Balanced network dynamics underlie slow autonomous oscillations

Numerous synaptic and intrinsic membrane mechanisms have been proposed for generating oscillatory activity in the hippocampus. Few studies, however, have directly measured synaptic conductances and membrane properties during oscillations. The time course and relative contribution of excitatory and inhibitory synaptic conductances, as well as the role of intrinsic membrane properties in amplifying synaptic inputs, remain unclear. To address this issue, we used an isolated whole hippocampal preparation that generates autonomous low-frequency oscillations near the theta range (3-12 Hz). Using 2-photon microscopy and expression of genetically-encoded fluorophores, we obtained on-cell and whole-cell patch recordings of pyramidal cells and fast-firing interneurons in the distal subiculum. Pyramidal cell and interneuron spiking shared similar phase-locking to LFP oscillations. In pyramidal cells, spiking resulted from a concomitant and balanced increase in excitatory and inhibitory synaptic currents. In contrast, interneuron spiking was driven almost exclusively by excitatory synaptic current. Thus, similar to tightly balanced networks underlying hippocampal gamma oscillations and ripples, balanced synaptic inputs in the whole hippocampal preparation drive highly phase-locked spiking at the peak of slower network oscillations. The timescale for hippocampal theta oscillations has been attributed to intrinsic membrane properties that impart resonance and rebound-spiking dynamics to neurons. Autonomous oscillations in the whole hippocampal preparation, which occur near the iv theta frequency band, are generated through a balanced excitation-inhibition mechanism that does not arise from rebound-spiking in pyramidal cells. In order to determine the timescale for autonomous oscillations, we injected pyramidal cells with artificial membrane conductance steps that emulate synaptic input during oscillations in order to evoke low spike rates mimicking those seen in autonomous oscillations. We find that refractory dynamics, particularly those influenced by potassium-channel conductance, inhibit spiking when a second conductance step is applied within 80 ms after the first conductance step. The delay at which spiking and potassium conductances recover matches the timescale of theta (80-320 ms). Thus, the timescale for autonomous oscillations likely arises from an increase in potassium-channel conductance that limits spiking during depolarization, rather than membrane resonance properties that amplify theta frequency inputs.

BALANCED NETWORK DYNAMICS UNDERLIE SLOW AUTONOMOUS OSCILLATIONS by Feliks M Royzen A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Interdepartmental Program in Neuroscience The University of Utah December 2019 Copyright © Feliks M Royzen 2019 All Rights Reserved The University of Utah Graduate School STATEMENT OF DISSERTATION APPROVAL Feliks M Royzen The dissertation of has been approved by the following supervisory committee members: John Andrew White , Chair 12/14/2018 Karen S. Wilcox , Member 12/14/2018 Alan Dale Dorval II , Member 12/14/2018 Dale Matthew Wachowiak , Member 12/14/2018 Alla R. Borisyuk , Member 12/14/2018 and by David Krizaj the Department/College/School of and by David B. Kieda, Dean of The Graduate School. Date Approved Date Approved Date Approved Date Approved Date Approved , Chair/Dean of Neuroscience ABSTRACT Numerous synaptic and intrinsic membrane mechanisms have been proposed for generating oscillatory activity in the hippocampus. Few studies, however, have directly measured synaptic conductances and membrane properties during oscillations. The time course and relative contribution of excitatory and inhibitory synaptic conductances, as well as the role of intrinsic membrane properties in amplifying synaptic inputs, remain unclear. To address this issue, we used an isolated whole hippocampal preparation that generates autonomous low-frequency oscillations near the theta range (3-12 Hz). Using 2-photon microscopy and expression of genetically-encoded fluorophores, we obtained on-cell and whole-cell patch recordings of pyramidal cells and fast-firing interneurons in the distal subiculum. Pyramidal cell and interneuron spiking shared similar phase-locking to LFP oscillations. In pyramidal cells, spiking resulted from a concomitant and balanced increase in excitatory and inhibitory synaptic currents. In contrast, interneuron spiking was driven almost exclusively by excitatory synaptic current. Thus, similar to tightly balanced networks underlying hippocampal gamma oscillations and ripples, balanced synaptic inputs in the whole hippocampal preparation drive highly phase-locked spiking at the peak of slower network oscillations. The timescale for hippocampal theta oscillations has been attributed to intrinsic membrane properties that impart resonance and rebound-spiking dynamics to neurons. Autonomous oscillations in the whole hippocampal preparation, which occur near the theta frequency band, are generated through a balanced excitation-inhibition mechanism that does not arise from rebound-spiking in pyramidal cells. In order to determine the timescale for autonomous oscillations, we injected pyramidal cells with artificial membrane conductance steps that emulate synaptic input during oscillations in order to evoke low spike rates mimicking those seen in autonomous oscillations. We find that refractory dynamics, particularly those influenced by potassium-channel conductance, inhibit spiking when a second conductance step is applied within 80 ms after the first conductance step. The delay at which spiking and potassium conductances recover matches the timescale of theta (80-320 ms). Thus, the timescale for autonomous oscillations likely arises from an increase in potassium-channel conductance that limits spiking during depolarization, rather than membrane resonance properties that amplify theta frequency inputs. iv TABLE OF CONTENTS ABSTRACT....................................................................................................................... iii LIST OF FIGURES .......................................................................................................... vii Chapters 1 INTRODUCTION ........................................................................................................... 1 The Hippocampus and Theta Oscillations ................................................................. 1 Mechanisms of Theta Rhythmogenesis ..................................................................... 2 Slow Autonomous Oscillations in the Whole Hippocampal Preparation .................. 4 Fast Frequency Oscillations in the Hippocampus ...................................................... 5 2 BALANCED SYNAPTIC CURRENTS UNDERLIE LOW-FREQUENCY OSCILLATIONS IN THE SUBICULUM ......................................................................... 7 Introduction ................................................................................................................ 7 Materials and Methods............................................................................................... 9 Ethics statement ................................................................................................ 9 Tissue preparation ............................................................................................. 9 Visualizing hippocampal pyramidal cells and parvalbumin-expressing interneurons..................................................................................................... 10 Electrophysiology ........................................................................................... 10 Data analysis ................................................................................................... 11 Results ...................................................................................................................... 14 Pyramidal cells and fast-firing interneurons share similar spike phase-locking profiles ............................................................................................................ 14 Membrane voltage is correlated to LFP oscillations....................................... 16 Pyramidal cells and interneurons experience a high conductance state during the peak of intrinsic oscillations ..................................................................... 18 Pyramidal cells, but not interneurons, are driven by balanced inputs during hippocampal oscillations................................................................................. 20 Oscillations generate a small steady-state conductance in pyramidal cells, but not interneurons .............................................................................................. 22 Discussion ................................................................................................................ 23 Balanced excitation and inhibition and the generation of network oscillations ...................................................................................................... 23 Comparisons to intrinsic oscillations in other hippocampal regions .............. 26 Comparison of intrinsic subicular oscillations to tightly balanced networks . 27 3 REFRACTORY DYNAMICS IN SUBICULAR PYRAMIDAL CELLS .................... 38 Introduction .............................................................................................................. 38 Materials and Methods............................................................................................. 40 Ethics statement .............................................................................................. 40 Tissue preparation ........................................................................................... 40 Visualizing hippocampal pyramidal cells ....................................................... 40 Electrophysiology ........................................................................................... 41 Data analysis ................................................................................................... 42 Results ...................................................................................................................... 44 Recurrent excitation drives synchronous pyramidal cell spiking ................... 44 Feedback inhibition limits excitatory drive in pyramidal cells ....................... 45 Excitatory drive is dependent on glutamatergic signaling .............................. 46 Conductance-evoked spiking recovers on a similar timescale to theta oscillations ...................................................................................................... 46 Theta timescale of spiking recovery matches recovery of potassium-channel conductances ................................................................................................... 47 Discussion ................................................................................................................ 49 Comparison of autonomous oscillations to established mechanisms of hippocampal theta rhythmogenesis ................................................................. 50 Role of fast-spiking interneurons in setting the timescale of hippocampal oscillations ...................................................................................................... 51 Role of potassium-channel conductance in setting the timescale of hippocampal oscillations................................................................................. 52 4 CONCLUSION .............................................................................................................. 63 Major Findings ......................................................................................................... 