A role for glial fibrillary acidic protein (GFAP)-expressing cells in the regulation of gonadotropin-releasing hormone (GnRH) but not arcuate kisspeptin neuron output

GnRH neurons are the final central neural output regulating fertility. Kisspeptin neurons in the hypothalamic arcuate nucleus (KNDy neurons) are considered the main regulator of GnRH output. GnRH and KNDy neurons are surrounded by astrocytes, which can modulate neuronal activity and communicate over distances. Prostaglandin E2 (PGE2), synthesized primarily by astrocytes, increases GnRH neuron activity and downstream pituitary release of luteinizing hormone (LH). We hypothesized GFAP-expressing astrocytes play a role regulating GnRH and/or KNDy neuron activity and LH release. We used adenoassociated viruses to target designer receptor exclusively activated by designer drugs (DREADDs) to GFAP-expressing cells to activate Gq or Gi-mediated signaling. Activating Gq signaling in the preoptic area, near GnRH neurons, but not in the arcuate, increases LH release in vivo and GnRH firing in vitro via a mechanism in part dependent upon PGE2. These data suggest astrocytes can activate GnRH/LH release in a manner independent of KNDy neurons.

kisspeptin, neurokinin B and dynorphin. KNDy neurons in brain slices exhibit changes in 25 spontaneous action potential firing activity that is on the time-scale of episodic LH release and 26 that is steroid sensitive 13 . In vivo, optogenetic activation of KNDy neurons induces LH release, 27 and increases in intracellular calcium in these cells measured by bulk photometry are 28 associated with spontaneous LH release 10,15 . Whether or not KNDy neurons endogenously 29 generate the release of GnRH or are a relay station for a signal generated in other cell types is 30 not known, nor is the list of other cell types that may alter GnRH neuron firing independent of 31 KNDy neurons complete . 32 In this regard, evidence suggests astroglia play a role in reproductive function [16][17][18][19][20][21] . Astrocytes 33 are a major glial type in the central nervous system. GnRH and KNDy neurons are contacted by 34 astrocytes and astrocytic coverage of neurons in the POA and ARC varies with estrous cycle 35 stage, gonadal status, and seasonal reproduction [22][23][24][25][26] . Astrocytes can propagate information to 36 adjacent astrocytes [27][28][29] making them intriguing candidates for coordinating episodic hormone 37 release over the distances among scattered GnRH neurons. A consensus has yet to emerge 38 regarding vesicle-mediated transmission by astrocytes 30 , but there are data indicating these 39 cells may release multiple gliotransmitters [31][32][33] . Among these is PGE2, a non-vesicular 40 transmitter that can increase GnRH neuron activity, GnRH release and LH release 34-36 . 41 Our working hypothesis is that local astrocytes modulate GnRH and/or KNDy neuron activity to 42 alter LH release in vivo. To investigate the role of astrocyte signaling in the control of GnRH and 43 KNDy neurons in vitro and LH release in vivo, we took advantage of designer receptors 44 exclusively activated by designer drugs (DREADDs). DREADDs were genetically targeted to 45 astroglia by putting them under the control of the glial fibrillary acidic protein (GFAP) promoter 46 coupled with local adeno-associated virus (AAV) delivery. GFAP is mainly expressed by 47 astrocytes in the central nervous system 37 , thus this approaches allowed temporal and spatial 48 (n=3/group). CNO had no effect on the mean LH levels (AAV-mCherry, control 4.5±0. 7

Gq activation in GFAP-expressing cells of the POA increases GnRH neuron firing rate in vitro 129
Extracellular recordings of GFP-identified GnRH neurons were made in brain slices from mice 130 infected in the POA with either AAV-mCherry or AAV-Gq. Bath application of CNO (200nM) had 131 no effect on firing rate of GnRH neurons from AAV-mCherry control mice ( Figure 4A, D). In 132 contrast, CNO increased firing rate of GnRH neurons within the infected area of AAV-Gq mice 133 as identified by mCherry signal ( Figure 4B, D, two-way repeated-measures ANOVA mCherry vs 134 Gq-mCherry F(2, 26)= 4.013, P=0.0303). Interestingly, GnRH neurons located outside of the 135 infected region in AAV-Gq mice, easily distinguishable because they were not surrounded by 136 mCherry signal, did not respond to CNO treatment ( Figure 4C). This suggests that activation of 137 Gq signaling in GFAP-expressing POA cells can increase the firing rate of GnRH neurons but 138 that the propagation of that signal to GnRH neurons in uninfected areas is limited. 