Copyright © 2001 by The Endocrine Society

IN ANALOGY WITH the nonhuman vertebrate brain ^{[1] [2]} , it is thought that in the human also the interaction between sex hormones and their receptors may play an important role in brain development (organizing effects) and may in adulthood alter brain function (activating effects), and that these two mechanisms lead to sex differences in behavior in adult life. Structural and functional sex differences in the brain may be related to reproduction, sexual orientation, gender identity ( i.e. the feeling of being male or female), cognition, and disease ^{[3] [4]} . In a number of areas of the human hypothalamus, structural and functional differences between the sexes and between homosexual and heterosexual men have been described ^{[5] [6] [7]} . In addition, our group has found that the central part of the bed nucleus of the stria terminalis (BSTc) is sexually dimorphic, i.e. smaller in women, with a female volume and neuron number in male-to-female transsexuals ^{[4] [8]} .

It has been shown that various areas of the preoptic area (POA) ^{[9] [10]} , BST ^[11] and suprachiasmatic nucleus ^{[12] [13]} are larger in men than in women, whereas the opposite was found for the anterior commissure ^[14] . Moreover, hypothalamic differences in relation to sexual orientation have been observed. The suprachiasmatic nucleus ^[15] and the anterior commissure ^[16] are larger in homosexual than in heterosexual men, whereas the interstitial nucleus of the anterior hypothalamus-3 is smaller in homosexual than in heterosexual subjects ^[17] . These data together with the abundant information showing that sexual orientation and gender identity do not vary with adult endocrine changes ^{[18] [19]} suggest that any possible clue to understanding the biological basis of sex differences, sexual orientation, or gender identity will require careful analysis of a large number of brain areas.

Recently we found that in a number of hypothalamic areas men showed stronger androgen receptor (AR) immunoreactivity (AR-ir) than women. Interestingly, in the anterior hypothalamus only moderate sex differences were found, whereas a conspicuous sex difference occurred in the posterior hypothalamus, i.e. the medial mamillary nucleus (MMN) and lateromamillary nucleus (LMN) of the mamillary body (MB) complex (MBC) ^[20] . Such sexual dimorphisms may be related to gender differences in certain aspects of reproduction or sexual behavior, as various studies

Recipient of a grant from the NWO (Grant 903-47004).

819

The present study was carried out to investigate whether the sex difference in nuclear AR-ir of the human MBC is related to sexual orientation, gender identity, or endocrine status. In many species, castration strongly reduces or even eliminates nuclear AR-ir, whereas testosterone, but not estrogen, injection restores such strong nuclear AR-ir ^{[28] [29] [30]} . As nuclear AR-ir is up-regulated by androgens ^{[28] [31] [32]} , we would expect decreased AR-ir in castrated/aged men. Therefore, we studied AR-ir in the MBC in groups of subjects with different testosterone levels ^{[33] [34]} , i.e. young heterosexual men/young homosexual men, young heterosexual women, aged heterosexual castrated and noncastrated men, and castrated and noncastrated transsexuals.

Subjects and Methods

Subjects

In the present study we included the area of the MBC from the posterior hypothalamus of the following 47 patients: 1) young heterosexual men (n = 9), 2) young homosexual men (n = 10), 3) old hetero sexual castrated men (n = 5), 4) castrated male-to-female transsexuals (n = 6), 5) old heterosexual intact men (n = 5), 6) young heterosexual women (n = 8), 7) a 36-yr-old noncastrated male-to-female transsexual, 8) a nontreated 84-yr-old male subject with strong cross-gender identity feelings, 9) a 51-yr-old female-to-male transsexual, and 10) a 46-yr-old woman with high levels of androgens.