63 Future Directions ..................................................................................................... 66 REFERENCES ................................................................................................................. 69 vi LIST OF FIGURES Figures 2.1 The whole hippocampal preparation generates stable and long-lasting autonomous oscillations. ....................................................................................................................... 30 2.2 PYR and PV cells spike primarily during the peak of LFP oscillations in the subiculum. ......................................................................................................................... 31 2.3 Intracellular current-clamp recordings of PYR and PV cells during intrinsic hippocampal oscillations................................................................................................... 32 2.4 Membrane properties of PYR and PV cells during autonomous oscillations............. 33 2.5 Membrane voltage of PYR and PV cells is highly correlated to the LFP. ................. 34 2.6 Quantification of membrane conductance as a function of LFP phase using voltage clamp across multiple holding voltages. ........................................................................... 35 2.7 A correlated rise and fall of excitation (gE) and inhibition (gI) in PYR cells results in balanced synaptic currents at the peak of the oscillation cycle, whereas in PV cells synaptic inputs are largely dominated by excitation. ........................................................ 36 2.8 Application of glutamatergic antagonist DNQX abolishes oscillations in both the LFP and membrane voltage of PYR and PV cells. ................................................................... 37 3.1 Bath applications of gabazine unveils the recurrent excitatory network in pyramidal cells that drives intrinsic hippocampal oscillations. ......................................................... 54 3.2 Gabazine transforms rhythmic activity in LFP and membrane voltage of pyramidal cells into random and more intense fluctuations............................................................... 56 3.3 Subsequent bath application of DNQX after gabazine blocks recurrent excitation in pyramidal cells. ................................................................................................................. 57 3.4 Injection of conductance to elicit spiking in quiescent pyramidal cells alters subsequent spiking response to injected conductance. ..................................................... 59 3.5 Injection of either current or conductance to elicit spiking in quiescient pyramidal cells has a similar effect on refractory dynamics. ............................................................. 60 3.6 Spiking characteristics that are determined by potassium-channel conductances recover within the timescale of the theta frequency band................................................. 61 viii CHAPTER 1 INTRODUCTION The Hippocampus and Theta Oscillations The hippocampus is a brain region commonly associated with learning and memory processes. Damage to hippocampal tissue through physical trauma, and diseases such as Alzheimer’s or epilepsy, substantially impair the encoding and retrieval of episodic memories (Burgess et al., 2002). Episodic or autobiographical memories are created through the encoding of spatial information associated to stimuli of various modalities (Ranganath, 2010). This can be observed behaviorally during conditioning paradigms, when stimuli such as fear and reward become bound to environmental cues (Maren et al., 2013). On a cellular level this binding of information occurs through mechanisms of long-term potentiation in dendritic spines (Lynch, 2004). In the hippocampus, spatial information is represented as a sequence of spiking neurons whose firing rate and spike timing is correlated to particular areas in the animal’s environment. The firing of these place cells occurs during exploratory behavior during which a prominent 3-12 Hz “theta” rhythm appears in the hippocampus (O’Keefe & Dostrovsky, 1971). Place cells fire at the peaks of intracellularly-recorded theta oscillations, but are phase-shifted to earlier parts of the extracellularly-recorded theta cycle as the animal traverses its environment (O’Keefe & Reese, 1993; Harvey et al., 2 2009). This spatially-correlated firing of hippocampal place cells is then thought to encode for spatial information, in an egocentric or autobiographical manner (Cei et al., 2014). Current mechanistic models posit that theta oscillations are generated via extrinsic sources that pace hippocampal neurons at theta frequencies; however, our study in Chapter 2 reveals local network properties within the subiculum of the isolated hippocampus that underlie self-generated theta-like oscillations. The hippocampal formation is divided into cornu ammonis (CA) areas 1-4, the dentate gyrus and subiculum which are further subdivided by layers containing distinct neuronal populations. The focus of this dissertation is on the subiculum which is bordered by area CA1 and entorhinal cortex and where in vitro theta-like oscillations are believed to originate (Jackson et al., 2011; Jackson et al., 2014). The subiculum contains recurrently connected pyramidal cells, which are the principal excitatory cells in the hippocampus (Harris et al., 2001). In addition, fast-firing interneurons, although a smaller percentage of the neuronal population, contact a large degree of pyramidal cell bodies and exert significant influence in pacing spontaneous pyramidal spiking (Cobb et al., 1995; Stark et al., 2013; Amilhon et al., 2015). Our data from these two main cell types are sufficient to determine the network dynamics underlying theta-like oscillations in the subiculum of the whole hippocampal preparation. Mechanisms of Theta Rhythmogenesis Our current understanding of theta rhythmogenesis is derived from in vivo observations that inactivation of the medial septum leads to a substantial drop in theta 3 power (Yoder & Pang, 2005; Boyce et al., 2016). As a result, GABAergic and cholinergic projections from the medial septum have been studied to determine their pacing action on the hippocampus. At certain levels of activation, acetylcholine receptors in the hippocampus can generate theta oscillations in vitro (Fellous & Sejnowski, 2000); however, the slow activation of these receptors indicates a modulatory, rather than pacemaking, role (Hasselmo & Fehlau, 2001). In contrast, stimulation of GABAergic projections paces spiking in post-synaptic targets at theta frequencies (Toth et al., 1997). Hippocampal neurons are paced at theta frequencies by rebound spiking after the cessation of inhibition (Cobb et al., 1995; Stark et al., 2013). This post-inhibitory rebound spiking is hypothesized to result from intrinsic membrane properties of hippocampal cells that contain the hyperpolarization-activated Ih current. The timescale of Ih activation matches the theta frequency band and it imparts subthreshold theta resonance and rebound-spiking dynamics to pyramidal cells and a subset of interneurons (Maccaferri & McBain, 1996; Hu et al. 2002; Zemankovics et al., 2010). Fast-spiking interneurons, however, do not exhibit these properties (Pike et al., 2000, Zemankovics et al., 2010) but have also been implicated in theta oscillations Cobb et al., 1995; Stark et al., 2013; Amilhon et al., 2015). Currently, only a single study has carried out simultaneous intracellular and LFP recordings during type I theta oscillations (Harvey et al., 2009). Because synaptic conductance values were not measured, it is difficult to confirm a rebound-spiking mechanism, especially since pyramidal cell spiking was shown to occur as a result of substantial depolarizations to spike threshold. In order to fully understand the local interactions of excitation and inhibition in generating theta oscillations, in Chapter 2, we 4 detail synaptic conductances and spiking behavior of both excitatory and inhibitory cells during theta-like oscillations in a reduced hippocampal preparation. Slow Autonomous Oscillations in the Whole Hippocampal Preparation In order to elucidate network structures and, to some degree, functions of the hippocampus, reduced in vitro preparations such as hippocampal slices and whole hippocampal tissue have been frequently studied. Under favorable conditions, but without the use of pharmacology, it was discovered that the whole hippocampal preparation can self-generate oscillations within the theta frequency band (Goutagny et al., 2009). Past studies of area CA1 of the whole hippocampal preparation (Huh et al., 2016), found almost no excitatory drive in pyramidal cells. Instead post-inhibitory rebound spiking was determined to be the mechanism by which pyramidal cells were driven to fire, similar to in vivo studies (Stark et al., 2013). Fast-firing interneurons exhibited outof-phase spiking in relation to pyramidal cells, contributing to the inhibitory conductance from which pyramidal cells rebound. The out-of-phase firing of these neuronal subtypes can also be seen during type II theta oscillations which occur during anesthetic states (Klausberger & Somogyi, 2008). In contrast, movement-related theta oscillations (type I) exhibit in-phase firing of pyramidal cells and interneurons (Fox et al., 1986; Skaggs et al., 1996; Csicsvari et al., 1999). Type I theta is also insensitive to cholinergic antagonists, unlike type II theta (Kramis et al., 1975). These differences could potentially occur from separate mechanism for theta rhythmogenesis, and requires further study. The atropine-insensitivity of the whole hippocampal preparation (Goutagny et al., 2009), along with in-phase spiking 5 behavior of both excitatory and inhibitory neurons detailed in Chapter 2, more closely resembles properties of type I theta. Fast Frequency Oscillations in the Hippocampus A common feature of network oscillations not previously associated with theta rhythmogenesis is balanced excitation-inhibition dynamics. During in vivo hippocampal oscillations at high frequencies such as gamma (30-80 Hz) and ripples (~200 Hz), pyramidal cells receive excitatory and inhibitory currents that are tightly balanced (Atallah & Scanziani, 2009; English et al., 2014; Gan et al., 2017). Feedback inhibition scales proportionally with recurrent excitation preventing runaway spiking behavior. The timescale for these oscillations likely arises from the fast kinetics of synaptic transmission. Brain regions associated with the hippocampus such as entorhinal cortex also display balanced synaptic inputs within excitatory cells, but at lower frequencies (< 1 Hz) compared to theta oscillations (Tahvildari et al., 2012). During slower oscillations, neurons exhibit more intense spiking behavior, and likely require additional feedback mechanisms, such as sodium-channel inactivation or synaptic depression, to quell the recurrent excitation. Another possible mechanism is the build-up of potassium-channel conductances during periods of spiking, which is posited to set the timescale of lowfrequency oscillations (Sanchez-Vives & McCormick, 2000; Compte et al., 2003). An increase in potassium conductance during spiking reduces neuronal excitability decreasing the probability of further action potentials, until the conductance recovers to baseline levels. The timescale of recovery sets a refractory period during 6 which spiking behavior, and thus recurrent excitation, is suppressed. In accordance, the degree of spiking in excitatory neurons during low-frequency oscillations has a strong correlation to the refractory period between bouts of depolarization, setting the timescale for oscillations (Sanchez-Vives & McCormick, 2000; Compte et al., 2003). In Chapter 3, we detail changes in the refractory dynamics