Forskolin reverses the O-GlcNAcylation dependent decrease in GABAAR current amplitude at hippocampal synapses possibly at a neurosteroid site on GABAARs

GABAergic transmission is influenced by post-translational modifications, like phosphorylation, impacting channel conductance, allosteric modulator sensitivity, and membrane trafficking. O-GlcNAcylation is a post-translational modification involving the O-linked attachment of β–N-acetylglucosamine on serine/threonine residues. Previously we reported an acute increase in O-GlcNAcylation elicits a long-term depression of evoked GABAAR inhibitory post synaptic currents (eIPSCs) onto hippocampal principal cells. Importantly, O-GlcNAcylation and phosphorylation can co-occur or compete for the same residue; whether they interact in modulating GABAergic IPSCs is unknown. We tested this by recording IPSCs from hippocampal principal cells and pharmacologically increased O-GlcNAcylation, before or after increasing serine phosphorylation using the adenylate cyclase activator, forskolin. Although forskolin had no significant effect on baseline eIPSC amplitude, we found that a prior increase in O-GlcNAcylation unmasks a forskolin-dependent increase in eIPSC amplitude, reversing the O-GlcNAc-induced eIPSC depression. Inhibition of adenylate cyclase or protein kinase A did not prevent the potentiating effect of forskolin, indicating serine phosphorylation is not the mechanism. Surprisingly, increasing O-GlcNAcylation also unmasked a potentiating effect of the neurosteroids 5α-pregnane-3α,21-diol-20-one (THDOC) and progesterone on eIPSC amplitude, mimicking forskolin. Our findings show under conditions of heightened O-GlcNAcylation, the neurosteroid site on synaptic GABAARs is accessible to agonists, permitting strengthening of synaptic inhibition.


Introduction
Gamma-aminobutyric acid type A receptors (GABA A Rs) are heteropentameric ligand-gated chloride channels composed of α, β, γ, and sometimes δ subunits that mediate both fast synaptic and tonic inhibition, depending on their synaptic vs extrasynaptic location, respectively 1 .Mutations in speci c subunits are linked to epilepsy syndromes 2 , and gene polymorphisms in speci c GABA A R subunits associate with neuropsychiatric disorders, including alcohol use disorder 3 , anxiety 4 , schizophrenia 5 , bipolar disorder 5 , and even major depressive disorder 6 , including postpartum depression 7 .
For decades it has been appreciated that neurosteroids mediate their sedative hypnotic and anxiolytic effects via positive allosteric binding to speci c GABA A R subunits, particularly extrasynaptic GABA A Rs containing α5, α4 or δ subunits 8,9 .GABA A R function is also potently modulated by serine phosphorylation 10 .For example, protein kinase A (PKA) mediated phosphorylation of speci c serines on synaptic GABA A Rs induced by application of the adenylate cyclase activator, forskolin, bidirectionally modulates GABA-gated current amplitude and induces endocytosis depending on the neuron type and phosphorylated residue 8, [11][12][13][14] .Importantly, the phosphorylation state has direct consequences on potency of allosteric modulation of GABA A Rs by neurosteriods, barbiturates, and benzodiazepines in a subunit-speci c manner 10 .
The post-translational modi cation of proteins by β-N-acetylglucosamine via an O-linkage on serine and threonine residues (O-GlcNAcylation), can modulate protein phosphorylation by competing directly with phosphorylation for the same residues, or indirectly via modi cation of other sites thereby changing protein structure and protein-protein interactions.In addition, many kinases are modi ed by O-GlcNAc and this can regulate their function 15 .Both O-GlcNAcylation and phosphorylation are dynamic, reversible, and ubiquitous.While many kinases and phosphatases exist, O-GlcNAc is tightly regulated by a single enzyme pair, OGT (O-GlcNAc transferase) and OGA (O-GlcNAcase), which add and remove O-GlcNAc from serine/threonine residues, respectively.Of note, these enzymes are highly expressed in hippocampus 16,17 .The OGT substrate, UDP-GlcNAc, is generated by the hexosamine biosynthetic pathway (HBP) via glucose metabolism potentially linking this modi cation to nutrient availability 18,19 .
