Characterization of Aromatase Expression in the Adult Male and Female Mouse Brain. I. Coexistence with Oestrogen Receptors α and β, and Androgen Receptors

Aromatase catalyses the last step of oestrogen synthesis. There is growing evidence that local oestrogens influence many brain regions to modulate brain development and behaviour. We examined, by immunohistochemistry, the expression of aromatase in the adult male and female mouse brain, using mice in which enhanced green fluorescent protein (EGFP) is transcribed following the physiological activation of the Cyp19A1 gene. EGFP-immunoreactive processes were distributed in many brain regions, including the bed nucleus of the stria terminalis, olfactory tubercle, medial amygdaloid nucleus and medial preoptic area, with the densest distributions of EGFP-positive cell bodies in the bed nucleus and medial amygdala. Differences between male and female mice were apparent, with the density of EGFP-positive cell bodies and fibres being lower in some brain regions of female mice, including the bed nucleus and medial amygdala. EGFP-positive cell bodies in the bed nucleus, lateral septum, medial amygdala and hypothalamus co-expressed oestrogen receptor (ER) α and β, or the androgen receptor (AR), although single-labelled EGFP-positive cells were also identified. Additionally, single-labelled ERα−, ERβ- or AR-positive cell bodies often appeared to be surrounded by EGFP-immunoreactive nerve fibres/terminals. The widespread distribution of EGFP-positive cell bodies and fibres suggests that aromatase signalling is common in the mouse brain, and that locally synthesised brain oestrogens could mediate biological effects by activating pre- and post-synaptic oestrogen α and β receptors, and androgen receptors. The higher number of EGFP-positive cells in male mice may indicate that the autocrine and paracrine effects of oestrogens are more prominent in males than females.


Introduction
It has been known for nearly forty years that oestrogens are produced in the brain by the local aromatisation of testosterone [1], and these locally synthesised oestrogens [2] may modulate neuronal functions and provide neuroprotection (reviewed by [3]). Regions of the brain that express aromatase were initially identified by in situ hybridization (see [4,5]) and measurements of aromatase activity. High levels of aromatase activity [6] and its mRNA [6][7][8][9][10] have been detected in the hypothalamus, preoptic nucleus, sexually dimorphic nucleus, bed nucleus of the striata terminalis (BST), hippocampal formation [2] and cerebellum [11].
While regions of the brain expressing aromatase are similar in both sexes, there are important differences in their levels of expression. Aromatase activity is four times higher in the hypothalamus of the postpubertal porcine male than female [12]. Aromatase transcript levels are also higher in the male rat amygdala, BST and medial septal nucleus than the corresponding nuclei in females [13,14], although these differences are not present in some species, including sheep [15].
Examination of the distribution of aromatase expression and sex hormone receptors in brain circuitry would be advanced if the full repertoire of investigative tools, including immunohistochemistry and Western blotting could be used. This is the case with avian models in which specific antibodies were available for avian aromatase. As a result the distribution of aromatase in avian brains is well defined, and has coincided with advances in understanding the effects of brain-derived steroids on behaviour of bird models. Indeed, studies in songbird point to sexual dimorphic distribution of aromatase and more importantly, to synaptic synthesis of neuroactive steroids such as oestrogens mediating fast, perisynaptic membrane actions (refer to reviews by [16,17]).
Unlike the avian models, immunohistochemical studies in rodents have provided inconsistent results that are at odds with results from studies using in situ hybridization or aromatase activity. For example, a polyclonal antibody raised against the rat aromatase sequence found little aromatase immunoreactivity in cell bodies within the BST, and medial basal hypothalamic and preoptic areas [18], whereas in situ hybridization [19], and other immunohistochemical studies, demonstrated high levels in these regions [20]. Of further concern is that Zhao et al. [21], using polyclonal antibodies raised against human placenta aromatase, found intense aromatase immunoreactivity in the male adult rat oval nucleus of the BST and lateral division of the central amygdaloid nucleus, but could not detect aromatase transcript by in situ hybridization in these same nuclei.
Evidence has accumulated that classical ERa can be anchored to the cell membrane through palmitoylation of ERa C447 and insertion to membrane caveoli [22,23]. This membrane bound ERa can elicit rapid, non-genomic actions [22][23][24]. In the view that palmitoylation of oestrogen receptors is essential for neuronal membrane signalling [25] and brain oestrogens produced by brain aromatase are 'neuroactive steroids' that alter neuronal excitability and rapid signalling (reviewed by [26]), it is imperative to describe the distribution of the neurites of aromatase-positive neurones relative to oestrogen receptor expressing neurones in mouse -a common animal model used to study behaviour due to the ease of manipulating gene expression.
