ÌÇÐÄvlog

Key Points

QuestionÌý Are the numbers of subcortical neurons, which often start degenerating early in neurodegenerative diseases, correlated with clinical sleep phenotypes in patients with Alzheimer disease (AD) or progressive supranuclear palsy (PSP)?

FindingÌý In this cohort study of 10 patients with AD and 9 patients with PSP, electroencephalographic and polysomnographic assessments of overnight sleep were performed, followed by a comprehensive immunohistochemical neuronal analysis of important subcortical neurons. The numbers of orexinergic lateral hypothalamic area, histaminergic tuberomammillary nucleus, and noradrenergic locus coeruleus neurons were significantly correlated with homeostatic sleep behavior in patients with AD and PSP.

MeaningÌý The study’s findings suggest that loss of subcortical wake-promoting neurons at the early stages of neurodegenerative disease may be sufficient to reduce regulation of sleep-wake homeostasis in patients with AD and PSP.

ImportanceÌý Sleep disturbance is common among patients with neurodegenerative diseases. Examining the subcortical neuronal correlates of sleep disturbances is important to understanding the early-stage sleep neurodegenerative phenomena.

ObjectivesÌý To examine the correlation between the number of important subcortical wake-promoting neurons and clinical sleep phenotypes in patients with Alzheimer disease (AD) or progressive supranuclear palsy (PSP).

Design, Setting, and ParticipantsÌý This longitudinal cohort study enrolled 33 patients with AD, 20 patients with PSP, and 32 healthy individuals from the Memory and Aging Center of the University of California, San Francisco, between August 22, 2008, and December 31, 2020. Participants received electroencephalographic and polysomnographic sleep assessments. Postmortem neuronal analyses of brainstem hypothalamic wake-promoting neurons were performed and were included in the clinicopathological correlation analysis. No eligible participants were excluded from the study.

ExposuresÌý Electroencephalographic and polysomnographic assessment of sleep and postmortem immunohistological stereological analysis of 3 wake-promoting nuclei (noradrenergic locus coeruleus [LC], orexinergic lateral hypothalamic area [LHA], and histaminergic tuberomammillary nucleus [TMN]).

Main Outcomes and MeasuresÌý Nocturnal sleep variables, including total sleep time, sleep maintenance, rapid eye movement (REM) latency, and time spent in REM sleep and stages 1, 2, and 3 of non-REM (NREM1, NREM2, and NREM3, respectively) sleep, and wake after sleep onset. Neurotransmitter, tau, and total neuronal counts of LC, LHA, and TMN.

ResultsÌý Among 19 patients included in the clinicopathological correlation analysis, the mean (SD) age at death was 70.53 (7.75) years; 10 patients (52.6%) were female; and all patients were White. After adjusting for primary diagnosis, age, sex, and time between sleep analyses and death, greater numbers of LHA and TMN neurons were correlated with decreased homeostatic sleep drive, as observed by less total sleep time (LHA: r = −0.63; P = .009; TMN: r = −0.62; P = .008), lower sleep maintenance (LHA: r = −0.85; P < .001; TMN: r = −0.78; P < .001), and greater percentage of wake after sleep onset (LHA: r = 0.85; P < .001; TMN: r = 0.78; P < .001). In addition, greater numbers of LHA and TMN neurons were correlated with less NREM2 sleep (LHA: r = −0.76; P < .001; TMN: r = −0.73; P < .001). A greater number of TMN neurons was also correlated with less REM sleep (r = −0.61; P = .01). A greater number of LC neurons was mainly correlated with less total sleep time (r = −0.68; P = .008) and greater REM latency (r = 0.71; P = .006). The AD-predominant group had significantly greater sleep drive, including higher total sleep time (mean [SD], 0.49 [1.18] vs −1.09 [1.37]; P = .03), higher sleep maintenance (mean [SD], 0.18 [1.22] vs −1.53 [1.78]; P = .02), and lower percentage of wake after sleep onset during sleep period time (mean [SD], −0.18 [1.20] vs 1.49 [1.72]; P = .02) than the PSP-predominant group based on unbiased k-means clustering and principal component analyses.

Conclusions and RelevanceÌý In this cohort study, subcortical wake-promoting neurons were significantly correlated with sleep phenotypes in patients with AD and PSP, suggesting that the loss of wake-promoting neurons among patients with neurodegenerative conditions may disturb the control of sleep-wake homeostasis. These findings suggest that the subcortical system is a primary mechanism associated with sleep disturbances in the early stages of neurodegenerative diseases.

