Cross sex hormone treatment is linked with a reversal of cerebral patterns associated with gender dysphoria to the baseline of cisgender controls

Abstract

Transgender persons experience incongruence between their gender identity and birth-assigned sex. The resulting gender dysphoria (GD), is frequently treated with cross-sex hormones. However, very little is known about how this treatment affects the brain of individuals with GD, nor do we know the neurobiology of GD. We recently suggested that disconnection of fronto- parietal networks involved in own-body self-referential processing could be a plausible mechanism, and that the anatomical correlate could be a thickening of the mesial prefrontal and precuneus cortex, which is unrelated to sex. Here, we investigate how cross-sex hormone treatment affects cerebral tissue in persons with GD, and how potential changes are related to self-body perception. Longitudinal MRI measurements of cortical thickness (Cth) were carried out in 40 transgender men (TrM), 24 transgender women (TrW) and 19 controls. Cth increased in the mesial temporal and insular cortices with testosterone treatment in TrM, whereas anti-androgen and oestrogen treatment in TrW caused widespread cortical thinning. However, after correction for treatment-related changes in total grey and white matter volumes (increase with testosterone; decrease with anti-androgen and oestrogen), significant Cth decreases were observed in the mesial prefrontal and parietal cortices, in both TrM and TrW (vs. controls) – regions showing greater pre-treatment Cth than in controls. The own body – self congruence ratings increased with treatment, and correlated with a left parietal cortical thinning. These data confirm our hypothesis that GD may be associated with specific anatomical features in own-body/self-processing circuits that reverse to the pattern of cisgender controls after cross-sex hormone treatment.

Graphical Abstract

This MR study of cortical thickness (Cth) investigates effects of cross sex hormone treatment in transgender men (TrM) and transgender women (TrW). While ‘feminizing’ treatment with antiandrogens and estradiol in TrW leads to widespread cortical thinning, testosterone treatment in TrM causes an increased thickness of the insular and mesial temporal cortex. However, when correcting for global estrogen and testosterone associated changes in the total white matter and grey matter volume, both transgender groups showed a relative tinning of the mesial frontal cortex and parietal cortex, regions in which we before treatment detected increased Cth. Thus, cross sex hormone treatment seems to be linked with a reversal of cerebral patters associated with gender dysphoria to the baseline of cisgender controls.

INTRODUCTION

Gender dysphoria (GD) is capturing a steadily increasing interest in our society. Yet, its underpinnings are largely unknown. For a long time, this condition has been theorized to be a consequence of altered cerebral sexual differentiation (Swaab, 2004). However, brain imaging studies addressing this hypothesis, while largely consistent with respect to structural and functional differences among cisgender male and female controls (Bramen et al., 2012; Filipek, Richelme, Kennedy, & Caviness, 1994; Goldstein et al., 2001; Lentini, Kasahara, Arver, & Savic, 2013; Luders et al., 2006; Savic & Arver, 2011, 2014), seem rather inconsistent with respect to brain-structure-findings among persons with GD. This especially applies to transgender men (TrM; persons with female sex assigned at birth and male gender identity), for which some studies have reported a cerebral pattern largely congruent with that of cisgender females (Hoekzema et al., 2015; Zubiaurre-Elorza et al., 2013), while others have found partly similar structural neural characteristics to those in cisgender males (Burke, Manzouri, & Savic, 2017; Simon et al., 2013), and still others have identified a pattern different from both cisgender male and cisgender female control groups (Junger et al., 2014; Kranz et al., 2014; Manzouri, Kosidou, & Savic, 2017; Rametti et al., 2011). Corresponding brain structure studies of transgender women (TrW; persons with male sex assigned at birth and female gender identity) seem more consistent, showing a cis gender ‘female’ pattern both with respect to fractional anisotropy (FA, reflecting white matter integrity), and cortical thickness (Cth) (Rametti et al., 2011; Zubiaurre-Elorza, Junque, Gomez-Gil, & Guillamon, 2014), although with some exceptions (Hahn et al., 2015; Luders et al., 2012; Spizzirri et al., 2018).

In summary, there has not been consistent evidence for systematical sex-divergent patterns among transgender populations. One conceptually important reason for the lack of clarity in GD is that most reports do not address the principal features of this condition – a strong perception of incongruence between one’s sense of self and one’s body, a discomfort with one’s own body, and a feeling of estrangement towards one’s physical sex (Cohen-Kettenis & Pfafflin, 2010). This incongruence may well be unrelated to cerebral sex dimorphism. Perception of one’s body in the perspective of self is believed to be mediated by a specific cerebral network, which includes the ventromedial prefrontal cortex (vmPFC) with pregenual anterior cingulate cortex (pACC), the temporo-parietal junction and fusiform body area (Feusner et al., 2017; Northoff & Panksepp, 2008).

One possibility is that, for individuals with GD, the typical physical traits of their sex assigned at birth are not incorporated into their own body image representation in the brain – and that this is associated with specific functional and structural cerebral signatures. Indeed, it was recently reported that TrM have a weaker response in the somatosensory cortex corresponding to the breast-thorax area of the homunculus than that in cisgender female controls (Case, Brang, Landazuri, Viswanathan, & Ramachandran, 2017), and that GD persons have higher degree of centrality in the parietal cortex (Lin et al., 2014), and smaller insular grey matter volume (GMV) in both untreated and oestrogen treated TrW in relation to cisgender control females (Spizzirri et al., 2018). Furthermore, in a series of studies by our group, we found weaker functional connections within the vmPFC and pACC of the default Mode Network (DMN) among TrM compared to cisgender controls (Burke et al., 2018; Feusner et al., 2017). In addition, Cth was greater in the vmPFC, cuneus and right parietal lobe in both TrM and TrW (Manzouri et al., 2017), in relation to cis-controls, irrespective of their sex and even when correcting for sexual orientation (Manzouri & Savic, 2018). Notably, there was no deviation from control data in the cortical surface area (SA), and, in other brain regions the Cth pattern followed the characteristics of the sex assigned at birth in both GD populations.

