SEXUALLY DIFFERENTIATED AND NEUROANATOMICALLY SPECIFIC CO-EXPRESSION OF AROMATASE NEURONS AND GAD67 IN THE MALE AND FEMALE QUAIL BRAIN

Abstract

Testosterone aromatization into estrogens in the preoptic area (POA) is critical for the activation of male sexual behavior in many vertebrates. Yet, cellular mechanisms mediating actions of neuroestrogens on sexual behavior remain largely unknown. We investigated in male and female Japanese quail by dual-label fluorescent in situ hybridization (FISH) whether aromatase-positive (ARO) neurons express glutamic acid decarboxylase 67 (GAD67), the rate-limiting enzyme in GABA biosynthesis. AROcells and ARO cells double labeled with GAD67 (ARO-GAD67) were counted at standardized locations in the medial preoptic nucleus (POM) and the medial bed nucleus of the stria terminalis (BST) to produce three-dimensional distribution maps. Overall, males had more ARO cells than females in POM and BST. The numberof double-labeled ARO-GAD67 cells was also higher in males than in females and greatly varied as a function of the specific position in these nuclei. Significant sex differences were however present only in the most caudal part of POM.Although both ARO and GAD67 were expressed in the VMN, no colocalization between these markers was detected.Together, these data show that a high proportion of estrogen synthesizing neurons in POM and BST are inhibitory and the colocalization of GAD67 with AROexhibits a high degree of anatomical specificity as well as localized sex differences. The fact that many preoptic ARO neurons project to the periaqueductal gray in male quail suggests possible mechanisms through which locally produced estrogens could activate male sexual behavior.

Graphical Abstract

We investigated by dual-label fluorescent in situ hybridization in male and female Japanese quail whether aromatase neurons also express glutamic acid decarboxylase 67 (GAD67), the rate-limiting enzyme in GABA biosynthesis. An extensive colocalization was found in the medial preoptic nucleus and bed nucleus of the stria terminalis and this colocalization exhibited a high degree of anatomical specificity as well as localized sex differences. Aromatase neurons are thus mostly inhibitory.

INTRODUCTION

Estrogens, such as 17β-estradiol (E₂), are textbook examples of steroid hormones synthesized and secreted by the ovaries that exert diverse effects on morphology, physiology and behavior, especially as these processes relate to reproduction and aggression, but also on cognitive processes including learning and memory(Balthazart & Ball, 2013; McEwen & Milner, 2017). In the second part of the 20th century, it was recognized that specific brain sites are important targets of estradiol action both for the negative and positive feedback effects of estradiol on endocrine functioning and for the activation of reproductive behaviors(Pfaff, 1980). Estrogen receptors were localized in brain regions critical for these processes (Pfaff & Keiner, 1973; Morrell et al., 1975; Pfaff, 1976)and these receptors were shown to act, when occupied by their ligand, as transcription factors modulating the expression of a wide variety of genes(O’Malley & Tsai, 1992; Tsai & O’Malley, 1994).

The ovaries are a main source of estrogens, but E₂ can also be synthesized in the brain of both males and females, so that it becomes locally available in high concentrations (Naftolin et al., 1975; Balthazart & Ball, 2013). Androgens, such as testosterone of gonadal or adrenal origin, pass into the brain where they can be converted to an estrogen via the action of the rate-limiting enzyme for estrogen synthesis called aromatase (or estrogen synthase).This enzymatic process plays a critical role in the activation of numerous physiological processes and behaviors in males of a variety of mammalian and avian species(Balthazart & Ball, 2013). Evidence also indicates a role of this enzyme in the brain of females (Cornil, 2018).

Subsequently, it became clear that, in addition to its relatively slow action as a regulator of gene transcription in the cell nucleus, E₂ acts more rapidly in the brain with a time course more similar to what has been described for modulatory neuropeptides (Balthazart & Ball, 2006). This work gave rise to the concept of dual action of estrogens indicating that brain E₂ actsin the cell nucleus with a time course typical for a steroid hormone (e.g., hours to days to weeks) as well as more rapidly at the cell membrane, like a neurotransmitter (e.g., within minutes), and these complementary actions control different aspects of a same physiological or behavior response(Saldanha et al., 2011; Cornil et al., 2015). In parallel, the activity of brain aromatase is also regulated in a slow manner (hours to days) via changes in enzyme concentration and more rapidly (minutes or even shorter) via changes in the phosphorylation status of the enzymatic protein (Balthazart & Ball, 2006; Charlier et al., 2011). This complex modulation of E₂ production and action in the brain makes an understanding of cells that are producing and are a target of this steroid especially important.

Aromatase has been localized in brain cells via immunohistochemical (IHC) methods to identify the protein as well as in situ hybridization histochemistry(ISH) to identify mRNA expression in specific brain regions in many species (Roselli, 2013; Saldanha et al., 2013).Most of the initial IHC work was actually performed on avian species in which aromatase detection appeared to be easier, presumably because they exhibit higher concentrations of brain aromatase (Balthazart et al., 1990; Foidart et al., 1995; Saldanha et al., 2000). Only a few studies using carefullyvalidated antibodies (Beyer et al., 1994; Garcia-Segura et al., 1999)and reporting systems (Wu et al., 2009; Stanic et al., 2014)have succeeded in localizing the aromatase protein in the brain of some mammals.

In birds, the medial preoptic nucleus (POM) constitutes one of the largest and best-documented populations of neurons expressing aromatase. Preoptic estrogen synthesis has long been known for its key role in the activation of male sexual behavior in birds (Panzica et al., 1996; Balthazart et al., 2004)and in rodents (Hull & Rodriguez-Manzo, 2009). It is also suspected to play a key role in the acute regulation of male sexual behavior as studies conducted in Japanese quail showed that (1) intracerebroventricular injection of an aromatase inhibitor in the close vicinity of this brain region induces within 30 min a prominent inhibition of appetitive sexual behavior (Seredynski et al., 2013; Seredynski et al., 2015), (2) sexual interaction results within minutes in changes in preoptic aromatase activity (Dickens et al., 2011; de Bournonville et al., 2013; Dickens et al., 2014)and (3) rapid changes in local estrogen concentration have been measured in the medial preoptic nucleus within 10 min of a sexual interaction with a female(de Bournonville et al., 2017b). However, how these brain-derived estrogens produced in the preoptic area acutely modulate sexual behavior and other behaviors regulated by estrogens is still largely unclear. In particular, the neurochemical nature of aromatase expressing neurons and its possible functional significance is not well understood.