139 To test if activating Gq signaling in GFAP-expressing cells alters firing rate of Tac-2 GFP  140 neurons, the above studies were repeated after targeting injection to the ARC. In this brain 141 region, infection of cells with neuronal morphology within experimental slices was noted. Some 142 of these neurons expressed GFP indicating they are Tac2-expressing neurons, which are 143 known to activate one another 44 . While these mCherry-expressing neurons were not recorded, 144 they could influence the response within the slice; data from slices with infected neurons were 145 excluded from statistical analyses but are shown in the individual data plots ( Figure 4E,F). 146 Neither the type of virus injected (AAV-mCherry vs AAV-Gq) nor the location of the cell inside or 147 outside the hit affected the firing rate of Tac2-GFP neurons in response to bath application of 148 200nM CNO ( Figure 4E-H, Table 4). The firing rate of Tac2-GFP neurons tends to increase over 149 time of recording 45 ; this was observed as an increase in firing rate during the wash vs control 150 period. 151

CNO-induced increase in GnRH neurons depend at least in part on activation of PGE2 152
receptors The putative gliotransmitter PGE2 is primarily produced by astrocytes in the 153 hypothalamus and can increase GnRH neuron firing rate 42,46 by acting on EP1 and EP2 154 receptors expressed by these neurons 34,47 . We hypothesized that the CNO-induced increase in 155 firing in AAV5-Gq mice is PGE2-dependent. We first examined the effects of the stable PGE2 156 analogue dimethyl-PGE2 (dmPGE2) on firing rate of GnRH and Tac2 neurons. In GnRH-GFP 157 neurons, pretreatment with either DMSO vehicle (0.3%) or a mix of EP1-and EP2-specific 158 antagonists (100M SC19220 and 20M PF04418948, respectively) had no effect on the firing 159 rate ( Figure 5A-C). Consistent with previous studies using PGE2 46 , dmPGE2 increased GnRH 160 neuron firing rate in cells pretreated with vehicle for the antagonists ( Figure 5A,C, n=9, 161 P<0.0001 two-way repeated-measures ANOVA/Bonferroni, Table 5). Pretreatment with 162 EP1/EP2 receptor antagonists blocked the effect of dmPGE2 on GnRH neuron firing rate 163 ( Figure 5B,C, n=10,   Figure 6A). 185 Despite the ability of CNO to increase intracellular calcium in infected cells, CNO had no effect 186 on firing rate of GnRH neurons in brain slices from female mice infected with either virus ( Figure  187 6B,C,E), unless infected cells with neuronal morphology were observed within the slice ( Figure  188 6D,E). These data suggest there may be a sex difference in the regulation of GnRH neurons by 189 GFAP-expressing cells. An important caveat to point out is that these recordings were made in 190 the afternoon (slices made 2:30-3:30pm, recordings 3:30-9:30pm), whereas recordings from 191 males were made in the morning (slices 9:30am-12pm, recordings 10:30am-6pm). This caveat 192 is mitigated by the consistent virus x treatment interaction for producing elevations in 193 intracellular calcium levels in both sexes. This suggests the lack of GnRH neuron response to 194 CNO in females is mediated subsequent to the GFAP-expressing cell. 195

Discussion 196
The central regulation of fertility depends on the secretion of appropriate patterns of GnRH. An 197 emerging dogma in the field is that this is substantially regulated by arcuate kisspeptin, also 198 known as KNDy, neurons. Here we present evidence that activation of Gq signaling in 199 astrocytes near GnRH neurons in males triggers increased LH release and increased GnRH 200 neuron firing. This is independent of KNDy neurons as similar activation of Gq signaling in 201 astrocytes near these cells fails to alter either LH release or their firing rate. Further, this effect 202 appears to be sexually differentiated as GnRH neurons from diestrous females did not respond. 203 This astrocyte signaling may be a key element in the modulation of GnRH neuron firing and LH 204 release in male mice. 205 The present findings support and extend previous work suggesting an involvement of astroglia 206 in the regulation of reproductive neuroendocrine function 21,48,49 . Anatomically, a substantial 207 proportion of GnRH neuron somatic membrane is contacted by glia and this contact varies with 208 reproductive state 22,50 . GnRH terminals are intimately connected with specialized tanycytes in 209 the median eminence; this interaction is dependent upon hormonal milieu in females 51,52 . 