Sesquiterpenes and sesquiterpenoids harbor modulatory allosteric potential and affect inhibitory GABAA receptor function in vitro

Naturally occurring compounds such as sesquiterpenes and sesquiterpenoids (SQTs) have been shown to modulate GABAA receptors (GABAARs). In this study, the modulatory potential of 11 SQTs at GABAARs was analyzed to characterize their potential neurotropic activity. Transfected HEK293 cells and primary hippocampal neurons were functionally investigated using electrophysiological whole‐cell recordings. Significantly different effects of β‐caryophyllene and α‐humulene, as well as their respective derivatives β‐caryolanol and humulol, were observed in the HEK293 cell system. In neurons, the concomitant presence of phasic and tonic GABAAR configurations accounts for differences in receptor modulation by SQTs. The in vivo presence of the γ2 and δ subunits is important for SQT modulation. While phasic GABAA receptors in hippocampal neurons exhibited significantly altered GABA‐evoked current amplitudes in the presence of humulol and guaiol, negative allosteric potential at recombinantly expressed α1β2γ2 receptors was only verified for humolol. Modeling and docking studies provided support for the binding of SQTs to the neurosteroid‐binding site of the GABAAR localized between transmembrane segments 1 and 3 at the (+α)‐(‐α) interface. In sum, differences in the modulation of GABAAR isoforms between SQTs were identified. Another finding is that our results provide an indication that nutritional digestion affects the neurotropic potential of natural compounds.


| INTRODUC TI ON
GABA A Rs represent the major inhibitory ligand-gated chloride ion channels in the central nervous system. GABA A Rs mediate fast phasic synaptic inhibition as well as tonic perisynaptic and extrasynaptic inhibition, and are expressed in brain areas, for example, cortex, hippocampus, and olfactory bulb. They contribute to the balance between excitatory and inhibitory neurotransmission processes by inhibiting incoming action potentials through chloride ion influx (Mortensen et al., 2012;Sieghart, 2006). GABA A Rs are heteropentameric receptors that belong to the superfamily of Cys-loop receptors also including glycine receptors, the 5HT 3 receptor, and nicotinic acetylcholine receptors (nAChRs).
Common to all Cys-loop receptors is their large N-terminal domain followed by four transmembrane domains (M1-4) with M2 representing the ion channel pore, and a short extracellular C-terminus.
The large intracellular loop between M3 and M4 is of the highest variability between Cys-loop receptors bearing binding sites for other structural proteins .
Recent structures revealed an arrangement of α-β-α-β-γ/δ in a clockwise manner (Laverty et al., 2017;Masiulis et al., 2019). The large extracellular N-terminus harbors the ligand-binding site for GABA at the interface of two adjacent subunits (α and β). The ion channel can be blocked by the antagonist picrotoxinin (PTX) which binds to residues located in M2 constricting the ion channel pore to about 1.5 Å. Following the binding of another antagonist, bicuculline, to the orthosteric-binding site, the GABA A Rs turn into a closed conformation similar to the PTX-bound state (Masiulis et al., 2019).
Diazepam, a benzodiazepine, is a potent allosteric modulator of GABAergic function binding specifically to the orthosteric-binding site at the interface of α 1 and γ 2 subunits. Diazepam is an established drug used as a muscle relaxant and a sedative substance (Rudolph & Knoflach, 2011). Recently, a second binding site for benzodiazepines was demonstrated responsible for the observed biphasic GABA A R potentiation at higher diazepam concentrations (Olsen, 2018;Walters et al., 2000). This binding site is localized between the transmembrane domains at the interface of subunits β (β 3 + ) and α (α 1 -). Similarly, neurosteroid binding has been demonstrated within a hydrophobic-binding pocket between adjacent transmembrane domains (Alvarez & Estrin, 2015;Alvarez & Pecci, 2018;Miller et al., 2017).
Several studies describe modulation of GABA A Rs by natural compounds taken up via nutrition or used in aroma therapy, however, most lack receptor subtype specificity (Johnston et al., 2006).
Positive and negative allosteric modulation of GABA A Rs has been shown, for instance, for flavonoids, terpenoids, phenols, and polyacetylenic alcohols, bearing convulsive, anticonvulsive, sedative or anxiolytic potential. Because of the partly high hydrophobicity of terpenoids, binding to the transmembrane subunit interfaces or modulation of the lipid surrounding have been postulated as underlying mechanisms (Manayi et al., 2016;Silva et al., 2019). Among different terpene subtypes, volatile bicyclic monoterpenoids carrying a hydroxy group revealed positive allosteric modulators at α 1 β 2 γ 2 but also at α 1 β 2 GABA A Rs (Kessler et al., 2012(Kessler et al., , 2014. Modulation of GABAergic function has also been ascribed to sesquiterpenes and sesquiterpenoids (SQTs) Manayi et al., 2016).
In this study, we investigated the modulatory effects of 11 SQTs occurring in different plants like hop and chamomile on GABA A R configurations present in the human brain. Therefore, the GABA A Rs expressed in transfected HEK293 cells and primary hippocampal neurons were characterized by electrophysiological whole-cell recordings. Our functional data were accompanied by structural modeling and molecular docking studies to further investigate possible binding modes of the SQTs studied. We found significant differences in the modulation of GABA A R isoforms between different SQTs as well as between transfected cells and hippocampal neurons. The results also indicate that structural changes because of digestion and biotransformation processes may affect the neurotropic potential of natural compounds.

