Obstructive sleep apnea (OSA) is a sleep-related breathing disorder characterized by the repetitive narrowing of the pharyngeal airway, resulting in nocturnal episodes of complete (apnea) or partial (hypopnea) obstruction (1). OSA prevalence in adults ranges from 6 to 17% but increases with aging, affecting up to 49% of the elderly population (2). By chronically fragmenting sleep and provoking intermittent episodes of hypoxemia, OSA may increase oxidative stress, neuroinflammation and cerebral edema, leading to complex and potentially non-linear changes in neuroimaging markers such as cerebral gray matter volume, metabolism, white matter diffusivity, as well as greater amyloid and tau levels (3–10). Unsurprisingly, most epidemiological studies show that untreated OSA increases the risk of cognitive decline and dementia (11–15).

Structural neuroimaging has been used to identify OSA-related brain changes suggestive of early neurodegenerative processes in late middle-aged and older adults. Some brain regions appear to be particularly sensitive to the pathological consequences of OSA, such as the medial temporal lobe (MTL), and other hubs of the default mode network (DMN), such as the posterior cingulate cortex and the precuneus (4, 5, 7, 9, 10). Whether OSA is associated with functional connectivity (FC) abnormalities in older populations is unclear, as most studies have been performed in young and middle-aged adults, often in male participants only. Studies among 20–70 years old participants with moderate to severe OSA show that they have lower resting-state FC between the hippocampus and other DMN regions, such as the posterior cingular cortex (PCC), medial prefrontal cortex (mPFC) and parahippocampal region (PHC) compared to non-apneic participants (16, 17). They also have lower FC between the hippocampus and other brain regions such as the insula, anterior cingulate cortex, thalamus, medial and superior temporal cortex (17, 18). FC increases have also been documented in association with OSA between the MTL and fronto-temporal areas (18, 19), the mPFC and parietal and MTL regions (15, 16, 19, 20), and the posterior insula and both the precuneus and anterior insula (21, 22).

MTL regions are among the earliest cortical regions to degenerate in Alzheimer’s disease (AD) (23, 24), and hub regions of the DMN are prone to accumulate amyloid pathology (25–27). In the context where OSA is increasingly recognized as a risk factor for AD, it is crucial to determine whether OSA is associated with disrupted FC patterns between these regions, especially in late middle-aged and older populations. To date, the only study which was specifically performed in this age group included 72 participants aged 66.6 ± 7.9 years old, at high risk for dementia (i.e., presenting subjective and/or objective cognitive impairments and seeking assessment for cognitive decline). The authors found that higher hypoxemia levels were associated with lower FC between the left and right parahippocampal gyri in participants with OSA, as compared to those without OSA (28). Therefore, these results suggest that MTL regions may be sensitive to OSA. It is therefore of interest to confirm and extend these findings to a cohort including both cognitively unimpaired and individuals with mild cognitive impairment (MCI) to better understand the effects of OSA on FC patterns in aging.

The present study aimed at characterizing the association between OSA severity and resting-state FC between regions of the DMN and MTL in late middle-aged and older adults (from 55 to 90 years old). We explored these associations according to cognitive status (i.e., in those with and without MCI). Moreover, we investigated the associations between FC and episodic memory performance. We hypothesized that more severe OSA would be associated with lower FC between MTL regions (i.e., the hippocampus and parahippocampal cortex) and DMN regions (i.e., the mPFC, lateral parietal cortex, precuneus and PCC). Furthermore, we expected these FC changes to be associated with lower episodic memory performance or the presence of MCI.

This study aimed at characterizing the associations between OSA severity and resting-state FC between brain regions that are sensitive to OSA and AD in late middle-aged and older adults. We found that more severe OSA, as measured with higher AHI, was associated with reduced FC between the hippocampus and the mPFC, the precuneus and PCC. All associations were driven by cognitively unimpaired participants and not replicated in MCI individuals. No association with global cognition or episodic memory performance was observed.

