The neurons expressing calcium-binding proteins in the amygdala of the guinea pig: precisely designed interface for sex hormones

The generation of emotional responses by the amygdala is determined largely by the balance of excitatory and inhibitory inputs to its principal neurons. These responses are often sex-specific, and any imbalance in excitatory and/or inhibitory tones leads to serious psychiatric disorders which occur with different rates in men versus women. To investigate the neural basis of sex-specific processing in the amygdala, relationships between the neurons expressing calbindin (CB), parvalbumin (PV) and calretinin (CR), which form in the amygdala main subsets of γ-aminobutyric acid (GABA)-ergic inhibitory system, and neurons endowed with oestrogen alpha (ERα), oestrogen beta (ERβ) or androgen (AR) receptors were analysed using double immunohistochemistry in male and female guinea pig subjects. The results show that in various nuclei of the amygdala in both sexes small subsets of CB neurons and substantial proportions of PV neurons co-express ERβ, while many of the CR neurons co-express ERα. Both these oestrogen-sensitive populations are strictly separated as CB and PV neurons almost never co-express ERα, while CR cells are usually devoid of ERβ. In addition, in the medial nucleus and some other neighbouring regions, there are non-overlapping subpopulations of CB and CR neurons which co-express AR. In conclusion, the localization of ERα, ERβ or AR within subsets of GABAergic interneurons across diverse amygdaloid regions suggests that steroid hormones may exert a significant influence over local neuronal activity by directly modulating inhibitory tone. The control of inhibitory tone may be one of the mechanisms whereby oestrogen and androgen could modulate amygdala processing in a sex-specific manner. Another mechanism may be thorough steroid-sensitive projection neurons, which are most probably located in the medial and central nuclei.


