Estradiol Replacement as a Potential Enhancer of Working Memory and Neuroplasticity in Hypogonadal Trans Women

In the last two decades, neuroimaging research has aimed to understand the neurobiology of gender dysphoria. However, studies investigating the implications of cross-sex hormone therapy (CSHT), henceforth referred to as CSHT, on brain plasticity are still limited. To date, research has primarily examined similarities and differences in brain anatomy and function with respect to biological sex and gender identity [1, 2] and the effects that sex hormones exert on brain structures [3]. In cisgender individuals, sex steroids can have a significant impact on neurotransmitter activity, as well as on structural macro- and microanatomy [4‒6]. Similarly, CSHT in gender-diverse individuals seems to lead to brain adaptations, making some structural characteristics, such as brain size and volume, more aligned with those typically seen in individuals of the same biological sex [7‒9]. Beyond sexual dimorphism, those brain changes might mark adaptive responses to the chemical and physical environment called neuroplasticity. Ultimately, neuroplasticity might be indirectly related via memory and cognition as a potential indirect measure of the neuronal response to damage or recovery [10‒12]. In cisgender women, postmenopausal changes in learning and memory appear to be related to ovarian failure, leading to physiological hypogonadism or a decline in hormones such as estradiol (E₂) [13‒15]. While research has been conducted investigating brain and cognitive changes following gender-affirming processes [16, 17], the potential impacts of correcting hypogonadism after trans women perform gender-related gonadectomy [17] are an area in need of further research.

Of importance, 17-beta-E₂, a neuroactive gonadal hormone used for gender-affirming purposes in trans women, appears to act on both intracellular estrogen receptors (ERs), alpha ER (ERα), and beta ER (ERβ), as well on extranuclear or membrane-localized receptors [18, 19]. Both binding sites are involved in promoting long-term potentiation (LTP), a neuroplastic event related to memory consolidation [20‒22]. Memory-enhancing and neuroprotective effects of E₂ are also seen, following its activation of a myriad of cellular membrane receptors such as metabotropic glutamate receptors (mGluRs), which belong to the larger class of G-protein coupled receptors [23‒25]. Although the precise mechanisms remain unclear, it has been suggested that alpha ER/mGluR1a signaling proceeds in an extracellular signal-regulated kinase (ERK)-dependent cascade, leading to cAMP response-element binding protein (CREB) phosphorylation [25]. This process can lead to increased gene expression of synaptic or protective proteins involved in neuronal growth and survival [19, 26].

The effects of E₂ are not limited to memory performance. Glutamate, the most abundant excitatory neurotransmitter, is present within tightly regulated concentrations in the brain, and its effects seem to be modulated by estrogen response elements associated with neuroprotection [27, 28]. Excessive levels of glutamate can overstimulate synaptic receptors and lead to an excitotoxic event, a phenomenon seen in brain damage/vascular injury, along with other severe neuropathologies [29, 30]. This event seems to be mediated by the N-methyl-D-aspartate receptor-type for glutamate, a class of ionotropic, fast-acting, and voltage-sensitive binding site involved in synaptic plasticity [31‒33]. E₂ has been shown to play a dual role in both enhancing glutamate release from neuronal cells via activation of the phosphatidylinositol 3-kinase (PI 3-kinase) and mitogen-activated protein kinase (MAPK) signaling mechanisms and attenuating glutamate-induced excitotoxicity [34, 35]. The relationship between E₂, glutamate, and LTP, a memory correlate, has also been seen in the CA1 hippocampal cells. Many studies have shown that E₂ augments dendritic spine density and regulates glutamate transporters and neurotransmission, enhancing the amplitude of excitatory postsynaptic potentials and LTP magnitude [20, 35‒45].