| Chemical Information
GABA, ZnCl 2 , picrotoxinin, diazepam, and gaboxadol were acquired from Sigma-Aldrich. Stock solutions of GABA (1 M) and ZnCl 2 (100 mM) were prepared in water. Picrotoxinin (50 mM), diazepam (10 mM), and gaboxadol (100 mM) were dissolved in ethanol. Guaiol, α-humulene, α-bisabolol, β-caryophyllene, and nootkatone were purchased from Sigma-Aldrich. β-Caryolanol and humulol were synthesized according to Heinlein & Buettner, (2012). Spathulenol, α-bisabolone oxide A, α-bisabolol oxide A, and α-bisabolol oxide B were isolated by a combination of different extraction and isolation steps. In short, dried chamomile flower heads were extracted with dichloromethane and the volatile fraction was isolated by means of solvent-assisted flavor evaporation (Engel et al., 1999). Centrifugal partition chromatography resulted in the direct isolation of αbisabolone oxide A. Silica gel chromatography and size exclusion chromatography provided further purification of spathulenol, αbisabolol oxide A, and α-bisabolol oxide B. Additionally, α-bisabolol oxide A was isolated by preparative two-dimensional GC. SQT stock solutions (100 mM) were prepared in ethanol. All stock solutions were stored at −20℃. Fresh solutions were prepared from stock solutions on the day of recording (final concentration of ethanol 0.6%).

| Extraction and isolation of SQTs from chamomile and hop for GC-MS analyses
Dried hop cones (2.43 g) and dried chamomile flower heads (2.62 g), both purchased from a local company (Wurdies Kräuter GmbH & Co. KG), were finely chopped for 1 min with a mini chopper (Kenwood CH180, Kenwood Limited). Solvent extraction was performed with dichloromethane (DCM, 50 ml, 30 min). After drying over sodium sulfate, the extracts were applied to solvent-assisted flavor evaporation (SAFE) at 60℃ in a vacuum (Engel et al., 1999). The distillates were concentrated to 100 µl by Vigreux and subsequent micro distillation at 50℃ (Bemelmans, 1979). As the last step, the distillates were diluted 1:10 (DCM) for GC-MS analysis.

| GC-MS analyses of the hop and chamomile distillates
The analyses were performed on a GC 6,890 (Agilent Technologies, Santa Clara, CA, USA) connected to a MSD 5,973 (Hewlett-Packard, Palo Alto, CA, USA), equipped with a GERSTEL MPS 2 multipurpose sampler and a GERSTEL CIS 3 injection system (GERSTEL GmbH & Co. KG, Mülheim an der Ruhr, Germany). An uncoated fused silica capillary pre-column (3 m × 0.53 mm i.D.) was fixed to a DB-5 capillary column (30 × 0.25 mm i.D., film thickness 0.25 µm; both from Agilent J&W Scientific). The oven program was started at 60℃ and was raised at 3℃/min until 246℃. The carrier gas (helium) flow was set to 1 ml/min. A split of 20:1 was used for the application of the sample (2 µl). EI mass spectra were recorded in full scan mode (40.0-400.0 amu) using 70 eV.

