Long-term potentiation of glycinergic synapses by semi-natural stimulation patterns during tonotopic map refinement

Before the onset of hearing, cochlea-generated patterns of spontaneous spike activity drive the maturation of central auditory circuits. In the glycinergic sound localization pathway from the medial nucleus of the trapezoid body (MNTB) to the lateral superior olive (LSO) this spontaneous activity guides the strengthening and silencing of synapses which underlies tonotopic map refinement. However, the mechanisms by which patterned activity regulates synaptic refinement in the MNTB-LSO pathway are still poorly understood. To address this question, we recorded from LSO neurons in slices from prehearing mice while stimulating MNTB afferents with stimulation patterns that mimicked those present in vivo. We found that these semi-natural stimulation patterns reliably elicited a novel form of long-term potentiation (LTP) of MNTB-LSO synapses. Stimulation patterns that lacked the characteristic high-frequency (200 Hz) component of prehearing spike activity failed to elicit potentiation. LTP was calcium dependent, required the activation of both g-protein coupled GABAB and metabotropic glutamate receptors and involved an increase in postsynaptic glycine receptor-mediated currents. Our results provide a possible mechanism linking spontaneous spike bursts to tonotopic map refinement and further highlight the importance of the co-release of GABA and glutamate from immature glycinergic MNTB terminals.


Results
Semi-natural stimulation patterns are effective in inducing LTP in the MNTB-LSO pathway. Recordings were performed in slices from mice aged postnatal day 5-8, an age when MNTB-LSO synapses are strengthened in vivo 11,12,14,15,29,51 . At this age, the activity of MNTB neurons in vivo reflects the pattern of cochlear generated activity, which consists of periodic bouts of action potentials that occur about 5 times per minute. During these activity bouts, action potentials are clustered together in 'mini bursts' that are separated by about 150 ms and contain 2-4 spikes at around 200 Hz 14,[30][31][32] . To explore the effects of these stereotypical spike patterns on immature MNTB-LSO synapses, we tested whether stimulating MNTB-LSO connections in vitro (Fig. 1A,B) with a pattern that mimicked the in vivo pattern changes the strength of these connections. These semi-natural stimulation patterns (SNPs) consisted of 4 bouts of electrical stimuli, each of which contained 78 stimuli that were grouped in pairs ("mini-bursts", 5 ms inter-stimulus interval) and separated by 150 ms (Fig. 1C). Stimulation of the MNTB-LSO pathways with SNPs elicited a potentiation of MNTB-LSO synapses in 86% of neurons (6 out of 7 neurons). The potentiation was long-lasting and on average increased the peak amplitude of responses to 140.0 ± 7.4% of baseline amplitudes (measured at ~ 23 min post induction, n = 7, p < 0.001, two-tailed paired t-test, Fig. 1D,E).
Synaptic potentiation can result from postsynaptic and/or presynaptic mechanisms, the latter of which involves an increase in neurotransmitter release due to an increase in the probability of release 52,53 . A change in the probability of release is expected to decrease the paired pulse ratio (PPR), which is defined as the ratio of the amplitude of the second response to two stimuli to the amplitude of the first response. In accordance with previous studies at this age 12,47 , the PPR of MNTB-LSO synapses before potentiation was close to 1 (0.94 ± 0.09, n = 7) and most importantly was not changed by synaptic potentiation (0.95 ± 0.07, n = 7, p = 0.949, two-tailed paired t-test, Fig. 1F). This suggests that SNP-induced LTP at MNTB-LSO synapses is primarily due to changes in the postsynaptic neuron.

Tetanic or low-frequency stimulation does not induce LTP in the MNTB-LSO pathway.