Immunohistochemistry depends on the sensitivity and selectivity of the antibody for its epitope, but it also has the advantage of identifying the specific cellular location of proteins, and allows for co-labelling with antibodies raised against aromatase, and oestrogen and androgen receptors. In this study we have circumvented the difficulties associated with antibodies raised against aromatase by using a transgenic mouse model in which transcription of enhanced green fluorescent protein (EGFP) occurs following the physiological activation of the Cyp19A1 gene, and thus labelling the soma as well as the fibres of the aromatase expressing cells. We have also compared the distribution of oestrogen receptor a and b, and the androgen receptor, with the EGFP expression in the brain of this mouse.

Distribution of EGFP-LI in the adult mouse brain
The distribution of EGFP-ir cell bodies and processes was examined in brains of adult male and female mice, in which transcription of EGFP occurred following the physiological   Table 1. Summary of GFP-like immunoreactivity in the cell bodies and nerve fibres throughout the adult male and female mouse brain.

AMYGDALOID COMPLEX
Amygdalohippocampal area -(+) -( +) Anterior amygdaloid area, dorsal and ventral parts -(+) -( +) activation of the Cyp19A1 gene. EGFP-LI was detected using the TSA+ immunohistochemical method, and a chicken antiserum raised against EGFP. Most EGFP+ cell bodies were round/ovoid in shape, and their size ranged from ,10 to 15 mm in diameter. EGFP-LI was typically distributed in the cytoplasm, and was particularly intense towards the main axon/dendrite. The EGFP-LI appeared confluent and densely packed in many regions. However, under high magnification, and in particular using confocal microscopy, this fluorescence was 'punctate' in appearance and often colocalized with synaptophysin-LI ( Fig. 1), thus was presumably confined to nerve endings. The expression pattern of the EGFP and aromatase has been confirmed by in situ hybridization [27,28]. The relation of EGFP+ cell bodies to other neurone populations expressing ERa2, ERb-, AR-or calbindin-LI is presented in merged colour micrographs of double-label immunohistochemistry, together with the corresponding blackand-white single channel micrographs. ERa2, ERb-, and AR-LI were confined to the nucleus of cell bodies, with the diameter of immunoreactivity ranging from ,8 to 12 mm. We will use abbreviations intermittently within subsections of the text to save space, and also to provide a 'direct' connection to the micrographs. For control purposes, EGFP fluorescence was detected in fresh testis and granulosa layer in follicles of snap frozen ovary collected from adult transgenic animals, while no EGFP fluorescence was detected in the liver of adult mice ( Fig. 2A-C).
The BST of male mice contained a greater density of EGFP-ir nerve fibres, and the number of EGFP+ cell bodies in the BST of female mice was 63% lower than the number estimated in male mice (Figs. 3H c.f. 3G, 6A; Table 2). In other regions of the basal forebrain, including the LS or Tu, the density/intensity of EGFPir fibres and EGFP+ cell bodies appeared similar in male and female mice (Fig. 3A c.f. 3B, 3E c.f. 3F, Table 1).
Cell bodies immunoreactive for ERa, ERb, and the AR were also distributed in the LS, BST and OT. In the LS, coexpression between EGFP+ and ERa+ cell bodies was minimal, with most EGFP+ cells located more laterally than those expressing ERa (Fig. 7A, 7B, 7G). There were, however, occasional cell bodies in the LS that coexpressed EGFP2 and ERa-LI. The LS contained a greater number of ERb-or AR-ir cell bodies compared to those that expressed EGFP-LI (Fig. 7C, 7D, 7H; 7E, 7F, 7I), with most     of the EGFP+ cell bodies also expressing ERb-LI or AR-LI ( Fig. 7H, 7I).
