Somatostatin contributes to long-term potentiation at excitatory synapses onto hippocampal somatostatinergic interneurons

Somatostatin-expressing interneurons (SOM-INs) are a major subpopulation of GABAergic cells in CA1 hippocampus that receive excitation from pyramidal cells (PCs), and, in turn, provide feedback inhibition onto PC dendrites. Excitatory synapses onto SOM-INs show a Hebbian long-term potentiation (LTP) mediated by type 1a metabotropic glutamate receptors (mGluR1a) that is implicated in hippocampus-dependent learning. The neuropeptide somatostatin (SST) is also critical for hippocampal long-term synaptic plasticity, as well as learning and memory. SST effects on hippocampal PCs are well documented, but its actions on inhibitory interneurons remain largely undetermined. In the present work, we investigate the involvement of SST in long-term potentiation of CA1 SOM-IN excitatory synapses using pharmacological approaches targeting the somatostatinergic system and whole cell recordings in slices from transgenic mice expressing eYFP in SOM-INs. We report that application of exogenous SST14 induces long-term potentiation of excitatory postsynaptic potentials in SOM-INs via somatostatin type 1–5 receptors (SST1-5Rs) but does not affect synapses of PC or parvalbumin-expressing interneurons. Hebbian LTP in SOM-INs was prevented by inhibition of SSTRs and by depletion of SST by cysteamine treatment, suggesting a critical role of endogenous SST in LTP. LTP of SOM-IN excitatory synapses induced by SST14 was independent of NMDAR and mGluR1a, activity-dependent, and prevented by blocking GABAA receptor function. Our results indicate that endogenous SST may contribute to Hebbian LTP at excitatory synapses of SOM-INs by controlling GABAA inhibition, uncovering a novel role for SST in regulating long-term synaptic plasticity in somatostatinergic cells that may be important for hippocampus-dependent memory processes.

A notable feature of CA1 SOM-INs is the long-term plasticity occurring at their excitatory synapses. These synapses show a Hebbian long-term potentiation (LTP) mediated by type 1a metabotropic glutamate receptors (mGluR1a) [8,12,13]. Excitatory synapses onto parvalbumin-expressing interneurons (PV-INs) do not display this form of long-term plasticity [8]. In addition, SOM-INs excitatory synapses show a late form of mGluR1a-dependent LTP, that can last from a few to 24 h and involves mammalian target of rapamycin complex 1 (mTORC1) mediated translation [9,14,15]. Interestingly, cell-specific conditional down-regulation of mTORC1 in SOM-INs impairs late mGluR1a-dependent LTP, as well as contextual fear and spatial memory consolidation [9]. Conversely, conditional up-regulation of mTORC1 activity in SOM-INs facilitates late mGluR1adependent LTP, as well as hippocampal-dependent memory [9]. Contextual fear learning induces mGluR1aand mTORC1-dependent LTP at SOM-IN excitatory synapses, suggesting a critical implication of SOM-IN long-term synaptic plasticity in hippocampal learning and memory [9]. More recently, cell-specific conditional knock-in of the non-phosphorylatable translation initiation factor eIF2α (eIF2α S51A ) in SOM interneurons was found to upregulate general mRNA translation in these cells and be sufficient to gate CA1 network plasticity and increase long-term contextual fear memory, further supporting a critical role of SOM-INs in hippocampal longterm memory consolidation [11].
Somatostatin (SST; also known as somatotropinrelease inhibitory factor, SRIF) is a peptide expressed in central nervous system. It was first discovered in the hypothalamus where it exerts an inhibitory action on growth hormone [16]. SST is implicated in multiple brain functions like olfaction, vision, cognition and locomotion, as well as in pathologies such as Alzheimer's disease, schizophrenia and major chronic depression [17]. SST acts via five metabotropic receptors (SST 1 R to SST 5 R) that are coupled to G proteins and target many effectors [18]. SSTRs have a wide distribution in brain with overlapping regional localization of receptor types [19]. In the hippocampus, mRNA for all five SSTRs is present, although expression is weaker for SST 5 R [20,21]. Subcellular localization of SSTRs is highly specific to the receptor type. SST 1 R is targeted pre-synaptically to axons, while SST 2,4,5 R are mostly distributed post-synaptically to neuronal somata and dendrites, and SST 3 R appears excluded from classic pre-and post-synaptic sites [19]. Consistent with the subcellular localization of its receptors, SST modulates neuronal activity via both pre-and post-synaptic mechanisms [17,18]. In the hippocampus, exogenous SST 14 hyperpolarizes PCs by activation of two distinct K + currents (M-current and voltage-insensitive leak current) [22][23][24]. SST 14 also induces a presynaptic inhibition of excitatory synaptic transmission in hippocampal PCs [25]. The presynaptic inhibition may involve a G-protein mediated inhibition of N-type voltage-gated Ca 2+ channels [26,27] and activation of presynaptic K + channels [25]. Although SST 14 presynaptic inhibition of hippocampal excitatory synaptic transmission is well documented, presynaptic inhibition of GABAergic inhibitory synaptic transmission has also been reported [28] as in other brain regions [29].
SST is critical for hippocampal long-term synaptic plasticity, as well as learning and memory. Depletion of SST by cysteamine treatment, or knock-out of the SST gene in transgenic mice, impairs contextual fear memory but not auditory fear learning [30]. The memory impairment is associated with a decrease in LTP in CA1 PCs [30]. Interestingly, blocking LTP at excitatory synapses of SOM-INs was found to impair contextual fear memory and facilitation of LTP in PCs by SOM-INs [9]. The analogous effects of manipulating SST or SOM-IN synaptic plasticity on contextual fear memory and PC synaptic plasticity, suggest a possible link between SST and longterm plasticity at SOM-IN excitatory synapses.
Here, we investigate the involvement of SST in longterm potentiation of CA1 SOM-IN excitatory synapses using pharmacological approaches targeting the somatostatinergic system and whole cell recordings in slices from transgenic mice expressing eYFP in SOM-INs. We report that application of exogenous SST 14 induces longterm potentiation of excitatory postsynaptic potentials (EPSPs) of SOM-INs via SST 1-5 Rs, but not of PC and PV-IN synapses. Also, Hebbian LTP in SOM-INs was prevented by inhibition of SSTRs and depletion of SST by cysteamine treatment, suggesting a critical role of endogenous SST in LTP. LTP of SOM-IN synapses induced by SST 14 was independent of NMDAR and mGluR1a, activity-dependent, and prevented by blocking GABA A receptor function. Our results indicate that endogenous SST may contribute to Hebbian LTP at excitatory synapses of SOM-INs by controlling GABA A inhibition, uncovering a novel role for SST in regulating long-term synaptic plasticity in somatostatinergic cells that may be important for hippocampus-dependent memory processes.