Together, these findings argue against a sex-atypical or sex-reversed cerebral sex dimorphism, reigniting the discussion about the possible neurobiology of GD. The findings also raise new questions. Given that transgender individuals report improved feelings of congruence between their gender identity and their body (Gooren et al., 2015) after cross-sex hormonal treatment, one may wonder to what extent (and also how) the described cerebral signatures of GD might change with sex hormone institution. Addressing this issue is important for understanding the dynamics of the described cerebral differences in relation to own body perception. To the best of our knowledge, only one study has thus far investigated this aspect (Burke et al., 2018). This previous study showed that testosterone treatment resulted in increased self-body congruence in TrM, and led to increased Cth in the insular cortex. This previous study was confined to a relatively small group of TrM (n = 24), hampering any general inference on the neurobiology of GD. We, therefore, carried out follow-up longitudinal investigations, with a larger number of TrM and the addition of a sufficiently large group of TrW. We related sex hormone effects on the brain to own-body perception as indexed from a so-called body morph test (Feusner et al., 2016, 2017). Furthermore, we distinguished between the overall effects on cerebral tissue and regional changes in Cth, which was, again, not previously elaborated. Because we previously detected differences between trans and cisgender populations in Cth but not SA (Manzouri et al., 2017; Manzouri & Savic, 2018; unpublished observations (I. S.), this report focuses on Cth.

Our primary hypothesis was that own-body-self congruence would increase with masculinizing treatment in TrM and with feminizing treatment in TrW. A second hypothesis was that such changes might have correlate in alterations of the Cth in the frontal and parietal cortex (Manzouri & Savic, 2018).

METHODS

Participants

This study was approved by the ethical committee of the Karolinska Institute (application number: Dnr 2011/281-31/4), and each participant provided a signed informed consent before entering the study. Forty TrM (mean age at the first scan, 24.9 ± 6.0 years), 24 TrW (mean age at the first scan, 26.6 ± 6.7 years) and 19 cisgender controls (mean age at the first scan, 29.7 ± 7.2 years; 11 females, 8 males) participated (Table 1).

Table 1.

Demographics, and ratings (mean ± SD)

	TrM (n = 40)a	TrW (n = 24)a	Controls (n = 19)a	p-values
				TrM versus TrW	TrM;TrW versus Controls
Age (years)	24.9 ± 6.0	26.6 ± 6.7	29.7 ± 7.2	0.10	0.004; 0.15
Years of education	13.5 ± 2.0	14.2 ± 2.3	16.3 ± 3.0	0.21	0.001; 0.02
Sexual orientation
Visit 1	4.3 ± 1.9	3.3 ± 1.6	0.3 ± 0.5	0.06	<0.001; <0.001
Visit 2	4.2 ± 1.6	3.1 ± 1.6	0.3±0.5	0.06	<0.001; <0.001
Rating of own body perception
Number of subjects	34	18
Visit 1	2.1 ± 1.0	2.0 ± 1.0
Visit 2	2.6 ± 1.0	2.5 ± 1.2

For TrM and TrW, Visit 1 and Visit 2 were conducted before treatment and during treatment, respectively.

^aThe body perception ratings were not obtained in all subjects; the n’s for these ratings are provided separately within the table.

^bPaired T-test showed increased self-ratings at Visit 2 compared to Visit 1; TrM: t(33) = 2.9, p = 0.006; TrW: t(17) = 2.0, p = 0.051.

^cKinsey scale, sexual orientation.

The transgender groups were recruited between 2013 and 2018 at ANOVA, the integrated center for Transgender Medicine, Karolinska University Hospital (Stockholm, Sweden). The diagnosis was based on consensus between specially trained endocrinologist, psychiatrist and psychologist as described previously (Manzouri et al., 2017). Adults aged 18–45 years who were diagnosed with transsexualism (according to the ICD-10 (World Health Organization, 1992)and referred for gender-confirming medical interventions were consecutively invited to participate in the study. Of the 80 transgender persons invited to participate both before and after institution of cross sex hormone treatment, 16 declined due to time constraints or the lack of a desire to participate in research. Except for four TrM, all the transgender participants described early onset of GD (prepuberty).

None of the TrM or TrW had received hormonal treatment at the time of the first scanning session, or gender confirmation surgery at the time of scanning sessions 1 and 2. Exclusion criteria for TrM and TrW were as follows: ongoing hormonal treatment for those scanned first time; known chromosomal or hormonal disorder; current psychiatric disorder, including body dysmorphic disorder (as confirmed by the Mini International Neuropsychiatric Interview (M.I.N.I.); Sheehan et al., 1998); neurological or other major medical disorder; and current use of psychopharmacological agents (antipsychotic or antiepileptic agents, lithium, benzodiazepines or opioid analgesics), with exception of antidepressants. However, none of the participant was treated with antidepressants.

