Damage to the hippocampus in obstructive sleep apnea: a link no longer missing

The findings by Owen and colleagues [1] complete a circle of evidence confirming brain injury in the hippocampus in people with obstructive sleep apnea (OSA). Before direct examination of the brain in OSA, circumstantial evidence such as cognitive and mood symptoms [2] suggested disrupted higher brain function [3]. Specifically, impaired memory and high levels of depressive symptoms are common in OSA [4–6], and both functions are strongly associated with the hippocampus [7, 8]. Animal models of intermittent hypoxia, one characteristic of OSA, showed neuronal death in the hippocampus [9, 10], consistent with the theory that OSA damages the hippocampus. The earliest direct measurements in the brain of OSA patients involved magnetic resonance spectroscopy (MRS), and these showed lower levels of N-acetylaspartate in white matter, consistent with lower metabolic activity and possible damage [11, 12]. Early magnetic resonance imaging (MRI) studies consisted of standard visual assessments, and some association was found between OSA and subclinical pathology (but not major pathology [13]), specifically white matter hyperintensities (WMH) [14]. Such hyperintensities, so named because they appear on some MRI scans as bright “spots” in white matter, reflect mostly vascular damage and are associated with normal aging as well as hypertension [15, 16]. The amount of WMH in OSA is beyond normal aging, but given the overlap between OSA and hypertension [17], it is not clear whether this brain pathology is specific to the sleep disorder. After these studies in the late 90s and early 2000s, new methodology allowed statistical analyses of MRI metrics, leading to multiple studies assessing brain structure and function in OSA. (A PubMed search for “obstructive sleep apnea brain MRI” will bring up dozens of these.) More recently, our group has used newer MRS techniques to measure concentrations of multiple chemicals in specific brain regions, including lower γ-aminobutyric acid in the insula [18], and higher myo-inositol (possibly indicative of glial activation) and glutamate (likely reflecting high excitation or possibly excitotoxicity) in multiple areas [18–21]. When considered in their totality, these MRI findings are consistent with OSA-related structural changes in the brain, including specific alterations in the hippocampus [22, 23], or changes in surrounding tissue (parahippocampal gyrus [24], temporal cortex [25, 26]) that may also reflect hippocampal deficits. In other words, signs of hippocampal damage in the literature of the past 30 years include symptoms indicating dysfunction, animal models of intermittent hypoxia, and many neuroimaging studies. The Owen’s findings now add a missing link, direct measures of hippocampal pathology in human OSA patients.

What is the nature of this injury, and is it preventable or reversible? While the pathology findings in the Owen’s study show a reduction in thickness and loss of myelin, in vivo human studies show mixed effects, with our recent findings indicating a combination of volume increases and decreases across different subfields of the hippocampus [27]. Treatment of OSA with continuous positive airway pressure (CPAP) can partially normalize both symptoms and hippocampal structure [23]. Reversibility of brain deficits with CPAP suggests that structural changes seen in OSA are not all permanent, which points to some adaptive or pathological changes short of neuronal death. However, given the Owen’s findings of atrophy in human tissue, and apoptosis in animal models of intermittent hypoxia, it seems reasonable to assume there is some permanent damage to neurons. The Owen’s study also suggests myelin is impacted and not protected by CPAP treatment. The most likely scenario is then a combination of irreversible cell death and reactive changes in the cerebral cellular environment. For example, inflammation [28] and more specifically hypoxia-induced astrocyte activation [29] could be adaptive changes that are reversed with CPAP treatment. Evidence consistent with such adaptive changes includes volume increases [27] and lower mean diffusivity [30], an MRI metric inversely related to cell size. The lesser damage that Owen and colleagues show with CPAP may reflect reversal or prevention, but longitudinal MRI studies confirm that treatment does lead to normalization of hippocampal structure [23] (suggesting reversal).

So, hippocampal damage occurs in OSA and is somewhat reversible, but does it matter? The question of what role altered brain function plays in the symptoms or genesis of OSA has been unclear for many years [31]. There appear to be links between symptoms and hippocampal injury [32, 33], so it is possible that OSA-induced injury causes hippocampal dysfunction and associated symptoms. However, there are few data that show directionality between symptoms and brain alterations, so the hippocampal damage may just be coincident or even downstream of the stresses associated with OSA and possibly its comorbidities. The fact that we use sophisticated neuroimaging markers or unique human tissue measures does not alter the fact that correlation is not causation. To further our understanding of the importance of hippocampal injury in OSA, we need longitudinal studies assessing the impact of CPAP on the relationship between brain injury and symptoms.

One more point: sex matters. An unexpected finding in our recent study of the hippocampus using precise MRI techniques was the dramatic difference in how females and males were affected by OSA [27]. We performed an initial analysis of a mixed group of 66 patients versus a control population and assessed the impact of OSA while, as we typically do, factoring sex with a categorical covariate. However, we also analyzed females and males in separate analyses, which revealed dramatically larger volume changes than seen in the mixed analysis. The disparate findings with the two analysis approaches indicate that sex in studies of the brain in OSA cannot be simply treated as a nuisance variable.

In summary, Owen and colleagues have provided an important validation of hippocampal injury in OSA. As with many imaging and other human studies, there are limitations that caution our interpretation of the source of the injury, and how specific it is to OSA. Nevertheless, the field is in a good position to address the broader question of does this injury matter; we suggest looking at this question in females and males separately.

Conflict of interest statement. None declared.