Results
The excitatory synapses onto CA1 SOM-INs show longterm plasticity [8,9]. Here we investigate if the peptide SST 14 , that is expressed specifically in SOM-INs, is involved in long-term plasticity of their excitatory synapses, using whole cell recordings in acute slices from SOM-eYFP mice (Sst ires−Cre ;Rosa26 lsl−EYFP ) that express eYFP in SOM-INs (Fig. 1a). Statistical results are summarized in the supplemental statistical table (Additional file 1).

SST 14 induces LTP via SSTRs
We examined with current clamp recordings the effects of application of exogenous SST 14 on EPSPs evoked in eYFP-expressing SOM-INs in CA1 stratum oriens. Bath application of 5 µM SST 14 for 5 min induced a gradual slow onset potentiation of EPSP amplitude that developed over 10-35 min after SST 14 application (201.9 ± 28.1% of control at 25-30 min; 185.4 ± 23.4% of control at 30-35 min; Fig. 1b, e, f ). Similar vehicle application did not induce change in EPSPs over the same time period, ruling out non-specific effects due to recording conditions (96.8 ± 9.4% of control at 25-30 min; 106.1 ± 13.4% of control at 30-35 min; Fig. 1c, e, f ). These results indicate that exogenous SST 14 induces long-term potentiation of EPSPs in SOM-INs.
Bath application of SST 14 affected EPSPs but had no effect on cell input resistance in the same cells. Cell input resistance was unchanged during and after SST 14 or vehicle application (96.1 ± 13.3% of control during and 95.3 ± 9.1% of control at 30-35 min after SST 14 ; 97.1 ± 8.5% of control during and 103.8 ± 8.0% of control at 30-35 min after vehicle; Fig. 1g). Thus, SST 14 effects on EPSPs may not involve postsynaptic changes in cell input resistance.

