In our initial experiments EGFP-positive interneurons (INs) with soma located in the superficial portion of stratum lacunosum-moleculare (SLM) of mid-ventral hippocampi and Layer I (LI) of cortex in P35-P60 Gad2-EGFP or Htr3a-EGFP mice were targeted (see Materials and methods for details; Chittajallu et al., 2017; Lee et al., 2010; Tricoire et al., 2011). Using the approach our focus was restricted to NGFCs derived from progenitors located in the caudal ganglionic eminence (i.e. CGE-NGFCs) (Overstreet-Wadiche and McBain, 2015). Approximately 80% of targeted CGE-NGFCs (124 out of 150 EGFP-positive INs) displayed late-spiking (LS; with varying degrees of latency) phenotype and an axo-dendritic morphology (38 EGFP-positive INs) characteristic of NGFCs (Overstreet-Wadiche and McBain, 2015; Figure 1A; Figure 1—figure supplement 1A). Using brief (2 ms) supra-threshold depolarizing current injection steps (1000–1200 pA) we elicited 10 action potentials at a frequency of 30 Hz repeated at 1 s intervals (Figure 1B). These conditioning APs resulted in the previously described phenomenon of barrage/persistent firing (BF) (Deemyad et al., 2018; Elgueta et al., 2015; Imbrosci et al., 2015; Jung and Hoffman, 2009; Krook-Magnuson et al., 2011; Rózsa et al., 2017; Sheffield et al., 2011; Suzuki et al., 2014) in approximately 70% of NGFCs tested (86 out of 124 NGFCs; Figure 1B; Figure 1—figure supplement 1C). BF constitutes a specific mode of intrinsic plasticity and is characterized by spontaneously generated action potential output in the absence of a depolarizing stimulus. In parallel, INs with soma located in layer I (LI) slices of human cortical tissue (resected from patients with pharmaco-resistant epilepsy in order to gain access to remove the hippocampus, the latter structure deemed the focal pathological region; see Materials and methods for details) were also targeted with a subpopulation demonstrating a late-spiking pattern (25 cells) identifying them as NGFCs (with additional morphological confirmation in 17 of these 25 cells; Figure 1C). Upon delivering the same induction protocol as described above, BF was elicited in 48% of human NGFCs (12/25 NGFCs tested; Figure 1D; Figure 1—figure supplement 1A,C). Interestingly, even with the number of conditioning APs far exceeding that required to induce BF in NGFCs (800–1500 APs; 30 Hz), BF could not be elicited in superficial cortical and hippocampal INs that were not NGFCs (non-NGFCs), including the recently described human rosehip cell (Boldog et al., 2018) (26 and 37 non-NGFCs tested for BF induction in mouse and human, respectively; Figure 1—figure supplement 1B,D). We tested induction protocols consisting of differing frequencies (8 Hz and 100 Hz) and these were also found to elicit BF with essentially similar induction requirements and duration (Figure 1E,F). Despite a similar number of conditioning APs necessary for BF induction being similar in both species, the mean BF duration in human NGFCs albeit more variable, was approximately double on average (Figure 1E,F). In some recordings, the induction protocol was continued immediately following cessation of BF to ascertain whether subsequent BF epochs of a longer duration could be triggered. However, NGFCs could not display another bout of BF using this experimental approach (Figure 1—figure supplement 2A). Interestingly, subsequent BF episodes could be elicited provided a sufficient refractory period had elapsed between delivery of subsequent induction protocols (Figure 1—figure supplement 2B). Circumventing such constraints in BF induction (i.e. allowing greater than 1 min before subsequent induction delivery; Figure 1—figure supplement 2C) multiple BF episodes (up to four were tested) could be reliably triggered (Figure 1—figure supplement 2B,D,E).

