Impact of gender-affirming hormone therapy on the endocrine axis and male germ cells

Abstract

Abstract

Gender-affirming hormone therapy (GAHT) facilitates the desired phenotypic changes consistent with an individual’s gender identity. However, inter-individual variability in treatment effects has been reported at the endocrine and testicular tissue levels. To address the hypothesis that treatment effects are not homogeneous, we selected 24 persons from a cohort of 453 persons assigned male at birth, based on a similar dose of cyproterone acetate (Androcur^®, 10 or 12.5 mg/day). Principal component analysis and hierarchical clustering on principal components were performed considering clinical parameters and identified four distinct clusters. Cluster 1 (n = 3) was characterized by the presence of tubules with round or elongated spermatids, the highest numbers of reserve spermatogonia (A_dark) and cells positive for UTF1, PIWIL4, NANOS3 and KIT. Persons in Cluster 2 (n = 4) showed a high percentage of tubules with spermatogonia as the most advanced germ cell type and high testosterone levels that were not suppressed to the desired female values. Persons in Cluster 3 (n = 10) had low testosterone but elevated AMH levels and suppressed spermatogenesis to the spermatogonial level. Finally, persons in Cluster 4 (n = 7) had the highest number of Sertoli cell-only tubules and tubular shadows, as well as the highest person age. Despite homogeneous inclusion criteria, we observed heterogeneous treatment effects. Consultation regarding contraceptives, as well as fertility preservation, therefore has to be considered in counselling, ideally prior to initiation of GAHT.

Lay summary

Hormonal therapy can help to align one’s physical appearance with gender identity. However, hormonal therapy does not work the same way for everyone. The effects on testis function, including hormone and sperm production, can vary considerably. In this study, we characterized persons who received comparable therapies to transition to their desired female gender. We found four subgroups: in Group 1, sperm was still produced by testicular tissues. In Group 2, only precursors of sperm were present, but male, rather than the desired female, hormone levels were observed. In Group 3, only precursors of sperm were present, and desired hormone levels were reached. Finally, Group 4 largely showed loss of germ cells and included the oldest people in the study. These results are relevant for counselling, as individuals need to be informed about the different effects that hormonal therapy can have with regard to sperm production and thereby fertility.

Introduction

Increasing numbers of transgender persons are seeking gender-affirming treatment (Goodman et al. 2019). Persons who wish to transition from their registered gender at birth to their desired gender are prescribed gender-affirming hormone therapy (GAHT) as first-line treatment. Specifically, to achieve physical feminization, persons assigned male at birth (AMAB) undergoing GAHT are prescribed a combination of anti-androgens and oestrogens or gonadotropin-releasing hormone (GnRH) analogues in combination with oestrogens (Hembree et al. 2017, Coleman et al. 2022). At the endocrine level, GAHT should lead to the desired decrease in testosterone to adult female values (<1.91 nmol/L) (De Sutter 2001) and to an increase in oestradiol. At the testicular level, endocrine changes lead to spermatogenic regression, thereby impacting fertility (Schneider et al. 2015).

In Germany, gender-affirming surgery (GAS), including orchiectomy, can be performed at least 12 months after initiation of GAHT, if desired (Hembree et al. 2017, Coleman et al. 2022). Based on tissue analyses of these testes obtained at the time of GAS, impacts with regard to spermatogenesis, spermatogonial numbers and composition of the spermatogonial compartment have been reported (Schneider et al. 2015, Dabel et al. 2023). Specifically, at the time of GAS, the level of spermatogenic regression was highly heterogeneous. Some tissues were completely devoid of germ cells (Sertoli cell-only phenotype or even tubular shadows), whereas other tissues in 11–24% of respective cohorts still contained tubules with complete spermatogenesis (Schneider et al. 2015, Jindarak et al. 2018, Matoso et al. 2018). In the majority of testicular tissues, however, spermatogenesis was suppressed, and tissues contained spermatogonia as the most advanced germ cell type (Jindarak et al. 2018, Matoso et al. 2018, Vereecke et al. 2020, Delgouffe et al. 2025). In those latter tissues, partial dedifferentiation of Sertoli cells has been reported, indicated by increased expression of AMH (Delgouffe et al. 2025). Interestingly, quantitative studies suggest not only regression of spermatogenesis but also a reduction in spermatogonial numbers compared with cis-gender controls (De Nie et al. 2022, Dabel et al. 2023). These studies highlight the inter-individual differences in the effects of GAHT on testicular tissues.