of pyramidal cells that occur as a result of their low-rate spiking behavior during autonomous oscillations. We show that this change in refractory dynamics suppresses spiking behavior within the time period of a theta cycle, potentially setting the timescale for slow autonomous oscillations. CHAPTER 2 BALANCED SYNAPTIC CURRENTS UNDERLIE LOW-FREQUENCY OSCILLATIONS IN THE SUBICULUM Introduction Balanced excitation-inhibition (E/I) during activity is a common feature of neuronal networks (Anderson et al., 2000; Haider et al., 2006; Rudolph et al., 2007; Okun & Lampl, 2008). During evoked activity and local field potential (LFP) oscillations, balanced E/I currents underlie depolarizations in hippocampal and cortical excitatory principal cells in vivo (Wilent & Contreras, 2005; Atallah & Scanziani, 2009; Gan et al., 2017). A key characteristic of balanced E/I inputs is that they maintain membrane voltage near spike threshold and permit precise control over spike output rate (Shu et al., 2003; Wehr & Zador, 2003; Gabernet et al., 2005; Higley & Contreras, 2006). Additionally, recruitment of inhibition over a range of network activity levels likely prevents runaway excitation in networks with strong recurrent connections such as the subiculum (Harris et al., 2001; Drexel et al., 2017). In the hippocampus, fast frequency oscillations such as gamma (30-80 Hz) and ripples (~200 Hz) have been shown to have properties consistent with generation through correlated and balanced synaptic currents (Atallah & Scanziani, 2009; English et al., 2014; Gan et al., 2017). For lower frequencies such as theta (3-12 8 Hz), however, balanced network activity has not been recognized as a potentially contributing factor to this oscillation range. Instead, mechanisms involving intrinsic membrane properties such as resonance and rebound spiking have been favored (Cobb et al., 1995; Hu et al., 2002; Zemankovics et al., 2010; Stark et al., 2013). We sought to better understand the biophysical factors generating low-frequency or theta-like rhythms in the isolated hippocampal preparation (Goutagny et al., 2009). Utilizing the hippocampal preparation, which generates autonomous low-frequency oscillations (3 – 7 Hz), we targeted pyramidal cells (PYR) and fast-firing parvalbumin (PV)-expressing interneurons to enable visually-guided patch-clamping using 2-photon microscopy. Our recordings focused on the distal subiculum, which appears to serve as the pacemaker of the theta-like rhythm in this preparation (Jackson et al., 2011; Jackson et al., 2014). During network oscillations, both PYR and PV cells in the hippocampal preparation spiked at the peak of the LFP. Our voltage- and current-clamp recordings indicate that intracellular depolarizations that correlate to LFP oscillations in PYR cells are produced through a concomitant increase in excitatory and inhibitory synaptic currents. A near simultaneous and balanced rise in both currents depolarizes PYR cells near spike threshold and leads to spike phase-locking at the peak of the network rhythm. In contrast, in PV cells, inhibition is nearly absent and spike activity is driven largely by excitatory currents. Nevertheless, PV cells fire in phase with pyramidal cells. The near simultaneous rise in excitation and inhibition, along with similar PYR and PV cell spikephase, indicates that a balanced and correlated increase in excitation and inhibition underlies spike phase-locking during network oscillations in subiculum. Autonomous 9 oscillations in this region, therefore, share similarities to previous network mechanisms establishing hippocampal gamma oscillations and ripples (Atallah & Scanziani, 2009; English et al., 2014; Gan et al., 2017). Materials and Methods Ethics statement All experimental protocols were approved by the Boston University Institutional Animal Care and Use Committee. Tissue preparation Transgenic mice (P12-P18) of either sex were anesthetized with isoflurane and decapitated into a solution of ice-cold artificial cerebrospinal fluid (aCSF), oxygenated, and buffered to pH 7.3 with 95% O2/ 5% CO2. ACSF consisted of (in mM): NaCl (126), NaHCO3 (24), D-glucose (10), KCl (4.5), MgSO4 (2), NaH2PO4 (1.25), CaCl2 (1), and ascorbic acid (0.4) (Sigma-Aldrich). Both hippocampi were extracted and incubated in room temperature aCSF for 30-120 minutes. Subsequently, one hippocampus was transferred to a custom recording chamber that was equipped to recirculate oxygenated/buffered aCSF (add. 1 mM CaCl2) at a flowrate of 20-25 mL/min and a temperature of 31.5°C ± 0.2°C. 10 Visualizing hippocampal pyramidal cells and parvalbuminexpressing interneurons To visualize cells and glass pipettes in the distal subiculum, we used a 2-photon microscope (Bergamo II Series, ThorLabs) equipped with a Ti:Sapphire laser (920 nm; Chameleon Ultra, Coherent) to enable excitation of fluorescent markers through a 20x, NA 0.95 objective lens (Olympus). To detect fluorescence from intracellular tdTomato and Alexa 488 in the pipette, we used two photo-multiplier tubes (Hamamatsu) that receive either red or green optically filtered emission, respectively. Transgenic mice were generated by crossing CaMKIIa- or PV-Cre promoter mice (Jackson Labs, stock # 005359 or 008069) with the loxP-flanked tdTomato reporter mice (Jackson Labs, stock # 007914)(Madisen et al., 2010). The CamkIIa- promoter is specific for expression in principal cells (Tsien et al., 1996), while the PV- promoter targets interneurons expressing parvalbumin (Hippenmeyer et al., 2005). Electrophysiology A pulled-glass electrode (0.5-1.5 MΩ) containing aCSF and Alexa Fluor 488 hydrazide (.0029% w/v; Thermofisher Scientific) was used to record LFP oscillations in the distal subiculum at the mid-septotemporal axis. The LFP electrode was advanced through the hippocampal tissue using a micromanipulator (Sutter Instrument) until oscillations reached peak amplitude, reportedly around the stratum radiatum (Goutagny et al., 2009). A second electrode (3.5-6.5 MΩ) with intracellular solution and Alexa 488 (.0014% w/v) was placed in close proximity to patch-clamp fluorescing pyramidal cells. The intracellular solution was comprised of (in mM): K-gluconate (136), HEPES (10), 11 KCL (4), EGTA (0.2), MgCl2 (2), diTrisPhCr (7), Na2ATP (4), and TrisGTP (0.3); buffered to pH 7.3 with KOH (Sigma-Aldrich). Signals were recorded with Clampex (v. 10.6; Molecular Devices) using a Multiclamp 700B amplifier and Digidata 1440A digitizer (Molecular Devices) at a sampling rate of 10 kHz, with the LFP filtered between 1-50 Hz. Signals were processed and analyzed using integrated MATLAB (R2017b; Mathworks) functions and custom scripts. Pipette offset was compensated prior to achieving on-cell patch recordings. Once the patch electrode was sealed to the cell membrane (>1 GΩ seal), pipette capacitance was compensated. Using this configuration, action potentials were recorded as capacitive current in voltage-clamp mode. During intracellular recordings, access resistance (15-30 MΩ) was compensated and monitored every 5-10 minutes to not exceed 65 MΩ. For measurements of intracellular currents (voltage-clamp), QX-314 chloride (1 mM, Tocris Bioscience) was added to the intracellular solution to block non-linearities associated with sodium channels. For voltage-clamp recordings, series resistance was compensated up to 75%. To eliminate oscillations in the hippocampal preparation, bath application of DNQX disodium salt (20 µM, Tocris Bioscience) was utilized. Data analysis Data in text and figures are reported as median and interquartile range (25-75%) due to the non-normal distribution (p < 0.05, one-sample Kolmogorov-Smirnov test). Likewise, the statistical tests used are non-parametric. Reported values of membrane 12 voltage are not adjusted for liquid junction potential; actual values are 11.5 mV more negative. All data from neurons used in this study (n = 38) had accompanying on-cell spiking behavior documented and analyzed using the last 100 oscillation cycles of the oncell recordings. Oscillation cycles were detected by peaks in the derivative of the LFP phase, which occurred when the phase wraps from -180 to 180 degrees (cycle beginning/end). The phase of the LFP was determined with a Hilbert transform function implemented in Matlab. Portions of the LFP trace that are relatively flat tend to create many of these peaks over a short duration, so data from inter-peak intervals less than 100 ms (cycles >10 Hz) were not considered. Recordings from the end of experimental sessions (>1hr) were discarded when the LFP became variable in amplitude and spiking was inconsistent (<1 spk/cycle). For the duration of the last 100 oscillation cycles of an on-cell recording, the LFP characteristics were also determined (Fig. 2.1). Frequency analysis of the LFP (Fig. 2.1A, 2.1Ai) utilized the multi-taper mtspecgramc function of the Chronux toolbox [http://chronux.org/], with an averaging window of 2.5 second moving every 1.67 seconds. This method was also applied to determine the frequency content of synaptic currents at -60 mV holding voltage. From whole-cell current-clamp recordings (Fig. 2.3), threshold values of spikes were determined using the peak of the second derivative of membrane voltage. Spike half-width was calculated as the time between midpoints (voltage between threshold and spike peak) of the rising and falling phase. If a neuron spiked more than once in a cycle, only the first spike was analyzed to avoid spike history dependence in the measurement. The mean from 10 cycles was used for each cell. Input resistance was calculated by 13 measuring the mean change in membrane voltage (~5 mV) in response to repeated steps of injected current (50x, 1 second on/off) at the resting membrane potential (Fig. 2.4). Membrane time constant was determined using a non-linear regression fit of the membrane voltage between the start of the current step and 100 ms after the start of the current step. Additionally, steps of various current (~30 steps; 10 seconds on/ 2 seconds off) were injected into neurons to drive membrane voltage throughout the cell’s dynamic range (Fig. 2.5). The Matlab function xcorr was used to determine the normalized crosscorrelation coefficient between the LFP and membrane voltage during current steps (Fig. 2.5Ai, 2.5Bi). To create averaged membrane voltage traces, for each current step, the membrane voltage trace was binned in accordance to the LFP phase (Fig. 2.5A, 2.5B). Similarly, for voltage-clamp recordings, membrane current traces were binned by the LFP phase for each holding-voltage step (10 steps, -75 mV to -30 mV; Fig. 2.6A, 2.6B). Then for every bin, the membrane current and voltage values were used to generate a linear I-V curve, whose slope is a measure of the cell’s conductance at a particular phase of the LFP (Fig. 2.6Ai, 2.6Bi). Total synaptic conductance as a function of LFP phase was determined by the difference in I-V slope, or conductance, between the first (baseline) bin and subsequent phase-dependent bins of conductance. The voltage at which the two I-V curves intersect is indicative of the net synaptic reversal potential, and the ratio of excitatory (AMPA-mediated) to inhibitory (GABAa -mediated) conductance as part of the total phase-dependent synaptic conductance (Fig. 2.7). Reversal potentials for GABAa were calculated from the Nernst potential equation for [Cl-] of the aCSF and intracellular solution. Similarly, the reversal potential for AMPA was calculated from the 14 mean reversal of [Na+] and [K+]. To extrapolate I-V curves, a first-order linear fit was achieved with the polyfit function in Matlab. Results Pyramidal cells and fast-firing interneurons share similar spike phase-locking profiles To assess the role of hippocampal neurons in generating intrinsic oscillations, we recorded the local field potential (LFP) in the distal subiculum of the