To explore the significance of the temporal structure of SNP in inducing LTP, we applied the same number of electrical stimuli and bursts but changed their interspike intervals (ISIs) (Fig. 2). Because tetanic stimulation (100 Hz) reliably elicits LTP in many systems [54][55][56] , we first tested whether delivering the same number of stimuli Scientific Reports | (2020) 10:16899 | https://doi.org/10.1038/s41598-020-73050-y www.nature.com/scientificreports/ within a burst at 100 Hz, while maintaining an inter-burst interval of ~ 4.1 s, could elicit LTP ( Fig. 2A). However, following tetanic burst stimulation, we observed no change in the PSC amplitudes (97.0 ± 7.8% of baseline PSC amplitude n = 9, p = 0.711, two-tailed paired t-test, Fig. 2A). This result indicates that temporal components of the SNP pattern, omitted in the 100 Hz tetanus, are essential for LTP in the MNTB-LSO pathway. We next tested whether a burst pattern which omitted the two distinct ISIs (5 and 150 ms) present in the SNPs can elicit LTP. In this induction protocol we maintained the overall burst duration (~ 5.9), stimulus number (78)  Inter-stimulus intervals of 5 ms are necessary and sufficient to elicit LTP in the MNTB-LSO pathway. To further explore the significance of the within-burst ISIs in LTP induction we tested whether an  Previous studies demonstrated that mice that lack the α9 subunit of nicotinic acetylcholine receptors and as a result, lack functional nicotinic cholinergic transmission to cochlear hair cells (α9 KO mice 57 ) have abnormal patterns of cochlea-generated spontaneous activity before hearing onset and an impaired developmental strengthening of MNTB-LSO connections 14 . To test whether the deficits in synaptic strengthening in α9 KO mice may result from deficits of this altered activity pattern to induce LTP, we stimulated the MNTB-LSO pathway using activity patterns mimicking those previously recorded in vivo in 8-day old α9 KO mice 14 . In particular, α9 KO mice have unchanged average activity levels but mini-bursts during an activity bout occur at shorter intervals, resulting in shorter overall burst duration and prolonged inter-burst intervals. To recapitulate this pattern in vitro, we stimulated the MNTB-LSO pathway with four stimuli bursts, 8.7 s apart, each of which contained 78 stimuli with a 5 ms ISI between the two stimuli of a mini-burst and a 30 ms ISI between mini-bursts (Fig. 2D). These α9 KO-like stimulus protocols lead to a significant increase of the PSC amplitude to an average of 130.1 ± 11.6% (n = 8, p = 0.021, two-tailed paired t-test, Fig. 2D), which was not significantly different from the magnitude of LTP observed after SNP stimulation (140.0 ± 7.4%, p = 0.498 between SNP-and α9-LTP, two-tailed unpaired t-test). This result indicates that the precise inter-burst interval is not crucial for the induction of LTP and further highlights the significance of stimuli with an ISI of 5 ms. In addition, they imply that the refinement deficits in α9 KO-mice are unlikely due to failure of their altered activity patterns to induce LTP in the MNTB-LSO pathway.
LTP induction does not require postsynaptic depolarizations or action potentials. Both in the MSO and LSO, induction of LTP after hearing onset requires coincident postsynaptic depolarizations 48,49 . In our experiments we did not expect that MNTB stimulation elicits significant postsynaptic depolarizations because we used a pipette chloride concentration (10 mM) with an estimated chloride reversal potential of − 66 mV, which is near the resting membrane potential and approximately matches the average native intracellular chloride concentration in LSO neurons at that age [16][17][18]58 . However, under high-frequency stimulation, hyperpolarizing GABAergic responses can become depolarizing due to a breakdown of the transmembrane chloride gradient and the generation of a bicarbonate mediated outward current 59 . This raises the possibility that the high frequency activity that is present in some of our induction stimulation protocols may lead to strong or even suprathreshold responses, which in turn may enable LTP induction. Analysis of synaptic membrane potential responses during induction protocols shows that most LSO neurons generated small, subthreshold depolarizations (1-5 mV from a Vrest of − 68.4 ± 0.7 mV) which could temporally summate to a plateau that lasted throughout the stimulation burst (Fig. 3). In only a small percentage of neurons (5 of 41) did these depolarizations trigger action potentials. Importantly, the direction and magnitude of postsynaptic depolarizations did not significantly differ between stimulus patterns ( Fig. 3F; V m change: SNP: 1.7 ± 0.8 mV, n = 7; 100 Hz: 5.6 ± 2.0 mV, n = 9; 200 Hz: 0.6 ± 2.3 mV, n = 9; low frequency: 2.9 ± 0.7 mV, n = 8; α9 pattern: 3.39 ± 1.92 mV, n = 8; p = 0.363; 1-way ANOVA). In addition, we found no correlation between the direction and magnitude of synaptic responses during induction and the presence or the magnitude of LTP (Fig. 3G). These results indicate that induction of LTP before hearing onset does not depend on postsynaptic membrane depolarizations or action potentials.

Induction of LTP in the MNTB-LSO is calcium-dependent but NMDA receptor-independent.
