative of mild depression, 9 to 11 as moderate depression, and 12 to 15 as severe depression. To assess depression history, participants completed a self‐reported retrospective measure that was additionally reviewed together with research coordinators. This depression history chart, developed specifically for the ADNI‐D study ¹³ and using the basic structure of the National Institute of Mental Health's life‐chart method, ³³ consisted of timeline charts of depressive symptoms, depressive episodes, and medication/therapy use across the participant's lifetime. Here, “lifetime depression” refers to the number of lifetime months participants reported experiencing depressive symptoms that met DSM‐IV criteria for a depressive episode, as verified by study staff. Similarly, “lifetime depressive episodes” refers to the number of distinct depressive episodes meeting DSM‐IV criteria. “Depression onset age” refers to the age at which the participant reported their first depressive episode meeting DSM‐IV criteria. “Lifetime untreated depression” refers to months with “lifetime depression” minus the number of months receiving treatment (medication or therapy), regardless of efficacy.

2.3. Statistical analyses

The primary goal of this analysis was to (1) evaluate whether the relationship between Aβ and sleep efficiency, sleep duration, and depression severity differs for APOE ɛ4 carriers versus non‐carriers, and (2) determine whether sleep symptoms and depressive symptoms are statistically unique in their relationship with Aβ for APOE ɛ4 carriers and non‐carriers.

A base model consisted of age, sex, education, and binarized APOE ɛ4 status regressed on Aβ PET SUVR. Age, sex, and education were included as covariates in all subsequent models due to their previously reported relationships with Aβ. ³⁴ , ³⁵ , ³⁶ Variables were normalized prior to analyses. All statistical analyses were conducted using RStudio version 2024.12.1.

The primary analyses consisted of three linear regression models examining the significance of interaction terms of: APOE ɛ4 status by sleep efficiency; APOE ɛ4 status by sleep duration; and APOE ɛ4 status by depression severity, each regressed on Aβ PET SUVR. We examined whether any of these significant moderation effects for sleep and depression were statistically unique by creating a final, combined model with multiple moderation effects simultaneously for any significant moderators identified in the first aim. This joint model was probed using the Johnson–Neyman technique to identify significant cut points of interaction terms.

To examine the relevance of these findings to currently used clinical cutoff scores for amyloid positivity, and to examine the robustness of our findings, the final combined model was repeated using logistic regression to regress significant sleep and depression interaction terms on a binary measure of amyloid‐positive versus amyloid‐negative individuals. We repeated our final combined model analysis after controlling for outliers by winsorizing amyloid, sleep, and depression variables at a two SD threshold.

As an exploratory analysis, we examined whether the presence of an APOE ɛ4 allele moderated the relationships between aspects of self‐reported depression history (total lifetime duration of depression in months, lifetime duration of untreated depression, number of lifetime depressive episodes, age of depression onset) and amyloid burden, additionally controlling for current self‐reported depression severity.

3. RESULTS

3.1. Demographics and descriptive statistics

The 110 participants had a mean age of 70.71 years (range = 65–88; SD = 5.03), 16.27 years of education on average (SD = 2.05), 66.36% were female, 19.09% were amyloid positive (for the full sample, M = 1.06, SD = 0.15), and 28.18% had at least one APOE ɛ4 allele (Table 1). The majority of participants identified as White (86.36%), followed by Asian (4.55%), more than one race (2.73%), Black or African American (1.82%), Native Hawaiian or Other Pacific Islander (1.82%), American Indian or Alaskan Native (0.91%), and undisclosed (1.82%); 80.90% of individuals identified as non‐Hispanic or Latin American, and 19.09% identified as Hispanic or Latino. Participants reported mild depression severity of 7.10 on the GDS (SD = 3.27). Sleep efficiency averaged 79.62% (SD = 16.31) with 40% of the sample demonstrating a clinically significant sleep efficiency of < 80%. Sleep duration was on average 6.66 hours per night (SD = 1.60).