In both female and male mice, the estimated number of ERa2, ERb-or AR-ir cell bodies in the BST was higher than the number of EGFP+ cells (Table 2). While many of the EGFP+ cells coexpressed ERa2, ERb-or AR-LI, EGFP+-only cell bodies were also found ( Fig. 8A, 8B, 8G; 8C, 8D, 8H; 8E, 8F, 8I). The BST of female mice contained 36% fewer EGFP/ERa coexpressing cell bodies than male mice, 60% fewer EGFP/ERb coexpressing cell bodies and 69% fewer EGFP/AR coexpressing cell bodies ( Fig. 6B; Table 2). The proportion of EGPF+ cell bodies in the BST that coexpressed ERa-LI was greater in female than male mice, while the proportion of EGFP+ cell bodies in the BST that coexpressed ERb-LI or AR-LI was similar in female and male mice ( Fig. 6C; Table 2). The proportion of ERa-ir cell bodies in the BST that coexpressed EGFP was similar in female and male mice, as was the proportion of AR-ir cell bodies in the BST that were EGFP+ ( Fig. 6D; Table 2). However, the proportion of ERb-ir cell bodies in the BST that coexpressed EGPF was greater in male mice ( Fig. 6D; Table 2).
Confocal analysis indicated that EGFP-ir fibres and terminallike structures were often distributed adjacent to single-labelled ERa2, ERb-or AR-ir cell bodies in the BST ( Fig. 8G-I, 8J-L). When the positioning of these structures was further examined, with the addition of immunohistochemistry against calbindin (which is distributed throughout the cytoplasm and thus outlines the boundaries of the cell), it became apparent that EGFP-ir fibre/ terminal-like structures were adjacent to the cell bodies in a way that suggested that they were in contact ( Fig. 8M-O).
In the Tu, many EGFP+ cell bodies coexpressed ERa-LI, although EGFP+-only cells were also present ( Cerebral Cortex. In the cerebral cortex, EGFP+ cell bodies were infrequently observed in the cingulate cortex and the secondary motor cortex.  Table 2. Estimated number of EGFP+, ERa-ir, ERb-ir and AR-ir cell bodies in the BST (mean 6 SEM). Hippocampal Formation. Occasional, sparsely distributed, EGFP+ cell bodies and EGFP-ir fibres were observed in the stratum radiatum (Rad) of the hippocampal formation (Fig. 4F, 4G), which did not coexpress ERa2, ERb-, or AR-LI.
Cell bodies immunoreactive for ERa, ERb, and the AR were also distributed through the MePD and MePV, and their number was greater than the number of EGFP+ cells (Table 3). In these regions of the amygdaloid complex, female mice contained 47% fewer EGFP/ERb coexpressing cell bodies than male mice and 41% fewer EGFP/AR coexpressing cell bodies, while the number of EGFP/ERa coexpressing cell bodies was similar in female and male mice ( Fig. 10B; Table 3). The proportion of EGPF+ cell bodies in the MePD and MePV that coexpressed ERa-LI was greater in female than male mice, while the proportion of EGPF+ cell bodies in the BST that coexpressed ERb-LI or AR-LI was similar in female and male mice ( Fig. 10C; Table 3). The proportion of ERa-ir cell bodies in the MePD and MePV that coexpressed EGFP was similar in female and male mice, as was the proportion of ERb-ir and AR-ir cell bodies that coexpressed EGFP ( Fig. 10D; Table 3).
As was the case in the BST, confocal analysis showed that EGFP-ir fibres and terminal-like structures were often adjacent to single-labelled ERa2, ERb-or AR-ir cell bodies in the MePD and MePV (Fig. 11G-I, 11J-L). As described previously, apparent contacts between EGFP-ir fibres and adjacent lying cell bodies were observed following immunohistochemistry against calbindin ( Fig. 11M-O).
Cell bodies immunoreactive for ERa and the AR were found in the hypothalamic nuclei containing EGFP+ cells, including the MPO, MPA, LH, PeF, MTu and Te (Figs. 12 and 13). ERb-ir cell bodies were also present in the MPO and MPA, as well as the arcuate nucleus, LH and adjacent MTu, PeF and Te. In doublelabelling experiments the few, sparsely distributed, EGFP+ cell bodies in the MPO did not coexpress ERa-LI (Fig. 12A, 12B, 12G-I). However, while some EGFP+ cells in the MPO coexpressed ERb-LI and AR-LI, EGFP+-only cell bodies were also identified (Fig. 12C, 12D Fig. 14J-O).
Thalamus. In the thalamus, some sparsely distributed EGFP+ cell bodies were found in the A13 region of zona incerta (ZI), in both male and female mice (Fig. 4M). A low density of weakly EGFP-ir nerve fibres was also observed in the ZI.