Sleep disturbance is common among patients with neurodegenerative diseases. For instance, patients with Alzheimer disease (AD) experience excessive daytime sleepiness and sundowning.^1-4 Progressive supranuclear palsy (PSP) features hyperarousal and decreased homeostatic sleep drive.⁵ Sleep disturbance generally precedes disease-defining symptoms, often by decades, suggesting that dysregulation of sleep is important in the early pathogenesis of neurodegeneration.^6-8 In experimental models, sleep disruption has been found to impair essential cellular and extracellular mechanisms, including redox homeostasis, proteostasis, and glymphatic flow.^6,9,10 In humans, studies of cerebrospinal fluid have reported that the sleep-wake cycle is associated with the accumulation and clearance of disease-associated proteins.^11-13 Moreover, the accumulation of neurodegenerative-related proteins has consequences for sleep.^14,15 Overall, sleep disturbances and early neurodegenerative processes have a bidirectional association, although it remains unclear which is the initial factor.

Identifying neural substrates of specific sleep-wake patterns may inform novel treatments to address early symptoms of neurodegenerative diseases and help to slow overall disease progression.^16-18 Sleep disturbances in patients with AD have been associated with the accumulation of β-amyloid in the cortical structures, even though subcortical nuclei involved in sleep-wake physiological processes accumulate tau before β-amyloid. Previous studies have reported that the groups of subcortical neurons associated with wakefulness, also known as wake-promoting neurons, are vulnerable to AD-tau copathology, producing extreme neuronal loss.^19-22 The same network is protected in other tauopathies, such as PSP and corticobasal degeneration, despite tau accumulation,¹⁹ which may result in clinical manifestations, such as arousal deficiencies in AD and sleep deficiencies in PSP. This contrast provides a distinct opportunity to develop tailored treatments for sleep disorders among patients with neurodegenerative diseases. Nevertheless, quantitative clinicopathological studies exploring the associations between clinical sleep phenotypes and subcortical lesions in tauopathies are lacking; these studies are important to understanding the early-stage sleep neurodegenerative phenomena.

To bridge this gap in knowledge, we examined the correlation between the number of important subcortical wake-promoting neurons (assessed using quantitative postmortem neuronal analysis) and clinical sleep phenotypes (assessed using objective sleep measurements) in patients with AD and PSP. We focused on a highly interconnected and interdependent subcortical system involving noradrenergic locus coeruleus (LC), orexinergic lateral hypothalamic area (LHA), and histaminergic tuberomammillary nucleus (TMN), given their substantial role in sleep-wake physiological processes in animal models and their early involvement in neurodegenerative diseases.^19,20 Furthermore, we compared sleep measurements among patients with AD and PSP with those of healthy individuals. We hypothesized that the number of subcortical wake-promoting neurons would be correlated with sleep duration and fragmentation among patients with AD and PSP. Alzheimer disease and PSP are well suited to providing insight into the neuronal basis of sleep in the human brain because they are on opposite ends of the spectrum of the clinical sleep-wake phenotype, and they exhibit a divergent pattern of neuropathological changes in the sleep-wake networks.

This cohort study was approved by the institutional review board of the University of California, San Francisco. All participants provided written informed consent. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology () reporting guideline.

Participants were recruited from other longitudinal studies conducted at the Memory and Aging Center of the University of California, San Francisco, between August 22, 2008, and December 31, 2020. A total of 33 patients with AD, 20 patients with PSP, and 32 healthy individuals were enrolled. All participants received electroencephalographic (EEG) and polysomnographic (PSG) sleep assessments; postmortem neuronal analysis of brainstem hypothalamic wake-promoting neurons was performed for 10 patients with AD and 9 patients with PSP, and the data were included in the clinicopathological correlation analysis. Although the goal was to recruit individuals of multiple races and ethnicities, all of those who consented to participate in both the clinical and pathological aspects of the study within the reporting period were White.

Exclusion criteria were head trauma with persistent deficits, encephalitis, neoplastic disease, active substance misuse, demyelinating disease, severe periventricular white matter disease or white matter lesions greater than grade 4, clinically meaningful lacunar infarctions, cortical stroke (including cortical microbleed), and the use of medications likely to impact central nervous system functions (with the exception of drugs approved by the US Food and Drug Administration for the treatment of AD). Cognitively healthy older adults were recruited for the control group (as previously described).^5,23 We included individuals diagnosed with possible or probable PSP (based on the Litvan criteria²⁴) or probable AD (based on criteria from the National Institute on Aging and the Alzheimer Association) who underwent a sleep study and had postmortem examinations. No eligible participants were excluded from the study. All neuropathological assessments were performed using standardized protocols and followed the internationally accepted criteria for neurodegenerative diseases.^25-27 Study recruitment is shown in Figure 1.