These results show that OSA severity is associated with a pattern of reduced FC between the hippocampus and several other DMN regions, such as the mPFC, PCC and precuneus. Interestingly, DMN regions are known to be structurally affected in relation to OSA. More specifically, animal studies show that the hippocampus is particularly sensitive to hypoxia, which is a major consequence of OSA (41–43). Hippocampal pyramidal neurons are especially vulnerable to hypoxia due to their large size, and their high pre-existing levels of reactive oxygen species (41–43). Consistently, studies in humans show that OSA is associated with hippocampal damage, with greater OSA severity being associated with lower hippocampal volume (9, 44). Interestingly, the MTL is well known to be among the first cortical areas to accumulate tau pathology, a hallmark of AD which partly underlies neurodegeneration in this population (45, 46). Patients with OSA also exhibit increased tau levels (6, 47, 48), which could explain the vulnerability of MTL regions to OSA. In the present study, FC was reduced between the hippocampus and several other DMN regions, namely the mPFC, PCC and precuneus. These DMN hubs are highly metabolically active, and known to be prone to amyloid accumulation in the early stages of the disease (25–27, 49). Previous studies have shown that higher OSA severity is related to structural and metabolic alterations in fronto-parietal areas (5). Moreover, several groups have reported that moderate-to-severe OSA patients exhibit a pattern of increased amyloid deposition in the PCC and precuneus (3, 50). More specifically, a multimodal neuroimaging study has shown that amyloid deposition topographically overlapped with a pattern of increased metabolic activity (3). This suggests that greater levels of neuroinflammation and neuronal hyperactivity could increase amyloid accumulation and precede neurodegeneration in the same brain regions (5, 51). Taken together, these studies suggest that OSA may enhance AD-related pathological mechanisms and neurodegeneration in DMN hubs, which may explain the vulnerability of this network to OSA. Longitudinal studies could bring important new information regarding this hypothesis.

The DMN is also sensitive to sleep disturbances. Indeed, DMN hubs are highly active during wakefulness, but this network is disengaged in deep sleep and exhibits the highest levels of slow wave activity (52–54). However, sleep quality is largely affected by the aging process, with increased fragmentation and wakefulness levels and decreased slow wave sleep (55, 56). Therefore, it is likely that age-related sleep changes may lead to an increase in neuronal activity levels (57, 58), which could in turn enhance amyloid deposition and disrupt brain functioning. Consistently, FC alterations in the DMN were observed by several groups in the context of sleep-deprivation protocols. In young adults, 24 h of sleep deprivation have been associated with a reduction of FC between DMN nodes (59–61). A study showed that the DMN and attentional network are less anti-correlated after sleep deprivation (62).

In older adults, self-reported sleep disturbances, as measured with greater scores on the Pittsburgh Sleep Quality Index, have been associated with disrupted FC in the DMN in patients with MCI (63, 64). Compared to our study, the authors found that the DMN was sensitive to global self-reported sleep disturbances in later cognitive stages (i.e., patients with MCI), but the regions affected included the mPFC, parietal and temporal regions, which parallels our observations. Here, we identify OSA as a condition impacting FC between DMN regions in late middle-aged and older adults. Specifically in older populations at risk for dementia, OSA is associated with FC abnormalities between MTL regions (28). We extend these findings to late-middle aged and older participants who were less at risk of cognitive decline, as 58% of our sample was composed of cognitively unimpaired participants. Interestingly, we found that OSA was associated with more widespread FC changes than those previously observed (28), involving the mPFC, PCC and precuneus. The effect of OSA on FC may be greater in cognitively healthy individuals and less evident in participants more advanced in the AD continuum. Consistent with this hypothesis, analyzes stratified by cognitive status show that in our sample, all associations were significant in cognitively unimpaired participants but not in individuals with MCI. It is possible that after the apparition of the first cognitive signs, the variance in FC explained by OSA severity may be lower than the variance explained by AD pathology.

FC alterations among DMN regions may partially result from microstructural alterations in white matter tracts or gray matter atrophy (65). Several studies show that key DMN nodes are directly connected by white matter tracts (66, 67). FC patterns between the PCC and hippocampus have been associated with the microstructure of white matter tracts connecting these regions, including the cingulum bundle (68). As OSA is known to affect white matter microstructure [for a recent review, see (5)], it is possible that OSA-related FC patterns may be partly explained by the alteration of white matter tracts. However, the mechanisms underlying FC are still unclear for some parts of the DMN, including the mPFC (66, 69, 70), such that the alteration of white matter tracts may not be the only mechanism underlying OSA-related FC changes. The adverse effects of OSA on FC in older adults may also be influenced by gray matter alterations and amyloid deposition, as they have both been associated with FC changes in preclinical and prodromal AD (71–73). In fact, both increases and decreases in FC have been reported in the Alzheimer’s continuum (i.e., from cognitively unimpaired amyloid-positive older adults to MCI patients), probably in reaction to underlying neurodegenerative processes and amyloid deposition (73, 74). In later stages of the disease, FC reductions between DMN regions are more marked (74–76), notably between the hippocampus, the mPFC and PCC in participants with MCI and AD dementia (71, 74, 77, 78). We here show that OSA is associated with disturbed FC patterns between the same regions, especially in cognitively unimpaired individuals. This reinforces the idea that OSA may alter the integrity of the DMN at an early stage (i.e., in participants with likely relatively mild levels of AD pathology), while in more advanced stages of the disease, FC reductions could be more explained by underlying neurodegenerative processes. Further studies including measures of AD biomarkers will need to verify this hypothesis. Of note, the fact that the associations between OSA and FC patterns were driven by cognitively unimpaired individuals in our sample could also be influenced by the lower sample size of the MCI group (i.e., 55 cognitively unimpaired versus 39 MCI participants).