Spiking events during NGFC BF episodes in mouse and human were comprised of both spikelets and overshooting action potential spikes (the latter defined as those in which the peak overshot 0 mV; Figure 1B,D,G). The spikelet waveform was typically nested within that of the spike as evidenced by a prominent inflection point in the voltage rise of the latter and in corresponding phase plots (Figure 1G,H). This biphasic rise, noted in previous studies (Elgueta et al., 2015; Sheffield et al., 2011), has led to the notion that BF consists of spontaneous AP originating at very distal axonal site(s) that travel retrogradely towards the soma and in some instances subsequently trigger an AP at more proximal sites (Elgueta et al., 2015; Imbrosci et al., 2015; Sheffield et al., 2011) (hence also referred to as retro-axonal BF Sheffield et al., 2013). Taken together, these data demonstrate that, of the IN subtypes residing in superficial lamina of cortical and hippocampal microcircuits, NGFCs are privileged with regard an ability to express BF in brief, repeatable episodes. Furthermore, this activity-dependent functional plasticity originating in the distal axonal compartments, is clearly conserved in NGFCs embedded in human cortical circuits.

During routine monitoring of recording stability before and immediately after BF cessation, we noted that small depolarizations delivered at the soma employed in voltage clamp to measure access resistance, in some cases, was sufficient in activating action currents. To investigate further, action potential output to depolarizations via a series of 1 s positive current injections before and after BF induction was examined (0 pA to 225 pA; 25 pA steps; Figure 2A). A leftward shift in the input/output curve and increased total number of action potentials were evident after excursions into the BF mode (Figure 2B–C). The increased excitability observed here occurred in the absence of changes in input resistance (208 ± 53 mΩ and 220 ± 47 mΩ for baseline vs. STP-SE, respectively; n = 11 mouse NGFCs; p=0.33).

To ascertain the duration of this increased excitability, we reverted to voltage-clamp for more precise control of both the holding membrane potential and the depolarizing stimulus. Action currents in any given NGFC were monitored over time following BF cessation in response to a constant test depolarizing voltage command that was found to be ineffective in eliciting any action currents prior to BF induction (+30 mV from a holding potential of −70 mV; Figure 2D,E). Using this approach, the potentiated NGFC output following cessation of BF1 was found to persist for a mean duration of approximately 60 s (Figure 2F,G). Interestingly, even though the duration of BFs expressed in individual NGFCs does not increase with repeated induction (Figure 1—figure supplement 2), the potentiation of somatic depolarization-induced output after the cessation of these subsequent BF episodes was markedly prolonged thus demonstrating the ability to prime this mode of plasticity by multiple rounds of induction (Figure 2G; Note: in three instances after cessation of BF4 heightened action potential output lasted 900 s at which point the experiment was terminated).

These data demonstrate that, in addition to the ability to undergo brief periods of distal axonal plasticity (i.e. BF), both mouse and human NGFCs also exhibit an activity-dependent increase in the sensitivity to proximal somatic depolarizing stimuli. For the remainder of the manuscript this mode of plasticity will be termed short-term potentiation of somatic-depolarization driven excitability (STP-SE).

Short periods of BF generated in distal axonal region(s) with subsequent driving of action potentials that likely back-propagate into the dendritic tree (Li et al., 2014) could conceivably trigger cellular pathways to influence intrinsic excitability resulting in STP-SE. Therefore, we sought to determine whether BF expression is a prerequisite for STP-SE by employing a protocol designed to assay its minimum induction requirements. This was achieved by eliciting action potential spiking in response to test rheobase current injection steps of 250 to 300 ms length via the patch pipette (typically 200 pA). After a baseline period, an induction paradigm was initiated by delivery of conditioning APs as previously described (10 action potentials at 30 Hz) while monitoring action potential output to the same test current pulse (Figure 3A). The induction protocol was maintained until commencement of a gradual loss of delay to spiking was noted at which point the conditioning APs were terminated and action potential output to the rheobase current injection continued to be assayed (Figure 3A). Notably, with this induction strategy, STP-SE was readily elicited in the absence of measurable BF in mouse hippocampal (n = 9) and human cortical NGFCs (n-4; Figure 3A–D) with the mean number of conditioning APs to elicit potentiation in excitability significantly lower than that required to induce BF (Figure 3E). As with BF, a subset of NGFCs (7/16 and 2/6 mouse hippocampal and human cortical NGFCs, respectively) and non-late spiking INs (4/4 and 3/3 mouse and human non-NGFCs, respectively) were unable to undergo STP-SE even under protracted period of conditioning APs (Figure 3D; Figure 3—figure supplement 1). Finally, this potentiation of intrinsic excitability was characterized by a significant hyperpolarization of action potential threshold (Figure 3F–H) and reduction in the delay to spiking (Figure 3B,C,I) which in combination likely contributes to STP-SE. Together, these data demonstrate the propensity of both mouse and human NGFCs to undergo an activity-dependent potentiation of their action potential output in response to somatic depolarizations without the requirement of entering into BF. For the remainder of the study, we focused our analyses to mouse hippocampal NGFCs to permit insight into the underlying cellular underpinnings for STP-SE expression.