In the present study we investigated the chemical phenotype of aromatase expressing cells in the POM and the medial part of the bed nucleus of the stria terminalis (BST). We focused on a marker of the inhibitory transmitter gamma aminobutyric acid (GABA), namely the synthesizing enzyme glutamic acid decarboxylase67 (GAD67). It has been hypothesized that, in rodents,steroid action on sexual behavior in thepreoptic area is mediated at least in partvia projections of this nucleus to the periaqueductal gray (PAG) in the midbrain(Murphy & Hoffman, 2001; Hull et al., 2002). PAG would exert a tonic inhibition on sexual behavior via its downstream projections (Brackett et al., 1986)and steroid action in the preoptic areacould suppress this tonic inhibition(Hull et al., 2002). Our tract-tracing studiesin quail demonstrating dense projections to the PAG of aromatase-immunoreactive neurons of the POM were clearly consistent with this possible function (Absil et al., 2001; Carere et al., 2007)and indicated that a better neurochemical characterization of these aromatase neurons would be potentially valuable. A few studies in recent years initiated investigations of the chemical phenotype of aromatase expressing cells in other brain areas. For example, in the auditory telencephalon of a songbird, the zebra finch (Taeniopygia guttata), it was found that about 35% of aromatase cells co-express GAD65 (Jeong et al., 2011b). A similar study also found in the zebra finch auditory forebrain that aromatase cells co-express the calcium binding protein parvalbumin, but not calbindin (Ikeda et al., 2017). In mice, it has recently been shown that most aromatase-positive cells in the principal component of the BST and medial amygdala co-express GAD1, the gene coding for GAD67 (Unger et al., 2015; Bayless et al., 2019). We therefore employed a double-label fluorescentin situ hybridization (FISH) histochemistry method to localize aromatase and GAD67 in two brain nuclei that contain dense populations of aromatase-positive cells, the POM and BST of male and female Japanese quail (Coturnix japonica)and test whetheraromatase expressing cells tend to be inhibitory neurons in the POM and/or BST. The ventro-medial nucleus of the hypothalamus that contains another dense population of aromatase-expressing neurons was also analyzed in a few brains(Foidart et al., 1995). Quail were used in these studies because this species has emerged as a useful model species to investigate the function of aromatase in relation to the expression of species-typical sexual behaviors (Ball & Balthazart, 2010; Cornil, 2019).

MATERIAL AND METHODS

Subjects and brain collection

Experiments were carried out on sexually mature males (n=10) and females (n=9) purchased from a local breeder at the age of 5–6 weeks. Birds were kept in individual cages in the laboratory until they reached sexual maturity, which is confirmed in males by the presence of a large cloacal gland indicative of high testicular activity (Sachs, 1967)and in females by regular egg laying. Birds were exposed to a photoperiod of 16 hours of light and 8 hours of dark per day which stimulates quail to be in a reproductive state. Food and water were always available ad libitum. Experiments complied with the Belgian laws on “Protection and Welfare of Animals” and on the “Protection of experimental animals”. Experimental procedures were approved by the Ethics Committee for the Use of Animals at the University of Liège (Ethic protocol #683).

Birds were injected in the wing vein with 200μl of heparin solution (20 mg/ml; sigma H3393) and were deeply anesthetized with an overdose of ketamine (40mg/kg) and xylazine (2.5 mg/kg) injected in the pectoral muscles. They were then perfused intracardially with saline solution (NaCl 0.9%) followed by 4% paraformaldehyde prepared in 0.1 M phosphate buffer (PB, pH 7.2). Brains were immediately dissected out of the skull and cryoprotected overnight in a 30% sucrose solution in 0.1 M PB (pH 7.2) at 4°C. Brains were then frozen on dry ice and stored at −80°C until used.

Riboprobes

The 489 bp aromatase sequence that matches nucleotides 260–748 of the previously cloned aromatase sequence of Japanese quail (Genbank no. AF533667) (Voigt et al., 2007)was expressed in a pGEM7 ZF plasmid. The GAD67 sequence was obtained from ARK genomics (Ref# ChEST886c5, BBSRC ChickEST Database, http://www.chick.manchester.ac.uk) and expressed in a pBluescript II KS+ plasmid. This 902 bp sequence matches nucleotides 350–1252 of the previously cloned GAD67 sequence of chicken (Genbank no. AF030355), covering 26% of the full length RNA sequence with an identity of 99.78%.

The sense and antisense probes were obtained following linearization of the plasmids and in vitro transcription with the appropriate restriction enzymes (Aromatase Antisense: EcoRI+SP6, Sense: BamHI+T7; GAD67 Antisense: Not1+T3, Sense: EcoR1+T7) and RNA polymerase. The transcription mixture included 1μg of linearized plasmid, 2μl of a mix of nucleotides containing ATP [10mM], GTP [10mM], CTP [10mM], UTP [6.5mM] and digoxigenin(DIG)-11-UTP [3.5mM] or fluorescein(FITC)-UTP [3.5mM] from Promega, 0.5μl of RNAse inhibitor and 1μl of RNA polymerase in 20μl of buffer. The transcription reaction was performed at 37°C for 2 hours after which the RNA probes were purified with microspin G-50 columns (Kit IllustraTM, GE Healthcare). The quality of the synthesis was monitored after migration on electrophoresis gels.

Fluorescent in situ hybridization (FISH)

Brains were sliced in 20μm thick coronal sections on a cryostat. Sections were collected from the level of the septopallio-mesencephalic tract (TSM) to the end of third nerve and mounted on superfrost® slides in 4 series so that consecutive sections on a same slide were 80μm apart.

These sections were labeled withFITC- or DIG-labeled RNA probes that were visualized respectively with Alexa 488 (green) and HNPP-Fast red (red). All steps were performed at room temperature unless otherwise indicated. Briefly, brain sections were post-fixed for 10 min in a 4% paraformaldehyde solution in 0.1 M phosphate-buffered saline (PBS). After rinses in 0.01 M PBS, brain sections were incubated in an acetylation solution, which consisted of 1.00 mM triethanolamine (pH8) and 0.25% acetic anhydride in water and were rinsed twice (5 min) in PBS with tween 0.1%. Sections were then pre-incubated in hybridization buffer (Lucron Amresco) for 60 min in a water bath at 70°C. Pre-heated hybridization buffer containing the RNA probes (1μg/ml) was applied onto each glass slide (200μl) that were then coverslipped and incubated overnightat 70°C. The next day, sections were rinsed twice in pre-warmed washing buffer (50% formamide, 2x SCC 0.15 M NaCl, 0.015 M Na-citrate and 0.1% Tween), 4 times in Tris Buffer (100mM HCl, 150mM NaCl and 0.1% Tween) and blocked with NGS 10% prepared in Tris buffer for 30 min. They were then incubated overnight at 4°C under coverslip with a sheep anti-DIG antibody conjugated with alkaline phosphate (1:500, Vector laboratories) and a rabbit anti-FITC antibody (1:250, Roche Applied Science) in Tris buffer containing 10% NGS. On the third day, slides were rinsed three times in Tris buffer for 10 min, then twice in PBS for 5 min. Endogenous peroxidases were inactivated with 1% hydrogen peroxide in PBS for 45 min in the dark and rinsed thrice in PBS. FITC probes were visualized by tyramide signal amplification following manufacturer instructions (Invitrogen, Eugene, OR, USA): 1 hour incubation with 1% blocking solution, 2 hour incubation with 1:100 goat anti-rabbit antibody conjugated with horseradish peroxidase (HRP; Molecular probes) in blocking solution, 3 rinses (5 min) in PBS and 1 hour incubation in Tyramide-Alexa 488 (1:100) in amplification buffer with 0.0015% hydrogen peroxide. After 3 rinses (5 min) in PBS, sections were incubated 3 times in detection buffer (100 mM Tris HCl, 100 mM NaCl and 10 mM MgCl₂) before visualization of DIG probes using HNPP fluorescence detection set (Roche Applied Science). Slices were incubated for 30 min in the dark in the HNPP/Fast Red TR mix prepared according to manufacturer instructions and rinsed twice for 3 min in detection buffer. This step was repeated twice. Slides were then washed for 10 min in distilled water and coverslipped with Vectashield Hard set fluorescence mounting medium containing DAPI (Vector laboratories). In addition to sense strand hybridization controls, control slices were incubated in the absence of anti-DIG or anti-FITC antibody to control for the specificity of the labeling.