The methods used for clinical sleep assessments have been previously described.^5,23 Throughout the sleep study, participants continued their usual medication regimen. Sleep among 20 patients with PSP and 18 healthy individuals in the control group was evaluated using nocturnal PSG and video, digitally recorded at 400 Hz using a portable sleep monitoring system (Beehive Horizon; Grass Technologies) and, in a single case, an ambulatory recorder (TREA; Grass Technologies) at the Clinical Research Center of the Memory and Aging Center. Signals were recorded from 6 EEG scans (F3, F4, C3, C4, O1, and O2) referenced to contralateral mastoids (M1 and M2), 6 electrooculographic scans (Fp1, Fp2, below, and lateral to E1/E2), 3 mentalis and submentalis electromyographic scans, 4 bilateral electromyographic scans of the anterior tibialis, 4 bilateral electromyographic scans of the extensor digitorum, a 2-point electrocardiographic scan, 2 respiratory effort belts using inductance plethysmography on the chest and abdomen, a thermistor and nasal cannula pressure transducer, and a pulse oximeter for detection of oxygen desaturation events.

Sleep among 33 patients with AD and 14 healthy individuals in the control group was evaluated at the Clinical Research Center of the Clinical and Translational Science Institute at the University of California, San Francisco, Moffitt Hospital. Monitoring included long-term recording using silver cup electrodes in the standard international 10 to 20 electrode array with an additional lead to record electrocardiography. The EEGs included video telemetry recordings. After initial setup, participants were asked to hyperventilate for 3 minutes, rest, and breathe normally for 7 minutes with eyes closed. The EEG recordings then continued overnight for a mean (SD) duration of 17.22 (3.40) hours among those with AD and 18.80 (3.80) hours among those in the control group.

Measures of interest were total sleep time (in minutes), wake after sleep onset (in percentages), sleep maintenance (in percentages), and percentage of time in rapid eye movement (REM) sleep and stages 1, 2, and 3 of non-REM (NREM1, NREM2, and NREM3, respectively) sleep during the sleep period time (the time between sleep onset and the last epoch of sleep). All measures were scored and calibrated by certified sleep experts (one of whom was L.Y.) using the American Academy of Sleep Medicine criteria.²⁸

The postmortem human brains of 10 patients with AD and 9 patients with PSP were processed at the Neurodegenerative Disease Brain Bank of the Memory and Aging Center. Detailed project-specific protocols for tissue processing and analysis have been previously described.^19,22 To summarize, celloidin-embedded brainstem and diencephalon blocks were cut in coronal or horizontal serial thick sections for stereological studies. All free-floating sections were double stained against phospho-Ser202 tau (mouse monoclonal antibody CP13 at 1:1000 dilution; Peter Davies donation) and one of the following: (1) for LC neurons, tyrosine hydroxylase (rabbit polyclonal antibody AB5935 at 1:500 dilution; MilliporeSigma), the rate-limiting enzyme in the synthesis of norepinephrine; (2) for LHA neurons, orexin A/hypocretin-1 peptide (orexin A rabbit polyclonal antibody H-003-30 at 1:500 dilution; Phoenix Pharmaceuticals); or (3) for TMN neurons, histidine decarboxylase (rabbit polyclonal antibody 03-16045 at 1:250 dilution; ARP American Research Products), the enzyme responsible for histamine synthesis. Sections were counterstained with gallocyanin (Nissl stain; Thermo Fisher Scientific).

The protocol for the optical fractionator probe has been previously described.^19,22,29 The LC neurons were quantified in 60-μm–thick sections at 1200-μm intervals. The hypothalamic tissues (ie, LHA or TMN) were quantified in 30-μm–thick sections at 300-μm intervals. Stereologically determined estimates were made for (1) the neurotransmitter-producing (ie, tyrosine hydroxylase⁺, orexin⁺, or histidine decarboxylase⁺) neuronal population, (2) the tau⁺ neuronal population, and (3) the total neuronal population. The proportion of neurons that were positive for different subpopulations was determined by dividing the number of neurons that were positive for a given marker (estimates 1 or 2) by the total neuron population (estimate 3), represented as a percentage.