Our lab recently reported that O-GlcNAcylation can be increased within minutes by exposing hippocampal brain slices to the HBP substrate glucosamine (GlcN), the OGA inhibitor thiamet-G (TMG) or in combination and this leads to depression of excitatory synaptic transmission 20,21 .More recently, we reported that pharmacologically increasing O-GlcNAc using GlcN and TMG in combination or GlcN alone induces a long-lasting depression of GABA A R-mediated inhibitory postsynaptic currents (IPSCs) at hippocampal synapses within minutes, and decreases the amplitude and frequency of spontaneous IPSCs (sIPSCs) 22 .In Oga +/mice, where O-GlcNAc is chronically elevated, inhibitory synaptic transmission was reduced in medial prefrontal cortex, and this was rescued by OGA overexpression via adeno-associated viral injection 23 .Additionally, the Oga +/mice exhibited an antidepressant-like behavior, which was also reversed by viral OGA overexpression.Collectively, acute and chronic elevation in O-GlcNAcylation in vitro and in vivo depresses GABA A R-mediated synaptic inhibition.These ndings highlight O-GlcNAcylation as a critical regulator of the e cacy of synaptic neuronal inhibition, in both physiological and pathophysiological conditions of elevated O-GlcNAcylation.
Although serine phosphorylation 10,11 and O-GlcNAcylation 22 are fundamental modulators of the strength of GABA A R-mediated synaptic inhibition, no study has examined whether these modi cations interact in modifying the e cacy of synaptic inhibition.We investigated this possibility using electrophysiology in hippocampal slices and pharmacologically increased O-GlcNAc using glucosamine (GlcN) and thiamet-G (TMG) in combination.We also increased serine phosphorylation using the adenylate cyclase activator forskolin.Unexpectedly, we found that a prior increase in O-GlcNAcylation, which induces depression of GABA A R-mediated eIPSC amplitude, unmasks a forskolin-dependent increase in eIPSC amplitude, even though forskolin had no effects when applied in the absence of increased O-GlcNAcylation.Surprisingly, inhibition of adenylate cyclase or PKA did not prevent the potentiating effect of forskolin on eIPSC amplitude, indicating that serine phosphorylation is not the mechanism.Similar to ndings in a study in carp amacrine-like cells showing that forskolin binds to a GABA A R neurosteroid site 24 , we found that the neurosteriods, THDOC and progesterone potentiate the IPSC amplitude following a prior increase in O-GlcNAcylation, mimicking the effect of forskolin.These ndings suggest that O-GlcNAcylation promotes neurosteroid site accessibility on GABA A Rs thereby reversing the depressive effect of O-GlcNAcylation and strengthening synaptic inhibition.

Results
An acute increase in O-GlcNAcylation reduces inhibitory post-synaptic currents on CA1 pyramidal cells.
The O-GlcNAc-induced depression of synaptic inhibition is not prevented by an actin stabilizer, but is partially dependent upon dynamin-mediated GABA A R endocytosis.
Because serine phosphorylation can cause GABA A R endocytosis in a subunit and serine speci c manor 10,25,26 by analogy, we speculated that the O-GlcNAc-induced LTD of synaptic inhibition (or O-GlcNAc iLTD) we previously reported 22 is occurring via GABA A R endocytosis.This possibility is supported by ndings from our lab and others that O-GlcNAcylation of the GluA2 AMPAR subunit leads to long-term depression of excitatory transmission (O-GlcNAc LTD) at hippocampal CA3-CA1 synapses 20 and causes AMPAR endocytosis 27 .
Next, we tested whether expression of O-GlcNAc iLTD requires dynamin-dependent GABA A R endocytosis.
Importantly, it was noted that the dataset with dynasore had greater variability during iLTD expression as indicated by the larger error bars between 20-25 mins (Fig. 3Aii).Upon further inspection of individual experiments, we recognized that in some cells (n = 4/8), the eIPSC amplitude was potentiated following GlcN + TMG, which would be expected if endocytosis is prevented.Therefore, when the dynasore dataset was separated into those with potentiation and those without, a clear population of cells was revealed that exhibited signi cant potentiation of the eIPSC amplitude (Fig. 3Aiv, 136.3 ± 10.2% of normalized eIPSC amplitudes following GlcN + TMG, application n = 4 cells, p = 0.039, paired t-test), while the remaining population exhibited no change in eIPSC amplitude from the previously depressed level following GlcN + TMG application (Fig. 3Aiv, 73.9 ± 13.5% of normalized eIPSC amplitudes following GlcN + TMG, application n = 4 cells, p = 0.13, paired t-test) and from the O-GlcNAC iLTD without dynasore (Fig. 3Ai) (p = 0.131, unpaired t-test).Additionally, there was a signi cant difference between the potentiated versus the non-potentiated population of normalized eIPSC amplitudes following GlcN + TMG (p = 0.03, paired t-test).This result is consistent with the interpretation that in some cells increasing O-GlcNAc induced a dynamin-dependent endocytosis of GABA A Rs that could underlie the synaptic depression and in others a different mechanism exists.Furthermore, these ndings rea rm the heterogeneity in GABA A Rs that exist at synapses in hippocampus.Additional experiments are needed to fully understand how O-GlcNAc impacts GABA A R tra cking.