In order to rule out an effect of the sensitivity of the fluorochrome used to detect one messenger as compared to the other, aromatase was detected with the FITC probe and thus visualized in red and GAD67 was detected with DIG probe and thus visualized in green in 4 females and 8 males, while aromatase was detected with the DIG probe and visualized in green and GAD67 was detected with the FITC probe and visualized in red in 5 females and 2 males. The total numbers of ARO and ARO-GAD67 cells as well as the percentages of ARO⁺/GAD⁺ cells at each of the 5 brain levels considered (3 levels in POM and 2 levels in BST) were analyzed by separate two-way ANOVA with the sex of the birds and the fluorochrome used for each type of mRNA as factors. These 15 ANOVA identified no significant effect of the fluorochrome and no interaction between fluorochrome and sex (all p>0.10, except for one p=0.0709). These two sets of results were therefore pooled in all subsequent analyses.

Quantification of signals

Sections were photographed with the 20x objective of an epifluorescence microscope (Leica DMRB) or with a Nikon Eclipse TE2000-U microscope coupled to a camera. The focus of each photomicrograph was adjusted so that a maximum of aromatase cells were in focus. Each field (590μm x 472μm) was photographed twice under the two illumination wavelengths activating the two fluorophores labeling aromatase and GAD67.

Photomicrographs were collected and aromatase-positive cells and of aromatase cells expressing GAD67 were quantified on both sides of the brainat three rostro-caudal levels of the medial preoptic nucleus (POM) and two rostro-caudal levels of the medial portion of the bed nucleus of the stria terminalis (BST)(see Figure 1). The ventromedial nucleus of the hypothalamus (VMN) was also investigated at one single rostro-caudal level in a few birds (2 males and 4 females).

Figure 1:

Illustration of the procedure used to determine the tridimensional distribution of the cells expressing aromatase mRNA and/or GAD67 mRNA in the medial preoptic nucleus (POM) and the medial portion of the bed nucleus of the stria terminalis (BST). The left column of the figure shows schematic drawings of the quail brain at 4rostro-caudal levels illustrating the position of photomicrographs that were collected. Rectangles on the left side of each figureillustrate the position of the photographed fields, but note that similar fields were photographed on both sides. Similarly the distribution of cells expressing aromatase mRNA is illustrated by black dots on the right side of each figure, but of course these cells were present on both sides. The right column of the figure illustrates how each quantification field was divided in 9 or 18(2×9)rectangles for plotting the position of positive cells organized in 3 columns in the medio-lateral axis (from lateral [Lat], intermediate [Int] to medial [Med]) and 3 or 6 rows in the dorso-ventral axis (Levels L1 to L3 or L1 to L6). Four rostro-caudal levels were investigated. Their position was based on the maximal extension of the commissure anterior (CA). Pictures were taken at CA, CA-2 (2 sectionsrostral to the section with CA), CA-4 (4 sections rostral to the section with CA) and CA+2 levels (2 sections caudal to the section with CA). For more information, see the method section. Abbreviations: FPL, fasciculus prosencephali lateralis (lateral forebrain bundle); TSM, septopallio-mesencephalicus tract; SM septum medialis; VL, ventriculus lateralis.

The three levels of the POM were selected based on their position relative to the section where the commissura anterior (CA) reaches its largest extension approximately corresponding to Plate A8.2 in the chicken atlas (Kuenzel & Masson, 1988). In this section containing the largest extent of the CA, two pictures were takenon each brain side: the first most dorsal picture was located in the corner formed by the ventral edge of the CA and the lateral edge of the third ventricle. The second picture was located immediately ventrally to the first one. Two adjacentpictures were similarly collected on each sideof the brain in the section located 160μm rostral to CA (i.e., two sections more rostral, since one section was stained every 80 μm, level CA-2).These two pictures were centered in the dorso-lateral axis on the aromatase-positive cell group and their lateral edge was aligned with the edge of the third ventricle. Finally, in the section located 320μm anterior to CA (where the aromatase cells assume a more ventral position dorsallyadjacent to the floor of the brain (level CA-4), the photographic fieldwas centered on the middle of the group of aromatase-positive cells in each side of the brain.

In the BST, one image was digitizedin the section located at the largest extension of the anterior commissure (CA). In this section, the aromatase cells are located dorsal to the anterior commissure in a position ventral to the tip of the lateral ventricle. The image was centered in each side of the brain on the middle of the cell group expressing the aromatase messenger. The maximal extension of the BST was located 160μm caudal to the maximal extension of the CA (CA+2). Two pictures were taken at this level on each sideof the brain to entirely cover the characteristic “V” shape assumed by this nucleus. A first field was placed at the corner determined by the ventral and medial edge of the aromatase cells group. The quantification area was then moved one field height more dorsally (472μm) along the third ventricle and one-half field laterally (295 μm).

In the VMN, one quantification field was digitized in each male and female subject in a section containing the dorsal supraoptic decussatio, corresponding approximately to plate A6.4 in the chicken brain atlas (Kuenzel & Masson, 1988). The field was centered on the middle of the area labeled for aromatase.

Digitized images obtainedat the exact same positionfor the two probes were automatically superimposed as two separate layers in Adobe® Photoshop CS. Brightness and contrast were slightlyadjusted in all images to optimize detectionof the cell boundaries, but without revealing new labeled cells that were not initially visible. Cells expressing aromatase mRNA were identified and marked by an arrow on a third layer in Photoshop. Cells were considered as positively labeled when fluorescent signal was detected in the cytoplasm and the round unlabeled shape of the nucleus was identifiable, indicating that the cell of interest was clearly in focus. Whether these aromatase-positive cells also expressed GAD67 mRNA was then determined by inspecting the corresponding GAD67 image layer.Co-labeling was accepted if and only if a) the black nucleus was also visible in the GA67 image and b) the specific shape of the perikarya was identical in both the aromatase and the GAD67 signal.