Wilcoxon tests were used to investigate the homogeneity of demographic and clinical variables in each cohort. Multiple linear regression analysis was used to assess the correlation between clinical and pathological variables. All models were adjusted for primary neuropathological diagnosis, age, sex, and time between clinical sleep study and death. The MASS::stepAIC function was used to perform a stepwise model selection using Akaike information criterion and evaluate the well-fitted covariate combination in multiple linear regressions. In addition, an unbiased clustering analysis was conducted based on pathological neuronal counts using k-means algorithms. To fill missing counts for LC, LHA, or TMN neurons, a linear regression model corrected by neurofibrillary tangle (NFT) Braak staging (stages 0-VI, with 0 indicating no cortical NFTs; stages I-II indicating NFTs confined mainly to the transentorhinal region; stages III-IV indicating NFT involvement in the limbic regions, such as the hippocampus; and stages V-VI indicating moderate to severe NFT involvement in the neocortical regions) was used.

To select the optimal k-means cluster number, the gap statistic for estimating the number of clusters algorithm with 10 000 bootstraps was used. The k-means function was performed with 10 000 bootstraps at a random nstart parameter value, and cluster compositions were identical in 99.33% of results. Two groups were found: AD-predominant and PSP-predominant. Spearman correlations were used for principal component analysis (PCA) of both clinical and pathological measurements. To compare results, z scores for clinical variables were derived based on normative data from the combined healthy control cohort.

All statistical analyses were conducted using R software, version 4.1.1 (R Foundation for Statistical Computing). The threshold for statistical significance was 2-tailed P &±ô³Ù; .05.

Among 19 patients with neurodegenerative diseases included in the clinicopathological correlation analysis, the mean (SD) age at death was 70.53 (7.75) years; 10 patients (52.6%) were female, and 9 (47.4%) were male. All patients were White. The mean (SD) time between the sleep study and death was 3.80 (2.63) years. The AD group (n = 10) had a significantly lower age at death (mean [SD], 66.80 [7.99] years vs 74.67 [5.17] years; P = .01) and a significantly higher test lag time (mean [SD], 5.16 [2.68] years vs 2.22 [1.60] years; P = .02) than the PSP group (n = 9). Among 32 healthy individuals in the control group, 18 participants (56.2%) were female, 14 (43.8%) were male, and all were White. Additional demographic, clinical, and pathological features of participants are shown in the Table.

Patients with PSP had a lower sleep drive than those with AD, as observed by significantly shorter total sleep time (mean [SD], 285.39 [106.83] minutes vs 399.90 [96.38] minutes; P = .04), lower sleep maintenance (mean [SD], 61.20% [18.21%] vs 80.16% [13.60%]; P = .01), and lower percentage of time in NREM2 sleep during the sleep period time (mean [SD], 29.15% [11.23%] vs 43.04% [9.94%]; P = .006). Patients with PSP vs AD also had a higher percentage of wake after sleep onset during the sleep period time (mean [SD], 183.72 [115.15] minutes vs 96.40 [56.28] minutes; P = .03). We found no significant differences in other clinical sleep measurements (Table).

We quantified subcortical neuropathological profiles among those with AD vs PSP. Similar to previous findings in an independent cohort,¹⁹ all 3 nuclei exhibited greater loss of wake-promoting neurons in patients with AD compared with patients with PSP. The percentage of tau⁺ LHA neurons was significantly higher in those with PSP vs AD (mean [SD], 44.14% [18.35%] vs 24.78% [10.38%]; P = .03), whereas the percentage of tau⁺ TMN neurons was higher in those with AD vs PSP (mean [SD], 24.23% [12.71%] vs 9.83% [4.40%]; P = .03). The percentage of neuron-synthesizing neurotransmitters was similar among those with AD and PSP in all nuclei (Table).

To examine the correlation between clinical sleep phenotypes and neuronal counts of individual nuclei, we first created a Spearman correlation coefficient plot as an exploratory analysis (Figure 2). A greater number of LHA or TMN wake-promoting neurons was correlated with decreased sleep drive, including shorter total sleep time (LHA: r = −0.63; P = .009; TMN: r = −0.62; P = .008), lower sleep maintenance (LHA: r = −0.85; P < .001; TMN: r = −0.78; P < .001), and greater percentage of wake after sleep onset during the sleep period time (LHA: r = 0.85; P < .001; TMN: r = 0.78; P < .001) (Figure 2). A greater percentage of tau⁺ TMN neurons was positively correlated with total sleep time (r = 0.51; P = .04), and a greater number of LC neurons was associated with shorter total sleep time (r = −0.68; P = .008) and greater REM latency (r = 0.71; P = .006).

