Medroxyprogesterone acetate positively modulates specific GABAA-receptor subtypes - affecting memory and cognition

Medroxyprogesterone acetate (MPA) is a progestin widely used in humans as hormone replacement therapy and at other indications. Many progestin metabolites, as the progesterone metabolite allopregnanolone, have GABAA-receptor modulatory effects and are known to affect memory, learning, appetite, and mood. In women, 4 years chronic treatment with MPA doubles the frequency of dementia and in rats, MPA causes cognitive impairment related to the GABAergic system. Activation of the membrane bound GABAA receptor results in a chloride ion flux that can be studied by whole-cell patch-clamp electrophysiological recordings. The purpose of this study was to clarify the modulatory effects of MPA and specific MPA metabolites, with structures like known GABAA-receptor modulators, on different GABAA-receptor subtypes. An additional aim was to verify the results as steroid effects on GABA response in single cells taken from rat hypothalamus. HEK-293 cell-lines permanently expressing the recombinant human GABAA-receptor subtype α1β2γ2L or α5β3γ2L or α2β3γ2S were created. The MPA metabolites 3α5α-MPA,3β5α-MPA and 3β5β-MPA were synthesised and purified for electrophysiological patch-clamp measurements with a Dynaflow system. The effects of MPA and tetrahydrodeoxycorticosterone were also studied. None of the studied MPA metabolites affected the responses mediated by α1β2γ2L or α5β3γ2L GABAA receptors. Contrary, MPA clearly acted both as a positive modulator and as a direct activator of the α5β3γ2L and α2β3γ2S GABAA receptors. However, in concentrations up to 10 μM, MPA was inactive at the α1β2γ2L GABAA receptor. In the patch-clamp recordings from dissociated cells of the preoptic area in rats, MPA increased the amplitude of responses to GABA. In addition, MPA alone without added GABA, evoked a current response. In conclusion, MPA acts as a positive modulator of specific GABAA receptor subtypes expressed in HEK cells and at native GABA receptors in single cells from the hypothalamic preoptic area.