In order to obtain a spatial representation of the distribution of aromatase cells and cells expressing both aromatase and GAD67, the location of all positive cells was mapped within eachphotomicrograph. A transparent grid was added as a fourth layer on the Photoshop file to allow counting positive cells in spatially identified subregions (see Figure 1, right side). The grid consisted of 3 X 3 rectangles in the medio-lateral axis (identified as Medial [Med], Intermediate [Int] and Lateral [Lat]) and the dorso-ventral axis (identified as levels 1, 2 and 3 [L1, L2 and L3]). Each of these rectangles was 197μm in widthand 157μm in height.

In the sections where two adjacent pictures were taken (POM at CA and CA-2 levels and BST at CA+2 level), both the dorsal and ventral fields were analyzed together so that data were distributed in 18 rectangles organized in 3 columns along the medio-lateral axis and 6 rows along the dorso-ventral axis (relabeled L1 to L6). The number of cells expressing aromatase and both aromatase and GAD67 were counted in each rectangle. Cells that were located at the border of two rectangles were included in the rectangle containing the largest portion of their surface.

Statistical analyses

Data were analyzed with Statistica 13.0 (Statsoft) or Prism 8 (Graphpad Software Inc). Quantifications obtained from both sides of the brain were averaged before all analyses. The total number of aromatase-positive cells and the number of aromatase cells also expressing GAD67 were used to compute a percentage of aromatase cells expressing GAD67 referred to as the percentage of colocalization (%ARO⁺/GAD⁺). This percentage was computed separately in each subject for each rectanglein the grid and also for the total data in POM or BST at each rostro-caudal level.

The first series of analyses compared the number of cells labeled for aromatase as well as the percentage of aromatase cells expressing GAD67 in the POM and BST based on a two-way ANOVA procedure in which the two sexes and the rostro-caudal position served as independent factors (Fig. 3). Similar analyses were also performed to test for the effect of the staining sequence on the final results (see above). A second series of analyses further investigated sex differences across the medio-lateral and dorso-ventral axes by separate three-way ANOVAs for each levelalong the rostro-caudal axis with sex as an independent factor and position on the two axes as repeated factors. All posthoc analyses were analyzed by Tukey HSD tests. Because the quantified fields were different at each rostro-caudal level (1 versus 2 photographic fields), data were analyzed separately for each level. Due to various technical problems (poor staining, torn or damaged sections) accurate counts could not always be performed for each subject and nucleus as reflected by the different degree of freedom among the analyses. Data are expressed as mean ± SEM and effects were considered significant for p < 0.05.

Figure 3:

Sex differences in the total number of cells expressing aromatase mRNA (A-B), the total number and the percentage of cells co-expressing aromatase and GAD67 mRNA (respectively C-D and E-F) at different rostro-caudal levels in the medial preoptic nucleus (POM) and the bed nucleus of the stria terminalis (BST). The numbers in the histogram bars represent the number of subjects. * and ** represent p < 0.05 and 0.01 vs CA-4 level within same sex, or otherwise mentioned, and Δ p < 0.05 vs male within same rostro-caudal level, or otherwise mentioned, (Tukey HSD post-hoc tests, following significant main effect of rostro-caudal level in the mixed design ANOVA whose results are summarized in the individual boxes where *, ** and *** representp < 0.05, 0.01 and 0.001, respectively and (*) represents a statistical trend (p<0.10). N.S.:= non-significant.

RESULTS

As illustrated in figure 2, the cluster of cells expressing aromatase mRNA defined on each side of the brain an almond-shaped structure characteristic of the POM with a higher density of cells located in the lateral part of the nucleus (Fig. 2A). Similarly, the dense cluster of aromatase-positive cells in the BST outlined the boundaries of a V-shaped structure centered on the brain midline and the left branch of this V is illustrated in Figure 2G showing positive cellsthat extend from the dorso-lateral to the ventro-medial part of the nucleus. This distribution closely corresponds to the pattern observed by immunohistochemical techniques analyzing neurons expressing the corresponding protein (Foidart et al., 1995). Interestingly, in both POM and BST aromatase-positive neurons seemed to form clusters of 2 to 4–5 cells similar to those described by Ikeda and colleagues in the auditory cortex of the zebra finch (see Fig. 2D and 2G), but no further attempt was made here to quantify this phenomenon (Ikeda et al., 2017).

Figure 2:

Representative photomicrographs illustrating the detection by in situ hybridization of aromatase mRNA (red signal in A, D, G) and GAD67 mRNA (green signal in B, E, H) in the POM (2 top rows) and BST (bottom row) of a female quail. The last column illustrates the superposition of the two signals and the white arrows represent cells that were counted as double labeled. The magnification bar is equivalent to 100μm in panels A-C (photomicrographs taken with the 10x objective) and 50μm in the 6 other panels (40X objective).

Cells densely expressing the GAD67 mRNA were present in high numbers in these two nuclei containing aromatase-positive cells (Fig. 2 B, E, H) and overlaying the two signals indicated that the two mRNA were often expressed in the same cells (Fig. 2C, F, I).

Quantification of aromatase cells

The total number of cells expressing aromatase mRNA at a given rostro-caudal level was first analyzed separately in the POM or BSTusing two-wayANOVAs. Because some sections had been lost or damaged during collection or processing (thus preventing repeated measure analysis), these datasets were analyzed using both sex and the rostro-caudal levels as independent factors. In the POM, the results confirmed that males had significantly more aromatase cells than females (F_1,41 = 9.842, p = 0.003) and that the number of these cells was different as a function of the position in the rostro-caudal axis (F_2,41 = 7.089, p <0.003; Fig. 3A). The interaction between these two factors was not significant (F_2,41 = 2.527, p = 0.092). Post hoc tests indicated that aromatase-positive cells were less abundant in CA-4 than at the two more caudal CA-2 and CA) levels.

In the BST, the ANOVA revealed a significant difference between the two rostro-caudal levels (F_1,30 = 66.56, p < 0.001) and suggested the existence of a sex difference (F_1,30 = 3.159, p = 0.086) and an interaction (F_1,30 = 3.135, = 0.087; Fig. 3B), but these two effects were not statistically significant. It should be noted that the presence of more cells in the caudal part of these two nuclei corresponds to the fact that the clusters of positive cells were larger at these levels which explains why two adjacent fields had to be photographed to cover the entire nucleus. If dataof POM_CA-2, POM_CA and BST_CA+2 were divided by 2, numbers would be roughly equivalent to those found in POM_CA-4 and BST_CA, respectively. The density of aromatase-positive cells is thus roughly similar along the rostro-caudal axis of the nuclei, but the size of both POM and BSTincreases in the caudal direction.

In the single section of VMN that was analyzed (Fig. 4), females displayed about twice as many aromatase-expressing cells (10.88 ± 2.77, n = 4) than males (5.50 ± 2.50, n = 2), but given the low number of subjects this difference was not significant (t₄=1.213, p = 0.2918).