Discussion
As the last step in oestrogen synthesis is aromatisation and as most oestrogen effects are mediated through oestrogen receptors, attempts have been made to compare the distributions of aromatase and oestrogen receptors in the brain [29][30][31][32]. However, a description of the distribution of aromatase in the rodent brain, and its coexpression with oestrogen and androgen receptors in individual cells, has been hindered by the lack of sensitive and specific antibodies. This problem has been circumvented by tagging aromatase expression to reporter gene expression following the physiological activation of the Cyp19A1 gene.
In the transgenic mouse model developed previously by Wu et al. [33], the nuclear LacZ reporting gene was integrated into the aromatase locus and the reporter protein is nuclear b-galactosidase, which identifies aromatase positive cell bodies but not the neurites extending from the soma. Since aromatase is a endoplasmic reticulum protein [34] and the endoplasmic reticulum has been reported to be present in synapses [35], we have adopted a different approach -to tag aromatase expression by a EGFP, a cytoplasmic protein, so that neurites extending from aromatase expressing neurones can be identified. This is made possible by inserting the EGFP cDNA before the ATG start codon of the aromatase coding region in the Cyp19A1 gene. EGFP cDNA is expressed (i.e. transcribed and translated) whenever the aromatase gene is expressed. However, the resulting protein molecule of EGFP is not linked to the aromatase protein molecule, as indicted by the small molecular weight (,32.7 kDa) of the expressed EGFP band on the Western blot analyses. Aromatase is an integral membrane protein of the endoplasmic reticulum, anchored to the membrane by an amino terminal transmembrane domain, whereas EGFP is a cytoplasmic protein. Although it is not our intention to describe the subcellular location of aromatase, EGFP is a good marker for neurites extending from aromatase positive neurones as it is present throughout the cytoplasm of aromatase-positive neurones. This logic is strengthened by the fact that EGFP fibres could not derive from neurones not expressing aromatase. A second advantage of not covalently linking EGFP to aromatase protein is to not alter aromatase activity.
We have identified cells expressing aromatase by EGFP immunoreactivity. Immunohistochemistry for EGFP was performed using the TSA + technique [36], a highly sensitive modification of the original indirect technique developed by Coons and collaborators [37]. Antibodies to the ERa, ERb and AR have also been validated, with Western blotting revealing a single band of predicted size, and staining patterns in agreement with previous immunohistochemical and in situ hybridization studies in mouse and rat [38][39][40][41][42][43]. Thus, the ERa, ERb and AR antisera do not seem to be associated with the type of unspecificity described for antibodies to, for example, many GPCRs [44].
EGFP-LI was completely absent from brain tissue of wild-type mice in which transcription of EGFP was not induced by physiological activation of the Cyp19A1 gene. Further support that this approach provides a valid representation of aromatase protein comes from the concordance between the distribution of EGFP-LI in the brain of this mouse and earlier in situ hybridization and aromatase activity studies performed in mouse, rat and monkey [2,[6][7][8][9][10]18,19,21]. Specifically, in both sexes, aromatase expression was highest in cells and fibres of the bed nucleus of the stria terminalis, olfactory tubercle, and the posterodorsal and posteroventral parts of the medial amygdaloid nucleus. Nevertheless, differences with previous descriptions of the distribution of aromatase in rats were found: EGFP+ cell bodies in the hippocampal formation were uncommon and sparsely distributed (c.f. [2,18]); infrequent EGFP+ cell bodies were found in the central amygdaloid nucleus (c.f. [21]); and EGFP-LI was not found in the cerebellum (c.f. [11]). These discrepancies may represent species differences. It is also possible that aromatase (or EGFP) protein is expressed in these regions but at a level that is below detection by our immunohistochemical approach. Note that a truncated form of aromatase has previously been detected in the rat cerebral cortex [45,46], where little or no aromatase activity has been reported [19]. The location of the EGFP promoter used in this study would prevent the identification of this truncated form of aromatase, which has not been identified in human or mouse.
Our data, for the most part, corresponds well with the distribution of aromatase obtained from in situ hybridization data in the GENSAT atlas [27] and the Allen Brain Atlas [28]. For example, EGFP+ cell bodies and those expressing the Cyp19A1 transcript were observed in the ventral striatum, BST and the medial amygdaloid nucleus. On the other hand, EGFP+ cell bodies were found in the lateral septal nucleus, but no Cyp19A1 transcript was reported in this nucleus. We contend that EGFP-LI allows a more precise description of the specific nuclei in which cell bodies expressing aromatase are distributed, and thus expands on descriptions provided by GENSAT and the Allen Brain Atlas. However, a direct comparison is difficult because distribution of the Cyp19A1 transcript in these atlases is obtained from sagittal sections, whereas our results were derived from examination of sections cut in the coronal plane.