TABLE 1.

Demographic and clinical characteristics of late‐life depression sample.

	APOE ɛ4 allele carriers (n = 31) Mean (SD)	APOE ɛ4 allele non‐carriers (n = 79) Mean (SD)	Total sample (n = 110) Mean (SD)
Age, years	71.16 (4.97)	70.53 (5.08)	70.71 (5.03)
Education, years	16.42 (2.35)	16.22 (1.93)	16.27 (2.05)
Sex, female (%)	74.19%	63.29%	66.36%
Geriatric Depression Scale ‐ 15	6.42 (3.27)	7.37 (3.25)	7.10 (3.27)
Sleep efficiency (%)	82.74 (14.63)	78.40 (16.86)	79.62 % (16.31)
Sleep duration (hours)	7.02 (1.49)	6.53 (1.63)	6.66 (1.60)
Amyloid SUVR	1.17 (0.38) ***	1.02 (0.09) ***	1.07 (.0.14)
Amyloid positive	45.16% **	8.86% **	19.09%

Note: APOE ɛ4 allele carriers demonstrated significantly higher amyloid burden (p < 0.001) and amyloid positivity rates (p < 0.01). No other significant group differences were found.

Abbreviations: APOE, apolipoprotein E; SD, standard deviation; SUVR, standardized uptake value ratio.

*p < 0.05; **p  < 0.01; ***p  < 0.001.

3.2. Base model and direct effects of sleep and depression on amyloid PET

A base model examining the relationships among age, years of education, sex, and APOE ɛ4 status accounted for an adjusted r ² of 19.2% of the variance in amyloid burden F(4, 105)  =  7.48, p   =  2.43e‐05, with APOE ɛ4 status as the only significant predictor variable (B  =  −0.47, p  =  4.81e‐07). In separate direct models with these same covariates, there was no significant relationship between sleep efficiency (B  =  −0.10, p  =  0.24; model‐adjusted r²   =  19.5%), total sleep duration (B  =  −0.07, p  =  0.44; model‐adjusted r²   =  18.9%), or depression severity with amyloid burden (B  =  −0.02, p  =  0.83; model‐adjusted r²   =  18.5%).

3.3. Primary analyses: individual moderation analyses

As hypothesized, the relationship between sleep efficiency and amyloid burden was significantly moderated by the presence of at least one APOE ɛ4 allele (B  =  −0.21, p  =  0.02; model‐adjusted r²   =  22.6%), controlling for the effects of age, years of education, and sex, such that individuals with lower sleep efficiency demonstrated higher amyloid burden only if they were APOE ɛ4 positive. This relationship was unaffected when additionally controlling for the effects of depression severity as measured by the GDS as shown in Figure 1A (B  =  −0.22, p  =  0.02; model‐adjusted r²   =  22.0%). The relationship between sleep duration and amyloid burden did not reach significance; however, the effect occurred in the expected direction, controlling for age, years of education, and sex (B  =  −0.16, p  =  0.09; model‐adjusted r²   =  20.3%) and when controlling for depressive symptoms (B  =  −0.16, p  =  0.09; model‐adjusted r²   =  19.7%).

FIGURE 1.

Individual moderation effects of APOE ɛ4 status by sleep efficiency and depression, regressed on PET global Aβ. A, Presence of at least one APOE ɛ4 allele moderated the relationship between sleep efficiency and amyloid burden, such that only APOE ɛ4 carriers exhibited a negative relationship between better sleep efficiency and higher amyloid burden. B, Similarly, the relationship between depression severity and amyloid burden was significantly moderated by APOE ɛ4 status, such that only APOE ɛ4 carriers exhibited a positive relationship between greater depression severity and higher amyloid burden. *p < 0.05; **p < 0.01; ***p < 0.001. Aβ, amyloid beta; APOE, apolipoprotein E; PET, positron emission tomography; SUVR, standardized uptake value ratio.