Figure 4:

Representative photomicrographs illustrating the absence of colocalization of the aromatase (red) and GAD67 (green) mRNA as by in situ hybridization in the ventromedial nucleus of the hypothalamus. These photomicrographs illustrate a section from a female, but a similar situation was observed in both sexes. The magnification bar is equivalent to 50μm in panels A-C (photomicrographs taken with the 40x objective).

Color-coded heat mapspresented in figure 5 provide a visual representation of the distribution of the number of aromatase-positive cells at each rostro-caudal level of the POM and the BST in both sexes (Fig. 5). The differences between males and females are illustrated in separate maps in the rightmost column of the figure, in which different shades of blue or red represent the degree of differencesfavoring males (blue) or females (red). Overall, the distribution of positive cells along the 3 axes of the two nuclei appeared fairly similar in males and females, although males tended to possess more positive cells than females, which is clearly reflected in the large dominance of blue color in the subtractive maps on the right side. Specifically, in the POM, the highest densities of aromatase-positive cells were observed in the intermediate (CA-2) and caudal (CA) level, in particular in males,where positive cells tended to concentrate in the four intermediate dorso-ventral levels (L2-L5) of the nucleus. On average, the number of cells in individual rectangles of the quantification grid ranged from 0.3 ± 0.2 to 13.1 ± 6.9 in males and from 0.2 ± 0.1 to 10.2 ± 2.5 in females. In the BST, the highest density of labeled cells was observed in its posterior part, where cells tended to be more numerous in the medial part of the nucleus in both the dorso-ventral and medio-lateral axes. On average the number of cells per rectangleof the quantification grid ranged from 1.0 ± 0.4 to 12.3 ± 1.9 in males and from 0.2 ± 0.1 to 11.9 ± 1.4 in females.

Figure 5:

Heat map representations of the spatial distribution of the average number of cells expressing aromatase mRNA in the 3 rostro-caudal levels (CA-4, CA-2 and CA) of the medial preoptic nucleus (POM) and the two rostro-caudal levels of the bed nucleus of the stria (BST; CA and CA+1) of both sexes as well as heat maps of the difference between the average number of cells counted in males and females (right column). Rostro-caudal levels are identified based on the number of sections separating them in the rostral (−) or caudal (+) direction from the section where the anterior commissura (CA) assumes its largest extension. The dorso-ventral axis of each nucleus was divided into 3 or 6 levels (L1 to L6) depending on the rostro-caudal level considered and the medio-lateral axis was divided in 3 levels (Lateral: Lat, Intermediate: Int and Medial: Med). The bars next to male and female heat maps present the color code for the number of aromatase-positive cells per rectangle ranging from 0 (yellow) to 13 (red), while bars next the subtractive maps range from 8 more cells in females (in red) to 8 more cells in males (in blue). (*), * and **denote p <0.1, 0.05 and 0.01 for the sex difference by Tukey HSD posthoc test following significant Sex x ML x DV interaction in 3-way ANOVA.

Each of the 5 male-female pairs of heat mapswas analyzed by a three-way ANOVAwith the sex of the subjects as independent factor and the position of each rectangle in the medio-lateral (ML) and dorso-ventral (DV) axis as two repeated factors. Each of these analyses generated 3 main effects, 3 primary interactions and one secondary interaction (Sex x ML x DV). This included a number of significant position effects, but these are of little interest and this presentation will focus on the effects of sex and their secondary interaction with positions (Sex x ML x DV).

Overall, males had significantly more aromatase-positive cells than females in the caudal portion of the POM (CA, F_1,17 = 17.615, p < 0.001), but not in the two more rostral levels of the nucleus (CA-2, F_1,13 = 2.762, p = 0.120; CA-4, F_1,13 = 0.002, p = 0.963). No overall effect of sex was detected in BST, evenif a statistical tendency was present in its most caudal part (CA, F_1,16 = 0.0007, p = 0.993; CA+2, F_1,14 = 3.429, p = 0.085). The secondary interaction of sex with positions (Sex x ML x DV) was significant only in the most caudal POM (F_10,170 = 3.447, p < 0.001), but not in any other section of the POM or BST (POM_CA-2 : F_10,130 = 0.643, p = 0.775; POM_CA-4 : F_4,44= 1.511, p = 0.215; BST_CA : F_4,64 = 0.629, p = 0.643; BST_CA+2 : F_10,140 = 1.434, p = 0.171).Posthoc analyses of the secondary interaction in POM_CAindicated that these differences arose mainly from higher numbers of cells expressing aromatase mRNA in the dorso-ventral levels 2 to 4 of the lateral column and the dorso-ventral level 5 in the intermediate column (Fig. 5).

Finally, as expected, many overall effects of position (ML in all regions except POM_CA-2 where 0.05<p<0.10; DV in all 5 regions) or position interactions (ML x DV in POM_CA and in the two BST levels) were additionally detected, but these never interacted significantly with sex, except in the POM_CA where there was a DV by sex interaction (F_5,85 = 4.157, p < 0.002). This DV by sex interaction was however also associated with a secondary interaction that has been analyzed by post hoc tests before and will not be further discussedhere.

Number of aromatase cells also expressing GAD67 mRNA

Large numbers of cells expressing GAD67 mRNA were detected in the POM and BST and some of them also expressed aromatase mRNA (Fig. 2C, F and I). The distribution of double-labeled cells in POM and BST was very similar to the distribution of cells single labeled for aromatase.

The analysis of the total number of double-labeled cells at each rostro-caudal level in POM or BST indicated that overall males have significantly more double labeled cells than females in the POM (F_1,41 = 15.303, p <0.001). In this region, there was also a significant effect of the rostro-caudal level (F_2,41 = 7.362, p <0.002) and a significant interaction between this factor and sex (F_2,41 = 3.873, p = 0.029). The posthoc analyses revealed that this interaction resulted from higher numbers of doublelabeled cells in males than females in the most caudal portion of the nucleus specifically, and from higher numbers of doublelabeled cells in the intermediate and caudal levels compared to the rostral level in males (Fig. 3C). In the BST, the analysis confirmed that there were more doublelabeled cells in the caudal portion of this region (F_1,30 = 46.398, p < 0.0001), but there was only a trend of sex difference (F_1,30 = 3.565, p = 0.069) and no interaction between the two factors (F_1,30 = 2.174, p = 0.151; Fig. 3D).

The distribution of these double labeled cells at each level was analyzed by three-way ANOVAs and is presented in heat maps as described before for the total number cells expressing aromatase (Fig. 6). On average the number of doublelabeled cells ranged from 0.1 ± 0.1 to 7.9 ± 1.2 in the POM of males, from 0.1 ± 0.1 to 5.6 ± 1.0 in the POM of females, from 0.6 ± 0.3 to 8.2 ± 1.1 in the BST of males and from 0.2 ± 0.1 to 8.7 ± 1.5 in the BST of females.