While the expression of EGFP-LI within the various cell bodies of the hypothalamus reflected previous in situ hybridization [13,47] and aromatase activity findings [6,19], there was frequently a high density of EGFP-LI outside cell bodies, whose distribution corresponded to fibres or terminals. We suspect that these EGFP-ir fibres/terminals may reflect fibre projections originating from the BST and medial amygdala [48] to the various hypothalamic nuclei, in particular, e.g. the medial preoptic nucleus/area, and the premammillary, medial tuberal and supraoptic nuclei, which contained the highest density of EGFPir fibres. Stronger EGFP-LI detected in the male hypothalamus may infer an increase in arbourisation of projections from distant aromatase positive neurones, such as the medial amygdaloid nucleus.
EGFP-LI and EGFP+ cell bodies were significantly higher in the BST and the medial amygdaloid nucleus of male mice than in the corresponding nuclei in female mice, and is in keeping with previous findings in a transgenic mouse with a LacZ reporter cassette inserted into the 39UTR of the Cyp19A1 gene [33]. Thus, it appears that there are more aromatase-expressing cells in these regions of the male mouse brain. Previously, differences in the number of cell bodies in the medial amygdaloid nucleus (and BST) of male and females was explained by sex-specific apoptosis, because more TUNEL-positive cells were found in females between P1 and P10 than in males of the same age [33,49]. To summarise our data, adult female mice have fewer EGFP+ cell bodies in the BST that coexpress ERa, ERb or the AR, and fewer EGFP+ cell bodies in the MePD/MePV that coexpress ERb or the AR than adult male mice. This suggests that cells in the BST and MePD/MePV of female mice at P1-P10 that undergo apoptosis may be aromatase-containing cell bodies that coexpress ERa, ERb and/or the AR. EGFP/ERa coexpressing cell bodies in the MePD/MePV may be an exception because their numbers in adult female and male mice were similar.
Of the three steroid receptors examined, ERb-LI was most commonly coexpressed in EGFP+ cell bodies in the BST and MePD/MePV of female and male adult mice. We propose two Table 3. Estimated number of EGFP+, ERa-ir, ERb-ir and AR-ir cell bodies in the MePD/V (mean 6 SEM). functional implications of this high incidence of ERb coexpression. The first relates to neurone survival. As discussed above, cells in the BST and medial amygdaloid nucleus of P1-P10 female and male mice undergo apoptosis [33,49]. Oestrogen signalling though the ERb may regulate the number of neurones in the BST and medial amygdaloid nucleus by promoting survival and reducing apoptotic cell death at this age. Indeed, Hisasue et al. [50] demonstrated that ERb specific agonist diarylpropionitrile (DPN) increased BST cell number although not volume in C57BL6 female mice. If so, we expect the proportion of EGFP+ cell bodies    expressing ERb-LI to be similar in the BST and MePD/MePV of adult and newborn mice. Whether this is the case should be clarified in future investigations as it had been reported that ERbKO but not ERaKO mice maintain a similar sexually dimorphic BST as wild-type mice [51]. The second implication that we propose is that oestrogens, signalling through ERb, influence aromatase expression in neurones in the BST and MePD/MePV of the adult mouse, and can thus promote male behaviours [52]. In view of the close proximity of the aromatase or rather EGFP labelling and oestrogen receptor labelling observed, this may be through an autocrine mechanism, where oestrogens produced by aromatase-expressing cells act on ERb expressed by these cells (,88% and ,75% of EGFP+ cells coexpress ERb in the female and male MePD/MePV, respectively), or through paracrine signalling, where oestrogens released by non-ERb/ aromatase-producing cells act on aromatase/ERb coexpressing cells (,12% and ,25% of EGFP+ cells were not found to express ERb in the female and male MePD/MePV, respectively). By discussing possible functional attributes of ERb, we do not preclude roles for ERa or the AR in these events. In fact, Wu and colleagues [33] reported that in newborn mice of both sexes, 99% of b-galactosidase positive (aromatase) cells co-express ERa in the BST and medial amygdaloid nucleus. In adult mice however, our current study demonstrates that only ,35% of EGFP+ cells coexpressed ERa in the male MePD/MePV (,29% in male BST), while ,62% of EGFP+ cells coexpressed ERa in the female (,60% in female BST); both are significantly lower than reported previously in newborn mice [33]. Changes in these proportions, and the higher levels of coexpressing EGFP/ERa cell bodies in female mice in particular, indicate that signalling through ERa is also important in regulating cell survival, levels of aromatase and male behaviours [52,53]. However, a single cell body can express more than one of the three (ERa, ERb or the AR) receptors, thus adding complexity to this discussion. In fact, a single cell may express all three receptor subtypes and regions of the female and male postnatal brain expressing ERa and ERb overlap [54]. Thus, to gain further insight on the precise roles for ERa, ERb and the AR, further studies are required to compare the number, and proportion, of aromatase cells expressing ERa, ERb and/or the AR in the newborn/postnatal and adult brain.