SST 1-5 R antagonist or cysteamine treatment prevent Hebbian LTP
Excitatory synapses onto CA1 SOM-INs show a mGluR1a-dependent Hebbian LTP [8,12]. Since SST 14 induces LTP of EPSPs in SOM-INs via SSTRs, next we examined if endogenous SST could be involved in Hebbian LTP.
We used whole cell current clamp recordings to monitor Hebbian LTP elicited in SOM-INs by theta burst stimulation (TBS), and bath application of the SST 1-5 R antagonist cyclosomatostatin to test for a possible role of SST (Fig. 2). Cyclosomatostatin or its vehicle (DMSO) were applied 10 min prior to and during TBS, and then washed out ( Fig. 2b- Fig. 2d, e). These results suggest that endogenous SST may be released and activate SSTRs in Hebbian LTP induced by TBS in SOM-INs.
To investigate the implication of endogenous SST in Hebbian LTP by a different approach, we used cysteamine, a compound that depletes SST levels in brain and other tissues [32,33]. Mice received an intraperitoneal (IP) injection of cysteamine (150 mg/kg) and hippocampal slices were harvested 4 h later. The effect of cysteamine injection on SST levels in hippocampus was verified with immunofluorescence. SST immunofluorescence in CA1 SOM-INs expressing eYFP was decreased in cysteamine injected mice relative to vehicle injected mice (58.3 ± 9.7% of control; Fig. 3a), suggesting an effective lowering of SST levels in SOM-INs. Next, we examined Hebbian LTP in SOM-INs in slices of mice after vehicle or cysteamine IP injection. In vehicle injected mice, TBS elicited long-term potentiation of EPSP amplitude (145.8 ± 12.9% of control at 15-20 min, 147.5 ± 11.8% of control at 20-25 min and 160.7 ± 9.6% of control at 25-30 min) (Fig. 3b,  To rule out extra-hippocampal effects of cysteamine IP injection, hippocampal slices from untreated mice were incubated in ACSF containing 200 µM cysteamine for 1 h. In cysteamine-treated slices, SST immunofluorescence was decreased in CA1 SOM-INs compared to vehicle-treated slices (76.9 ± 6.1% of control; Fig. 3d). Taken together, the results of these experiments with cyclostomatostatine and cysteamine suggest that TBS may lead to the release of endogenous SST and the activation SSTRs in Hebbian LTP in SOM-INs.

SST 14 induced potentiation is independent of NMDAR and mGluR1a
NMDA receptors (NMDAR) and metabotropic glutamate 1a receptors (mGluR1a) are involved in synaptic plasticity in hippocampal interneurons [12,34]. We examined a possible implication of these receptors in SST 14 -induced LTP in SOM-INs. First, we examined if NMDARs were implicated by including the antagonist DL-APV (50 µM) in the ACSF for the duration of the experiment. Bath application of SST 14 in presence of DL-APV, induced a gradual LTP of EPSPs (183.9 ± 30.5% of control at 30-35 min) that was similar to the LTP elicited by SST 14 without DL-APV (183.8 ± 17.8% of control at 30-35 min) (Fig. 4a, b). Thus, SST 14  Therefore, SST 14 -induced LTP does not involve mGluR1a.
In the above series of experiments with DL-APV and LY367385, we used paired stimulation to measure the paired-pulse ratio of EPSPs in cells that received SST 14 in the absence of antagonists. Paired-pulse facilitation was similar before (-5 to 0 min; 2.116 ± 0.113) and after (30 to 35 min; 2.254 ± 0.184) SST 14 application (Fig. 4e). These results suggest an absence of presynaptic changes during SST 14 -induced LTP.
To shed further light on the mechanisms involved in the SST 14 -induced LTP, we examined if LTP is dependent on synaptic activity during SST 14 application. EPSPs were recorded during a 5 min baseline period and stimulation was interrupted. SST 14 was applied for 5 min and washed-out for another 5 min, without stimulation. Stimulation was resumed and EPSPs recorded for 30 min (Fig. 5a-c). Application of vehicle (127.2 ± 16.7% of control at 25-30 min) or SST 14   control) (Fig. 5c, d). These results suggest that, although interruption of stimulation induces a rebound potentiation in both groups, SST 14 -induced LTP of EPSPs may require synaptic stimulation in the presence of the peptide.
The mGluR1a-dependent Hebbian LTP found at excitatory synapses onto SOM-INs is not observed at synapses onto PV-INs [8]. Therefore, we examined if application of SST 14 affects EPSPs of PV-INs using a similar recording paradigm from PV-INs of PV-eYFP mice (Fig. 6a). After   Fig. 6b-d). These results suggest that, under our recording conditions, EPSPs of PV-INs show spontaneous run-up over time. Moreover, application of SST 14 produced similar results as vehicle (Fig. 6h), suggesting that SST 14 Fig. 6e-g, i), indicating a lack of SST 14 effect on PC EPSPs. These results suggest that SST 14 -induced LTP of EPSPs in SOM-INs may not arise from a di-synaptic effect via CA1 PC excitatory synapses.