We also excluded individuals with known autism spectrum disorder (ASD) (diagnosed before being referred to the team), as well as individuals who showed clinical signs of ASD when being assessed by the team. Exclusion criteria for the control group included GD, neurological or psychiatric disorder, substance use disorder, family history of psychiatric disorders and ongoing medication. Additionally, cisgender controls were excluded if they experienced any major disease, life trauma or initiated ongoing medication between the two scanning sessions. All female control participants had regular menstrual cycles and were investigated during the second week after menstruation. None of the female controls used oral contraceptives, none had known polycystic ovarian syndrome or other conditions associated with altered sex hormone levels. Notably, the major purpose to include the control population was to control for possible treatment unrelated changes (such as repetition of brain scans, ageing processes). The control population was gender mixed and limited in size, as we did not primarily opt to evaluate possible sex differences between trans- and cis-gender populations. This aspect of GD has been specifically addressed in our earlier studies (data from the first scan in 56 of the transgender participants were presented in Manzouri & Savic, 2018). We found that the regional pattern of cortical thickness (Cth) was largely in accordance with the sex assigned at birth, with the exception of frontal and parieto-occipital cortex for which both TrW and TrM showed greater values than both male and female controls, see also (Burke et al., 2017, 2018; Manzouri et al., 2017).

Procedures

Participants were scanned twice using magnetic resonance imaging (MRI): TrM were scanned before starting testosterone treatment (Visit 1) and again at least 6 months after the institution of testosterone (average inter-scan interval 14.8 ± 5.5 months; range, 6.1–26.9 months; TrW were scanned before starting feminizing treatment (Visit 1) and again at least 6 months after the institution of anti-androgen and oestrogen (average inter-scan interval, 14.9 ± 9.1 months; range, 7.0–37.9 months); and controls were scanned before and after a period of time without any intervention (average inter-scan interval, 34.8 ± 3.2 months; range, 7.3–38.7 months) (Visit 2).

Sexual orientation was assessed using the self-report Kinsey scale (Kinsey et al. 1948), which is a 7-point scale ranging from 0 (indicating a heterosexual orientation; exclusively gynephilic [in cis males], i.e., sexually attracted to females; and exclusively androphilic [in cis females], i.e., sexually attracted to males) to 6 (indicating a homosexual orientation; exclusively androphilic [in cis males]; and exclusively gynephilic [in cis females]). Prior to completing the questionnaire, individuals with GD were informed that the Kinsey scale was constructed for cisgender individuals and were asked to interpret “homosexual” and “heterosexual” in relation to their sex assigned at birth. Twenty-six of the 40 TrM identified as gynephilic (score, 4–6), eight identified as bisexual (score, 3) and six identified as androphilic (score, 0–2). Fourteen of the 24 TrW identified as androphilic (score, 4–6), nine identified as gynephilic (score, 0–2) and one identified as bisexual (score, 3).

All participants were tested for handedness according to the methods by Oldfield (1971). All were right handed.

Body perception test

To explore behavioural responses in own-body perception, we carried out a “body perception test”, in which participants, outside of the scanner, viewed photographs of their bodies morphed in 20% increments towards either cisgender male or cisgender female bodies and were asked to respond “to what degree is this picture you?” (for details of the procedure, see Feusner et al., 2016, 2017). This allowed us to obtain ratings that index own-body identification with images of the participant’s body morphed to appear more masculine or more feminine, as well as with one’s actual image (the 0% morph image). In the present report we only used the ratings of the un-morphed own-body image presented during 2 s (the 0% morph image) before and during treatment (Visit 1 and Visit 2), as an index of own-body identification with one’s actual image. Ratings were on scale of 1–4, with 1 corresponding to 0%–25% me, 2 corresponding to 25%–50% me, 3 corresponding to 50%–75% me and 4 corresponding to 75%–100% me. The images of different morph degree were presented in a randomized and balanced order (Visit 1 and Visit 2), and the ratings used for statistical analyses (paired t- test, p < 0.05, within group comparisons) were averaged over thirty 0% morph trials for each test occasion. Not all participants completed the body perception test, as this part of the experiment was added after MRI data collection had begun; 34 of the 40 TrM, and 18 of the 24 TrW performed the body perception test at both time points. 0% morph images were presented without repetition.

In TrM and TrW, serum hormone levels were assessed by routine clinical checkups and the assessments closest in time to the MRI sessions were used for the purpose of this study. Serum testosterone and serum estradiol were analysed at the Department of Clinical Chemistry at Karolinska University Hospital and determined by electrochemiluminescense immunoassay (Roche Diagnostic GmbH, Mannheim, Germany). No blood samples were collected in cisgender controls. Naturalistic testosterone treatment, per clinical protocol, consisted of intra- muscular injections of testosterone undecanoate (Nebido, Bayer); 1,000 mg every 12 weeks with the first two injections given 6 weeks apart). TrW received a combination of anti-androgen treatment and estradiol treatment, initiated in parallel. Anti-androgen treatment consisted of 50 mg cyproterone acetate (Androcur, Bayer) daily, or injections of Gnrh-analogue every 3rd month in the form of 11.25 mg of triptorelin (Pamorelin, Ipsen) or leuprorelin (Enanton Depot Dual, Orion Pharma). Estradiol treatment was given as tablets (Progynon, Bayer), gel (Divigel, Orion Pharma), patches (Estradot, Novartis) or intra-muscular injections of polyestradiolphosphate (Estradurin, Pharmanovia). The dosage and choice of treatment was decided with regard to the clinical situation and the patient’s preferences. For all treatment forms, dosage adjustment was performed at the clinical follow-up after 3 months of treatment, depending on the clinical response and blood levels. For the observed hormone levels in TrM and TrW before and during cross-sex hormone therapy (Table S1).