Introduction
Medroxyprogesterone acetate (MPA) is widely used by women, both as contraceptive, and as a progestin during hormone therapy. In the year 2004, about 30 million women in the United States were prescribed medications including MPA.
MPA was used in the Women´s Health Initiative Memory Study where postmenopausal women received long-term continuous hormone treatment (5 -7 years), either with only conjugated equine estrogens (CEE), in women without a uterus (Shumaker et al., 2004), or with CEE plus MPA (Shumaker et al., 2003). The study was disrupted because of complications, among them the group treated with CEE plus MPA showed an increased risk for development of dementia, mostly Alzheimer's disease (AD) (Shumaker et al., 2003). This increased risk was not shown in the group receiving CEE alone (Shumaker et al., 2004) and the increased dementia frequency was not due to stroke or other cardio-vascular events (Coker et al., 2010). The increased dementia risk in the MPA treated group was interpreted as due to a biological effect of MPA in the brain of the treated patients (Coker et al., 2009). In animal studies, worse performance in learning and memory has been shown after long-term treatment (6 month) with estradiol plus MPA, or after treatment with only MPA (Braden et al., 2010(Braden et al., , 2011Lowry et al., 2010).
MPA has in addition to the effect related to cognition, also an orexigenic effect and increases the food intake in animals and humans (Steward et al., 2016;Le et al., 2009). MPA has even been used as treatment for cachexia (Simons et al., 1998). We have earlier shown that the neuroactive steroid allopregnanolone has a pronounced effect in hypothalamic cells and it is therefore of interested to use the same brain region for confirming MPA experiment as the region is very sensitive to allopregnanolone with significant effects already at 2 nM (Löfgren et al., 2019). This effect is at least partly mediated by the GABA A receptor subtype α2 in the hypothalamus and is involved in stimulating hyperphagia. (Holmberg et al., 2018;Bauer et al., 2012).
MPA can induce anesthesia in rats, an effect also seen with progesterone, which has a chemical structure like MPA ( Fig. 1) (Meyerson, 1967;Bixo and Bäckström, 1990). The anesthetic effect of progesterone has been attributed to the strong positive GABA A -receptor modulation by allopregnanolone, a 3α-hydroxy-A ring 5α reduced progesterone metabolite. Allopregnanolone is active at the GABA A receptor, the major inhibitory receptor within the brain, and anesthesia is easily induced by allopregnanolone (Majewska et al., 1986;Norberg et al., 1987). Continuously raised allopregnanolone concentration, as during chronic stress or at frequent repeated injections, enhances the development of AD-like symptoms (decreased memory and learning) in transgenic AD-mice, already after one month's treatment (Bengtsson et al., 2012(Bengtsson et al., , 2013. With 5 months exposure to allopregnanolone, even wild type mice showed a permanent memory impairment and a reduction of the hippocampal volume (Bengtsson et al., 2016).
MPA and progesterone are both metabolized at the A-ring, and several metabolites of MPA have been detected in human plasma and bile, among these are A-ring reduced tetrahydro-MPA variants. Steroids as progesterone with a delta4-3keto configuration are first metabolized by a 5α-reduction catalyzed by 5α-reductase (5αR). Thereafter, a 3αhydroxylation by 3α-hydroxysteroid dehydrogenase (3αHSD) may occur which modifies the steroids to contain a hydroxy group at 3α position. Steroid metabolites with this configuration can act as positive modulators of GABA A receptors (Melcangi et al., 1999;Fukushima et al., 1979;Zhang et al., 2008;Chen et al., 2009). Therefore, the 3α-hydroxy metabolites of MPA were of interest to investigate for effects on the GABA A receptor subtypes.
The 3αΟН A-ring structure in endogenous steroids is important for binding to the GABA A receptor. Allopregnanolone and tetrahydrodeoxycorticosterone (THDOC) are both 3α5α− reduced-pregnane-steroid metabolites thought to be among the most potent positive GABA A -receptor modulators (Purdy et al., 1990(Purdy et al., , 1992. THDOC and allopregnanolone bind to the same binding site on the GABA A receptor (Hosie et al., 2006) and have similar efficacy and potency at the receptor (Stromberg et al., 2005).
The aim of this study is to clarify if tetrahydro-MPA metabolites or MPA itself affect the currents mediated by different GABA A receptor subtypes.

Synthesis of MPA metabolites
The syntheses of the medroxyprogesterone acetate metabolites were made at Umecrine AB as described below. 300 mg (0.78 mmol) of MPA was dissolved in 150 ml ethanol (95%, Kemetyl), 0.45 ml of glacial acetic acid (Merck, Sigma-Aldrich), and 5% Rhodium on carbon catalyst (5%, Alfa Aesar) were added. A stream of hydrogen gas from a separate flask with sodium borohydride (Fisher) in methanol (pro analysis (p.a.), Merck, Sigma-Aldrich) was bubbling overnight in the MPA solution under stirring. Thin-layer chromatography (TLC) (detection with UV light and/or vanillin stain) monitored the reaction mixture. At completion, the mixture was filtered, the solvent evaporated in vacuo, and the white residue extracted with dichloromethane (Sigma) and water. The organic phase was washed with aqueous sodium bicarbonate (Merck, Sigma-Aldrich) yielding 290 mg of a white solid material.
Thus, MPA was hydrogenated with the use of Rhodium-catalyzed reactions, leading to the formation of four isomers of MPA. The synthesized compounds were purified by use of preparative high pressure liquid chromatography (HPLC), Waters: a 1515 Isocratic Pump, 717 plus Auto-sampler equipped with a 1500 Column Heater, a Symmetry-Prep C18 7 µm 78 × 300 mm column, 2487 Dual λ Absorbance Detector, and Fraction Collector II. The detector output was recorded on a PCbased Waters Breeze Chromatography Software (version 3.20). The conditions used were T (column) 45 • C; detection at 206 nm; flow rate:  Spectra of the crude reaction mixture showed the following proportions among reduced products: 3α5α-MPA ː 3β5α-MPAː3β5β-MPAː3α5β-MPA, 23:23:45:9.
GABA (Sigma Chemical Co.) was dissolved in EC-solution, while THDOC (Sigma Chemical Co.), MPA and MPA-metabolites were dissolved in DMSO (Sigma Chemical Co., 6 mM stock concentration). During all electrophysiological measurements, the DMSO concentration was 0.1%. Whole-cell voltage-clamp recordings were made with the Dynaflow® system (Dynaflow Pro II Platform Zeiss Axiovert 25) and the DF-16 Pro II chip (Cellectricon AB, Göteborg, Sweden) that allows rapid solution exchanges. The patch-clamp instruments used were an Axonpatch 200B amplifier, a Digidata 1322 A digitizer and pCLAMP software (Axon Instruments, Foster City, CA, USA).