The distribution of aromatase positive neurones observed in our study is similar to that described in Wu et al. [33] and we proceeded to examine how this EGFP-positive fibre distribution pattern integrated with oestrogen and androgen receptors. As the EGFP is not fused to the aromatase protein, which is endoplasmic reticulum bound, there is the advantage that fibres of aromatase expressing cells can be identified. This provides clues to the location of terminals and fibres of aromatase expressing cells with respect to the cell somata that were immunoreactive for ERa, ERb or the AR. Confocal microscopy revealed EGFP-ir structures, with a fibre/terminal-like appearance, lying in close proximity to cell bodies immunoreactive for ERa, ERb or the AR. This observation supports the current view that oestrogens derived from neural aromatisation of testosterone have paracrine, intracrine or autocrine influences on ERa and ERb [26,[55][56][57]. The coexpression of AR-LI and EGFP-LI (i.e. aromatase) in cell bodies supports previous reports that androgen regulates brain aromatase expression [14,21,47,58].
As neural aromatisation of testosterone is thought to have organisation roles i.e. to masculinise and defeminise the brain [55,56], future analyses of aromatase-EGFP expression in our transgenic mouse embryonic brains will address whether there is a sexually dimorphic expression pattern similar to the adult brains. The high levels of aromatase in the BST and the medial amygdaloid nucleus of adult male mice suggest that these nuclei are important in masculine behaviours, i.e. activational roles as supported by reports that male sexual behaviour is impaired in the male aromatase knockout (ArKO) mouse [59] but can be restored in adulthood by oestrogen replacement [60]. Furthermore, we can postulate that the sexual differences of aromatase, oestrogen receptors and androgen receptor described here in gonadally intact animals may be altered by circulating hormones. Future studies will be conducted to investigate the changes in the distribution during hormonal manipulation either by gonadectomy or hormone/agonist/antagonist administration.
There were moderate levels of EGFP expression in the caudomedial regions of the caudate putamen and the A13 region of the zona incerta of the thalamus. Aromatase expression has been previously described in the caudate and putamen and thalamus of the human brain in a positron emission tomography study using the radiolabelled aromatase inhibitor [N-methyl-(11)C]vorozole [61].
In summary, the widespread distribution of EGFP+ cell bodies and immunoreactive fibres suggests that aromatase expression is common in the mouse brain and is in close proximity to oestrogen receptors. This supports the long standing notion that locally synthesized oestrogens mediate biological effects by paracrine and autocrine activation of pre-and post-synaptic oestrogen a and b receptors, and androgen receptors, and also by projections to distant nuclei, particularly in the hypothalamus. The sex differences in aromatase expression in regions such as the amygdala and BST suggest that these nuclei may be important in producing differences in male and female behaviours.

Animals
All experimental procedures performed in this study conformed to the Australian National Health and Medical Research Council published code of practice, and were approved by the Animal Ethics Committee of The Florey Institute of Neuroscience and Mental Health (#09-024). The transgenic CYP19A1-EGFP BACmouse strain (Swiss 6X FVB/N) was obtained from the Mutant Mouse Regional Resource Centers at University of California, Davis, USA. An EGFP gene was inserted upstream of the ATG start codon of the Cyp19A1 gene of the selected BAC clone [62]. The transgenic mouse line was back-crossed onto FVB/N background for .8 generations and maintained under specific pathogen-free (SPF) conditions on a 12 h day/night cycle, with water and soy-free food (Glen Forrest Stockfeeders, Glen Forrest, Western Australia, Australia) ad libitum. PCR genotyping was performed using primers: EGFP-forward 59 CCTACGGCGTGCAGTGCTTCAGC 39 EGFP-reverse 59 CGGCGAGCTGCACGCTGCCGTCCTC 39.
