Morphological assessment of GABA and glutamate inputs to GnRH neurons in intact female mice using expansion microscopy

The roles GABAergic and glutamatergic inputs in regulating the activity of the gonadotrophin‐releasing hormone (GnRH) neurons at the time of the preovulatory surge remain unclear. We used expansion microscopy to compare the density of GABAergic and glutamatergic synapses on the GnRH neuron cell body and proximal dendrite in dioestrous and pro‐oestrous female mice. An evaluation of all synapses immunoreactive for synaptophysin revealed that the highest density of inputs to rostral preoptic area GnRH neurons occurred within the first 45 µm of the primary dendrite (approximately 0.19 synapses µm‐1) with relatively few synapses on the GnRH neuron soma or beyond 45 µm of the dendrite (0.05‐0.08 synapses µm‐1). Triple immunofluorescence labelling demonstrated a predominance of glutamatergic signalling with twice as many vesicular glutamate transporter 2 synapses detected compared to vesicular GABA transporter. Co‐labelling with the GABAA receptor scaffold protein gephyrin and the glutamate receptor postsynaptic density marker Homer1 confirmed these observations, as well as the different spatial distribution of GABA and glutamate inputs along the dendrite. Quantitative assessments revealed no differences in synaptophysin, GABA or glutamate synapses at the proximal dendrite and soma of GnRH neurons between dioestrous and pro‐oestrous mice. Taken together, these studies demonstrate that the GnRH neuron receives twice as many glutamatergic synapses compared to GABAergic synapses and that these inputs preferentially target the first 45 µm of the GnRH neuron proximal dendrite. These inputs appear to be structurally stable before the onset of pro‐oestrous GnRH surge.


| INTRODUC TI ON
The gonadotrophin-releasing hormone (GnRH) neurons represent the final output cells of the neural network controlling fertility in mammals. Driven by pulse and surge generators, GnRH neurons release GnRH into the median eminence in an episodic manner to create pulsatile or surge profiles of circulating gonadotrophin hormones. 1 The surge generator is only usually found in female mammals with the mid-cycle GnRH surge generating a luteinising hormone (LH) surge that triggers ovulation. 2 Precisely how the GnRH neurons become suddenly and intensely activated to create the massive outpouring of GnRH that occurs at the surge remains unclear. 3,4 There is little doubt that sustained high levels of circulating oestradiol are obligatory for the GnRH surge in spontaneously ovulating mammals 5 and this is relayed to the GnRH neurons by afferent inputs that express the key oestrogen receptor, ESR1. [6][7][8] These neurons very likely target the GnRH neuron cell body and proximal dendrites to trigger the intense neuronal activation required for the surge. 9,10 The prime candidate for triggering the surge is presently the ESR1-expressing kisspeptin neurons of the rostral periventricular area of the third ventricle (RP3V) that directly innervate the GnRH neuron cell body and dendrites. 11,12 However, there has been a long-standing interest and focus on the role of GABAergic and glutamatergic inputs to the GnRH neuron in surge generation. 3,4,13 For example, the deletion of ESR1 from all GABAergic or vesicular glutamate transporter 2 (VGLUT2)-glutamatergic neurons abolishes the LH surge likely independent of kisspeptin. 14 Electrophysiological studies have also reported that GABA A and glutamate receptor transmission at the GnRH neuron can change at the time of the LH surge, 15,16 although this depends on the animal model used. 17,18 Hence, it remains possible that steroidinduced plasticity in GABA and glutamate signalling to the GnRH neuron cell body may play a role in the generation of the GnRH/ LH surge.
There is considerable evidence for gonadal steroids to evoke structural plasticity within the brain. 19 As such, the elevated follicular-phase levels of oestradiol may result in morphological re- In the present study, we have taken advantage of recent developments in super-resolution microscopy to re-examine the issue of structural GABA and glutamate synaptic plasticity in relation to the GnRH neurons at the onset of the GnRH surge and done so in intact female mice. Expansion microscopy (ExM) utilises isotropic tissue swelling to expand the sample so that regular confocal imaging can achieve the necessary resolution required to identify bona fide synapses in the brain. 22

| Mice
Female C57BL/6 GnRH-green fluorescent protein (GFP) mice 23 aged between 2 and 4 months old were used for immunohistochemistry and ExM. The mice were maintained under a 12:12-hour light/dark photocycle (lights off 6.00 pm) with access to food and water available ad lib. Experimental procedures were undertaken in accordance with the University of Otago Animal Welfare and Ethics Committee. Female GnRH-GFP mice exhibiting at least 3 regular oestrous cycles were killed for experiments between 2.00 pm and 4.00 pm on dioestrus (N = 4) or pro-oestrus (N = 4). The pro-oestrous LH surge commences at 5.00 pm (1 hour before lights off) in this mouse colony. 24