Data acquisition

MRI data were acquired on a 3-Tesla MRI scanner (Discovery 3T GE-MR750, General Electric, Milwaukee, Wisconsin), equipped with a 32-channel or 8-channel phased array receiving coil. 3D T1-weighted Spoiled Gradient Echo pulse sequence (SPGR) images were acquired with 1 mm3 isotropic voxel size (TE = 3.1 ms, TR = 7.9 ms, TI = 450 ms, FoV = 24 cm, 176 axial slices, flip angle 12°). We used an 8-channel coil because it provided better demarcations between white and grey matter in the occipital cortex for the purposes of the FreeSurfer analyses.

Calculation of cortical thickness

Cortical reconstruction and volumetric segmentation was performed using the FreeSurfer image analysis suite, version 6.0 (Fischl & Dale, 2000) (www.surfer.nmr.mgh.harvard.edu) to derive Cth and SA, as well as the total grey and white matter volume (WMV), and the total intracranial volume (ICV). Data from various subcortical structural volumes will be presented in a separate manuscript. The T1-weighted images were processed using the FreeSurfer Longitudinal Stream, in which an unbiased within-subject template (Reuter & Fischl, 2011) is created using robust, inverse consistent registration (Reuter et al., 2010). The resulting surface models were visually inspected for accuracy and manually edited for all participants.

Statistical analysis

Group differences in the percentage of change (pc1) in Cth from Visit 1 to Visit 2 (Cth Visit 2-Cth Visit 1/inter-scan interval)/Cth Visit 1 was evaluated using the Qdec toolbox of FreeSurfer. Interscan interval was calculated as Scan 2 at Visit 2 – Scan 1 at Visit 1. Within-group changes in Cth were evaluated using mri_glmfit in FreeSurfer. Bilateral maps were analysed at a vertex-wise level using a general linear model approach, after registration to the MNI standard space and smoothing with a Gaussian kernel of 15 mm FWHM and a minimum cluster size of 5.0 cm2. To control for multiple comparisons, cluster-wise p-value correction was performed using Monte Carlo simulation with 5,000 iterations (vertex-wise threshold of p < 0.05). Within- and between-group comparisons of the change in Cth were performed with and without correction for overall cerebral effects (change in the total grey and WMV); demeaned age and years of education were used as nuisance covariates for all analyses. In addition, a conjunctional analysis was performed to identify shared differences in Cth among both transgender groups vis á vis controls.

Differences in subject characteristics and brain volumetric data extracted with the FreeSurfer software, such as total grey and WMV, were evaluated in _SPSS version 24 (Armonk, NY: IBM Corp.) using general linear model with and without age as covariate of no interest (p < 0.05). In addition, Spearman partial correlation coefficients between the change in the 0% body morph ratings (Visit 2 – Visit 1) and changes in Cth (Visit 2 – Visit 1) were analysed in clusters showing significant between scan changes in both transgender groups (p<0.05).

RESULTS

Hormonal changes

Testosterone levels increased with treatment in all TrM, and were significantly higher at Visit 2 than at Visit 1, whereas oestrogen levels decreased and were significantly lower at Visit 2 compared to Visit 1 ([s-testosterone, t = 5.0, p < 0.001; s-oestrogen t = −11.9, p < 0.001]). Likewise, testosterone levels decreased and oestrogen levels increased in all TrW from Visit 1 to Visit 2 (s- testosterone, t = −10.2, p < 0.001; s-oestrogen t = −5.6, p < 0.001; Table S1). TrM were investigated, on average 10.4 ± 5.8 months after institution of testosterone treatment, and TrW 11.1 ± 7.8 months after institution of oestrogen+anti-androgen treatment.

Pre-treatment (Visit 1) brain differences between trans- and cis-gender groups

Cerebral volumes

Significant group differences in total cerebral volumes were observed before treatment, at Visit 1 (ICV: F(2,80) = 9.345, p < 0.001; GMV: F(2,80) = 9.873, p < 0.001; WMV: F(2,80) = 6.992, p < 0.001), but not in cerebrospinal fluid volume: F(2,80) = 2.558, p = 0.084), with significantly larger total GMV, and WMV, in TrW than in TrM, and larger total GMV, WMV, and ICV in TrW than in gender-mixed controls (Table 2). No other group differences in total cerebral volumes were detected.

Table 2.

Brain volumetric data (mean ± SD)

	TrM	TrW	Controls	p-values
				TrM versus TrW	TrM, TrW versus Controls
ICV (cm3) Visit 1	1521 ± 108	1684 ± 128	1573 ± 163	<0.001	0.55; 0.071
ICV (cm3) Visit 2	1532 ± 109	1679 ± 193	1571 ± 163	<0.001	0.27; 0.031
GMV (cm3)
Visit 1	688 ± 50	750 ± 56	693 ± 70	<0.001	0.732; 0.005
Visit 2	690 ± 51	734 ± 57	687 ± 70	0.009	0.882; 0.021
Visit 2–Visit 1				<0.001	0.009; 0.002
WMV (cm3)
Visit 1	442 ± 48	492 ± 55	457 ± 55	<0.001	0.311; 0.041
Visit 2	445 ± 49	488 ± 54	457 ± 56	0.002	0.431; 0.081
Visit 2–Visit 1				<0.001	0.041; 0.022
CSF (cm3)
Visit 1	115 ± 25	130 ± 27	124 ± 28	0.842	0.212; 0.500
Visit 2	115 ± 25	128 ± 28	126 ± 27	0.102	0.121; 0.801
Visit 2 – Visit 1				0.411	0.212; 0.421
Delta GMV+WMVa, cm3	5.5 ± 13.9	−20.8 ± 15.8	−5.5 ± 11.2	<0.001	0.004; 0.001

CSF, cerebrospinal fluid; GMV, grey matter volume; ICV, intracranial volume; WMV, white matter volume.