Numerous cellular conductances operating at subthreshold voltages, including those carried by K⁺-channels, are well suited to influence both the delay to spiking and action potential threshold in a variety of neuronal subtypes. Thus, dynamic modulation of their function constitutes a potentially rapid and powerful cellular mechanism to modulate intrinsic excitability. However, little is known about the complement of K⁺-channels in positively identified NGFCs (Overstreet-Wadiche and McBain, 2015). Under conditions of voltage-gated Na⁺- and low threshold Ca²⁺-channel block, a depolarizing ramp, indicative of an NGFC identity, is clearly apparent in a subpopulation of Htr3a-EGFP SLM interneurons in response to somatic current injections (Figure 4A; note for these experiments this parameter was solely employed to define NGFC identity without further morphological confirmation). Restricting our analyses to this subset of recorded neurons, standard voltage-clamp protocols in combination with a pharmacological approach were performed to investigate K⁺-channel expression. Depolarizing steps from −70 mV to membrane potentials close to AP threshold (AP threshold in NGFCs = −29 ± 0.3 mV; n = 70) revealed both inactivating and to a lesser extent sustained K⁺-conductances operating at subthreshold membrane potentials (Figure 4B,C). Activity of subthreshold Kv1 conductances, in particular Kv1.1 (often referred to as D-type) have been linked to the phenomenon of late-spiking in excitatory and inhibitory neuronal subtypes (Bekkers and Delaney, 2001; Campanac et al., 2013; Dehorter et al., 2015; Storm, 1988). However, the subthreshold K⁺-conductance(s) in NGFCs was not significantly inhibited by α-dendrotoxin (DTX; 100–200 nM), a specific blocker of Kv1 (Kv1.1, Kv1,2 and Kv1.6) channels (Figure 4D). A low concentration of 4-AP (low 4-AP; 100 μM), which also blocks Kv1 channels, caused only a modest yet statistically significant inhibition of subthreshold K⁺-conductance(s) (Figure 4D). In contrast, millimolar concentrations of 4-AP (high 4-AP; 1–3 mM) produced a significant robust inhibition of the subthreshold K⁺-conductance with this digitally isolated conductance exhibiting a distinct inactivating time profile reminiscent of A-type K⁺-channels (I_KA; Figure 4D,E). We next examined the extent to which these pharmacological manipulations affected delay to spike and action potential threshold. Although DTX and low 4-AP reduced delay to spike and hyperpolarized action potential threshold, the extent to which this occurs was quite variable and not as marked to that seen following induction of STP-SE (Figure 4F,I). In contrast, high 4-AP and 4-AP delivered to NGFCs via the intracellular patch pipette (intra 4-AP; 3 mM) (Jung and Hoffman, 2009), modulated the delay to spike and hyperpolarized the AP threshold to an extent that more closely mimics the shift in both these parameters observed during STP-SE (Figure 4G–I).

Next, the ability of these pharmacological manipulations to increase intrinsic excitability in response to suprathreshold somatic current injections (1.5 × rheobase) was analyzed. In agreement with the propensity to modulate spike latency and AP threshold (Figure 4I), DTX and low 4-AP increased intrinsic excitability to a much lesser extent than high 4-AP or intra 4-AP (Figure 5A-C). Interestingly (in the presence of TTX) pharmacological inhibition of I_KA with millimolar concentrations of 4-AP results in a transformation of the membrane voltage response characterized by the emergence of a depolarizing hump (Figure 5D) as would be predicted by the time-dependent profile of I_KA (Figure 4B, E). This increased envelope of depolarization, particularly at the onset of the depolarizing stimulus step, must contribute to the loss of delay to spike and increased excitability seen during STP-SE (Figures 2A-E, 3D,H,I, 4G-I).