Introduction
The amygdala is composed of a set of nuclei and cortical regions within the temporal lobe which are thought to be critically important for emotional behaviour and emotional learning (LeDoux 2003;Sah et al. 2003;Pape and Pare 2010), mediation of pheromonal and reproductive functions (Baum 2009;Kevetter and Winans 1981;Martel and Baum 2009;Morris et al. 2008;Segovia et al. 2006), drug addiction mechanisms (Buffalari and See 2010) and stress response (LeDoux 1994). Although the general mechanisms of all these functions are similar in both sexes, there are also sexspecific details which make these functions sexually dimorphic (Hamann 2005;Fagergren and Hurd 1999;Gruene et al. 2015;Rollins and King 2000;Swaab et al. 2003). For example, women on average retain stronger and more vivid memories of emotional events than men (Seidlitz and Diener 1998;Canli et al. 2002). In addition, the fact that emotional memories tend to be stronger in women may be linked to the greater prevalence of depression and some types of anxiety disorders in women (Bekker and van Mens-Verhulst 2007;Davidson et al. 2002). Indeed, there are many other psychological disorders involving emotional function which occur with substantially different rates in men versus women. For example, anorexia is another disorder associated with amygdala dysfunction, which is observed more often in women than in men (Swaab et al. 2003). In contrast, schizophrenia, autism and drug addiction, also associated with abnormalities in the functioning of the amygdala, are quoted more frequently in men than in women (Swaab et al. 2003). It is worth mentioning that various aspects of amygdala anatomy are sexually dimorphic (Morris et al. 2008;Równiak 2013;Równiak et al. 2015;Segovia et al. 2006).
Circulating levels of sex hormones in the bloodstream constitute additional factors which may influence amygdala physiology and may drive amygdala responses in a sexspecific manner by acting through receptors expressed in specific amygdala nuclei. Indeed, the amygdala has a high density of oestrogen and androgen receptors, as well as high aromatase expression and activity (Shughrue et al. 1997;Lephart 1996). Moreover, these receptors and aromatase were found in both the glutamatergic principal neurons (Kiss et al. 2013) and c-aminobutyric acid (GABA)ergic interneurons (Blurton-Jones and Tuszynski 2002), suggesting that excitation and inhibition in the amygdala may be substantially modulated by sex hormones. There is a growing body of evidence that oestrogens and androgens enhance glutamatergic neurotransmission and down-regulate GABAergic neurotransmission (Barth et al. 2015;Frye et al. 2008), although opposite effects have also been reported (Tan et al. 2012;Zhou et al. 2005). For example, Murphy et al. (1998) reported that oestradiol down-regulated brain-derived neurotrophic factor (BDNF) in cultured hippocampal cells and this effect was associated with an increase in excitatory tone leading to a twofold increase in dendritic spine density and a decrease in inhibition, as evidenced by decrease in glutamate decarboxylase (GAD) and GABA level. Moreover, as both glutamatergic and GABAergic cell types express GABA receptors, an oestradiol-induced decrease in GABAergic transmission to glutamatergic cells could effectively increase excitatory drive and thus favour the formation of new dendritic spines (Murphy et al. 1998;McDonald and Mascagni 2004). Further studies have confirmed that oestradiol enhances the magnitude of longterm potentiation in the hippocampus through facilitation of the N-methyl-D-aspartate receptors (NMDA) (Foy 2001;Córdoba Montoya and Carrer 1997;Smith et al. 2009). On the other hand, oestrogen has an ability to alter the frequency of GABAergic miniature inhibitory postsynaptic currents (Zhou et al. 2005). Moreover, other experiments have shown that ovariectomy and subsequent oestrogen replacement can alter mRNA levels of GABA receptor subunits, GABA transporter, and GAD, the rate-limiting enzyme in GABA synthesis (Nakamura et al. 2004;Herbison and Fénelon 1995).
The generation of responses by the amygdala is determined largely by the balance of excitatory and inhibitory inputs to its principal projecting neurons. The activity of these neurons is tightly controlled by GABAergic interneurons, especially these expressing calbindin (CB), parvalbumin (PV) and calretinin (CR) (Sorvari et al. 1996;Muller et al. 2007;Woodruff and Sah 2007a). It is worth mentioning that each of these subpopulations, apart from GABA, also uses different neuropeptides. For example, CB-containing neurons often co-express somatostatin (SOM) and/or neuropeptide Y, whereas CR-positive cells often co-express cholecystokinin and/or vasoactive intestinal peptide (McDonald and Mascagni 2002;Mascagni and McDonald 2003). In contrast, PV neurons are devoid of any of these neuropeptides (Mascagni et al. 2009). Moreover, it should be kept in mind that some of the CB and CR neurons are not GABAergic interneurons but glutamatergic projecting neurons (Moryś et al. 1999;Kemppainen and Pitkänen 2000;McDonald et al. 2012). If the activity of various subpopulations of GABAergic neurons was strongly influenced by circulating steroid hormones via their receptors, inhibitory control in the amygdala could easily be driven in a sex-specific manner. Such modulation may have a huge impact on the amygdala's overall processing, and this way promotes sexually dimorphic responses. Although the amygdala has a large population of GABAergic neurons co-expressing CB, PV or CR (Pitkänen and Amaral 1994) and a large population of cells bearing oestrogen alfa (ERa), oestrogen beta (ERb) and androgen (AR) receptors (Shughrue et al. 1997;Warembourg and Leroy 2004;Wood and Newman 1999), little is known about relationships among all these populations. There are only two studies on the relationship between neurons expressing calcium-binding proteins and steroid receptors in the amygdala, and only one of them is quantitative (Blurton-Jones and Tuszynski 2002;Perez et al. 2004). However, in both of these studies the amygdala was just one of many brain centres studied. The quantitative study of Blurton-Jones and Tuszynski (2002) describes the co-expression pattern of PV interneurons with ERb-bearing cells in the selected nuclei of rat amygdala, where the counts were limited to the lateral and basomedial nuclei of female subjects. Perez et al. (2004) reported a lack of association between CB or PV-expressing neurons and ERa in monkey forebrain, which was also studied in female subjects only. Thus, the present study investigates for the first time the co-localization pattern of CB, PV or CR with ERa, ERb or AR in the amygdala of the guinea pig. Female and male subjects were taken into account to examine possible sex differences as well. To make the text easier to read, the neurons expressing CB, PV, CR, SOM, GABA, ERa, ERb and AR will be uniformly described as CB?, PV?, CR?, SOM?, GABA?, ERa?, ERb? and AR? neurons.

Subjects
Ten adult Dunkin-Hartley guinea pigs (Cavia porcellus, L.), five males and five females, at the age of 6 months and weighting 0.7-0.8 kg were used. All these animals were purchased from the professional supplier (Polish Mother's Health Centre in Łódź, Poland) to ensure their proper parameters such as the same strain, the same age, similar weight and proestrus phase in female subjects. The phase of the estrous cycle was determined directly from vaginal smears. Animal care and handling were in accordance with the European Union Directive for animal experiments (2010/63/EU) and were approved by the Local Ethical Commission of the University of Warmia and Mazury in Olsztyn (No. 27/2009). All efforts were made to minimize animal suffering and to use the minimum number of animals necessary to produce reliable scientific data.

Tissue preparation
All animals, i.e. males and females, were deeply anesthetized with an intraperitoneal injection of Morbital (Biowet, Poland; 2 ml/kg body weight, 133.3 mg/ml of pentobarbital sodium salt and 26.7 mg/ml of pentobarbital), and after cessation of breathing, immediately perfused transcardially with saline (0.9%) followed by 4% paraformaldehyde (pH 7.4) in phosphate-buffered saline (PBS). Following perfusion, brains were dissected out from the skulls, postfixed overnight in the same fixative, washed twice in 0.1 M phosphate buffer (pH = 7.4, 4°C) and then stored for 3-5 days in graded solutions (19% and 30%) of sucrose (Sigma Aldrich) at 4°C until they sank. Finally, the brains were frozen and then coronally sectioned at a thickness of 10 lm using a cryostat. The sections were mounted on object slides and stored at -80°C until further processing.