^aDelta GMV+WMV refers to between-scan difference in grey and white matter volume (Visit 2–Visit 1).

Cortical thickness

As in our previous studies, which used partly the same populations, we found significant pre-treatment (Visit 1) group differences in Cth, with thicker cortex in the frontal lobe, parts of occipital lobe and left temporo-parietal lobe, in both TrW and TrM than controls (Table 3; Figure S1).

Table 3.

Group differences in cortical thickness at Visit 1 (Scan 1)

Region	TrM–Controls			TrW-Controls
	Max – log 10(p)	Size (cm2)	Coordinates	Max – log 10(p)	Size (cm2)	Coordinates
Cortical thickness
L inferior parietal cortex	3.6	65.4	−38 −59 23
L lingual gyrus	3.0	15.3	−30 −42 −4
L middle temporal cortex				3.4	17.1	−59 −36 −7
R superior frontal cortex	2.8	14.0	10 48 11	2.6	7.1	23 43 7

Statistical threshold is p< 0.05, corrected for multiple comparisons (according to Monte Carlo permutations). The filter was 15 mm. The “Region” column describes the coverage of the respective cluster. The Talairach’s coordinates correspond to the peak value in the respective cluster. Demeaned age and level of education, and the total grey and white matter volume were used as covariates of no interest. R = right; L = left; Italics indicate a subsignificant cluster, which, in addition, was significant if using a 10-mm filter.

Effects of cross-sex hormone treatment

Cerebral volumes

At Visit 2, the total GMV differed between groups (F(2,80) = 5.041 p = 0.009), as did the WMV (F(2,80) = 5.005, p = 0.009), with significantly larger volumes in TrW than in TrM (Table 2). In TrM, the total GMV was numerically (although not significantly) higher in Visit 2 compared to Visit 1 (t(39) = 1.349, p = 0.185), while the total WMV significantly increased from Visit 1 to Visit 2 (t(39) = 3.581, p = 0.001). In controls, on the other hand, there was a significant decrease in GMV from Visit 1 to Visit 2 (t(18) = −2.514, p = 0.022), whereas no significant change was detected in WMV (t(18) = 0.299, p = 0.768). Finally, in TrW, both the total GMV (t(23) = −7.005, p < 0.001) and WMV (t(23) = −2.909, p = 0.008) decreased during treatment. There was a significant difference in the degree of change in GMV between TrM and TrW (t(62) = 42.318, p < 0.001), and between the trans- gender groups and controls [(TrM vs. controls: t(57) = 2.084, p = 0.042; TrW vs. controls: t(41) = −2.479, p = 0.017)]. Similarly, there was a significant difference in the degree of change in GMV between TrW and TrM (t(62) = 6.591, p < 0.001), between TrM and controls (t(57) = 2.726, p = 0.009), and between TrW and controls (t[41] = −3.273, p = 0.002; Table 2). Differences between the transgender and cis gender groups remained when adding age as covariate.

Within-group analyses of the changes in Cth

In TrM, the Cth increased significantly with testosterone treatment (Visit 2–Visit 1) in the left medial orbitofrontal-, cuneus- and parahippocampal-cortices, right anterior cingulate cortex, as well as in the left superior temporal cortex (Table S2a). In contrast, there were decreases in Cth with treatment in TrW (Visit 2–Visit 1), the precuneus and insular cortices, with greater decreases in the right hemisphere than in the left hemisphere (Table S2b; Figure S2a,b). Additionally, there were increases in Cth in the controls, located in the left occipital and frontal pars triangularis, and a decrease in Cth in the right mid-temporal cortex (Table S2c).

Between-group analyses of the changes in Cth

Group comparisons without correction for the change in total GMV and WMV revealed significant increases in Cth in TrM compared to those in controls, bilaterally in the insular and superior temporal cortex. In contrast, in TrW there were significantly greater Cth decreases compared to those in controls in large parts of the frontal, temporal and parietal lobes (Table 4; Figure 1a,b). However, when controlling for the change in total GMV + WMV, the group differences followed an entirely different pattern. There were Cth reductions in both transgender groups in relation to that in controls, confined to the left mesial prefrontal/superior frontal cortex and parietal cortex (partly covering the extra striatal body area) (Table 5; Figure 2). Moreover, in the conjunctional analysis, clusters showing a significant decrease in Cth in both transgender groups compared to that in controls were detected in the left inferior parietal cortex and mPFC, as well as a sub-significant decrease in the precuneus cortex (Table 6; Figure 3). Direct comparison between TrM and TrW in GMV+WMV corrected change in Cth showed one cluster (size 5 cm2) in the left insular-superior temporal lobe.

Table 4.

