Hilar Somatostatin Interneurons Contribute to Synchronized GABA Activity in an In Vitro Epilepsy Model

Epilepsy is a disorder characterized by excessive synchronized neural activity. The hippocampus and surrounding temporal lobe structures appear particularly sensitive to epileptiform activity. Somatostatin (SST)-positive interneurons within the hilar region have been suggested to gate hippocampal activity, and therefore may play a crucial role in the dysregulation of hippocampal activity. In this study, we examined SST interneuron activity in the in vitro 4-aminopyridine (4-AP) model of epilepsy. We employed a multi-disciplinary approach, combining extracellular multi-electrode array (MEA) recordings with patch-clamp recordings and optical imaging using a genetically encoded calcium sensor. We observed that hilar SST interneurons are strongly synchronized during 4-AP-induced local field potentials (LFPs), as assayed by Ca2+ imaging as well as juxtacellular or intracellular recording. SST interneurons were particularly responsive to GABA-mediated LFPs that occurred in the absence of ionotropic glutamatergic transmission. Our results present evidence that the extensive synchronized activity of SST-expressing interneurons contribute to the generation of GABAergic LFPs in an in vitro model of temporal lobe seizures.


Introduction
Temporal lobe epilepsy is the most common type of adult pharmacoresistant focal seizure disorder, characterized by excessive and abnormally synchronous activity in the hippocampus and surrounding cortex [1]. GABAergic interneurons of the hippocampal hilus are thought to act as a gate for runaway excitation [2], and have therefore been implicated in the pathogenesis of temporal lobe epilepsy. The two major subtypes of interneurons in this area are the parvalbumin (PV)-positive fast-spiking interneurons and the somatostatin (SST)-positive, low-threshold spiking interneurons [3;4]. As SST-interneurons are strongly implicated in gating hippocampal activity [2], we investigated the role of SSTexpressing interneurons in the generation of epileptiform synchronization using mice that express Cre recombinase in this specific neuronal population [5].
Using a multidisciplinary approach that combined extracellular and intracellular recording with optical imaging, we studied the activity of SST interneurons during epileptiform activity. We used Cre recombinase-driven expression of the GCaMP3 optical Ca 2+ sensor [17] in SST-Ires-Cre neurons [5] to selectively express GCaMP3 in SST-positive interneurons. Combining this optical imaging with extracellular recordings using pMEA and patch-clamp recordings from visually identified SST-interneurons, we found that SST interneurons are strongly synchronized during all LFPs. We also found that SST interneurons are driven more extensively by neuronal activity resulting from the combined activation of dentate granule (DG) granule cells and CA3 pyramidal neurons. Although SSTinterneurons all behaved similarly during 4-AP-induced epileptiform activity, upon blockade of glutamatergic transmission we revealed distinct action potential firing patterns of these neurons, which might be related to the generation of longlasting, GABA-mediated LFPs [9].

Slice Preparation
All experiments were performed in accordance with the Georgetown University Animal Care and Use Committee (GUACUC) and in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH Publications No. 8023, revised 1978

pMEA Recordings
Neuronal network activity was recorded using a perforated multi-electrode array (pMEA, Multi Channel Systems, Reutlingen, Germany) with 59 embedded titanium nitrite electrodes, each with a diameter of 30 mm and an inter-electrode spacing of 200 mm. Recorded electrical activity from each channel was digitized at a sampling rate of 5 kHz and acquired using a MEA2100 amplifier and MC Rack 4.0 software (Multi Channel Systems, Reutlingen, Germany). The temperature in the recording chamber was constantly monitored using an analog YS1 Thermometer (Yellow Spring Instruments, Yellow Springs, OH) and maintained at 26 to 28uC by a heated perfusion cannula (ALA Instruments, Farmingdale, NY).

Single and Dual Patch-clamp Recordings
Single-and dual-cell patch-clamp recordings were performed from visually identified fluorescent SST-positive interneurons in the hilar region of the dentate gyrus in 16 slices derived from 10 mice using a fluorescent microscope (Axioskop, Zeiss, Jena, Germany). Juxtacellular, cell attached, recordings to record action potentials firing rates from outside the membrane without breaking in the cell were performed from 22 SST-interneurons and whole cell recordings after breaking the membrane to evaluate action potentials and of synaptic activity from 16 SST-interneurons.Patch-clamp microelectrodes were fabricated from borosilicate capillary tubes and pulled using a two-step pipette puller (PP-83, Narishige, and Tokyo, Japan). Pipettes with tip resistances between 4-6 MV were filled with internal solution (in mM): 145 Kgluconate, 1.1 EGTA, 5 MgATP, 0.2 NaGTP, and 10 HEPES, adjusted to pH 7.2 with KOH. Recordings were performed using Axopatch 1D and Axopatch 200B amplifiers, a Digidata 1440A interface, and pClamp software (all from Molecular Devices, Sunnyvale, CA). Signals were low-pass filtered at 2kHz, sampled and processed at 5 kHz, and acquired with two analog channels of the MCS amplifier MEA-2100. Data are shown as mean 6 SEM.