Immunofluorescence experiments
Amygdala sections were processed for routine doubleimmunofluorescence labelling using primary antisera raised in different species and species-specific secondary antibodies (Table 1). As each slide contained two sections, five neighbouring slides (ten neighbouring sections) were always used for staining with the use of seven different primary antibodies. Nine of these ten sections were used for the main experiment (co-localization of 3 calcium-binding proteins and 3 steroid receptors) and one spare section was used for pan-neuronal marker to visualize amygdaloid nuclei. In addition, some sections in three male and three female subjects were destined for additional staining described in the present study.
All samples were washed three times in PBS and then incubated for 1 h in humid chambers with blocking buffer (0.1 M PBS, 10% normal donkey serum, 0.01% bovine serum albumin, 1% Tween, 0.05% thimerosal, 0.01% NaN 3 ). The sections were then rinsed in PBS and incubated overnight at room temperature with a mixture of primary antibodies, namely a combination of the appropriate primary antibody to one of the calcium-binding proteins (CB or PV or CR) and one to the steroid receptor (ERa or ERb or AR) ( Table 1). Additional incubations included neuronspecific nuclear protein NeuN (pan-neuronal marker) or mixtures consisting of CB and PV, CB and CR or SOM with one of the steroid receptors ( Table 1). The antibodies were diluted in PBS containing Triton X-100 (0.3-0.5%) and 1% normal donkey serum. After incubation with the primary antibodies, the sections were rinsed in PBS and incubated for 1 h with a mixture of species-specific secondary antibodies (Table 1). Finally, all samples were rinsed in PBS and then mounted with carbonate-buffered glycerol (pH 8.6) and coverslipped.

Controls
The specificity of the primary antisera used in this study has been shown by various researchers using these products in multiple previous studies (Airaksinen et al. 1997;Drexel et al. 2011;Meszar et al. 2012;Zimmermann and Schwaller 2002;Stanić et al. 2014;Kritzer 2004;Wood and Newman 1999). In addition, product descriptions of rabbit anti-calbindin (CB-38), mouse anti-calbindin (300), and mouse anti-calretinin (6B 3 ) antisera include immunoblots of the guinea pig brain homogenates, which were specifically stained by these antibodies, showing bands at 28, 28 and 29 kDa, respectively. The same documents show also absence of specific immunohistochemical staining in brains sections of CB or CR knock-out mice using these antibodies. The rabbit antisera towards ERa (06-535), ERb (05-824) and AR (06-680) were tested by immunoblots and these antibodies specifically stained the mouse brain extracts showing single bands at 66, 59 and 110 kDa, respectively (Stanić et al. 2014). The specificity of secondary antibodies was controlled by the omission and replacement of all primary antisera by non-immune sera or PBS. Lack of any immunoreactions indicated specificity.

Counts and statistics
Colocalizations of CB?, PV? and CR? cells with particular steroid receptors were analysed in the lateral, basolateral, basomedial, central, medial and cortical amygdaloid nuclei using an Olympus BX51 microscope equipped with Cell-F image analysis software (Olympus GmbH, Germany). Only the numbers of single-and double-labelled CB?, PV? and CR? cells were counted. Single-labelled ERa?, ERb? or AR? cells were excluded from the investigation. For each nucleus in each animal of both sexes, particular combinations of antigens were counted on 10 evenly spaced sections arranged from the rostral to the caudal extent. The space between sections analysed per specific nucleus depended on the nucleus length, and these values were as follows: LA: 250 lm, BL: 200 lm, BM: 200 lm, CE: 200 lm, ME: 200 lm and CO: 250 lm. In order to confirm the localization of the individual nuclei on the sections, neighbouring sections stained with mouse anti-NeuN (pan-neuronal marker) were used. All counts on the single section were made at 209 magnification using 347.6 lm 9 260.7 lm regions. The test frames were arranged in a way that ensured coverage of the whole cross-sectional area of each nucleus studied. The numbers of test frames studied per nucleus were as follows: LA: 5-7, BL: 2-4, BM: 2-3, CE: 2-4, ME: 3-4 and CO: 2-3. Within test frames single-labelled and double-labelled neurons were counted separately. Such separate counts made within the test frames in the single nucleus in the subject were totalled. Finally, counts from each nucleus were averaged in each sex group and expressed as mean ± standard deviation (SD). All counts were made on coded slides by the author. To avoid fluorescence fading, test frame was digitally recorded before counting. Such digital frames were in the form of stacks which were composed of two microphotographs representing red and green immunofluorescence channels. Saved stacks were also evaluated by two independent experimenters, being blind to the parameters of the studied tissues (sex, nucleus, antigens, etc.). The results of these counts showed high inter-rater reliability (Pearson R = 0.86, p \ 0.001). Statistical differences between the two experimental groups (males vs. females) were analysed with a two-tailed t test (*p B 0.05, **p B 0.01 and ***p B 0.001) using Graph-Pad Prism 5 software (GraphPad Software, La Jolla, CA, USA).
As the population of CB? neurons in the amygdala consists of two separate and non-overlapping subpopulations such as CB?/PV? and CB?/SOM? cells, additional staining was performed to clarify the relationships between CB? neurons and either PV? or CR? cells, as well as between SOM? neurons and either ERb? or AR? cells. All these sections were analysed using an epifluorescence microscope. However, the SOM signal was present only in cytoplasm, while the AR signal was present only in cell nuclei, and thus in many double-stained cells the colocalization was not obvious. Therefore, to unequivocally ascertain that SOM and AR co-exist in the same neurons, additional analysis was performed with the use of a confocal microscope. All additional sections were also evaluated quantitatively in a way described above (5 sections per nucleus in 3 male and 3 female subjects). In the case of CB staining, only single and double-labelled CB? neurons were counted, whereas single-labelled PV? or CR? cells were excluded from analysis. In the case of SOM preparations, only single-and double-labelled SOM? neurons were counted, while single-labelled ERb? or AR? cells were excluded from investigation.

