Detrimental effects of hypercortisolism on brain structure and related risk factors

Brain structural abnormalities are often observed on magnetic resonance imaging (MRI) scans of Cushing's syndrome patients, but the pathogenesis is not fully understood. To understand the relationship between brain structural abnormalities and potential risk factors in active Cushing's disease (CD) patients, a total of 101 treatment-naïve CD patients and 95 sex-, age- and education matched controls with non-functioning adenomas (NFA) underwent clinical evaluation and MRI investigation, and the relative risk factors were analyzed. 14 patients in sustained remission after transsphenoidal surgery were followed. Compared with the NFA subjects, the patients with CD had more cortical (P < 0.01) and subcortical atrophy (P < 0.01) and a higher prevalence of white matter hyperintensity (WMH) (P < 0.01). WMH severity in CD patients positively correlated with age (r = 0.532, P = 0.000), disease course (r = 0.257, P = 0.009), postprandial glucose (r = 0.278, P = 0.005), frequency of left ventricular hypertrophy (r = 0.398, P = 0.001) and hypothyroidism (r = 0.246, P = 0.014). The markers of cortical and subcortical atrophy (sylvian fissure ratio, bifrontal ratio, bicaudate ratio and third ventricle width) were positively associated with the progression of WMH in the CD patients. In the follow-up of 14 patients with CD, brain atrophy and WMH was partially reversible after correction of hypercortisolism. In conclusions, brain atrophy and WMH were more likely to appear in CD patients and were possibly partially reversible following correction of hypercortisolism.

Brain atrophy assessment in CD patients and NFA subjects. The results of the cerebral atrophy rating scale assessment before and after adjusting for SBP and BMI are shown in Table 2. Compared with NFA after adjustment, the subjects in the CD group showed decreased hippocampal height (P < 0.01) and increased temporal horn width (P < 0.01), suggesting temporal lobe atrophy in the CD patients (Fig. 1A, B). Increasing sylvian fissure ratio (SFR) (P < 0.01) and frontal interhemispheric fissure ratio (FFR) (P < 0.01) indicated cortical atrophy, while increasing bicaudate ratio (BCR) (P = 0.012) indicated subcortical atrophy in the CD patients, accompanied by widened third ventricle (P < 0.01) (Fig. 1C). Furthermore, the patients with CD tended to have more lacunar infarcts than NFA subjects (8/101 vs. 2/95, P = 0.055).
White matter lesions in the CD patients and NFA subjects by Scheltens scale. White matter lesions were common in the CD patients. We used Scheltens rating to evaluate the WMH ( Fig. 2A,B). The prevalence of WMH was 68% in the CD patients and 27% in the NFA subjects. The deep white matter hyperintensity (DWMH) scores and periventricular hyperintensity (PVH) scores were obviously higher in the CD group than in the NFA group. In the periventricular region, the CD group had a higher likelihood of receiving www.nature.com/scientificreports/ a score of 2 (4) than a score of 0 (1) in the NFA group (P < 0.001). In the deep white matter region, the frontal lobe (P < 0.001), parietal lobe (P < 0.001), occipital lobe (P < 0.001), temporal lobe (P < 0.001) and basal ganglia (P < 0.001) showed higher Scheltens scores in the CD subjects than in the NFA group, while there was no difference in the infratentorial area between the two groups (P = 0.051). Figure 2C shows a representative FLAIR image of the PVH and DWMH in a CD patient and a NFA subject.

Analysis of risk factors for WMH severity in CD patients by Fazekas scale.