The moderation effect of APOE genotype on the relationship between depressive symptoms and amyloid burden was significant (B  =  0.22, p  =  0.02; model‐adjusted r²   =  22.2%), controlling for the effects of age, years of education, and sex, such that amyloid burden increased with higher depressive symptoms for APOE ɛ4 carriers (Figure 2). The significance of this relationship remained unchanged when additionally controlling for the effects of sleep efficiency and sleep hours (B  =  0.22, p  =  0.008; model‐adjusted r²   =  21.9%), as shown in Figure 1B.

FIGURE 2.

Johnson–Neyman technique probing of moderation effects. A, The Johnson–Neyman technique showed that the effect of APOE ɛ4 carrier status on global amyloid burden was significant (p < 0.05) when sleep efficiency was < 98%. The negative relationship between presence of an APOE ɛ4 allele on amyloid burden (slope) becomes weaker as sleep efficiency increases. B, The effect of APOE ɛ4 carrier status on global amyloid burden was significant (p < 0.05) when depression severity on the GDS was > 3.11. The positive relationship between presence of an APOE ɛ4 allele on amyloid burden (slope) becomes weaker as depression severity decreases. Note: Plotting has been limited to show the estimated effect only within the range of observed data. APOE, apolipoprotein E; GDS, Geriatric Depression Scale.

3.4. Primary analyses: joint moderation analyses

To more definitively establish independence of the relationships of sleep and depression with amyloid, we examined a model with interaction terms for sleep efficiency and depression simultaneously. Moderation effect sizes were increased for both variables (sleep efficiency: B  =  −0.24, p  =  0.009; depression severity: B  =  0.24, p  =  0.006; model‐adjusted r²   =  27.0%) as shown in Table 2. Probing these interactions using the Johnson–Neyman technique showed that the slope of the relationship between APOE ɛ4 carrier status and amyloid burden was significant for individuals with a reported sleep efficiency < 98% (Figure 2A). Similarly, the slope of the relationship between APOE ɛ4 carrier status and amyloid burden was significant for individuals with depression severity over 3.11 on the GDS (Figure 2B). These thresholds suggest that even mild sleep and mood disturbance may indicate the presence of greater amyloid burden in APOE ɛ4 carriers.

TABLE 2.

Linear regression joint model of sleep efficiency and depression severity's interaction terms with APOE ɛ4, regressed on PET global Aβ.

Term	B	SE	t	P	Partial eta2	95% CI
(Intercept)	0.02	0.14	0.13	.898	–	[−0.27, 0.30]
Age	−0.00	0.08	−0.01	.994	0.00	[−0.17, 0.17]
Education (years)	0.03	0.09	0.31	.758	0.00	[−0.15, 0.20]
Sex (female)	0.06	0.18	0.36	.720	0.00	[−0.29, 0.41]
APOE ɛ4 status	0.54	0.09	6.36	<  .001***	0.24	[0.37, 0.71]
Depression severity (GDS)	0.04	0.08	0.53	.598	0.00	[−0.12, 0.21]
Sleep efficiency (PSQI)	−0.14	0.08	−1.69	.093	0.02	[−0.31, 0.02]
APOE ɛ4 status × depression severity	0.24	0.09	2.81	.006**	0.06	[0.07, 0.41]
APOE ɛ4 status × sleep efficiency	−0.24	0.09	−2.65	.009**	0.07	[−0.42, −0.06]

Note: In a model testing the independence of the effects of interaction terms for depression severity and sleep efficiency with APOE ɛ4 carrier status, both interactions terms remained significant, suggesting non‐overlapping relationships with amyloid burden; F(8,101)  =  6.03, p  =  2.3e‐06, adjusted r²   =  27.0%.

*p < 0.05; **p  < 0.01; ***p  < 0.001.