Moreover, there is also a suspected impact on the composition of the spermatogonial compartment, which is traditionally classified into A_dark (reserve) and A_pale (progenitor) spermatogonia (Clermont 1963). Based on single-cell (sc) RNA profiling, a higher resolution of the spermatogonial compartment has been achieved, which subclassifies the undifferentiated spermatogonia (UTF1⁺) into PIWIL4⁺ (State 0), FGFR3⁺ (State 0A), NANOS2⁺ (State 0B) and NANOS3⁺/GFRA1⁺ cells (State 1). Differentiating spermatogonia are characterized by expression of KIT⁺ (State 2) (Guo et al. 2018, Sohni et al. 2019, Di Persio et al. 2021). ScRNA analyses of testes from two persons AMAB undergoing GAHT showed a low proportion of differentiating spermatogonia (KIT⁺), spermatocytes and spermatids but a potential enrichment in the proportion of undifferentiated UTF1⁺ spermatogonia (Guo et al. 2020). The impact of GAHT on the spermatogonial compartment, based on morphological characteristics and marker expression, remains hitherto unknown.

Despite increased research efforts, the inter-individual differences regarding the effects of GAHT remain poorly understood. In addition to regression of spermatogenesis and reduction of spermatogonial numbers, there may also be alterations in the spermatogonial compartment’s cellular composition including Leydig cells, peritubular cells, Sertoli cells and spermatogonia. This warrants further analysis, as these alterations impact the potential long-term recovery of spermatogenesis after cessation of GAHT and thereby fertility.

To address the hypothesis that persons AMAB undergoing GAHT show inter-individual variability in their response to GAHT, this study set out to scrutinize persons (n = 24) exposed to comparable GAHT regimens. Specifically, this study employed cluster analysis using a combination of andrological parameters, such as endocrine levels, histological parameters and quantitative parameters of spermatogonial subpopulations.

Materials and methods

Ethical approval

Testicular tissues of persons AMAB undergoing GAHT were collected by gonadectomy as part of their GAS, performed in one of three cooperating clinics (Osnabrück Hospital (clinic A); University Hospital Essen (clinic B); Asklepios Westklinikum Hamburg (clinic C); Supplementary Table S1 (see section on Supplementary materials given at the end of the article)). For the collection and use of these testicular tissue samples, ethical approval was obtained from the Ethical Committee of the Medical Board Münster (Ärztekammer Westfalen-Lippe; no. 2012-555-f-S) and the Medical Board Hamburg (Ärztekammer Hamburg; no. MC-131/13). Testicular biopsies from male infertility patients with obstructive azoospermia (full spermatogenesis but no sperm in the ejaculate due to congenital bilateral absence of the vas deferens (CBAVD) or vaso vasostomy) served as controls. Exclusion criteria for the controls were prior gonadotoxic or hormonal treatment, a history of cryptorchidism or oncological disease and genetic diseases affecting spermatogenesis. These research biopsies were obtained during microsurgical testicular sperm extraction, which was carried out at the Department of Andrology at the Centre for Reproductive Medicine and Andrology (CeRA), University Hospital Münster. Collection of biopsies for research was approved by the Ethics Committee of the Medical Faculty of Münster and the State Medical Board (No. 2008-090-f-S). Written informed consent was provided by all persons. To exclude the impact of morphological changes at the testicular tissue level due to age, we selected samples from men with a mean age of 34.14 years (range: 27–42 years) to match the cohort of persons AMAB undergoing GAHT. Moreover, samples were selected based on a Bergmann–Kliesch score (Bergmann & Kliesch 2009) of at least 8, indicating that >74% of analysed seminiferous tubules contained elongated spermatids.

Selection of the cohort of persons AMAB undergoing GAHT

A group of persons AMAB undergoing GAHT (n = 24) was selected from a cohort of 453 persons based on age at the time of GAS (mean age: 28.9 years, range: 17–40 years) and a comparable dose of cyproterone acetate (Androcur^®, 10 or 12.5 mg/day), because this is the suggested treatment pathway according to the German guideline for treating transgender patients. Oestrogen intake was reported by 22 persons AMAB undergoing GAHT (mean: 4.77 mg/day, range: 1.43–10 mg/day). Treatment duration ranged from 10 to 65 months, with a mean of 26.7 months (n = 24) (Supplementary Table S1). Slight differences persisted between the collaborating clinics with regard to hormone therapy discontinuation strategies prior to gender-affirming surgery. Persons AMAB undergoing GAHT treated at the Asklepios Westklinikum Hamburg (clinic C, n = 4) continued GAHT until the day of GAS. At the University Hospital Essen (clinic B, n = 16), hormone intake was discontinued 2 weeks prior to surgery, and at the Osnabrück Hospital (clinic A, n = 4), it was discontinued 4–6 weeks prior to surgery Supplementary Table S1.