Proton nuclear magnetic resonance spectroscopy was used for the in vivo quantification of brain glutamate levels [46, 47]. Studies applying this noninvasive neuroimaging technique have shown that the measurements of specific metabolites are effective in the detection of cognitive diseases and associated loss of function [48, 49]. Some investigations have also shed insight into important similarities and differences between trans and cisgender individuals when investigating common brain metabolites, such as creatine and glutamate [50, 51]. For instance, Collet et al. [50] reported significant, elevated metabolite ratios (e.g., glutamate/creatine) in cisgender over trans persons in the amygdala-anterior hippocampal region, following magnetic resonance spectroscopic analysis. Combining nuclear magnetic resonance spectroscopy with clinical and neurocognitive data may provide a more comprehensive understanding of the role of sex hormones in the brain, particularly for their protective or deleterious effects [52]. Recently, our group showed that E₂ replacement in hypogonadal trans women strengthens the functional connectivity between the thalamus and primary sensorimotor cortex and affects cortical thickness in E₂-receptive brain regions [17, 53]. However, the underlying biochemistry of sex hormones that translates into neuroimaging and its relationship with a neurobiological basis of neurocognition is still unclear.

The objective of the present study was to investigate how E₂ administration in trans women post-gender-affirming-related gonadectomy can affect glutamate concentration in regions of the brain known to be involved in semantic and working memory. Following E₂ replacement, we expect glutamate levels to change in pre- and post-measures, which will be related to memory performance. Three distinct brain regions were selected accordingly, including the left dorsolateral prefrontal cortex (L DLPFC) [54‒56], the left hippocampus (posterior region) [57, 58], and the pregenual part of the anterior cingulate cortex (ACC) [59, 60]. These regions, only assessed in the dominant hemisphere, are associated with glutamatergic transmission [56, 57, 61, 62] and known to be involved in semantic and working memory, as demonstrated by previous neuroimaging investigations [58, 63‒65]. To our knowledge, few studies have investigated the correlation between glutamate and cognitive performance in hypogonadal individuals in the aforementioned regions, which could help elucidate the impact of sex hormone deficiency in other populations as well (i.e., menopausal women). Glutamate levels in these regions were assessed using single voxel spectroscopy (SVS). Cognitive information was also collected to corroborate spectroscopy findings, as the combination of clinical measures with brain metabolites can strengthen the meaning of the neuroimaging findings.

Participants: eighteen trans women (assigned men at birth) who underwent post-gender-affirming-related gonadectomy and were not using antiandrogens were invited to participate in the present study at routine follow-ups at our clinic. One individual was excluded due to encephalomalacia, resulting in a final sample of 17 participants. To ensure sample homogeneity, all participants were measured for handedness at baseline. The inclusion criteria were (1) age range of 18–59; (2) at least 1 year after gonadectomy (related to gender-affirming surgery), all the patients went through pubertal development before surgery; and (3) agreement to stop hormone therapy for a minimum of 30 days. Exclusion criteria were (1) past or current neuroendocrine disease; (2) current neuropsychiatric illness such as epilepsy, psychosis, or current or past 90 days of moderate mood or anxiety disorder; (3) use of psychotropics (for the last 30 days); (4) current substance use disorder; (5) brain injury within the last 3 years or severe head trauma in lifetime with loss of consciousness; and (6) major untreated medical illness (including untreated HIV/AIDS or in treatment with detectable viral charge or low CD4).

Following enrollment in the current study, participants were instructed to stop CSHT for a period of 30 days to reach a hypogonadal state. At the end of this period (t1), we assessed working and semantic memory, hormonal assays, and brain imaging. Subsequently, individuals resumed CSHT (oral or transdermal formulations of E₂ without progesterone) for the next 60 days (day 60 = t2), after which they underwent the same assessment battery conducted at the end of the 30-day washout (WO) period. Decisions regarding optimal E₂ formulations were made by endocrinologists according to individual clinical profiles.

Mood and anxiety were assessed using the clinician-administered Hamilton Rating Scales for Depression (HAM-D) [66] and Anxiety (HAM-A) [67]. Working memory was assessed using the Operational Memory Index (OMI) from the Wechsler Adult Intelligence Scale (WAIS-3), while semantic memory was assessed with the Rey Auditory Verbal Learning Test (RAVLT). Using the RAVLT, recall after 1 min, recall after 7 min, and learning over trial were used to represent immediate and late recall, and verbal learning, respectively. The tests were administered by a research assistant with a background in psychology, and the results were scored and reviewed by two independent psychologists. The same set of assessments was applied at t1 and t2.