SST actions on SOM-IN excitatory synapses mediated by GABA A inhibition
To gain more insight on the mechanisms of SST 14 -induced LTP, we investigated the effects of SST 14 on putative single fiber excitatory postsynaptic currents (EPSCs) evoked by minimal stimulation [12,36]. Non-NMDAR-mediated EPSCs were recorded in whole cell voltage clamp mode from SOM-INs in the presence of NMDA and GABA A receptor antagonists, DL-APV and gabazine respectively (Fig. 7a-c). EPSC amplitude was not affected by bath application of 5 µM SST 14 (77.8 ± 13.3% of control at 30-35 min) or vehicle (99.9 ± 40.3% of control at 30-35 min) (Fig. 7c-g), suggesting a lack of lasting effect of SST 14 on excitatory postsynaptic currents in these conditions. Given this lack of effect of SST 14 on EPSCs, we examined if the SST 1-5 R antagonist cyclosomatostatin affected TBS-induced LTP of EPSCs evoked by minimal stimulation. TBS given in the presence of DMSO elicited a slow onset LTP of EPSC amplitude (153.6 ± 20.9% of control at 30-35 min; Fig. 7d, f, g). Application of TBS in the presence of cyclosomatostatin produced LTP of EPSC amplitude that was not different from LTP elicited in DMSO (184.1 ± 17.6% of control at 30-35 min; Fig. 7e, f,  g). These results indicate that blocking SST 1-5 Rs did not affect LTP of pharmacologically isolated non-NMDARmediated EPSCs.
As the non-NMDAR-mediated EPSCs in the previous experiments were pharmacologically isolated in the presence of the GABA A receptor antagonist gabazine and in slices with CA3-CA1 surgical cuts, we examined if the long-lasting effects of SST 14 on EPSPs was due to an indirect action via GABA A receptors. We tested the effects of SST 14 on EPSPs of SOM-INs in normal slices, slices with a CA3-CA1 cut, or slices with a CA3-CA1 cut and gabazine (Fig. 8). Application of SST 14  receptors blocked the SST 14 -induced potentiation of EPSPs. Thus, SST actions in LTP of excitatory synapses of SOM-INs may be indirectly mediated via GABA A inhibition (Fig. 8 f ).

Discussion
In the present work, we uncover a critical involvement of the peptide SST in long-term potentiation at excitatory synapses of hippocampal SOM-INs. We found that application of exogenous SST 14 induces long-term potentiation of EPSPs in SOM-INs via somatostatin type 1-5 receptors (SST 1-5 Rs) (Fig. 1). Application of SST 14 did not affect EPSPs in PCs or parvalbuminexpressing interneurons (Fig. 6). TBS-induced Hebbian LTP of EPSPs in SOM-INs was prevented by inhibition of SST 1-5 Rs (Fig. 2) and by depletion of hippocampal SST by cysteamine treatment (Fig. 3), suggesting a significant role of endogenous SST in LTP. LTP of SOM-IN EPSPs induced by SST 14 did not involve changes in paired-pulse ratio of synaptic responses (Fig. 4), was independent of NMDAR and mGluR1a (Fig. 4) and was dependent on concomitant synaptic activity (Fig. 5). Importantly, we observed that SST 14 did not affect non-NMDAR-mediated EPSCs recorded during GABA A receptor blockade, and that the SST 1-5 R antagonist cyclosomatostatin did not affect TBS-induced LTP of these EPSCs (Fig. 7). Finally, pharmacological block GABA A receptor function prevented SST 14 -induced potentiation of EPSPs (Fig. 8

SST 14 -induced LTP of SOM-INs excitatory synapses
The observed long-lasting potentiation of EPSPs in SOM-INs by SST 14 contrasts with the previously described SST actions in hippocampal PCs. Bath-applied SST 14 was reported to decrease evoked and spontaneous EPSCs in pyramidal cells [25]. These effects were acute, occurring within 2-4 min of application onset, and rapidly (2-4 min) reversible. Consistent with these previous observations, we did not find long-lasting effects of SST 14 on EPSPs in PCs, suggesting that SST 14 long-term effects on SOM-IN excitatory synapses do not arise from disynaptic actions at PC afferent synapses, but are specific to synapses onto interneurons. Moreover, SST 14 did not affect excitatory synaptic responses of PV-INs, indicating that SST 14 long-term effects may be specific to excitatory synapses onto somatostatinergic cells. However, in experiments with recordings from PV-INs, the same basal recording and stimulating conditions that elicited stable synaptic responses in SOM-INs and pyramidal cells, resulted in a spontaneous run-up over time. Previous work has shown that excitatory synapses of PV-INs display an anti-Hebbian form of LTP mediated by Ca 2+ -permeable AMPARs [34]. Since in our recording conditions we maintained the postsynaptic PV-IN at hyperpolarized level during stimulation, the run-up over time may have been caused by anti-Hebbian plasticity. But anti-Hebbian plasticity is independent of GABA A inhibition [34], and, thus, unlikely to occlude SST 14 actions via GABA A inhibition. Thus, it would be important to re-examine the SST 14 effects on PV-INs excitatory synapses using different recording/stimulation conditions that elicit stable basal responses, to rule out a possible occlusion of SST 14 effects by the response run-up over time. Hippocampal PCs are also hyperpolarized by bath application of SST 14 via activation of postsynaptic K + conductances [22][23][24]. These are also acute effects, reversible in minutes, that are unlikely to contribute to the slow onset and long-lasting potentiation of EPSPs in SOM-INs. Furthermore, SST 14 -induced hyperpolarization of PCs reduces action potential firing [22], and thus would be expected to reduce presynaptic activation of PCs and decrease EPSPs in SOM-INs. Although SST 14 has been reported to depolarize and excite PCs [35], these effects are produced by local application of SST 14 and are not observed with bath application [24]. Thus, these direct membrane effects of SST 14 on PCs are distinct from the long-lasting actions of SST 14 on SOM-IN excitatory synapses observed here.