isolated hippocampal preparation. Due to the thickness of the hippocampal preparation, traditional visualization using light microscopy is difficult. Instead, using 2-photon imaging, we targeted excitatory pyramidal (PYR) cells and fast-firing interneurons (PV) for cellattached and subsequent whole-cell patch recordings using selective expression of the red fluorophore tdTomato under control of CaMKIIa and PV promoters, respectively. Although oscillations in the isolated hippocampus tend to slow down over time (Fig. 2.1A, and Goutagny et al., 2009), spiking in PYR and PV cells remained strongly phase-locked (Fig. 2.2Ai, 2.2Bi) to the peak of the LFP (PYR & PV Mdn = -12.3°, IQR = 12.4°; n = 38, Fig. 2.1Bi) regardless of oscillation frequency (PYR & PV Mdn = 3.02 Hz, IQR = 1.13 Hz, n = 38; Fig. 2.1Ai and Fig. 2.2Aii, 2.2Bii). PYR and PV cells have been shown to be highly coherent with theta oscillations in area CA1, both in vivo and in the hippocampal preparation (Csicsvari et al., 1999; Klausberger & Somogyi, 2008; Huh et al., 2016). To understand the synaptic and intrinsic mechanisms establishing spike-phase in PYR and PV cells, we first carried out on-cell recordings and used the capacitive current associated with spikes (Fig. 2.2A, 2.2B) to determine the spike-phase relative to 15 the LFP without perturbing the intracellular environment. In both PYR (n = 22) and PV cells (n = 16), spikes were phase locked (-20.7°, Fig. 2.2Ai; -23.3°, Fig. 2.2Bi) near the peak of the LFP, with the spike-phase not differing significantly between the two cell types (p > .10, Kuiper’s test), and independent of the LFP frequency (PYR Spearman’s rho = 0.09, p = 0.69, Fig. 2.2Aii ; PV Spearman’s rho = 0.07, p = 0.80, Fig. 2.2Bii). The spike rate in PYR cells (Mdn = 1.73 spk/cycle, IQR = 1.19 spk/cycle; Fig. 2.2Aiii) and PV cells (Mdn = 3.65 spk/cycle, IQR = 4.48 spk/cycle; Fig. 2.2Biii) was not significantly different (p = 0.08, Wilcoxon rank-sum test). Unfortunately, the low spike rate did not allow us to establish a burst-firing phenotype in subicular PYR cells, which have been shown to be differentially modulated during oscillations compared to regularspiking pyramidal cells (Bohm et al., 2015). Furthermore, burst-firing may be substantially modulated by synaptic conductances during oscillations. Therefore, a more targeted approach using synaptic blockers and responses to step currents is necessary to determine which pyramidal cells exhibit the burst-firing phenotype and whether those cells differ during intrinsic hippocampal oscillations. Although we encountered a few pyramidal cells that exhibited bursting in response to synaptic excitation, like the majority of other pyramidal cells, these cells also spiked at the peak of oscillations. Next, we used whole-cell patch clamp to quantify membrane properties of neurons actively participating in the network oscillation. Consistent with previous work (Huh et al., 2016), spikes in PYR cells (n = 14, Fig. 2.3Ai) had significantly (p < .01, Wilcoxon rank-sum test) longer half-widths (Mdn = 0.80 ms, IQR = 0.10 ms) and a lower voltage threshold (Mdn = -48.8 mV, IQR = 2.2 mV) compared to PV cells (Mdn = 0.30 ms, IQR = 0.10 ms; Mdn = -42.3 mV, IQR = 5.4 mV; n = 8, Fig. 2.3Bi). To determine 16 input resistance and membrane time constant, we applied repeated presentations of a small hyperpolarizing current step at the resting membrane voltage (PYR Mdn = -61.4 mV, IQR = 4.4 mV, Fig. 2.4A, 2.4Ai; PV Mdn = -54.7 mV, IQR = 6.4 mV, Fig. 2.4B, 2.4Bi). To mitigate the effects of spiking non-linearities on measures of resistance, we averaged the membrane voltage response using non-spiking regions during the LFP trough (~180°, Fig. 2.4A, 2.4B). When assessed near the resting membrane voltage, properties such as input resistance (PYR Mdn = 83.8 MΩ, IQR = 47.9 MΩ; PV Mdn = 73.2 MΩ, IQR = 18.0 MΩ) and membrane time constant (PYR Mdn = 9.65 ms, IQR = 2.91 ms; PV Mdn = 7.32 ms, IQR = 11.9 ms) were not significantly different between the two cell types (p > 0.1, Wilcoxon rank-sum test; Fig. 2.4Ai, 2.4Bi). Given the similarities between PYR and PV cells in terms of spike phase and membrane properties, the significant differences in spike characteristics provides confirmation of cell-specific targeting of tdTomato. Membrane voltage is correlated to LFP oscillations In the absence of any current injection, the membrane potential in both PYR and PV cell at 0 ms lag was positively correlated with the LFP (PYR Mdn = 0.69, IQR = 0.13, Fig. 2.5A; PV Mdn = 0.58, IQR = 0.13, Fig. 2.5B). This indicates that membrane voltage in both cell types depolarizes with the rise of the LFP, and that spiking, phaselocked to the peak of intrinsic oscillations, results from this strong depolarization. In the above experiments it was unclear, however, whether cells were being excited or released from inhibition. To address the synaptic mechanisms establishing spike activity during oscillations, we injected cells with different amounts of positive or 17 negative current and measured the cross-correlation coefficient between intracellular membrane voltage and the LFP across a range of different membrane voltage values. We hypothesized that the correlation coefficient would vary with membrane voltage (Fig. 2.5Ai, 2.5Bi) depending on the nature of synaptic inputs driving spike activity in PYR and PV cells. Furthermore, the amount of synaptic conductance will determine how strongly the membrane voltage is clamped at the net synaptic reversal potential. From data shown in Figure 2.5, it is clear that in both PYR and PV cells, a strong synaptic conductance drives depolarizations during the peak of the LFP. In contrast, during the trough of the LFP, the membrane voltage is easily changed with different levels of DC current. When the membrane voltage at the trough is made to match the synaptic reversal potential at the peak, the resulting membrane voltage trace does not correlate to the LFP because the driving force associated with net synaptic activity is zero. Using this method, we assessed the net synaptic reversal potential of LFP correlated inputs in PYR and PV cells. Data from PYR cells showed that the normalized cross-correlation coefficients (Fig. 2.5Ai) crossed zero at a membrane voltage value (-52.3 mV) between the reversal for GABAa -mediated inhibition (-62.2 mV) and AMPA-mediated excitation (-1.5 mV). Thus, synaptic conductance onto PYR cells was a mixture of excitatory (gE) and inhibitory conductance (gI). On the other hand, PV cells were subjected to a synaptic conductance that clamped the membrane voltage to a value closer to the excitatory reversal potential. This is apparent in the example PV recording (Fig. 2.5B), where the trough voltage is unable to be depolarized with injected current beyond the reversal potential at the peak. Thus, the LFP remains positively correlated to the membrane 18 voltage of PV cells even at highly depolarized voltages (Fig. 2.5Bi), suggesting a mostly excitatory synaptic conductance. Our direct measurements of relatively large synaptic currents (Fig. 2.6A, 2.6B) closely match the phase-dependent behavior of the membrane voltage (Fig. 2.5A, 2.5B) suggesting that synaptic inputs largely drive spiking output directly, rather than being dependent on intrinsic membrane properties to amplify certain input frequencies. Pyramidal cells and interneurons experience a high conductance state during the peak of intrinsic oscillations To achieve a precise measurement of synaptic conductances, as well as the time course for excitatory and inhibitory inputs, we relied on a separate set of experiments in which we voltage-clamped PYR (n = 8) and PV (n = 8) cells to determine the time course and amount of excitatory (gE) and inhibitory conductance (gI) during an oscillation cycle. For these experiments, it was necessary to add QX-314 to the intracellular solution to block Na+ channels (Connors & Prince, 1982) and eliminate any spike-induced non-linearities in the I-V curves. Utilizing average current values measured at the trough and peak from multiple holding voltages (Fig. 2.6A, 2.6B), we constructed I-V curves at the peak and trough of the oscillation cycle (Fig. 2.6Ai, 2.6Bi). For all clamping voltages, the average current values were greater at the peak of the LFP compared to the trough. The larger changes in current as a function of voltage indicated a greater slope and hence conductance values near the peak of oscillations. This was true for both PYR and PV cells, and consistent with our observations in current clamp mode (Fig. 2.5A, 2.5B). In addition, synaptic 19 currents oscillate at theta-like frequencies (Fig. 2.6Aiii/i, 2.6Biii/i) directly influencing membrane voltage, in contrast to mechanisms that preferentially amplify theta frequencies in the membrane voltage due to voltage-gated ion channels (Cobb et al., 1995; Hu et al., 2002; Zemankovics et al., 2010). The intersection of the two I-V curves corresponds to the net synaptic reversal potential at the peak of oscillations (Fig. 2.6Ai). Using the change in slope, intersection, and assuming synaptic activity arises largely from AMPA and GABAa -mediated activity, we calculated the relative contribution and time course of gE and gI during an oscillation cycle. This approach is identical to past studies that have used these measures to extract synaptic conductance changes during anesthesia-induced up-states (Haider et al., 2006). During the peak conductance change in PYR cells, the average synaptic reversal (Mdn = -51.9 mV, IQR = 23.9 mV, Fig. 2.6Aii) was peri-threshold. This was consistent with our measures in current clamp mode in which the cross-correlation between the membrane voltage and the LFP fell to values of zero near -50 mV. These measures are indicative of balanced inputs in which precise amounts of excitatory and inhibitory conductance sum to produce a membrane voltage depolarization that drives precise spiking behavior (Wehr & Zador 2003; Haider et al., 2006; Atallah & Scanziani, 2009; Gan et al., 2017). For PV cells, even at highly depolarized holding voltages, an inward current was always present at the peak of the LFP. This indicated a synaptic reversal much more depolarized than that observed in PYR cells. Consequently, depolarizing PV cells to voltages where the change in current was zero between trough and peak of the LFP (i.e. where the two I-V curves intersect) was not possible due to the extremely large 20 amount of current required to hold the cell at depolarized potentials. Therefore, for PV cells, the synaptic reversal potential (Mdn = -13.2 mV, IQR = 21.1 mV, Fig. 2.6Bii) was extrapolated from linear portions of the I-V curves generated at the trough and peak (Fig. 2.6Bi). These measures suggest that PYR cells receive a mixture of excitatory and inhibitory synaptic conductance inputs, while an excitatory conductance is largely dominant in PV cells. Pyramidal cells, but not interneurons, are driven by balanced inputs during hippocampal oscillations Next, we quantified changes in conductance, relative to the trough, for every 4° phase bin of the LFP, as well as the corresponding reversal potentials, to determine the relative contribution of gE and gI to phase-specific conductance changes during LFP oscillations. PYR cells experienced a concomitant increase and subsequent decrease in gE and gI, with nearly a 100% increase in total conductance during the peak (Mdn increase = 8.14 nS, IQR = 8.36 nS), relative to the trough (Mdn = 8.11 nS, IQR = 4.06 nS). The high conductance state during spiking is largely due to gI. Nevertheless, inhibition does not prevent firing because of the much larger driving force associated with gE at a similar LFP phase (H [2] = 2.11, p = .35, Kruskal-Wallis test; Fig. 2.7Ai). We calculated the phase-dependent excitatory and inhibitory currents during an oscillation cycle (Fig. 2.7aii) using the mean excitatory and inhibitory conductances calculated from voltage-clamped PYR cells (n = 8) and the mean membrane voltage of current-clamped PYR cells (n = 14) with no injected current (Fig. 2.3). The net synaptic current tends to be inward or depolarizing at the beginning of an oscillation cycle, and 21 becomes balanced at the peak of the cycle as membrane voltage reaches spike threshold and the net synaptic reversal. We normalized the amount of gI and gE in each PYR cell to its total peak conductance to ascertain the phase-dependent gE to gI ratios (Fig. 2.7Aiii). The rise and fall of gE and gI is similar to results in Fig. 2.7A, indicating balanced synaptic currents during peak conductance, and leading to peri-threshold membrane voltage values at the peak of LFP oscillations (Mdn gE:gI= 1 : 7.04). Hence, despite variability in the absolute conductance levels, excitatory and inhibitory synaptic currents are correlated and balanced in individual PYR cells. Although PV cells also experienced a high conductance state during the peak of intrinsic oscillations (Mdn = 9.87 nS, IQR = 6.78 nS), the conductance change was the result of excitatory conductance, with only a small contribution from inhibition (Fig. 2.7B). The resulting net inward current (Fig. 2.7Bii) easily drives spiking in PV cells (H [2] = 0.06, p = .97, Kruskal-Wallis test; Fig. 2.7Bi). The normalized ratio of gE to gI in PV cells, therefore, was skewed largely in favor of excitatory conductance inputs (Fig. 2.7Biii). Together, these data suggest that circuitry in the subicular region of hippocampus is comprised of recurrent excitation between PYR cells that simultaneously also excites PV cells, which fire spikes and feedback inhibition onto PYR cells, but not to other PV cells. 