Using the same approach, we examined the correlation between different components of sleep (percentage of NREM1, NREM2, NREM3, or REM sleep during the sleep period time) and the neuronal numbers of LC, LHA, and TMN. A greater number of LHA neurons was correlated with lower percentage of time in NREM2 sleep during the sleep period time (r = −0.76; P < .001) and lower percentage of time in NREM3 sleep during the sleep period time (r = −0.62; P = .01). A greater number of TMN neurons was correlated with a lower percentage of time in NREM2 sleep during the sleep period time (r = −0.73; P < .001) and a lower percentage of time in REM sleep during the sleep period time (r = −0.61; P = .01). The LC neuronal count was not correlated with any sleep measures assessed (Figure 2). After adjusting for all covariates (primary diagnosis, age, sex, and test lag time), all of these correlations remained statistically significant, with the exceptions of the correlation between LHA neuronal count and percentage of time in NREM3 sleep during the sleep period time and the correlation between tau⁺ percentage of TMN neurons and total sleep time.

Given the interconnectivity and interdependency of the 3 wake-promoting nuclei in governing physiological sleep processes, we sought to examine how summative pathological lesions of the 3 nuclei were correlated with clinical sleep measurements. We inputted neuronal counts (ie, neurotransmitter⁺ neurons, tau⁺ neurons, and total neurons) from all 3 nuclei in patients with AD and PSP into a k-means clustering algorithm. Two robust clusters were found: AD-predominant (n = 11) and PSP-predominant (n = 8).

The AD-predominant group had significantly greater sleep drive, including higher total sleep time (mean [SD], 0.49 [1.18] vs −1.09 [1.37]; P = .03), higher sleep maintenance (mean [SD], 0.18 [1.22] vs −1.53 [1.78]; P = .02), and lower percentage of wake after sleep onset during the sleep period time (mean [SD], −0.18 [1.20] vs 1.49 [1.72]; P = .02) compared with the PSP-predominant group. Regarding sleep architecture, the AD-predominant group exhibited a higher percentage of time in NREM2 sleep during the sleep period time (mean [SD], 0.32 [1.05] vs −1.06 [1.26]; P = .003) than the PSP-predominant group. We did not detect differences between the groups with regard to REM latency or percentage of time in NREM1, NREM3, or REM sleep during the sleep period time (Figure 3).

We conducted a PCA, a multivariate analysis that maximizes variance, to confirm the correlation between summative clinical and pathological sleep measurements. Two PCAs were generated using clinical or pathological measurements (Figure 4). The first principal component (PC1) represented the overall summary of the main variations. The PC1 composition of the clinical PCA was most associated with total sleep time (11.38%), sleep maintenance (22.82%), percentage of wake after sleep onset (20.55%), percentage of time in NREM2 sleep (18.35%), and percentage of time in NREM3 sleep (10.71%). The PC1 composition of pathological PCA was most associated with total LC (18.74%), LHA (17.74%), and TMN neurons (18.77%) and with the percentage of tau⁺ TMN neurons (15.90%) and tau⁺ LHA neurons (9.18%). Using Spearman correlation analysis, we detected a significant negative correlation (r = −0.59; P = .008) between the clinical and pathological PC1s, suggesting that patients with a low hypothalamic neuronal count and higher hypothalamic tau burden (predominantly patients with AD) had higher sleep drive (ie, higher total sleep time and sleep maintenance).

In this clinicopathological cohort study, we conducted objective sleep measurements using PSG or EEG and postmortem neuronal analysis among patients with AD and PSP. Results suggested that a greater number of subcortical wake-promoting neurons, particularly the hypothalamic orexinergic and histaminergic neurons, were correlated with decreased sleep drive among individuals with neurodegenerative diseases, as observed in lower total sleep time and sleep maintenance and greater percentage of wake after sleep onset, even after adjusting for covariates. In addition, a greater number of hypothalamic wake-promoting neurons was associated with a lower percentage of time in NREM2 and REM stages. The number of LC neurons was associated with lower total sleep time and greater REM latency. Using unbiased pathological clustering of all 3 nuclei, we found that the AD-predominant group had significantly greater sleep drive than the PSP-predominant group because of neuronal loss in the subcortical wake-promoting hub, which was consistent with previous findings.¹⁹ The PCA using summative clinical and pathological measurements corroborated these findings. Altogether, the results of our study suggest that subcortical wake-promoting neurons have a substantial role in sleep-wake physiological processes in humans by decreasing sleep promotion. Among individuals with neurodegenerative conditions, loss of wake-promoting neurons at the early stages of disease may be sufficient to decrease regulation of sleep-wake homeostasis.