Possible interactions between phosphorylation and O-GlcNAcylation
Next, we wanted to determine if serine phosphorylation and O-GlcNAcylation interact to affect GABA A Rmediated synaptic inhibition and whether an order effect exists.For decades, forskolin has been used to activate adenylate cyclase to drive protein kinase A (PKA) dependent phosphorylation of AMPARs at excitatory synapses in hippocampus, leading to synaptic potentiation [29][30][31] .PKA-dependent serine phosphorylation also modulates synaptic inhibition, but the effect is variable depending on the GABA A R subunit con rmation 10 .Therefore, to test whether serine phosphorylation has an impact on subsequent induction of O-GlcNAc iLTD, we recorded from CA1 pyramidal cells (Cs Gluconate pipette solution; E Cl− = −60 mV), and bath applied forskolin (50µM) for 10 min to drive activation of adenylate cyclase and PKA followed by 10 min application of GlcN + TMG to induce O-GlcNAc iLTD.The eIPSC amplitudes during forskolin and GlcN + TMG were normalized to baseline and statistically compared by averaging 30 events during (a) baseline, (b) forskolin and (c) GlcN + TMG using repeated measures (RM) RM-ANOVA and Šídák's multiple comparisons post hoc test (Fig. 4Ai-iii, p = 0.0001, RM ANOVA).We found no signi cant effect of forskolin compared to baseline transmission (Fig. 4Aii, 89.2 ± 6.9% of baseline transmission, p = 0.39), and subsequent application of GlcN + TMG induced signi cant iLTD (Fig. 4Aii, 64.2 ± 4.4% of baseline transmission, p < 0.0001).Despite no signi cant effect of forskolin on the eIPSC amplitude in the averaged dataset compared to baseline, we want to ensure there was no effect on the magnitude of subsequently induced O-GlcNAc iLTD.Therefore, we normalized the eIPSC amplitudes at the end of the 10 min forskolin application, thereby establishing new baseline from which to measure the O-GlcNAc iLTD magnitude.We found that from this new baseline, subsequent application of GlcN + TMG induced a signi cant iLTD (74.9 ± 3.8% of new baseline transmission (b-c comparison), p < 0.0001, paired t-test) that is not different from the iLTD magnitude under control conditions obtained in Fig. 2Ai in the absence of jasp (Fig. 2Ai, 67.6 ± 5.6% of baseline transmission versus 73.5 ± 4.2% of new baseline transmission p = 0.53, unpaired t-test).In reviewing the data, it is important to note that there was high cell-to-cell variability in eIPSC amplitude during forskolin application, with some cells displaying potentiation and some depression of the eIPSC amplitude, as can be seen by inspection of the individual data points in the bar chart in Fig. 4Aii.
To determine if this nding in CA1 is generalizable, we performed the same experiment in dentate granule cells (DGCs).Similar to CA1, we recorded from DGCs (Cs Gluconate pipette solution; E Cl− = −60 mV), and bath applied forskolin (50µM) for 10 min followed by 10 min of GlcN + TMG to induce O-GlcNAc iLTD.
To determine if there is any effect of forskolin or GlcN + TMG on presynaptic release probability, we analyzed the paired-pulse ratio (PPR), an indirect measure of presynaptic release probability, during baseline, GlcN + TMG and forskolin.No signi cant differences were detected in CA1 (Fig. 4Aiv, p = 0.69, RM-ANOVA) or in the dentate gyrus (Fig. 4Biv, p = 0.46, RM-ANOVA), indicating that a presynaptic mechanism is not involved.