Determination of serum and intratesticular hormones

Venous blood samples (heparinized blood and ethylenediaminetetraacetic acid (EDTA) blood) were taken on the day of GAS. Serum hormone values for LH, FSH, testosterone, free testosterone, oestradiol, prolactin, SHBG, AMH and inhibin B were measured at the CeRA. In-house male and female reference values are provided in Supplementary Table S1. Therefore, serum and blood cells were separated by centrifugation. Following previously published immunofluorometric assays (Ragab et al. 2018), serum concentrations of LH (Cat No. 02P40-25), FSH (Cat No. 07 K75-25), testosterone (Cat No. 02P13-28), oestradiol (Cat No. 07 K72-25), prolactin (Cat No. 07 K76-25) and SHBG (Cat No. 08 K26-20) were analysed with the Abbott Architect i1000SR automated analyser (Abbott Diagnostics). For the calculation of free testosterone, the Vermeulen method was used (Vermeulen et al. 1999, Schneider et al. 2015). Briefly, free testosterone was calculated considering total testosterone and the fraction bound to SHBG using the published formula (https://www.issam.ch/freetesto.html).

Commercially available enzyme-linked immunosorbent assays (ELISAs) were used for measurement of serum AMH (AMH GEN II ELISA, Cat No. A79765, Beckmann Coulter, USA) and serum inhibin B (inhibin B GEN II ELISA, Cat No. A81303, Beckmann Coulter, USA). The intra-assay coefficients of variation were 3.0% (AMH) and 4.3% (inhibin B) (Schneider et al. 2021).

The measurement of intratesticular AMH and inhibin B was performed using a modified protocol adapted from (Simoni et al. 1994). A 50 mg sample of testicular tissue was placed in 250 μL of lysis buffer (sucrose 0.25 M, sodium chloride 0.14 M) and homogenized for at least 60 s by ultrasound on ice. After 1 h of centrifugation (25,000×, Hettich centrifuge EBA 15, Hettich GmbH, Tuttlingen, Germany), intratesticular AMH and inhibin B levels were measured in the supernatant. The measurements were performed using the same ELISAs as for serum hormones. Values were calculated as the mean of three different dilutions (1:5, 1:10, 1:20). The intra-assay coefficients of variation were 2.7% (AMH) and 6.2% (inhibin B). The inter-assay coefficients of variation were 3.3% (AMH) and 3.7% (inhibin B) (Schneider et al. 2021). Results were calculated per milligram of testis.

Collection and preparation of human testicular tissues and slide preparation

Following GAS at one of the three cooperating clinics, testes of persons AMAB who underwent GAHT were incised, placed in Dulbecco’s modified Eagle medium (Life Technologies, USA) and transported to the CeRA at 4°C. During tissue preparation, the epididymis was removed, and testes were weighed prior to removal of the capsule to determine testicular weight, reflecting testicular volume (Behre et al. 1989). Testicular tissue fragments from both persons AMAB and controls were fixed in Bouin’s solution and 4% PFA overnight for immunohistochemical and immunofluorescence staining, respectively. Paraffin-embedded testicular tissue fragments were sectioned to 5 μm using a Leica SM 2000R microtome.

Periodic acid–Schiff staining and histological evaluation of human testicular tissues

For routine clinical diagnostics, the histological status is determined by evaluation of periodic acid–Schiff-stained tissue sections from two independent biopsies per testis according to the Bergmann–Kliesch (BK) score (Bergmann & Kliesch 2009). The score (0–10) describes the percentage of tubules containing elongated spermatids. A score from 8 to 10 indicates intact spermatogenesis, with at least 75% of tubules containing elongated spermatids.