Essentially all forms of synaptic potentiation described so far require an initial increase in the intracellular Ca 2+ concentration 41,60 . To test whether this also holds true for LTP in the MNTB-LSO pathway, we chelated intracellular Ca 2+ by including the Ca 2+ chelator BAPTA (20 mM) in the recording pipette solution. LSO neurons perfused with BAPTA failed to exhibit LTP in response to the SNP stimulation (92.5 ± 7.5%, n = 6, p = 0.337, twotailed paired t-test, Fig. 4A) indicating that LTP is calcium-depended. To address how MNTB inputs increase postsynaptic calcium concentrations, we next tested whether NMDA receptors, which are responsible for inducing LTP in a large variety of synapses, including glycinergic synapses in the MSO 49 , mediate LTP-inducing calcium influx. In developing MNTB-LSO synapses, NMDA receptors are activated by the co-release of glutamate before hearing onset, and induce a local rise in the dendritic intracellular calcium concentration without accompanying noticeable membrane depolarizations 22,25 . However, these dendritic NMDA receptor-mediated calcium responses are not necessary for induction of LTP because blocking NMDA receptors by bath application of the NMDA receptor antagonist AP-5 (100 µM) had no effect on LTP (Fig. 4B). The magnitude of LTP by SNP stimulation during application of AP-5 (122.8 ± 8.7% of baseline PSC amplitudes, n = 11, p = 0.016, two-tailed paired t-test, Fig. 4B LTP at MNTB-LSO synapses requires activation of g-proteins by mGlu and GAGA B receptors acting cooperatively. Having ruled out the requirement of NMDA receptor-mediated Ca 2+ influx for the induction of LTP, we next tested the contribution of g-protein intracellular signaling that can lead to the release of Ca 2+ from intracellular stores. In a first step, we blocked all g-protein-dependent signalling by substituting GTP in the recording electrode solution with the broad-spectrum g-protein blocker GDP-β-s (2 mM) 61,62 . In the presence of GDP-β-s we observed no significant change in PSC amplitudes in response to SNP stimulation (96.5 ± 7.0% of baseline PSC amplitude, n = 7, p = 0.622, two-tailed paired t-test, Fig. 5A), indicating that LTP depends on the postsynaptic activation of g-protein coupled receptors. Due to their co-release of glutamate and    21,28 and that GABA B receptor activation is necessary for the induction of LTP. In summary, our results establish the necessity of a cooperative activity of GABA B and mGluR-dependent g-protein-mediated signaling in inducing glycinergic LTP at MNTB-LSO synapses.

LTP involves an increase in glycine receptor-mediated responses. LTP did not change PPR and
therefore most likely is expressed by postsynaptic changes (Fig. 1D). Because before hearing onset MNTB-LSO synapses co-release glycine, GABA and glutamate, LTP could involve an increase in currents mediate by glycine, GABA A , or AMPA receptors. To examine these possibilities, we isolated glycine receptor-mediated responses by bath application of the AMPA receptor antagonist CNQX (20 µM) and the GABA A receptor antagonist SR 95,531 (20 µM). Due to the voltage-dependent block of NMDA receptors at our negative holding potentials (V hold − 84 mV) during test responses and due to the very small amplitude of NMDA receptor mediated currents with somatic recordings 22 , NMDARs are highly unlikely to contribute significantly to EPSCs under these conditions. SNP stimulation under these conditions induced LTP (PSC amplitude 120.3 ± 6.4% of baseline, n = 10, p = 0.005, two-tailed paired t-test, Fig. 6), with a magnitude that was not significantly different from that induced in control conditions (p = 0.065, two-tailed unpaired t-test, compare Fig. 6 with Fig. 1B,C). In a separate group of neurons, we also tested the response to SNP stimulation in the presence of CNQX alone. In these neurons, we also observed LTP as indicated by an increase of responses to 135.4 ± 7.8% of baseline (n = 5, p = 0.002, paired t-test), which was not significantly different to the magnitude of LTP in control experiments (p = 0.681, twotailed unpaired t-test). Taken together, these results indicate that LTP in the MNTB-LSO pathway potentiates glycine receptor-mediated currents. Due to the very small amplitude of isolated GABA A receptor-mediated currents, we were unable to reliably evaluate their contribution to the overall magnitude of potentiation, leaving open the possibility that the potentiation of GABA A receptor-mediated responses contribute to LTP to a minor degree.