Brains were obtained by autopsy (for clinicopathological information and ages, see Table 3 ). Unless stated otherwise, patients had no primary neurological or psychiatric diseases. The sexual orientation of the subjects was presumed to be heterosexual ^[15] unless stated otherwise, whereas the sexual orientation of the homosexual group was documented in the clinical records ^[15] . All homosexual patients died of acquired immunodeficiency syndrome or related diseases. The patient data have previously been reported ^[15] . General pathology and neuropathology were performed either at the Free University of Amsterdam (Dr. W. Kamphorst, Prof. F. C. Stam or Prof. P. van der Valk) or at the Academic Medical Center of the University of Amsterdam (Dr. D. Troost). The subjects had no primary endocrine illnesses, except for those who had undergone orchidectomy, had been given hormonal treatment, or had had abnormal hormone fluctuations that are mentioned in Table 2 . The pathologically high levels of androgens in a 46-yr-old woman, [androstenedione, 48.0 ng/mL (normal values for women, 0.4-3.5 ng/mL); testosterone, 26.82 nmol/L (normal values for women, 1.04-3.30 nmol/L)] were due to an adrenal cortex carcinoma.

Histology and immunohistochemistry

After autopsy, the hypothalamus was fixed for about 1 month in 4% formaldehyde at room temperature, dehydrated, and embedded in paraffin. Serial 6-mum frontal sections were cut on a Leitz microtome (Rockleigh, NJ).

The immunohistochemical protocol followed for the AR staining has been previously described in detail ^[20] . Briefly, this protocol consisted of mounting paraffin-embedded sections of the posterior hypothalamus onto SuperFrost Plus (Menzel, Darmstadt, Germany) slides. The sections were deparaffinized and rehydrated in a series of ethanol concentrations. To retrieve antigenicity, sections were microwaved (10 min at 700 watts) in 0.1 mol/L citric acid monohydrate buffer (pH 6.0) ^{[35] [36]} , after which they were rinsed with TBS buffer (0.05 mol/L Tris-0.9% NaCl, pH 7.6). To decrease background, the slides were preincubated for 1 h with TBS-milk [5% milk-TBS solution with commercially available powdered milk (ELK, Campina Melkunie, Eindhoven, The Netherlands)] before incubation with the primary antibody PG21 (donated by Drs. Gail Prins and Geoffrey Greene; 1:1000) for 1 h at room temperature and subsequently kept overnight at 4 C. After rinsing in TBS-milk buffer, sections were incubated for 1 h with a goat antirabbit biotinylated second antibody (1:200), followed by another hour of incubation in the avidin-biotin complex (1:800). The subsequent signal amplification method consisted of an incubation in biotinylated tyramine (1:1000) and 0.01% peroxide (Merck & Co., Darmstadt, Germany) for 20 min ^[37] . Thereafter, sections were rinsed with TBS, and the avidin-biotin complex procedure was repeated. After rinsing in 0.05 mol/L Tris-HCl (pH 7.6), slides were developed by incubation for 10 min in 0.05 mol/L Tris-HCl containing 0.05% 3,3-diaminobenzidine (Sigma, St. Louis, MO), 0.01% hydrogen peroxide, and 0.3% nickel ammonium sulfate. Developed sections were dehydrated in alcohol, cleared with xylene, and coverslipped with Entallan (Merck & Co.).

Analysis of AR staining intensity

The sections were rated for staining intensity by three independent investigators blind to the details of the patients. The few differences in rating were concurred by settlement ^[20] . The category assigned to the MMN and LMN corresponded to the predominant cell type within that area according to the following scale: 0 = no staining, 1 = staining diffuse and transparent, and 2 = intense staining with individual granules of the reaction product distinguishable. The staining range was established for both the cytoplasm and the nucleus. The estimates were made at three different microscopic magnifications: ×2.5, ×10, and ×40 objectives ^[20] . The identification of MMN and LMN was made with the aid of maps of coronal sections of the human brain published by Mai et al. ^[38] and using alternating thionine-stained sections for orientation.

Statistics

The assigned categories of AR-ir in the MMN and LMN were compared using the Kruskal-Wallis ANOVA, followed by the Mann-Whitney U test. The fixation time and postmortem delay were analyzed by the Kruskal-Wallis test. Differences were considered statistically significant at P < 0.05 (two-tailed).