We next took advantage of a characteristic biophysical property of I_KA that manifests as a temporal delay in recovery of activation following inactivation to further investigate its role in NGFC excitablility. Thus, by delivering depolarizing subthreshold pulses in quick succession the extent of I_KA influence in NGFCs could be experimentally ‘titrated’ (Figure 5E,G). In a separate series of experiments, we assessed the spike delay under conditions where the extent of I_KA activation was manipulated in this manner (Figure 5F,G). We reveal that the time course of the recovery from inactivation of I_KA is mirrored by that seen in the recovery of the delay to spike (Figure 5G). Furthermore, a close relationship exists between the extent of I_KA inhibition in NGFCs (mediated by the differing magnitudes of recovery from inactivation) and corresponding increase in intrinsic excitability as measured by action potential output to the depolarizing test current injection (Figure 5H). Together, these pharmacological/biophysical analyses demonstrate that a transient, subthreshold K⁺-conductance(s) sensitive to high concentrations of 4-AP is a major contributor in dictating action potential latency, threshold and ultimately intrinsic excitability in NGFCs.

We next examined the extent to which pharmacological block of the I_KA alters the ability of NGFCs to express STP-SE. For these experiments, I_KA was blocked by introduction of 4-AP via the recording pipette as previously described (Figure 4H; Figure 5A) to prevent circuit wide perturbations that could confound the interpretation of results (Deemyad et al., 2018). Interleaved experiments (K-gluconate control vs. K-gluconate with intra 4-AP) clearly demonstrate that STP-SE could be induced more readily 4-AP containing internal solution as indicated by a leftward shift in the time course of potentiation that corresponds to a significant reduction in the number of required conditioning APs (Figure 5I,J). This gain of function provides further evidence for a reduction in 4-AP sensitive conductance(s) particularly those that resemble I_KA, as a contributing factor in STP-SE. However, due to our inability to fully occlude STP-SE we cannot rule out a role for additional modulation of other channels that are insensitive to intra-4-AP block.