22 Oscillations generate a small steady-state conductance in pyramidal cells, but not interneurons Given the intensity of synaptic activity and continuous spiking, we speculated that there may be an additional amount of conductance that is present during the trough of the LFP oscillation. To test this, we used the AMPA-R antagonist DNQX (20 µM) to eliminate oscillations in the whole hippocampal preparation, as well as intracellular activity that underlies spiking (Fig. 2.8A). Subsequently, using a small step hyperpolarization to measure input resistance and time constant values at the resting membrane potential, we assessed whether the presence of oscillations had a steady-state impact on membrane properties (Fig. 2.8B). Compared to the membrane properties gathered during the trough of LFP oscillations (Fig. 2.4Ai, 2.4Bi), DNQX significantly altered the resting membrane potential, membrane time constant, and input resistance of PYR cells (n = 7, p = .016, Wilcoxon signed-rank test; Fig. 2.8Ai), due to an additional 3 nS of leak conductance present during oscillations. In contrast, PV cells remained unchanged in their membrane properties after DNQX application (n = 4, p = .65 or .13 for resting voltage, Wilcoxon signed-rank test; Fig. 2.8Bi). As a result, during oscillations, PYR cells received balanced synaptic inputs that drive peak depolarizations, as well as a steady-state background conductance, while PV cells lack the steady-state conductance and are driven almost exclusively by excitation. 23 Discussion To address the synaptic mechanisms of intrinsic hippocampal oscillations, we gathered LFP and intracellular recordings from specific cell types in the whole hippocampal preparation. Our recordings reveal that pyramidal cells and fast-firing interneurons exhibit simultaneous, phase-locked firing similar to those observed during movement-evoked theta oscillations in the hippocampus (Fox et al., 1986; Skaggs et al., 1996; Csicsvari et al., 1999, Stark et al., 2013). The synaptic mechanisms of these slow autonomous oscillations, however, resembled those generating in vivo gamma oscillations and ripples (Atallah & Scanziani, 2009; English et al., 2014; Gan et al., 2017). Similar to these high-frequency hippocampal oscillations, as well as lower frequency cortical oscillations (Shu et al., 2003; Haider et al., 2006; Tahvildari et al., 2012), we find that balanced and correlated inputs underlie the periodic depolarization in pyramidal cells, and drive spiking locked to the peak of network oscillations. Balanced excitation and inhibition and the generation of network oscillations From our recordings, we infer that recurrent excitation from pyramidal cells provides positive feedback that induces simultaneous firing throughout the entire network. This includes fast-firing interneurons that contribute strong negative feedback by shunting excitatory drive to pyramidal cells in the network. What results is a large depolarization that drives pyramidal cell membrane voltage to spike threshold during the peak of inhibition. Overall, our data and analyses are inconsistent with spike phaselocking in pyramidal cells resulting from rebound spiking or membrane resonance as 24 suggested previously for theta oscillations. In particular, any rebound spiking would require that spike generation follows a bout of inhibition (Cobb et al., 1995, Stark et al., 2013), something that was never observed since peak spiking probability coincided with peak synaptic excitatory and inhibitory conductance. Therefore, our results are similar to extracellular recordings from movement-associated (type I) theta, where pyramidal cells and interneurons fire in-phase with each other (Fox et al., 1986; Skaggs et al., 1996; Csicsvari et al., 1999, Stark et al., 2013). This is in contrast to anesthesia-induced theta (type II) where PV cells fire prior to PYR cells (Klausberger & Somogyi, 2008) in support of the post-inhibitory rebound spiking mechanism. Additionally, membrane resonance mechanisms would require a significant amplification of theta inputs by intrinsic membrane properties (Hu et al., 2002, Zemankovics et al., 2010). Our measures of synaptic currents in voltage clamp, however, indicated that a large net synaptic input drove periodic depolarization and spiking in pyramidal cells, suggesting membrane properties that amplify synaptic inputs would be secondary to network properties that maintain balanced currents. Further, any membrane resonance would likely be overwhelmed by the high conductance associated with synaptic-mediated inputs (Fernandez & White, 2008). The theta-like activity we observed in hippocampus, however, likely differs from that observed in vivo (Fox et al., 1986; Skaggs et al., 1996; Csicsvari et al., 1999, Stark et al., 2013). Most notably, we observed lower frequencies in network oscillations than those typically associated with theta activity in awake, behaving animals. Along with inputs from medial septum (Freund & Antal, 1988; Lee et al., 1994), theta activity in vivo likely involves many additional factors. Nevertheless, the activity we observed indicates 25 that the network architecture in hippocampus supports a tight balance that maintains remarkably similar levels of excitatory and inhibitory currents in pyramidal cells during network-wide oscillations. In our study, the network activity underlying intrinsically generated oscillations shares many similarities to other in vivo oscillations. Hippocampal gamma oscillations and ripples, as well as cortical up-down state transitions, have been shown to be generated through a mechanism involving balanced excitation and inhibition in principal excitatory cells (Shu et al., 2003; Haider et al., 2006; Atallah & Scanziani, 2009; Tahvildari et al., 2012; Gan et al., 2017). A simultaneous rise in excitatory and inhibitory synaptic current produces depolarization in pyramidal cells that drives spiking phaselocked to the peak of the network oscillations. In the case of hippocampal oscillations at gamma (Atallah & Scanziani, 2009) and ripple frequencies (English et al., 2014; Gan et al., 2017), interneurons provide negative feedback that is tightly coupled to the rise of excitatory current in individual pyramidal cells. During both the oscillations we recorded and these faster hippocampal oscillations, peak inhibitory and excitatory currents are approximately balanced and drive pyramidal cells to membrane voltage values near spike threshold. Similarly, reoccurring up-down states in the cortex (Haider et al., 2006), often oscillatory in the range of 0.5-2 Hz, have been shown to be generated through a near-simultaneous rise in excitatory and inhibitory currents. Although inhibitory conductance was greater than excitation, the reduced driving force for inhibition leads to excitatory and inhibitory currents that are balanced near spike threshold. Our results extend the ability for balanced activity in pyramidal 26 cells, and tightly coupled negative feedback from fast-firing interneurons, to generate hippocampal oscillations at low theta frequencies. Comparisons to intrinsic oscillations in other hippocampal regions Previous studies of the isolated hippocampal preparation did not find significant excitatory drive in area CA1 pyramidal cells. Instead, inhibitory currents in CA1 pyramidal cells at the peak of intrinsic oscillations resulted in rebound spiking during the trough (Goutagny et al., 2009, Huh et al., 2016), in line with juxtacellular recordings during anesthesia-induced theta (Klausberger & Somogyi, 2008). This is in contrast to our recordings from pyramidal cells in the distal subiculum, which spiked at the peak of balanced E/I currents and the LFP oscillation, with no synaptic or spiking activity during the trough. Our data, however, match previous extracellular recordings of pyramidal cells in distal CA1/proximal subiculum from mice as old as 5 weeks, which spike during the peak of intrinsic oscillations (Amilhon et al., 2015). On the other hand, we found PV cells exhibited similar spiking behavior in our subicular recordings compared to those from CA1. In both areas, PV cells spiked at the peak of oscillations and excitatory synaptic drive. However, we recorded only excitatory conductance in subicular PV cells, whereas PV cells in CA1 exhibited some, albeit small, inhibitory conductance as well (Huh et al., 2016), likely from significant reciprocal PV connections (Bartos et al., 2002). The lack of inhibition in PV cells we observed is surprising in light of studies showing substantial reciprocal connections between fastfiring interneurons in the subiculum (Bohm et al., 2015) and presubiculum (Peng et al., 27 2017). It is uncertain, however, whether past results are specific to parvalbuminexpressing interneurons or include non-PV cells. Additionally, PV interneurons have been shown to undergo synaptic depression as a result of sustained activity (Stark et al., 2013), which may limit their recurrent inhibition during theta oscillations. The region-specific differences in pyramidal cell behavior, and thus proposed mechanisms underlying intrinsic oscillations, could potentially result from differences in network architecture. Both the subiculum and CA3 are independent oscillators in the isolated hippocampus (Jackson et al., 2011; Jackson et al., 2014), and both areas share extensive axon collateralization of pyramidal cells (Harris et al., 2001; Witter, 2007). In contrast, area CA1 lacks this degree of recurrent architecture (Deuchars & Thomson, 1996), which may preclude strong positive feedback mechanisms. Instead, pyramidal cells in CA1 project their axons throughout the subiculum (Amaral et al., 1991). However, a recent study in the whole hippocampal preparation revealed the existence of pyramidal cell-to-cell connections within the longitudinal CA1 axis, in contrast to their absence from transverse hippocampal slices (Yang et al., 2014). Comparison of intrinsic subicular oscillations to tightly balanced networks The neuronal behavior we recorded in subiculum shares characteristics with tightly balanced networks in vivo (Wehr & Zador, 2003; Wilent & Contreras, 2005; Haider et al., 2006; Okun & Lampl, 2008). In particular, excitation and inhibition are strongly correlated on very short timescales, and exhibit strong phase-locked spiking at the peak of synaptic input. This is in contrast to loosely balanced networks in which 28 overall network firing rates of E and I cells are balanced, but synaptic currents in individual neurons may not be (Deneve & Machens, 2016). Tight balance imparts networks with an integration window in which spiking behavior can be influenced by the network state (Gabernet et al., 2005; Higley & Contreras, 2006). Changes in this balance correspond to changes in LFP oscillations and individual neuronal spike activity (Atallah & Scanziani, 2009). In contrast to hippocampal brain slices that require pharmacological application to induce theta-like activity (Fellous & Sejnowski, 2000), oscillations in our preparation occurred spontaneously. The propensity for hippocampal oscillations without pharmacology is likely the result of increased recurrent network connectivity provided by the whole hippocampal tissue. For example, cortical slices have been shown to generate intrinsic oscillations as a result of recurrent excitation and balanced inputs (Shu et al., 2003, Tahvildari et al., 2012). Additionally, the subiculum can independently generate oscillations (Jackson et al., 2014) even when removed from the whole hippocampal preparation (Jackson et al., 2011). This suggests that recurrent architecture is an important component for in vitro rhythmogenesis, which may be absent from transverse hippocampal slices. A potential source for the timescale of slower cortical slice oscillations are potassium conductances that increase during depolarization or spiking (Sanchez-Vives & McCormick, 2000; Tahvildari et al., 2012). In cortical slices undergoing oscillations, the increased conductance was measured once spiking had ceased, during the down state or trough. Consistent with this scenario, our current clamp recordings did indicate a significant difference in conductance between the trough state during oscillations and the 29 neuron’s intrinsic conductance when oscillations were pharmacologically blocked. This increased conductance at the trough is likely the result of continuous spiking during the peak when balanced inputs drive membrane voltage to spike threshold. In cortical slices, repetitive spiking creates long refractory periods during the trough of the oscillation (Compte et al., 2003), whereas in the hippocampal preparation, fewer spikes could generate a shorter refractory period resulting in faster oscillations. Although potassium conductances likely play a role in setting the refractory period during intrinsic hippocampal oscillations as well, further investigation will be needed to determine how much of the trough conductance we recorded is due to cycle to cycle spiking versus increased potassium conductance due to consistent spiking over long periods of time. The latter could explain the increase in refractory period/decrease in oscillation frequency that occurs over time in the hippocampal preparation. 30 Figure 2.1 The whole hippocampal preparation generates stable and long-lasting autonomous oscillations. A. Example spectrogram of an LFP oscillation in the distal subiculum and its progression over the course of 1 hr. Ai. Distribution of mean LFP frequency across all hippocampal preparations (PYR & PV, n = 38), recorded from the subiculum during concurrent on-cell patch clamp experiments (Fig. 2.2). B. Phase-averaged LFP cycle from all recordings. Bi. distribution of LFP phase corresponding to the LFP peak amplitude across all recordings. 31 Figure 2.2 PYR and PV cells spike primarily during the peak of LFP oscillations in the subiculum. A. Example recording from the fastest frequency LFP (top, black) during concurrent oncell recording of a PYR cell (red, bottom). PYR cell spikes are highly phase-locked to the peak of LFP oscillations. Ai. Distribution of spike-phase and vector strength from each PYR cell (yellow, vertices) and average spike-phase/vector strength from all PYR cells (red, n = 22). Aii. Preferred spike phase in PYR cells (red) is independent of LFP frequency and Aiii. PYR cell spiking occurs at a slow rate. B. Example of fastest frequency LFP (top) during concurrent on-cell recording of PV cell (blue, bottom). Bi. Distribution of spike-phase and vector strength from each PV cell (yellow, vertices) and average spike-phase/vector strength from all PV cells (blue, n = 16). Bii. Preferred spike phase in PV cells (blue) is independent of LFP frequency and Biii. PV spiking occurs at a similar rate to PYR cells. 32 Figure 2.3 Intracellular current-clamp recordings of PYR and PV cells during intrinsic hippocampal oscillations. A. Example LFP recording (black) and corresponding membrane voltage from PYR cell (red) and B. PV cell (blue). Ai. Distribution of spike half-width and spike threshold for PYR cells (red, n = 14). and Bi. PV cells (blue, n = 8). C. 2-photon image, during whole-cell patch clamp, of a PYR cell (yellow) filled with Alexa 488 (green) in addition to endogenous tdTomato (red). D. Fill of a PV cell (yellow) during whole-cell patch clamp. 33 Figure 2.4 Membrane properties of PYR and PV cells during autonomous oscillations. A. Example PYR cell membrane voltage in response to hyperpolarizing steps (black), and average from all steps (red) during non-spiking trough phase. B. as in A, but from PV cell (blue) Ai. Distributions of intrinsic membrane properties from PYR cells (red, n = 14) and Bi. PV cells (blue, n = 8). 34 Figure 2.5 Membrane voltage of PYR and PV cells is highly correlated to the LFP. A. Example of a PYR cell’s membrane voltage (red) averaged over multiple oscillation cycles at various steps of applied current. B. As in A, but for a PV cell. Ai. Crosscorrelation (0 ms lag) between the LFP and membrane voltage of PYR cells (red, n = 14) over multiple membrane voltages. With applied current, as in A (I=125pA), when the membrane voltage matches the reversal potential of synaptic input, the driving force to move membrane potential up or down drops to zero as does the LFP to membrane voltage cross-correlation (dotted line). A sign reversal in cross-correlation that occurs between the values of Eexc (-1.5mV) and Einh (-61.5mV) indicates synaptic input that is a combination of both excitation and inhibition. Bi. As in Ai, but for PV cells (blue, n = 8). The cross-correlation for PV cells does not reverse sign indicating a lack of hyperpolarizing input. 35 ++++++ Figure 2.6 Quantification of membrane conductance as a function of LFP phase using voltage clamp across multiple holding voltages. A. Example voltage clamp of a PYR cell at various holding voltages averaged over multiple oscillation cycles. Ai. The slope of the I-V curve represents the neuronal conductance at a particular phase of the LFP. The increase from baseline (orange) to peak (red) conductance results from synaptic input. The reversal potential of synaptic input is the holding voltage where the difference between baseline and peak current equals 0. A synaptic reversal potential that falls between the reversals of GABAa (62.2mV) and AMPA (-1.5mV) provides the ratio of inhibitory (gI) to excitatory conductance (gE). As in Aii, a reversal potential that is perithreshold is considered the result of balanced currents. Aii. Distribution of reversal potentials from all voltage clamped PYR cells (n = 8) during their respective phase of peak conductance. Aiii. Example spectrogram of PYR cell current at a holding voltage similar to resting membrane potential (-60 mV) reveals synaptic inputs at theta-like frequencies. Aiiii. Distribution of membrane current frequency in all voltage-clamped PYR cells (n = 8) at -60 mV holding voltage. B. As in A, but for PV cells. Inward current is present even at suprathreshold voltages, suggesting a lack of hyperpolarizing GABAa conductance. Bi. Inability to depolarize past the reversal of synaptic conductance in PV cells, the reversal potential between baseline (light blue) and peak conductance (dark blue) must be extrapolated. Bii. Distribution of reversal potentials from all voltage-clamped PV (n = 8) cells during their respective phase of peak conductance. Biii. Example spectrogram of PV cell showing theta-modulated synaptic currents at -60 mV. Biiii. Distribution of membrane current frequency in all voltage-clamped PV cells (n = 8) at -60 mV holding voltage. 36 Figure 2.7 A correlated rise and fall of excitation (gE) and inhibition (gI) in PYR cells results in balanced synaptic currents at the peak of the oscillation cycle, whereas in PV cells synaptic inputs are largely dominated by excitation. A. Across PYR cells (n = 8), the time course of gE (iqr=red) is correlated to gI (iqr=blue) (median=black) and current-balanced at the peak of high conductance synaptic input. Ai. Distribution of LFP phase during which PYR cells achieved peak gE (left) and peak gI (middle). Neither gE or gI are significantly different from the phase of spiking (right), recorded from these cells during on-cell configuration. Spiking results from a peak in gE despite a concurrent peak in gI. Aii. Ratio of gE:gI in PYR cells results in excitatory (red) and inhibitory (blue) synaptic currents that in summation (black), balance at the peak of an average oscillation cycle. Aiii. Most PYR cells exhibit a current-balanced ratio of gE:gI on a cell-by-cell basis when gE and gI in each cell is normalized to the peak of the cell’s total conductance. B. As in A, but for PV cells. Unlike PYR cells, PV cells (n = 8) are dominated by an unbalanced amount of high gE to low gI. Bi. Similar to Ai, the phase of peak gE (left) and gI (middle) in PV cells is not statistically different from the phase of on-cell spiking (right) in these cells. Bii. Net synaptic current (black) in PV cells is largely excitatory (red) with little inhibition (blue). Biii. The normalized ratio of gE:gI in PV cells is similar to B such that most PV cells exhibit greater amount of gE compared to gI. 37 Figure 2.8 Application of glutamatergic antagonist DNQX abolishes oscillations in both the LFP and membrane voltage of PYR and PV cells. A. Example LFP (top, black) and membrane voltage of a PYR cell (bottom, red) after bath application of DNQX. Within minutes, spiking and voltage fluctuations in the PYR cell cease. B. Small hyperpolarizing current steps (black) were repeated to obtain an average voltage trace for PYR and PV cells, examples in (red) and (blue), respectively. Ai & Bi. Comparison of intrinsic membrane properties at baseline (some cells from Fig. 2.4), and after DNQX application. DNQX application significantly changes the intrinsic properties of Ai. PYR cells (red, n = 7) but not Bi. PV cells (blue, n = 4). CHAPTER 3 REFRACTORY DYNAMICS IN SUBICULAR PYRAMIDAL CELLS Introduction Autonomous oscillations in the isolated, whole hippocampal preparation (Goutagny et al., 2009) occur within the theta-frequency band (3-12 Hz). The properties that set the timescale for oscillations in the isolated hippocampus, however, have not been determined. Theta oscillations in vivo and the whole hippocampal preparation have been suggested to occur through a mechanism involving post-inhibitory rebound spiking and membrane resonance (Stark et al., 2013; Huh et al., 2016). The timescale for these mechanisms is considered to arise from intrinsic membrane properties, particularly the activation and deactivation kinetics of the Ih current (Hu et al., 2002; Zemankovics et al., 2010). In our previous work (Royzen et al., 2019), we showed that autonomous oscillations in the subiculum of the whole hippocampal preparation are associated with a tight balance of synaptic current in which excitation and inhibition rise and fall together. This result is inconsistent with spiking resulting from post-inhibitory rebound. Therefore, Ih-mediated dynamics may not be involved in setting the timescale of autonomous oscillations. 