Between-scan (between visit) group differences in cortical thickness (Scan 2 – Scan 1), without adjustment for global cerebral changes

Region	TrM–controls			TrW-controls
	Max – log 10(p)	Size (cm2)	Coordinates	Max – log 10(p)	Size (cm2)	Coordinate
R superior frontal cortex				−5.5	82.3	12 43 15
R precuneus				−6.4	29.4	17 −58 31
R parahippocampal cortex				−4.8	45.5	35 4 −26
L Supramarginal cortex	−2.4	−21.0	−54 −26 −18
L insular+superior temporal cortex	−2.6	13.1	−51 −4 −7	−6.2	188.0	−56 −2 −8
L mesial frontal cortex	2.8	4.9	−6 10 −10
L lateral occipital cortex				−4.8	69.1	−24 −87 8

Statistical threshold is p< 0.05, corrected for multiple comparisons (according to Monte Carlo permutations). The filter was 15 mm. The ‘Region’ column describes the coverage of the respective cluster. The Talairach’s coordinates correspond to the peak value in the respective cluster. Demeaned age, and level of education were used as covariates of no interest.

Figure 1.

Between-group changes in Cth in TrM, TrW, compared to Controls. TrM – Controls. TrW– Controls. Only age and years of education were used as covariate, and there was no additional correction for changes in the WMV+GMV. Observe the differences compared to Figures 2 and 3, illustrating clusters when the calculation included correction for global cerebral changes. The contrasts were calculated at p < 0.05, corrected for multiple comparisons (Monte Carlo permutation). The projection of cerebral hemispheres (MR images of the FreeSurfer atlas) is standardized. Scale is logarithmic and shows –log 10(P), with cool colours indicating negative contrast, warm colours indicating positive contrast

Table 5.

Group differences for Visit 2 (Scan 2) – Visit 1 (Scan 1) in Cth, taking into account global cerebral changes

Cluster	TrM-Controls (positive values)
	Maximum vertex-wise −log10(p)	Cluster size (cm2)	Talairach’s Coordinates
Controls – TrM (negative values)
L Inferior parietal cortex	−3.8	65.1	−31 −73 37
L Superior frontal cortex	−2.9	5.7	−10 49 10
Controls – TrW (negative values)
R superior frontal cortex	−7.4	19.1	9 43 20
R insular cortex	−2.4	9.2	35 −18 13

Statistical threshold is p< 0.05, corrected for multiple comparisons (according to Monte Carlo permutations). Demeaned age and demeaned total grey matter volume were as nuisance covariates. The filter was 15 mm. The Talairach’s coordinates indicate peak values in clusters showing significant group between differences in the between Scan 2 and Scan 1 change in Cth, using age, education and the total Grey matter+White matter volume as covariate of no interest; the ‘Region’ column describes the coverage of the respective cluster. Lowering the significance threshold revealed that reductions were bilateral in both TrW and TrM, R = right; L = left.

Figure 2.

Between-group differences in cortical thickness, with correction for global cerebral changes. The contrasts were calculated at p < 0.05, corrected for multiple comparisons (Monte Carlo permutation), using age, education, and, in addition, the total WMV+GMV, as the covariate of no interest. The projection of cerebral hemispheres (MR images of the FreeSurfer atlas) is standardized. Scale is logarithmic and shows –log 10(P), with cool colours indicating negative contrast, warm colours indicating positive contrast. TrW = transgender women; TrM = transgender men

Table 6.

Shared between-scan changes in cortical thickness (Visit 2–Visit 1), obtained in a conjunction analysis of TrM versus Controls and TrW versus Controls

Region	Max –log 10(p)	Size (cm2)	Coordinates
L superior frontal cortex	−2.7	13.1	−6 48 11
L inferior parietal cortex*	−2.6	54.2	−41 −74 32

Statistical threshold is p< 0.05, corrected for multiple comparisons (according to Monte Carlo permutations). The filter was 15 mm. The Talairach’s coordinates indicate peak values for significant conjunctional clusters indicating shared decreases in Cth in TrW and TrM compared to controls (Scan 2 – Scan 1), using age, education and the total Grey + White matter volume as covariates of no interest; the ‘Region’ column describes the coverage of the respective cluster. R = right; L = left. *Covers part of the precuneus.

Figure 3.

Cth, conjunctional analysis. The figure illustrates clusters in which both TrM and TrW reduced their Cth with treatment Visit 2 – Visit 1 (Scan 2 – Scan 1) in relation to the Visit 2 – Visit 1 (Scan 2 – Scan 1) differences in controls. The contrasts were calculated at p < 0.05, corrected for multiple comparisons (Monte Carlo permutation), using age and education, and, in addition, the total WMV + GMV, as the covariate of no interest. The projection of cerebral hemispheres (MR images of the FreeSurfer atlas) is standardized. The blue colour indicates negative contrast, illustrating reductions in relative Cth in both transgender groups. There were no areas with increases in relative Cth.

