Expression of type one cannabinoid receptor in different subpopulation of kisspeptin neurons and kisspeptin afferents to GnRH neurons in female mice

Adult, female CD1 mice (n = 20) were used for the RNAscope in situ hybridization histochemistry (ISHH), and adult, female CD1 (n = 10) and Kiss1-Cre-GFP mice (Kiss1^{tm1.1(cre/EGFP)Steiner)}, Jackson Laboratory, stock No. 017701, n = 20) were used for the immunohistochemical studies. The animals were housed under controlled lighting [12:12 h light–dark cycle, lights on at 07:00 h, and temperature (22 ± 2 °C)] conditions, with access to food and water ad libitum. The CD1 mice were ovariectomized (OVX) with 3-week recovery time, and thereafter, 15 of them received a two-step subcutaneous estrogen substitution or vehicle, as described before (Bosch et al. 2013). Briefly, the animals (n = 30) received on the first and the second days 0.25 or 1.5 μg 17β-estradiol benzoate (EB) in 50 μl oil (n = 15) or 50 μl oil vehicle (n = 15) at Zeitgeber time (ZT) 4–5 (Table 1). The uterus weight of the OVX + Oil mice differed significantly from those treated with EB (0.049 ± 0.014 g versus 0.16 ± 0.014 g, respectively, p < 0.001). The Kiss1-Cre-GFP animals were all OVX and received a viral construct AAV-EF1a-DIOhChR2 (H134R)-EYFP injection [200 nl 1:2 solution of the viral stock) into the anteroventral periventricular nucleus (AVPe, Bregma AP 0.25 mm, ML 0.2 mm, DV 4.6 mm, n = 5)], ARC (Bregma AP −2.15 mm, ML 0.25 mm, DV 5.25 mm, n = 8), or posterodorsal subdivision of the medial amygdaloid nucleus (MePD Bregma AP −1.9 mm, ML 2.0 mm, DV 4.9 mm, n = 7) to label the KP neurons. Three weeks after the surgery, these animals were also treated with EB similarly to the above protocol.

Five slides of the rostral periventricular area of the third ventricle (RP3V) [containing the AvPe and/or the periventricular hypothalamic nucleus (Pe)], and five slides of the ARC regions (altogether 10 sections of five OVX-EB and 10 sections of five OVX + oil mice for each regions) were selected for each hybridization. The pre-treatment (#3,20,513-USM) and hybridization (#3,20,293) protocols of the RNAscope technique (Advanced Cell Diagnostics Inc., Newark, CA, USA) were used, which allowed the application of the Multiplex reagent kit. The hybridization cocktail contained the following probes: for channel 1, it was Kiss1 (catalog #476291_O1, XM_006529679, region 121 – 1376); for channel 2, it was either SLC32A1 (VGAT, catalog #3,19,191-C2, NM_009508.2, region 894–2037) or SLC17A6 (VGLUT2; catalog #3,19,171-C2, NM_080853.3, region 1986–2998); and for channel 3, it was CNR1(CB1, catalog # 4,20,721-C3, NM_007726.3, region 530–1458). The “B” variant of Amp4 was used, which labeled channel 1–3 with Alexa488 (green), Atto550 (red), and Atto647 (far red), respectively. The sections were coverslipped with Prolong Antifade kit (Molecular Probes, Leiden, The Netherlands) after the amplification/labeling steps and counterstaining with DAPI.

The quadruple-labeled sections were scanned in a Nikon C2 confocal microscope (Nikon, Japan) using the 20 × objective. Multiple stacks of optical slices (1024 × 1024 pixels, z-steps 0.6 µm) were obtained from the −0.6 to −2.4 µm layer of the sections by scanning the full RP3V and ARC regions on one side. The fluorochromes were excited with laser lines 488, 561, and 641 nm. The DAPI nuclear staining was also detected using the 405 nm laser. Laser intensities and other acquisition parameters were kept the same during the whole scanning. Using the Image J software, the different channels in the image stack were combined into maximum intensity projections, and binarized and merged into single TIFF images. The DAPI channel was also saved as separate TIFF images. Lit pixels were counted within regional and cellular borders, which were determined manually based on the DAPI image and kisspeptin signals. A cell was deemed as positive for a given mRNA when the number of lit pixels within the cell area were higher (containing more than five pixels) than the number of lit pixels in identical sized area of the white matter (maximum five pixels). Data were analyzed by one-way ANOVA, and significant difference was determined by the Holm–Sidak method.