Electrophysiological conditions
Whole-cell currents from Human embryonic kidney (HEK-293) cells permanently transfected to constitutively express recombinant human GABA A receptors (see below) were recorded with the whole-cell patch technique at room temperature (21-23 • C). The glass pipettes used had a resistance of 2-6 MΩ when filled with IC solution. After compensation for liquid-junction potential, a steady holding potential of − 17 mV was used in all experiments. The cells were added to the chip and kept in EC solution. EC solution with or without steroids and GABA were applied by the Dynaflow® system. The syringe pump flow rate was 26 µl/min. The standard study protocol included cell exposure to THDOC, MPA or MPA metabolites for 20 s before GABA was applied. Thereafter followed a washout period of 2 min in EC solution to completely remove studied steroids before the next application. The human GABA A receptor subtypes studied are endogenously found in different compartments of the neurons. Thus, the α5β3γ2 L GABA A receptor is typically found outside the synapse, i.e. extra-synaptic (Belelli and Lambert, 2005). Such receptors are activated by lower GABA concentrations during a longer period than receptors located in the synapse. The α1β2γ2 L and α2β3γ2 S are the most common synaptic GABA A receptors. They are exposed to high GABA concentrations (> 100 µM) for short periods (ms). Thus, to resemble natural conditions, these receptors were exposed to 30 and 45 μM GABA (EC 75 ). GABA was applied together with the studied steroid for 40 ms. Studies of the α5β3γ2 L receptor was with 0.3 μM GABA together with the studied steroid for 6 s. The responses to 1.0 µM THDOC on HEK-293 α5β3γ2 L and α2β3γ2 S are shown as reference.

HEK-293 cells expressing recombinant human GABA A receptors
The effect of THDOC, MPA or MPA metabolites was measured as the area under the curve (AUC) of evoked currents (corresponding to charge transfer). AUC was analysed to include both potential effects on the peak amplitude and on the desensitization and deactivation phases. The AUC was measured semi-manually by using cursors and the curve-fitting procedures of the pCLAMP software. Direct activation of the GABA A receptor by THDOC or MPA (in the absence of applied GABA) was measured as change in steady-state current. Steroid effects on AUC were normalized to control in order to avoid effects of inter-and intracellular (e.g. due to slow drift) variation in measured parameters.
HEK-293 cells were permanently transfected with cDNA to constitutively express the human α1β2γ2L, α5β3γ2L or α2β3γ2S GABA A receptor subtypes, respectively. The GABA A receptor subunits α1 ) with introduced Kozac sequences just before the start codons were sub-cloned into the mammalian expression vectors (Invitrogen); pcDNA3.1(+) for α1 and α5, pcDNA3.1(-) for α2, pcDNA3.1/ Hygro(-) for β2 and β3, and pcDNA3.1/Zeo(+) for γ2S and γ2L. A HEK-293 cell line stably expressing the three GABA A -receptor subunits was produced by transfection of the subunits one at a time. The transfections were followed by selection with the appropriate antibiotics, cell separation with the use of subunit-specific antibodies, and production of single cell colonies. Produced cell lines were analyzed with immunocytochemistry for the GABA A receptor subunits, followed by selection of a suitable cell line showing good reactivity towards GABA and THDOC in electrophysiological recordings. For electrophysiological experiments, HEK-cell pellets were resuspended in 2-3 ml EC-solution and kept at 37 • C for 15-30 min. The cells were then resuspended and added to the chip bath.