| Cell lines
HEK293 human embryonic kidney cells (CRL-1573, RRID:CVCL_0045, ATCC; passages 33 lot 70016364 and 34 lot 61714301) were grown in minimal essential medium (MEM) supplemented with 10% fetal calf serum, 200 mM GlutaMAX, 100 mM sodium pyruvate, 100 U/mL penicillin, and 100 µg/ml streptomycin (Thermo Fisher Scientific) under standard growth conditions at 37℃ and 5% CO 2 . Following thawing, cells used for experiments were between passages 6 and 20. The cell line is not listed as a commonly misidentified cell line by the International Cell Line Authentication Committee.

| Neuronal preparation and culture
Hippocampal neurons were prepared from mouse embryos at embryonic day 17 (E17) from pregnant female wild-type CD1 mice (Strain code: 022, RRID:IMSR_CRL:022, Charles River). Female pregnant mice were subjected to a 10-min-deep CO 2 anesthesia. Experiments were authorized by the local veterinary authority and Committee on the Ethics of Animal Experiments (Regierung von Unterfranken, license FBVVL 568/200-324/13). In short, hippocampi from all embryos (n = 10-16) were pooled and trypsinated for 30 min by incubation in 0.5 mg/ml trypsin, 0.2 mg/ml EDTA, and 10 µg/µl DNase I in PBS for 30 min at 37℃. After adding 10% fetal calf serum, neurons were dissociated by trituration, counted, and seeded on polyl-lysine-coated coverslips. Hippocampal neurons were cultured under standard growth conditions at 37℃ and 5% CO 2 in a neurobasal medium containing 2 mM GlutaMAX and 2% (v/v) B27 supplement (Thermo Fisher Scientific). Neurons were used for patch clamp experiments after 18-21 days in culture.

| Electrophysiology
Whole-cell recordings of transfected HEK293 cells and hippocampal neurons were obtained using the patch clamp technique.

| Molecular modeling and docking
The docking study was performed at two different locations at the transmembrane domain of the GABA A -receptor: the diazepambinding site in the β + α --interface and the pregnanolone-binding site in the αα-interface. For the former, a CryoEM structure of the human full-length α 1 β 3 γ 2 -GABA A -receptor was used (PDB: 6HUP) (Masiulis et al., 2019), for the latter (i.e., the pregnanolone-binding site) the crystal structure of an α 5 β 3 -chimera (PDB: 5O8F) (Miller et al., 2017). The preparation of the structures as well as the setup of the docking calculations with AutoDock (Morris et al., 1998) are described in detail in the Supplemental Information.

| Statistical analysis
GraphPad Prism 9.0.0 (GraphPad Software) was used to calculate mean values, standard deviation, standard error of the mean, and statistical significance. The two-tailed paired t test was used to estimate significance values with *p <.05, **p <.01, ***p <.001, ****p <.0001. Data were not assessed for normality, no test for outliers was conducted, and no sample size calculation was done.
The sample size is based on previous studies of a similar design (Milanos et al., 2018).

| Study and experimental design
This study was not a pre-registered study. Therefore, no randomization was performed to allocate subjects in the study. Moreover, this study did not include animal experiments. Hippocampal cultures were obtained from pooled hippocampi of embryos from pregnant female CD1. Three independent cultures were used for the electrophysiological recordings from mature neurons between DIV 18-21 in culture. No blinding of the experimenter was performed in the electrophysiological data acquisition and analysis.
Exclusion criteria for transfected HEK293 cells used for electrophysiological measurements: We defined exclusion criteria as transfected cells might not have picked up all plasmids (3 to express a defined type of GABA A R). Among the cells measured, one cell did not show a picrotoxinin block arguing that the expression of the GABA A R was not achieved. Thus, this cell was excluded from analysis (8 of 9 cells were used for analysis). For neurons, not all neurons in the hippocampal culture represent GABAergic cells. Picrotoxinin blocks GABAergic neurons and was used as a selection criterion. Synaptic GABA A Rs were verified by potentiation with diazepam. Gaboxadol, a superagonist at tonic GABA A Rs, was used to show that cells express tonic GABA A Rs. Using these criteria, we excluded two cells from our initial recordings from hippocampal neurons not exhibiting a picrotoxin block and concomitantly no diazepam potentiation.