The animals were perfused transcardially with phosphate-buffered saline (PBS 0.1 M) containing 4% paraformaldehyde (PFA). The brains were removed, post fixed for 24 h, and transferred into 30% sucrose for cryoprotection, and then, 30 µm thick coronal sections were cut on a freezing microtome by collecting every third sections into the same well.

A group of sections were mounted from each brain with RP3V (n = 5), ARC (n = 8), or MEA (n = 7) injections of the viral tracer. Brains showing successful labeling of KP neurons in consecutive sections were selected for subsequent multiple-label immunofluorescence staining of YFP in KP cells, GnRH, and CB1.

After the endogenous peroxidase activity had been quenched with 0.5% hydrogen peroxide (20 min), sections were permeabilized with 0.5% Triton X-100 (23,472–9, Sigma, 20 min), and treated with 2% normal horse serum (20 min) to reduce non-specific antibody binding. All treatments and interim rinses in PBS (3 × 5 min) were carried out at room temperature, except for incubation in the primary antibodies and fluorochromes, which was carried out at 4 °C. Sections were incubated in a cocktail of the primary antibodies for 72 h and the fluorochrome-labeled secondary antibodies overnight with a 2 h rinse in TRIS at RT in between. Primary and secondary antibodies used in the immunohistochemical procedure are listed in Table 1.

Thus, guinea pig anti-GnRH (#1018, gift from Dr. Erik Hrabovszky, 1:50,000) (Hrabovszky et al. 2011), sheep anti-KP (#053, gift from Dr. Alain Caraty, 1:1000) and rabbit anti-CB1, ImmunoGenes-ABS Zrt, 1:1000) primary antibodies, and FITC-donkey-anti-guinea pig IgG (H + L) (#706-095-148, Jackson ImmunoResearch Laboratories, 1:500), CY3-donkey-anti-sheep IgG (Fab)2 (#713-165-147, Jackson ImmunoResearch Laboratories, 1:3000), and CY5-donkey-anti-rabbit IgG (H + L), 711-175-152, Jackson ImmunoResearch Laboratories, 1:2000) secondary antibodies were used to stain sections of CD1 mice. To maximize the visualization of the axonal boundaries of KP neurons, mice expressing YFP in KP cells of the virus injected KP-CRE animals were used for the immunofluorescent detection of CB1 in afferents of GnRH neurons. In this case, guinea pig anti-GnRH (#1018, 1:50,000) (Hrabovszky et al. 2011), rabbit anti-GFP (for KP, AB10145 Millipore, 1:2000), and goat anti-CB1, 1:600 (Makara et al. 2007), primary antibodies, and CY5-conjugated donkey anti-guinea pig IgG (#706-175-148, Jackson ImmunoResearch Laboratories, 1:2000, 2 h), FITC-conjugated donkey anti-rabbit IgG (#711-095-152, Jackson ImmunoResearch Laboratories, 1:1000, 2 h), and CY3-conjugated donkey anti-goat IgG (#706-165-147, Jackson ImmunoResearch Laboratories, 1:2000, 2 h) secondary antibodies were used. Sections were then rinsed in TRIS (2 h), mounted onto glass slides and cover slipped with Moviol.

The triple-labeled sections were scanned in a Nikon A1R confocal microscope (Nikon, Japan) using the  × 20 and the 60 × oil immersion objectives. Multiple stacks of optical slices (1024 × 1024 pixels, z-steps 0.15 µm) were generated, which contained the KP-IR fibers in apposition to the cell bodies and the processes of GnRH-IR neurons. The separately recorded green, red, and far-red channels were merged and displayed with the ImageJ software running on an IBM compatible personal computer. Orthogonal views from different planes (x/y, x/z or y/z) of the confocal microscopic images were used to analyze the KP-immunoreactive fibers for apposition to GnRH neurons and CB1-immunoreactivity. To enable three-dimensional (3D) analyses, the images were further processed using the software Amira (6.0, Visual Imaging Group). The stacks of the optical slices were loaded into the visualization program and rendered in three dimensions with surfaces generated from above threshold immunoreactivity. The threshold was set individually for each image and color channel to minimize any noise, while maintaining the proper cellular boundaries. The surfaces generated from the three channels in the same optical volume were visualized to check for cell-to-cell contacts, and the presence of CB1-immunoreactivity in KP fibers associated with GnRH neurons. This enabled verification of the findings from the two-dimensional confocal image analyses.