210 Functional interactions among GnRH neurons and glia have been postulated to be primarily 211 mediated by PGE2, as blocking central prostaglandin synthesis reduces gonadotropin release 212 53 , whereas injection of PGE2 into the third ventricle or implantation into the preoptic area 213 enhances LH release 36,54 . More recently, PGE2 was shown to increase firing rate of GnRH 214 neurons in both sexes 18 . By activating Gq signaling within GFAP-expressing cells directly, the 215 present work extends these findings to the in vivo situation in which the entire hypothalamo-216 pituitary gonadal axis can interact, and indicates a role for these cells within the preoptic area in 217 ultimately increasing LH release in males. In brain slices from these mice, CNO increased 218 intracellular calcium with a similar timecourse to that observed in other brain regions when 219 astrocytes were activated with either DREADDs or by agonists native GPCRs 55,56 . In the present 220 work, the increase in astrocyte intracellular calcium occurred with a shorter lag and had a 221 shorter duration than the changes in neuronal firing rate. This suggests that the GnRH neuron 222 response is dependent upon the elevation of calcium in astrocytes, and also that the signal 223 conveyed to the GnRH neuron produces a prolonged response consistent with activation of a G-224 protein-coupled receptor within the neurons. This postulate was confirmed as the CNO-induced 225 increase in firing rate response was mimicked by treatment with a stable form of PGE2 and was 226 largely dependent upon activation of EP1 and EP2 receptors. 227 Astroglia perform several functions within the central nervous system, including release of 228 multiple substances that can serve as gliotransmitters. Astrocytes also are able to communicate 229 over broad areas, propagating elevations in intracellular calcium 29 . This latter observation was 230 key to sparking our interest in how astroglia may function in GnRH release, in particular we 231 postulated that the propagation of signals to distal astrocytes may serve as a mechanism for 232 coordinating activity among the soma of GnRH neurons, which are spread over a wide area 233 from the diagonal band of Broca through the medial basal hypothalamus 57 . It was with this 234 postulate in mind that we recorded from GnRH neurons outside the hit region as defined by 235 mCherry fluorescence. These neurons outside the area of infection did not respond to bath 236 application of CNO with an increase in firing rate, however, suggesting that astroglia do not 237 effectively propagate a signal capable of activating distal GnRH neurons in brain slices. 238 Interestingly, the activation of Gq signaling in GFAP-expressing cells within the arcuate nucleus 239 was without effect on either LH release or firing rate of KNDy neurons, indicating that the effects 240 on GnRH neurons were independent of this cell type that appears to play a key role in other 241 aspects of GnRH neuron activation. Consistent with a role for PGE2 in the putative astroglia-242 GnRH communication, treatment of KNDy neurons with PGE2 had no effect upon their firing 243 rate. The lack of effect of astroglia Gq-signaling on KNDy neurons is not attributable to a 244 technical failure as CNO treatment increased intracellular calcium levels in GFAP-expressing 245 cells in this region. Interestingly, the area infected by virus includes tissue through which the 246 processes of GnRH neurons pass en route to the median eminence. This suggests that this 247 region of the GnRH neuron axon may not be sensitive to mediators released by astroglia. Of 248 interest in this regard, treatment of explants containing the median eminence along with the 249 medial basal hypothalamic region with PGE2 induced GnRH release 58 . Our arcuate hits 250 typically did not extend into the median eminence itself. As was the case with GnRH neurons 251 outside the hit, it is possible there is not sufficient propagation of astroglial signals effective in 252 activating GnRH neuron terminals outside the local region of infection. It is also possible that 253 astrocytes exhibit regional variation in the gliotransmitters they produce. Of note, however, 254 activation of Gq signaling in these cells would be postulated to induce release of whatever 255 substances are downstream of that signaling pathway in the cells and not be limited to 256 prostaglandin synthesis. Together these observations suggest the effects of Gq signaling within 257 astroglia are at least in part specialized to preoptic regions of GnRH neurons for the 258 reproductive neuroendocrine system. 