| Positive and negative allosteric modulation of GABA A Rs of the α1β2 subtype by SQTs
In this study, the modulatory effect of 11 different SQTs was investigated on GABA A Rs ( Figure 1a). β-Caryophyllene and α-humulene are major compounds of the volatile fraction of hop.
In previous studies it was shown, that they are transformed into β-caryolanol and humulol, respectively, when applied to an in vitro digestion model (Heinlein & Buettner, 2012)   Previous studies demonstrated that GABAergic modulation by terpenoids is most probably independent of the presence of the γ 2 subunit in the pentameric GABA A R complex (Kessler et al., 2014). Therefore, the modulatory effect of SQTs was first analyzed on heteromeric GABA A Rs composed of ⍺ 1 and the β 2 subunits. Cells were co-transfected with a green fluorescent protein (GFP) to distinguish transfected from untransfected cells during patch clamp experiments by using a fluorescence microscope.
Multiple receptor configurations in the pentameric receptor complex are possible resulting from ⍺ 1 β 2 subunit transfection ( Figure 2a). Their effects on agonist potency are unknown, but expression differences seem to influence receptor stoichiometry, with 2:3 being the most likely one (Wagoner & Czajkowski, 2010).
For analysis of SQT modulation, 10 µM GABA (referring to EC 10-30, (Milanos et al., 2018) was applied for 50 ms, followed by a 50 ms coapplication with 600 µM of the appropriate modulator (Figure 2b, c). A concentration of 600 µM for the SOTs was chosen because of previous results on other terpenes starting with modulation at concentrations of 100 µM and being most effective between 300 µM and 1 mM (van Brederode et al., 2016;Kessler et al., 2014;Milanos et al., 2018). Obtained maximum current (I max ) values were divided by the mean GABA current for better comparability. Co-application with β-caryolanol (146 ± 24%, p = .017) and humulol (138 ± 29%, p = .033) resulted in significantly increased I max , whereas α-bisabolone oxide A (89 ± 22%, p = .023), α-bisabolol oxide B (90 ± 20%, p = .009), (E)β-farnesene (83 ± 19%, p = .010), and guaiol (62 ± 16%, p = .0039) significantly reduced GABA-induced I max values (Table 1). Representative current traces for these recordings are shown in Figure 2d. The significantly different effects of β-caryophyllene and α-humulene and their respective derivatives β-caryolanol and humulol suggest that the additional hydroxy group may account for the observed differences in receptor modulation. Furthermore, α-bisabolone oxide A, α-bisabolol oxide B, (E)β-farnesene, and guaiol exhibited a negative allosteric modulatory effect independent of the presence of a hydroxy group. As a single concentration of 600 µM SQTs was used for recordings, we In our initial screen, a modulator concentration of 600 µM was used. As it is known that hydrophobic chemicals can bind to plastic tubes (Heinlein et al., 2014) as used in the application system of the electrophysiological setup, samples of the applied solutions were taken before and after passage through the application system. These samples were analyzed by GC-MS to reveal concentration differences between the start and end of the recording session (Supporting information Figure S1). Our data revealed a reduction of the used substances by a mean of 52% (~300 µM released by the application system) while transported through the tubing system, suggesting that the real modulatory effect is probably much more pronounced. With a reduction of the concentration of the substances released by the perfusion system, a concentration ~300 µM should still be sufficient to modulate GABA A receptors (Milanos et al., 2018).

| Modulation of the GABAergic response by co-applied SQTs is difficult to predict
As some SQTs modulate GABA A Rs of the ⍺ 1 β 2 subtype positively and others modulate negatively, we hypothesized additive or neutralizing effects for co-applications of two modulators. β-Caryolanol showed the largest positive modulation. It was co-applied with humulol, another positive modulator to investigate additive effects.
A second round of co-application analyzed competing effects between SQTs, respectively. Here, β-caryolanol together with the negative modulators α-bisabolol and guaiol, were used (Figure 3a, b). 10 µM GABA were co-applied with 500 µM of each modulator.