For histological evaluation of testicular tissues from persons AMAB who underwent GAHT, Bouin’s-fixed testicular tissue sections were deparaffinized, dehydrated and subjected to periodic acid–Schiff (PAS) staining was performed following previously published protocols (Brinkworth et al. 1995, Di Persio et al. 2021). Within PAS-stained cross-sections, seminiferous tubules were classified according to the most advanced germ cell type (elongated spermatids, round spermatids, spermatocytes, spermatogonia) or the absence of germ cells (Sertoli cell-only, tubular shadows). The evaluation of germ cells was based on their morphology. The data for this cohort of persons AMAB who underwent GAHT were previously published (Dabel et al. 2023) with the number of evaluated tubules per section ranging from 65 to 559 tubules and an average of 219 evaluated tubules per tissue section.

The control cohort consisted of testicular tissues from 14 adult men. Of these 14 control tissues, n = 8 were fixed in Bouin’s solution for subsequent immunohistochemical analyses and n = 6 were fixed in 4% paraformaldehyde (PFA) for immunofluorescence analyses, respectively.

Immunohistochemical staining of testicular tissues

Immunohistochemical staining was performed using the antibodies listed in Supplementary Table S2 and following published protocols (Albert et al. 2012, Di Persio et al. 2021). Species-specific IgG antibodies served as negative controls. In brief, primary antibodies were incubated overnight. The sections were incubated with corresponding biotin-linked secondary antibodies for 1 h and afterwards with streptavidin–horseradish peroxidase (streptavidin–peroxidase from Streptomyces avidinii, Cat No. S5512, Sigma-Aldrich, Munich, DE) for 45 min to connect streptavidin to biotin. For detection of peroxidase activity, 3,3′ diaminobendizine tetrahydrochloride (DAB; Cat No. A0596.0001, Applichem, Germany) solution was added, and sections were counterstained with Mayer’s haematoxylin (Cat No. 1.092.490.500, Sigma-Aldrich). The stained sections were scanned with the PreciPoint M8 Microscope and Scanner (PreciPoint GmbH, Germany) at 60× magnification and were evaluated using ViewPoint software (Cat No. 1.0.0.9442, PreciPoint GmbH, Germany).

Quantification of spermatogonia and spermatogonial subpopulations

The number of immune-positive cells per round seminiferous tubule was determined for each marker (MAGEA4, UTF1, PIWIL4, FGFR3, NANOS3) by analysing tissues from persons AMAB undergoing GAHT (n = 24) and Bouin’s fixed controls (n = 8). At least 25 round seminiferous tubules per tissue were evaluated. Tubules were classified as round if their maximum diameter ratio was between 1 and 1.5 and if they were enclosed by peritubular cells by >80%. For assessment of alterations among morphologically defined A_dark spermatogonia, the proportion of these reserve stem cells was quantified among 100 randomly selected MAGEA4⁺ cells per tissue. For the controls, two independent sections (>15 μm distance) were evaluated.

Immunofluorescence staining of testicular tissue sections for Sertoli cell maturity and quantification of spermatogonial subtypes

Immunofluorescence staining was performed using PFA-fixed slides and primary antibodies against EGR3, SOX9, KIT, MKI67, MAGEA4 and GFRA1, analysing tissues from persons AMAB undergoing GAHT (n = 24) and controls (n = 6). Again, species-specific IgG antibodies served as negative controls (Supplementary Table S2). Following incubation with the primary antibodies overnight, tissue sections were counterstained with corresponding secondary antibodies and placed in a dark chamber for 60 min. For nuclear staining, 4′,6-diamidin-2-phenylindol (DAPI) was added to the sections for 10 min prior to application of Vectashield. For image acquisition, the Evident V120 microscope and scanner was used and quantification was performed using QuPath software (Bankhead et al. 2017). To determine the proliferation rate of State 1 (GFRA1⁺) and State 2 (KIT1⁺) spermatogonia, the following marker combinations were evaluated: GFRA1/MKI67/MAGEA4 and KIT/MKI67/MAGEA4. For this purpose, 400 MAGEA4⁺ cells per sample were quantified, and the proportion of KIT⁺ or GFRA1⁺ cells was determined. Finally, MKI67⁺ cells among KIT⁺ or GFRA1⁺ cells were assessed.

