Sex hormones as correlates of oxidative stress in the adult brain

Abstract

Oxidative stress, an imbalance between the production of reactive oxygen species and available antioxidant capacity, is implicated in multiple psychiatric disorders and neurodegenerative conditions. Peripheral and preclinical studies suggest oxidative stress differs by biological sex and covaries with estrogens. However, limited knowledge exists on the effect of circulating sex hormones on oxidative stress in the brain in humans in vivo. We aimed to examine the relationship of circulating estrogen with regional concentrations of brain glutathione (GSH) as a marker of oxidative stress. GSH was measured using magnetic resonance spectroscopy (MRS) at 7 Tesla in the dorsal anterior cingulate cortex (ACC), ventromedial prefrontal cortex (VMPFC), and left dorsolateral prefrontal cortex (DLPFC) in 34 individuals (18 females and 16 males). We observed an inverse correlation of estradiol with DLPFC GSH, as well as a trend inverse correlation of estrone with DLPFC GSH, in the combined sample of males and females and in females only. No significant sex differences were observed for GSH levels in the brain. Our study provides evidence of diminished DLPFC GSH in females with higher estradiol, suggesting circulating sex hormones may be important factors to consider in future studies examining brain GSH levels related to psychiatric and other disorders.

1. Introduction

Oxidative stress is a biological mechanism that has been implicated in multiple psychiatric and neurodegenerative conditions. Oxidative stress is an imbalance between the production of reactive oxygen species and available antioxidant capacity, which can lead to RNA and DNA damage, and downstream damage to proteins and cells including from apoptosis, cellular senescence and telomere shortening. Evidence from peripheral and postmortem studies in psychiatric disorders have shown increased markers of DNA damage, RNA damage, lipid peroxidation, protein peroxidation, and reactive oxygen species, and decreased antioxidant levels, across multiple disorders including psychotic disorders, mood disorders, and dementia.^1–3

Given the accumulation of evidence from preclinical, postmortem, and peripheral studies, there is great interest in examining evidence of oxidative stress as a mechanism in psychiatric disorders in the brain in vivo. Glutathione (GSH) is an endogenous antioxidant and the most prevalent antioxidant in the brain. Magnetic resonance spectroscopy (MRS) methods allow for regional examination of GSH levels in the brain in vivo and multiple MRS studies have now identified lower GSH levels in the brain in depression, schizophrenia, anxiety, and bipolar disorders.^1,4–6 However, results are not consistent as others have found either higher levels or no difference in GSH levels in the brain in these populations.^1,4,7

We propose that influences of sex hormones may at least in part contribute to inconsistencies in the literature. Pre-clinical and peripheral studies suggest that oxidative stress differs by biological sex and covaries with circulating estrogens.^8,9 Estrogen is an endogenous antioxidant and can influence oxidative stress via direct and indirect pathways. For example, premenopausal women have been shown to have decreased systemic oxidative stress compared to women in postmenopause¹⁰ and a study in women undergoing medically-induced menopause has shown that estrogen replacement therapy restores blood levels of GSH, thus protecting against oxidative stress.¹¹ In humans, brain GSH levels have been shown to be higher in young healthy females than in young healthy males.¹² However, there exists limited knowledge of the relationship between sex hormones and GSH in the brain in humans in vivo, which may be important in the context of many psychiatric and neurodegenerative disorders that also have sex differences in their prevalence or etiology. Therefore, we aimed to examine the relationship of estrogens with regional concentrations of GSH in the brain measured using MRS. We hypothesized that females would have higher levels of GSH in the brain than males, and that estrogens would positively correlate with GSH.