Figure 6:

Heat map representations of the spatial distribution of the average number of cells expressing both aromatase and GAD67 mRNA in the 3 rostro-caudal levels (CA-4, CA-2 and CA) of the medial preoptic nucleus (POM) and the two rostro-caudal levels of the bed nucleus of the stria (BST; CA and CA+2) of both sexes as well as heat maps of the difference between the average number of cells counted in males and females (right column). Rostro-caudal levels are identified based on the number of sections separating them in the rostral (−) or caudal (+) direction from the section where the anterior commissura (CA) assumes its largest extension. The dorso-ventral axis of each nucleus was divided into 3 or 6 levels (L1 to L6) depending on the rostro-caudal level considered and the medio-lateral axis was divided in 3 levels (Lateral: Lat, Intermediate: Int and Medial: Med). The bars next to male and female heat maps present the color code for the number of aromatase-positive cells per rectangle ranging from 0 (yellow) to 13 (red), while bars next the subtractive maps range from 8 more cells in females (in red) to 8 more cells in males (in blue). (*) and * denote p <0.1 and 0.05 for the sex difference by Tukey HSDpost-hoc test following significant Sex x ML x DV interaction in 3-way ANOVA.

The analysis by three-way ANOVA of the heat maps describing the tridimensional distribution of the double labeled cells confirmed that overall males have more doublelabeled cells than females in the caudal portion of the POM (CA; F_1,17 = 15.973, p <0.001). A trend, albeit non-significant, was also found in the POM CA-2 (F_1,13 = 4.625, p = 0.051). There was however no sex difference in the most rostral portion of this nucleus (F_1,13 = 0.228, p = 0.643) and in both subdivisions of the BST (CA: F_1,16 = 0.363, p = 0.5554; CA+1: F_1,14 = 2.889, p = 0.111). Significant secondary interactions between sex and positions(Sex x ML x DV) were also detected in the most caudal portion of the POM (CA: F_10,170 = 2.669, p <0.005), but not in the intermediate and more rostral portions (CA-2: F_10,130 = 1.121, p = 0.289; CA-4: F_4,44 = 1.239, p = 0.309),nor in any portion of the BST (CA: F_4,64 = 1.042, p = 0.393; CA+2: F_10,140 = 1.200, p = 0.310). As observed for the total number of cells expressing aromatase, the Sex x ML x DV interaction in the number of doublelabeled cells in the caudal POM is mainly explained by higher cell numbers in the dorso-ventral levels 2 to 4 of the lateral column and the dorso-ventral level 5 in the intermediate column (Fig. 6).

Finally, in agreement with the distribution of all cells expressing aromatase, significant position effects were also detected here (ML in all regions except POM_CA-4 where 0.05<p<0.10; DV in all 5 regions). We also detected ML x DV position interactions in all 5 regions,but interactions involving the sex were all non significant except for DV by Sex in POM_CA (F_5,85 = 3.896, p < 0.003). This interaction was also associated with a secondary interaction that has been analyzed by post hoc tests before and will thus not be further discussed.

Quite interestingly, although many aromatase-positive cells were present in the VMN, the co-labeling of aromatase and GAD67 wasdrastically different in this nucleus where no single aromatase-expressing cells were found to be co-labeled for GAD67 in both males and females (Fig. 4).

Percentage of aromatase cells co-expressing GAD67

The percentage of aromatase-positive cells co-expressing GAD67 ranged between 23 and 84 % in the entire POM (52 ± 4% on average)and between 10 an 82% in the entire BST (63 ± 4% on average). The highest percentages of co-expression were detected in the dorso-lateral portions of the POM at the intermediate and caudal levels.Given the complete absence of GAD67 in all aromatase-expressing cells of the VMN of both males and females, the percentage of co-expression in this nucleus was zero with novariance. These data could thus not be subjected to any statistical analysis.

The analysis of the total percentage of aromatase cells expressing GAD67 across rostro-caudal levels of the POM suggested an overall sex difference (male>female), but it did not reach statistical significance (F_1,41 = 3.902, p = 0.055; Fig 7). No position effect (F_2,41 = 1.713, p = 0.193) and no interaction were detected (F_2,41 = 0.053, p = 0.949). In the BST, no main effect of sex (F_1,30 = 1.177, p = 0.287), position (F_2,30 = 0.0002, p = 0.988) or interaction between the two factors (F_2,30 = 0.201, p = 0.657) were present.

Figure 7:

Heat map representations of the spatial distribution of the percentage of cells expressing both aromatase and GAD67 mRNA in the 3 rostro-caudal levels (CA-4, CA-2 and CA) of the medial preoptic nucleus (POM) and the two rostro-caudal levels of the bed nucleus of the stria (BST; CA and CA+2) in both sexes as well as heat maps of the difference between the percentage of double labeled cells in males and females (right column). Rostro-caudal levels are identified based on the number of sections separating them in the rostral (−) or caudal (+) direction from the section where the anterior commissura (CA) assumes its largest extension. The dorso-ventral axis of each nucleus was divided into 3 or 6 levels (L1 to L6) depending on the rostro-caudal level considered and the medio-lateral axis was divided in 3 levels (Lateral: Lat, Intermediate: Int and Medial: Med). The bars next to male and female heat maps present the color code for the number of aromatase-positive cells per rectangle ranging from 0 (yellow) to 13 (red), while bars next the subtractive maps range from 48 more percent in females (in red) to 48 more percent in males (in blue).

The heat maps illustrating the differences confirmed that males tended to have a higher proportion of doublelabeled cells than females in both POM and BST (dominance of blue color in the subtractive maps), but differencesin the other direction were also locally present. The statistical analyses of spatial distribution of the percentage of double labeled cells identified a main effect of sex only in the most caudal section of the POM (CA: F_1,17 = 6.360, p = 0.022), but not at the more rostral levels (CA-2: F_1,13 = 1.465, p = 0.248; CA-4: F_1,11 = 1.016, p = 0.335), nor in the BST (CA: F_1,16 = 1.909, p = 0.186; CA+2: F_1,14 = 0.291, p = 0.598). No Sex x ML x DV interaction was detected in any region (POM CA; F_10,170 = 0.646, p = 0.773; CA-2; F_10,130 = 0.292, p = 0.982; CA-4: F_4,44 = 0.767, p = 0.552; BST CA: F_4,64 = 0.605, p = 0.660; BST _CA+2: F_10,140 = 0.927, p = 0.511).

Again, overall significant position effects were detected at all rostro-caudal levels of both nuclei (ML in all 5 areas, DV in in all areas except BST_CA), as well as ML by DV interactions (in POM_CA, POM_CA-2, BST_CA+2), but these position effects never interacted significantly with the effect of sex.