Animals
Male Sprague-Dawley rats (n = 19, weight < 130 g), bred at Umeå Centre for Comparative Biology (UCCB) Umeå University, Umeå, Sweden were used. They were housed two or four rats per cage. All housing was at constant temperature, 22 • C, and with artificial daylight. Rats had free access to water and standard food. The regional ethics committee for animal research ("Umeå djurförsöksetiska nämnd"; approval no. A44-2018) approved the experimental protocols. All efforts were made to minimize the number of animals used and their suffering, and all procedures including animals were conducted in accordance with the ethical guidelines set by the European Union.

Drugs and chemicals
Medroxyprogesterone (MPA), GABA, Picrotoxin and amphotericin B were obtained from Sigma Chemical Co. (St. Louis, MO, USA). Other chemicals used were purchased from a local supplier.
Electrophysiological experiments: The methods used for electrophysiological recording and cell preparation have been described in detail elsewhere (Haage and Johansson, 1999). A summary of the procedures is given below.

Cell preparation
The rats were killed by decapitation without anesthesia. The brain was quickly removed and placed in ice-cold incubation solution (see above). A block/part of brain tissue from the rostral part containing the hypothalamus was dissected. Brain slices (thickness 250 µm) from the hypothalamic region were prepared and incubated for at least 1 h in oxygenated incubation solutions at 28 • C. Single cells with adhering synaptic nerve terminals from the medial preoptic area were isolated by vibrodissociation (Vorobjev, 1991) using a thin glass rod (~0.5 mm in diameter). The cell bodies of obtained neurons were round / oval, some with parts of dendrites attached.

Electrophysiological recordings
The amphotericin-B-perforated patch-clamp technique (Strömberg et al., 2009) was used to record whole-cell currents from the postsynaptic neurons under voltage-clamp conditions. In the perforated patch technique, the internal cellular components of the cell are preserved while still allowing control over internal Na + , K + , and Clconcentrations (Strömberg et al., 2009). The steady voltages used for recording from preoptic neurons were (after correction for liquid-junction potential) − 51 mV and − 11 mV, the former implying a voltage close to the resting potential and in this respect physiological, while the latter voltage was used to improve the signal-to-noise ratio as a consequence of the larger driving force for the GABA-evoked Clcurrents. (The equilibrium potential for Clin these recordings was − 100 mV.) Haage and Johansson (1999) describe the techniques used in detail. Cells were bathed in an EC solution and the patch pipettes were filled with intracellular solution (IC). EC-solutions, with or without test substances, were applied by a gravity-fed fast perfusion system controlled by solenoid valves. All patch-clamp experiments were performed at room temperature (21 -23 C). Borosilicate glass pipettes (GC150, Clark Electromedical Instruments, Pangbourne, UK), with a resistance of 3-4 MΩ when filled with intracellular solution and immersed in extracellular solution, were used.