We used the Fazekas classification method to divide the CD patients into 3 groups according to the WMH grade by univariate analyses (Table 3) and correlation analysis (with significant difference, Table 4). The CD patients with moderate to severe WMH were older and had a longer disease course, higher blood glucose and TG levels, higher frequency of diabetes, LVH and arrhythmia than the patients in the other categories. WMH severity in CD patients positively correlated with age (r = 0.532, P = 0.000), disease course (r = 0.257, P = 0.009), fasting blood glucose(r = 0.212, P = 0.033), postprandial glucose (r = 0.278, P = 0.005), frequency of diabetes (r = 0.245, P = 0.013), LVH (r = 0.398, P = 0.001) and hypothyroidism (r = 0.246, P = 0.014). The markers of cortical and subcortical atrophy: SFR (r = 0.197, P = 0.049), BFR (r = 0.200, P = 0.044), BCR (r = 0.278, P = 0.005) and third ventricle width (r = 0.242, P = 0.015) were positively associated with the progression of WMH in the CD patients.
During the 25.4 ± 10.7 months after surgery, all 14 CD patients showed biochemical evidence of remission with normal morning cortisol suppression (23.35 ± 9.40 nmol/L). 8 CD patients experienced improvement in WMH scoring, while the rest 6 patients had no significant changes. The remitted CD patients experienced increased hippocampal height (P = 0.047), BCR (P = 0.024) and third ventricle width (P = 0.002), suggesting an improvement in brain atrophy. In addition, the PVH (P = 0.017) and DWMH scores, especially in the parietal lobe (P = 0.027) and temporal lobe (P = 0.042), substantially decreased compared with baseline (Table 5, Fig. 3).

Discussion
Hypercortisolemia causes brain atrophy changes. This study demonstrated that brain structural abnormalities are common in patients with active CD. We detected temporal lobe atrophy, cortical atrophy, subcortical atrophy and a widened third ventricle in CD patients when compared with age-and sex-matched controls. These detrimental effects of chronic glucocorticoid excess have been described in previous clinical and experimental studies 3,15 . The data obtained from 14 CD patients in remission with a mean follow-up of 25.4 ± 10.7 months showed that hippocampal height, BCR and third ventricle width increased after the correc- www.nature.com/scientificreports/ tion of hypercortisolemia, suggesting an improvement in brain atrophy. Our results supported the suggestion that chronic glucocorticoid excess causes brain atrophy changes. Some clinical studies have confirmed that the structural abnormalities revealed by MRI in patients with active Cushing's syndrome might be related to psychological morbidity and cognitive impairment 16 , and following the successful treatment of hypercortisolism, both the physical features and the psychiatric symptoms tended to substantially improve 17,18 . The neurotoxic effects of corticosteroid excess on the CNS have been well recognized in experimental animal studies 19,20 . To develop an effective medical treatment for these harmful effects of hypercortisolism on brain structure and psychological/cognitive impairment, it is important to gain more insight into these pathological processes. www.nature.com/scientificreports/ White matter hyperintensities indicate that hypercortisolemia affects the entire brain. Widespread white matter hyperintensities throughout the brain in the CD patients was another finding in this study. Although the average age was only 37.4 years old in our CD patients, the prevalence of WMH was 67%, much higher than that in NFA (27%) and healthy community-based populations 21 . WMH is a characteristic of white matter injury. We found that the DWMH scores and PVH scores were obviously higher in the CD group. Predictably, we observed that the markers of cortical atrophy and subcortical atrophy were closely associated with the progression of WMH in CD patients. These white matter lesions improved in 14 patients, especially in the parietal lobe and temporal lobe, following the successful treatment of hypercortisolism. These findings are in line with the recent studies which examined white matter structural changes in patients with Cushing's syndrome, suggesting hypercortisolism affects the entire brain with indications of demyelination of the white matter tracts 12,13 . However, in a cross-sectional study, Werff reported that CD patients with long-term remission showed widespread reduction of white matter integrity in the brain, suggesting persistent structural effects of hypercortisolism 10 . In a latest research, the alterations of white matter in 35 CD patients seem to be persist after remission 22 , which is in contrast with our results. The difference may due to the limitation of our small sample size. More research with sufficient cases is needed to clarify how hypercortisolism affect white matter tissue injury.