Photomicrographic production
Low-and high-magnification photomicrographs of immunoreactive cells were taken with a CC-12 digital camera (Soft Imaging System GMBH, Germany) on an Olympus BX51 microscope. These digital images were slightly modified to optimise the image resolution, brightness and contrast using CS4, version 11.0, software (Adobe Systems Inc., San Jose, CA, USA). Low-and high-magnification confocal fluorescent images ( Fig. 4d-f) were obtained with a Leica TCS SP5 microscope (Wetzlar, Germany). Sequential excitation at 488 nm and 561 nm was provided by argon and DPSS 561 lasers, respectively. Schematic drawings demonstrating the distribution patterns of ERa?, ERb? and AR? cells were created by marking with a dot each neuron on the digitalized section and superimposing outlines from the adjacent anti-NeuN sections.

Results
Distribution of CB1, PV1, CR1 and SOM1 neurons in the amygdala of male and female guinea pig The position of the various nuclei in the guinea pig amygdala as well as the distributions of CB?, PV? and CR? neurons in this region were precisely described in both sexes in our previous reports concerned with sex differences existing in this species (Równiak 2013;Równiak et al. 2015). The distribution pattern of SOM? neurons in the guinea pig amygdala was very similar to that described in the rat (McDonald 1989) and pig (Równiak et al. 2008). In the present study, the same six main amygdaloid regions, i.e. the lateral, basolateral, basomedial, central, medial and cortical nuclei were chosen for detailed investigation. In all these regions, the relationships between CB?, PV? or CR? neurons and ERa?, ERb? or AR? cells were analysed and compared between the sexes.

Distribution of ERa1, ERb1 and AR1 neurons in the amygdala of male and female guinea pig
An analysis of the distributions of ERa?, ERb? and AR? immunoreactivities in the amygdala of male and female guinea pig revealed large numbers of weakly to strongly stained cells throughout the entire rostral to caudal continuum of the amygdala (Figs. 1, 2). In both sexes, ERa? cells were the most numerous, while ERb? neurons were the least numerous. The subcellular distribution of these immunoreactivities was predominantly nuclear in most neurons (Fig. 2). However, in some cells, the cytoplasm and neuronal processes were also immunostained. Although, populations of ERa?, ERb? and AR? neurons had slightly different patterns of distribution in the amygdala, these three patterns also had many common features (Figs. 1, 2). For example, in both sexes, the ERa?, ERb? and AR? cells were the most numerous and the most densely packed in the medial nucleus (Figs. 1, 2). Moreover, in the posterodorsal portion of the medial nucleus, the ERa?, ERb? and AR? cells were among the most intensely labelled, whereas in the other nuclei of the amygdala the intensity was weaker. Another characteristic and common feature was clusters of intensely stained ERa?, ERb? and AR? cells in the intra-amygdaloid division of the bed nucleus of the stria terminalis and scattered neurons in the central nucleus. In the basolateral complex of the amygdala, only the basomedial nucleus contained many ERa?, ERb? and AR? cells. On the other hand, in the lateral and basolateral nuclei fewer ERa?, ERb? and AR? signals were found (Figs. 1, 2). However, in the lateral nucleus, the density of ERa? nuclei was quite moderate. In the amygdalohippocampal area, ERa? and AR? cells were abundant, but relatively few ERb? cells were found (Figs. 1, 2). In contrast, ERa? cells were much more numerous than ERb? or AR? neurons in cortical nuclei (Figs. 1, 2).
Co-existence pattern of CB1, PV1 and CR1 neurons with ERa, ERb or AR in the amygdala of male and female guinea pig All details concerning the co-existence of CB?, PV? and CR? neurons with ERa?, ERb? and AR? in the amygdala of male and female guinea pig are shown in Tables 2, 3, 4 and 5 and Figs. 3, 4, 5 and 6. These details clearly demonstrate that: CB?, PV? and CR? neurons have different and unique patterns of co-expression with ERb, ERa and AR; there is cellular separation of ERb, ERa and AR within non-overlapping subpopulations of CB? PV? and AR? cells; the patterns and extents of co-expression are very similar in both sexes.