2. Methods

2.1. Participants

Thirty-four participants (18 females and 16 males) were recruited and included individuals with and without a mood disorder, ages 35–63 (M = 48, SD = 9) years old. There was no difference in the number of females and males with a history of mood disorder (p = 0.218). Inclusion criteria for individuals with mood disorder consisted of mood disorder research diagnosis from the MINI for DSM-IV,¹³ eligibility for a brain MRI scan, and native English speaker. We excluded mood disorder participants with a current mood episode based on the Montgomery-Asberg Depression Rating Scale (MADRS) and Young Mania Rating Scale (YMRS),^14,15 history of intellectual disability, history of medical conditions that would interfere with the study protocol, history of head injury resulting in loss of consciousness, cognitive impairment as assessed by the Montreal Cognitive Assessment (MoCA),¹⁶ or history of epilepsy. Other exclusion criteria included alcohol or substance abuse or dependence during the past 12-months, currently pregnant, currently in crisis, current or recent N-acetylcysteine treatment of 500mg or more daily, or positive toxicology screen. Inclusion and exclusion criteria for healthy control participants paralleled that of the participants with mood disorders with the exception that healthy controls were excluded if they met criteria for a research diagnosis of bipolar disorder, schizophrenia, recurrent major depression, or other psychiatric disorder.

2.2. Procedures

This study was approved by the Mass General Brigham Human Research Committee. Eligible participants provided written informed consent prior to study participation. Following an initial prescreening for eligibility for study participation, a psychiatrist or MA-level clinical rater administered the MINI, YMRS and MADRS to assess psychiatric research diagnoses and mood symptoms. Eligible participants were invited for an onsite study visit at Brigham and Women’s Hospital. During the onsite study visit participants completed a toxicology screen for recent use of opiates, methamphetamines, or cocaine. Participants able to become pregnant were administered a urine pregnancy test. If the pregnancy test or toxicology screen was positive then the participant was excluded from further participation in the study. A trained clinical phlebotomist conducted a fasted morning blood draw. Participants completed self-rated instruments including questionnaires to assess ingestion in the past 24 hours. Study staff collected vital signs, including blood pressure, height, and weight. An MRI technician screened participants for MR safety. Participants then completed a fasted 1-hour 7.0 Tesla MRI scanning session during which structural MRI and MRS measurements were collected. Procedures contributing to the analyses presented herein are described in more detail next.

2.3. MRI acquisition and processing

Imaging was performed on a clinical 7.0 Tesla MR scanner (Siemens Magnetom Terra) with a 32 channel receiver head coil. T1-weighted images were acquired with a 3D magnetization-prepared-rapid-gradient-echo (MPRAGE) sequence (TR=2290 ms, TE=2.95 ms, voxel size=0.7×0.7×0.7 mm³, acquisition matrix=224×168, flip angle=7°, slice thickness 0.70 mm, 240 total slices) and were used to place each MRS voxel. Single-voxel MRS utilized a STEAM sequence (TR/TE/TM = 3000 ms/20 ms/13 ms, 2048 points, 3000 Hz spectral bandwidth, 128 averages) with WET water suppression.¹⁷ In addition, a non-suppressed water reference spectrum was obtained with 16 averages. Voxels were positioned in the dorsal anterior cingulate cortex (ACC, voxel size=40×20×20mm³), ventromedial prefrontal cortex (VMPFC, voxel size=20×20×20mm³), and left DLPFC (voxel size=20×20×20mm³). Voxels underwent automated optimization that included 3 dimensional shimming, transmit gain, frequency adjustment, and water suppression. Manual shimming was performed to optimize the magnetic field homogeneity of the selected spectroscopy volume of interest to a line width of <30 Hz FWHM of the magnitude water signal when the initial full width at half maximum (FWHM) of the water signal was ≥30 Hz.