Hypothyroidism may be involved in brain structural changes in CD patients. The functions of the pituitary-thyroid axis are suppressed in patients with Cushing's syndrome because of a direct effect of cortisol on TSH secretion 23 . Hypothyroidism is a recognized cardiovascular risk factor, and reduced cardiac output and tissue perfusion as well as decreased tissue oxygen utilization are common in patients with hypothyroidism. www.nature.com/scientificreports/ Additionally, cognitive dysfunction is a common feature of hypothyroidism, and there have been reports that hypothyroidism may be a risk factor for Alzheimer's disease 24,25 . In this study, we found that the incidence of central hypothyroidism in the patients with WMH was at least double that of the patients without WMH, and the FT 4 levels were much lower in the CD patients with moderate or severe WMH than in those without WMH. It is logical to speculate that hypothyroidism may be involved in brain structural changes in CD patients.
Cerebral small vessel disease might be one of the potential pathophysiological links between hypercortisolism and brain structural abnormalities. It is known that WMH, indicating white matter tract demyelination, is an imaging manifestation of denervation of nerve conduction fibers caused by diffuse cerebral ischemia and belongs to a type of cerebral small vessel disease (CSVD) 26 . Many studies have confirmed that CSVD is a key pathological link in the early stages of many brain diseases 27 . Given that WMH and brain atrophy are characteristic imaging changes in CSVD, we hypothesize that brain structural abnormalities on MRI images of patients with Cushing's syndrome may be related to CSVD. We analyzed the relationship between a variety of traditional cardiovascular risk factors and WMH in CD patients, and the results showed that age and fasting and postprandial blood glucose and TG levels tended to rise in accordance with WMH grade. These Table 3. Characteristics in subgroups based on WMH severity in CD patients. www.nature.com/scientificreports/ well-established factors were as applicable to CD patients as they were to the general population. In particular, we found a higher prevalence of LVH and arrhythmia with WMH progression, raising the hypothesis that excess cortisol-induced cerebral ischemia changes may share the same well-established mechanism of cerebrovascular endothelial dysfunction and inflammatory response 28 . Vascular conditions can, at least partially, contribute to brain structural abnormalities in CD patients. Not only absolute blood pressure levels, but even their fluctuations over time can have detrimental effects on brain parenchyma 29 and are associated with cognitive dysfunction 30 . Additionally, cardio-vascular risk factors in CD patients, including but not limited to hypertension and diabetes mellitus, can significantly impair cerebral hemodynamics, which is a contributor to neurocognitive dysfunction and is also amenable to reverse 31 . The relation between WMH and diabetes is still debated, a recent study found that WMH volume and number of WMH lesions were significantly associated with diabetes 32 . It is still not clear whether WMH is related to hypercortisolism regardless of diabetes in our research due to lack of evidenced analysis.
Mineralocorticoid receptor (MR) overactivation may also play a role in brain structural abnormalities in Cushing's syndrome. Glucocorticoids, for instance cortisol, usually bind with glucocorticoid receptors (GRs) and mineralocorticoid receptors (MRs). 11β-HSD2 enzyme converts cortisol into inactive cortisone (which can't bind with MR) to protect from the MRs effect. However, elevated cortisol levels overcome the 'protective' role of 11β-HSD2, leading to MR excess and successive multiple effects 33 . The incidences of hypertension and hypokalemia in our CD patients were 77.2% and 39.6%, respectively, suggesting MR hyperactivation  Table 5. Neuroradiological structure and WMH scores (Scheltens scale) of CD prior to treatment (baseline), after surgical remission and sex-and age-matched NFA subjects. www.nature.com/scientificreports/ at the renal level. Moreover, the incidences of arrythmia and LVH in our CD patients were 13.8% and 54.4%, respectively, suggesting MR hyperactivation at the cardiovascular level 34 . It is worth noting that 11β-HSD2 is not expressed in the hippocampus or other limbic structures 35 ; therefore, hypercortisolism increases the occupation of MRs/GRs. Studies on hippocampal cell cultures showed that supraphysiological doses of glucocorticoids lead to a reversible phase of atrophy of the apical dendrites of pyramidal neurons 36 . Accumulating evidence has indicated an important role for the activation of MRs in the pathophysiology of vascular damage in the heart and kidney 37 . Given that MRs are highly expressed in many brain tissues 38 and cerebral vessels 39 , we speculated that MR overactivation may also play a role in the brain structural abnormalities in Cushing's syndrome. In a latest animal research, we observed that MR and GR overactivation can cause brain atrophy and vascular apoptosis in a cortisol-excess animal model 40 . Further exploration of the cerebral molecular mechanisms underlying the detrimental effects of Cushing's disease could help prevent nervous system complications and cardiovascular disorders.