Calbindin neurons
Small proportions of CB? neurons in the amygdala were costained for ERb or AR, but these cells almost never coexpressed ERa (   As population of CB? neurons in the amygdala consists of two separate and non-overlapping subpopulations such as CB?/PV? and CB?/SOM? cells, additional staining was performed to verify which CB? neurons may co-express ERb or AR (Fig. 4). Double-immunofluorescence with antibodies against CB and PV revealed the substantial overlap of both markers, suggesting that at least a part of CB?/ERb? neurons may contain PV as well (Fig. 4a). For example, in the lateral and basolateral nuclei, approximately 30 and 40% of CB? neurons co-expressed PV in male and female subjects, respectively (Table 3). Doubleimmunofluorescence with antibodies against SOM and ERb demonstrated the lack of any co-existence of both these markers in the same neurons (Table 3; Fig. 4c). In contrast, SOM and AR co-existed in the same cells in the medial nucleus and neighbouring regions (Fig. 4d-f). In the medial nucleus approximately 40% of SOM? neurons were simultaneously AR? (Table 3). In the basomedial, cortical and central nuclei, only single cells were doublelabelled for SOM and AR (Table 3).

Parvalbumin neurons
A substantial proportion of PV? neurons in the amygdala co-expressed ERb, but these cells almost never co-expressed ERa or AR (Table 4; Fig. 5). Within the lateral and basolateral nuclei, approximately 15 and 7% of PV? neurons co-expressed ERb, respectively, whereas within the basomedial nucleus one-third of these cells were simultaneously ERb?. Within the medial and central nuclei, there were no neurons double-labelled for PV and ERb due to the limited number of PV? cells in both these regions. Within the cortical nuclei, double-labelled cells Note in both sexes extensive co-localization of PV? neurons with ERb and the lack of associations of these cells with ERa or AR. Data were expressed as mean ± standard deviation (SD). As PV? neurons almost never co-express ERa or AR, only the counts concerning relationships of PV? neurons with ERb were statistically evaluated by the two-tailed t test (* p B 0.05, ** p B 0.01 and *** p B 0.001) Note that in both sexes CR? neurons show extensive co-localization with ERa and modest co-expression with AR, however, these cells almost never co-express ERb. Data were expressed as a mean ± standard deviation (SD). As CR? neurons almost never co-express ERb, only the counts concerning relationships of CR? neuron with ERa or AR were statistically evaluated by the two-tailed t test (* p B 0.05, ** p B 0.01 and *** p B 0.001) were rarely observed in more anterior portions of the nucleus. However, approximately one-third of PV? cells co-expressed ERb within the periamygdaloid cortex and posterior cortical nucleus.

Calretinin neurons
In contrast to CB? and PV? neurons, CR? cells often coexpressed ERa, but these cells almost never co-expressed ERb (Table 5; Fig. 6). Moreover, across diverse regions of the amygdala, there were also dispersed CR? neurons immunoreactive to AR (Table 5; Fig. 6). Within the lateral nucleus, approximately 20% of CR? neurons co-expressed ERa. Within the basolateral, nucleus the density of ERa? cells was low, so the percentage of CR? neurons co-expressing ERa was one-third of that in the lateral nucleus. Within the basomedial and medial nuclei more than onethird of CR? neurons co-expressed ERa, and one quarter of such CR?/ERa? cells were observed within cortical nuclei. Double-labelled neurons for CR and AR were observed within the medial, basomedial and cortical nuclei (Table 5; Fig. 6). In the medial and basomedial nuclei approximately 16 and 7% of CR? neurons co-expressed AR, respectively. In the cortical nuclei less than 5% of CR? cells were AR?. Statistical evaluation of numbers of cells expressing various calcium-binding proteins, cells double-labelled for both calcium-binding proteins and steroid receptors, as well as percentages of double-labelled cells indicated the sex diversified sensitivity of various amygdala regions to sex hormones. For example, the lateral and basomedial nuclei are mostly controlled by oestrogens via PV? neurons using ERb and CR? cells using ERa. Such impact is stronger in females (Tables 4, 5). On the other hand, the medial nucleus is mostly controlled by androgens via CB? and CR? neurons using AR and by oestrogens via CR? neurons using ERa. In both these cases, such influence is stronger in males (Tables 2, 5).