Protocol and analysis
A standard protocol was used for the experiments. Firstly, the current was recorded in EC solution for 30 -60 s, followed by a 60 -300 s test period with or without GABA alone, MPA alone, GABA + MPA or GABA + MPA + picrotoxin. The aim in this study was to use a GABA concentration as in natural situation. In vivo, the intra-synaptic receptors are exposed to a much higher GABA concentration than the extra-synaptic receptors. GABA A receptors α1 and α2 are considered intra-synaptic and are exposed to high GABA concentrations while α5 receptor is considered extra-synaptic and is thus exposed to much lower GABA concentrations. In addition, we used GABAs EC 75 of the receptor subtype to be able to detect both positive and negative modulation by MPA. We have therefore used different GABA concentrations depending on the receptor subtype.
To allow for near steady-state conditions, the first 30 s interval of the test period was not used for analysis. The test period was followed by a 2 -4 min washout period in EC solution at steady rate during recording. The protocol was applied repeatedly for recordings from individual cells. The time-course and amplitude of the evoked currents were measured semi-manually by using cursors and were fitted to an exponential decay curve by the curve-fitting routine of the pCLAMP software (versions 7 -9; Axon Instruments, Foster City, CA, USA). The peak amplitudes of the currents were analyzed for events of amplitudes ~1.5 times the peak-to-peak noise (in the solution with the largest noise and with lowpass Bessel filtering at 2.0 kHz). To avoid intra-and intercellular variations in the time-course and peak amplitude, all events recorded in the presence or absence of MPA in the individual cells were indicated relative to a control. Finally, all the relative values from the recorded cells were pooled and subjected to statistical analysis, Origin and SPSS (IBM statistical package for the social sciences) software were used. To quantify the effect of MPA, measures of time constants, amplitudes, and area under the curve from each cell were expressed relative to the mean values under control conditions.

Statistical analysis
In experiment 1, SPSS was used for statistical analyses and GraphPad Prism, (GraphPad Software Inc., San Diego, CA, USA) was used for making illustrations. Area under the curve (AUC), peak amplitude and decay time constant (tau) were measured.
All values obtained from recordings in a certain condition were

MPA metabolites evaluated with different GABA A receptor subtypes
MPA metabolites with A-ring structures identical to those in compounds known to affect the GABA A receptor, i.e. a 3α-or 3β-hydroxyl group in the 3rd position, were tested with patch-clamp recordings for effect at the human GABA A receptor subtypes α1β2γ2L and α5β3γ2L. § No statistical testing made, due to few observations. NS = not significant. None of the 3α-hydroxy-MPA metabolites applied in a concentration of 0.1 -3.0 µM showed any significant modulation of the GABA response at α1β2γ2L or at α5β3γ2L GABA A receptors (Table 1). A high concentration (20 µM) of 3β-hydroxy-MPA metabolites, tested in four α1β2γ2Lexpressing HEK-cells, reduced the GABA-evoked current in all cells tested (Table 1), but due to the low number of cells tested with this concentration, no statistical analysis was made.

MPA effects on α5β3γ2L and α2β3γ2S GABA A receptors
When MPA was tested together with GABA at the α5β3γ2 L GABA A receptor, the results clearly show that MPA is a positive modulator at this receptor subtype. As shown in Fig. 3

Effects by MPA on GABA-evoked currents in dissociated neurons from the rat preoptic area
The effect of MPA on GABA-evoked currents was tested at a steady holding potential of − 51 mV (12 cells) and − 11 mV (7 cells). Compensation for liquid junction potential was maintained. At both holding potentials, MPA increased the amplitude of GABA-evoked current (− 51 mV: 343% of control, p = 0.011, Friedman test, Fig. 4, top; − 11 mV: 229% of control, p = 0.018, Wilcoxon test, Fig. 4, middle). The effect of MPA was reversible (Fig. 4, top). In the ad hoc test of recordings at − 51 mV holding potential, co-application of MPA (10 µM) and GABA (100 µM) resulted in higher response current amplitude compared to that at GABA alone before (p = 0.008) and GABA alone after (p = 0.034) co-application of GABA and MPA.

Effect by MPA on baseline current suggesting direct activation
At − 51 mV holding potential, MPA alone did not evoke any significant baseline shift (presumed "direct activation" of the GABA A receptor). However, at − 11 mV, where the driving force for Clcurrents through GABA A receptors is larger, MPA alone did induce a baseline shift in 6 out of 13 cells (46% of tested cells) (Fig. 4, bottom shows traces in a single cell).