Importantly, the DMN is involved in interoception and episodic memory functioning (79). Lower FC among the DMN and limbic regions has consistently been associated with episodic memory impairment in older adults and patients with MCI or mild AD (71, 74, 80, 81). Moreover, disrupted FC in the DMN precedes the emergence of significant cognitive impairment and cortical atrophy in AD (71). In our sample, we did not observe significant associations between cognitive performance and OSA-related FC alterations. Cognitively unimpaired individuals performed in the normal range for their age and education levels on MoCA and RAVLT scores, such that there was overall little inter-individual performance variability. Therefore, these participants are able to maintain cognitive performance in the normal range despite exhibiting FC-related disruption in relation to OSA. This suggests that OSA may have an early and relatively asymptomatic impact on cerebral connectivity at first, with participants being cognitively resilient to the adverse effects of OSA on cerebral FC. However, OSA may still precipitate cognitive decline assessed with longitudinal designs. Interestingly, OSA may represent a modifiable risk factor for cognitive decline, as some studies suggest that CPAP treatment may improve cognitive functioning in patients with OSA (82–86). However, the results of large multicenter studies are still mixed (87–89). Interestingly, some studies suggest that CPAP treatment may increase FC in adults with OSA, notably in DMN regions such as the PCC and precuneus (82, 90). Taken together, the potential beneficial effects of CPAP treatment to preserve brain and cognitive integrity still need to be confirmed in older populations.

The relatively low proportion of women (i.e., 25.5% of the sample) is a limitation to the present study, preventing from robustly exploring the effect of sex on OSA-related FC disruptions. This question is however important and should be addressed in larger samples. Another limit is the use of one standard neuropsychological test (i.e., the RAVLT) to measure memory, as cognitively unimpaired individuals performed in the normal range for their age and education levels, leading to little statistical variability. A sleep-dependent memory task could have been better suited to detect subtle memory disturbances consequently to OSA in cognitively unimpaired individuals.

This study shows that OSA is associated with reduced FC between major nodes of the DMN (i.e., MTL, mPFC, PCC and precuneus). As these regions are sensitive to both OSA and AD pathological processes, and crucial for cognition, future studies will need to assess whether OSA-related FC alterations are associated with cognitive decline, and the potential of OSA treatment to maintain the integrity of neural networks.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethics committees of the Center intégré universitaire de santé et de services sociaux du Nord-de-l’île-de-Montréal and the Functional Neuroimaging Unit of the Institut universitaire de gériatrie de Montréal (#2012–697, #12–13-008 and #2010–468). The patients/participants provided their written informed consent to participate in this study.

VD, GMV, J-ML, JM, JC, NG, and CA contributed to the design and conceptualization of the study. JM, JC, and NG secured study funding. VD, GMV, M-EM-D, A-AB, KG, CL, DG, CT, NM, and CA contributed to data acquisition and preprocessing. VD and GMV performed statistical analyzes. VD, GMV, CA, and NG interpreted the data and drafted the manuscript. All authors critically revised and approved the final version of the manuscript.

This research was supported by the Canadian Institutes of Health Research [Grant numbers FDN154291, MOP123294, and MOP102631] and the Fonds de la Recherche du Québec–Santé. NG holds a Canada Research Chair in sleep disorders and brain health. JM holds a Canada Research Chair in Sleep Medicine. M-EM-D is supported by a doctoral fellowship from the CIHR and the Fonds de Recherche du Québec–Santé. A-AB is supported by the Banting Fellowship Program (CIHR). NM is supported by the Swiss National Science Foundation Postdoctoral Mobility fellowship. CA is supported by postdoctoral fellowships by the Fonds de Recherche du Québec–Nature et Technologies through the Merit scholarship program for foreign students (PBEEE), the Fonds de Recherche du Québec–Santé and Quebec Bio-Imaging network.

JC is the current recipient of an investigator driven grant from Eisai (not linked to the present study). NG has received honoraria from Eisai and sponsorships from Jazz Pharmaceuticals, Eisai and Paladin (not linked to the present study).

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

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