Principal component analysis and hierarchical clustering

Principal component analysis (PCA) and clustering were performed as previously described (Schülke et al. 2024) using R 4.3.1 and specifically the R packages FactoMineR (Version 2.9) (Lê et al. 2008) and factoextra (version 1.0.7). Briefly, the function PCA() from the FactoMineR package (version 2.9) was used to reduce the dimensionality of the data into a few continuous variables. Prior to the analysis, the data were scaled to avoid dominance by variables with large measurement units. Five dimensions were retained in the output. The results of the PCA were visualized using the factoextra (version 1.0.7) package. The FactoMineR (version 2.9) package, and specifically the function HCPC(), was used to perform hierarchical clustering by applying Ward’s criterion to the selected principal components. The partitioning into different clusters was performed in two steps: initially, we cut the hierarchical tree, and then, the K-means clustering was used to improve the initial partition. The function fviz_cluster() in the factoextra (version 1.0.7) package was used to visualize individual clusters. The stability of the clusters was tested using the clusterboot() function in the fpc package (version 2.2–14), which assesses cluster stability by resampling (n = 10,000) and computes the Jaccard similarity index (Hennig 2007).

Clinical data used for the analyses included age of persons AMAB undergoing GAHT at the time of gender-affirming surgery, GAHT dosage (of cyproterone acetate and oestrogens), treatment duration, testicular weight, serum levels of LH, FSH, prolactin, testosterone, free testosterone, SHBG, oestradiol, inhibin B and AMH, in addition to intratesticular levels of inhibin B and AMH. Furthermore, histological information listed in Supplementary Table S1 was taken into consideration.

Data representation

All quantitative data are shown as box plots. Box plots elements are defined as follows: centre line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; and points, outliers. The control values included in the box plots were used to facilitate a comparison of different spermatogonial subpopulations in the testes of individuals with obstructive azoospermia. These values were not part of the cluster analysis; no additional statistical comparisons were made between the controls and the different AMAB groups undergoing GAHT.

Results

Clustering of persons AMAB undergoing GAHT based on clinical parameters

To assess whether there are heterogeneous responses of persons AMAB to GAHT, we selected 24 persons with comparable CPA treatment regimens (10–12.5 mg/day) plus oestrogens and aged from 17 to 40 years from a cohort of 453 persons AMAB. We performed PCA followed by hierarchical clustering on principal components (HCPCs) based on clinical and histological parameters (Fig. 1A, Supplementary Fig. S1A, Supplementary Table S1). The PCA showed a clear separation of persons on the first (PC1) and second (PC2) principal components, which accounted for 27.4 and 12.5% of total variation, respectively. The HCPC method identified four distinct clusters: Cluster 1, n = 3; Cluster 2, n = 4; Cluster 3, n = 10; Cluster 4, n = 7 (Fig. 1B, Supplementary Fig. S1). The main driving factors are related to testicular tissue composition, spermatogonial subpopulations and endocrine parameters (Supplementary Table S3). There was no specific assignment of persons to clusters based on the clinic of origin (Supplementary Fig. S1B). The stability of the clusters was measured using the Jaccard stability index (Hennig 2007). All clusters had a Jaccard stability index higher than 0.70 (Cluster 1 = 0.72; Cluster 2 = 0.81; Cluster 3 = 0.76; Cluster 4 = 0.72).

Figure 1.

Selection of the study cohort with parameters and results of cluster analysis. (A) Schematic representation of the study cohort and parameters included in the cluster analysis. (B) Hierarchical clustering on principal components separated the 24 persons into four clusters. Each dot represents one individual, each colour represents one cluster (purple = Cluster 1; light blue = Cluster 2; magenta = Cluster 3; orange = Cluster 4), and the lines that group the dots indicate Clusters 1–4. Percentages on the x- and y-axes depict the variation in the data. CPA, cyproterone acetate; AMH, anti-Müllerian hormone; LH, luteinizing hormone; FSH, follicle-stimulating hormone; SHBG, sex hormone-binding globulin; Dim, Dimension.

Histological characteristics of testicular tissues in identified clusters

Parameters driving Cluster 1 were the presence of tubules with elongated spermatids (mean 8.2%) and round spermatids (mean 14.6%), which were solely present in tissues from this cluster (Fig. 2A, Supplementary Table S1). Moreover, another driving factor of Cluster 1 was the presence of tubules with spermatocytes (mean 61.2%), which were also present in Cluster 2 but with reduced frequency (mean 20.2%). In Cluster 2, the majority of tubules contained spermatogonia as the most advanced germ cell type (positive driving factor, mean 66.2%; overall mean 33.9%). The low percentage of tubules with spermatocytes (mean 0.5%) was a negative driving factor of Cluster 3, which was otherwise characterized by tubules containing spermatogonia (mean 37.5%). The number of tubules with germ cells was lowest in Cluster 4, with a mean of 20.6% of tubules containing spermatogonia. Cluster 4 showed the highest number of Sertoli cell-only tubules (positive driving factor, mean 60.6%) and tubular shadows (mean 18.4%); however, the latter parameter was not a driving factor in the analysis. In line with the continuous depletion of germ cell populations, mean testicular weight decreased from Cluster 1 to Cluster 4. It is noteworthy that all clusters showed testicular weights below the reference values of 20–30 g (Supplementary Fig. S2A).