Effects of MPA on decay of GABA-evoked currents in dissociated cells from rat preoptic area
MPA did neither significantly affect the desensitization (in the presence of GABA) nor the deactivation (after the end of GABA application) time constants. At a steady holding potential of − 11 mV, the desensitization time constant in GABA alone was 2599 ± 427 ms (mean ± SE) and in GABA + MPA it was 3457 ± 1646 ms (n = 5). At − 51 mV steady holding potential, the corresponding values were 5054 ± 708 ms, n = 7, for GABA alone and 3602 ± 773 ms, n = 7, for GABA + MPA (not significantly different, Wilcoxon's signed rank test).

Effects of picrotoxin on MPA-induced current changes
We tested the effect of picrotoxin (PTX, a blocker of the chloride channel in the GABA A receptor) to confirm that MPA was indeed acting at the GABA A receptor in preoptic neurons. In all eight cells tested, PTX (200 µM) blocked 94.5% of the current response to co-application of GABA (100 µM) and MPA (10 µM) (Fig. 5), suggesting that the additional current evoked by adding MPA to GABA is indeed mediated by GABA A receptors.

Discussion
In the results above, we have four new findings: 1). MPA directly, in the absence of GABA, evokes a current in cells expressing α5β3γ2L and α2β3γ2S GABA A -receptor subtypes but not in those expressing the α1β2γ2L subtype, 2). MPA is a positive GABA A -receptor modulator. 3).
MPA is a selective positive GABA A -receptor modulator, as it does not affect the α1β2γ2L GABA A receptor subtype. Thus, the MPA effect on the GABA A receptor depends on the α subunit, and 4). Surprisingly, MPA's 3α-OH metabolites do not affect the investigated GABA A receptors, in spite of their 3α-OH structure, a similar A-ring steroid structure as of e. g., the potent GABA A -receptor modulator allopregnanolone.
The current induced by MPA alone, similar to what is often termed "direct activation" of GABA A receptors by steroids (Chen et al., 2019), may in principle be due either to MPA-induced activation, i.e., opening, of closed GABA A receptors, but may also be due to potentiation of GABA A receptors that are open in the absence of externally applied GABA (Birnir et at, 2000). Indeed, PTX-sensitive GABA A receptors have been shown to account for a major fraction of the resting Clconductance in preoptic neurons (Yelhekar et al., 2017).
We and others have earlier shown that positive GABA A receptor modulating steroids have negative effects on memory and learning (Johansson et al., 2002;Vallee et al., 2001). The findings of the present study implicate that MPA might influence memory and cognitive processes via positive modulation of specific GABA A receptor subtypes, such as those investigated in this study, α5β3γ2L and α2β3γ2S. Our hypothesis was that especially the 3α-hydroxy-5α-reduced MPA-metabolite should be active at GABA A receptors, as steroid compounds with a similar 3α-OH structure have positive modulatory effects at GABA A receptors but not steroids with a ketone at 3-position (Hogenkamp et al., 1997). To our surprise, 3α5α-MPA or 3α-OH-MprogA did not modulate the α1β2γ2Lor the α5β3γ2L-GABA A receptors, while MPA itself, which have a ketone at the 3-position, was found to be a positive modulator of the GABA A receptor subtypes α5β3γ2L and α2β3γ2S.
As shown here, MPA itself acts as a positive GABA A -receptor modulator and "direct activator" (see above) at specific GABA A -receptor subtypes. MPA has been shown to inhibit learning and memory (Braden et al., 2010(Braden et al., , 2011. This is interesting as continuous long-term exposure to low stress concentrations of another positive GABA A receptor-modulating steroid, allopregnanolone, permanently deteriorates memory and learning in transgenic Alzheimer mice (Bengtsson et al., 2012(Bengtsson et al., , 2013 and, after a longer exposure (5 months), also in wild type mice (Bengtsson et al., 2016). However, there are also results showing a positive memory enhancing effect of allopregnanolone when given in high dosages intermittent, once per week, to trans genic AD mice Singh et al., 2012). However, when the allopregnanolone injections were given with short interval, that is 3 times/week, a deteriorating effect is noted in transgenic AD mice  similar to what is noted after continuous exposure in AD-mice. This issue has been described in a review paper were the dual effect of allopregnanolone is discussed in greater detail (Bengtsson et al., 2020).