P value CD baseline versus in remission
Our study has some limitations. First, our sample size was too small to allow sufficiently powered statistical such as ordinal logistic regression to be performed. The number of patients in the comparative analysis of clinical characteristics between subgroups based on WMH was unbalanced and small. Further examination with larger numbers of CD patients is needed to confirm our findings. Moreover, due to the retrospective design, no measures were used to assess psychological morbidity or cognitive impairment and we could not use more advanced MRI techniques to measure brain volume, such as voxel-based morphometry in which 3D sequence is needed.

Conclusion
Brain atrophy and white matter hyperintensities are characteristic manifestations of brain MRI images in Cushing's syndrome. The pathogenesis of the brain structural abnormalities induced by chronic glucocorticoid excess appears to be multifactorial and is not yet fully understood. In addition to glucocorticoid excess itself, cerebral small vessel disease might be one potential pathophysiological link between long-term hypercortisolemia and brain structural abnormalities. High blood pressure, elevated blood glucose, associated hypothyroidism and MR overactivity may all be involved in this pathological process. To develop an effective medical treatment for www.nature.com/scientificreports/ the detrimental effects of glucocorticoid excess, it is imperative to understand the underlying pathways through which hypercortisolemia leads to brain structural changes.
Patients and methods. Participants. One hundred and one consecutive patients who had undergone evaluation and treatment for Cushing's disease (CD) at West China Hospital between 2013 and 2019 were involved in this study. We retrospectively reviewed our data. The study was approved by the Ethics Committee of West China Hospital and conformed to the tenets of the Declaration of Helsinki. Written informed consent was obtained from all subjects in our study. All 101 CD patients were diagnosed based on agreed guidelines 41 , with clinical manifestations and positive biochemical tests. The mean age of the patients at the time of diagnosis was 37.4 ± 12.6 years. Eighty-four patients were female (83%). Ninety-five sex-, age-and education matched treatment-naïve subjects diagnosed with nonsecreting pituitary microadenoma (NFA) were recruited in West China hospital as controls.
Of these 101 CD patients, 14 subjects (all females) who were in sustained remission after transsphenoidal surgery were followed for one to four years (the median follow-up period was 25.4 ± 10.7 months). Biochemical remission after surgery was defined as suppression of plasma cortisol to less than 50 nmol/l after 1 mg overnight dexamethasone and a normal urinary free cortisol (UFC) excretion 8 . 14 sex-, age-and education matched treatment-naïve subjects diagnosed with nonsecreting pituitary microadenoma (NFA) were recruited in West China hospital as controls. No subjects had a history of stroke, long-term physical illness, or other neuropsychiatric diseases. Only right-handed subjects were included.