SST 14 -induced potentiation is mediated by GABA A inhibition
SST 14 -induced potentiation of EPSPs was prevented by the GABA A antagonist gabazine, indicating that SST 14 actions are mediated indirectly via GABA A inhibition. The mechanism by which SSTR activation acts via GABA A inhibition to increase synaptic excitation remains to be clarified. However, SSTRs are coupled to presynaptic inhibition via inhibition of voltage-gated Ca 2+ channels and activation of K + currents [17,18]. Thus, via such mechanisms, activation of SSTRs could inhibit release from GABAergic terminals, blocking GABA A inhibition at the PC synapse onto SOM-IN, resulting in potentiation at this synapse via disinhibition. Inhibition of GABAergic synaptic transmission in PC was reported by local application of SST 14 [28]. However, bath application of SST 14 was reported to inhibit selectively synaptic excitation without affecting inhibition in PCs [25]. SST 14 actions in hippocampal interneurons also appear complex with report of depolarization and hyperpolarization [35]. In other brain regions, SST was found to decrease GABA release [17].
Whether SST actions are mediated via GABA A inhibition acting pre-or post-synaptically at the PC to SOM-IN synapse remains unclear. Paired-pulse ratio was unchanged during and after SST 14 application (Fig. 4), suggesting no presynaptic changes in transmitter release. However, the excitatory synapses from PC to SOM-INs are composed of calcium-permeable AMPA receptors, and paired stimulation of synaptic responses can be affected also by postsynaptic AMPAR mechanisms [36]. Thus, a lack of change in paired-pulse ratio may not be a reliable indication of a lack of pre-synaptic GABA A inhibition at these synapses. Further experimentation assessing additional parameters, such as coefficient of variation of synaptic currents [37], may be useful to help resolve this issue. Another possibility to assess if the SST-mediated indirect GABA A receptor inhibition occurs only at the presynaptic PC terminal would be to observe if the NMDA-component of synaptic responses recorded from SOM-INs is also potentiated by SST 14 .
SOM-INs receive postsynaptic GABA A mediated inhibition [13], notably from interneuron-selective interneurons expressing vasoactive intestinal peptide [38,39], and these could be targeted by SST. However, we did not observe changes in cell input resistance during and after SST 14 application (Fig. 1), suggesting no postsynaptic change. However, cell input resistance was measured at the soma and synaptic inhibition may occur at more remote dendritic sites [39]. Interestingly, local application of somatostatin depresses inhibitory postsynaptic potentials recorded in CA1 pyramidal cells, without affecting postsynaptic responses to exogenously applied GABA, indicating somatostatin-induced presynaptic inhibition of GABA synaptic responses in pyramidal cells [28]. Likewise, bath application of somatostatin presynaptically inhibits GABA synaptic transmission onto basal forebrain cholinergic neurons [29]. Similar experiments on inhibitory synaptic transmission onto SOM-INs would be important to clarify the mechanisms of SST 14 actions via GABA A inhibition. Our results also raise the question of which type of GABAergic interneuron expresses SSTRs pre-synaptically? SST 1-4 R, and to a lesser extent SST 5 R, are present in CA1 hippocampus and in pyramidal cells [40][41][42]. Generally, SSTR subtypes preferentially occupy specific cell compartments. SST 1 R is mainly pre-synaptic, SST 2,4,5 R post-synaptic, and SST 3 R extra-synaptic (neuronal cilia) [43]. However, which inhibitory cell type in the hippocampus expresses the receptors and whether they are pre-or post-synaptic remains largely to be determined [42,43]. Interestingly, SST 5 R and CB1 receptors co-localize in some CA1 interneurons [21]. CB1 receptors are highly expressed mostly in inhibitory interneurons that co-express the neuropeptide cholecystokinin (CCK) [44], suggesting that these interneuron subtypes may mediate SST actions on presynaptic GABA A inhibition. Further experiments will be required to elucidate the pre-and/or post-synaptic GABA A mechanisms involved in the disinhibitory actions of SST at SOM-IN excitatory synapses.
Our results with the antagonist cyclosomatostatin are also consistent with an effect of endogenous SST released after theta burst stimulation contributing to long-term potentiation at the PC to SOM-IN synapse indirectly via GABA A inhibition. Application of SST 14 failed to modify non-NMDAR-mediated EPSCs recorded in the presence of gabazine (Fig. 7). In addition, the SSTR antagonist cyclosomatostatin did not affect TBS-induced LTP of non-NMDAR-mediated EPSCs recorded in the presence of gabazine (Fig. 7). However, SST 14 and cyclosomatostatin showed effects on EPSPs recorded with GABA A inhibition intact (Figs. 1 and 2). These results suggest that, under physiologically relevant conditions, release of endogenous SST by theta burst stimulation contributes to long-term potentiation at PC to SOM-IN synapses indirectly via GABA A inhibition.