Changes in own-body perception

We used both front and sagittal view of the 0% morphed images, in a randomized order. 12 transgender participants did not provide response to 0% morphed images in both front and sagittal view for both scan 1 and 2. Own-body recognition ratings increased between the two visits in both transgender groups (TrM: t[33] = 2.9, p = 0.006; TrW: t[18] = 2.0, p = 0.051). Furthermore, there was a significant negative partial correlation (controlling for age, education and change in GMV + WMV) between the change in BM ratings and the change in Cth of the left inferior parietal cluster obtained from the conjunction analysis among all transgender participants (rho = −0.372), p = 0.022, df = 36; Figure 4). Clusters in conjunctional analysis (see Figure 3) were defined as change in Cth (TrW- controls) and (TrM-controls) including global correction. When the two transgender groups were considered separately, the correlations reached significance only in TrW (rho = −0.710, p = 0.0.003, df = 13), but not in TrM (rho = −0.121, p = 0.581, df = 20). The change in the Cth of the left mesial prefrontal/superior frontal cluster from the conjunction analysis was not significantly correlated with the change in 0% body morph ratings among all transgender participants (rho = 0.131, p = 0.438, df = 35), nor within TrM (rho = 0.167, p = 0.471, df = 19) and TrW groups (rho = 0.112, p = 0.690, df = 13). In Figure 4, we only present correlation between changes in Cth and ratings to 0% morph images of participants who gave response to all 0% images.

Figure 4.

Partial correlation between change in the left inferior parietal cortical thickness (Cth) and change in the own body perception ratings. The change in the Cth between Scan 2 and Scan 1 corrected for time difference between the scans, and after regressing out age, education, and grey/white volumes and expressed as unstandardized residual is plotted (y-axis) against the change in own body perception rating (x-axis). Significant negative partial correlation (rho = −0.351, p = 0.021, df = 35) was detected when including data from both TrM (filled rings) and TrW (open rings)

DISCUSSION

The present study was conducted to investigate how cross-sex hormones affect the brain in transgender persons, with an interest in whether these effects were widespread or localized, and, in particular, whether they affect regions which in our previous studies showed difference from both male and female cis-gender controls. We did not specifically compare sex and gender interaction of cross sex hormone effects, as this would require matching treatment in controls. Thus, the control population was scanned twice only to avoid influence of scanner related factors (e.g., drift, less novelty during the second scan), and included both males and females. We found that oestrogen+anti-androgen treatment led to widespread decreases in Cth in TrW, similar to what has been reported with oestrogen treatment in regard to Cth and GMV in postmenopausal women (Zhang et al., 2016), and congruent with the few hitherto published studies of oestrogen effects on Cth in TrW (Mueller, Landre, Wierckx, & T’Sjoen, 2017; Seiger et al., 2016; Zubiaurre-Elorza et al., 2014). In contrast, testosterone treatment led to increases in Cth, and these increases were detected in testosterone-receptor abundant regions, the insular and superior temporal cortices (Table S2), and, to a lesser extent, also in the frontal cortex. This accords with previous reports in TrM participants (Burke et al., 2017; Zubiaurre-Elorza et al., 2014). The generated data are also in accord with previous reports about structural volumes in relation to testosterone treatment in TrM (Mueller et al., 2017; Seiger et al., 2016; Zubiaurre-Elorza et al., 2014). However, several of these previous, as well as the present study, also showed a general increase of the GMV and WMV due to testosterone treatment, and corresponding decreases, especially in the GMV, with anti- androgen and oestrogen treatment. The present report adds to the previous literature by paying attention to these global changes in cortical and subcortical volumes, motivating a recalculation of the group comparisons using the change in the total GMV and WMV as a nuisance covariate, thus, employing a correction for global changes to reveal possible selective changes. Interestingly, this approach modified the results considerably. First of all, in both transgender groups, we then detected a reduction in the Cth. Secondly, the reduction was restricted to those regions that showed significantly greater Cth at Visit 1 in both TrM and TrW compared to that in controls, without a significant difference between the two transgender populations – the right mPFC, left lateral temporo-occipital and parietal cortices. Thus, the initially detected divergence from controls along the cerebral midline in both TrW and TrM, disappeared with cross-group difference between the scans.

Between scan reduction in GMV was detected also in controls, albeit less pronounced, and was, possibly, effect of ageing. No changes of the MR scanner were done during the course of the study and the subjects were investigated during the same period, thus the changes in controls are unlikely to be linked to a technical factor.

To summarize, the present study demonstrated general cross-sex hormonal effects that appear congruent with the previously described effects of sex hormones on cerebral grey and WMV, and Cth among testosterone and oestrogen treated cisgender study groups. In addition, and of note, it also showed cross-sex hormonal effects that are more specific for GD, located along the cerebral midline, and similar in both transgender groups.

How to interpret the region-specific changes in Cth in both transgender populations?

One possibility is that we are dealing with features that are characteristic of GD, are part of the neurobiology of GD, and linked to its cause. Hypothetically, the greater Cth along the midline could reflect a difference in cortical maturation from that associated with the sex assigned at birth, which may be linked to a different own-body image in the networks processing the perception of self in relation to one’s own body. Indeed, the anatomical locations of the regions in these networks correspond well with the locations of thicker cortex among the transgender study groups. Furthermore, we found that the degree to which one’s own, unmorphed, body image was perceived as ‘self’ increased significantly with cross-hormonal treatment. The GD participants also verbally indicated that they felt more congruent with their bodies after the institution of cross-sex hormones. This could be due to a direct hormonal effect on the brain. It is also possible that sex hormone-induced changes of the body, towards the perceived gender, led to a more congruent perception of the self with one’s own body and reduced rumination and suffering about one’s own body, resulting in a thinning of the cortex in networks mediating self-body perception (Greenough & Volkmar, 1973). Whereas oestrogen effects on spine density and length are reported foremost in the hippocampus, the prefrontal cortex (Hao et al., 2007), the primary sensory-motor cortex and cerebellum (Chen et al., 2009), studies of human development suggest that testosterone contributes to a global increase of the WMV (Perrin et al., 2008), increase in the occipital and superior temporal lobe GMV and thinning of the parietal cortex (Bramen et al., 2012). These regionally differentiated effects of oestrogen and testosterone could provide an explanation to the presently observed fontal and occipital reduction of Cth in both transgender population when corrected for the respective cross hormone related changes in total GMV+WMV.