259 We conducted limited studies of activating Gq signaling in astroglia in diestrous females. CNO 260 treatment effectively induced increases in intracellular calcium, but failed to alter GnRH neuron 261 firing rate unless there were overtly infected neurons within the slice. There are several possible 262 explanations for these observations. First cycle stage may affect the response, although the 263 elevation of intracellular calcium in the present study suggests that astroglia were activated in 264 the diestrous mice studied. Second different gliotransmitters may be released upon astroglia 265 activation in females, perhaps also dependent upon cycle stage. In prior work, over two-thirds of 266 GnRH neurons increased firing rate in response to PGE2 regardless of sex or of cycle stage in 267 females 18 , indicating responsiveness to this mediator is consistent among groups. Third, it is 268 possible that the relevant astroglia regulating GnRH release are located in different brain 269 regions, for example more caudal within the anteroventral periventricular region known to be 270 important for induction of the female-specific preovulatory GnRH and LH surge 59 . 271 In contrast to the robust serum LH and GnRH neuron firing response to activation of Gq 272 signaling in astroglia, activation of Gi signaling had no effect under the conditions studied. This 273 included no effect on intracellular calcium levels in POA astroglia in either intact or castrated 274 mice, and no effect on LH in castrated mice. The effects of activating Gi signaling in GFAP-275 expressing cells varies with brain region. In the hippocampus, increases in intracellular calcium 276 were reported 60 , whereas in the hypothalamic arcuate region, activation of Gq and Gi signaling 277 have opposing effects on food intake 41  In summary, the present work defines that activation of Gq signaling in astrocytes can increase 298 GnRH neuron activity and LH release in a sex-dependent manner. Activation of this pathway 299 may activate this system independent of arcuate kisspeptin neurons or may sculpt the response 300 to that or other critical inputs to GnRH neurons to affect ultimately reproduction. 301

Materials and Methods 302
All reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless noted. 303 Animals Mouse strains used for this work are summarized in Table 1 Tail tip blood collection To examine the effect of DREADD activation on LH levels, mice were 357 handled daily for 2wks before CNO administration studies or 5wks before sampling LH pulses 358 and habituated to IP injection of 0.9% saline for the last 3-4d. The tip of the tail was nicked and 359 6L of blood was collected and mixed immediately with 54µl of assay buffer (PBS, 0.05%Tween 360 and 0.2% BSA). Sampling regimen and animal models were selected based on postulated LH 361 response to activating Gq or Gi signaling within GFAP-expressing cells. To test the effects of 362 activating Gq, postulated to be activating, gonad-intact males were sampled every 10min for 2h, 363 with IP injection of 0.9% saline vehicle at 30min, and IP injection of 0.3mg/kg CNO at 60min. To 364 test the effect of activating Gi, postulated to be inhibitory, males were castrated to elevate LH 365 release; one week after castration, these mice were sampled every 6min for 3h, with IP injection 366 of 0.3mg/kg CNO at 90min. 367 LH assay Tail blood diluted with assay buffer was kept on ice until the end of sampling then 368 stored at -20°C until LH assay by the University of Virginia Ligand Assay and Analysis Core 48 . Denmark; D048701-2). Mouse LH reference prep (AFP5306A; NHPP) is used as the assay 374 standard. The limit of quantitation (functional sensitivity) was 0.016 ng/ml, defined as the lowest 375 concentration that demonstrates accuracy within 20% of expected values. Coefficient of 376 variation (%CV) was determined from serial dilutions of a defined sample pool. Intraassay CV 377 was 2.2%; interassay CVs were 7.3% (low QC, 0.13 ng/mL), 5.0% (medium QC, 0.8 ng/mL) and 378 6.5% (high QC, 2.3 ng/mL). 379 Brain Slice Preparation All solutions were bubbled with 95% O2/5% CO2 for at least 30 min 380 before exposure to tissue and throughout the experiments. At least one week after sampling for 381 LH, brain slices were prepared through the hypothalamus as described 69  Recordings were made in voltage clamp with a 0mV pipette holding potential, data acquired at 399 10kHz and filtered at 5kHz using a one amplifier of an EPC10 dual patch clamp amplifier 400 controlled with PatchMaster software (HEKA Elektronic, Lambrecht, Germany). 