Diurnal Patterns for Cortisol, Cortisone and Agouti-Related Protein in Human Cerebrospinal Fluid and Blood

Abstract

Context

Cortisol in blood has a robust circadian rhythm and exerts potent effects on energy balance that are mediated in part by central mechanisms. These interactions involve orexigenic agouti-related protein (AgRP) neurons that are stimulated by glucocorticoids. However, diurnal changes in brain or cerebrospinal fluid (CSF) cortisol and cortisone, which are interconverted by 11ß-HSD1, have not been characterized in humans.

Objective

To conduct a secondary analysis of existing samples to characterize diurnal changes in cortisol and cortisone in CSF and examine their relationships to changes in AgRP.

Methods

Stored CSF and plasma samples were obtained from 8 healthy subjects who served as controls for a sleep study. CSF was collected every 2h for 36h via indwelling lumbar catheter; plasma was collected every 2h.

Results

There was a diurnal rhythm for cortisol and cortisone in CSF that closely followed the plasma rhythm by 2 h with peak and nadir levels at 0900h and 0100h. The ratio of cortisol (active) to cortisone (inactive) in CSF was 48% higher at the peak versus nadir. There was a diurnal rhythm for AgRP in plasma that was out of phase with the cortisol rhythm. There was a less distinct diurnal rhythm for AgRP in CSF that oscillated with a similar phase as cortisol.

Conclusions

There is a robust diurnal rhythm for cortisol and cortisone in CSF. Diurnal changes were noted for AgRP that are related to the cortisol changes. It remains to be determined if AgRP mediates adverse metabolic effects associated with disruption of the cortisol circadian rhythm.

Glucocorticoids, primarily cortisol in humans and corticosterone in rodents, play a key role in modulating the response to stress and have a major impact on energy balance and metabolism as well as on other physiological activities including immunity, behavior and cognition (1, 2). The hypothalamic–pituitary–adrenal (HPA) axis has a well-established diurnal rhythm that is regulated by a circadian pacemaker in the suprachiasmatic nucleus (2, 3). The cortisol rhythm can be regulated at different levels of the HPA axis and contributes to the synchronization of a number of secondary clocks located in the brain and in multiple peripheral tissues (2). Cortisol activity is regulated at the tissue level by the enzyme 11ß-hydroxysteroid dehydrogenase (11ß-HSD) that functions in the bidirectional interconversion of active cortisol to inactive cortisone (4, 5). The robust diurnal rhythm for cortisol in blood has been well characterized but diurnal changes in brain or cerebrospinal fluid (CSF) cortisol and interconversion with cortisone by 11ß-HSD1, which is widely expressed in brain and regenerates active cortisol from cortisone (6), have not been characterized in humans. This is important given the known physiological and pathophysiological effects of cortisol on the brain that are potentially impacted by these diurnal changes.

An important physiologic effect of glucocorticoids (GCs) is on the regulation of energy balance. GCs exert potent effects on appetite, body composition, and metabolism that are mediated in part by central mechanisms (7). When infused into the brain, low doses of GCs increase food intake and decrease energy expenditure in rats; this occurred at doses that were ineffective when infused peripherally (8). Glucocorticoid receptor (GR) expressing neurons in the hypothalamus are well situated to mediate the central metabolic effects of GCs. Specifically, the hypothalamic melanocortin system, which plays a critical role in regulating energy balance and metabolism in animals and humans, has been implicated as a mediator of the central effects of GCs on energy balance and metabolism. This system consists of proopiomelanocortin (POMC) and agouti-related protein (AgRP) expressing neurons which interact at downstream melanocortin receptor (MC4-R) expressing targets (9). The POMC-derived peptide, α-melanocyte-stimulating hormone (α-MSH), inhibits feeding and stimulates energy expenditure while AgRP antagonizes the effects of α-MSH causing an increase in food intake and a decrease in energy expenditure. AgRP neurons express GR and are stimulated by GCs (10, 11). Furthermore, specific deletion of GR from AgRP neurons results in a beneficial metabolic phenotype (12). A diurnal rhythm for AgRP has also been reported in the rodent hypothalamus that is abolished by adrenalectomy (13) and more recently a clock within AgRP neurons has been described that coordinates feeding and glucose metabolism (14). However, diurnal changes in AgRP have not been characterized in humans.