Abbreviations: Aβ, amyloid beta; APOE, apolipoprotein E; CI, confidence interval; GDS, Geriatric Depression Scale; PET, positron emission tomography; PSQI, Pittsburgh Sleep Quality Index.

3.5. Sensitivity analyses

To examine the robustness of our final model and examine the clinical relevance of these findings, we performed the same joint model in a logistic regression after dichotomizing amyloid positive and negative individuals (Figure 3). This increased the significance of the moderation effect for sleep efficiency (B  =  −0.36, p  =  0.0002) but reduced it for depression severity (B  =  0.19, p  =  0.03; model‐adjusted r²   =  27.0%).

FIGURE 3.

Odds ratios for model predictors of logistic regression predicting amyloid positivity status. Standardized coefficients for a logistic regression model identifying predictors of amyloid burden above the standard clinical cutoff score of 1.11; F(8,101)  =  6.02, p  =  2.62e‐06, adjusted r²   =  27.0%. The factors that significantly affected odds of amyloid positivity included: presence of at least one APOE ɛ4 allele, which increased odds of amyloid positivity; followed by higher sleep efficiency, which decreased the odds of amyloid positivity; followed by higher depression severity, which increased the odds of amyloid positivity. *p < 0.05; **p < 0.01; ***p < 0.001. APOE, apolipoprotein E;

We additionally tested our final joint model after winsorizing amyloid, sleep, and depression datapoints. These analyses revealed reduced, but similar, effects for both sleep efficiency and depression (sleep efficiency: B  =  −0.22, p  =  0.019; depression severity: B  =  0.21, p  =  0.018; model‐adjusted r²   =  24.7%).

Total lifetime depression burden in months (B  =  0.22, p  =  0.022) and untreated lifetime depression burden (B  =  0.35, p  =  0.005) were significantly moderated by APOE status in their relationship to amyloid burden, while lifetime depressive episodes (B  =  0.06, p  =  0.43) and age of depression onset (B  =  −0.15, p  =  0.078) were not (Figure 4). After winsorizing datapoints, only months of untreated depression remained significant (B  =  0.27, p  =  0.014), though age of depression onset (B  =  −0.16, p  =  0.074) and total months of depression (B  =  0.17, p  =  0.074) demonstrated a small effect size.

FIGURE 4.

Moderation effects of APOE ɛ4 status depression history, regressed on PET global Aβ. Examination of whether the presence of at least one APOE ɛ4 allele moderated the relationship between amyloid burden and self‐reported (A) total lifetime depression burden in months (B  =  0.22, p  =  0.022), (B) untreated lifetime depression burden (B  =  0.35, p  =  0.005), (C) lifetime depressive episodes (B  =  0.06, p  =  0.43), and (D) age of depression onset (B  =  −0.15, p  =  0.078), controlling for age, education, sex, and current self‐reported depression severity. *p < 0.05; **p < 0.01; ***p < 0.001. Aβ, amyloid beta; APOE, apolipoprotein E; PET, positron emission tomography; SUVR, standardized uptake value ratio.

4. DISCUSSION

Depression and sleep disturbance are common, and often comorbid, health conditions in late life that are associated with cognitive impairment and increased dementia risk. ¹⁰ Despite this, linking depression to mechanisms of dementia, particularly the accumulation of amyloid, has been fraught with mixed findings. ¹³ , ¹⁴ Increasing evidence suggests that genetic risk factors for AD have varied effects on amyloid accumulation through a plethora of complex pathways, potentially including altered mood and sleep states. ⁸ , ¹⁹ , ³⁶ In the current study, we demonstrate that the relationship of depression and sleep with amyloid is additive, but dependent on underlying genetic risk factors, highlighting the importance of a precision psychiatry approach to examining risk factors of amyloid pathology.