As previously reported in the methodology section in Schneider et al. [53], venous samples were collected between 8 a.m. and 10 a.m. at each of the time points. E₂ was measured using an electro-chemiluminescence immunoassay (ECLIA; Roche Diagnostics, Mnaheim, Germany) with assay sensitivity of 5.0 pg.mL and intra- and inter-assay CV of 5.7% and 6.8%, respectively. Sex hormone binding globulin (SHBG) was measured by CLIA (Immulite 2000 Siemens). Follicle stimulant hormone (FSH) and luteinizing hormone were measured by CLIA (CentaurXP; Roche Diagnostics, Mannheim, Germany).

Anatomical protocol: Structural images were acquired in a Philips Ingenia 3.0T MR system (Best, The Netherlands, 2015) located at Hospital de Clínicas de Porto Alegre with a 32-channel head coil at each of the two time points. A T1-weighted 3D magnetization-prepared rapid acquisition with gradient echoes sequence with 200 sagittal orientation slices was applied using the following parameters: TE = 3.9 ms, TR = 8.5 ms, TI = 900 ms, flip angle = 8°, 256 × 256 matrix, matrix size = 272 × 272, in-plane voxel size 0.94 × 0.94 mm, slice thickness 0.94 mm, no gap, FOV = 256 mm.

A single-voxel PRESS pulse sequence was applied to three regions: the posterior left hippocampus, L DLPFC, and subgenual ACC (subACC). The acquisition parameters were TR 2,000 ms; TE 35 ms; number of excitations 128; flip angle 90°; 15 × 15 × 20 mm³ voxel size. The size of the voxels was determined to respect the anatomy of the structure without contaminating with the noise of surrounding tissues. For metabolite quantification, another acquisition without water suppression was acquired with the same parameters but NEX 16.

Proton spectra were fitted and quantified using the LCModel algorithm [68]. Briefly, it analyzed an in vivo proton brain spectrum as a linear combination of model in vitro spectra from individual metabolite solutions. After metabolite quantitation, partial volume corrections were applied to correct for brain tissue concentration. Partial volume correction is the gold standard to estimate the absolute amount of the metabolite in spectroscopy [69]. Partial volume correlation was performed using a Matlab R2015b (The Mathworks, Natick, USA) algorithm based on the Gasparovic algorithm [70]. Estimations of the amount of gray matter, white matter, and cerebrospinal fluid present in the selected voxels were obtained to properly correct partial volume. This approach reduces the probability of underestimation of absolute metabolite concentration given their different concentrations according to tissue type [71, 72].

The FMRIB software library was used to separate the fractions of the three main brain regions using a standardized routine. Brain extraction tool was used initially [73] followed by segmentation function for anatomical image [74]. Following brain segmentation, data including peak concentration for each metabolite and proportion of GM, T1, and T2 acquisition’s relaxation time were input into Matlab, adding angular and spatial coordinates to adjust for each voxel rotation in native space. This latter approach was used to reduce errors due to misalignment of voxel positioning.

In addition to glutamate, choline, n-acetyl aspartate (NAA), creatine, and myoinositol amounts were also extracted for each voxel. T1 relaxation time for those metabolites in gray and white matter were based on Mlynárik and colleagues’ experiment [75], while their relaxation time at T2 (GM and WM) were based on Choi and colleagues’ results [76]. The final concentration of metabolites coming from the Gasparovic equation was informed by nmol/kg of water for GM.

All data were initially assessed for normality using the Shapiro-Wilk test (not shown) and corrections for multiple comparisons were conducted when applicable. Paired sample t tests were used to compare pre- versus post-intervention mean values for E₂, FSH, luteinizing hormone, and SHBG levels. Similarly, paired sample t tests were used to compare semantic and working memory (RAVLT and OMI, respectively) [77, 78] and to compare mood and anxiety symptoms (HAM-D and HAM-A, respectively) [66, 67] between time points.

Paired sample t tests were used to compare mean values of glutamate between the two time points (t1: WO and t2: CSHT) among the three single voxels within the ACC, L DLPFC, and left hippocampus.

Linear regression analyses were further used to investigate the relationship between semantic memory and the concentration of glutamate in the left hippocampus during WO and CSHT time points independently. Similarly, the potential relationship between working memory and concentrations of glutamate in the L DLPFC and subACC was also tested as pertinent to our a priori hypothesis. All analyses were adjusted for age.

