Cortisol homeostasis is important for healthy cognitive aging. Dysregulation of the hypothalamus pituitary adrenal (HPA) axis, characterized by chronically elevated basal cortisol levels, is often associated with accelerated aging (Prenderville et al., 2015) and declined cognitive functioning in the elderly (Lupien et al., 2007, 2009; Comijs et al., 2010). Understanding the role of cortisol on neuro-cognitive functions in non-clinical elderly population would bring insight into the prevention of the development of neuro-cognitive disorders, e.g., dementia. Furthermore, it could also provide the scientific basis for stress reduction programs that improved cognitive functions in healthy aging adults (e.g., Galvin et al., 2006).

The role of endogenous cortisol on cognitive function in healthy elderly subjects is relatively less studied. Thus far, only a few recent studies have demonstrated the association between cortisol and cognitive functions in healthy elderly subjects. For instance, Geerlings et al. (2015) observed that the morning cortisol levels measured 45 min after awakening were positively associated with cognitive processing speed and executive function, whereas the evening cortisol levels, sampled prior to sleep, were negatively associated with memory, cognitive processing speed, and executive function in healthy elderly subjects. Furthermore, low morning to evening cortisol ratio was shown to be associated with cognitive impairment in elderly men (Johar et al., 2015). In addition, Evans et al. (2012) also reported that executive function was positively linked to cortisol awakening response magnitude in a population of healthy older adults. These studies highlight the role of morning cortisol levels in cognitive function in elderly people. In fact, cortisol release is normally elevated in the morning to enhance energy, which has been reported to be associated with increased anticipatory states in 70 middle-class mothers (Adam and Gunnar, 2001), supporting the role of cortisol in facilitating cognitive performance. Nevertheless, there are also studies reporting no associations between diurnal cortisol levels, sampled at waking and 10 pm, and cognitive functions (e.g., Cox et al., 2015).

Cortisol has also been reported to be associated with changes in regional brain volumes. For instance, brain atrophy in bilateral hippocampal and cerebellar gray matter (GM) expressing high levels of glucocorticoid receptors (de Kloet et al., 2005; Hawrylycz et al., 2012) was observed in patients with uncontrolled Cushing’s disease, a state with high cortisol levels in the blood (Burkhardt et al., 2015). A recent meta-analysis showed that the damaging effect of high cortisol levels in Cushing’s syndrome on the human brain was irreversible (Andela et al., 2015). In healthy subjects, a brief hydrocortisone challenge was shown to significantly reduce hippocampal volume in a group of healthy adults (Brown et al., 2015). In addition, a recent study demonstrated a positive association between the morning cortisol levels and white matter volume (WMV) in the temporal, parietal and occipital lobes in a large sample of elderly subjects without dementia (Geerlings et al., 2015). In contrast, the morning plasma cortisol levels (at 9 am) have been reported to have no associations with age-related brain atrophy in a group of healthy elderly men (MacLullich et al., 2005), which may be due to the different sampling time.

There has been an increase in the number of studies reporting the relationship between cortisol, brain volumes and cognitive functions. The interaction between cortisol and its related regional brain volumes on cognitive functions is, however, less understood. In the present study, we aim to investigate the role of serum cortisol levels in the relationship between the regional brain volumes and cognitive status using moderation analyses in a group of non-depressed, cognitively normal elderly subjects. The cognitive processing speed was used to reflect the general state of cognitive functioning because it is an index sensitive to the general cognitive status (Salthouse, 1996). We hypothesized that serum cortisol significantly interacted with the regional brain volumes to predict processing speed outcomes. The findings from this study provided the first piece of evidence on the role of cortisol levels in neural-cognitive associations, which is important for monitoring the brain and cognitive aging.

The current study applied moderation analyses to understand the role of serum cortisol levels in the relationship between regional brain volumes and cognitive processing speed. Our study emphasized on the moderating role of serum cortisol levels in the neural-cognitive association. We also demonstrated associations between serum cortisol levels and regional brain volumes, including both the GMV and WMV, which corroborates some of the previous observations from the literature. Our findings support the importance of cortisol homeostasis in monitoring morphological and cognitive changes seen in healthy elderly people.

In a previous study on a large sample of elderly subjects without dementia, the morning saliva cortisol levels (sampled 45 min after awakening) were shown to be positively associated with the WMV in different lobes of the brain as well as processing speed and executive function, but not with the GMV (Geerlings et al., 2015). In this study, using a whole brain voxel wise approach, we found that the late morning serum cortisol levels were positively associated with the WMV of the right thalamus, the GMV of the left thalamus and right cerebellar tonsil, and were negatively associated with the GMV and WMV of the left MTG (Figure 1, Table 2). The differences in the findings between Geerlings’s and our study may be due to the use of average volume of bilateral brain regions in their study, which may render the investigation of the association between morning cortisol levels and GM less sensitive, especially if the effect is laterality-specific. Also, the difference in cortisol sampling time between the two studies could potentially affect the outcomes. The brain regions that we found to be correlated with serum cortisol levels express glucocorticoid receptors in humans (Sarrieau et al., 1988; Yu et al., 2002) and have been reported to be associated with cognitive functions. For example, higher high-gamma power in the left thalamus was associated with better processing speed among healthy elderly adults using magnetoencephalogram (Akimoto et al., 2014). Hypometabolism in the right cerebellar tonsil has been reported in illiterate elderly subjects with poorer cognitive functions compared to age- and gender-matched literate subjects (Kwon et al., 2012). The role of cerebellum in controlling incoming working memory information was also demonstrated in patients with cerebellar ischemic stroke (Baier et al., 2014). The positive associations between serum cortisol levels and these regions may imply a facilitating role of cortisol in cognitive functions via these brain regions. On the other hand, atrophy of the bilateral thalamus and left MTG were consistently found in patients with mild cognitive impairment or dementia (Yang et al., 2012a,b). The negative association between serum cortisol levels and regional brain volumes of the left MTG indicates a region-specific response to cortisol in the MTG, which is different from the cortisol response in the thalamus and cerebellum. This observation requires more future studies to confirm. Taken together, our findings suggest the close relationship between the morning serum cortisol levels and the WMV/GMV of brain regions that are important for cognitive functions in cognitively normal elderly people.