Next, we performed the experiment in reverse order, increasing O-GlcNAc with GlcN + TMG prior to driving phosphorylation with forskolin.We recorded from both from CA1 pyramidal cells and DGCs and applied GlcN + TMG for 10 min followed by forskolin for 10 min.Similar to above, eIPSC amplitudes during forskolin and GlcN + TMG were normalized to baseline and compared (Fig. 5Ai, p < 0.0001, RM-ANOVA; 5Bi; p = 0.009 RM-ANOVA).A 10 min exposure to GlcN + TMG induced O-GlcNAc iLTD in CA1 pyramidal cells (Fig. 5Aii: 65.4 ± 5.2% of baseline transmission, p = 0.0002; Šídák's post hoc test) and in dentate granule cells (Fig. 5Bi, Bii: 82.3 ± 2.6% of baseline transmission, p = 0.002, Šídák's post hoc test).Surprisingly, subsequent application of forskolin reversed the O-GlcNAc iLTD and elicited a potentiation of the eIPSC amplitude in recordings from both CA1 pyramidal cells and DGCs (Figs. 5Ai-iii and 5Bi-iii).To analyze the magnitude of the forskolin-induced eIPSC potentiation, we re-normalized eIPSC amplitudes at the end of the 10 min GlcN + TMG application, establishing a new baseline, and then normalized forskolin values to the new baseline.We found a signi cant potentiation in CA1 (138.5 ± 7.8% of new baseline (b-c comparison), p = 0.006, paired t-test) and in dentate (143.4 ± 8.8% of new baseline (b-c comparison), p = 0.003, paired t-test) that reverses O-GlcNAc iLTD and in dentate, the potentiation overshoots the original baseline.Furthermore, these results suggest that a prior increase in O-GlcNAc unmasks a possible PKA dependent potentiation of synaptic inhibition that is absent under control conditions.
The forskolin dependent increase in eIPSC amplitude is not PKA dependent.
Since we were unable to block the forskolin mediated potentiation via PKA inhibition, we next targeted adenylate cyclase using the inhibitor, SQ22536 (100 µM).SQ2253 was bath applied for 10 mins before and during forskolin application, and experiments with and without SQ2253 were interleaved.eIPSC amplitudes during GlcN + TMG and forskolin amplitude were normalized to baseline (Fig. 6Ci) and statistically compared (Fig. 6Ci-Ciii).GlcN + TMG induced O-GlcNAc iLTD and subsequent application of forskolin reversed the O-GlcNAc iLTD and elicited a potentiation of the eIPSC amplitude with (Fig. 6Cii, p = 0.031, RM ANOVA) and without SQ22536 (Fig. 6Ciii, p = 0.03, RM ANOVA).In the dataset with SQ22536, post hoc Šídák's multiple comparisons test showed a signi cant difference between baseline vs. GlcN + TMG (Fig. 6Cii, p = 0.031).In the dataset without SQ22536, Šídák's multiple comparisons test showed a signi cant difference between baseline vs. GlcN + TMG (Fig. 6Ciii, p = 0.020).To measure the magnitude of the forskolin-induced potentiation, we re-normalized the eIPSC amplitudes at the end of the 10 min GlcN + TMG application to establish a new baseline, and found signi cant potentiation with (122.7 ± 7.1%, p = 0.017, paired t-test) and without (134.8± 13.3%, p = 0.047, paired t-test) SQ22536, but similar to CA1, there was no signi cant difference between groups (Fig. 6Ci, p = 0.42, unpaired t-test).Being that neither the adenylate cyclase nor PKA inhibitor prevented the forskolin dependent increase in eIPSC following a prior increase in O-GlcNAc, we concluded that this potentiation occurs through another mechanism.
The inability to prevent the forskolin-induced eIPSC potentiation following O-GlcNAc iLTD with adenylate cyclase and PKA inhibitors, and the partial mimic of forskolin's effect by the inactive adenylate cyclase analog 1,9 dideoxyforksolin, was very puzzling.In searching for a possible explanation, we were intrigued by a report where both forskolin and 1,9-dideoxyforskolin accelerated desensitization of GABA A R currents in recordings from amacrine-like cells in carp (Carassius auratus) retina that was resistant to PKA inhibition 24 .Surprisingly, the neurosteroid, 5α-pregnane-3α,21-diol-20-one (THDOC), which is a structural analog to forskolin, also accelerated GABA A R desensitization, mimicking the effect of forskolin 24 .Further experiments led to the conclusion that forskolin is acting at an allosteric neurosteroid site on GABA A Rs.