The RNAscope in situ hybridization technique detected mRNA signals for KP, CB1, VGLUT2, and VGAT concurrently in ovariectomized mouse models treated for 2 days either with EB or oil vehicle. In agreement with the previous reports (Gottsch et al. 2009), the KP mRNA signal showed an estrogen-dependent alteration in the RP3V (Fig. 1) and ARC (Fig. 2), and this has validated the animal models used in this study. Thus, there was a profound and area-specific effect of estrogen-treatment on the level of KP mRNAs in the RP3V and ARC (Fig. 3A). While EB-treatment significantly increased the mRNA levels for KP in the RPRV region (from 1.83 ± 0.41 to 4.51 ± 0.8, p < 0.02), the same treatment reduced it in the ARC (from 1.95 ± 0.56 to 0.63 ± 0.05, p = 0.02). The expression levels of VGAT and VGLUT2 were relatively low in both regions compared to the neighboring medial preoptic area (MPA) and/or the ventromedial hypothalamic nucleus (VMH) (Figs. 1 and 2). No significant differences were detected between the two animal models for VGAT mRNAs in the ARC and VGLUT2 mRNAs in both the RPRV and ARC. The VGAT mRNA signal was, however, significantly lower (p = 0.028) in the RP3V region of OVX-EB animals (0.29 ± 0.26) than in the OVX + OIL group (0.59 ± 0.07) (Fig. 3A). The CB1 mRNA was also relatively low in the RP3V and ARC compared to the neighboring MPA and VMH regions (Figs. 1 and 2), respectively. Additionally, CB1 mRNA showed an estrogen-sensitivity with signal being lower both in the RP3V (0.35 ± 0.14 vs 1.2 ± 0.19, p = 0.02) and ARC (0.33 ± 0.05 vs 0.16 ± 0.04, p = 0.05) of EB-treated mice compared to mice receiving vehicle only (Fig. 3A).

The same sections were analyzed at cellular levels in the RP3V and ARC for co-expression of KP with CB1 and either VGAT or VGLUT2 (Fig. 3B–E). In the RP3V, 93.4 ± 2.7% of KP neurons expressed CB1 in the OVX + OIL group, and 32 ± 6.9% of them in the OVX + EB group, whereas in the ARC, these values were 15.5 ± 2 and 49.3 ± 4.4%, respectively (Fig. 3F).

KP neurons express VGAT and VGLUT2 mRNAs both in the RP3V and the Arc. In the RP3V, 48.4 ± 10.6% of KP neurons expressed VGAT in the OVX + OIL group, and 69.2 ± 13.4% of them in the OVX + EB group, whereas in the ARC, these values were 76.5 ± 2.8 and 73.5 ± 5.9%, respectively. Concerning the co-expression of VGLUT2 with KP, the value was 23.5 ± 1.9% in the RP3V of OVX + OIL group, and 31.4 ± 2.5% in the same region of the OVX + EB group, whereas these values were in the ARC 83.7 ± 2.4 and 88.3 ± 3.8%, respectively. The estrogen level of the animals had no significant influence on the percentage of KP neurons co-expressing VGAT or VGLUT. The percentage of VGLUT2-positive KP neurons was, however, significantly higher in the ARC than in the RP3V (p < 0.001) (data not shown). When CB1 mRNA expression was analyzed in the VGAT- or VGLUT2-positive KP neurons, a profound site-specific effect of EB was revealed in the RP3V and ARC. Thus, the percentage of the CB1 expressing VGLUT2-positive KP neurons was significantly increased in the ARC of estrogen-treated OVX mice, compared to the oil-treated animals (43.5 ± 5 vs. 14.8 ± 2.4%, p < 0.005). The same treatment resulted in a significant decrease of CB1 expression in VGAT-positive KP neurons in the RP3V region of the brain (19.1 ± 3.5 vs. 67.3 ± 10.9%, p < 0.05) (Fig. 3F).