Figure 2.

Comparison of testicular tissue and spermatogonial compartment composition among Clusters 1–4. (A) Representative images of periodic acid-Schiff-stained tubules and corresponding box plots depicting the percentage of the most advanced germ cell types in seminiferous tubules of controls and Clusters 1–4. Scale bars represent 50 μm. ES, elongated spermatids; RS, round spermatids; SC, spermatocytes; SG, spermatogonia; SCO, Sertoli cell only; TS, tubular shadows. (B) Representative images of immunohistochemical staining for pan- and un-differentiated spermatogonial markers in tubules from control patients. Staining for melanoma-associated antigen 4 (MAGEA4), undifferentiated embryonic cell transcription factor 1 (UTF1), Piwi-like RNA-mediated gene silencing 4 (PIWIL4), fibroblast growth factor receptor 3 (FGFR3), Nanos C2HC-type zinc finger 3 (NANOS3) and the proportion of A_dark spermatogonia among MAGEA4-positive cells are shown. Scale bars represent 50 μm. Box plot elements are defined as follows: centre line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers.

Undifferentiated spermatogonial subpopulations in persons AMAB undergoing GAHT

The total number of spermatogonia per tubule marked by MAGEA4 showed the highest numbers in Cluster 1 (positive driving factor, mean: 5.6 positive cells per tubule) and the lowest numbers in Cluster 4 (mean: 0.8; Fig. 2B). Moreover, the proportion of A_dark reserve spermatogonia among MAGEA4⁺ cells followed this pattern, with the highest numbers in Cluster 1 (positive driving factor, mean: 9%) and reduced frequency in Cluster 4 (mean: 2%).

Undifferentiated spermatogonial subpopulations showed a similar pattern with the highest numbers of cells per tubular cross-section in Cluster 1 (positive driving factors), with a mean of 1.5 UTF1⁺, 0.5 PIWIL4⁺, 1.4 FGFR3⁺ and 0.9 NANOS3⁺ cells per tubular cross-section. Again, numbers decreased in Clusters 2–4 (Fig. 2B).

Differentiating spermatogonial subpopulations in persons AMAB undergoing GAHT

The proportion of differentiating spermatogonia, determined by the percentage of KIT⁺ cells within the MAGEA4⁺ cell population, was the highest in Cluster 1 (positive driving factor, mean 6%, Fig. 3A and B, Supplementary Table S3), as was the subpopulation of proliferating KI67⁺ cells among these KIT⁺ cells (both positive driving factors, Supplementary Table S3).

Figure 3.

Quantitative data on differentiating spermatogonial populations among Clusters 1–4. (A) Representative images of immunofluorescence staining of controls and Clusters 1–4 for MAGEA4, KIT, MKI67 and the respective IgG controls. In the bottom row, merged images for all markers are shown. Scale bars represent 5 μm. (B) Box plots depicting the percentage of KIT⁺ cells among MAGEA4⁺ spermatogonia and MKI67⁺ cells within the KIT⁺/MAGEA4⁺ spermatogonia (x-axis) for controls and Cluster 1–4 (y-axis). The percentage of KIT⁺ cells was a positive driving factor for Cluster 1. The percentage of MKI67⁺ among KIT⁺ cells was highest in Clusters 1 and 2, but was not a driving factor. Positive box plot elements are defined as follows: centre line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers.

The percentage of KIT⁺ spermatogonia decreased from Cluster 1 to Cluster 4. With regard to proliferative behaviour, Clusters 1 and 2 had the highest percentage of MKI67⁺ cells among KIT⁺ spermatogonia (MKI67⁺/KIT⁺/MAGEA4⁺). Nevertheless, MKI67 was not a significant driving factor for any of the clusters.