Because GABA A Rs in mammalian hippocampus are potently modulated by neurosteroids containing speci c subunit combinations 9 , we sought to determine if the hippocampal neurosteroid, THDOC, also mimics the forskolin-induced eIPSC potentation following O-GlcNAc iLTD.
However, similar to the cell-to-cell variability we observed with forskolin (e.g., Fig. 4Aii), we noted variability in the response to THDOC (Fig. 7Bii), with eIPSCs recorded from some cells having a clear potentiation (Fig. 7Ci-iii, n = 7/12), while others had no change (Fig. 7 Di-iii, n = 5/12 cells).To measure the effect of THDOC in these two populations, we re-normalized eIPSC amplitudes at the end of the 10 min GlcN + TMG application to establish a new baseline, then THDOC values were normalized to the new baseline.Similar to forskolin (Fig. 4Ai-iii), we found a signi cant potentiation of the eIPSC amplitude in this subset of cells (116.4 ± 5.3%, p = 0.021, paired t-test, n = 7/12).In the remaining cells, (Fig. 7Di-iii), we found no further change in eIPSC amplitude (90.0 ± 5.5%, p = 0.15, paired t-test) beyond what occurred following GlcN + TMG application.

Discussion
Understanding the mechanisms that modulate the strength of inhibitory transmission at GABAergic synapses is essential to understanding the excitation/inhibition balance critical for brain function.We previously reported that pharmacologically increasing the post-translational modi cation O-GlcNAcylation rapidly depresses spontaneous IPSC frequency and amplitude, and the amplitude of miniature IPSCs, suggesting the mechanism underlying the synaptic depression is postsynaptic 22 .We also reported a long-lasting depression of electrically evoked IPSC amplitude 22 , representing a novel form of LTD of synaptic inhibition and referred to here as O-GlcNAc iLTD.In the current study, we extend these initial ndings by investigating whether increasing O-GlcNAcylation triggers GABA A R endocytosis during expression of O-GlcNAc iLTD and how O-GlcNAcylation and serine phosphorylation might interact in the modulation of GABAergic inhibition on principle cells in hippocampus.
A growing list of mechanisms regulate GABA A R membrane stability and tra cking during long-term changes in strength of inhibitory transmission [36][37][38][39] .Under some conditions, GABA A R receptors can undergo clathrin/dynamin-mediated endocytosis involving PKA-dependent phosphorylation of serine 408 (S408) and S409 on β1 and β3 subunits [40][41][42][43] .Similarly, depending upon the subunit composition, PKC dependent serine phosphorylation also modulates endocytosis, leading to decreased plasma membrane GABA A R density [44][45][46][47] .Because O-GlcNAcylation also occurs on serine residues, and O-GlcNAc LTD of excitatory transmission likely involves endocytosis of AMPARs 20,27 , it seemed probable that GABA A R endocytosis occurs during expression of O-GlcNAc iLTD.Surprisingly, the actin stabilizer, jasplakinolide did not prevent expression of O-GlcNAc iLTD, but pharmacologically inhibiting dynamin-dependent tra cking with dynasore prevented O-GlcNAc iLTD expression in about 50% of recorded cells.In those speci c cells, a signi cant potentiation of the eIPSC amplitude was observed which is reminiscent of previous results where dynamin inhibition in cultured hippocampal neurons resulted in accumulation of postsynaptic GABA A Rs and increased mIPSC amplitude 42 .The variable effect of dynasore on O-GlcNAc iLTD suggests that the expression mechanism is complex.This is not too surprising since the speci c combination of scaffolding proteins that interact with GABA A Rs varies among inhibitory synapses in speci c brain regions, neuron types, and even within speci c regions within the same neuron 37 .Therefore, it is possible that the precise mechanism underlying O-GlcNAc iLTD may also be variable.Like AMPARs, GABA A Rs are highly dynamic within inhibitory synapses and can rapidly undergo lateral diffusion, causing depression of inhibitory transmission [48][49][50][51] .Whether increasing O-GlcNAcylation alters GABA A R subunit interaction with gephyrin or other scaffolding proteins to stimulate lateral diffusion that underlies expression of O-GlcNAc iLTD is currently unknown and is an area of needed future investigation.