After testing for a correlation between glutamate and memory (semantic and working memory) in the L DLPFC and the left hippocampus, moderation analyses were employed to investigate whether these relationships were intermediated by E₂ levels in a linear model. To assess the moderation effect of E₂, two models were compared: (1) including E₂ as a predictive variable in the original linear model, and then (2) adding the product between E₂ and glutamate as a predictor.

Exploratory analyses investigating mean differences between 4 other major metabolites (i.e., myoinositol, creatine, NAA, and choline) for each of the single voxels were also conducted using paired t tests. The results are shown in the online supplementary Appendix A (for all online suppl. material, see www.karger.com/doi/10.1159/000527130) since they were not part of a priori established analyses.

Following CSHT, E₂ levels were found to increase during the reinstatement phase as compared to the WO/hypogonadism phase (p = 0.0002), whereas the mean FSH decreased following E₂ replacement as expected (t1 = 91.75 mUI/mL, t2 = 56.56 mUI/mL; p = 0.02) (Table 1). Performance (pre- and post-intervention) in both the first- and seventh-minute auditory memory recall were also statistically significant (RAVLT-A1: t2 – t1 = 0.80, p_A1 = 0.0026; RAVLT-A7: t2 – t1 = 1.00, p_A7 = 0.0402), with better recall at time point 2. There were no significant changes in mood (t1 = 4, t2 = 2; p_HAM-D = 0.51) and anxiety rating scores (t1 = 5, t2 = 2; p_HAM-A = 0.16).

Results comparing the concentration of glutamate at the WO and CSHT time points within three single voxels showed no statistically significant difference (p_{Glu Hippo} = 0.14; p_{Glu DLPFC} = 0.92; p_{Glu ACC} = 0.35) (Table 2). Despite non-significance, trends toward larger decreases in glutamate were observed in the posterior part of the left hippocampus (t2–t1 = −0.53) and ACC voxels (t2–t1 = −0.31), in contrast to small increases of glutamate in the DLPFC (t2–t1 = 0.03) (Table 2).

Models assessing the relationships between glutamate in the SVS and neurocognitive memory outcomes adjusted for age showed a significant correlation between glutamate and working memory (OMI) in the L DLPFC following E₂ replacement (β = 0.60, p_{GLU DLPFC} = 0.01; p-model: 0.04; R² = 0.28; Table 3) (shown in Fig. 1a). During the WO phase, the same model was found to be significant (p-model: 0.03; R² = 0.11) (shown in Fig. 1b), although the relationship between the predictor variable Glu within the DLPFC and OMI was statistically nonsignificant (β = 0.10, p_{GLU DLPFC} = 0.72) (Table 3). In regard to semantic memory, there was a positive relationship between Glu in the left Hippo and RAVLT-A1 scores (first 1-min recall) for both time points (CSHT: β = 0.48; p_{GLU Hippo} = 0.04; p-model = 0.04; R² = 0.27) (WO: β = 0.62; p_{GLU DLPFC} = 0.01; p-model = 0.03; R² = 0.30). No significant correlation was observed between Glu Hippo and the RAVLT-A7 for both the WO (p_{GLU Hippo} = 0.20) and reinstatement phases (p_{GLU Hippo} = 0.21). Additionally, no significant relationships between glutamate levels in the subACC voxel and OMI in either the WO or CSHT phases were seen (WO: β = −0.07; p_{GLU Hippo} = 0.80; p-model = 0.85; R² = −0.12) (CSHT: β = 0.19; p_{GLU Hippo} = 0.49; p-model = 0.74; R² = −0.09).

The moderation analysis showed no superiority of the model including the interaction between Glu DLPFC and E₂ levels during the E₂ restatement phase (CSHT R²_{model without interaction} = 0.37; R²_{model with interaction} = 0.32). Moderation analysis of E₂ in the relationship between A1/glutamate hippocampus was also not significant during the CSHT (phase R²_{model without interaction} = 0.18; R²_{model with interaction} = 0.14) (Table 4).