The current study design does not allow us to determine the causal mechanisms. One way to address this issue in the future would be to investigate Cth in prepubertal children who express GD and determine whether Cth is different from that in matched control children, before the rumination process has become more established. Another approach would be to correlate Cth data and the functional and structural connections within the DMN to the individual responses in the body morph test in large populations of TrM and TrW, and evaluate the possible dynamics of the response pattern in relation to cross-sex hormonal treatment. Indeed, our recently published preliminary data on the effects of testosterone in TrM suggest specific effects on the DMN, possibly implying that this network is causally involved in GD (Burke et al., 2017).

Finally, it is worth commenting that not all subjects gave higher scores in own-body perception with cross hormonal treatment (Figure 4). In six of the nine persons with a lower ‘self’ rating after treatment, there was also an increase in Cth in the left inferior parietal cluster, which could support the notion that changes in cortical thickness in this particular region may be related to own-body perception, an issue deserving further attention.

Limitations

The present data need to be viewed in the scope of their limitations. Although our study adds to the literature by the inclusion of longitudinal data for controls and has a larger total number of participants than those in previous studies, the investigated populations should optimally be larger, allowing greater power in analyses of the correlations between changes in 0% own body morph scores and changes in Cth. For the same reason we could not evaluate a possible impact of sexual orientation (Manzouri & Savic, 2018); However, based on our previous studies of sexual orientation and GD (Burke et al., 2017; Manzouri & Savic, 2018) we have no reason to assume that degree of homo/heterosexuality influenced the observed fronto-inferotemporal parietal cortical changes in our transgender participants. Furthermore, Cth in these particular regions was in our previous study (Manzouri & Savic, 2018) different among transgender versus cis-gender participants independently of their sexual orientation, and also of the sex of controls. Of particular interest that necessitates a larger study would be the determination of whether a combination of the body morph-test and measurements of Cth can help identify poor responders to cross-sex hormonal treatment. Perhaps, it might be possible to predict treatment responses from the own body perception in the context of self (body morph test scores) and Cth patterns already present at the first visit, before the institution of treatment.

Additionally, we did not carry out correlational analyses between sex hormone levels and the degree of change in Cth, as the effects of sex hormones on the brain are complex and plasma levels do not linearly reflect the specific effects on the brain receptors (e.g., the effect of same amount of testosterone on the androgen receptor differs, depending on the gene polymorphism; Raznahan et al., 2010). Furthermore, it is important to consider that hormone effects in TrW are not only due to the increase of estrogen levels, but are also related to a reduction of testosterone levels, and possibly also affected by the gonadotropin analogue. A post hoc analysis of individual changes in Cth in the two regions of interest detected in conjunctional analysis (Figure 3) showed no significant correlations between testosterone levels at Visit 2 in TrM, nor between estrogen levels at Visit 2, and relative changes in the Cth.

The time between Visit 1 and Visit 2 varied among participants, even though our initial intention was to have all participants investigated within 6–8 months of treatment initiation. This was not possible due to logistical difficulties. However, we managed to maintain a lower time limit of 6 months and avoid significant group differences in the between-scan interval. In addition, all of the calculations of group differences in Cth were corrected for the between-scan interval (see the calculation of pc1 in the FreeSurfer longitudinal pipeline within FSL).

In the present work we focused on Ch, as SA in our previous studies was not found to differ between transgender and cisgender study groups (Manzouri et al., 2017; Manzouri & Savic, 2018; unpublished data (I. S.).

Finally, it should be emphasized that the control group was gender mixed and the study design does not allow for specific comparisons between cis and transgender populations of the same at birth assigned sex with respect to sex hormone treatment. Such comparisons will require larger control groups. This particular issue was, however, beyond the scope of the present study, which primarily focused on the cross-sex hormone effects in the respective transgender population. Specific sex and gender identity interactions with respect to Ch have been reported in our previous study of GD, including 40 male and 40 female cis gender controls investigated in identical manner (Manzouri & Savic, 2018), however, without a longitudinal design. This previous study allowed us to discuss how the pattern of Ch changes related to the described sex differences in Cth among these controls and deduct how oestrogen and testosterone treatment was linked to GD specific effects on the fronto-parietal Cth.

CONCLUSIONS

In conclusion, we reproduced our previous observation of increased Cth in the mPFC, pACC and precuneus in TrW, as well as in TrM. The novel finding of specific reductions after cross-sex hormonal treatment in these same regions, and, independently of the sex assigned at birth, is rather intriguing and accords with our hypothesis about a specific link between GD, the neuronal networks along the cerebral midline processing the perception of one’s own body in the context of self.

Abbreviations:

BM

bodymorph

CSF

cerebrospinal fluid

Cth

cortical thickness

DMN

default mode network

FA

fractional anisotropy

FBA

fusiform body area

GD

gender dysphoria

GMV

grey matter volume

ICV

intracranial volume

MRI

magnetic resonance imaging

pACC

pregenual anterior cingulate cortex

SA

surface area

TPJ

temporo-parietal junction

TrM

transgender men

TrW

transgender women

vmPFC

ventromedial prefrontal cortex

WMV

white matter volume