DISCUSSION

The goal of the present study was to determine whether aromatase expressing cells in the brain of male and female Japanese quail were GABAergic. With the use of fluorescent in situhybridization (FISH) procedures we successfully localized cells expressing the mRNA for the enzyme aromatase in the POM and the BST of both sexes. We also localized cells expressing the mRNA for one isoform of the synthesizing enzyme GAD67 that serves as a marker of GABAergic cells. By quantifying cells in subregions of the POM and BST, we investigatedthe tridimensional distribution of labeled cells and determined with neuroanatomical precision where the expression of these two genes differs between males and females. We identified a clear sex difference in the number of cells expressing aromatase in the POM that was localized in the caudal and lateral part of the nucleus. Despite a statistical trend suggesting that these cells might be more abundant in males than in females, no sex difference in the number of cells expressing aromatase was identified in the BST. We also detected a large number of cells expressing the mRNA for GAD67 in the POM and the BST and this mRNA was expressed in a large number of aromatase-positive cells. On average, there was a high percentage of co-localization of these two mRNAs in both the POM and the BST (over 50% in both cases) and this co-localization rate even reached higher values within specific subregions of both nuclei (up to around 80%; see Fig. 7). In the VMN, where we did a less thorough investigation, no single cellwas detected that exhibited a co-localization between GAD67 and aromatase.

Comparison with previous studies

These findings are consistent with previous worklocalizing cells expressing the aromatase protein in quail by immunohistochemistry (Foidart et al., 1994; Balthazart et al., 1996; Balthazart et al., 2000). One study in particularanalyzed the 3 dimensional distribution of aromatase-immunoreactive cells with a higher degree of anatomical resolution (10 rostro-caudal levels quantified with a grid of 10 × 14 squares), but a similar sex difference in number of positive cells was detected with males exhibiting larger numbers than females specifically in the lateral part of the POM, mostly at the CA-2 level(Balthazart et al., 1996). By contrast, in the medial POM, males had fewer aromatase-immunoreactive cells than females and numerical differences in the same direction were observed here (red squares in figure 5, right column), even if these differences did not reach statistical significance. The BST was not analyzed in that previous study.

The expression of aromatase mRNA in the quail brain was also studied by in situ hybridization utilizing a radioactive probe and film autoradiography to localize the areas of hybridization (Voigt et al., 2007). The mRNA was quantified by measuring the optical density of and the area covered by the radioactive signal rather than counting individual cells as done in the present study. That study detected a female-biased sex difference in optical density of the radioactive signal in the BST, but not in the POM. There was also alarger volume covered by radioactive signal in the POM of males as compared to females. Due to the different methods used, it is however impossible to directly relate these results to the present one.

The major new finding of the present study is obviously the high degree of co-localization between aromatase and GAD67 that was observed in the POM and BST, but not in the VMN, of quail. This observation is consistent with recent findingsdemonstrating that in male and female mice the vast majority of aromatase neurons of the postero-dorsal medial amygdala and in the principal component of the BST also express GAD1 the gene coding for GAD67 (Unger et al., 2015; Bayless et al., 2019). Studies of aromatase cells by FISH in the auditory telencephalon of zebra finches also revealed that these cells have a heterogeneous neurochemical phenotype, but that around 35% of them co-express GAD65, the other isoform of glutamic acid decarboxylase(Jeong et al., 2011a). Further work on the zebra finch auditory telencephalon showed that aromatase is co-expressed with parvalbumin, but not with calbindin, two other markers of inhibitory neurons, in the caudomedial nidopallium, but not in other regions, such as the caudolateral nidopallium or the hippocampus(Ikeda et al., 2017). There is thus in this species, as in the quail studied here, a neuroanatomical specificity of the aromatase-expressing neurons that can be inhibitory or not. In the quail POM and BST, the colocalization with GAD67 is largely the rule, but the situation is different in the VMN, where no co-localization at all could be detected in both males and females. More studies in other species and brain sites would however be needed to determine to what extent aromatase-expressing cells are generally inhibitory neurons.

Interestingly, a qualitative look at the sections labeled for aromatase revealed that these cells were often present in dense somato-somatic clusters of 2–3 up to 5–6 cells (see Fig. 2D and G). A similar situation was previously identified and quantitatively analyzed in great detail in the zebra finch telencephalon (Ikeda et al., 2017). This anatomical organization raises the possibility that these cells might be coupled by gap junctions and thus synchronize their activity and their production/release of estrogens. Gap junctions between clusters of cells have been observed in the song control nucleus HVC of canaries (Gahr & Garcia-Segura, 1996) and were serendipitously detected during electrophysiological studies by observing transfer of biocytin between adjacent aromatase-positive cells in the quail POM (Cornil et al., 2004). The physiological implications of this neuroanatomical organization would clearly deserve additional investigation.

Functional significance of aromatase neurons in the lateral part of the caudal POM

As discussed in the previous section, the topographic distribution of cells expressing aromatase mRNA in the POM is similar to what has been reported previously based on detailed immunohistochemical studies of the aromatase protein (Balthazart et al., 1996). Interestingly, the major sex difference in the number of aromatase-positive cells was mainly observed in both cases in the dorso-lateral portion of the caudal POM. This subregion of the POM was previously identified as being a key neuroanatomical site for the sexually differentiated activation by testosterone of male-typical copulatory behavior in quail (Panzica et al., 1996; Voigt et al., 2007). It has indeed been clearly established that testosterone activates male-typical copulatory behavior in male, but not in female, quail (Adkins, 1975; Balthazart et al., 1983)and this difference correlates with a sex difference in the volume of the POM (males>females) even if this difference is no longer observed in gonadectomized birds (Panzica et al., 1996). However, cells in the dorso-lateral part of the caudal POM are significantly larger in male than in female quail (Panzica et al., 1991)but, if this sex difference in cell size partly reflects activational effects of testosterone (Panzica et al., 1991),it is also organized byembryonic sex steroids. Indeed, the size of these cells is decreased in adult male birds that were treated with estradiol in ovo and this effect is only observed in birds in which copulatory behavior was also inhibited (Aste et al., 1991)

This sex difference in cell size thus positively correlates with the sex difference in the activational effects of testosterone on behavior and a substantial proportion of the cells that are concerned mustexpress aromatase(Aste et al., 1994), although this idea has never been tested.In addition, selective lesions of the caudal parts of the POM, where these steroid-sensitive aromatase-expressing cells are located(Aste et al., 1994),specifically inhibit male-typical consummatory sexual behaviors,while leaving relatively intact the appetitive sexual behaviors(Balthazart et al., 1992; Balthazart et al., 1998).Conversely, the induction of aromatase expression in this subregion by local stereotaxic implants of testosterone positively correlates with the intensity of the behavioral activation(Balthazart et al., 1992). Finally, the neuronal activation as measured by an induction of c-fos expression following sexual interactions with a female is also maximal in this caudal subregion of the POM (Meddle et al., 1997; Taziaux et al., 2006). All these data thus concur to supportthe notion that aromatase-expressing cells in the dorso-lateral part of the caudal POM are a key site of testosterone action in the control of male copulatory behavior (for an overview see (Balthazart & Ball, 2007)). The present results confirming a local sex difference in aromatase expression based on the analysis of the corresponding mRNA further reinforce this conclusion.