Discussion
In this study we demonstrate that semi-natural stimulation patterns that mimic the temporal structure of spontaneous activity in vivo before hearing onset elicit a novel form of inhibitory glycinergic LTP in the developing MNTB-LSO pathway in vitro. The induction of this LTP did not require postsynaptic membrane depolarization or action potentials, but depended on an increase in the postsynaptic Ca 2+ concentration and the activation of g-proteins through a cooperative action of mGluR and GABA B receptors. LTP seems to be expressed postsynaptically as it was not associated with a change in the PPR. The properties of this inhibitory LTP make it a plausible mechanism for activity-dependent strengthening of MNTB-LSO connections and tonotopic refinement before the onset of sound driven activity.

Semi-natural temporal patterns elicit LTP.
Before hearing onset, the topographic precision of the MNTB-LSO pathway is increased by the silencing of most initial connections and the strengthening of maintained ones [11][12][13]15 . This synaptic remodeling critically depends on patterned spontaneous spike activity that originates in immature cochlear hair cells from which it is faithfully transmitted along the ascending auditory   14,30,32 . In this study we demonstrate that stimulation of the MNTB-LSO pathway with a temporally structured stimulation pattern that reproduces the main temporal features of the spontaneous activity pattern in vivo, reliably elicits LTP (Fig. 1), which could be a plausible synaptic mechanism by which patterned activity contributes to the maturation and tonotopic refinement of the MNTB-LSO pathway. Although the magnitude of LTP that we observed in our acute experiments is far from the magnitude of the developmental increase in the strength of MNTB-LSO connection occurring in vivo, in many other systems LTP can lead to long-lasting functional and structural changes, which include the insertion of postsynaptic receptors [68][69][70] and the formation of new synapses [71][72][73] . It is thus at least conceivable that such structural changes may also be triggered by glycinergic LTP and lead to the increase in the number of release sites and quantal amplitudes that underlie the 8-12 fold strengthening of single MNTB-LSO connections before hearing onset 12 .
Our results also reveal that stimulation at 200 Hz, the frequency of action potentials during mini-bursts in vivo 14,32 , is necessary and sufficient to induce LTP. 200 Hz stimuli presented as mini-bursts (Figs. 1, 2D) or as a continuous train (2C), induced LTP, while stimulation patterns that omitted stimulus intervals of 5 ms failed to induce LTP ( Fig. 2A,B). Interestingly, 200 Hz stimulation is also highly effective in inducing LTP in MNTB axon collateral synapses on MSO neurons after hearing onset 49 . A sensitivity to 200 Hz (theta burst) stimulation was also reported for GABA-LTP in adult CA1 pyramidal cells, a form of inhibitory LTP that depends on the coactivation of GABA B and mGluR receptors 61 by neurotransmitter spillover [74][75][76] . It is likely that neurotransmitter spillover also underlies the activation of GABA B and mGluRs for triggering LTP of MNTB-LSO synapses. In support of this, previous studies of immature MNTB-LSO synapses demonstrated GABA spillover to extrasynaptic GABA A receptors 28 and glutamate spillover to extrasynaptic NMDARs 77 . Along the same lines, high-frequency, but not low-frequency stimulation, of excitatory inputs elicits mGluR-mediated calcium responses in immature LSO neurons, consistent with the idea of extrasynaptic glutamate spillover to mGluRs 65 .

Synaptic mechanisms mediating MNTB-LSO LTP. Inhibitory synapses across the central nervous
systems can express a wide variety of activity-dependent plasticity with diverse stimulation requirements and mediated by distinct induction and expression mechanisms 45,[78][79][80][81] . For glycinergic synapses formed by MNTB axons, previous studies demonstrated two distinct forms of age-dependent LTP in the MSO 49 and LSO 48 , which occur after the onset of hearing. Similar to LTP reported in this study, these forms of inhibitory LTP are expressed postsynaptically and involve an increase in GlyR mediated currents. However, they differ from LTP before hearing onset in respect to their stimulation requirements as well as their induction and expression mechanisms. LTP in the MSO of gerbils after hearing depends on the activation of dendritic NMDA receptors and LTP at MNTB-LSO synapses after hearing onset is induced by low-frequency stimulation combined with glutamate-mediated postsynaptic depolarizations and GABA B receptor activation 48 . Thus, after hearing onset, LTP in MNTB pathways requires coincident activity of MNTB inputs with glutamatergic inputs from the cochlea nucleus, which in vivo would occur under acoustic stimulation. In contrast, LTP in the LSO before hearing onset, despite also being dependent on glutamate release, can occur independently from cochlear nucleus inputs due to the transient co-release of glutamate from MNTB terminals that is present during this age 22 . Thus, a possible developmental function of co-release of glutamate may be to enable LTP in the MNTB-LSO pathway at an early age when cochlear-elicited activity between both sides is not yet coordinated by sound 50 , yet when the MNTB-LSO pathway is remodeled by spontaneous activity. Interestingly, mice that have a genetic loss of the vesicular glutamate transporter 3 and, as a consequence, lack glutamate release from inner hair cells and MNTB terminals 24,82 , exhibit an impaired strengthening of MNTB-LSO connection before hearing onset. This impaired strengthening may reflect deficits in LTP due to the loss of mGluR activation by MNTB terminals and/or may be due to the absence of 5 ms inter-spike intervals in the spontaneous auditory nerve activity in these mice before hearing onset 83 . Further investigations of LTP or LTD in mouse models lacking glutamate release or having altered spontaneous activity patterns can address these open questions.