Individuals with AD and PSP, while harboring pathological tau lesions, exhibit sleep-wake disturbances differently. From a clinical perspective, patients with AD exhibit increasing severity of arousal deficiencies. In contrast, those with PSP, a 4-repeat tauopathy,³⁰ exhibit sleep deficiencies that are characterized by hyperarousal and difficulty falling and remaining asleep.^5,31,32 Patients with PSP have difficulty recuperating sleep debt after a poor night of sleep, unlike older adults with intact homeostatic sleep drive.⁵ Consistent with previous findings, our study found significantly lower sleep drive among those with PSP vs those with AD, as observed in lower total sleep time and sleep maintenance and greater percentage of wake after sleep onset.

The neuropathological hallmarks of AD are β-amyloid plaques and tau accumulation, with tau appearing in subcortical nuclei before β-amyloid accumulates in the neocortex.³³ Sleep disturbances in AD have been associated with pathological changes from cortical β-amyloid. However, the early emergence of sleep disturbances in AD more closely aligns with the time course of subcortical tau accumulation and neuronal loss, as is observed in LC neurons.^1,20,22 Tau accumulation is sufficient to cause sleep disturbances, which is exemplified in patients with PSP, a pure tauopathy³⁴ that lacks β-amyloid accumulation. The results of the present study corroborate previous findings suggesting that the subcortical arousal network involving LC, LHA, and TMN neurons is vulnerable to severe neuronal loss after AD-tau accumulation. Notably, the same network is relatively spared in those with PSP, even at late stages.¹⁹ This differential vulnerability of wake-promoting neurons provides a clear rationale for the differences in clinical sleep phenotypes observed in individuals with AD vs PSP.

The quest to understand sleep-wake circuits in animal models has evolved as a result of the availability of new tools, such as those used for in vivo calcium imaging and optogenetics.^35-37 However, in the sleep field, the translational gap between animal models and humans is substantial. Personalized sleep medications for the treatment of patients with neurodegenerative conditions remain underdeveloped.³⁸ Although several studies have used neuroimaging and cerebrospinal fluid,^17,39-41 only a few clinicopathological studies to date have combined objective sleep measurements with stereological postmortem neuronal quantification of subcortical sleep nuclei in those with neurodegenerative diseases,^42,43 and no studies have investigated wake-promoting neurons. Lim et al⁴² used wrist actigraphy, finding that individuals with more neurons in the hypothalamic intermediate nuclei had less fragmented sleep, suggesting intermediate nuclei are the human homologues of the sleep-promoting ventrolateral preoptic nuclei in mice. From the same research group, Wang et al⁴³ reported a correlation between the number of vasoactive intestinal peptide-expressing neurons in the suprachiasmatic nuclei and circadian rhythm amplitude of motor activity in patients with AD and healthy individuals. The findings of both studies^42,43 suggested that the number of specific hypothalamic neurons was significantly correlated with homeostatic behavior in humans. In the current study, we used the clinical sleep phenotypes of AD vs PSP, which represent opposite ends of the spectrum, and their stereotypical patterns of neuronal loss, finding that greater numbers of hypothalamic orexinergic and histaminergic neurons were correlated with lower sleep drive. In addition, we found that the number of LC neurons was positively correlated with REM latency. Thus, our data suggest these subcortical neurons play a substantial role in facilitating wakefulness.

The LC, LHA, and TMN neurons facilitate sleep-wake behavior in an interconnected and interdependent manner.^35,36 Thus, we also examined the wake-promoting hub as a whole. We inputted summative pathological variables of the 3 nuclei into an unbiased k-means clustering algorithm. We found 2 clusters that were identified as AD-predominant and PSP-predominant. The AD-predominant group had significantly higher sleep drive than the PSP-predominant group, suggesting overall unbalanced degeneration of wake-promoting neurons and relatively spared sleep-promoting centers, even as tau progressed. Further PCA corroborated these findings. We also hypothesized that the decrease in homeostatic sleep drive observed in patients with PSP may have been associated with degeneration of subcortical sleep-promoting nuclei in the setting of spared wake-promoting nuclei.⁴⁴ Given that quantitative data on pathological changes in subcortical sleep-promoting neurons in tauopathies is limited, further evaluation is warranted.^42,44,45