| Expansion microscopy
The ExM was carried out as previously reported in detail. 10  proportions. Next, the gel-embedded sections were trimmed and digested overnight with proteinase K at 37℃. The samples were incubated with strepavidin-568 (dilution 1:1500) at 37℃ for 3 hours.
The 568-fluorophore was not compatible with the gelation process; therefore, it was added after the gelation step. Increasing the incubation temperature improved the diffusion of streptavidin molecules in the gel. Finally, expansion was undertaken by adding water every 20 minutes, up to 5 times. Expanded samples were placed in an imaging chamber and cover slipped using #1.5 (0.17 mm) glass.

| Image analysis
The definition of a synapse with ExM relies on the detection of sufficient overlap between the fluorescent labels of the pre-and postsynaptic markers. 10 Although the tissue expansion increases the apparent final resolution of ExM into the superresolution realm, the confocal microscopy used for imaging the specimen is still diffractionlimited. This means that pre-and postsynaptic structures divided by a synaptic cleft even with this technique show some overlap of their image intensity profiles. Using VGAT and gephyrin immunofluorescence to define GABA synapses on GnRH-GFP neurons, we have previously calculated that, with our microscope (see details below), the overlap between a presynaptic marker and cytoplasmic GFP in a GnRH-GFP neuron must be > 0.95 μm (0.23 μm pre-expansion) in the "side-on" fluorescence profile view to be considered a synapse 10 (ie, every VGAT-gephyrin synapse has an overlap between VGAT and GnRH-GFP fluorescence greater than 0.95 μm). Not all synapses can be rotated to provide a side-on view and, in those cases, a "z-stack/ face-view" profile of multiple imaged synapses is required and with this orientation, the fluorescence needs to overlap by > 1.75 μm TA B L E 1 List of antibodies and software (0.42 μm pre-expansion) to account for decreased resolution along this imaging axis and to be considered a synapse. 10 To determine the degree of tissue swelling, the diameter of the nuclei of individual GnRH neurons were measured before and after gelation.
All imaging was performed using an A1R upright confocal microscope (Nikon, Tokyo, Japan) with images captured using a water-

| Statistical analysis
Statistical analyses were performed using Prism (GraphPad Software Inc., San Diego, CA, USA). Two-way ANOVA with repeated-measures was applied to compare the mean number of synapses on soma and on dendrite segments every 15 µm between pro-oestrous and dioestrous animals. Full statistical details for each analysis are provided in Table 2.

| RE SULTS
By assessing the size of the GnRH neuron nuclei before and after expansion, we determined that the ExM procedure enlarged brain sections by a factor of 4.17 ± 0.01 (n = 43).

| Distribution of synaptic inputs along the GnRH neuron soma and proximal dendrite
We first assessed the distribution of all synapses along the GnRH neuron using synaptophysin as a general presynaptic marker. 21,30 All mice exhibited GnRH neurons surrounded by typical punctate synaptophysin immunoreactivity ( Figure 1) that was completely absent when the primary antibody was omitted from the immunohistochemistry (not shown). Many close appositions were detected between synaptophysin boutons and GnRH neurons but only some satisfied the criteria for a synapse. Figure 1 shows examples of bona fide synapses in the z-stack face-on view ( Figure 1i) and sideon view ( Figure 1ii) and a close apposition that is not forming a direct synapse (Figure 1iii). Quantitative analysis revealed that the average density of synapses peaked in the 0-45 µm segment of the proximal dendrite (0.19 ± 0.01 synapses µm -1 ) compared to the soma (0.08 ± 0.01 synapses µm -1 ) and beyond 45 µm on the dendrite (0.05 ± 0.01 synapses µm -1 ) ( Figure 2 and Table 3). No

| Distribution of GABAergic and glutamatergic inputs along the GnRH neuron soma and proximal dendrite
We assessed glutamatergic and GABAergic inputs to the GnRH neuron by labelling for synaptophysin and GFP combined with the presynaptic markers VGAT or VGLUT2. The specificity of VGAT and VGLUT2 antibodies have been described previously. 25,26 The absence of primary antibodies in the present experiments resulted in no detectable immunofluorescence (not shown).
Glutamatergic synapses were defined as being boutons that contained both synaptophysin and VGLUT2 immunofluorescence that formed a sufficient ExM overlap with the cytoplasmic GFP of the GnRH neuron ( Figure 3A). Spatial analysis revealed a distribution of VGLUT2 synaptic density that was similar to that of  Table 3.   with very few synapses detected elsewhere ( Figure 3D and Table 3).