In addition to the activation of mGluR receptors, LTP at MNTB-LSO synapses requires the activation of postsynaptic GABA B receptors (Fig. 7), a feature it shares with all other known forms of activity-dependent long-term www.nature.com/scientificreports/ plasticity at these synapses 48,67 and with LTP of GABAergic synapses in the cortex 78 and hippocampus 61,84 .
Although postsynaptic GABA B receptors are best known for their g-protein-mediated activation of potassium channels (GIRK channels), GABA B receptors also have been linked to important signaling pathways that are tied to inducing synaptic plasticity 85 . For example, in cortical, cerebellar and hippocampal neurons, activation of GABA B receptors can lead to the activation of phospholipase C (PLC), the generation of IP3, and the release of Ca 2+ from intracellular stores 78,[86][87][88] . PLC activation and an increase in intracellular Ca 2+ concentration has been shown to be necessary for GABA B receptor-dependent GABA-LTP in cortical neurons 78 90 . In the developing hippocampus, GABA B receptor activation can also lead to a Ca 2+ -dependent release of BDNF, which in turn promotes the membrane expression of GABA A receptors 91,92 . Because BDNF is also expressed in the developing LSO 93 and is necessary for inducing GABA B receptor-mediated LTD of MNTB-LSO synapses 67 , it may also participate in the induction of LTP. Finally, another possibility is that GABA B receptors contribute to LTP induction by augmenting mGluR-mediated signaling 94,95 . In this scenario, one would expect that stronger activation of mGluRs, perhaps by the coincident activation of glutamatergic synapses that occurs after hearing onset, would make LTP GABA B independent, increase its magnitude, or widen the spectrum of stimulation patterns that can induce it. Although the exact signaling cascades that link GABA B receptors to LTP remain to be determined, the fact that both LTP and LTD of MNTB-LSO synapses critically depend on GABA signaling points to an important role of GABA co-release at these developing glycinergic synapses.
Expression site of LTP. The expression of LTP was not accompanied by a change in the PPRs suggesting an unchanged probability of presynaptic vesicle release, which argues in favor of a postsynaptic expression of LTP. LTP resulted in an increase in pharmacologically isolated GlyR-mediated currents (Fig. 6). Due to the small amplitude of pharmacologically isolated GABA-mediated responses at this age and our experimental conditions, we were unable to reliably determine whether synaptic GABA responses also increased during LTP and thus cannot rule out a change in GABA A receptor mediated currents, as is the case during LTD at these synapses 47 . However, because of the small amplitude of GABAA receptor-mediated currents and the fact that the amplitude of LTP of isolated GlyR mediated responses was similar to the amplitude of LTP under control conditions, any contribution of GABA A receptor-mediated currents to the overall LTP would be minor. An increase in glycine receptor-mediated responses could result from an increase in the number of postsynaptic receptors and/or an increase in their open probability or single-channel conductance (Fig. 7). A rise in intracellular Ca 2+ concentration and activation of calcium-/calmodulin-dependent protein kinase II (CaMKII) increases synaptic clustering of glycine receptors in Mauthner cells in fish 96 , which can result in larger quantal sizes of synaptic glycinergic currents 97 . Protein kinases regulate the strength of glycinergic synapses via phosphorylation of the glycine recep-   [98][99][100][101][102] or via glycine receptor anchoring protein gephyrin 99,[103][104][105] . Due to the fast onset of LTP, modifications of existing synaptic glycine receptors likely underlies at least the early phases of LTP while gephyrinmediated receptor clustering could play a more important role in later phases of LTP and the developmental increase of functional release sites in vivo 12 .