In this study, we used CSF measurements as a surrogate to assess diurnal changes in brain cortisol and cortisone levels, and their relationship to changes in plasma cortisol and cortisone levels. We postulated that cortisol and cortisone levels in CSF would reflect brain GC exposure and that the ratio of cortisol to cortisone in CSF could be used to asses brain 11ß-HSD1 activity. These changes were then related to diurnal changes in AgRP levels. AgRP was measured in both CSF and plasma as there is evidence that plasma AgRP reflects hypothalamic AgRP levels. We have previously shown in rats that plasma AgRP correlates with hypothalamic AgRP and that both increase during fasting and during treatment with GCs (10). Similarly, in humans we have shown that plasma AgRP increases during fasting and decreases during refeeding, consistent with a hypothalamic origin. Furthermore, plasma AgRP levels are elevated in patients with hypercortisolism and decrease when cortisol levels are normalized (15). AgRP has also been shown to stimulate the HPA axis via MC4-R mediated effects on CRH neurons (16). Thus, there is evidence for bidirectional interactions of AgRP with the HPA axis. The purpose of this study was to characterize the diurnal rhythm for cortisol and cortisone in human CSF for the first time and to examine its relationship to AgRP. Accordingly, we have measured cortisol, cortisone and AgRP in CSF and plasma collected every 2 h over a 36 h period in 8 healthy subjects.

Materials and Methods

Subjects

Stored CSF and plasma samples were obtained from 8 subjects in good general health (2 males, 6 females), mean age 42.2 years (range, 25–57 years), who were recruited from both a longitudinal study (Adult Children Study (17)) at the Knight’s Alzheimer Disease Research Center and a research volunteer registry at Washington University (Volunteers for Health), and served as controls for a sleep study conducted at Washington University School of Medicine, St Louis, Missouri (18). Individual body mass index values and fasting plasma leptin levels are depicted in Table 1. The study was conducted at Washington University School of Medicine in St Louis, Missouri, as previously described (18). The study protocol was approved by the Washington University Institutional Review Board and the General Clinical Research Center Advisory Committee. The Clinical Trials number was NCT02063217. All participants completed written informed consent and were compensated for their participation in the study.

Table 1.

Subject characteristics

Subject	Gender	Age (Years)	BMI (kg/m2)	Plasma leptin (ng/mL)
1	M	52	23.2	0.93
2	F	33	32.1	52.3
3	F	56	32.1	26.2
4	F	57	29.2	27.6
5	F	31	27.9	62.8
6	F	47	20.9	4.62
7	F	39	31.6	47.9
8	M	25	24.3	2.81

Abbreviation: BMI, body mass index.

Protocol

All participants were in good general health and had no clinical sleep disorders, neurological diseases, or contraindications to a lumbar catheter. All participants were screened to exclude sleep-disordered breathing with a home sleep apnea test (Apnealink, ResMed, San Diego, CA, US). Prior to admission to the Clinical Research Unit (CRU), participants wore an actigraph for up to 7 days (Actiwatch2, Respironics, Bend, OR, US). Polysomnography was performed as previously reported throughout each participant’s admission to the CRU (18). After an acclimation night, an intrathecal lumbar catheter was placed at the L3–L4 interspace using a Sprotte needle for chronic CSF sampling. Starting at 0700h, CSF samples (6 mL) and blood samples were obtained every 2 h throughout the study for 36 h. All participants were kept awake from the time of the lumbar catheter placement until 2100h when they were permitted to sleep. The lumbar catheter was removed on day 2 at 1900h. All participants had meals served at 0900h, 1300h, and 1800h. Snacks were served at 1100h, 1500h and 2000h. CSF and plasma samples were centrifuged and frozen immediately at −80°C. Aliquots of CSF and plasma (0.5 mL) were obtained over a 36 h period for the current study. The CSF samples had been previously thawed once and then quickly refrozen at −80°C.

Assays

Cortisol was measured in CSF by sensitive ELISA (Salimetrics, State College, PA, US). The assay detection level is 0.07 ng/mL; cross-reactivity with cortisone is 0.13%; interassay coefficient of variation is 7.9%. Cortisol was measured in plasma by chemiluminescent immunometric assay (Siemens Healthcare Diagnostics, Tarrytown, NY, US). The assay detection level is 10 ng/ml; cross-reactivity with cortisone is <0/.1%. Cortisone was measured in both CSF and plasma by a highly sensitive ELISA (Arbor Assays, Ann Arbor, MI, US). The assay detection level is 0.06 ng/mL; cross-reactivity with cortisol is <0.1%; interassay coefficient of variation is 6.1%.