Pre-processing was performed in OpenMRSLab including coil combination, frequency and phase correction, as well as eddy current correction and residual water removal.^{18 1}H single voxel spectroscopy spectral fitting was performed using LCModel software,¹⁹ with a customized basis set of simulated spectra for 19 metabolites which included glutamate, N-acetylaspartarte (NAA), N-acetylaspartateglutamate (NAAG), glutamine, alpha/beta glucose resoannces, beta-glucose, glutathione, scyolloinositol, alanine, lactate, glycine, myoinositol (mI), aspartate, phosphocholine (PCh), creatine (Cr), glycerophosphocholine (GPC), phosphocreatine (PCr), taurine, methyl resonances of creatine. Total choline (PCh+GPC), total NAA (NAA+NAAG), total creatine (Cr+PCr) and total myoinositol (mI+Glyc) were used for metabolite comparisons. Also a full set of standard lipid and macromolecures were included in the macromolecule analysis (Lip13a/b, Lip09, Lip20, MM09, MM12, MM14, MM17, MM20). Visual inspection of all spectra was performed for quality control according to the following criteria: 1) a FWHM linewidth of the NAA <30Hz, 2) signal-to-noise ratio (SNR) >5, and 3) Cramer-Rao Lower Bound (CRLB) of NAA <5. Spectra not meeting these criteria were re-processed with eddy current or frequency shift correction removed as there were several cases where either processing step induced reduced spectral quality. One DLPFC voxel and one VMPFC voxel did not meet quality control metrics after re-processing were eliminated from further analyses. FWHM, SNR, and CRLB are described in the Supplemental Table and males and females did not significantly differ on these quality control measures (all p > 0.248). GSH levels in institutional units (i.u.) were calculated in the LCModel output and echo time correction of 0.860708 was applied.²⁰ Representative voxel placement and MRS spectra are shown in Figure 1.

Figure 1. Representative voxel location (left) and MRS spectrum (right).

Blue line in spectrum shows the GSH fit from LCModel.

2.4. Peripheral sex hormones and GSH measurements

Peripheral sex hormones and GSH were assayed from a fasted morning blood draw. Processing and analysis were conducted by trained staff at the Brigham Research Assay Core (BRAC). For sex hormones, serum was stored at −80°C until assayed. Sex hormones were measured using liquid chromatography-tandem mass spectrometry (LC/MS/MS) and included estradiol, estrone, progesterone, total testosterone and free testosterone. For peripheral GSH, a fasted whole blood sample was stored at room temperature and assayed for total GSH within one week of collection utilizing LC/MS/MS.

2.5. Statistical analyses

T-tests were used to test relationships involving dichotomous variables, including to test group differences based on self-reported biological sex. Relationships between continuous variables of interest, including relationships of peripheral and brain GSH with circulating sex hormones, were tested using Pearson’s correlations. We conducted separate analyses for each voxel location in relation to circulating sex hormones. Based on extensive prior evidence of an antioxidant role of estrogens, estradiol and estrone were our primary sex hormones of interest. Given strong correlation between estradiol and estrone (r=0.909, p<0.001), we do not correct for multiple comparisons and report results as significant at nominal p < 0.05 and a trend-level at nominal p < 0.1. Progesterone and testosterone were examined as secondary variables of interest. Given potential confounding effects of age and mood symptoms, we assessed the influence of these variables in validation analyses using linear regression models, though we are unable to fully disentangle the effects of these variables in the current sample. Study data were managed using REDCap electronic data capture tools hosted at Brigham and Women’s Hospital.^21,22 Statistical analyses were scripted and conducted using R and RStudio.^23,24

3. Results

3.1. Demographic and clinical characteristics

Demographic and clinical features of participants are shown in Table 1.

Table 1.