K⁺-channels comprised of distinct subunits can contribute to I_KA in neural cell types (Rudy, 1988). Channels containing alpha subunits Kv1.1 (when in combination certain auxiliary subunits for example Kvβ1), Kv1.4, Kv3.3, Kv3.4 and Kv4 subunits can all mediate I_KA and regulate neuronal excitability (Carrasquillo et al., 2012; Rudy, 1988). The complement of channels responsible for this conductance in NGFCs, however, has not been previously explored. Our previous pharmacological analyses to this point (Figure 4) suggest a relatively small contribution of Kv1.1 in increasing intrinsic excitability of NGFCs that itself is insufficient to fully recapitulate the changes observed in STP-SE (Figure 5B) thus implicating additional transient K⁺-channels. As a first pass to identify the molecular identity of I_KA in NGFCs, we probed the publicly available Allen Brain Institute mouse dataset (Tasic et al., 2018) (https://portal.brain-map.org/atlases-and-data/rnaseq#Mouse_Cortex_and_Hip; see Materials and methods for details) containing single-neuron mRNA profiles of ~76,000 pan-glutamatergic and pan-GABAergic cortical and hippocampal neurons. Applying unsupervised graph-based clustering approaches using Seurat v3 (Butler et al., 2018; Stuart et al., 2019), we observed the cardinal neuronal classes (Figure 6; Figure 6—figure supplement 1A,B), including a clear segregation of MGE and CGE-derived GABAergic interneurons. Focusing attention on the 3626 neurons within the cluster corresponding to cortical and hippocampal CGE-NGFCs (Figure 6A; Figure 6—figure supplement 1C; see Materials and methods for details) these analyses reveal that essentially no mRNA encoding for Kv1.4 (Kcna4), Kv3.3 (Kcnc3) and Kv4.1 (Kcnd1) subunits indicating their unlikely contribution to I_KA in CGE-NGFCs (Figure 6B). In contrast, appreciable mRNA levels of Kcnc4, Kcnd2 and Kcnd3 mRNA that encode for Kv3.4, Kv4.2 and Kv4.3 subunits, respectively, were apparent (Figure 6B). We employed standard voltage-step protocols and digital subtraction to isolate sustained (I_KDR) and I_KA currents in mouse hippocampal NGFCs (Figure 6C) to determine the relative contributions of these subunits in mediating I_KA in NGFCs. As expected, TEA at the concentrations used (2–10 mM) reliably inhibits I_KDR (Figure 6C,D). However, I_KA in NGFCs is TEA insensitive indicating that Kv3.4 does not appreciably mediate NGFC I_KA (Rudy and McBain, 2001) despite the presence of detectable levels of mRNA (Figure 6C,D). In contrast, I_KA is largely abolished by the selective Kv4 blocker, AmmTx3 (Maffie et al., 2013) (% baseline I_KA peak after 200 nM AmmTx3 = 29.6 ± 9.9%; n = 4; Figure 6C,D). Since the expression and functional role of Kv4-channels and their associated auxiliary subunits have been well characterized in principal cells, particularly those of the hippocampus (Jerng and Pfaffinger, 2014), we compared mRNA expression in these principal cells to that seen in CGE-NGFCs. In general, Kcnd2/Kv4.2 is predominantly expressed in PCs, particularly those in the CA1 region (Figure 6E,F; see Materials and methods for details), whereas Kcnd3/Kv4.3 is mainly restricted to INs (Lien et al., 2002; Menegola et al., 2008; Rhodes et al., 2004; Serôdio and Rudy, 1998). Here, the molecular profile analyses demonstrate that at least at the transcript level revealed by RNAseq analyses, NGFCs possess comparable levels of combined mRNA encoding Kcnd2 and Kcnd3 to that seen in PCs (Figure 6E). In agreement with the RNAseq and functional data, in situ hybridization using the RNAscope platform (see Materials and methods for details) revealed the presence of detectable Kcnd2 and Kcnd3 mRNA in putative SLM hippocampal NGFCs (identified by NDNF mRNA expression; Figure 6G–I). Finally, expression of mRNA for well-known Kv4 auxiliary subunits KChIPs and DPLPs in NGFCs (particularly Kcnip1, Kcnip4, Dpp6 and Dpp10) albeit with a differential complement, is essentially similar if not higher to that observed in hippocampal pyramidal cells (Figure 6J). Interestingly, the presence of DPP6 or DPP10 subunits in the macromolecular complex of Kv4 channels is a requirement in conferring high AmmTx3 sensitivity (Maffie et al., 2013). Taken together the combined molecular, physiological and pharmacological approach demonstrate that the majority of mouse hippocampal CGE-NGFC I_KA which strongly influences intrinsic excitability of this IN subtype, is most likely mediated via a combination of Kv4.2 and Kv4.3 containing channels.