| The GABA A R subunit γ 2 plays a major role in the modulation by SQTs
SQTs modulate ⍺ 1 β 2 GABA A Rs in transfected HEK293 cells, but they might have a different effect in vivo, where multiple GABA receptor configurations are present in the same neuron (phasic receptors containing the γ 2 subunit and tonic receptors containing a δ subunit). Among possible GABA A R subunit combinations, ⍺ 1 β 2 γ 2 is the most abundant receptor configuration in the brain (Figure 4a) (Whiting, 2003 and α-humulene (93 ± 19%) when co-applied with 10 µM GABA (Figure 4b-g). However, a significant decrease in current for 10 µM GABA + humulol (78 ± 19%, p = .0056, Table 2) as well as a significant increase for guaiol (120 ± 27%, p = .0008). These data are in contrast to the effects observed for ⍺ 1 β 2 receptors in transfected HEK293 cells and might argue for differences in GABAergic modulation by GABA A Rs containing γ or also possibly δ subunits or other ⍺/β subunits than ⍺ 1 /β 2 . At GABA concentrations of 10 µM, tonic GABA A receptors including the δ subunit are saturated and may less contribute to the observed effects than synaptic receptor compositions including the γ subunit (Karim et al., 2013;Mortensen et al., 2012). Therefore, GABA A R modulation was generally tested in hippocampal neurons by diazepam, a positive allosteric modulator, the antagonist picrotoxinin, or the agonist gaboxadol. Diazepam (138 ± 33%, p ≤ 0.0001) and gaboxadol (176 ± 41%, p = .0025) application resulted in significantly increased currents and hint for the presence of γ and ẟ containing receptors (Chua & Chebib, 2017). In contrast, GABAergic currents were significantly reduced by co-application of GABA and picrotoxinin to 41%-59% (Figure 4c, f, h-j). These data suggest that the presence of different GABA A R configurations in one neuron harboring γ and ẟ subunits may play a role in GABA A R modulation by SQTs. GABA A Rs containing the γ 2 subunit enable phasic inhibition, receptors harboring a ẟ subunit mediate tonic inhibition (Olsen & Sieghart, 2009).
To investigate the SQT modulation on phasic GABA A Rs, HEK293 cells were transfected with ⍺ 1 β 2 γ 2 receptors to see if the presence of the Taken together, these results indicate that the γ 2 subunit plays a major role in the modulation of GABA A Rs by SQTs.
The presence of the expression of tonic receptors was verified by F I G U R E 3 Co-application of two SQTs generates competitive rather than additive modulation at GABA A Rs. (a) Currents acquired by patch clamp recordings from HEK293 cells expressing ⍺ 1 β 2 GABA A Rs. Currents were recorded during the application of 10 µM GABA (black dots) or with co-application of two modulators with a concentration of 500 µM per modulator (red dots). Each pair of dots represents one patched HEK293 cell. Experiments from 3 recording sessions using independent cell batches and transfections are shown, number of cells recorded was 9 for every condition, significance level ***p ≤ .001. (b) Differences between each current pair in a are shown gaboxadol which increased current amplitudes for ⍺ 6 β 3 δ receptors.
In sum, δ-subunit containing GABA A Rs are most probably rather negatively modulated by SQTs.