Demographic and Clinical Characteristics, Sex Hormones, and GSH by Sex^a

	Female (N=18)	Male (N=16)	Statistical test, p-valueb
Age	47 (8)	49 (10)	t(32) = −0.62, p = 0.540
Race (% White)	94%	88%	X2(3) = 3.18, p = 0.364
YMRS	1.61(2.50)	0.63 (1.59)	t(29.1) = 1.39, p = 0.176
MADRS	3.61 (6.99)	0.25 (0.58)	t(17.3) = 2.03, p = 0.058
Estradiol	57.52 (69.52)	19.06 (8.69)	t(18.3) = 2.30, p = 0.033
Estrone	29.52 (29.34)	23.69 (12.90)	t(22) = 0.50, p = 0.621
Progesterone	0.90 (2.36)	0.39 (1.21)	t(31) = 0.77, p = 0.446
Total Testosterone	22.97 (9.33)	468.56 (175.74)	t(15.1) = −10.12, p < 0.001
DLPFC GSH	3.45 (0.41)	3.71 (0.75 )	t(31) = −1.25, p = 0.222
ACC GSH	3.50 (0.72)	3.24 (0.4 5)	t(32) = 1.21, p = 0.237
VMPFC GSH	3.93 (0.90)	3.94 (0.86)	t(31) = −0.04, p = 0.972

^aResults are presented as mean(s.d.) or %. YMRS=Young Mania Rating Scale, MADRS=Montgomery-Asberg Depression Rating Scale

^bWhere the assumption of equal variance was not met we report the t statistic, df, and p-value with unequal variances assumed.

As expected, females exhibited higher concentrations of estradiol than males (p = 0.033) and males exhibited higher concentrations of total testosterone compared to females (p < 0.001). No significant differences between males and females were observed for age, race, current mood symptoms, estrone, progesterone, free testosterone, or peripheral GSH. Peripheral GSH did not correlate with brain GSH in any of the MRS voxels in the full sample (all p > 0.582) nor separately in females (all p > 0.236) or males (all p > 0.582). Subsequent analyses focus on our primary variables of interest, brain GSH markers.

3.2. Sex differences and brain GSH

We examined GSH concentrations in the DLPFC, VMPFC and ACC. No significant differences between males and females were observed for GSH levels in the DLPFC, VMPFC, or ACC (see Table 1). In addition, we did not observe significant differences between males and females for GSH when controlling for YMRS and MADRS scores (all p > 0.198) or when controlling for age (all p > 0.229).

3.3. Sex hormones and brain GSH

Estradiol and estrone

We observed a nominally significant inverse correlation of estradiol with DLPFC GSH (r = −0.414, p = 0.044, Table 2, Figure 2) as well as a trend inverse correlation of estrone with DLPFC GSH (r = −0.358, p = 0.094) in the combined sample of males and females. No significant effects were observed for relationships of estrogens with either ACC or VMPFC GSH.

Table 2.

Correlations Between GSH and Sex Hormones in the Full Sample and by Sex^a

		Estradiol	Estrone	Total Testosterone	Progesterone
Full sample	VMPFC GSH	0.318 (.129)	0.290 (.180)	0.237 (.224)	0.008 (.965)
	DLPFC GSH	−0.414 (.044)	−0.358 (.094)	0.104 (.599)	−0.093 (.611)
	ACC GSH	−0.085 (.687)	−0.182 (.394)	−0.095 (.623)	0.018 (.921)
Females	VMPFC GSH	0.377 (.136)	0.345 (.190)	0.171 (.596)	0.119 (.661)
	DLPFC GSH	−0.495 (.043)	−0.450 (.081)	−0.603 (.038)	−0.112 (.680)
	ACC GSH	−0.197 (.432)	−0.186 (.474)	−0.189 (.535)	−0.070 (.788)
Males	VMPFC GSH	0.490 (.264)	0.079 (.866)	0.327 (.216)	−0.231 (.390)
	DLPFC GSH	−0.398 (.376)	−0.326 (.475)	−0.080 (.769)	−0.049 (.856)
	ACC GSH	−0.020 (.966)	−0.718 (.069)	0.433 (.094)	0.192 (.477)

^aResults are shown as correlation (p-value). Results with p<0.05 in bold.

Figure 2. Relationship between DLPFC GSH with a) estradiol in the combined sample of males and females b) estradiol in females and c) total testosterone in females.