AgRP was measured in CSF and in plasma by 2-site ELISA (R&D Systems, Minneapolis, MN, US) that uses full-length human AgRP as standard as previously described (19). POMC was assayed using an in house 2-site ELISA with antibodies provided by Dr. Anne White, with the capture monoclonal antibody directed against ACTH_10-18 (20, 21) and the detection antibody directed against γ-melanocyte-stimulating hormone (22, 23). There is 100% cross-reactivity with 22K pro-ACTH. There is no cross-reactivity with ACTH, α-MSH, γ-melanocyte-stimulating hormone, or β-endorphin (24). Affinity purified human 31K POMC was used as standards. Plasma leptin were measured by ELISA (R&D Systems) (25).

Statistics

Cosinor analysis was performed as previously described and was used to analyze the 36 h patterns of cortisol, cortisone, and AgRP in CSF and plasma and of POMC in CSF (25, 26). A cosine transformation was applied to the time variable using 24 h as the default circadian cycle and the GraphPad Prism version 6.0b for Mac (Graphpad Software) was used to estimate the parameters of the circadian patterns for leptin and POMC fluctuations. The mesor (midline of the oscillation), amplitude (distance between the peak and mesor), and acrophase (the time corresponding to the peak of the curve) were calculated for each subject and averaged for the group. Mean peak and nadir cortisol, cortisone, and cortisol-to-cortisone ratios were compared by paired t test. Statistical significance was set at P < .05.

Results

Concentrations of cortisol and cortisone in plasma and CSF over a 36 h period

Plasma cortisol and cortisone concentrations oscillated in a similar manner with peak and nadir concentrations occurring at the same times (Fig. 1A). There was a diurnal rhythm for plasma cortisol with peak of 136 ± 20 ng/ml (13.6 ± 0.20 µg/dl) at 0700h and a nadir at 2100h to 2300h. A similar rhythm for plasma cortisone was also observed with a peak of 26.1 ± 3.8 ng/mL (Fig. 1A). Mean CSF cortisol and cortisone concentrations are shown in Fig. 1B. CSF cortisol had a peak of 6.76 ± 0.85 ng/mL at 0900h and nadir of 1.71 ± 0.34 ng/mL at 0100 h. Similarly, CSF cortisone had a peak of 2.69 ± 0.41 ng/mL and nadir of 1.16 ± 0.35 ng/mL. The CSF cortisol and cortisone rhythms closely paralleled the plasma rhythms with peak and nadir levels occurring 2 h after the respective plasma peak and nadir levels (Figs. 1C and 1D).

Figure 1.

Concentrations of cortisol and cortisone in plasma and CSF in 8 subjects over a 36 h period. A: Mean plasma cortisol and cortisone (±standard error of the mean) levels. B: Mean CSF cortisol and cortisone levels C: Relationships between mean plasma and CSF cortisol levels. D: Relationship between mean plasma and CSF cortisone levels, E. Mean levels of cortisol (F), cortisone (E), and their ratios (F:E) in plasma at the peak and nadir points. F: Mean levels of F and E and the F:E ratios in CSF at the peak and nadir points **P = .002.

Cosinor analysis was used to access the circadian pattern of changes in cortisol, cortisone in plasma, and CSF. The cortisol data from plasma and CSF fit a cosine function with r = 0.67 and r = 0.75 respectively (Figs. 2A and 2B). A similar cosine fit was observed for the cortisone data from plasma and CSF with r = 0.62 and r = 0.62 respectively (Figs. 2C and 2D).

Figure 2.

Cosinor fit of plasma and CSF cortisol (A and B), cortisone (C and D) and AgRP (E and F) and CSF POMC (G) levels over time. r values indicate how well the data fit a cosine function.

Diurnal changes in the ratio of cortisol to cortisone in CSF

The mean ratio of cortisol to cortisone was calculated in plasma and CSF at the peak and nadir points during the diurnal cycle. The mean cortisol to cortisone ratio in CSF at 0900h was 2.74, significantly higher than the ratio of 1.85 measured at 0100h (P = .002), consistent with increased 11ß-HSD1 activity in the morning compared to the night. The ratio of cortisol to cortisone in plasma was not significantly different at the peak and nadir points (Figs. 1E and 1F).