Methods
Animals and slice preparation. Mice of the strain C57BL/6J (Charles River) of either sex were used between postnatal days (P) 5-8. All animal procedures were performed in accordance with NIH guidelines and were approved by the IACUC of the University of Pittsburgh. Mice were deeply anaesthetized with isoflurane, decapitated and brains were removed and transferred to ACSF (composition in mM: NaCl 124, NaHCO 3 26, glucose 10, KCl 5, KH 2 PO 4 1.25, MgSO 4 1.3, CaCl 2 2, pH = 7.4, aerated with 95% O 2 /5% CO 2 ). Coronal brain slices (300 µM thick), were cut using a vibrating microtome as previously described 23 . Slices containing both the LSO and MNTB were incubated for 1 h at 32 °C in an interface-style chamber and subsequently transferred to a recoding chamber where they were submerged and continuously perfused with aerated ACSF at room temperature throughout the recording session.
Electrophysiology. Whole-cell patch clamp recording were made from borosilicate glass pipettes (tip resistances of 3-6 mΩ) and filled with internal solution containing in mM: 140 K-gluconate, 10 NaCl, 10 Hepes, 0.6 EGTA, 2 MgATP, 0.3 NaGTP and 2.5-10 phosphocreatine-Tris. When indicated, NaGTP was replaced by 2 GDP-β-s. If the internal solution contained 20 mM BAPTA, EGTA was omitted, K-gluconate lowered to 60 mM and 60 mM sucrose added. All internal solutions were pH adjusted using KOH to a final pH of 7.2-7.4 with an osmolarity of 280-290 mOsm. The MNTB-LSO pathway was electrically stimulated (pulse duration 0.2 ms) using an ACSF filled pipetted (resistance ~ 1-3 mΩ), which was placed at the lateral edge of the MNTB. Paired-pulse stimuli with an interstimulus interval (ISI) of 50 ms were delivered at 0.1 Hz to elicit postsynaptic currents (PSCs). Stimulus strength was adjusted to evoke maximal responses 11 . Synaptic responses were recorded in voltage clamp (V clamp ) at a holding potential of − 84 mV. This potential was chosen to increase the chloride driving force (calculated E cl -65.5 mV) and to minimize the activation of NMDA receptors by the co-release of glutamate from MNTB terminal 22,25 .
At least 20 successive stimuli were acquired during baseline (up to 15 min) after which recording were switched to current clamp (I = 0) and one of five induction stimulus protocols was delivered. Consecutive iterations (4 times) of induction protocols were triggered manually. They are described in the results and illustrated schematically in Figs. 1 and 2. The average series resistance (r s ) was 14.9 ± 0.4 MΩ (n = 98) and was not compensated. Recordings with r s > 25 mΩ or if r s changed by more than 20% were excluded from analysis. Only cells with recordings ≥ 23 min post-induction stimulation were included in analysis.

Statistical analysis.
To assess the expression of LTP in individual neurons in response to stimulation paradigm or pharmacological treatment a minimum of 20 pre-stimulation measures (baseline amplitudes or baseline paired-pulse ratios) were compared to a minimum of 10 consecutive post-stimulation measures. Consecutive post-stimulation measures were taken starting at 2 min after the induction stimulation and every 7 min thereafter. Within-neuron pre-and post-stimulation measures were compared using 1-way ANOVA with Dunnett's multiple comparisons test. Neurons that maintained a significantly elevated amplitude at ~ 23 min post-stimulation (and during subsequent timepoints when obtained) were considered to express LTP.
For comparing response magnitudes within a group of cells amplitudes were normalized to the average prestimulation amplitude. Two-tailed paired t-test was used to compare average pre-stimulation responses (baseline) to the average of a minimum of 10 responses at ~ 23 min (23:59 ± 00:05 min) post induction. The means of normalized responses between two groups were compared using a two-tailed unpaired t-test. When comparing the means between more than 2 groups a 1-way ANOVA was performed. The normality of measures within a group was established using a Shapiro-Wilk test. One group (200 Hz Tetanic stimulus) was not normally distributed but was tested using a parametric test to maintain consistency across studies and to take a more conservative statistical approach. Testing the same dataset with a non-parametric test essentially yielded the same results. All results were considered statistically significant with a p < 0.05. Errors are reported as ± SEM.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.