Mechanisms of SST-induced LTP
Although synaptic plasticity in some hippocampal interneurons involves NMDARs [34], mGluR1a-mediated Hebbian LTP in SOM-INs does not [12]. Consistent with this notion, our results indicate that SST 14 -induced LTP in SOM-INs is unaffected by the NMDAR antagonist DL-APV, and thus does not involve NMDARs.
Hebbian LTP requires mGluR1a activation [12] and our results indicate that SST 14 actions that lead to LTP of EPSPs occur downstream of mGluR1a activation since the antagonist LY367385 does not prevent SST 14 -induced potentiation. Moreover, the long-lasting actions of SST 14 are activity-dependent and require concomitant synaptic activity during application of SST 14 . In these experiments we observed that when synaptic stimulation was re-initiated after a period of interruption a rebound potentiation of synaptic responses was observed. Previous work in oriens-alveus interneurons has shown that, during recordings with intracellular BAPTA to buffer postsynaptic Ca 2+ levels, LTP is blocked [37]. However, in addition, the injection of BAPTA induces a long-lasting depression of synaptic responses [37]. Thus, postsynaptic Ca 2+ mechanisms are necessary for LTP induction and for maintenance of intact transmission at these synapses. Moreover, activation of excitatory synapses of SOM-INs involve calcium-permeable AMPARs [36] and Ca 2+ influx [45]. These results suggest that inactivation and subsequent reactivation of synapses may influence postsynaptic Ca 2+ homeostasis, resulting in rebound potentiation. Importantly, in comparison to SST-induced LTP, the magnitude of rebound potentiation was variable and transient, only reaching significance at 15-20 min after resuming stimulation, suggesting different mechanisms at play. Since SST-induced potentiation via GABA A inhibition occurs downstream of AMPARand mGluR1a-mediated Ca 2+ signals, the mechanisms of rebound potentiation are unlikely to have interfered with, or occluded, the SST effects. Thus, SST 14 actions on GABA A inhibition may require synaptic activity during SSTR activation to lead to long-lasting changes. Intriguingly, long-lasting reduction of synaptic inhibition by local application of SST was previously reported in PCs [28] but not with bath application [25]. Further experiments focusing on GABA A inhibition of SOM-INs will be necessary to explain the activity-dependent disinhibitory actions of SST 14 in SOM-INs.