For APOE ɛ4 allele carriers only, greater depression severity and lower sleep efficiency were associated with increased global amyloid burden. Furthermore, even mild depressive symptoms and sleep disturbance were indicative of higher amyloid burden in APOE ɛ4 carriers. While recent focus has shifted away from amyloid and toward other possible mechanisms to explain neurodegeneration and cognitive decline in LLD, ¹³ these findings suggest depression and sleep are related to amyloid burden for genetically high‐risk individuals. However, the degree to which amyloid produces cognitive decline in LLD is still unknown, and there are likely multifactorial mechanisms at play, possibly including vascular disease, tau pathology, and neuroinflammatory processes independent of amyloid and APOE. Future research is needed to explore the degree to which sleep and depression‐related amyloid increases in APOE ɛ4 carriers is a mechanism, or symptom, of neurodegeneration and cognitive decline in this population.

Previous work has reported a correlation between sleep disturbance and amyloid burden dependent on the APOE ɛ4 allele, ²⁵ , ³⁷ but has not compared between sleep features or accounted for depressive symptoms. Here, we observed that lower sleep efficiency was associated with higher amyloid burden only for APOE ɛ4 allele carriers. Despite sleep duration and sleep efficiency being related constructs, sleep duration demonstrated a weaker effect than sleep efficiency and failed to reach statistical significance. Sleep duration has previously been shown to reliably correlate with brain volume and white matter integrity, though this relationship may not be linear, ³⁸ and has also preliminarily been associated with amount of amyloid burden. ³⁹ It is possible that sleep efficiency is more linear in its relationship with amyloid, or better parses individuals with disrupted sleep (due to stress, anxiety, depression, environment disruption, pain, poor bladder function, etc.) from individuals with short, but adequately restful, sleep. This also suggests that cognitive behavioral therapy for insomnia (CBT‐I, which reliably improves sleep efficiency but does not as reliably improve total sleep time compared to control interventions) may be a possible intervention to address the sleep contribution to amyloid burden ⁴⁰ , ⁴¹ , ⁴² if sleep is a mechanism of amyloid accumulation and not simply a symptom. However, the effect of the APOE ɛ4 allele on response to CBT‐I, and whether CBT‐I can meaningfully reduce amyloid burden in this subpopulation by improving sleep is still unknown.

Poor sleep and depression appear additive in their relationship to volumetric brain changes in aging, ⁴³ but no study to date has examined their simultaneous relationship to amyloid burden as moderated by APOE in LLD. Here, we observed that joint depression severity and sleep efficiency increased overall model fit, suggesting unique, independent relationships with amyloid. Our exploratory findings preliminarily suggest greater lifetime depression burden may be a risk for greater amyloid accumulation; however, whether sleep and depression are modifiable risk factors or symptoms in APOE ɛ4 carriers is still unknown.

Thus, a key question in the field is the directional nature of the relationships among sleep disturbance, depression, and amyloid accumulation. Mechanistically, there is plausibility in bidirectionality for both sleep and depressive symptoms. Exogenously induced sleep disturbance disrupts amyloid clearing processes in the brain that occur during deep sleep; sleep disturbance may also arise as a symptom of amyloid pathology in the brain, ²⁰ perhaps mechanistically linked to the effects of the APOE ɛ4 allele. ⁹ Similarly, depressive symptoms may exacerbate inflammatory processes and vascular disease, increasing dementia risk. ⁴⁴ , ⁴⁵ LLD has also been suggested as a prodrome of dementia, driven by neurodegenerative processes altering emotion‐regulation circuits in the brain. ¹⁵ It is also possible that the APOE ɛ4 allele affects depression through unknown pathways that are either related to or independent of amyloid, as has been preliminarily investigated in mouse models. ⁸

Therefore, the key limitation of this study is its cross‐sectional nature and its inability to elucidate the degree to which sleep and mood dysregulation in these high‐risk individuals are symptomatic, mechanistic, or both. Additionally, participants recruited from this study were highly educated and not racially or ethnically representative of the US population; therefore, the generalizability of these findings may be limited. It is also unclear whether these relationships are limited to individuals with diagnosable depression, or whether they would extrapolate to individuals with sub‐threshold depressive symptoms. Furthermore, sleep data and depressive history data used in this study were based on self‐report only and therefore may be less accurate than either objective measures (such as polysomnography) or measures that may reduce consistency bias (such as ecological momentary assessment). More research is needed to examine the sensitivity of these findings to different depression phenotypes, samples with greater depression severity, assessment methods, and subpopulations of LLD.