Comparison of endocrine parameters between the clusters

With regard to the endocrine axes, Cluster 1 was characterized by the lowest prolactin levels (negative driving factor, mean 831 mU/L) (Supplementary Fig. S2). Cluster 2 was characterized by testosterone (positive driving factor, mean 24.3 nmol/L) and free testosterone levels (mean 503.5 pmol/L; Fig. 4A) within the normal male range, rather than the desired female levels (<1.4 nmol/L for testosterone and <2.03 pg/mL for free testosterone, respectively; Supplementary Table S1). Interestingly, in Cluster 3, FSH levels were slightly below the reference values and negatively drove this cluster, which also displayed increased intratesticular inhibin B levels (positive driving factor, mean 23.8 pg/mL; Fig. 4B). For Cluster 3, serum AMH and intratesticular AMH levels were elevated beyond the normal reference range (positive driving factors, mean 69.1 ng/mL). Cluster 3 also showed the highest mean values with regard to treatment duration (positive driving factor, mean 33.6 months; Fig. 4C). Cluster 4 was positively driven by the highest age (mean 35.7 years) and treatment dose of cyproterone acetate (mg/day) Fig. 4D). Furthermore, Cluster 4 showed the highest FSH (mean 10.6 U/L) and LH (mean 7.8 U/L) values (positive driving factors).

Figure 4.

Comparison of clinical driving factors among Clusters 1–4. Box plots comparing endocrine parameters of persons AMAB undergoing GAHT in Clusters 1–4. (A) Testosterone and free testosterone as driving factors in Cluster 2. (B) FSH as a negative and intratesticular inhibin B as a positive driving factor in Cluster 3. (C) Treatment duration, serum AMH and intratesticular AMH as driving factors in Cluster 3. (D) Age and CPA intake as positive driving factors in Cluster 4. In-house reference values are indicated by dotted lines.

Discussion

Persons AMAB undergoing GAHT show heterogeneous effects at the testicular tissue and endocrine levels. As comparative studies are hampered by varying treatment regimens, this study sought to assess the existence of subgroups within a cohort of persons AMAB with comparable CPA treatment regimens.

This selected homogeneous group was separated into four clusters, with the main driving factors related to the composition of testicular tissues, the spermatogonial compartment and endocrine parameters. In contrast, oestrogen dosage and discontinuation of treatment prior to surgery did not differ among the clusters. Clusters 1 and 3 showed the expected suppression of testosterone levels (mean 1.6 nmol/L), in line with the desired suppression to adult female values (<1.91 nmol/L) (De Sutter 2001, Schneider et al. 2015). Nonetheless, there were important differences between these two clusters. Cluster 1 constituted the smallest cluster (n = 3), and persons had the lowest mean treatment duration (14 months). This cluster was largely characterized by ongoing germ cell differentiation up to the stage of elongated spermatids, facilitating validated clinical measures of fertility preservation. Previous studies also reported a proportion of persons AMAB undergoing GAHT with complete spermatogenesis in 11–24% of respective cohorts (Schneider et al. 2015, Jindarak et al. 2018, Matoso et al. 2018). These findings are in line with data from contraceptive trials in which men with normal and impaired sperm parameters were treated with testosterone undecanoate, in response to which approximately 25% of individuals did not show the expected suppression of spermatogenesis to <1 mill sperm/ejaculate, indicating inter-individual differences in suppressibility of spermatogenesis (Nieschlag et al. 2011). Hence, there is a need to counsel persons undergoing GAHT with regard to contraception (Hembree et al. 2017, Coleman et al. 2022) as well as options for fertility preservation (Schneider et al. 2019). Regarding fertility preservation, on the day of GAS either microsurgical epididymal sperm aspiration or microsurgical testicular sperm extraction could be performed to cryopreserve sperm. If this is not successful, cryopreservation of spermatogonial stem cells can be offered on the day of GAS. This method, however, remains experimental. Furthermore, stopping GAHT before GAS can be discussed to increase the chances of cryopreserving sperm. It is also noteworthy that persons from Cluster 1 displayed relatively normal adult AMH levels in serum and testicular tissues, which does not indicate a dedifferentiation of Sertoli cells. Finally, samples from Cluster 1 showed the highest numbers of A_dark spermatogonia, as well as undifferentiated (UTF1⁺, PIWIL4⁺, FGFR3⁺, NANOS3⁺) and differentiating spermatogonial subpopulations (KIT⁺). We consider it likely that persons from this cluster would recover spermatogenesis upon treatment termination, as reported by (De Nie et al. 2022). Cluster 3 (n = 9) was the largest cluster and persons had the longest treatment duration (mean 33.6 months). Their testicular tissues displayed the expected regression of spermatogenesis to the spermatogonial level, which is in line with reduced testosterone levels and previously reported results (Delgouffe et al. 2025). Moreover, samples in this cluster displayed the reported dedifferentiation of Sertoli cells, with the highest serum (64.1 ng/mL) and intratesticular (12.6 ng/mL) AMH levels (Vereecke et al. 2020, Delgouffe et al. 2025), indicating that serum AMH reflects suppression of spermatogenesis. In Clusters 2 and 4, testosterone levels were close to male reference values (24.3 and 12.0 nmol/L, respectively) and were not reduced to the desired female values (<1.9 nmol/L). Persons in Cluster 2 (n = 4) had a mean treatment duration of 26 months and were characterized by individual tubules containing spermatocytes and the highest percentage of tubules with spermatogonia. Interestingly, these persons had serum (6.9 ng/mL) and intratesticular AMH levels (2.8 ng/mL) that did not indicate dedifferentiation of Sertoli cells. We speculate that in these persons, GAHT may have been terminated weeks prior to surgery, which is still advised by individual clinics to minimize the risks of thrombosis (Hembree et al. 2017, 2009). This discontinuation of GAHT has previously been demonstrated to lead to re-virilized hormone levels (Schneider et al. 2015) and re-initiation of germ cell differentiation. While we cannot exclude that discontinuation of GAHT may have facilitated re-initiation of germ cell differentiation from spermatogonia to spermatocytes in tissues from Cluster 2, the clinic performing the surgery and the corresponding instruction for GAHT discontinuation were not relevant factors for clustering of persons undergoing GAHT as shown in Supplementary Table S1. Finally, persons in Cluster 4 (n = 7) had a mean treatment duration of 22.6 months. Tissues from these persons displayed the most drastically impacted testicular phenotype, with a low percentage of tubules containing spermatogonia and the highest percentage of tubules with a Sertoli cell-only phenotype or tubular shadows. As the treatment duration was of intermediate length and persons were characterized by the highest age (35.7 years), we speculate that testicular tissues may become more sensitive to GAHT with age (Dabel et al. 2023), leading to the most severe testicular phenotype observed.