Within the sample of females the same pattern was observed. Estradiol was nominally significantly and inversely correlated with DLPFC GSH (r = −0.495, p = 0.043, Table 2, Figure 2) and estrone was negatively correlated with DLPFC GSH at a trend level (r = −0.450, p = 0.081) in females. No significant effects were observed for relationships of estrogens with ACC or VMPFC GSH in females.

In males, no significant relationship was observed between estradiol and GSH in any brain region (Table 2). However, we did observe a trend-level correlation of estrone (r = −0.718, p = 0.069) with ACC GSH in males.

Progesterone and total testosterone

In secondary analyses, we were not able to detect significant correlation between brain GSH and either progesterone or testosterone in the combined sample of females and males. In females, only total testosterone was significantly inversely correlated with DLPFC GSH (r = −0.603, p = 0.038). We were not able to detect significant correlation between progesterone and DLPFC GSH or for relationships of either sex hormone with ACC or VMPFC GSH in females. In males, no significant relationship was observed between progesterone and GSH in any brain region (Table 2). However, we did observe a trend-level positive correlation of testosterone (r = 0.433, p = 0.094) with ACC GSH in males.

Validation analysis controlling for age

While age was nominally significantly negatively correlated with estradiol (r = −0.501, p = 0.011), age was not a significant covariate in any of the models tested (p > 0.375). In models controlling for age, the observed relationships between estradiol and DLPFC GSH remained nominally significant (p = 0.043) and we continued to observe a trend in association of estrone with DLPFC GSH (p = 0.084). In females, associations of DLPFC GSH with each of estradiol, estrone, and testosterone were qualitatively similar after adjusting for age, but were no longer statistically significant (p = 0.163, p = 0.162, p = 0.119, respectively). The relationships of estrone and total testosterone with ACC GSH in males were no longer trend-level significant after adjusting for age (p = 0.210 and p = 0.272, respectively).

Validation analysis controlling for current mood symptoms

In models controlling for YMRS and MADRS scores, the observed relationship between estradiol and DLPFC GSH remained nominally significant (p = 0.013) and we continued to observe a trend in association of estrone with DLPFC GSH (p = 0.067). In both models, YMRS scores emerged as a significant control variable (all p < 0.029). In females, when controlling for YMRS and MADRS scores, the observed association between estradiol (p = 0.042) and total testosterone (p = 0.030) with DLPFC GSH remained nominally significant. The relationship between estrone and DLPFC GSH still showed a trend-level negative association (p = 0.096). YMRS and MADRS scores did not emerge as significant control variables in any of the models tested in females (p > 0.460). In males, the relationship of estrone with ACC GSH was no longer trend-level significant after controlling for YMRS and MADRS scores (p = 0.200). However, we did observe a nominally significant positive association of total testosterone with ACC GSH (p = 0.041). YMRS and MADRS scores were not significant in any models tested in males (p > 0.138).

4. Discussion

In this study, we investigated relationships sex hormones with oxidative stress in the brain by testing for associations of estrogens with regional brain concentrations of GSH. Secondary analyses examined relationships of GSH with progesterone and total testosterone. We found a significant inverse correlation of estradiol with DLPFC GSH as well as a trend inverse association of estrone with DLPFC GSH in the combined sample of males and females, as well as in the sample of females only. Additionally, in females total testosterone correlated with GSH in the DLPFC, suggesting that the DLPFC is a key region that may be impacted by the interplay of sex hormones and GSH.

Estrogens have antioxidant properties,²⁵ upregulate antioxidant and longevity-related genes in females,²⁶ and upregulate the expression of antioxidant enzymes including GSH peroxidase.^27,28 For example, estradiol has been shown to prevent oxidative stress in mitochondria by preventing reactive oxygen species from forming.²⁸ Preclinical research indicates that the prefrontal cortex is a target of estrogen action with downstream effects on dendritic branching and interaction with the dopaminergic system.²⁹ As further evidence for a role of estrogen in prefrontal cortex, higher endogenous levels of estrogen are related to greater neuroplastic response in the DLPFC following HF-rTMS and tDCS in females when estrogen is high relative to males and females when estrogen is low.^30,31 The relationship of GSH with sex hormones in females has previously been shown primarily in the context of how estrogens relate to peripheral GSH and GSH-related enzyme activity, demonstrating for example a correlation between estradiol and menstrual-cycle dependent changes in GSH peroxidase activity.^9,32 In the current study, we extend this line of investigation to the brain with our observation of a negative association between estradiol and DLPFC GSH levels in females, as well as in the combined sample of females and males.