The presence of CB1 in KP afferents of GnRH neurons was studied in CD1 and KP-CRE mice (Fig. 4) by immunohistochemical labeling of preoptic sections for CB1, GnRH, and KP (Fig. 5A–C) or alternatively YFP (Fig. 5D–F), expressed after transmission of its gene by viral-infection of KP-CRE neurons. Because of its dominant membrane localization, CB1 appeared rarely in overlap with KP-immunoreactivity marking primarily secretory granules in the axon terminals (Fig. 5A–C). Therefore, axonal projections of KP-CRE-expressing neurons were traced with yellow fluorescent proteins, which also marked the borders of axon terminals in contact with GnRH neurons (Fig. 5D–F).

The viral construct (AAV-EF1a-DIOhChR2 (H134R)-EYFP) was delivered to all major KP-populations of the mouse brain, where it was translated to EYFP and transported together with the ChR2 to the processes of KP-CRE cells including the cell membrane (Fig. 4). YFP-positive processes in apposition to GnRH neurons were rarely seen in the POA of mice, if the viral construct was injected into the ARC or the MePD. In contrast, such afferents were seen more often in mice, which were targeted with the viruses in the RP3V region.

The YFP expressed by AAV-infected RP3V neurons identified the cellular borders of KP neuronal processes in the MPA, including those, which were in juxtaposition to GnRH-IR perikarya and processes (Figs. 5D–F). CB1-IR was found in about two-thirds of these KP neuronal processes (Table 2, 70 ± 2.9 and 59 ± 4.1% of them (n = 187) were in apposition to GnRH perikarya and processes, respectively).

Levels of CB1 expression are highly variable among different brain locations and cell types. While the cerebral cortex and hippocampus contain a very high level of CB1 protein, expression levels are relatively low in hypothalamic regions (Wittmann et al. 2007). Despite its low levels, a large body of evidence supports strong cannabinoid-dependent signaling in the regulation of hypothalamic functions (Gammon et al. 2005; Pagotto et al. 2006; Tasker 2006; Brents 2016). Concerning the reported role of CB1 in the regulation of reproduction, and the critical functions of kisspeptin neurons in this regard, in the current study, the KP cell-specific expression of CB1 was investigated using multiple labeling in situ hybridization and immunofluorescent histochemical techniques. During the ovarian cycle, the hypothalamic regulatory circuits operate under the influence of estrogens’ negative and positive feedback effects. Therefore, we have used the animal models with reduced serum ovarian hormone levels achieved by 3-week time gonadectomy (Baumgartner et al. 2019), and the hormone-replaced pairs treated subsequently with a priming and a surge-inducing dose of estrogen (Bosch et al. 2013). As expected, this treatment regime resulted in a significant difference between the uterine weights of the two models by increasing the weights fourfold compared to controls. In addition, the appearance of KP transcript levels showed high estrogen-dependency in the RP3V and ARC regions, as it was reported earlier (Lehman et al. 2010, Navarro et al. 2009, Smith et al. 2005). Thus, in the RP3V, KP transcript levels were high in the estrogen-treated animals, whereas strong signals were detected in the ARC in the absence of estrogen (Fig. 3A). The estradiol-treated model was chosen also to investigate the afferents on GnRH neurons based on the observation of Chan et al. (2011). They reported that the positive feedback levels of estradiol stimulate a robust increase in spine density and most likely in synaptic inputs of GnRH neurons.

Our previous mapping studies in male mice (Hrabovszky et al. 2012) revealed a differential expression of mRNAs for CB1, the GABAergic marker GAD65, and the glutamatergic marker VGLUT2 in the RP3V and ARC regions, where the two major subpopulations of KP neurons are located. Relatively low abundance of the hybridization signal and high-to-moderate-to-low number of labeled cells were found for CB1 in the AVPe, the Pe, and the ARC nucleus, respectively. CB1 mRNA was detected in both GABAegic and glutamatergic neurons in these regions. In the current study, the mRNA signals for CB1, VGAT and VGLUT2 were relatively low in both the RP3V and ARC of female mice and their relative abundance compared to the neighboring brain areas (i.e., MPA or the VMH) was similar to the one detected previously in male mice (Hrabovszky et al. 2012).