There is also a vast literature on effects of neuroactive steroids in the brain and a deep review of that literature is out of the scope of this paper. Especially the effect of allopregnanolone has been studied concerning mood, depression, cognition, food intake and appetite with studies in animals and humans. Some of the results are reviewed in these papers (Bengtsson et al., 2020;Bäckström et al., 2011Bäckström et al., , 2021Rasmusson et al., 2021;Bortolato et al., 2021;Yilmaz et al., 2019;Holmberg et al., 2018).
The GABA A receptor subtypes shown to be modulated by MPA were α5β3γ2L and α2β3γ2S. To our knowledge, the effect of MPA on these receptor subtypes has not been described earlier. MPA has been reported not to affect recombinant rat α1β3γ2 GABA A receptors expressed in Xenopus laevis oocytes (Belelli and Herd, 2003), as supported in the present study by the lack of MPA effect at the recombinant human α1β2γ2 GABA A receptors. In the α5β3γ2L receptor subtype, there was an increased deactivation time constant, but no effect was noted in the α1β2γ2 GABA A receptors. Both in the recombinant human receptors and in dissociated cells from rat hypothalamus, MPA showed a "direct activation" effect. In the α2β3γ2S GABA A receptor this effect was concentration dependent. In the α5β3γ2 L subtype, 10 μM MPA gave a maximal effect, likely due to saturation. In the dissociated hypothalamic neurons, the modulatory effect of MPA was mainly noted as a potentiation of the GABA-evoked current amplitude and not as a large effect on the deactivation time constant. We have not measured the receptor subtype content of the hypothalamic cells studied and thus the GABA A receptor subtype composition in the investigated cells is unknown. In the recombinant cells, the GABA A receptor subtypes are well defined and minimal contribution of unknown GABA-evoked current mediators is expected. In the dissociated neurons studied, however MPA effects could hypothetically be mediated by unknown channels and receptor types. Thus, we tested the effect of PTX, a well-known GABA A -receptor channel blocker. The blocking effect of PTX on the majority of current evoked by GABA + MPA suggests that the additional current evoked by adding MPA to GABA is indeed mediated by GABA A receptors.
Other possible effects of MPA on GABA A receptors via indirect mechanisms has been reported in the literature. In vitro, MPA has been shown to be an inhibitor of the enzyme involved in both synthesis and degradation of allopregnanolone (Penning et al., 1985). In the rat dentate gyrus of the hippocampal area, MPA inhibits the degradation of allopregnanolone and thereby claimed to enhance GABA A -receptor mediated inhibitory neurotransmission (Belelli and Herd, 2003;Bernardi et al., 2006). In the rat spinal cord, on the contrary, MPA is shown to reduce the endogenous concentration of allopregnanolone and decrease the effect on the GABA A receptor (Meyer et al., 2008). The results in these studies show disparate effects of MPA via action on enzymes that either increase or decrease concentration of allopregnanolone. It is thus unclear how MPA affects endogenous allopregnanolone concentrations and to what extent in vivo effects of MPA can be attributed to allopregnanolone.
The GABA A -receptor subtypes studied here were chosen for several reasons. The α1 subunit containing GABA A receptor is the most common receptor subtype in the brain and is an intrasynaptic receptor with rapid response (Belelli, and Lambert, 2005;Olsen and Sieghart, 2008;Ghit et al., 2021). The α1 subunit is thought to be involved in sedation and anesthesia and, 3α5α-reduced progesterone metabolites are potent positive modulators of GABA A receptors with sedative effect (Norberg et al., 1987), but as MPA can induce anesthesia, the α1 subunit was an obvious choice (Meyerson, 1967). GABA A receptors including the α5 subunit are mainly found in the hippocampus, have been implicated in memory and learning and are discussed in relation to dementia development (Shumaker et al., 2003;Johansson