Regarding the spermatogonial compartment, GAHT does not seem to selectively impact specific subpopulations of spermatogonia. There was a general decrease in undifferentiated (UTF1⁺, PIWIL4⁺, FGFR3⁺, NANOS3⁺) as well as differentiating spermatogonial subpopulations (KIT⁺) from Clusters 1–4. This finding is in line with the representation of all undifferentiated spermatogonial subpopulations in testicular tissues of two persons AMAB undergoing GAHT analysed by scRNA-seq analyses (Guo et al. 2020). The general reduction of spermatogonia is furthermore paralleled by a continuous reduction in A_dark spermatogonia, implying that the spermatogonial compartment as a whole is shrinking under GAHT. This finding is in line with the reported correlation between A_dark spermatogonia and the percentage of tubules showing spermatogenesis (Schülke et al. 2024).

To conclude, even in this selected cohort of persons AMAB undergoing GAHT with comparable CPA regimens, there was heterogeneity with regard to treatment effects. The majority of persons (16/24, Clusters 3 and 4) showed regression of germ cell differentiation, with spermatogonia as the most advanced germ cell type. The associated and proportional downsizing of the spermatogonial compartment may, in principle, facilitate the reinitiation of spermatogenesis upon termination of GAHT. Interestingly, in samples of older persons, the impact of GAHT appeared to be even more drastic, with high numbers of Sertoli cell-only tubules and tubular shadows, which will likely limit chances for spermatogenic recovery upon termination of treatment. Relevantly for the clinical care is the finding that 2/24 persons AMAB undergoing GAHT showed full germ cell differentiation, and one had round spermatids on the day of GAS, which is of relevance for counselling regarding contraception as well as options for fertility preservation.

Declaration of interest

The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Funding

This work was supported by the IMF (I-SC212412) of the University of Münster. We acknowledge the support from the open-access publication fund of the University of Münster.

Author contribution statement

NN, SDP and FS performed the study conception and design. NN, SDP, FS, SK and StS supervised the study. JH, FS, SDP and SK acquired and evaluated clinical data. KS, FSi, FS and SDP performed data and bioinformatic analyses. KS, FSi, FS, SDP and NN wrote the original draft of the manuscript. All authors contributed to editing, read and approved the final version of the manuscript.