Discussion
To the best of our knowledge, the present study describes for the first time a systematic analysis of relationships between populations of CB?, PV? and CR? neurons, which form in the amygdala main subsets of GABA? system and neurons bearing steroid receptors such as ERa or ERb or AR. The results show that in the amygdala of male and female guinea pig, subpopulations of the CB? and PV? neurons co-express ERb, while many of the CR? neurons co-express ERa. In addition, in the medial nucleus and some other neighbouring regions there are also subpopulations of CB? and CR? neurons which co-express AR. The localization of ERa, ERb or AR within subsets of GABAergic interneurons across diverse amygdaloid regions suggests that steroid hormones may exert a significant influence over local neuronal activity by directly modulating inhibitory tone. The control of inhibitory tone may be a mechanism whereby oestrogen and androgen modulate amygdalar processing in a sex-specific manner. Moreover, as the numbers of CB?, PV? and CR? neurons (Równiak et al. 2015) and the numbers of such cells co-expressing steroid receptors (present results) are sexually dimorphic in various nuclei of the guinea pig amygdala, sensitivity to sex hormones and their impact on many regions may substantially differ in male and female subjects. Thus, GABA? neurons co-expressing CB?, PV? or CR? provide a precisely designed inhibitory interface which is steroid-sensitive and sexually dimorphic. This interface could be the natural basis for sex-specific processing in the amygdala. The present results demonstrate that in the amygdala of the guinea pig, subpopulations of CB?, PV? and CR? neurons are oestrogen or androgen-sensitive as many of these cells co-express ERa, ERb or AR. The patterns of coexpression are similar in both sexes. However, these patterns are quite different in the ''cortex-like'' nuclei (the cortical and basolateral amygdala) and ''striatum-pallidumlike'' nuclei (the central and medial amygdala). In the ''cortex-like'' regions, CB?, PV? and CR? neurons use mostly oestrogen receptors, and the pattern of co-expression is similar to relationships observed in the cerebral cortex (Blurton-Jones and Tuszynski 2002;Kritzer 2002). In the ''striatum-pallidum-like'' nuclei PV? neurons are not present, while CB? and CR? cells are much more influenced by androgens. However, there are no data for comparisons in the available literature. As the amygdala is composed of ''cortex-like'' and ''striatum-pallidum-like'' nuclei (Swanson and Petrovich 1998), one important fact should be kept in mind. In the ''cortex-like'' regions, most projection neurons (pyramidal and semi-pyramidal cells) use glutamate as a fast, excitatory neurotransmitter (McDonald 1996), whereas, in contrast, many projection cells of the ''striatum-pallidum-like'' nuclei use GABA as a fast, inhibitory neurotransmitter (Swanson and Petrovich 1998;Saha et al. 2000;Seo et al. 2016). In the ''cortexlike'' regions, GAD is expressed in interneurons (McDonald 1985;Pitkänen and Amaral 1994), whereas in the ''striatum-pallidum-like'' nuclei, it is often expressed in projection neurons (Swanson and Petrovich 1998;Saha et al. 2000;Seo et al. 2016). The finding that substantial subpopulations of PV? neurons and small subsets of CB? neurons in the amygdala of the guinea pig co-express ERb, while these cells almost never co-express ERa coincide well with some previous studies on the ''cortex-like'' amygdala (Blurton-Jones and Tuszynski 2002; Perez et al. 2004) and cerebral cortex (Kritzer 2002) of the rat. For  Kritzer 2002). The co-expression of CB? neurons with ERb in the sensorimotor and association cortices was rather limited. However, in the piriform and enthorinal cortices, such co-expression was much more extensive (Kritzer 2002). On the other hand, there is no or very weak co-expression between CB? or PV? neurons with ERa in the amygdala (present results; Perez et al. 2004) and various areas of the cerebral cortex (Perez et al. 2004;Kritzer 2002). The present results also demonstrate that substantial populations of CR? neurons in the amygdala of the guinea pig co-express ERa, while these cells were almost without exception immunonegative for ERb. These findings are in agreement with observations in the juvenile rat cerebral cortex, which revealed substantial overlap between CR? and ERa? immunoreactivity but virtually no co-localization of CR? and ERb? immunolabelling (Hayashi et al. 2001;Kritzer 2005). The cells double-labelled for CR and ERa were also described in the adult rat hippocampus (Nakamura and McEwen 2005). Finally, the present results demonstrate that in the medial nucleus and in some other neighbouring regions, there are also subpopulations of CB? and CR? neurons which co-express AR. Data concerning relationships between CB? neurons and AR in the amygdala are not as yet available. In the cerebral cortex AR is mostly present in the pyramidal neurons, and in the most cortical areas virtually none of the non-pyramidal CB? or PV? neurons were found to be immunoreactive for AR (Kritzer 2004). Only in the proisocortical areas of the piriform and entorhinal cortices, many of the nonpyramidal CB? or PV? cells were simultaneously AR? (Kritzer 2004). Data concerning relationships between CR? neurons and AR in the amygdala or cerebral cortex are not yet available.