Kv4 containing channels are located along the somatodendritic axis in numerous neuronal subtypes particularly studied in pyramidal cells where they influence membrane voltage responses to incoming inputs impinging on these domains (Alfaro-Ruíz et al., 2019; Menegola et al., 2008; Trimmer, 2015). Although the subcellular expression profile has not been resolved in NGFCs to date, we evaluated their potential functional role on the dendritic integration of EPSPs elicited in NGFCs. Electrical stimulation of SLM afferents elicited excitatory input onto NGFCs that displayed paired pulse facilitation in agreement with previous observations (Chittajallu et al., 2017; Mercier et al., 2019; Price et al., 2005; Figure 7A). Bath application of AmmTx3 had no significant effect on either the amplitude of EPSCs or paired pulse ratio, whereas a marked significant increase in the depolarization elicited by the same EPSCs when monitored as EPSPs in current clamp (Figure 7A–D). These data demonstrate that pharmacological modulation of Kv4 mediated I_KA conductance in NGFCs alone, produces EPSP amplification facilitating the extent of temporal summation of excitatory afferent inputs onto NGFCs, independent to any changes in presynaptic release probability or postsynaptic glutamate receptor function. To this point in our study, we have assayed intrinsic excitability only in response to ‘artificial’ depolarizing current injection steps delivered to the soma. Based on our data thus far we would predict that if STP-SE occurs partly from a reduction of somatodendritic I_KA, an increase E-S coupling would be one physiological manifestation. Therefore, in a final set of experiments we probed the effect STP-SE on the integration of synaptic inputs onto NGFCs. EPSPs were evoked in NGFCs by electrical stimulation of SLM afferent fibers (Figure 7E). The NGFCs were then exposed to the induction protocol as previously described (Figure 3A,B) which was terminated immediately upon expression of STP-SE (Figure 7F). Electrical stimulation was recommenced at the same intensity as that used during baseline. In 10 out of 17 NGFCs tested, an increase in action potential output in response to synaptically evoked EPSPs was clearly evident (Figure 7G,H). This potentiation of EPSP-spike (E-S) coupling possessed a similar mean duration to that seen in increased intrinsic excitability upon somatic depolarization following a single round of conditioning APs indicating similar underlying mechanisms (Figure 7H vs. Figure 2G; 63 ± 11 s vs. 77 ± 23 s for duration of STP-SE and E-S coupling potentiation, respectively; n = 10, 37; p=0.9; Mann-Whitney U -Test). In most recordings a biased current was imposed to ensure resting membrane voltage was equivalent between baseline and post STP-SE induction (Vm = −63.0 ± 0.7 mV vs. −62.7 ± 0.9 mV for baseline and post-induction, respectively; n = 10; p=0.52, Mann-Whitney U-Test). It must be noted that in these experiments, synaptic parameters such as probability of release and postsynaptic receptor function could potentially influence the extent of E-S coupling. Therefore, we also interrogated spontaneous EPSC parameters following induction of multiple BF episodes that have been shown to robustly condition the expression of STP-SE (Figure 2G). Interestingly, a small but significant decrease in sEPSC amplitude after the 4^th BF episode (post-BF4) was noted, however, this would in fact negatively impact any potentiation noted and serve to minimize heightened E-S coupling. In addition, no significant changes in sEPSC frequency or tau decay were observed (Figure 7—figure supplement 1). Together analyses of spontaneous EPSCs following induction of STP-SE confirm a modulation of intrinsic properties as the main underlying cause for the increased E-S coupling observed.

Adult (p35-p60) male and female Gad2-EGFP (Tg(Gad2-EGFP)DJ31Gsat/Mmucd; GENSAT; Cat. No. 011849-UCD) or Htr3a-EGFP (Tg(Htr3a-EGFP)DH30Gsat/Mmnc; GENSAT; Cat. No. 000273-UNC) mice were used and all experiments were conducted in accordance with animal protocols approved by the National Institutes of Health (ASP# 17–045). For human data, we collected surgical specimens from ten participants (seven male; age range 22–69 years) with drug-resistant epilepsy who underwent an initial surgical procedure in which intracranial recording electrodes were implanted subdurally on the cortical surface as well as within the brain parenchyma for seizure monitoring followed by a second surgical procedure in which a resection was performed. In each case, the clinical team determined the initial placement of the contacts to localize the epileptogenic zone. Based on the intracranial recordings during the monitoring period, the clinical team then determined which electrode contacts lay over brain regions exhibiting ictal or inter-ictal activity and which contacts were electrographically normal. The surgical resections during the second surgery were designed to remove the areas of seizure onset, but in every case, also involved removal of some brain tissue determined to be electrographically normal. In eight participants, we collected electrographically normal brain specimens from the lateral temporal lobe cortex which was removed as part of a standard anterior temporal lobectomy for resection of the hippocampus. In the two remaining participants, we obtained parietal and frontal lobe brain specimens that were immediately adjacent to the epileptogenic zone but that were also determined to be electrographically normal. The NINDS Institutional Review Board (IRB) approved the research protocol (ClinicalTrials.gov Identifier NCT01273129), and informed consent for the experimental use of resected tissue was obtained from each participant and their guardians.