Concentrations of AgRP in plasma and of AgRP and POMC in CSF over a 36 h period and relationship to changes in CSF cortisol

Mean plasma AgRP concentrations, as related to changes in CSF cortisol, are shown in Fig. 3A. Mean plasma AgRP and CSF cortisol levels presented as percent of mean levels are depicted in Fig. 3C. There was a diurnal rhythm for plasma AgRP with levels peaking (112% of mean) in the evening between 1700 and 2300h and a nadir (86% of mean) at 0900h. The plasma AgRP data fit a cosine function r = 0.50 (Fig. 2E). Mean CSF AgRP concentrations, as related to changes in CSF cortisol, are shown in Fig. 3B. Mean CSF AgRP and CSF cortisol levels presented as percent of mean levels are depicted in Fig. 3D. There was a less distinct diurnal rhythm for AgRP in CSF that was out of phase with the plasma rhythm. The CSF AgRP data fit a cosine function r = 0.36 (Fig. 2F). Thus, plasma AgRP peaked at night when CSF cortisol was low and nadired in the morning when CSF cortisol was high. In contrast CSF AgRP oscillated with a similar phase as CSF cortisol; this was most evident during the second day when CSF AgRP and CSF cortisol peaked at the same time at 0900h. An example of the relationship between cortisol in CSF and AgRP in both plasma and CSF in a single subject is shown in Figs. 3E and 3F. CSF POMC was also measured over the 36 h period and as shown previously did fit a cosine function (Fig. 2G).

Figure 3.

A to D: Changes in AgRP in plasma and CSF as related to changes in CSF cortisol in 8 subjects over a 36h period. A: Mean plasma AgRP (±standard error of the mean) and CSF cortisol levels. B: Mean CSF AgRP and CSF cortisol levels. C: Mean plasma AgRP and CSF cortisol levels presented as percent of mean levels. D: Mean CSF AgRP and CSF cortisol levels presented as percent of mean levels. E and F: Changes in CSF cortisol and AgRP in plasma and CSF in an individual subject over a 36h period presented as percent of mean levels. E: Changes in plasma AgRP and CSF cortisol levels. 2F: Changes in CSF AgRP and CSF cortisol levels.

Discussion

In this study, we demonstrate a clear diurnal rhythm for cortisol and cortisone in human CSF. Although the CSF rhythms for cortisol and cortisone closely parallel their plasma rhythms, significant changes in the ratio of cortisol to cortisone were noted in CSF that may be indicative of diurnal changes in brain 11ß-HSD activity. However, other explanations such as relative differences in transport into CSF also need to be considered. In addition, we provide evidence for a diurnal rhythm for AgRP and demonstrate a relationship between CSF cortisol and AgRP that may have important metabolic implication. CSF cortisol and cortisone were measured by sensitive and specific ELISAs. The values obtained were very close to those reported previously in CSF using LC/MS methodology (27, 28). Cortisol was 20- to 30-fold higher in plasma than in CSF; however, we only measured total cortisol in plasma and more 90% of plasma cortisol is bound to plasma proteins (29). In contrast, 90% of cortisol in CSF is free and not bound to proteins (30). In a prior study that compared free plasma cortisol levels to CSF cortisol levels, only a 3-fold difference in concentration was noted (30). The CSF rhythms for cortisol and cortisone followed the plasma rhythms closely with peak levels occurring in CSF 2 h after the plasma peak. However, since CSF and plasma levels were only measured at 2 h intervals we cannot be precise about the exact timing of achieving peak levels in CSF relative to plasma. The CSF cortisol rhythm is consistent with previous reports in monkeys using older assay techniques (31, 32).

Although it was expected that diurnal changes in CSF cortisol would parallel the well-established peripheral cortisol rhythm, the extent to which this rhythm might be amplified or diminished by central cortisol metabolism was unknown. Tissue specific GC metabolism by11-ß-HSD is an important modulator of GC activity both peripherally and centrally has been implicated in obesity and metabolic dysfunction (4). In the adult brain, cortisol activity is primarily mediated by 11ß-HSD1 that functions as a ketoreductase to reactivate cortisol from cortisone (6). We therefore measured CSF cortisone and calculated the ratio of cortisol to cortisone in CSF in order to assess potential diurnal changes in brain 11ß-HSD1activity. We found that the mean ratio of cortisol to cortisone is almost 50% higher at 0900h versus 0100h. Thus, at the morning peak, there is relatively more active cortisol compared to the nadir at 0100h. No significant difference was noted for the respective plasma ratios. It remains to be determined if this ratio changes under other physiological or pathophysiological circumstances that impact brain cortisol metabolism.