| SQTs with their lipophilic scaffold fit well to the binding sites of neurosteroids and diazepam localized in the transmembrane segment of GABA A Rs
A binding site for SQTs is not described, yet. SQTs are highly hydrophobic. Hence, they might favor a localization within the Moreover, naturally occurring substances, such as terpenoids, are able to modulate GABAergic function (Manayi et al., 2016). In this study, SQTs isolated from chamomile were found to harbor allosteric potential at GABA A Rs in transfected cell lines and primary neurons. Previously, we have identified odorous monoterpene structures from Sideritis extracts with a preference of a bicyclic character in combination with the presence of a hydroxy group harboring positive allosteric potential at GABA A Rs of the ⍺ 1 β 2 configuration in vitro (Kessler et al., 2012(Kessler et al., , 2014. Terpenoids represent a diverse group of natural compounds formed by the condensation of isoprene units with the potential to modulate GABA A Rs and thus discussed for treatment of CNS disorders (Manayi et al., 2016). Valerenic acid, a sesquiterpenoid from Valeriana officinalis L., is used to treat anxiety and sleep disorders.
It binds to the GABA A Rs dependent on the β isotype present in the complex (Khom et al., 2007). For bilobalide, present in Gingko biloba, different actions at the GABA A R have been described.
Anticonvulsant but also antagonistic action at ⍺ 1 β 2 γ 2L GABA A Rs similar to bicuculline and picrotoxin have been found (Huang et al., 2003;Kiewert et al., 2007). Here, we studied the neurotropic potential of SQTs assumed to be taken up by nutrition and digested in the human body. Biotransformation processes might alter the modulatory potential, for example, gastrointestinal processes were identified to modulate the chemical composition of ingested aroma constituents thus changing bioactivity of essential oils (Heinlein & Buettner, 2012). However, about the underlying mechanism of the gut-brain axis there is little systematized effort.
⍺-humulene, a monocyclic, and β-caryophyllene, a bicyclic sesquiterpene, are major components of the volatile fraction of hop but also part of many other essential oils. As shown in previous studies, metabolization processes taking place during intake and digestion of these compounds lead to changes in the structure. Both, ⍺-humulene and β-caryophyllene are hydroxylated because of the low pH in the gastric phase of digestion which TA B L E 2 Electrophysiological data obtained from hippocampal neurons may affect the modulatory potential of the substance (Heinlein & Buettner, 2012). Indeed, in transfected HEK293 cells with GABA A Rs of the ⍺ 1 β 2 subtype, the hydroxylated metabolites humulol and β-caryolanol showed positive allosteric potential by significantly increasing the GABAergic current upon co-application of GABA in a low concentration. In contrast, the parent compounds ⍺-humulene and β-caryophyllene had no effect on GABAergic current. These data are in line with previous reports showing positive allosteric modulation of the GABAergic currents by terpenes and terpenoids harboring a hydroxy group (Kessler F I G U R E 5 SQT modulation of phasic ⍺ 1 β 2 γ 2 GABA A Rs. (a-d) Absolute currents of patch clamp recordings during application of 10 µM GABA (black dots) or with co-application of 600 µM modulator (red dots), recorded from HEK293 cells transfected with ⍺ 1 β 2 γ 2 GABA A Rs. 10 µM GABA was co-applied with 10 µM diazepam, 10 µM ZnCl 2 , or 100 µM picrotoxinin. Each pair of dots represents one patched HEK293 cell. Patch clamp experiments were performed on at least 3 days with different cell batches, total numbers of recorded cells (n = 8-12) are shown in Table 3 Nevertheless, the diazepam site as well as the alternative ⍺ 1 + β 3 -site also contain some polar amino acids. As these would presumably interact with water molecules in the unbound state, binding at these sites would be expected to result in a strong desolvation penalty, TA B L E 3 Electrophysiological data obtained from ⍺1β2γ2, ⍺ 4 β 3 δ, and ⍺ 6 β 3 δ GABA A Rs This difference in burial is markedly reflected in the scoring, which favors direct interactions, but does not consider favorable exposure to a lipophilic environment. Accordingly, better scores of the poses in the diazepam-binding site should not be overinterpreted. Rather, the structure of the neurosteroid-binding site would better match the scaffold of hydroxylated SQTs, which is also supported by the observation that the hydroxylated SQTs, when docked to the entire αα-interface, were located in the neurosteroid-binding site, whereas the non-hydroxylated ones were not found at that location. Hence, binding to the lipophilic interface forming the neurosteroid-binding site is possibly preferred over the diazepam-binding site, especially for humulol and β-caryolanol. In addition, it is known that the hydroxy group of neurosteroids is crucially involved in their GABApotentiation effects (Miller et al., 2017), which could possibly explain the effects of SQTs. Nevertheless, an interaction to other modulatory proteins of the GABA A R cannot be excluded.