Both mouse and human brain tissue were sectioned at 300 μm using a Leica Vibratome 1200s in ice-cold sucrose-based cutting solution of the following composition (mM): 90 sucrose, 80 NaCl, 3.5 KCl, 24 NaHCO₃, 1.25 NaH₂PO₄, 4.5 MgCl₂, 0.5 CaCl₂, and 10 glucose, saturated with 95% O₂ and 5% CO₂. Slices were allowed to equilibrate in warm (31–33°C) sucrose cutting solution for 20–30 min after which the solution was allowed to return to room temperature where they were maintained until electrophysiological analyses. Electrophysiological recordings were performed on slices perfused with extracellular solution of the following composition (mM): 130 NaCl, 24 NaHCO₃, 1.25 NaH₂PO₄, 3.5 KCl, 1.5 MgCl₂, 2.5 CaCl₂ and 10 glucose saturated with 95% O₂ and 5% CO₂. In certain experiments (see results section for details) the extracellular solution was supplemented with one or a combination of the following drugs; 50 μM picrotoxin (Millipore Sigma; Cat. No. 80410), 2–5 μM CGP556845A (Abcam; Cat. No. ab120337), 5–10 μM bicuculline methobromide (Abcam; Cat. No. ab120109), 100 μM or 1–3 mM 4-aminopyridine (Millipore Sigma; Cat. No. A78403), 100–200 nM α -DTX (Alomone Labs; Cat No.D-350), 200 or 500 nM AmmTx3 (Alomone Labs; Cat No.STA-305). Whole cell voltage and current-clamp recordings were performed at 32–34°C with electrodes (3–5mΩ) pulled from borosilicate glass (World Precision Instruments, Sarasota, FL) filled with intracellular solution of the following composition (in mM); 150 K–gluconate, 3 MgCl₂, 0.5 EGTA, 2 MgATP, 0.3 Na₂GTP, 10 HEPES and 3 mg/ml biocytin; pH was adjusted to 7.3 with KOH and osmolarity adjusted to 280–300 mOsm. All electrophysiological recordings were performed using Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA). Signals were filtered at 4–10 kHz and digitized at 10–20 kHz (Digidata 1322A and pClamp 10.2 Software; Molecular Devices, Sunnyvale, CA, USA). For analyses of K⁺-channel currents in NGFCs, voltage clamp recordings were performed in the presence of 0.5 μm TTX and 200 μM NiCl₂ with voltage steps delivered as indicated in the manuscript. Digital leak subtraction was performed online (P/4) to remove capacitance transient and leak currents and series resistance compensation was not performed. Glutamate receptor-mediated synaptic responses (EPSCs and EPSPs) were evoked with A360 constant-current stimulus isolator (World Precision Instruments, Sarasota, FL, USA) and with stimulation electrodes pulled from borosilicate glass filled with aCSF placed in SLM in the presence 50 μM picrotoxin, 2–5 μM CGP556845A and 5–10 μM bicuculline to block GABA receptor-mediated responses. In experiments assessing the effect of AmmTX3 on EPSP integrations 5 mM QX-314Cl^- was included in the intracellular solution to prevent activation of voltage-gated sodium channels.

For neuron morphology recoveries after electrophysiological recordings, slices were post-fixed in 4% paraformaldehyde in 0.1M phosphate buffer at least overnight followed by fluorescent conjugation to biocytin and confocal images attained using a Zeiss 780 microscope (NICHD Microscopy and Imaging Core) and reconstructions drawn using Neurolucida software (MBF Bioscience, VT).

In this study, two mice aged p35 and p40 were anesthetized with isoflurane and intracardially perfused with 0.1M phosphate buffer, Ph7.4, followed by 4% paraformaldehyde/0.1M phosphate buffer. Brains were removed and post-fixed in 4% paraformaldehyde/0.1M phosphate buffer overnight, then embedded in Tissue-Tek O.C.T. compound {Sakura Finetek, USA Cat. No 4583} and frozen. 10–14 μM horizontal sections containing mid-ventral hippocampus were cut using a Leica CM3050-S cryostat, mounted on microscope slides and stored at −80°C until use. Fluorescent in situ hybridization was performed according to the RNAscope (Advanced Cell Diagnostics USA, Newark, CA) protocol using the Multiplex Fluorescent Reagent Kit v1 (Cat. No 320851). Probes employed were Mm-Ndnf (Cat No. 447471-C2), Mm-Kcnd2 (Cat No. 452581) and Mm-Kcnd3 (Cat No. 573641). Sections were cover-slipped and confocal images taken using Zeiss 780 microscope (NICHD Microscopy and Imaging Core). Z-stack images were acquired using a 20x/0.8 Plan Apochromat or 40x/1.3 Plan Neofluar objective and were rendered as maximum intensity injections.