Estrogen treatment resulted in a down-regulation of CB1 expression in both regions (Fig. 3A). This agrees with the observation of Riebe and Gorzalka, (Riebe et al. 2010), showing that OVX females have higher amounts of hypothalamic cannabinoid receptor binding relative to both cycling and OVX + E2 females. The current result, therefore, might indicate a reduced involvement of CB1 mediating suppression in local circuit activity, which generate higher frequency postsynaptic events in target cells in the presence of estrogen (Glanowska and Moenter 2011).

Similarly to CB1, expression of VGAT mRNA was lower in RP3V of the OVX + EB mice compared to the oil-treated ones. This seems to be congruent with the observation of Ottem et al. (Ottem et al. 2004) showing a decrease of the VGAT-containing vesicles in the RP3V of female rats at the time of the surge.

Gonadal hormones were reported to alter also the VGLUT2 expression in these regions. E2-treatment was shown to increase the VGLUT2 immunoreactive vesicles in the afferents of GnRH neurons in ovariectomized rats (Ottem et al. 2004). An increase of VGLUT2 mRNAs by estrogen-treatment was reported also in the ARC of ovariectomized mice (Qiu et al. 2018). The positive effect of gonadal hormones on VGLUT2 expression is, however, not unambiguous, since gonadectomy led to a significant elevation of VGLUT2 mRNA in male mice, compared to the levels observed in gonad intact conditions (Nestor et al. 2016).

The comparison of current RNAscope signals of the EB or oil-treated animals, however, did not reveal a significant difference for VGLUT2 signals neither in the RP3V nor in the ARC. This may indicate the presence of mixed glutamatergic cell populations in these regions in which expression of VGLUT2 is differently or even antagonistically regulated by estrogen.

Analyses of the RNAscope data at cellular levels revealed the presence of CB1 transcripts in KP neurons in both the RP3V and ARC regions, but the hormonal conditions had a great impact on the level of co-expression. Nearly all preoptic KP neurons expressed CB1 in the OVX + OIL group (93.4 ± 2.7%), but this value fall to one-third in the OVX + EB group (32 ± 6.9%). The opposite was observable for ARC KP neurons, showing low levels of co-expression in the OVX + OIL group (15.5 ± 2%), whereas CB1 mRNA expression was detectable in nearly half of the KP neurons in the OVX + EB group (49.3 ± 4.4%). By involving data on the mRNA levels for VGAT and VGLUT2 in the analyses, the presence of CB1 mRNA was also examined separately in the GABAergic and glutamatergic subpopulations of KP neurons. Both subpopulations proved to be positive for CB1 mRNA and the co-localization levels varied again according to the hormonal status of the animals. The percentage of the GABAergic KP neurons expressing CB1 was significantly lower in the RP3V of OVX + EB-treated mice, indicating that CB1 signaling might be reduced or suspended in this cell population at high estrogen levels.

Conversely, the percentage of the glutamatergic KP neurons expressing CB1 (Fig. 3F) was significantly higher in the ARC of the estrogen-treated animals compared to controls. The dramatic differences seen in the preoptic versus arcuate co-localization levels can be explained by the antagonistic regulation of KP transcript in these regions, which is combined with the reduced expression of CB1 mRNA in response to estrogen-treatment in both locations. Thus, lower co-localization levels can be found in RP3V in the OVX + EB model and in the arcuate nucleus in the OVX + Oil model, in which models the KP expression is high and the CB1 message is downregulated.