Post hoc exploratory analyses including other metabolites (creatine, myoinositol, NAA, and glycine) showed no significant mean differences for almost all of the metabolite sites aside from the myoinositol and creatine in the DLPFC (mI; p-unc = 0.013; Cre, p-unc = 0.019). Overall, there was a trend that depicted a decrease in the absolute amount of the metabolites in most of the SVS across the brain (online suppl. Appendix A).

The present study is the first to investigate the impact of sex hormone deficiency, following surgically induced hypogonadism on brain neurotransmitters by means of E₂ replacement (CSHT) in trans women. This research adds a new perspective on the cognitive impacts associated with hormone replacement therapy for trans women. Although there were no significant pre- and post-differences in the concentration of glutamate within our regions of interest, there was a significant relationship between the amount found in the DLPFC and working memory performance during the E₂ reinstatement phase. This relationship was not observed during the WO phase. Contrary to our hypothesis, E₂ was not found to moderate the relationship between glutamate levels and memory performance. In addition, there was a positive relationship between glutamate in the posterior left hippocampus and first-minute recall for the verbal memory test.

Despite limited studies investigating the impact of hypogonadism among transgender individuals, this has been a topic of increased interest with respect to menopause, a physiological event marked by cessation of ovarian function in cisgender women. In a related cohort of surgically menopausal women, it was found that scores for the immediate verbal recall test were superior in a sample treated with E₂ as compared to the placebo-control group, thereby showcasing the short-term memory enhancement effects of the hormone [79]. Both the immediate and delayed tests also showed that decreased E₂ levels were correlated with poorer recall. Although the exact mechanism by which E₂ might enhance memory performance is still not fully understood, an increase in dendritic spine density in the prefrontal cortex and hippocampus or possible epigenetic alterations has been proposed [36, 80‒83]. Research suggests that E₂ is likely to be implicated in promoting synaptic plasticity by up-regulating synaptic proteins and stimulating glutamate exocytosis [20, 35, 46].

Due to the lack of statistical significance in pre- and post-testing for changes in absolute glutamate, these preliminary results are still inconclusive when it concerns the specific roles of E₂. Nonetheless, our findings suggest that glutamate may be involved in promoting neuroplasticity when correcting hypogonadism with E₂ as observed by the linear models (Fig. 1a). Therefore, we believe that E₂ may reinforce and influence brain metabolic homeostasis through glutamate dynamics. As for future directions in neuroendocrinology sciences, the present findings show potential neuronal adaptation at the synaptic level due to hormonal interference in neurotransmitters seen in early age hypogonadism. Therefore, our preliminary evidence also has potential implications for the improved management of hypogonadism-related challenges in menopausal women and transgender individuals.

Studies coupling glutamate measures with functional magnetic resonance imaging have brought more robust evidence of the role of this neurotransmitter in cognition using real-time paradigms. Increased glutamate-induced metabolic activity during task activation has been found to reflect increased excitatory activity in different cortical areas of the brain, such as the ACC [61, 81], hippocampus [62], and DLPFC [49]. For instance, Woodcock et al. [49] showed an increase of 2.7% in glutamate levels in the DLPFC during the passive fixation stage of a visual working memory and at the 2-backstage memory recall. This supports previous findings about the role of glutamate in working memory with an expected increase in metabolic and synaptic activity during memory elicitation. The study also found that the possible modulatory effect of glutamate in this working memory paradigm was attenuated during the “stress-induced” phase (yohimbine + hydrocortisone administration) [49]. This suggested decreased glutamate activity when memory demands were made in stressful conditions. In our experimental model, it is possible to hypothesize that hypogonadism represents a “synaptic stress” condition. Impacts of synaptic stress may be a possible reason why we did not find a relationship between glutamate in the DLPFC and working memory as opposed to in the E₂ replacement phase.

With reference to the significant relationship between glutamate levels in the left hippocampus and verbal memory performance (A1 RAVLT-first-minute recall) as seen in both the washout and hormone replacement phases, our results do not allow us to draw any definite conclusions. First-minute recall has been associated with attention rather than working memory itself, and such a significant association has not been seen in late phases of verbal recall (memory consolidation). In previous literature, on in vitro models of hypogonadism, E₂ administration led to the activation of a cascade of events involving group I of the mGluR and subunits of N-methyl-D-aspartate receptors through LTP in dentate gyrus neurons [32]. This supported hypotheses on the enhancing effect of E₂ on semantic memory. In vivo rodent and primate models have helped validate the knowledge that hypogonadal animals undergoing E₂ replacement outperform the controls who do not receive the hormone [84]. However, this effect might be dependent on the time that E₂ replacement began [85]. Moreover, in ovariectomized mice, the administration of a glutamate-agonist seems to increase visual retention during a T-maze paradigm, which was augmented with intra-hippocampal coadministration of 17-beta E₂ [86].