Endogenous SST contributes to mGluR1a-mediated Hebbian LTP
Our results with the SST 1-5 R antagonist cyclosomatostatin suggest that, under physiologically relevant conditions, theta burst stimulation causes release of endogenous SST which contributes to LTP at SOM-IN excitatory synapses via GABA A disinhibition. The release of endogenous SST is frequency-dependent, as EPSPs elicited at 0.1 Hz are unaffected by the antagonist (Fig. 1). During the LTP induction protocol, theta burst stimulation elicits EPSPs that cause action potential burst firing in SOM-INs [8], conditions that are sufficient to cause release of endogenous SST. Such an activity-dependent release of SST is consistent with recent evidence that release of endogenous SST in acute prefrontal cortex slices is induced by frequency-dependent (> 10 Hz) optogenetic stimulation of SOM-INs [46] and that release of endogenous somatostatin in cultured hippocampal neurons is stimulated by AMPA receptor activation [47].
Our results with SST depletion by cysteamine also support a role of endogenous SST in Hebbian LTP. We found that systemic injection of cysteamine or in vitro treatment of slices with cysteamine prevented TBS-induced Hebbian LTP, providing further support for a role of release of endogenous SST in LTP at SOM-IN excitatory synapses. We observed that TBS resulted in longlasting depression of EPSPs after systemic cysteamine treatment. Since no lasting depression was observed after in vitro treatment of slices with cysteamine, the depression may be the result of extra-hippocampal effects of cysteamine treatment. In previous work during recordings with intracellular BAPTA to prevent postsynaptic Ca 2+ rise, LTP was blocked in oriens-alveus interneurons and replaced by LTD [37]. Moreover, in this previous work, injection of BAPTA alone induced a long-lasting decrease in EPSC amplitude, indicating that postsynaptic Ca 2+ mechanisms are necessary for LTP induction and for maintenance of intact transmission at these synapses [37]. Thus, extra-hippocampal effects of systemic cysteamine treatment may have interfered with Ca 2+ homeostasis in SOM-INs and resulted in LTD.
SST was previously shown to be critical for hippocampal long-term synaptic plasticity, as well as learning and memory. Depletion of SST by cysteamine treatment, or knock-out of the SST gene in transgenic mice, impairs hippocampus-dependent contextual fear memory but not hippocampus-independent auditory fear learning [30]. The memory impairment is associated with a decrease in LTP in CA1 PCs [30], as well as at mossyfiber CA3 PC synapses [48]. SST-induced LTP in SOM-INs may be the link between the role of SST in regulation of hippocampal network plasticity and hippocampal memory. Firstly, contextual fear learning was shown to induce a persistent LTP at excitatory synapses of SOM interneurons mediated by mGluR1 and mTORC1 [9]. Our finding that SST contributes to mGluR1a-mediated Hebbian LTP in SOM-INs, suggests that SST-induced LTP may be induced by contextual learning. Secondly, SOM cell-specific transgenic mouse approaches have shown a functional role of LTP at SOM interneuron excitatory synapses in hippocampal learning and memory [9]. Genetic down-regulation of mTORC1 activity impaired, whereas up-regulation facilitated, mGluR1amediated LTP at SOM interneurons excitatory synapses [9]. At the network level, SOM interneurons, and most notably OLM cells, are dendrite projecting inhibitory interneurons that differentially regulate Schaffer collateral (SC) and temporo-ammonic (TA) pathways onto CA1 pyramidal cells [7]: they suppress the distal TA pathway and facilitate the more proximal SC pathway [7]. Thus, LTP at excitatory synapses onto SOM interneurons causes long-term changes in their output firing [36] and inhibition of pyramidal cells [37], resulting in differential long-term regulation of plasticity at SC and TA synapses onto pyramidal cells: up-regulation of plasticity of the SC pathway [8,9] and down-regulation of plasticity of the TA pathway [11]. At the behavioral level, genetic loss of mTORC1 function specifically in SOM interneurons impaired contextual fear and spatial long-term memories, whereas genetic upregulation of mTORC1 augmented spatial and contextual fear memories [9]. Thus, learning-induced LTP at SOM-IN excitatory synapses is linked to regulation of CA1 network metaplasticity and hippocampal long-term memory consolidation [9]. Our findings that endogenous SST plays a critical role in LTP at SOM-IN excitatory synapses, suggest that impairments in LTP in CA1 pyramidal cells and deficits in contextual fear memory caused by SST depletion/ knockout [30] may be due to loss of SST-mediated LTP at SOM-IN synapses [9]. Thus, the role of SST in longterm synaptic plasticity of SOM-INs uncovered here may be crucially implicated in SST regulation of hippocampal learning and memory.

Animals
All animal procedures and experiments were performed in accordance with Université de Montréal Animal Care Committee (Comité de déontologie de l'expérimentation sur les animaux, CDEA) and followed the guidelines of the Canadian Council on Animal Care. Experiments were carried out on mice (5-8 week-old males for electrophysiology, and from both sexes for immunofluorescence). Mice were housed 2-5 per cage and given ad libitum access to food and water, in temperature (~ 22 °C) and humidity (~ 55%) controlled rooms with a normal 12 h light/dark cycle.

Cysteamine injection
To study somatostatin depletion, SOM-eYFP mice were injected intraperitoneally (IP) with 150 mg/kg cysteamine (Sigma-Aldrich; M6500) diluted in bacteriostatic NaCl 0.9% (Hospira) or vehicle [49]. After 4 h, mice were anaesthetized with isoflurane inhalation and then decapitated to obtain acute hippocampal slices, as described below. Some mice were deeply anesthetized with sodium pentobarbital (MTC Pharmaceuticals, Cambridge, Ontario, Canada) and were perfused transcardially, first with ice-cold 0.1 M phosphate buffer (PB), then with 4% para-formaldehyde in 0.1 M PB (PFA) and the brain isolated. Post-fixed brains were cryoprotected in 30% sucrose and coronal brain sections (50 µm thick) were obtained for immunofluorescence.

Acute hippocampal slice preparation
SOM-eYFP or PV-eYFP mice were anaesthetized with isoflurane inhalation and then decapitated. The brain was rapidly removed and placed in ice-cold sucrose-based solution containing (in mM): 75 sucrose, 87 NaCl, 2.