The present results demonstrate that there is strict cellular separation of ERb in CB? and/or PV? neurons and ERa in CR? cells. In contrast, AR is expressed in non-overlapping subpopulations of CB? and CR? cells, because CR and CB were not co-localized in amygdala neurons of the guinea pig. Such separaration of ERa, ERb or AR within subsets of nonoverlapping subpopulations of GABA? interneurons does not seem to be accidental, and this suggests that these steroidsensitive subsets form a kind of precisely designed interface for sex hormones' actions. For example, there is no or very little overlap in adult life between immunoreactivity for ERb and ERa in any major region of the cortex, hippocampus or amygdala (Kritzer 2002;Perez et al. 2004), although there are brain regions where such co-existence exists (Gréco et al. 2001). There is also no or very little overlap between CR? cells either with CB? or PV? neurons in the cortex, hippocampus or amygdala (present results, and Miettinen et al. 1992;Mascagni et al. 2009). However, in all these areas there is extensive co-localization of CB and PV in the same neurons (present results, and Mascagni et al. 2009). Moreover, there are intriguing parallels in the development of ERb? and PV? neurons on the one hand and ERa? and CR? cells on the other, suggesting that the cellular separation observed in adult life is established during brain development (Kritzer 2005). Specifically, in the postnatal rat cerebrum, like ERb? cells, PV? neurons are identified initially in low numbers around PND 7 (7th day of postnatal life), and from that time steadily increase in number and density to reach mature patterns of labelling by the end of the third postnatal week (Alcántara et al. 1993;Solbach and Celio 1991;Sánchez et al. 1992;Hof et al. 1999). In contrast, like ERa? cells, CR? neurons peak in density early, at about PND 7, transiently include large numbers of pyramidal and non-pyramidal cells, and sharply decrease from peak densities in early postnatal life to much lower, near-adult values by PND 15-21 (Fonseca et al. 1995;Schierle et al. 1997;Hof et al. 1999). Finally, double-labelling studies in the rat and monkey indicated that in the adult mammalian amygdala, there are at least three non-overlapping subpopulations of GABA? interneurons: (1) PV? neurons (most of which also contain CB); (2) SOM? neurons (most of which also contain CB and/or neuropeptide Y); and (3) CR? neurons (most of which also contain cholecystokinin and/or vasoactive intestinal polypeptide) Mascagni 2001, 2002;Mascagni and McDonald 2003;Mascagni et al. 2009). Interestingly, the present results indicate that the first population uses ERb to respond to sex hormones. The second one is endowed with AR. The third subpopulation co-expresses usually ERa, but some of these neurons also utilize AR. It is worth mentioning that the present work was done in the guinea pig, which could be a good model system for studying brain sexual dimorphism, as well as neuroendocrine and developmental processes because of the many similarities to primates and humans (Wallen and Baum 2002;Bartesaghi andSeveri 2004, Kaiser andSachser 2005). As the results in the guinea pig are quite similar to available data obtained from the rat (Blurton-Jones and Tuszynski 2002) and monkey (Perez et al. 2004), it seems that in the amygdala the co-expression patterns of neurons containing calciumbinding proteins and steroid receptors may be similar across various mammalian species.
The present results seem to be very interesting in the light of synaptic studies which indicated that PV?, CB? and CR? neurons perform quite different roles in the inhibitory mechanism of the amygdala (Muller et al. 2007;Miles et al. 1996;Woodruff and Sah 2007a). It should be added, however, that available data are limited to the ''cortex-like'' amygdala only. For example, PV? neurons are considered to be a critical component of the inhibitory circuitry in the basolateral and cortical amygdala, where they form characteristic basket-like plexuses around the unstained somata and cartridges, which represent axo-axonic contacts on initial segments of axons (McDonald and Mascagni 2001;Muller et al. 2005;Woodruff and Sah 2007b). Thus, as in the cortex, many PV? interneurons in both the rodent and primate ''cortex-like'' amygdala appear to be basket or chandelier cells that provide a strong perisomatic inhibition of local pyramidal neurons (Pitkänen and Amaral 1993;Sorvari et al. 1995Sorvari et al. , 1996Woodruff and Sah 2007a, b). Such inhibition is considered to control the output, most notably the synchrony of action potentials of large principal cell populations (Cobb et al. 1995;Miles et al. 1996). Taking into account that many PV? neurons in the amygdala coexpress ERb (present results, and Blurton-Jones and Tuszynski 2002), and that the numbers of these cells in various amygdaloid nuclei are sexually dimorphic (Równiak et al. 2015;Stefanova et al. 1997), the mechanism described above may be strongly modulated by circulating steroid hormones and sex-specific. CB? neurons perform a dual role in the intrinsic inhibitory mechanism of the amygdala. In rodents (but not primates) some