Results

Staining specificity

Negative controls, i.e. without the first antibody, and positive control sections, i.e. tissue of mouse testes and human anterior

TABLE 1 -- Androgen receptor intensity staining in the medial mamillary nucleus (MMN) and the lateromamillary nucleus (LMN) of the mamillary body of heterosexual young men, heterosexual young women, homosexual young men, and castrated transsexual men

	N	LMN		MMN
		c	n	c	n
Heterosexual men	9	1	2	1	2
Heterosexual women	8	1	1	1	0
Homosexual men	10	1	2	1	1
Transsexual men	6	1	0	1	0
The data of these 33 subjects that were used for the statistical analysis (see Results and Table 3 ) are summarized here with the median values for nuclear and cytoplasmic AR-ir. The table shows median values. The AR-ir intensity was assigned according to the following scale: 0 = no staining; 1 = staining diffuse and transparent, and 2 = intense staining with individual granules of the reaction product still distinguishable. c, Cytoplasmic; n, nuclear. The Kruskal-Wallis and Mann-Whitney U tests for nuclear staining indicated statistically significant differences among the groups (see Results). The cytoplasmic staining in both brain areas studied was relatively weak for the four groups and did not differ statistically.

820

Figure 1. Photomicrographs showing AR-ir in neurons of the MMN of the mamillary body of a heterosexual man (A), a heterosexual woman (B), a homosexual man (C), and a woman with high levels of androgens (D). Note that in the mamillary body there is a clear sex difference in AR-ir (see A and B), whereas there is no difference in the intensity of AR staining between the representative heterosexual man (A), the homosexual man (C), and the virilized (androgenized) woman (D). Scale bar, 150 mum.

821

Staining pattern

ANOV As for fixation time and postmortem delay among the heterosexual men, heterosexual women, homosexual men, and castrated transsexual men did not exhibit statistically significant differences ( P > 0.1 and P > 0.9, respectively).

The immunohistochemical staining for AR in the MMN and LMN revealed cells with nuclear or cytoplasmic labeling or with both types of staining ( Tables 1 and 3 and Figs. 1, 2, and 3 ).

The heterosexual men showed strong nuclear AR-ir in both brain regions ( Tables 1 and 3 and Fig. 1A ). In contrast, the women revealed much less intense labeling in the nucleus of neurons of the LMN and MMN ( Tables 1 and 3 and Fig. 2B ). This sex difference was statistically significant for nuclear staining in both areas ( P < 0.05). The homosexual men showed a similar staining to that of the heterosexual men for both areas ( P > 0.2) with a more moderate staining in the MMN ( Tables 1 and 3 and Fig. 1C ). Women differed significantly from homosexual men in nuclear AR-ir in both the MMN and LMN ( P < 0.05). The castrated male-to-female transsexual group had a lack of nuclear staining in both brain areas, but had cytoplasmic labeling in the LMN and MNN ( Tables 1 and 3 and Fig. 2B ). This group was statistically different in the LMN from the heterosexual and homosexual men group ( P < 0.05) and similar to that in women ( P > 0.5). The castrated male-to-female transsexual group had significantly less nuclear AR-ir in the MMN than the heterosexual male group ( P < 0.05). This difference showed only a trend when compared with homosexual men ( P = 0.10). When

Figure 2. Illustration of the staining intensity of nuclear AR-ir in neurons of a noncastrated 36-yr-old male-to-female transsexual (A) compared with the lack of such staining in a 26-yr-old castrated male-to-female transsexual. Scale bar, 150 mum.

In the 36-yr-old noncastrated male-to-female transsexual, strong nuclear and cytoplasmic staining was observed in both areas ( Fig. 2A ). Similarly, the 46-yr-old woman with high levels of androgens revealed strong nuclear labeling and weak to intermediate cytoplasmic labeling in the LMN and MMN ( Fig. 1D ). The female-to-male transsexual who did not receive androgen replacement therapy during the last 3 yr before death ( Table 2 ) showed less intense nuclear and weak to intermediate cytoplasmic staining in both the LMN and MMN.