Cysteamine treated acute hippocampal slices
Hippocampal slices were obtained as above and transferred in oxygenated ACSF containing cysteamine (200 µM) or ACSF alone, for 1 h at room temperature. Slices were then used for electrophysiological whole cell recording (as described below) or fixed overnight at 4 °C with 4% PFA, rinsed with PB 0.1 M, cryoprotected in 30% sucrose/PB 0.1 M and re-sectioned (50 µm) using a freezing microtome (Leica SM200R, Germany) for immunofluorescence.

Whole-cell current clamp recordings
Slices were transferred to a submersion chamber perfused with ACSF (3-4 ml/min) at 31 ± 0.5˚C. CA1 interneurons expressing eYFP, or pyramidal cells not expressing eYFP, were identified using an upright microscope (Zeiss Axioskop, Toronto, Canada) with a water-immersion long-working distance objective (40X N-Achroplan, Zeiss, Toronto, Canada), epifluorescence lamp (FluoArc N HBO 103, Zeiss, Toronto, Canada) and an infrared digital video camera (Infinity 3, Lumenera, Ottawa, Canada). Whole-cell current clamp recordings were obtained using borosilicate glass pipettes (2-5MΩ; WPI) filled with intracellular solution containing (in mM): 120 KMeSO 4 , 10 KCl, 10 HEPES, 0.5 EGTA, 10 Na 2 -phosphocreatine, 2.5 MgATP, 0.3 NaGTP (pH 7.4, 300 mOsmol/L). Data were acquired using a Multiclamp 700B amplifier (Molecular Devices), digitized at 20 kHz using Digidata 1440A and pClamp 10 (Molecular Devices). Recordings were lowpass filtered at 2 kHz. Access resistance was regularly monitored during experiments and data was included only if the holding current was stable and access resistance varied less than 20% of initial value. Excitatory postsynaptic potentials (EPSPs) were evoked using constant current pulses (50 µs duration) via a concentric bipolar Pt/Ir electrode (FHC) placed in stratum oriens near the alveus, 100 µm lateral from the recorded cell soma. Membrane potential was held at -60 mV by constant current injection. EPSPs were evoked during a hyperpolarizing current step (5-10 mV, 0.5-1 s duration) to avoid action potential generation. Paired stimulations (50 ms inter-event interval) were given at 0.1 Hz. LTP was induced by three episodes (at 30 s intervals) of theta-burst stimulation (TBS) of afferents (five bursts, each consisting of four pulses at 100 Hz, with a 250 ms interburst interval).

Somatostatin immunofluorescence
Sections were permeabilized with 0.3% Triton X-100 in 0.1 M PB (15 min) and unspecific binding was blocked with 10% normal goat serum in 0.1% Triton X-100 and 0.1 M PB (1 h). Sections were incubated 24-48 h at 4ºC with mouse monoclonal somatostatin antibody (1/500; Santa Cruz Biotechnology; Dallas, TX), and subsequently at room temperature with Rhodamine-Red ™ X-conjugated goat anti-mouse IgG2b (1/200; 90 min; Jackson Immunoresearch Labs; West Grove, PA). Sections were mounted in ProLong ™ Diamond (Life technologies; Carlsbad, CA) and images were acquired using a confocal microscope (LSM880; Carl Zeiss, Oberkochen, Germany) at excitation 488 and 543 nm. Images were acquired using the exact same parameters fixed on control slices (ACSF) or mice (saline). The intensity of the somatostatin immunofluorescence, in oriens-alveus region of the CA1 hippocampus was quantified using ImageJ software (National Institute of Health; https:// github. com/ imagej/ image j1) by comparing integrated density in cells corrected for background fluorescence. For experiments with cysteamine IP injection, cell fluorescence was measured typically in 44-63 fields of view per animal coming from 3-4 sections and averaged per animal. A total of 3 animals per group coming from 3 independent experiments were analyzed (total of 397 cells for saline; 441 cells for cysteamine). For acute slices, cell fluorescence was measured typically in 3-49 fields of view per slice coming from 2-4 sections and averaged per slice. A total of 3 animals coming from 3 independent experiments were analyzed (total of 714 cells for ACSF; 713 cells for cysteamine).

Pharmacology
The neuropeptide somatostatin (SST 14 ; Abcam #141206) was diluted daily in ACSF at 5 µM and perfused for 5 min and then washed-out during whole cell recordings of EPSPs/EPSCs. In some experiments, cyclosomatostatin (Abcam #141211), a non-selective SST 1-5 R antagonist was dissolved in DMSO and applied at a final concentration of 1 µM in ACSF. It was perfused for 5 min before and during somatostatin application, and then washed out. In TBS LTP experiments, cyclostomatostatin was perfused for 10 min before and during TBS, and then washed out. In some experiments, 40 µM LY367385 (Tocris #1237), a mGluR1a selective antagonist, or 50 µM DL-APV (Abcam #120004), a NMDAR selective antagonist, were diluted in ACSF and perfused throughout the recording period (40 min).