It is also worth highlighting that the neural anatomy of trans women may be affected by biological sex, and these physiological differences may lead to altered responses toward E₂ replacement. Early in development, sex-specific gonadal hormones may directionally influence brain structures, such as the hippocampus [87‒90]. Further, through animal model investigations, it is apparent that exogenously supplemented E₂ has memory-enhancing effects in gonadectomized and intact male and female rodents [23, 85, 91‒93]. For example, Gibbs and Johnson demonstrated that ovariectomized rats receiving E₂ had fewer working memory errors than both intact and ovariectomized rats with no E₂ in certain training blocks [93]. Thus, there is supporting evidence to suggest that E₂ could have a positive effect on cognitive functions and in shaping brain structures [16, 20, 40, 82, 89, 93].

This study has a small sample size that may potentially underpower the findings. Hence, it is likely that the power to detect a significant difference or refute the hypothesis that E₂ modulates the relationship between memory and glutamate is reduced. Further, there were no between group comparisons. Although individuals were compared to themselves (pre and post) following a within-group design. In addition, there are age differences between participants that might elicit a varied response to exogenous E₂ administration owing to (1) an aging factor and (2) differences in the time of exposure to E₂ prior to surgery. Evidence suggests that replacing E₂ early in life or during middle age triggers different responses to its replacement. In addition, lifetime exposure to androgens and estrogens varied among participants, along with the age at which they received gender-affirming surgery, which would impact the study’s findings.

Despite the abovementioned limitations, there were several strengths of this study. Notably, all cognitive tests were taken immediately before the client started their scans. Although we did not perform functional spectroscopy representing a real-time glutamate synaptic response, maintaining this protocol allowed us to reduce variability in glutamate response to stimuli. This is of utmost importance since functional magnetic resonance spectroscopy demonstrated a time effect along task blocks in glutamate concentration, which means that late stages of the test in healthy controls (for instance, 0-back vs. 2-back) were associated with a higher concentration of glutamate [94]. To our knowledge, there are only two papers investigating glutamate differences in transgender brains [50, 95]. Thus, this is an innovative study in a population who has been marginalized with a significant need for the continued promotion of health care following affirming procedures.

The research protocol was approved by the internal Ethics Review Board from Hospital de Clínicas de Porto Alegre, Brazil. (CEP 15-0199). No financial compensation was offered to participants. All participants provided written informed consent in accordance with the Helsinki Declaration.

The authors have no conflicts of interest to declare.

The current study was funded by the Brazilian National Institute for Hormones and Women’s Health, Research and Events Incentive Fund of the Hospital de Clínicas de Porto Alegre (FIPE/HCPA) and the Conselho Nacional de Desenvolvimento Científico e Tecnológico (INCT/CNPq: 465482/2014-7).

Maiko Abel Schneider and Maria I.R. Lobato: conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; resources; software; supervision; validation; visualization; roles/writing-original draft; and writing-review & editing. Devon Malhotra, Taylor Hatchard^a, Sasha A. Haefner^a, and Sabrina K. Syan: writing-original draft and writing-review & editing. Poli M. Spritzer: conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; resources; supervision; validation; visualization; roles/writing-original draft; and writing-review & editing. Luciano Minuzzi and Benicio N. Frey: formal analysis; visualization; and supervision. Mauricio Anes: data curation and visualization. Andrew Nicholson and Margaret McKinnon: writing-review & editing. Taiane de Azevedo Cardoso and Alejandro Santos-Díaz: formal analysis and data curation. Karine Schwarz: data acquisition and writing-original draft.

Date can be made available through direct request to the authors or to the institution that holds the date under their protection in accordance with data privacy policy. Further inquiries can be directed to the corresponding author.