Both male and female hemizygous transgenic animals have a gross anatomy, fertility, life span and hormonal profile similar to wild-type animals (data not shown). Eleven mature gonadally intact 16-20-weeks old hemizygous mice (6 male and 5 female), and six wild-type mice which do not have the EGFP transgene (3 male and 3 female), weighing between 35-40 g, were used. The weight of the transgenic mice was similar to wild-type littermates. In addition, the FVB/N mouse strain is not classified as obese, and these mice are less susceptible to obesity development than C57BL6 mice [63].

Western blot analysis
Tissue preparation. All animals were deeply anaesthetised with pentobarbitone sodium (Lethabarb, Virbac, Milperra, NSW, Australia, 100 mg/kg i.p.) and killed by cervical dislocation. Brains were quickly removed and dissected, and samples snap frozen on dry ice and stored at 280uC until further processing.
Sample preparation. Tissue was homogenized by ultrasonication in TNE buffer (5 mM Tris-HCl, 0.5 mM EDTA, 0.1 M NaCl) also containing 0.32 M sucrose and protease inhibitors. Cell debris and nuclei were removed by centrifugation (800 g, 10 min at 4uC). Membrane and cytosol fractions were separated by centrifugation at 14,000 g for 30 min at 4uC. Membrane fractions were washed in TNE buffer containing protease inhibitors, and digested in TNE buffer supplemented with 1% Triton X-100 and protease inhibitors for 90 min. Undigested tissue was removed by centrifugation at 14,000 g for 30 min at 4uC. Protein concentrations were determined by Bradford's colorimetric method [64].
Gel electrophoresis and protein detection. Samples were diluted to a final protein concentration of 2 mg/ml in Laemmli buffer, and analyzed by SDS-PAGE on 10% resolving acrylamide gels. After transferring onto Odyssey nitrocellulose membranes (LI-COR), protein samples on membranes were blocked in 5% (w/v) non-fat milk in Tris-buffered saline solution containing 0.5% Tween-20 (TBST), pH 8.0, and exposed to primary antibodies  (Table 4)] diluted in blocking solution, at 4uC overnight. Hybridized membranes then were washed 3 times in TBST for 5 mins, and probed with Odyssey secondary antibodies, IRDye 800CW Goat anti-Rabbit IgG and IRDye 680RD Goat anti-Mouse IgG (LI-COR) (1:10,000). Immunoreactivity was visualised and analysed on the Odyssey Classic infrared imaging system (LI-COR; Fig. 15).

Primary antisera used
Enhanced Green Fluorescent Protein. Chicken anti-green fluorescent protein (GFP) polyclonal antibody was raised against the recombinant full length protein (Abcam, Cambridge, United Kingdom; Code No. ab13970; Lot No. 902074; Table 4). Immunoblotting recognises a single band of approximately 27-30 kDa (Abcam), and the antibody did not immunolabel brains of wildtype mice which do not have the EGFP transgene (data not shown). Western blot analysis of mouse brain extracts revealed a single band of approximately 30 kDa. Staining patterns obtained using this antibody were consistent with those previously reported [67].
Oestrogen receptor a. Rabbit anti-oestrogen receptor a (C1355) polyclonal antibody, raised against a KLH-conjugated synthetic peptide representing the last 15 amino acids (TYYIP-PEAEEGFPNTI) of rat ERa (Millipore, Billerica, MA; Code No. 06-935; Lot No. NG1838275; Table 4). Western blot analysis of mouse brain extracts revealed a single band of approximately 66 kDa, the size of the rat ERa (Fig. 15). The staining patterns obtained using this antibody were consistent with those previously reported in rat [38] and mouse [39,40].
Oestrogen receptor b. Rabbit anti-oestrogen receptor b, clone 68-4 protein A purified monoclonal antibody, raised against a KLH-conjugated synthetic peptide corresponding to amino acids  Table 4). Western blot analysis of mouse brain extracts revealed a single band of approximately 59 kDa, the size of the rat ERb (Fig. 15). The staining patterns obtained using this antibody were consistent with those previously reported in mouse [40,41], and were similar to patterns reported by Shughrue et al. [68], whose antibody did not stain the brain of mice lacking ERb.