We further investigated the moderating role of serum cortisol levels in the neural-cognitive association that involved the abovementioned brain regions. From the moderation analyses, we found a significant moderating role of the serum cortisol levels in the relationship between the GMV of the left MTG and PSI. Increased cortisol levels changed the relationship between the GMV of the left MTG and PSI from negative to positive (Figure 2A), suggesting that cortisol facilitates the left MTG in performing the cognitive task. The MTG is associated with many different functions, such as language processing (Mirz et al., 1999), observation of motion (Rizzolatti et al., 1996), deductive reasoning (Goel et al., 1998) and dynamic facial expressions (Sato et al., 2012). Recently, different parts of the MTG have been reported to functionally and anatomically connect with other brain regions that are involved in memory retrieval (e.g., hippocampus) and executive controls (e.g., inferior frontal gyrus; Xu et al., 2015), which is believed to have substantial influence on processing speed (Cepeda et al., 2013). Interestingly, the moderating effect of serum cortisol levels was specific to the digit symbol sub-scales of PSI, of which the performance is significantly contributed by memory functions assessed by incidental learning and Wechsler Memory Scale III (immediate and delayed recall for visual and auditory domains; Joy et al., 2004). Our findings suggest that the effect of the interaction between cortisol and the MTG on processing speed involves a memory component. Unfortunately, memory domains were not captured in the current study. Therefore, such speculation requires further studies to confirm.

We also found a significant moderating role of the serum cortisol levels in the association between the GMV of the left thalamus and the PSI. In contrast to the finding in the MTG, positive associations between the GMV of the left thalamus and PSI were observed in subjects with low to medium serum cortisol levels (from the 1st to the 3rd quartile, Figure 2B), whereas a negative association between the GMV of the left thalamus and PSI was observed in subjects with high serum cortisol levels (the 4th quartile, Figure 2B). Furthermore, such moderating effect was specific to the other sub-scale of PSI—the symbol search task—that does not rely on memory as much as the digit symbol task. One speculation for this observation is that glucocorticoid receptors are more densely expressed in the thalamus. Within the optimal range of cortisol, the bigger volume of the thalamus (plausibly with more glucocorticoid receptors) could bind more cortisol, which subsequently influences perceptual detection thresholds. This can then enhance focused attention on the perceived stimuli and excludes irrelevant stimuli (Fehm-Wolfsdorf and Nagel, 1996; Erickson et al., 2003), resulting in better processing speed. In contrast, excess cortisol can down-regulate glucocorticoid receptors in the brain (Sapolsky and McEwen, 1985), which reduces the regional brain responsiveness to cortisol stimulations, and hence, minimizes the effect of cortisol on processing speed. Specifically, the thalamic region that we found to be associated with cortisol levels was located approximately at the lateral thalamic nuclei that were functionally related to spatial attentional weighting, selective attentional control and visual perceptual processing speed (Kraft et al., 2015). We, therefore, speculate that the cortisol-thalamic interaction effect on cognitive processing speed may be mediated through visual attentional processing. This speculation requires more empirical studies to confirm. Consistent with the literature, our findings further corroborate the non-linear relationship between cortisol levels and cognitive function (for review see Belanoff et al., 2001) by interacting with the regional brain volumes. The differential interaction effects between serum cortisol levels, the MTG and the thalamus might be due to the regional-specific reduction in glucocorticoid receptors in the brain during aging, which requires more empirical studies to confirm.

There are several limitations in the current study. First, interpretations on results are limited by the current study design. Although we adopted moderation models, the causality among those markers of interest cannot be confirmed without a longitudinal design. Second, serum cortisol was sampled only once in the late morning (~11 am). The effect of cortisol awakening response and the diurnal effect of cortisol on the regional brain volumes and cognitive function cannot be deduced in the current study. Third, the absence of information on other domains of cognitive ability or other brain imaging modalities would limit our interpretations on the results. Furthermore, cautions should be taken when interpreting the results due to the limited sample size in the current study. Last, blood sampling was done beyond the morning period (12–1 pm) in four subjects. Nonetheless, all significant findings maintained after these four subjects were removed in a sensitivity test, suggesting the effect of this small delay on our findings is minimal. Future studies that directly manipulate the level of cortisol with a greater sample size could provide a more direct impact of this stress marker on the brain and cognitive aging.

To conclude, our study provided the first piece of evidence to support the moderating effects of serum cortisol levels on the relationship between regional brain volumes and processing speed. Our findings provide insight into the role of cortisol in modulating brain morphology and cognitive functions in healthy elderly subjects.

TMCL and ACKL conceived and designed the study. TMCL supervised data acquisition. WKWL and MKL analyzed the data. WKWL interpreted the data and drafted the article. All authors revised the article critically for important intellectual content and have approved the final article.

The 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.