of them appear to be basket cells, which provide a strong perisomatic inhibition of local pyramidal neurons, and such cells co-express PV but do not contain somatostatin and/or neuropeptide Y (Pitkänen and Amaral 1993;Sorvari et al. 1995Sorvari et al. , 1996Woodruff and Sah 2007a, b). The other CB? neurons are considered as dendrite-targeting interneurons, providing a robust innervation of dendrites in the rat and primate amygdala (Mascagni et al. 2009). Such neurons contain somatostatin and/or neuropeptide Y (but not PV) (McDonald and Mascagni 2002;Muller et al. 2007), and are responsible for the control of the efficacy and plasticity of inputs from specific sources that terminate in the same dendritic domain (Cobb et al. 1995;Miles et al. 1996). Interestingly, it seems probable that Brain Struct Funct (2017)  However, such studies in the neocortex revealed that CR? neurons are mostly dendrite-targeting, and their buttons typically innervate dendritic shafts, and less often dendritic spines or somata (Meskenaite 1997). Moreover, at least in some cortical areas and layers, CR? neurons innervate preferentially other GABA? neurons (Melchitzky and Lewis 2008), and by inhibiting other classes of inhibitory interneurons they may exert a significant disinhibitory effect on the pyramidal neurons (Meskenaite 1997;Gonchar and Burkhalter 1999). Analogous relationships were also described in the hippocampus, where CR? interneurons were found to be specialized to control other interneuronal types . Taking into account that many CR? neurons in the amygdala co-express ERa or AR (present results), the mechanisms described above may be strongly modulated by both oestrogen or testosterone, and thus sex-specific.
The present results and some other studies (Krȩ_ zel et al. 2001;Tan et al. 2012;Blurton-Jones and Tuszynski 2002) provide evidence that there is a relationship between steroid hormone receptors and inhibitory tone in the amygdala. Thus, the question arises what may be the functional significance of such a relationship. There are several lines of evidence that modulation of inhibitory tone by oestrogen and/or androgen via steroid receptors on GABAergic cells may be important for proper amygdala physiology, and any alterations in such modulation may lead to various psychiatric disorders, which differentially affect men and women (prevalence, severity) (Krȩ_ zel et al. 2001;Kinrys and Wygant 2005). For example, studies in ERb knock-out mice revealed that mutant females displayed increased anxiety and/or depression-like behaviour associated with a reduced threshold for the induction of synaptic plasticity in the basolateral amygdala (Krȩ_ zel et al. 2001). A reduced threshold in this structure could result in exaggerated responses to normally innocuous stimuli or environments, a characteristic feature of fear and anxiety disorders. Whether a reduction of GABA responses in the basolateral amygdala is the mechanism by which anxiety is increased in ERb mutant mice remains to be tested, but some evidence indicates that pharmacological manipulation of GABA type A receptors in wild type mice can mimic some of the effects of ERb null mutation (Krȩ_ zel et al. 2001). Moreover, selective ERb agonists typically exert potent anxiolytic activity (Lund et al. 2005;Oyola et al. 2012).
Such anxiolytic effects are mediated by the cortex and hippocampus, and most probably also by the basolateral amygdala, by elevation of GABAergic inhibitory signalling and reduction of glutamatergic excitatory drive (Tan et al. 2012). On the other hand, male ERa knock-out mice display reduced aggression (Ogawa et al. 1997(Ogawa et al. , 1998, while selective ERa agonists typically elevate anxiety and/or aggression (Nomura et al. 2002). All these data from ERb and ERa knock-out mice are in accordance with recent physiological studies (Adhikari et al. 2015;Hong et al. 2014) and the present description of ERa, ERb and/or AR enrichment in several amygdala nuclei. For example, the basomedial nucleus is the region in the amygdala, which is directly engaged in anxiolysis, however, the medial and cortical nuclei cannot be excluded as well (Adhikari et al. 2015). Interestingly, in all these nuclei, ERb is very abundant according to present results. On the other hand, in the medial nucleus, which is especially rich in ERa and other steroid receptors, there are many aggression-promoting GABAergic neurons (Hong et al. 2014).

Conclusions
The results of the present study provide evidence that in the amygdala of the guinea pig subpopulations of CB? and PV? neurons co-express ERb, while many of CR? neurons co-express ERa. In addition, in the selected nuclei, there are also subpopulations of CB? and CR? neurons which co-express AR. The localization of ERa, ERb or AR within subsets of GABAergic interneurons across diverse amygdaloid regions suggests that steroid hormones may exert a significant influence over local neuronal activity by directly modulating inhibitory tone. The control of inhibitory tone may be one of the mechanisms whereby oestrogen and androgen could modulate amygdala processing in a sex-specific manner. Another mechanism may be thorough steroid-sensitive projection neurons, which are most probably located in the medial and central nuclei.