While GCs have well characterized direct effects on peripheral metabolic tissues, including liver, fat, and skeletal muscle, effects on appetite and some of the effects on body composition and metabolism are mediated by central mechanisms (7, 8). There is evidence to support a role for GR expressing hypothalamic AgRP neurons as mediators of the effects of GCs on appetite, body weight gain and composition, energy expenditure, and glucose and fat metabolism (12). AgRP gene expression in the rodent hypothalamus is stimulated by corticosterone (10, 11). Furthermore, a diurnal rhythm for AgRP has been reported in the rodent hypothalamus that is abolished by adrenalectomy (13). We hypothesized that similar diurnal changes in AgRP would occur in humans and that there would be an association between the diurnal cortisol and AgRP rhythms. Our results demonstrate a diurnal rhythm for AgRP in plasma that was out of phase with the CSF cortisol rhythm such that plasma AgRP peaked when cortisol was lowest. In contrast, CSF AgRP and cortisol oscillated in the same phase consistent with a positive relationship. The explanation for this discrepancy is unclear. It could be argued that the source of circulating AgRP is not the hypothalamus. However, a number of findings suggest that circulating AgRP is of hypothalamic origin. Plasma AgRP has been shown to correlate with hypothalamic AgRP expression in rats (10). AgRP increases in both the hypothalamus and in plasma during fasting in rats (33, 34). Similarly, in humans, plasma AgRP increases during fasting (19, 34) and after dieting (19) consistent with the expected changes in hypothalamic AgRP during these conditions. Furthermore, plasma AgRP levels are elevated in patients with hypercortisolism caused by Cushing disease and the degree of elevation correlates with elevations of 24 h urine free cortisol levels (15). Plasma AgRP levels then decreased in these patients as early as one week after surgical cure of their disease (15). Stimulation of cortisol by the opioid antagonist, naltrexone, has also been shown to correlate with increases in plasma AgRP (35). Although AgRP is moderately expressed by the adrenals, plasma AgRP still increased when adrenalectomized rats were treated with corticosterone (15). Thus, assuming that plasma AgRP is of hypothalamic origin, potential explanations for the differences noted in AgRP levels in CSF and plasma could relate to different pools of AgRP in the hypothalamus with anatomical differences in fiber tracks that gain access to CSF and blood and/or kinetics of release of AgRP into CSF and blood (36, 37). Distinct populations of AgRP neurons have been identified within the arcuate and one of these populations is outside the blood–brain barrier (38). It is possible that the population that resides outside the blood brain barrier is the source of AgRP in blood. The difference between the cortisol and plasma AgRP rhythms does not preclude a positive relationship between cortisol and AgRP and may relate to a phase shift with cortisol peaking first followed by a delayed increased in AgRP in plasma.

Disruption of the normal cortisol rhythm has metabolic consequences (2). The metabolic effects of short-term cortisol excess are more pronounced in the evening compared to the morning (39). There is also evidence that patients with adrenal insufficiency requiring GC replacement have better metabolic outcomes when the replacement mimics the normal circadian rhythm (40, 41). Our findings provide evidence for a link between the diurnal changes in cortisol and AgRP and are consistent with rodent studies showing that GCs are necessary for the normal diurnal changes in hypothalamic AgRP (13). It has recently been shown that a clock network within the hypothalamus coordinates appetite and metabolism with the sleep–wake cycle and that AgRP neurons play an important role in this process that includes coordination of the transcriptional response to leptin (14). These finding suggest that disruption of the normal cortisol rhythm could impact AgRP levels leading to adverse metabolic consequences. It has been shown that specific deletion of GR from AgRP neurons ameliorates the adverse metabolic effects of a high-fat diet (12). Thus, the role of AgRP in mediating adverse metabolic effects associated with conditions that have a disrupted cortisol rhythm such as with chronic sleep deprivation or stress deserves further study (42–44).

This study has several limitations related to the small sample size and the fact that this was a secondary analysis of previously collected sample. Although it is unlikely that the robust CSF cortisol and cortisone rhythms would change significantly with a larger sample size, it is possible that a better defined AgRP rhythm would emerge with a larger sample size. It would also be of interest to examine changes in the AgRP rhythm related to body mass index and leptin levels. However, despite these limitations, our study provides unique new data about brain cortisol dynamics in human subjects and demonstrates a relationship with the neuropeptide AgRP. Further study is necessary to determine if changes in the ratio of cortisol to cortisone in CSF reflect physiological meaningful changes in brain 11ß-HSD1 activity and to determine if AgRP mediates some of the adverse metabolic effects associated with disruption of the normal cortisol circadian rhythm.

Additional Information

Disclosure Summary : The authors have nothing to disclose.

Data Availability

The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.