Co-localization of Aromatase and GAD67: potential impact on male sexual behavior

It is usually assumed that inhibitory neurons are for the most part interneurons. These interneurons can be schematically divided into local interneurons (that most people have in mind), but also relay interneurons that have long distance projection axons. We have previously shown, by retrograde tracing coupled to immunohistochemistry, that a large number of aromatase-immunoreactive cells of the POM send projections to the periaqueductal gray, PAG (Absil et al., 2001; Carere et al., 2007). These projections predominantly originate from aromatase cells located in the dorso-lateral part of the nucleus,particularly in the most caudalextent of the nucleus. They are sexually differentiated, i.e. present inlarger number in males than in femalesespecially in the caudo-lateral part of the nucleus(Carere et al., 2007). This sex difference in connectivity represents another functional link to the control of copulatory behavior that is not displayed by females.

In rodents,it is well established that the PAG is a pre-motor relay nucleus as far as copulatory behavior is concerned. The PAG has extensive bidirectional connections with the preoptic area(Simerly & Swanson, 1986; Rizvi et al., 1992; Rizvi et al., 1996; Murphy & Hoffman, 2001)and it contains dense populations of cells expressing sex steroids receptors including estrogen receptors of the alpha subtype (ERα) and androgen receptors (Murphy et al., 1999b; Murphy & Hoffman, 2001). The PAG seems to play a chronic inhibitory role on the expression of male sexual behavior(Brackett et al., 1986; Murphy & Hoffman, 2001)via its projections to more caudal structures including prominently the nucleus paragigantocellularis (Normandin & Murphy, 2008). It is indeed well established that electrolytic or cell body lesions of the PAG markedly facilitate male sexual behavior as measured by an increased mounting and ejaculation frequency and decreased ejaculation latency (Brackett et al., 1986).

The projection from the preoptic area to the PAG thus potentially modulates this behavioral role of PAGand could play an important role in the control of male sexual behavior(Murphy et al., 1999a; Murphy & Hoffman, 2001; Hull et al., 2002). Knowing that, in quail at least, this projection largely originates from aromatase-expressing cells(Absil et al., 2001; Carere et al., 2007)and that a large proportion of these cells are GABAergic in nature (present work) opens the question of the specific function of this input to PAG.

The most parsimonious explanation of the endocrine controls of male sexual behavior in this context would imply that estrogens derived from local aromatization upregulate the activity of the GABAergic projections to the PAG and that the resulting increased GABA release then suppresses the tonic inhibition exerted by the PAG on male sexual behavior. This regulation could however take place at different levels in the neural circuits and two main groups of mechanisms can be postulated.

On the one hand, the axons of the aromatase cells terminating in PAG could release in this site substantial amounts of estrogens since the enzyme is known to be present(Naftolin et al., 1996; Peterson et al., 2005; Rohmann et al., 2007)and enzymatically active (Schlinger & Callard, 1989; Cornil et al., 2012) in pre-synaptic boutons. These estrogens could then bind to the nuclear ERαknown to be present in the PAG or to membrane ER at that level and in this way modulate the PAG inhibitory output towards the lower brain stem and spinal cord.

On the other hand, it is conceivable that the estrogens produced by preoptic aromatase neuronsmodulate the excitability and the synaptic properties of these very same neuronsin an intracrine orparacrine manner. Theclustering of these aromatase neuronsis obviously an anatomical organization that would favor such paracrine interactions. We do not know at this point in time what is/are the specific mechanism(s) through which estrogens derived from local aromatization of testosterone in the brain affect the activity of local neurons. The simplest option is that aromatase produces estrogens in the cytoplasm of neuronsthat also contain nuclear ER (ERα or ERβ) and that these estrogens modulate the transcriptional activity of ER in the nucleus of the cells where they were produced. This modulation of ER transcriptional activity could then secondarily result in a change of the GABAergic tone impacting the PAG. However, even if aromatase and ERα are often co-localized in other brain nuclei (nearly 70% of the neuronsin VMN), thisis rather the exception in the POM (e.g. 8% colocalization) and BST (4%colocalization; (Balthazart et al., 1991)).

More recent research has also demonstrated that estrogens can act more rapidly at thelevel of the cell membrane byactivatingeither nuclear receptors that are translocated at the membrane from where they can signal via transactivation of other membrane receptors (Acconcia et al., 2005; Bondar et al., 2009; Mermelstein, 2009)or membrane specific G-protein coupled receptors (e.g. GPER1 and Gq-mER)(Revankar et al., 2005; Qiu et al., 2008). In this context, estrogens produced in the brain could act in a paracrine manner both pre- and post-synaptically i.e., by interacting withreceptors located atthe membrane of the cell that produced them or atthe membrane of an adjacent cell (see (Cornil, 2009) for a more detailed discussion).These possibilitiesare supported by recent work showing that both pre- and post-synaptic actions of E₂ rapidly modulate hippocampal neurotransmission (Smejkalova & Woolley, 2010; Huang & Woolley, 2012; Oberlander & Woolley, 2016).

Other more complex scenarios involving more that one POM neuron are obviously also possible. Such scenarios wouldimplicate preoptic excitatory glutamatergic inputs,which have been shown to be modulated by estrogens in a number of experimental systems, the hippocampus in particular (e.g., (Wong & Moss, 1992; Kramar et al., 2009; Smejkalova & Woolley, 2010; Huang & Woolley, 2012; Oberlander & Woolley, 2016)). Regardless of whether these glutamatergic inputs are from local or long distance origin, they could modulate the activity of the aromatase/GABAergic projections and as a result the PAG output. In this context,it isinteresting to note that in vivomicrodialysis studies have demonstrated that copulation in male quail is associated with a major release of glutamate in the preoptic area (de Bournonville et al., 2017a) as previously observed in male rats (Dominguez et al., 2006). More work combining electrophysiological and chemical neuroanatomical approaches would clearly be needed to fully uncover the mechanisms underlying the impact of sex steroids action in the POM on the expression of male copulatory behavior.

ABBREVIATIONS

ARO

aromatase

BST

bed nucleus of the stria terminalis

CA

commissure anterior

DIG

digoxygenin

DV

dorso-ventral

E2

estradiol

ERα

estrogen receptor alpha

ERβ

estrogen receptor beta

FISH

fluorescent in situ hybridization

FITC

fluorescein

GAD65

glutamic acid decarboxylase 65

GAD67

glutamic acid decarboxylase 67

HRP

horseradish peroxidase

IHC

immunohistochemistry

ML

medio-lateral

PAG

periacqueductal gray

PB

phosphate buffer

PBS

phosphate buffered saline

POM

medial preoptic nucleus

VMN

ventromedial nucleus of the hypothalamus