Both VGAT or VGLUT2 mRNA-containing KP cells could be detected in the preoptic and ARC regions, which is in concordance with studies detecting VGAT/or VGLUT2 mRNA in KP-CRE mice (Cravo et al. 2011) or KP-immunoreactivity in VGAT-CRE or VGLUT2-CRE mice (Cheong et al. 2015). Thus, about half to two-third of the KP neurons were positive in the current study for VGAT mRNA, and about one quarter to one-third of KP neurons were positive for VGLUT2 mRNA in the RP3V. Most of the ARC KP neurons were also positive for VGLUT2 mRNA and three quarter of them also expressed the VGAT mRNA. Of note, the level of co-localization of KP and VGAT or VGLUT mRNAs may depend on the animal model used, since quantitative PCR studies were able to find relatively few, or no RP3V KP cells with Slc17a6 (VGLUT2) mRNA or no ARC KP cells with Slc32a1 (VGAT) mRNA (Qiu et al. 2018) in Kiss1Cre:GFP mice (Gottsch et al. 2011). Furthermore, translation of neither amino acid transporter markers has been proved to functional transporter proteins, respectively, in the RP3V or ARC KP cells of these animals (Qiu et al. 2018).

KP neurons in all major locations are involved in the regulation of GnRH secretion. We have traced projections of KP neurons located in the RP3V, ARC, and MePD (Fig. 4), and in concordance with the previous observations (Kallo et al. 2012; Pineda et al. 2017; Yeo 2013; Yeo and Herbison 2011; Yeo et al. 2019; Wintermantel et al. 2006), found KP fibers in apposition to GnRH cells to originate primarily from the RP3V cells.

CB1-IR was found in the majority of these KP neuronal processes independent of being in apposition to the perikaryon or proximal processes of GnRH neurons (Fig. 5C and F). This suggests that neurotransmitter release from the KP afferents to GnRH neurons is also under regulation of endocannabinoids.

The endocannabinoid control is inhibitory by reducing the frequency of mPSCs and the firing activity of GnRH neurons in slices of metestrus female mice (Balint et al. 2016). This phase of the cycle is characterized by low serum estrogen levels, which triggers the synthesis and release of 2-AG from GnRH neurons via activating membrane-associated, rapid ERβ-mediated processes in these neurons. Acute administration of a high dose of estrogen to the slice had no effect on this inhibition, indicating that rapid, non-genomic actions of estrogen might not interfere with the CB1 signaling. The inhibition, however, does not seem to operate at high serum estrogen levels in proestrus GnRH neurons (Farkas et al. 2018). A possible explanation for this could be a longer exposure to and most likely genomic action of estrogen seen in the current study, which resulted in the down-regulation of CB1 expression in both the RP3V and ARC KP neurons. Whether it has involved both the GABAergic and glutamatergic KP afferents of GnRH neurons needs to be further investigated. However, the down-regulation of CB1 in the glutamatergic KP afferents may have contributed to the appearance of the glutamate receptor mediated PSC-es in the GnRH neurons seen on the proestrus afternoon (Farkas et al. 2018).

Glutamate released by ARC KP neurons targets KP neurons in this brain region and may contribute to the synchronized activity increase in populations of KP neurons during pulse generation. Glutamate released by ARC KP neurons also targets preoptic KP neurons, which in turn can enhance stimulation of GnRH neurons leading ultimately to a GnRH surge (Qiu et al. 2016). Therefore, a down-regulation of CB1 mRNA expression in the glutamatergic KP neurons by the preovulatory estrogen rise would facilitate an increased firing of KP afferents to other KP neurons, which in turn could enhance the activity of GnRH neurons.

KP neurons also innervate the Arc POMC neurons. The glutamatergic and GABAergic afferents of POMC neurons are under tonic, endocannabinoid-induced inhibition (Hentges et al. 2005), which seems to be differentially regulated by estrogen. The estrogen-treatment, given 24 h prior to experimentation, increased the mEPSC frequency, and markedly decreased the potency of CB1 agonists to decrease mEPSC frequency in POMC neurons. In contrast, estrogen potentiated the cannabinoid-induced decrease in mIPSC frequency (Nguyen and Wagner 2006). The opposing effects of estrogen on the cannabinoid regulation of amino acid neurotransmission lead ultimately to the excitation of POMC neurons. This shows similarities with the effect of estrogen on the tonic endocannabinoid signaling of GnRH neurons. As the GABAergic input of GnRH neurons is facilitatory due to their high intracellular chloride levels (DeFazio et al. 2002), an estrogen-mediated down-regulation of CB1 in both GABAergic and glutamatergic afferents, including those originating from KP neurons, could potentially increase the excitation of GnRH neurons. These needs, however, to be confirmed by electrophysiological recordings.