To analyze the Allen Institute mouse dataset of the single-cell transcriptomes of ~76,000 cells from >20 areas of mouse cortex and hippocampus, we downloaded the transcript.tome/HDF5 file (https://portal.brain-map.org/atlases-and-data/rnaseq) and subsequently converted into Seurat v3-compatible format based on the instructions provided in the Allen Institute Portal (https://portal.brain-map.org/atlases-and-data/rnaseq/protocols-mouse-cortex-and-hippocampus) and custom scripts in R package. Basic processing and visualization of the scRNA-seq data were performed with the Seurat v3 in R (Butler et al., 2018; Stuart et al., 2019). Briefly, data were preprocessed by removing cells with less than 200 detectable genes and with reads mapping to mitochondrial genes comprising more than 0.05% of total reads. Gene expression counts were then log normalized and the resulting expression matrix was scaled to 10,000 transcripts per cell. Next, the expression values of each gene across all the cells (z-score transformation) were utilized for dimensionality reduction. These were performed using the NormalizeData() and ScaleData() functions in Seurat with default parameters. Next, we identified the top 2000 genes that exhibited high variability across cells and therefore represent unique features of the cell clusters, for downstream analysis using the FindVariableGenes() function, by applying selection.method = ‘vst’ feature. Next, principal component analysis (PCA) was performed, and the top 20 principal components were stored. We clustered the cells using the FindClusters() with a clustering resolution set to 0.6. This approach generated a total of 38 clusters. Using the FindAllMarkers() function at cutoff of genes expressed in 25% of cells, at a logfc.threshold = 0.25, we identified the genes that are expressed in each cluster. For each cell type we used the well-established marker genes that uniquely established the cell identity, as previously described in the literature. Notably, Snap25 positive clusters were annotated as neurons, and the Snap25 negative clusters were annotated as non-neurons; among the neurons, Slc17a7 was indicative of glutamatergic and Gad1 expression indicative of GABAergic neurons (Figure 6—figure supplement 1A). More specifically, the Gad1 clusters expressing a combination of Lhx6, Pvalb, Sst, Hapln1 are indicative of MGE-derived GABAergic interneurons, and a combination of Lhx6-negative, Prox1, Vip, Calb2, Cnr1, Hapln1, Ndnf are indicative of CGE-derived GABAergic interneurons (Figure 6—figure supplement 1B). Relevant for the present study, specific-markers include, but are not limited to: Gad1, Prox1, Hapln1, Lamp5, Ndnf, but Lhx6 negative, for the CGE-NGFCs (Mayer et al., 2018; Tasic et al., 2018); Slc17a7, Wfs1, Fibcd1, Crym for the hippocampal pyramidal neurons (Cembrowski et al., 2016a; Cembrowski et al., 2016b). The expression of Kcn genes and auxiliary subunits were obtained for the above clusters, and the Violin/Split-violin plots were plotted via custom scripts in the R package using ggplot. All scripts employed are available at GitHub https://github.com/mahadevanv/Chittajallu-et-al-Allen-Mouse-CGE-NGFCs (Chittajallu, 2020; copy archived at https://github.com/elifesciences-publications/Chittajallu-et-al-Allen-Mouse-CGE-NGFCs).

Each dataset that underwent statistical analyses was first tested for normal distribution using the Shapiro-Wilk test. For data that were deemed normally distributed parametric paired or unpaired t-tests were employed, otherwise Mann-Whitney U-tests or Wilcoxin-sign Rank tests were used. Due to the low n numbers in some human datasets (Figure 2C, n = 3; Figure 3H,I, n = 4), Shapiro-Wilk or non-parametric tests could not be performed and in these cases we defaulted to parametric Students T-test. In the figure panels statistical significance is denoted as follows; *p<0.05, **p<0.01, ***p<0.001, ns not significant p>0.05. The type of statistical test and exact p-values are indicated in the figure legends where appropriate. n values throughout the manuscript correspond to the number of cells tested and for each set of experiments and are from a minimum of 3 mice and two human subjects. Throughout the manuscript figures, data in black/grey and orange were from mouse and human, respectively. In box plots, the data are represented as the median and mean (line and symbol, respectively), the 25^th and 75^th percentile (bottom and top of the box, respectively), 10th and 90^th percentiles (bottom and top of whiskers, respectively). In scatter and line plots all data are expressed as the mean ± s.e.m.