The results of staining in the posterior hypothalamus of old patients are illustrated in Fig. 3 (A and B) . In old castrated heterosexual men almost no nuclear and weak cytoplasmic AR-ir were found ( Table 3 and Fig. 3B ). Thus, in this group of five, two subjects had very weak nuclear and cytoplasmic AR-ir in both areas, whereas three of five had no nuclear but weak cytoplasmic AR-ir ( Table 3 and Fig. 3B ). A similar trend of weak AR staining with more, but less intense, nuclear AR-ir was observed in five old intact men ( Table 3 and Fig. 3A ). In this group four of five of the individuals had very weak nuclear as well as cytoplasmic AR-ir ( Table 3 ). Between the castrated old men and the intact old men no statistical significant differences were found.

Discussion

The present study confirms the clear sex differences in nuclear AR-ir expression in neurons of the MBC ^[20] and shows, for the first time, that this sex difference is related to circulating levels of testosterone rather than to sexual orientation or gender identity.

822

823

Heterosexual and homosexual men had the most intense nuclear AR-ir in the LMN and MMN which were statistically not different from each other but showed significantly more AR-ir than women. A similar male-like staining intensity was also found in a 36-yr-old bisexual noncastrated male-to-female transsexual and in a 46-yr-old heterosexual virilized woman with high circulating levels of testosterone. In all cases studied, a close relationship was found between the endocrine status and the intensity of AR staining. High levels of testosterone went together with high nuclear AR-ir and low levels of testosterone with weak or no nuclear AR-ir. The fact that homosexual men showed more variability in their nuclear AR-ir profiles compared with heterosexual men, whereas in the MMN the AR-ir did not differ statistically from the transsexual group might be due to their acquired immunodeficiency syndrome status, as some of these patients have subnormal testosterone levels ^{[45] [46] [47]} . The strong decrease in nuclear MBC AR-ir in five old male heterosexual intact subjects also fit the idea of an androgen-dependent nuclear expression of the AR, as decreased circulating levels of androgens occur with aging ^{[48] [49]} . The weak AR-ir in an 84-yr-old man, who was gynecophilic and who had well documented strong cross-gender identity feelings but never received hormonal treatment or sex reassignment therapy, fits with his age. The fact that the 51-yr-old female-to-male transsexual who did not receive testosterone replacement during the last 3 yr before death still had a female-like pattern of AR-ir is also fully in agreement with the assumption that circulating testosterone is crucial for nuclear AR-ir.

From our data there appeared no relationship between AR-ir and sexual orientation or gender identity. Regardless of sexual orientation or gender identity, a female pattern of AR-ir, i.e. low nuclear staining in the LMN and MMN, was observed in women, castrated male-to-female transsexuals, a female-to-male transsexual, and old men.

Animal studies show that castration induces a shift from strong nuclear to weak cytoplasmic AR-ir in hypothalamic neurons, which can be reversed by treatment with androgens ^{[29] [30]} . Regarding this point it seems of particular interest to note that ongoing testosterone treatment in female-to-male transsexuals was accompanied by up-regulation of AR in peripheral ectocervix tissue, resulting in increased nuclear AR-ir ^[32] . In addition to the specificity tests (Refs. ^{[29] [36] [39]} and 43 and our additional specificity data), the analogy between data on ARs in animals and humans under different levels of testosterone ^{[33] [34]} now also seems to provide biological evidence that in neurons of the human brain, PG21 indeed recognizes ARs. Our data from postmortem tissue are in agreement with the experimental data reported by Wood and Newman ^{[29] [30]} , as we also show that gonadectomy in transsexuals ^[34] goes together with cytoplasmic AR-ir, in contrast with the mainly nuclear AR-ir pattern in the male

824

Figure 3. Illustration of the almost complete lack of nuclear AR-ir in a noncastrated old man (A) and in a castrated old man (B). Scale bar, 50 mum.