Imaging
To determine the maximal distribution of GCaMP3 expressing neurons, confocal fluorescence imaging of hippocampal slices from SST-GCaMP3 mice was performed in the presence of 40 mM KCl ( Figure 1B). For all other experiments, cells were imaged in aCSF using a fluorescent microscope (Axioskop, Zeiss, Jena, Germany) equipped with Plan-Neofluar 2.5x/0.075, N-Achroplan 20x/0.5, and N-Achroplan 40x/0.75 objectives (Zeiss, Germany). Images were captured at a sampling rate of 20 Hz using a chargecoupled device camera (CoolSNAP HQ 2 , Photometrics, Tucson, AZ) and Nikon Elements software (Nikon, Japan). Calcium responses were recorded using the intrinsic fluorescent properties of the fluorescent Ca 2+ sensor in the SST-interneurons following 488 nm centered wavelength excitation (PhotoFluor II, 89 North, Burlington, VT). To monitor the changes of the calcium signal, image sequences were taken every 50 ms for a period of 2 minutes and analyzed using ImageJ (National Institutes of Health, Bethesda, MD). After background subtraction, regions of interest (ROI) were selected manually and changes in intensity levels (DF/ F) over time were computed for each ROI with Clampfit (Molecular Devices, Sunnyvale, CA).