Though the role of testosterone in oxidative stress has been less studied than the role of estrogens, studies have shown that testosterone levels positively relate to total antioxidant capacity in men³³ and regulate GSH homeostasis in male and testosterone-treated female mice.³⁴ In the current study, we found a significant inverse correlation of total testosterone with DLPFC GSH in females in our study.

Previous studies have noted sex differences in peripheral GSH-related activity.³⁵ For example, healthy young females have been shown to have higher GSH/GSSG ratio than healthy young men, an indicator of healthier redox state that may be related to estrogen-related upregulation of antioxidant defenses.^36,37 However, others have reported that sex-differences in GSH metabolism decrease with menopause.^36,38 In a previous study of central GSH, healthy young females showed higher overall mean GSH in the parietal and frontal cortices than healthy young males, though neither sex hormone levels nor menstrual cycle status of the females were reported in that study.¹² In the current study, we were not able to detect a sex-difference in brain GSH levels, possibly related to the inclusion of a wider age range in our sample that includes for example females both pre- and post-menopause, although menopause status was not significantly associated with GSH levels in our sample.

Peripheral and preclinical studies of GSH indicate that GSH levels decrease with age.^34,39,40 In the brain, age-related effects of GSH are mixed and have been shown to vary by region.^41,42 For instance, GSH in the occipital and parietal lobes is higher in young adults relative to elderly adults;^12,41 however, GSH in the medial frontal and sensorimotor cortices has been shown to be higher in older adults compared to younger adults when accounting for age-related atrophy.⁴² Our work extends these studies, finding age negatively correlated with VMPFC GSH (r = −0.438, p = 0.011) and at a trend level with ACC GSH (r = −0.326, p = 0.060), but not DLPFC GSH (r = −0.015, p = 0.933).

Strengths of our study include utilization of 7 Tesla MRS allowing us to reliably measure GSH levels in multiple brain regions due to increased SNR and spectral dispersion,⁴³ and inclusion of participants across the adult reproductive lifespan to capture a range of sex hormones levels. Another major strength of our study is that we have accounted for several key factors known to impact on GSH and its measurement in our study protocol, including fasted measurement to address food intake, standardized time of day of assessment to address circadian influences, exclusion of individuals with positive toxicology screen, and exclusion of individuals taking the antioxidant NAC. We also conducted validation analyses accounting for additional factors that may impact on GSH including age and mood symptoms. Nonetheless, an exhaustive look at all such factors is beyond the scope of this study. The results of this study should be interpreted in the context of several additional limitations. Our data are cross-sectional and so inferences cannot be drawn about causal relationships or the effect of within-person changes in sex hormones on GSH. The MRS data are not corrected for tissue fractions. The sample size is also modest, limiting our ability to detect small effect sizes and test for interaction effects. Most females in our sample were in a low estrogen state on the day of their study visit, which may contribute to the similar relationship with GSH we observed in males and females.

In conclusion, our study provides evidence of diminished GSH in DLPFC, consistent with higher oxidative stress, in females with higher estradiol and with higher total testosterone. Future studies should examine relationships between sex hormones and oxidative stress in the brain longitudinally within individuals and assess additional potential moderators of the relationships, including menstrual cycle and menopause-related changes.

Highlights.

• 7 Tesla MRS allows for reliable measurement of regional brain glutathione levels

• Glutathione levels are diminished in the DLPFC in women with higher estradiol

• The DLPFC is a key region that may be impacted by the interplay of sex hormones and oxidative stress