Hormones and their receptors in relation to sexual orientation and transsexuality

The possible role of steroid hormones in the development of sexual orientation has been studied in various animal models ^[51] and in humans ^[52] . The animal data show that steroid hormones during the neonatal period contribute to the organization of the brain and influence sexual preference ^[53] . In contrast, the exposure to sex steroids during adulthood stimulates sexual behavior, but does not modify sexual orientation in animals or humans ^{[18] [54]} . In the present study no statistical difference was found in AR-ir between heterosexual and homosexual men. The lack of differences in the AR-ir in the posterior hypothalamus and in the sequence variation in the AR gene ^[55] between homosexual and heterosexual men suggests that neither the intensity of AR-ir nor a variation in AR structure is related to the expression of homosexuality. Although possible differences between homosexual and heterosexual men in the AR-ir in other brain areas have not yet been systematically studied, the data obtained to date reinforce the idea that homosexuality does not depend on differences in activational effects of testosterone in adulthood ^[18] .

In the present study we found no clear relationship between MBC AR-ir and gender identity as we did, for instance, find for the size of the BSTc and gender identity ^{[4] [8]} . Recent studies ^{[56] [57] [58] [59]} and our observation that the volume and neuron number of the BSTc in male-to-female transsexuals in adulthood is independent of sex hormone levels ^{[4] [8]} support the idea that steroids do not act in adulthood but, rather, earlier during development to establish gender identity.

Functional implications

The results obtained in the present study fully agree with the idea that the AR-ir sex differences in the MBC of the posterior hypothalamus are due to differences in circulating levels of androgens and give additional support to the paradigm that endocrine features during adulthood do not contribute to sexual orientation or gender identity.

Anatomical and functional studies in rats have shown that the MMN and LMN are larger in males than in females ^[60] , a sex difference that is accompanied by an increase in the rate of global protein synthesis ^[61] . In addition, experimental data in animals show a sex-specific involvement of the MB in aspects of sexual behavior such as sexual motivation, penile erection, and sexual activity ^{[21] [22] [23] [24] [27]} . Also in humans, the MBC has been implicated in the regulation of reproduction, possibly by inhibiting the release of gonadotropins, as lesions in the posterior hypothalamus go together with precocious puberty ^[62] . Our findings of gonadal hormone

825

826

In addition to reproduction, the MBC plays a crucial role in memory function ^[63] . Mamillary bodies atrophy with age ^[64] and even more so in Alzheimer's disease ^[65] and are damaged in alcohol-associated Wernicke-Korsakoff's disease ^[66] . The decline with age in nuclear AR-ir in the male MBC as found in the present study may also be reflected in functional changes. Whether the observed changes in the MBC play a role in the relationship between low levels of sex hormones and impairment in sexual and cognitive functioning ^{[48] [67] [68]} or in the increased prevalence of nonfamiliar Alzheimer's disease in the elderly ^{[69] [70]} should be further investigated. Protective actions of androgens on neurons ^[71] and memory loss ^{[72] [73]} have been described. It may in this connection also be of interest to investigate the possible neuroprotective effects of androgens in age-related diseases in men, in a similar way as is done for estrogen replacement therapy in postmenopausal women with reported beneficial effects on physical status, mood, cognition, and the prevention of Alzheimer's disease ^{[67] [74] [75]} , although the latter certainly requires more investigation.

In conclusion, here we show for the first time that the sex differences in nuclear AR-ir in the MBC of the posterior hypothalamus reflect differences in circulating levels of androgens rather than differences in sexual orientation or gender identity. The functional implications of these alterations should be studied in the future.

Acknowledgments

Brain material was obtained from The Netherlands Brain Bank (Anne Holtrop, Michiel Kooreman, and Jose Wouda; coordinator Dr. R. Ravid). We thank Bart Fisser for his technical assistance, Dr. F. W. van Leeuwen for the antibody PG21, Dr. S. Kaiser for critically reading the manuscript, G. van der Meulen for photography, and W. Verweij for secretarial help.

References