TA B L E 2 Statistical reporting
Again, no differences were detected between dioestrous (16 cells, ( Figure 3B). The densities per 15µm segment are given in Table 3.
An alternative and potentially more definitive method for determining glutamatergic and GABAergic synapses is to combine the  the latter are restricted to proteins that have antisera well characterised to work in mouse brain tissue. We used Homer1, a scaffold protein at glutamatergic synapses 31 and gephyrin that clusters GABA A receptors to synapses. 32 Synapses were defined as locations where Homer1/gephyrin immunoreactivity within the GFP-tagged cytoplasm overlapped with the presynaptic VGLUT2/VGAT immunofluorescence as defined above (Figure 4). The specificity of the gephyrin antibody has been validated previously. 27 Homer1 antisera preadsorbed with Homer1 control peptide eliminated all labelling.
For glutamate, VGLUT2-Homer1 synapses were detected with increasing density along the primary dendrite to peak at the 31-45μm segment with an abrupt decline thereafter ( Figure 5). The average density of this segment in dioestrous mice is 0.08 ± 0.02 synapses μm -1 and 0.07 ± 0.02 synapses μm -1 in pro-oestrous mice.
Very few VGLUT2-Homer1 synapses were detected on the soma ( Figure 5C). No differences were detected between dioestrous  Table 3.

For GABA inputs, a low density of VGAT-gephyrin synapses was
detected along the dendrite with very few (0-0.01 synapses μm -1 ) on the soma ( Figure 6 and Table 3

| D ISCUSS I ON
In the present study, we report the density and spatial distribution of synaptic inputs to the mouse GnRH neuron cell body and proximal dendrite using ExM. This methodology has gained rapid popularity in neuroscience because of its ability to overcome the inadequate resolution afforded by conventional confocal microscopy to define a synapse. 22 Serial ultrastructural reconstructions have demonstrated that only approximately 10% of axons apposed to a dendritic spine actually form a synapse, raising caution when interpreting confocal apposition data. 33 Accordingly, we note that the densities of syn- body of GnRH neurons. 35 The more extensive analysis of the GnRH neuron dendrites afforded by dye-filling GnRH neurons in situ has similarly shown that the highest density of spines occurs with the first 50 µm of dendrite with values halving beyond this. 34 We have recently reported that the highest density of synaptic inputs to the GnRH neuron occurs at its distal-most projections, the dendrons, just before they pass into the median eminence. 10 This indicates that GnRH neurons have two sites of high density synaptic input; one in the proximal dendrite near the cell body and the other at the distal dendrons.
We find here that GABA and glutamate inputs within this den- suggest that the excitatory "hot-spot" on the dendrite is proximal to, rather than at, the action potential initiation site.
Although we cannot necessarily assume equal efficiencies of our vGLUT2 and VGAT immunohistochemistry, it appears that there are twice as many glutamatergic as GABAergic inputs to the GnRH neuron cell soma and proximal dendrite. This has not been detected previously. Although the frequency of GABA A receptor-mediated postsynaptic currents recorded from the GnRH neuron cell body is typically much higher than that of AMPA receptor-mediated events, 4,38 it is important to consider that these events reflect GnRH neurons. 21 Thus, there appears to be a curious and presently These results are very similar to those observed in ovariectomised, oestrogen-treated mice killed at the peak of the LH surge. 21 This suggests that neither the pro-oestrous, nor ovariectomised, oestrogen-evoked LH surge is associated with morphological changes in GABAergic or glutamatergic inputs to the GnRH neuron.
This is consistent with the electrophysiological findings and, taken together, this indicates a lack of any robust synaptic plasticity at the GnRH neuron cell body and proximal dendrite immediately prior to or during the pro-oestrous GnRH/LH surge in mice.
Although synapse density may not change, it nevertheless remains likely that presynaptic transmitter release and postsynaptic receptor dynamics and at the GnRH neuron are altered at the time of the surge. For example, the subunit composition of AMPA receptors expressed by GnRH neurons has been reported to change at the time of the LH surge in rats. 16 The primary driver of the abrupt increase in receptor expression had no effect on the LH surge mechanism 44 and a functional role for RP3V GABA neuron transmission in surge generation has not been identified. 11 At present, the ability to examine kisspeptin, GABAergic or glutamatergic transmission specifically at the GnRH neuron in vivo remains a significant technical challenge.
In summary, we have used ExM to assess the distribution of synapses on proximal elements of the GnRH neuron. We find that a "hot-spot" of excitatory GABAergic and glutamatergic innervation

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/jne.13021.