Patch clamp experiments were performed on at least 3 days with different cell batches, number of recorded cells were between 8 and 10. *p ≤ .05, **p ≤ .01. (e-g, l-m) Differences between each current pair in b-d, i-k are shown sesquiterpenoids also harboring a hydroxy group, for example, bisabolol, bisabolol oxide B, and spathulenol did not follow this pattern.
Therefore, we investigated sesquiterpene modulation at isolated primary neurons of the hippocampus rich in GABAergic interneurons (Pelkey et al., 2017). The main receptor configuration is the ⍺ 1 β 2 γ 2 subtype (Olsen & Sieghart, 2008). Those GABA A Rs enable phasic inhibition whereas subtypes including the ⍺4, ⍺6, or δ subunit regulate long-lasting inhibition and thereby control the excitability of the neurons (Lee & Maguire, 2014). The neurotropic activity of the SQTs was different on primary hippocampal neurons in comparison to the controlled over-expression of the ⍺ 1 β 2 subtype in HEK293 cells. While βcaryolanol showed no longer a significant modulatory effect, humulol turned into a negative and guaiol into a positive allosteric modulator.
Control measurements using the superagonist gaboxadol and the benzodiazepine agonist diazepam confirmed the presence of phasic as well as tonic GABA A Rs in the hippocampal neurons. Back to controlled over-expression, the presence of the γ 2 subunit in the ⍺ 1 β 2 γ 2 subunit composition led to the same observation as in hippocampal neurons with humulol as a negative allosteric modulator. Hence, the presence of the γ 2 subunit in the majority of GABA A Rs (65%) in primary neurons is an important component for sesquiterpene modulation. At tonic GABA A Rs in the ⍺ 4 β 3 δ configuration, β-caryophyllene and ⍺-humulene as well as the sesquiterpenoids humulol and β-caryolanol but also guaiol exhibited negative allosteric modulation.
Positive allosteric modulation of tonic GABA A Rs has been reported for monoterpenoid structures, verbenol and myrtenol, measured on hippocampal slices (van Brederode et al., 2016). In contrast to monoterpenes, our results show the negative modulatory potential of SQTs at GABA A Rs mainly at tonic receptor subtypes. Bilobalide, a sesquiterpene lactone, antagonizes GABA A Rs similar to bicuculline and picrotoxinin and the SQTs used here, but also exerts anticonvulsant activity via a presynaptic route maintaining the GABA levels and thus the GAD (glutamate decarboxylase) activity in the hippocampus and cerebral cortex (Sasaki et al., 2000). The action of sesquiterpenes at postsynaptic GABA A Rs but also at presynaptic sites regulating the activity of the GABA synthesizing enzyme GAD has been supported recently for the malaria drugs artemisinins. Artemisinins modulate GABA A Rs via targeting the binding epitope of the scaffold protein gephyrin by hindering the critical interaction of gephyrin and GABA A Rs F I G U R E 7 Molecular docking results at the neurosteroid-binding site localized in the TM domain of the GABA A R. As a reference, the experimentally observed binding mode of pregnanolone in the αα-interface (PDB: 5O8F) is shown in (a). Binding modes proposed by docking of the SQTs guaiol (b), β-caryolanol (c), and S-humulol (d) are illustrated based on the top-scoring docking solutions and thus decreasing GABAergic inhibition .
Moreover, sesquiterpenes interact with PDXK (pyridoxal kinase), an enzyme regulating the synthesis of GABA via GAD activity in presynaptic terminals (Kasaragod et al., 2020). Both of the abovementioned actions, however, require a longer presence of the SQTs enabling uptake into the cell to facilitate their action.
Based on a qualitative analysis of the binding modes obtained from our docking studies, preferential binding of the hydroxy group carrying SQTs to the neurosteroid-binding site and the diazepam site seems possible. Furthermore, we cannot exclude an indirect interaction of SQTs with associated proteins of the GABA A R or interactions with presynaptic enzymes and thereby modulating GABAergic activity. In our approach, the SQTs were co-applied together with the neurotransmitter GABA enabling monitoring of short and direct effects on receptor opening and closing only.
Together with the recently observed shared potential of SQTs at postsynaptic and presynaptic sites, short-term and long-term aspects at GABAergic neurotransmission in the presence of SQTs should be investigated in more detail in the future, combined with affinity and binding studies to differentiate direct from indirect actions of SQTs on GABAergic inhibition processes. A better understanding of these mechanisms will further define their therapeutic potential for CNS disorders. All experiments were conducted in compliance with the ARRIVE guidelines.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R S' CO NTR I B UTI O N S
CV and AB participated in research design. Electrophysiology was performed by DJ. GC-MS and isolation of SQTs was performed by BS. DJ, CV, BS, and HL carried out data analysis. MZ and CS carried out modeling and docking analysis. CV, DJ, and BS contributed to the writing of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.