The known interplay, and sometimes competition, between serine O-GlcNAcylation and phosphorylation
on key proteins 15,[52][53][54] led us to further explore how a possible interaction might impact the strength of inhibitory transmission.A notable example of this O-GlcNAcylation-phosphorylation interaction is competition for the same serines on Tau where increasing O-GlcNAc prevents hyperphosphorylation of Tau and development of tangles in Alzheimer's disease 53 .Serine phosphorylation has complex effects on GABA A R function, including impacting how channel function is modulated by benzodiazepines, barbiturates, and neurosteriods 8, 10,12 .An interaction between O-GlcNAcylation and phosphorylation would add to the complexity, and we speci cally focused on PKA dependent phosphorylation using the adenylate cyclase activator, forskolin.As mentioned previously, this strategy has been used to investigate PKA dependent potentiation of excitatory transmission in hippocampus [29][30][31] .While forskolin had a variable effect on the amplitude of the eIPSC during baseline transmission, causing depression in some cells and potentiation in others with no statistically signi cant overall effect, it did not impact the magnitude of subsequently induced O-GlcNAc iLTD.This is consistent with no interaction between a prior increase in PKA dependent phosphorylation with subsequent O-GlcNAc modi cation.However, forskolin unexpectedly reversed the polarity of eIPSC amplitude when applied during expression of O-GlcNAc iLTD, eliciting a signi cant potentiation of the eIPSC amplitude that in some cells even overshot the original baseline.It is important to note that this unexpected potentiation of the eIPSC amplitude occurs at inhibitory synapses in both CA1 and dentate gyrus, suggesting a general mechanism that may not be too dependent on a speci c subunit composition.Moreover, this highly interesting nding suggests the possibility that direct O-GlcNAc modi cation of GABA A R subunits, and/or speci c scaffolding proteins, changes the GABA A R con rmation in a way that unmasks this potentiating effect of forskolin.A further surprise was that this forskolin-dependent eIPSC potentiation was not prevented by pharmacological inhibition of either adenylate cyclase nor PKA, and was mimicked by the adenylate cyclase inactive forskolin analog 1,9 dideoxyforskolin.Thus, this interaction of O-GlcNAcylation and forskolin is not a consequence of PKA mediated phosphorylation.
Clearly, forskolin is working through some other mechanism to potentiate inhibitory transmission following a prior increase in O-GlcNAcylation.Ironically, a previous report demonstrated forskolin-dependent increase in GABA A R desensitization in carp retina through a non-PKA dependent mechanism that involved acting directly on a neurosteroids site 24 .Neurosteroids and metabolites of progesterone positively modulate GABA A Rs in a dose dependent manner by acting on synaptic and extrasynaptic GABA A Rs, and their mechanism of action can be enhanced or diminished depending on the activation or inhibition of serine phosphorylation of GABAARs as well as GABA A R subunit expression 10,12,55 .Furthermore, neurosteroids, such as THDOC, occupy a binding pocket in the transmembrane region that can involve conserved threonines in the α1 and α5 subunits.Importantly, threonines can also undergo O-GlcNAc modi cation similar to serines.
Our data showing that both THDOC and progesterone reversed O-GlcNAc iLTD in recordings from CA1 pyramidal cells, while having no signi cant effect on baseline eIPSC amplitude, precisely mimics the effect of forskolin.These ndings supports the interpretation that forskolin is acting at the neurosteroid site on synaptic GABA A Rs. Perhaps most exciting is the observation that these steroids induced no signi cant effect on baseline eIPSC amplitude, but could only modulate the strength of synaptic inhibition following a prior increase in O-GlcNAcylation.Thus, O-GlcNAc modi cation enables synaptic GABA A Rs to be modulated by neurosteroids and potentiate the eIPSC amplitude thereby reversing the polarity of the iLTD.It is important to point out that not all recorded cells exhibit this reversal in polarity of the eIPSC amplitude, suggesting that there may be a GABA A R subunit con rmation preference and/or subunit combination preference, a concept supported by the varying subunit composition of GABA A Rs across the same cell types and across different locations on the same cell, leading to different responses upon exposure to allosteric modulators 56 .In addition, our ndings indicate that somehow O-GlcNAcylation enhances access to the neurosteroid site on GABA A Rs for both forskolin and allosteric modulators to act.Because GABA A Rs are therapeutic targets for drugs used in the treatment of neurological and neuropsychiatric conditions, understanding how they are modulated by O-GlcNAcylation has clinical implications, particularly if O-GlcNAc interferes with or enhances their e cacy.Future work is needed to determine whether GABA A R subunits and/or scaffolding proteins are directly O-GlcNAc modi ed, and if so, which speci c serines and/or threonines are modi ed.Furthermore, understanding how elevated O-GlcNAc impacts synaptic stability and lateral diffusion within the membrane, and how the neurosteroid site becomes more accessible at synaptic receptors.