This study has limitations. First, although we performed a combined analysis of wake-promoting neurons, we lack data on the sleep-promoting and circadian-regulating neuronal counterparts, thereby limiting interpretation. In PSP, while wake-promoting neurons are relatively spared, sleep-promoting neurons may be selectively vulnerable. In addition, it would be of high interest to examine how the upstream circadian suprachiasmatic nucleus neurons differ in AD vs PSP given emerging evidence of associations between circadian rhythms and neurodegeneration.^43,46,47 Second, our sample is relatively small because of the time it takes to obtain postmortem brain tissues from well-characterized participants with objective sleep measurements in longitudinal studies. However, we compensated for this limitation by using in-depth objective sleep measurements from PSG and EEG rather than self-reported questionnaires or actigraphy and by performing postmortem stereological neuronal quantification. Pure PSP cases are rare (approximately 8%),⁴⁸ and most of the participants with PSP in our study have low AD-tau copathology. However, there are no reported differences in the main clinical PSP milestones when comparing patients with pure PSP with those who have PSP with a low burden of AD-tau copathology.⁴⁸ In fact, the PSP Braak stage IV cases were clustered with the AD group.

Third, the AD and PSP cohorts underwent sleep studies using different methods (PSG vs long-term EEG monitoring). However, the same sleep experts thoroughly calibrated and scored clinical sleep data, and all brains received identical processing and neuronal analysis. Fourth, all of the participants in this study were White, and we recognize the need for diversity in neurodegenerative research to represent the general population. Despite these limitations, our longitudinal study is the first, to our knowledge, to examine the correlation between subcortical wake-promoting neurons and sleep phenotypes in patients with neurodegenerative disease using highly quantitative methods.

This cohort study found that the numbers of subcortical wake-promoting neurons were correlated with sleep phenotypes in humans, and the loss of wake-promoting neurons among patients with neurodegenerative conditions may disturb the control of sleep-wake homeostasis. In addition, the study’s findings may bring renewed attention to the neuromodulatory subcortical system as a primary mechanism in sleep disturbances among patients with neurodegenerative diseases. Further research on the specific patterns of neurodegeneration within the subcortical network, including sleep-promoting nuclei and the circadian system, may inform tailored treatment strategies for sleep disturbances and early symptoms of neurodegenerative diseases, which may help to slow overall disease progression.

Accepted for Publication: January 13, 2022.

Published Online: April 4, 2022. doi:10.1001/jamaneurol.2022.0429

Corresponding Author: Lea T. Grinberg, MD, PhD, Memory and Aging Center, Department of Neurology, Sandler Neurosciences Center, Box 1207, 675 Nelson Rising Ln, Room 211B, San Francisco, CA 94158 (lea.grinberg@ucsf.edu).

Author Contributions: Drs Neylan and Grinberg had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Mr Oh and Dr Walsh were equal first authors. Drs Neylan and Grinberg were equal senior authors.

Concept and design: Oh, Walsh, Kramer, Miller, Vossel, Ranasinghe, Neylan, Grinberg.

Acquisition, analysis, or interpretation of data: Oh, Walsh, Mladinov, Pereira, Petersen, Falgàs, Yack, Lamore, Nasar, Lew, Li, Metzler, Coppola, Pandher, Le, Heuer, Heinsen, Spina, Seeley, Rabinovici, Boxer, Miller, Vossel, Ranasinghe, Neylan, Grinberg.

Drafting of the manuscript: Oh, Walsh, Nasar, Li, Coppola, Pandher, Grinberg.

Critical revision of the manuscript for important intellectual content: Oh, Walsh, Mladinov, Pereira, Petersen, Falgàs, Yack, Lamore, Lew, Metzler, Le, Heuer, Heinsen, Spina, Seeley, Kramer, Rabinovici, Boxer, Miller, Vossel, Ranasinghe, Neylan, Grinberg.

Statistical analysis: Oh, Pereira, Petersen, Nasar, Metzler, Coppola, Pandher.

Obtained funding: Walsh, Seeley, Boxer, Miller, Vossel, Neylan, Grinberg.

Administrative, technical, or material support: Walsh, Mladinov, Falgàs, Lamore, Lew, Li, Coppola, Pandher, Heuer, Heinsen, Seeley, Boxer, Miller, Ranasinghe, Neylan, Grinberg.

Supervision: Oh, Mladinov, Heinsen, Miller, Vossel, Neylan, Grinberg.