Results
In this study we took advantage of a combination of imaging, patch-clamp recordings and multi-electrode array recordings to determine the activity of the SST neurons, an important subclass of GABAergic interneurons in the hippocampus, during the 4-AP model of epileptiform activity. We used SST-Ires-Cre mice [5] to drive the Cre recombinase-driven expression of either the red fluorescent reporter td-Tomato ( Figure 1A) or the calcium sensor GCaMP3 ( Figure 1B) in SST-interneurons. Within the hippocampus, the hilus contained a high number of SSTinterneurons (Figure 1), consistent with previous reports [5,19]. The presence of a fluorescent probe allowed us to perform patchclamp recordings from visually identified interneurons, while recording extracellular activity throughout the slice with a perforated multi-electrode array (pMEA).
We used the previously characterized 4-AP model to induce epileptiform activity in hippocampal slices [15,16] (Figure 2). Consistent with our previous reports [16], we observed two distinct classes of LFPs, the first originating and remaining for the large part in the CA3 region (CA3-restricted LFPs; LFP R-CA3 ) and the second propagating from CA3 to the DG (LFP P-CA3/DG, indicated with asterisk in Figure 2A1, B1). We performed juxtacellular recordings from 22 SST-interneurons to compare the firing of individual SST neurons to these two classes of LFPs. SSTinterneurons increased action potential firing in response to both types of LFPs, with the majority of neurons (12/16) showing more pronounced responses to the LFP P-CA3/DG . (Figure 2A2, B2). Three neurons had comparable firing frequencies with both LFP P-CA3/DG and LFP R-CA3 and one neuron showed an increase in its action potential frequency only with LFP R-CA3 .
We also observed increased fluorescence signal with respect to LFPs using optical imaging using the genetically encoded Ca 2+ sensor, GCaMP3 ( Figure 3A). To confirm that the increased Ca 2+ fluorescent corresponded to action potential firing, we performed juxtacellular recordings from SST-interneurons ( Figure 3B3). Both GCaMP3 fluorescence intensity (DF/F) ( Figure 3B1) and AP firing rate ( Figure 3B3) increased in hilar SST interneurons during the LFP P-CA3/DG and the LFP R-CA3 . However, the DF/F was consistently larger in correspondence to the LFP P-CA3/DG than to the LFP R-CA3 (LFP R-CA3 :8.160.01%, N = 122 neurons; LFP P-CA3/DG : 23.260.05%, N = 46 neurons; Average LFPs in neurons pooled from 9 slices, p,0.05, paired t test).
We therefore investigated the role of the SST interneurons in the generation of the large, GABAergic CPP/NBQX LFP. Upon perfusion with iGluR blockers, GCaMP3 imaging revealed a Ca 2+ transients matching the CPP/NBQX LFP ( Figure 3C1) with the average DF/F of 4.860.01% (N = 39 neurons from 9 slices), a considerable drop from the average DF/F of the LFP P-CA3/DG (see above). In 12 of 22 SST-interneurons, iGluR blockade induced spontaneous action potential firing (3.8160.76 Hz) in the absence of mean activity recorded with LFP as shown in the example in Fig. 3C bottom. In 9 of these neurons, action potential firing briefly increased at the onset of the CPP/NBQX LFP (9.662.2 Hz), followed by a silent period of variable length ( Figure 3C3). In three other interneurons action potential firing was present only in correspondence with the CPP/NBQX LFP. The remaining 10 cells did not increase basal firing frequency and did not respond or had occasional spikes correlated with the CPP/ NBQX LFP.
The Ca 2+ transients evoked in SST-interneurons by the CPP/ NBQX LFPs were often distinct between neurons in the same slice (Fig. 4). We observed three types of Ca 2+ transients depicted in Figure 4B-C: i) cells with increased Ca 2+ signal (grey fill), ii) cells with biphasic Ca 2+ transients, consisting of an increased and then decreased Ca 2+ signal (brown shade), or iii) cells displaying intermediate behavior (pink shade).
To gain insights into the events occurring in SST-interneurons during LFPs induced by 4-AP, we performed whole-cell recordings from visually identified neurons before and during glutamatergic blockade. GCaMP3 imaging again revealed consistently larger Ca 2+ responses to the LFP P-CA3/DG ( Figure 5A, orange traces), corresponding to the increased AP firing observed in juxtacellular recordings ( Figure 5A2). Intracellular current-clamp recordings revealed spontaneous PSPs with large synchronous depolarization during LFPs ( Figure 5A3). We also performed simultaneous juxtacellular, whole cell, and MEA recordings together with Ca 2+ signal. In 3 neurons from distinct mice, both the duration of the depolarization measured in current clamp recordings (1.260.1 s, n = 21 events), and the spike burst duration measured with juxtacellular recordings (1.360.8 s, n = 21 events) in correspondence to the LFP P-CA3/DG were longer (p,0.05, unpaired t test) than those measured in correspondence to the LFP R-CA3 (0.2860.12 s and 0.2960.14 s, n = 65 events), suggesting a stronger activation of SST-interneurons during the LFP P-CA3/ DG . In two of these neurons, perfusion with CPP/NBQX induced sustained spiking in the SST interneuron recorded in juxtacellular mode ( Figure 5B2), although no such firing was observed in current-clamp mode ( Figure 5B3). Interestingly, the long pauses in action potential firing seen in juxtacellular recordings after the CPP/NBQX LFP corresponded with a small, long-lasting synaptic potentials revealed in current clamp recordings,that was investigated further in voltage-clamp recordings (see below). In some of the recorded cells as illustrated before, the Ca 2+ transient observed in the juxtacellular recorded cell was clearly biphasic ( Figure 5B2), suggesting that the increase and subsequent pause in the firing rate during the CPP/NBQX LFP are the underlying mechanism. The transient however, was observed even in cells were no spiking was detected either in juxtacellular or in current clamp recordings (n = 13). Interestingly, the long pauses in action potential firing seen in juxtacellular recordings after the CPP/ NBQX LFP corresponded with a small, long-lasting synaptic potentials revealed in current clamp recordings ( Figure 5B3), that were not investigated further.
To elucidate these findings, we compared optical ( Figure 6A2) and electrical recordings in juxtacellular ( Figure 6A3) and voltage-clamp mode at 250 mV ( Figure 6A4). In this recording configuration we could observe at the same time EPSCs and IPSCs as inward and outward currents, respectively. The inward excitatory postsynaptic currents (EPSCs) 16 SST interneurons in 16 slices from 10 mice were on average significantly longer in correspondence to the LFP P-CA3/DG but not larger, (duration: 1.2360.14 s, amplitude: 391690 pA, n = 21 events), than to the LFP R-CA3 (0.2860.1 s, 190642pA, n = 75 events; p,0.001 and p = 0.078 respectively, unpaired t-test versus LFP P-CA3/DG ), suggesting that SST-interneurons receive a robust barrage of excitatory synaptic input from the hippocampal network during  the LFP P-CA3/DG . In 9 neurons studied, perfusion with CPP/ NBQX abolished the EPSCs corresponding to the LFP P-CA3/DG and to the LFP R-CA3 ( Figure 6B), revealing the occurrence of sustained outward inhibitory postsynaptic currents (IPSCs 47621 pA, n = 23 events) in association with the CPP/NBQX LFP ( Figure 6C). In addition, we observed a significant increase in spontaneous IPSCs frequency (from 1.0960.04 Hz to 5.3660.07 Hz) in the presence of iGluR blockers ( Figure 6B,D), possibly deriving from spontaneous activation of other interneurons (see above and [16]).