Collectively, these current results, together with our previously published results [20][21][22] not only solidi es O-GlcNAcylation as a critical regulator of both synaptic inhibition and excitation, but also provides highly novel information that O-GlcNAc dictates the polarity of the change in GABA A R synaptic current amplitude mediated by endogenous neurosteroids THDOC and progesterone, highlighting O-GlcNAcylation's ability to modify the effectiveness of allosteric modulators on GABAergic transmission.Thus, our current ndings uncovers how protein O-GlcNAcylation can possibly serve as a gauge for the potency of synaptic inhibition and its modulation by allosteric modulators and novel therapeutic agents.

Materials and Methods
All experimental procedures were approved by the Medical University of South Carolina Institutional Animal Care and Use Committee and follow the National Institutes of Health experimental guidelines and is reported in accordance with ARRIVE guidelines.Animals were housed in approved animal care facilities in the Division of Laboratory Animal Research (DLAR) at MUSC under the direction of 3 fulltime veterinarians who are fully accredited by the AAALAC and meet all standards prescribed by the "Guide for the Care and Use of Lab Animals".All animal facilities are administered by DLAR to ensure compliance with federal, state, and standards of care and use of animals.

Electrophysiology
All recordings were performed in a submersion chamber with continuous perfusion of oxygenated standard ACSF.The blind patch technique was used to acquire interleaved whole-cell recordings from CA1 pyramidal neurons and dentate granule cells.Neuronal activity was recorded using an Axopatch 200B ampli er and pClamp10.7 acquisition software (Molecular Devices, Sunnyvale, CA).Signals were ltered at 5 kHz and digitized at 10 kHz (Digidata 1440).Patch pipettes (BF150-110 HP; Sutter Instruments, Novato, CA) were pulled on a Sutter P-97 (Sutter Instruments, Novato, CA) horizontal puller to a resistance between 3-5 MΩ.Spontaneous IPSCs were recorded using CsCl internal solution (in mM: GABA A R currents were pharmacologically isolated with bath perfusion of DNQX (10µM; Hello Bio) and R-CPP (5 µM; Hello Bio).Recordings were performed at a 10 mV test pulse at the end of each sweep to monitor series resistance and was excluded if there was more than a 20% change during the experiment.Stability of series resistance was veri ed using post-hoc scaling of averaged waveforms before and after pharmacologically increasing O-GlcNAcylation and after Forskolin, KT5720, SQ22536, 1,9 Dideoxyforskolin, Jasplakinolide and THDOC exposure.

Chemicals
Forskolin (Hello Bio) and 1,9 Dideoxyforskolin (Sigma-Aldrich) were prepared as a 50 mM stock in DMSO and stock was added to external solution for a nal concentration of 50 µM.KT5720 (Hello Bio) was prepared as a 25 mM stock in DMSO and stock was added to external solution for a nal concentration of 3 µM.SQ22536 (Tocris), 5α,21-pregnane-3α,21-diol-20-one (THDOC) (Sigma-Aldrich, mixed and sonicated), Progesterone (Sigma-Aldrich) was prepared as a 100 mM stock in DMSO and the stock was added to external solution for a nal concentration of 100 µM, 10 µM, and 1 µM respectively.Jasplakinolide (Hello Bio) was prepared as a 1 mM stock in DMSO and stock was added to Cs-gluconate internal solution for a nal concentration of 2 µM.Dynasore (Sigma-Aldrich) was prepared as a 100 mM stock in DMSO and stock was added to external solution for a nal concentration of 80 µM.The actin stabilizer jasplakinolide (2µM, Hello Bio) 57,58 was included in the pipet solution to prevent GABA A R endocytosis.PKA dependent serine phosphorylation was triggered by bath application of the adenylate cyclase activator, forskolin (50µM, Hello Bio) 31 determine if the O-GlcNAc induced synaptic depression is prevented.

Statistical analysis
Recordings were analyzed using Clamp t 11.

Figure 3 O
Figure 3