Conflict of Interest Disclosures: Dr Yack reported receiving personal fees from the University of California, San Francisco, during the conduct of the study. Dr Coppola reported receiving grants from the National Institute on Aging, the Rainwater Charitable Foundation, and the Tau Consortium, during the conduct of the study. Dr Pandher reported receiving grants from the National Institutes of Health (NIH), the Rainwater Charitable Foundation, and the Tau Consortium during the conduct of the study. Dr Spina reported receiving consulting fees from Acsel Health, Precision Xtract, and Techspert.io outside the submitted work. Dr Rabinovici reported receiving grants from the NIH during the conduct of the study; grants from the Alzheimer's Association, the American College of Radiology, Genentech, and the NIH; personal fees from Eisai, Eli Lilly and Company, Genentech, Johnson & Johnson, and Roche; and being employed as an associate editor of JAMA Neurology outside the submitted work. Dr Boxer reported receiving grants from the NIH and the Rainwater Charitable Foundation during the conduct of the study; grants from Biogen, Eisai, and Regeneron Pharmaceuticals; personal fees from Denali Therapeutics, GlaxoSmithKline, Humana, Oligomerix, Oscotec, Roche, Transposon Therapeutics, and Wave Life Sciences; nonfinancial support from Eli Lilly and Company and Novartis; and owning stock options in Alector, Arvinas, AZTherapies, and TrueBinding outside the submitted work. Dr Miller reported receiving grants from the NIH; serving as an advisor for the Arizona Alzheimer's Disease Center, the Association for Frontotemporal Degeneration, the Bluefield Project to Cure FTD, the Buck Institute for Research on Aging, the John Douglas French Alzheimer’s Foundation, the Larry L. Hillblom Foundation, the Massachusetts Alzheimer’s Disease Research Center, the National Institute for Health Research Cambridge Biomedical Research Centre (and its subunit, the Biomedical Research Unit in Dementia), Stanford University, the Tau Consortium, the University of Southern California, and the University of Washington; serving as a director or codirector of the Bluefield Project to Cure FTD, the Global Brain Health Institute, and the Tau Consortium; and being employed as affiliated faculty at the Institute for Neurodegenerative Diseases outside the submitted work. Dr Vossel reported receiving grants from the Alzheimer's Association, the John Douglas French Alzheimer's Foundation, the NIH, and the S.D. Bechtel, Jr. Foundation during the conduct of the study. Dr Grinberg reported receiving grants from the NIH and the Rainwater Charitable Foundation during the conduct of the study and personal fees from CuraSen Therapeutics and the Simons Foundation outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by grants R01AG060477 (Drs Neylan and Grinberg), R01AG064314 (Drs Neylan and Grinberg), RO1AG038791 (Dr Boxer), U54NS092089 (Dr Boxer), K24AG053435 (Dr Grinberg), and K23AG038357 (Dr Vossel) from the NIA; grant PCTRB-13-288476 from the Alzheimer’s Association, made possible by Part the Cloud (Dr Vossel); grant K08AG058749 from the NIH (Dr Ranasinghe); grants 2015-A-034-FEL and 2019-A-013-SUP from Larry L. Hillblom Foundation (Dr Ranasinghe); a grant from the Rainwater Charitable Foundation and the Tau Consortium (Drs Grinberg, Neylan, and Walsh); and funding from the John Douglas French Alzheimer’s Foundation (Dr Vossel) and the S.D. Bechtel, Jr. Foundation (Dr Vossel).

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank the participants, caregivers, and family members who were associated with this study. We also thank Tau Consortium colleagues and the Rainwater Charitable Foundation family for their continued advice and support. Several research coordinators referred and helped to schedule participants for this study, including Nik Block, BA, Ali Dallich, BA, Phi N. Luong, BSc, Natanya Russek, BA, and Lisa Voltarelli, BA, of the University of California, San Francisco (UCSF). Alaisa Emery, BA, Jonathan Varbel, BA, and Kathleen Walker, BA, of UCSF generated the clinical sleep data. We thank the UCSF clinical research coordinator nursing staff and the clinical fellows who assisted in the health care management of the participants, including Jee Bang, MD, Carolyn Fredericks, MD, Joanna Hellmuth, MD, Serggio Lanata, MD, Zachary Miller, MD, Anitha Rao-Frisch, MD, Julio Rojas, MD, and Richard Tsai, MD, of the Memory and Aging Center, UCSF. Grace Oh, BS, of WorkDay Inc, assisted with the figure designs.