et al., 2002;Collinson et al., 2002;Maubach et al., 2003;Birzniece et al., 2006). MPA is also known to induce negative mood, and α2 subunit containing GABA A receptors are present in the amygdala and seem to be related to disturbed mood (Löw et al., 2000;Korpi et al., 2002). The enzymes 5αR and 3αHSD are present in the brain and the metabolism of MPA into GABA A -receptor-active metabolites seems a possible explanation to MPA-induced changes seen in women (Shumaker et al., 2003) and in rat (Meyerson et al., 1967). However, as 3α5α-MPA in this study was devoid positive modulating activity at GABA A receptors this hypothesis is not supported. Rather, it seems likely that unmetabolized MPA directly induced the observed effects.
It is well known that progesterone can be metabolized in the brain. Less is known about the metabolism of MPA in the brain. The metabolism of MPA in the liver is mainly via hydroxylation, at positions C6β, C21, C2β, and C1β, mediated primarily via CYP3A4, (Kuhl, 2005). But it has been found that the major hepatic, adrenal and gonadal CYP isozymes contribute very little to the overall steroid metabolism in brain (Miksys et al., 2004). However, 3-hydroxy and 5-dihydro and 3,5-tetrahydro metabolites of MPA are also formed in the brain. (Kuhl, 2001;Stanczyk and Bhavnani, 2015).
MPA concentration in plasma has been determined by liquid chromatography-atmospheric-pressure ionization tandem mass spectrometry (LC-APCI-MS/MS). MPA given subcutaneously with Alzet osmotic pumps gave a plasma concentration of 6.67 ± 1.08 ng/ml = 21 nM (Braden et al., 2010). The concentration of allopregnanolone needed to activate GABA A receptors in the hypothalamus of the rat is as low as 2 nM (Löfgren et al., 2019). However, as seen in the results above, THDOC is more potent than MPA and a higher concentration of MPA was required for a similar effect. However, a continuous concentration of 21 nM MPA for 66 days gives an impairment of memory and learning in rats (Braden et al., 2010).
The present experiments on dissociated neurons from rat hypothalamus confirm the effect of MPA as a positive modulator of PTX-sensitive GABA A receptors. MPA also likely has a direct activating effect alone as seen at least in some of the cells analysed. We do not know the receptor subtypes in the hypothalamic cells studied, but we assume that some of these cells may contain the α1 and do not have α5 or α2 receptor subunits as α1 is highly expressed in hypothalamus (Korpi et al., 2002;Wisden et al., 1992).
Surprisingly, the results from the hypothalamic neurons show that MPA have different pharmacokinetic properties compared to allopregnanolone. MPA mainly affects the amplitude of GABA-evoked currents while allopregnanolone and THDOC mainly affects the deactivation time constant of GABA-evoked currents in the human recombinant GABA A receptors as well as the decay time constant of miniature inhibitory postsynaptic currents (Haage and Johansson, 1999). Our expectation was that MPA would prolong the effect of GABA on the GABA A receptors, e.g. like the neurosteroids allopregnanolone and THDOC, but MPA did not. The different effects of MPA and of allopregnanolone on GABA A -receptor kinetics will be investigated in more detail in forthcoming studies.

Conclusions
These results show that MPA directly affects α5and α2-subunitcontaining GABA A receptors and positively modulates the effect of GABA on such receptor subtypes. MPA does not affect the α1β2γ2L GABA A -receptor subtype. Unexpectedly, MPA's 3α-OH metabolites did not affect the studied receptors, rather the effects were ascribed to MPA as such. These findings imply that MPA is also likely to directly influence memory and cognitive processes while MPA's 3α-OH metabolites, not effective as GABA A receptor modulators, are less likely to modulate memory function.

Conflict of Interest
This study is supported by an EU-grant, H2020 project number 721802, to Umecrine via the consortium SYNDEGEN for RD, Umecrine AB, Umecrine Cognition AB, and Umeå University Medical Faculty (Karin and Harald Silvander fund). TB has shares in Umecrine AB. JS, DH and GR has been employed by Umecrine AB and MJ by Umecrine Cognition AB. Other investigators have no conflict of interest.