401 Slices were transferred to a recording chamber with constant perfusion of carboxygenated 402 ACSF at 29-32°C at a rate of approximately 3ml/min. GFP-positive cells were targeted for 403 recording in the POA (GnRH neurons) or ARC (KNDy neurons). Cells that were surrounded by 404 mCherry signal were considered within the infected area (hit) and those with no detectable 405 peripheral mCherry signal within 150µm were considered outside the infected area (miss). 406 Recordings consisted of a 5-10min stabilization period, a 5-min control period, bath-application 407 of treatment for six minutes, followed by a wash. Mean firing rate was calculated for the last 3-408 min baseline period, for min 5-7 of treatments, and min 8-10 after treatment (wash). At the end 409 of each recording, cells that were inactive throughout the recording were treated with 20 mM 410 potassium in ACSF; cells that exhibited action currents in response were verified to be alive and 411 recordable, and all data were used. Cells that did not respond to elevated potassium were 412 excluded. No more than three cells from the same animal were included and at least four 413 animals were studied per group for electrophysiology studies. 414 To test the hypothesis PGE2 affects GnRH and KNDy neuron firing, we used 16,16-dimethyl 415 prostaglandin E2 (dmPGE2, 200nM), a stable form of PGE2 (Cayman Chemical 14750) in 416 0.00076% methyl acetate vehicle. Antagonists for prostaglandin receptor EP1 (SC-19220, 417 100M) and EP2 (PF-04418948, 20M) were diluted into fresh bubbled ACSF (final DMSO 418 concentration 0.3%) and bath applied before CNO treatment. The recording paradigm was a 6-419 min baseline, 6-min pretreatment with methyl acetate vehicle containing either DMSO vehicle or 420 EP1 and EP2 antagonists, a 6-min treatment that added 200nM dmPGE2, followed by wash in 421 ACSF. Mean firing rate was calculated for the last 3-min baseline and pretreatment periods, 422 minutes 5-7 following addition of dmPGE2 for treatment, and min 8-10 after wash began. 423 Calcium imaging Calcium imaging was performed on brain slices from mice co-injected with 424 AAV5-gfaABC1D-cyto-GCaMP6f and either AAV-mCherry control, the AAV-Gq or the AAV-Gi 425 (ratio 1:1). Calcium was imaged for a 50ms exposure time at 0.5s intervals for 20min with an 426 Hammamatsu ORCA-FLASH4.0 V3 CMOS camera. Imaging began after a 5-min stabilization 427 period (not included in the analysis), and consisted of a 5-min control period, a 6-min treatment 428 with 200nM CNO, and 9-min wash. Regions of interest (ROIs) were isolated from astrocytic 429 domains for which signal fluctuations were identified prior to CNO treatment. Each ROI likely 430 represents the soma or proximal regions of a single astrocyte. For each slice, a background ROI 431 was isolated from a region in which no signal fluctuations were detected; a single region was 432 used as the background reference for all the ROIs of a slice. ΔF/F0 [(F(t) -F0(t)) / F0(t)] was 433 calculated using custom routines in Igor Pro (Wavemetrics) and plotted as a function of time for 434 each ROI. Briefly, the background ROI was subtracted from all other ROIs from that slice. To 435 account for the signal decay related to photobleaching, we used a double exponential curve fit 436 for each individual ROI that was considered F0; the CNO treatment period was excluded from 437 the curve fit using the masking capability in the Igor Pro curve fitting function. We used the 438 resulting curve fit as an estimate of F0 as a function of time. Each curve fit was verified by eye 439 to conform to the timecourse of the bleaching for each ROI. F0 was subtracted from the 440 background-subtracted traces to calculate ∆F. ∆F was divided by the fit to obtain ΔF/F0. To 441 quantify the effect of CNO on the astrocytic calcium signals, the area under the curve (AUC) 442 from each ΔF/F0 curve was calculated for different periods (control, the last 3-min baseline 443 period; CNO from the first 3min of CNO exposure; wash AUC from min 5-7 after CNO treatment 444 ended. The AUCs obtained from 27 independent ROIs were average to obtain the average 445 AUCs of the slice (n=1). Three to six slices from 1-2 mice were used for each group (Table 3). 446 Analysis and statistics Data were analyzed using code written in IgorPro 8 (Wavemetrics). 447 Action currents (events) were detected and confirmed by eye. Statistical analyses were 448 conducted with Graphpad Prism 9. Distribution normality was checked with Shapiro-Wilk; most 449 experiments included one or more groups that were not normally distributed.      Table 3.    Supplementary Figure S1. GFAP and S100 signals are co-expressed in the majority of cells (n=3 mice, 5 fields/mouse).