Discussion
Previous studies have suggested the occurrence of specialized hippocampal LFPs in the 4-AP model of epilepsy [7-10;15,16]. In particular, there is strong evidence for ''GABA-mediated LFPs'' that occur synchronously in different areas of the hippocampus and in several other hippocampal subregions, [7-10;15,16]. The application of iGluR blockers reveal that these events originate primarily in the DG area and that excitatory synaptic transmission is not required. Together, this suggests that synchronous activity of GABAergic interneurons is the likely underlying mechanism of these LFPs. Our findings bring experimental evidence to this hypothesis with a unique approach combining the strengths of pMEA, whole-cell patch clamp and Ca 2+ imaging from genetically modified mice. We took advantage of the previously developed SST-Ires-Cre mouse model [5] to study SST interneurons tagged with fluorescent markers in the hilar region of the hippocampus. Using these methods, we present two novel findings. First, using juxtacellular recordings from SST-interneurons, we find that the majority of SST neurons fire more robustly with the LFP P-CA3/DG than the LFP R-CA3 , as recorded with pMEA electrodes in the DG and CA3. This result was extended to a larger number of neurons with the study of fluorescence changes in SST neurons. Secondly, our data suggest that the excitatory innervation of SST interneurons is strongly induced by neurons activated during the LFP P-CA3/DG , which include both CA3 pyramidal neurons and DG granule neurons. Together, these data bring evidence to the hypothesis that the synchronous activation of SST interneurons contributes to epileptiform activity.
SST interneurons were well synchronized with the two types of LFPs and little firing was seen in between LFPs. The combined application of two blockers of glutamatergic synaptic transmission increased the basal firing rate in over half of the SST neurons, resulting in pauses during the CPP/NBQX-induced LFP. This effect is likely due to the intrinsic firing properties of the neuropeptide Y (NPY)-expressing interneurons that hyperpolarize after bursts of activity [19] and have large overlap in the expression with SST interneurons [5,19]. In some SST-interneurons however, we observed either no increase in basal firing or higher rates of basal firing with the CPP/NBQX LFP. This suggests that either the CPP/NBQX LFPs had a dual effect on these neurons or that two distinct populations of SST cells are present. The results of Ca 2+ imaging further illustrate that there are various types of Ca 2+ signaling in SST interneurons, with responses ranging from mono-to bi-phasic signaling. Additionally, combined data from pMEA, current clamp recordings and Ca 2+ imaging indicate that small Ca 2+ transients were present even without spiking of the recorded neuron. We speculate that dendritic calcium transients propagating to the soma might be the underlying cause of this effect. Nevertheless, this creates a limitation in the interpretation of the transients observed with Ca 2+ imaging as they may not be exclusively attributable to spiking activity.
Our results demonstrate that in SST-interneuron, barrages of EPSCs are synchronized by the LFP P-CA3/DG fields and the application of iGluR blockers uncovers IPSCs barrages. We also observed an increase in the baseline of spontaneous IPSCs (sIPSCs), which correlates with the increase in spontaneous firing observed from SST-interneurons in cell attach mode. Our study, however, failed to observe the progressive and larger shifts in holding current that one would expect if the proposed shifts in the extracellular K + concentration were operative [10,20]. The presence of strong IPSCs during the CPP/NBQX LFP suggest reciprocal innervation between SST-interneurons. This will explain the finding of distinct populations of SST interneurons, one that fire action potentials only during CPP/NBQX LFPs and the other that fire spontaneously and pause during the CPP/ NBQX LFPs. This effect may be also related the rhythmic bursting of hilar NPY neuron evoked by 4-AP as shown in [19]. Our hypothesis of a reciprocal innervation between SSTinterneurons would also imply that the balance of activity between the different populations of SST-interneurons might be involved in the mechanisms generating the CPP/NBQX LFP.

Conclusion
Using multiple approaches, we present clear evidence to support the hypothesis that the synchronization of hilar SST GABAergic interneurons contribute to the generation of CPP/NBQX LFPs. While our findings do not expose the SST-interneurons to be the main players in the generation of these LFPs, we believe that other subtypes of interneurons may be the mastermind allowing for this synchronization. Further studies with other Cre-expressing mice may shed light on this important quest.

Author Contributions
Conceived and designed the experiments: SG BQ RL SV. Performed the experiments: SG BQ RL SV. Analyzed the data: SG SV RL. Contributed reagents/materials/analysis tools: SV. Wrote the paper: SG BQ RL SV. Conception and design of the work, acquisition, analysis and interpretation of data, Drafting the article and revising it, Final approval of the version to be published: SG BQ RL SV.