Our findings suggest that the relationship between amyloid and sleep and mood symptoms is dependent on the APOE ɛ4 allele, which is known to alter the accumulation of amyloid and increase risk of AD. While it is unclear whether lower sleep efficiency and higher depressive symptom burden are a consequence of amyloid accumulation or risk factors, these findings nevertheless suggest that genetic testing may tangibly inform preventative health measures. Further, this study suggests that sleep and depression are each uniquely related to amyloid burden, and therefore may require distinct intervention efforts.

CONFLICT OF INTEREST STATEMENT

Dr. Shankar has no conflicts of interest to disclose and no funding sources to report. Dr. Kryza‐Lacombe has no conflicts of interest to disclose and is funded by the VA VISN21 Early Career Award Program. Dr. Insel has consulted with Roche and Merck and has no funding sources to report. Dr. Harris has no conflicts of interest to disclose and no funding sources to report. Dr. Kassel has no conflicts of interest to disclose and no funding sources to report. Dr. Goldstein has no conflicts of interest to disclose and receives funding from the National Institutes of Health (NIH) and Deparment of Defense (DoD). Dr. Krystal reports serving as a consultant for Axsome Therapeutics, Abbvie, Big Health, Eisai, Evecxia, Harmony Biosciences, Idorsia, Janssen Pharmaceuticals, Jazz Pharmaceuticals, Neurocrine Biosciences, Neumora, Neurawell, Otsuka, Sage, and Takeda. He receives stock options from Neurawell and Big‐Health. He receives research funding from Alkermes, Janssen Pharmaceuticals, Axsome Pharmaceutics, Attune, Eisai, Harmony, Neumora, Neurocrine Biosciences, The Ray and Dagmar Dolby Family Fund, Weill Institute for Neurosciences, and the National Institutes of Health. Dr. Landau reports serving on DSMBs for KeifeRX and the NIH IPAT study. She has received speaking honoraria from Eisai, Johnson & Johnson, IMPACT‐AD and serves as a consultant for Banner Health. She is on the JAMA Neurology editorial board and receives research funding and other support from NIH, Johnson and Johnson, the Alzheimer's Association, and ATRI. Dr. Tosun has no conflicts of interest to report and receives funding from the National Institute on Aging (NIA). Dr. Butters has no financial disclosures to report and receives funding from the NIH. Dr. Aisen consults with Merck, Roche, Genentech, Abbvie, Biogen, ImmunoBrain Checkpoint, and Arrowhead, and receives funding from NIH and Lilly and Eisai. Dr. Raman reports receiving funding from Eisai and receiving grant funding through her University from the National Institute on Aging, Alzheimer's Association, American Heart Association. Dr. Toga has no financial disclosures to report and receives funding from the NIA, Alzheimer's Associate, and Fox Foundation. Dr. Jack has no financial disclosures to report and receives funding from the NIA. Dr. Nelson reports being an advisor and consultant to Biohaven, Clexio Biosciences Ltd., Johnson and Johnson, Neurocrine Biosciences Inc., Novartis, and Otsuka and received royalties from UpToDate. He receives research support from the National Institute of Mental Health (NIMH). Dr. Mackin receives research support from Johnson and Johnson, and receives research funding from the NIA and NIMH. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

The study was conducted in accordance with the Declaration of Helsinki for protection of human subjects, with procedures approved by the institutional review boards of each study site. All participants provided written informed consent upon enrollment in the study.

Supporting information

Supporting Information: dad270478‐sup‐0001‐ICMJE.pdf