Androgen receptor. Rabbit anti-androgen receptor, PG-21 affinity purified polyclonal antibody, raised against a KLHconjugated synthetic peptide corresponding to amino acids 1-21 (MEVQLGLGRVYPRPPSKTYRGC) of the human AR (Millipore; Code No. 06-680; Lot No. DAM1661059; Table 4). Western blot analysis of mouse brain extracts revealed a single band of approximately 110 kDa, the size of the rat AR (Fig. 15). The staining patterns obtained using this antibody were consistent with those previously reported in mouse [42,43].
Calbindin. Mouse anti-calbindin D-28k monoclonal antibody, raised against calbindin D-28k purified from chicken gut [Swant, Bellinzona, Switzerland; Code No. 300; Lot No. 07 (F); Table 4]. Immunoblotting recognizes a single band of 27-28 KDa, and the antibody does not stain the brain of calbindin D-28k KO mice. The staining pattern of this antibody is consistent with results described previously [69,70].
Synaptophysin. Mouse anti-synaptophysin monoclonal antibody (clone SVP-38), derived from the hybridoma produced by the fusion of mouse myeloma cells and splenocytes, and using a synaptosome preparation from rat retina as the immunogen (Sigma; Code No. S 5768; Table 4). Immunoblotting recognises a single band of approximately 38 kDa (Sigma) [71,72].

Image Processing
After processing, sections were examined using a Leica DM LB2 fluorescence microscope (Leica, Wetzlar, Germany), equipped with a dark field condenser and epi-polarization, and epifluorescence with appropriate filter combinations, and with The FITC labelling was excited using the 473 nm diode laser. For the detection of AlexaFluor 555, a 559 nm diode laser was used. Z-stack images were captured with multiple images, each separated by a stepwise depth of 0.8-1.0 um in the z-plane.
Digital images from the microscopy were slightly modified to optimise for image resolution, brightness and contrast using Adobe Photoshop CS5 extended, version 12.0, software (Adobe Systems Inc., San Jose, CA), so as to best represent the immunohistochemistry observed at the microscope.

Cell Quantification
Estimates of the number of EGFP-positive (+), ERa immunoreactive (-ir), ERb-ir and AR-ir cell bodies in the bed nucleus of the stria terminalis (BST) were obtained from four 14 mm-thick sections, each 280 mm apart, ranging from Bregma +0.38 mm to 20.46 mm [73]. Components of the BST from which these cell bodies were quantified included its: lateral division -dorsal, intermediate, posterior and ventral parts; and medial divisionanterior, posterointermediate, posterolateral, posteromedial and ventral parts. Estimates were made using a fractionator sampling design according to optical disector rules [74][75][76][77][78], with guard zones of 1 mm (top) and 1 mm (bottom). Regular predetermined intervals (x = 150 mm, y = 175 mm) and counting frame dimensions (50650 = 2500 mm 2 ) were derived by means of a grid program (Stereoinvestigator v.7.0, MicroBrightField, Williston, VT, viewed through a microscope and 660 oil objective, Leica). The total number of cell bodies in the BST was estimated by multiplying the number of cell bodies counted within the sampling regions with reciprocals of the fraction of sections sampled, fraction of sectional area sampled, and the fraction of the section thickness sampled. EGFP-positive (+), ERa-ir, ERb-ir and AR-ir cell bodies were also quantified in the medial amygdaloid nucleus -posterodorsal (MePD) and posteroventral (MePV) parts. Regions of the MePD/ MePV from which cell bodies were quantified corresponded to Bregma 21.34 mm to 22.18 mm [73]. Estimates were made according to the procedure outlined for the BST above, except that regular predetermined intervals of x = 125 mm and y = 150 mm were employed.
All comparisons were conducted by student t-tests and a value of P#0.05 was considered statistically significant. Values are expressed as the mean 6 SEM.

Data Analysis
The presence of EGFP-like immunoreactivity (-LI) in cell bodies and nerve fibres throughout the mouse brain is summarised in Table 1. EGFP-LI in fibres was described as (+) low, (++) moderate, (+++) high, or (++++) very high density/intensity [79]. The density of EGFP-immunoreactive/positive (+) cell bodies is indicated as (#) sparse, (##) low, (###) medium, or (####) high. The Paxinos and Franklin Stereotaxic Atlas of the Mouse Brain [73] was used to define and label the specific regions expressing EGFP. In Figs. 3-5, 7-9 and 11-13, the stereotaxic coordinates indicating distance from Bregma are used only as a guide, and are given to allow the reader to refer to the Atlas mentioned above.