Clinical interview and biochemistry. All subjects underwent a complete clinical interview, physical examinations and blood biochemical tests, including electrolytes. Clinical variables related to cardiovascular risk, including age, smoking history, drinking history, systolic blood pressure (SBP), diastolic blood pressure (DBP), hemoglobin A1c (HbA1C), fasting and 2-h postprandial plasma glucose, and lipid profiles were collected. All CD patients received 24-h dynamic electrocardiogram and ultrasonic cardiogram examinations. Hypertension was defined as a blood pressure greater than or equal to 140/90 mmHg 32 . Prediabetes was defined by a fasting glucose level between 5.6 mmol/L and 6.9 mmol/L or a 2-h postprandial load glucose level between 7.8-11.0 mmol/L. Diabetes was determined as a fasting glucose level ≥ 7.0 mmol/L or 2-h postprandial load glucose levels ≥ 11.1 mmol/L, according to the World Health Organization (WHO) diagnostic criteria for diabetes 42 .
In addition to plasma/urine cortisol and ACTH level measurements, dynamic endocrine tests (including dexamethasone suppression tests, desmopressin (DDAVP) stimulation tests) and in some cases inferior petrosal sinus sampling required for establishing CD diagnosis, the following pituitary-target gland hormones were also tested in all participants: thyroid-stimulating hormone (TSH), free thyroxine (FT 4 ), luteinizing hormone (LH), follicle-stimulating hormone (FSH), estrogen (E 2 ), testosterone (T), growth hormone (GH) and prolactin (PRL). Fourteen patients who were in remission and had follow-up after surgery were reevaluated at 6-month to 12-month intervals, and the endocrine evaluation consisted of measurement of serum cortisol after 1 mg overnight dexamethasone and 24-h UFC excretion along with assessment of anterior pituitary hormone function.
MRI scans. All subjects were examined with a 3-T Signa MRI scanner (GE, DISCOVERY MR750). A line connecting the anterior commissure and posterior commissure (AC-PC) was drawn on the midsagittal slice and used for orientation of the remaining series. T1-weighted, T2-weighted, and FLAIR axial sections (repetition/echo/inversion time (TR/TE/TI): 9,000/95/2,475 ms, matrix 256 × 256) were obtained through the entire brain, parallel to the AC-PC line, with 5-mm contiguous slices. Custom graphics software was locally developed by using Win10 Windows. The slices containing regions of interest were first identified on digital images. Corresponding cuts were then displayed on the graphics workstation and interactively outlined with a mousecontrolled cursor. The MRI images were annually reexamined and reviewed in 14 CD patients with persistent remission after surgery every 6 months on the same scanner.
Cerebral atrophy rating scale. To assess cerebral atrophy, we used linear measurements 43 . We adopted the bicaudate ratio (BCR) and bifrontal ratio (BFR) 44 as measures of internal cerebral atrophy, the sylvian fissure ratio (SFR) and frontal interhemispheric fissure ratio (FFR) 44 as measures of external cerebral atrophy, the uncotemporal index 45 , and hippocampal formation height and width of the temporal horn as a measure of temporal atrophy 46 . Detailed descriptions of the linear measurements above are listed in Table 6. Furthermore, a lesion was considered a lacunar infarct if its score was hypointense on T1 and FLAIR images and if its appearance was unlike a perivascular space.
White matter hyperintensities by Scheltens rating. WMH were considered present if these were hyperintense on FLAIR images and not hypointense on T1WI 47 . To measure WMH in the cross-sectional (part 1) and longitudinal design (part 3) in different subregion, we used the modified Scheltens rating scale WMH from two different areas, including periventricular hyperintensity (PVH) and deep white matter hyperintensity (DWMH), were rated in a semiquantitative way 48 . The total PVH score was added with a range of 0-6, and DWMH was calculated with a range of 0-36 (Table 7). All MRI scans were independently rated by two raters blinded to the disease status of the participants. In the case of disagreements, consensus readings were held. The interrater reliability coefficient was 0.96-0.98 for the linear measures, and a paired t-test revealed no significant differences between the means obtained by the two raters.  Table 7. White matter hyperintensities by Scheltens rating.