39 Given the similarities of our previous results in subiculum to other tightly balanced networks underlying oscillations (Shu et al., 2003; Haider et al., 2006; Atallah & Scanziani, 2009; Tahvildari et al., 2012; Gan et al., 2017), we tested whether similar dynamics are involved in setting the timescale of autonomous oscillations. For example, during slow-wave sleep and anesthetic states, the timescale for oscillations is considered to be dependent on potassium-channel conductances that increase as a result of spiking during the peak of balanced inputs (Sanchez-Vives & McCormick, 2000; Compte et al., 2003; Shu et al., 2003; Tahvildari et al., 2012). Potassium-channel conductances decrease neuronal excitability, and the timescale of their recovery to baseline levels is considered to set the refractory period between cycles, and thus oscillation frequency. Therefore, we tested whether recovery of potassium-channel conductances in hippocampal pyramidal cells share a similar timescale to intrinsic theta frequency oscillations. By injecting artificial membrane conductance into quiescent pyramidal cells, we determined that conductance-evoked spiking increased potassium-channel conductances after spiking had ceased. Importantly, this potassium conductance recovered to baseline levels on a similar timescale to theta frequencies. On the contrary sodium-channel conductance recovered on much longer timescales. Additionally, we show that the timescale of oscillations is dependent on the rate of pyramidal cell firing. Bath application of a GABAa antagonist transforms one or two spikes per cycle, during autonomous oscillations, into sustained firing leading to depolarization block. This behavior is highly correlated to fluctuations in the LFP, which as a result take much longer to recover between bouts of recurrent excitation. 40 Materials and Methods Ethics statement All experimental protocols were approved by the Boston University Institutional Animal Care and Use Committee. Tissue preparation Transgenic mice (P12-P18) of either sex were anesthetized with isoflurane and decapitated into a solution of ice-cold artificial cerebrospinal fluid (aCSF), oxygenated, and buffered to pH 7.3 with 95% O2/ 5% CO2. ACSF consisted of (in mM): NaCl (126), NaHCO3 (24), D-glucose (10), KCl (4.5), MgSO4 (2), NaH2PO4 (1.25), CaCl2 (1), and ascorbic acid (0.4) (Sigma-Aldrich). Both hippocampi were extracted and incubated in room temperature aCSF for 30-120 minutes. Subsequently, one hippocampus was transferred to a custom recording chamber that was equipped to recirculate oxygenated/buffered aCSF (add. 1 mM CaCl2) at a flowrate of 20-25 mL/min and a temperature of 31.5°C ± 0.2°C. Visualizing hippocampal pyramidal cells To visualize cells and glass pipettes in the distal subiculum, we used a 2-photon microscope (Bergamo II Series, ThorLabs) equipped with a Ti:Sapphire laser (920 nm; Chameleon Ultra, Coherent) to enable excitation of fluorescent markers through a 20x, NA 0.95 objective lens (Olympus). To detect fluorescence from intracellular tdTomato and Alexa 488 in the pipette, we used two photo-multiplier tubes (Hamamatsu) that receive either red or green optically filtered emission, respectively. Transgenic mice were 41 generated by crossing CaMKIIa-Cre promoter mice (Jackson Labs, stock # 005359) with the loxP-flanked tdTomato reporter mice (Jackson Labs, stock # 007914) (Madisen et al., 2010). The CamkIIa- promoter is specific for expression in principal cells (Tsien et al., 1996). Electrophysiology A pulled-glass electrode (0.5-1.5 MΩ) containing aCSF and Alexa Fluor 488 hydrazide (.0029% w/v; Thermofisher Scientific) was used to record LFP oscillations and fluctuation in the distal subiculum at the mid-septotemporal axis. The LFP electrode was advanced through the hippocampal tissue using a micromanipulator (Sutter Instrument) until oscillations reached peak amplitude, reportedly around the stratum radiatum (Goutagny et al., 2009). A second electrode (3.5-6.5 MΩ) with intracellular solution and Alexa 488 (.0014% w/v) was placed in close proximity to patch-clamp fluorescing pyramidal cells. The intracellular solution was comprised of (in mM): K-gluconate (136), HEPES (10), KCL (4), EGTA (0.2), MgCl2 (2), diTrisPhCr (7), Na2ATP (4), and TrisGTP (0.3); buffered to pH 7.3 with KOH (Sigma-Aldrich). Signals were recorded with Clampex (v. 10.6; Molecular Devices) using a Multiclamp 700B amplifier and Digidata 1440A digitizer (Molecular Devices) at a sampling rate of 25 kHz, with the LFP filtered between 1-50 Hz. Signals were processed and analyzed using integrated MATLAB (R2017b; Mathworks) functions and custom scripts. Pipette offset was compensated prior to achieving on-cell patch recordings. Once the patch electrode was sealed to the cell membrane (>1 GΩ seal), pipette capacitance 42 was compensated. During intracellular recordings, access resistance (15-30 MΩ) was compensated and monitored every 5-10 minutes to not exceed 65 MΩ. For analysis of pyramidal cell refractory dynamics (n = 12 cells), oscillations in the whole hippocampal preparation were eliminated with bath application of both gabazine (10 µM, Abcam) and DNQX disodium salt (20 µM, Tocris Bioscience). Paired conductance steps were then applied to pyramidal cell inputs to emulate inputs during intrinsic hippocampal oscillations. Using real-time dynamic clamp (RTXI), 10 nS of conductance was injected for 150 ms in the first step to elicit a single or doublet action potential. Thus, reversal potential of the first conductance step was tuned for each cell to be peri-threshold similar to balanced currents during autonomous oscillations (Royzen et al., 2019). The second conductance step had the same parameters as the first except the step was increased to 500 ms to examine spiking dynamics over an extended period of time. The delay between the two steps was increased logarithmically and recovery of spiking was quantified. In the same cells, the first conductance step was modified to evoke the same rate of spiking over 150 ms using current-based injection, to examine if the method used to elicit spiking had any significance on refractory dynamics. Refractory dynamics were examined in depth using current-evoked spiking during the second 500 ms step. Data analysis Data in text and figures are reported as median and interquartile range (25-75%) due to the non-normal distribution (p < 0.05, one-sample Kolmogorov-Smirnov test). Likewise, the statistical tests used are non-parametric. Reported values of membrane 43 voltage are not adjusted for liquid junction potential; actual values are 11.5 mV more negative. For pharmacological experiments (Fig. 3.1-3.3, n = 6 cells), the phase of the LFP was determined with a Hilbert transform function implemented in Matlab. The phase of the LFP was used to bin and average membrane voltage and LFP data (downsampled to 50 Hz to eliminate spiking) during oscillation/fluctuation cycles (Fig. 3.2A, 3.2B). Cycles were detected by peaks in the derivative of the LFP phase, which occurred when the phase wraps from -180 to 180 degrees (cycle beginning/end). Portions of the LFP trace that are relatively flat tend to create many of these peaks over a short duration, so data from inter-peak intervals less than 100 ms (cycles >10 Hz) were not considered. Frequency analysis of the LFP and membrane voltage (Fig. 3.1B, 3.3B) utilized the multi-taper mtspecgramc function of the Chronux toolbox [http://chronux.org/], with an averaging window of 2.0 seconds moving at 1.33 seconds. The MATLAB function xcorr was used to determine the normalized cross-correlation coefficient between the (downsampled to 50 Hz) LFP and membrane voltage before (0-40 seconds) and after (1-5 minutes) gabazine had taken effect (Fig. 3.1C). Membrane voltage skew and standard deviation were also determined with MATLAB’s included functions before (0-40 seconds) and after (80-120 seconds) DNQX had taken effect (Fig. 3.3C). In order to quantify refractory dynamics as a function of delay between the paired steps (Fig. 3.4-3.6), spiking characteristics of the first action potential were analyzed during the second step of current. Rate of action potential rise and fall were analyzed using the maximum derivative of membrane voltage during the second step. Threshold values were similarly determined using the peak of the second derivative of membrane 44 voltage. Spike half-width was calculated as the time between midpoints (voltage between threshold and spike height) of the rising and falling phase. The time constant of recovery for these spike characteristics was determined using a non-linear regression fit; however, some could not be properly fit (Fig. 3.5C, bottom; Fig. 3.6A, 3.6B, bottom). Spiking characteristics during the first step of current or conductance were also analyzed, to determine if the method used to evoke spiking had a significant effect (Fig. 3.5B). Results Recurrent excitation drives synchronous pyramidal cell spiking In our previous work (Royzen et al., 2019), we found that theta oscillations were associated with balanced excitation and inhibition in pyramidal cells, with excitation increasing rapidly during the start of a theta cycle. Further, during a theta cycle, increases in excitation were matched by a proportional increase in inhibition. A potential source for the increase in excitation associated with intracellular depolarization are recurrent connections between pyramidal cells in subiculum (Harris et al., 2001); positive feedback from recurrent excitation is kept in check by a proportional increase in inhibition. In the absence of inhibition, therefore, it would be expected that runaway excitation leads to larger events that drive pyramidal cells to depolarization block. To test whether recurrent excitation or inhibition synchronizes pyramidal cell spiking in the whole hippocampal preparation, we bath-applied the GABAa antagonist gabazine to eliminate inhibition in the oscillatory network. The absence of inhibitory drive did not suppress LFP activity, indicating that synchronous population spiking was still present. Instead, strongly cross-correlated (Mdn 45 = 0.80, IQR = 0.072; Fig. 3.1C, top) oscillations in the LFP and membrane voltage of pyramidal cells transformed into aperiodic fluctuations (Fig. 3.1A, 3.1B), marked by much larger depolarizations. Inhibition was not necessary to synchronize spiking across the pyramidal cell population, as seen by a strong cross-correlation (Mdn = 0.58, IQR = 0.24; Fig. 3.1C, bottom) between the LFP and membrane voltage of pyramidal cells (n = 6) during post-gabazine fluctuations. Feedback inhibition limits excitatory drive in pyramidal cells Consistent with a loss of feedback inhibition and presence of strong recurrent excitatory connections, depolarizations in the presence of gabazine were significantly larger (p= 0.031, Wilcoxon signed-rank test) in pyramidal cells (n = 6), leading to depolarization block (Mdn peak Vm = -18.7 mV, IQR = 14.0 mV) in most cases (Fig. 3.2A, 3.2Ai). This contrasts with depolarizations during autonomous oscillations, which only produced 1-2 spikes per cycle (Royzen et al., 2019) as a result of feedback inhibition that maintains membrane voltage at peri-threshold values (Mdn peak Vm = -41.6 mV, IQR = 5.6 mV). As spiking ceased during the peak of LFP fluctuations or oscillations, the LFP fell to baseline levels. This indicates that feedback mechanisms, which limit pyramidal cell spiking either by intrinsic or network properties, have a significant effect on terminating LFP oscillations or events. The effect of gabazine on LFP amplitude was not as dramatic as on membrane potential. Although the LFP peak amplitude appeared to increase after gabazine application (Mdn peak Vm = 0.35 mV, IQR = 0.16 mV), it was not significantly different 46 (p = 0.22, Wilcoxon signed-rank test; Fig. 3.2B, 3.2Bi) from the pre-gabazine LFP amplitude (Mdn peak Vm = 0.25, IQR = 0.15). Excitatory drive is dependent on glutamatergic signaling In the absence of inhibition, acetylcholine has been shown to generate hippocampal oscillations at various frequencies depending on concentration (Fellous & Sejnowski, 2000). However, the addition of DNQX to block excitatory AMPA receptors, eliminated post-gabazine fluctuations in the LFP and membrane voltage of pyramidal cells (Fig. 3.3A, 3.3B). This result indicates that pyramidal cell spiking was driven solely by glutamatergic input. Thus, cholinergic signaling was unlikely to contribute to pyramidal cell depolarizations, as pyramidal cells became quiescent with a significant drop in membrane voltage skew and standard deviation after DNQX application (Fig. 3.3C). Conductance-evoked spiking recovers on a similar timescale to theta oscillations Next, we sought to determine the factors that give rise to the network oscillation timescale. Given the strong recurrent feedback between subicular pyramidal cells (Harris et al., 2001), we hypothesized that timescale of theta-like oscillations emerges from intrinsic membrane properties that reduce pyramidal cell excitability after strong depolarization. To test the effect of pyramidal spiking on network refractory dynamics, we blocked excitation and inhibition to quieten the network and injected quiescent pyramidal cells in the distal subiculum with steps of artificial conductance to elicit 1-2 47 spikes (Fig. 3.4A), similar to their behavior during autonomous oscillations (Royzen et al., 2019). Artificial conductance more accurately simulates synaptic inputs compared to current injection and could have a significant effect on spiking dynamics (Chance et al., 2002; Prescott & De Koninck, 2003; Fernandez & White, 2010). The reversal potential of these conductance steps was chosen to be peri-threshold for each cell (~ -50 mV to -45 mV) to elicit the minimum amount of spiking within 150 ms, similar to depolarizing events driven by balanced excitation-inhibition currents during autonomous oscillations (Royzen et al., 2019). The length of the step is approximately the amount of time synaptic inputs are active during autonomous oscillation cycles. A large conductance (10 nS) was injected to approximate the average synaptic input in pyramidal cells during the peak of autonomous oscillations. By varying the delay between the two conductance steps, we found that steps delivered at delays shorter than 80 ms generated no spiking in pyramidal cells (Fig. 3.4B). At delays longer than 320 ms, all pyramidal cells had recovered spiking behavior. Thus, despite only depolarizing cells to peri-threshold levels and generating one or two spikes, recovery from the initial step took approximately 80-320 ms, and matched the theta frequency timescale. Theta timescale of spiking recovery matches recovery of potassium-channel conductances To determine the cellular properties that set the refractory timescale, we analyzed features of the first action potential during the second step, as a function of delay between the two steps. Current injection was used for the second step instead of conductance in 48 order to more reliably elicit action potentials regardless of delay. In particular, the derivative, height, and half-width of a spike can be used as indirect measures of the magnitude of the underlying voltage-gated conductances during a spike. Recovery of delay-dependent conductances that match the timescale of theta, therefore, would more likely be responsible for the recovery of spiking behavior and contribution to network oscillations at theta. The timescale of recovery was determined by quantifying the change in specific spike characteristics as a function of delay, relative to baseline values (delay of 5 seconds), and quantifying the non-linear fit of that relationship. Spike characteristics indicative of sodium-channel conductivity such as action potential rate of rise, and to a lesser degree the spike height (Fig. 3.5C), recovered on timescales that appear too slow to determine the 80-320 ms theta period. In contrast, spike characteristics such as threshold and those dependent on potassium-channel conductances, such as action potential rate of fall and half-width, recovered on timescales within the theta range (Fig. 3.6). These results suggest that both voltage-gated Na+ and K+ channels contribute to reduced excitability of subicular pyramidal cells after a bout of theta-related activity, but that the recovery of voltage-gated K+ channels is more likely responsible for the theta timescale refractory period quantified in Figure 3.4. In addition to using current injection during the second step, we also modified the first step so that a minimal 1-2 spikes were evoked by either a conductance or current injection (Fig. 3.5A) to determine if there were any differences in spiking (Fig. 3.5B) or refractory dynamics (Fig. 3.5C, 3.6) between the two methods. As seen in previous studies (Fernandez & White, 2010), spike threshold during the first step significantly (p= 0.0093, Wilcoxon signed-rank test) increased in pyramidal cells (n = 12, Fig. 3.5B) when 49 using conductance compared to current injection (Mdn change = 1.29 mV, IQR = 1.28 mV). In contrast, refractory dynamics, as measured by the changes in spike characteristics during the second step, did not appear to differ in the timescale of recovery based on whether the first step was conductance- or current-based (blue or green, respectively; Fig. 3.5C, 3.6). The important factor appeared to be the number of spikes evoked during the first step and not whether the underlying depolarization was driven by conductance or current injection. Discussion Using pharmacological manipulations and injection of artificial membrane conductance, we uncovered properties of the subicular network that could determine the timescale of theta-like autonomous oscillations. Our results indicate that intracellular depolarizations associated with theta-like oscillations induce a refractory period in individual pyramidal cells that diminished the likelihood of spike generation and requires a timescale of recovery compatible with theta oscillations. Our results differ from previously established mechanisms of theta oscillations that have suggested amplification via intrinsic membrane resonance at theta to amplify theta inputs from extrinsic sources (Cobb et al., 1995; Toth et al., 1997; Hu et al. 2002; Zemankovics et al., 2010; Stark et al., 2013). Instead, our mechanism and network behavior resemble oscillatory networks of balanced excitation-inhibition that generate oscillations both at lower and higher frequencies than theta (Shu et al., 2003; Haider et al., 2006; Atallah & Scanziani, 2009; Tahvildari et al., 2012; Gan et al., 2017). 50 Comparison of autonomous oscillations to established mechanisms of hippocampal theta rhythmogenesis Selective inactivation of neuronal subtypes in the medial septum, often regarded as a pacemaker of hippocampal theta oscillations, dramatically reduced the amplitude of in vivo theta oscillations in the hippocampus (Yoder & Pang, 2005; Boyce et al., 2016). These results provide evidence for the involvement of cholinergic and GABAergic projections (Freund & Antal, 1988) in driving hippocampal theta rhythms. However, the slower timescale of muscarinic receptors (Hasselmo & Fehlau, 2001) and experimental evidence that antagonists do not affect movement-related theta oscillations (Kramis et al., 1975) suggest that GABAergic projections play a greater role in pacing (Toth et al., 1997). In the whole hippocampal preparation, devoid of septal input, our results point to local, rather than extrinsic, GABAergic transmission as being necessary for the maintenance of theta-like oscillations in the whole hippocampal preparation. Also, we find that cholinergic signaling is unlikely to provide the recurrent excitation that is necessary to drive autonomous oscillations in the subiculum. GABAergic projections from medial septum preferentially target the oriensalveus layer in the hippocampus, which contains largely interneuronal populations (Freund & Antal, 1988). Intrinsic membrane properties of these interneurons, as well as pyramidal cells, endow these neurons with a preference for theta frequency inputs. Theta frequency preference is abolished with application of ZD7288, a channel blocker of Ih current (Hu et al. 2002; Zemankovics et al., 2010). Thus, the timescale of theta oscillations in the hippocampus has been suggested to occur as a result of the Ih current, which imparts theta resonance and rebound-spiking dynamics to neurons (Cobb et al., 51 1995; Stark et al., 2013). In contrast, our results indicate that autonomous theta-like oscillations in the subiculum do not depend on resonance or rebound-spiking dynamics to set the timescale of oscillations. Instead, network properties that control the recovery of spike discharge probability in pyramidal cells influence the refractory period and thus timescale of oscillations. Role of fast-spiking interneurons in setting the timescale of hippocampal oscillations Fast spiking interneurons, which can have a substantially higher frequency preference than pyramidal cells (Pike et al., 2000, Zemankovics et al., 2010), have also been shown to be highly active during the theta rhythm (Klausberger & Somogyi, 2008). Furthermore, activation of fast-spiking interneurons can pace pyramidal cells to theta frequencies in brain slices (Cobb et al., 1995), in vivo (Stark et al., 2013) and in the whole hippocampal preparation (Amilhon et al., 2015). Our study found that the activity of fastfiring interneurons was key to setting the timescale of autonomous oscillations in the whole hippocampal preparation. We show that without the large inhibitory conductance normally provided by fast-spiking interneurons, the resultant activity is more irregular with a much lower average period. Without inhibition, runaway excitation in the recurrent pyramidal network is responsible for initiating population spiking, which ceases once pyramidal cells encounter depolarization block, and enter a long-lasting refractory period. Blocking glutamatergic receptors completely suppressed this activity, suggesting that cholinergic or GABAergic signaling is unlikely to drive synchronous pyramidal cell spiking during autonomous oscillations in the subiculum. 52 With inhibition intact, recurrent excitation still initiates concurrent population spiking, but is tightly balanced by fast-spiking interneurons to maintain smaller depolarizations and spiking at much lower rates in pyramidal cells. Pyramidal cell spiking at low or high rates was associated with shorter and longer refractory periods, respectively. This indicates that balanced network properties which keep pyramidal cell spiking at low rates contribute to setting the timescale of autonomous oscillations by preventing runaway excitation that leads to prolonged refractory period and longer, more irregular network inter-event times. Role of potassium-channel conductance in setting the timescale of hippocampal oscillations In other tightly balanced networks, the refractory period, or timescale of oscillations, is set by potassium-channel conductances that increase proportionally to the rate of spiking during the peak of synaptic input, and recover to baseline levels prior to the next oscillation cycle (Sanchez-Vives & McCormick, 2000; Compte et al., 2003; Tahvildari et al., 2012). Potassium-channel conductance has been known to influence refractory dynamics by increasing the afterhyperpolarization period and inter-spike interval. We show that during the peak of injected conductance, an increase in spikinginduced potassium-channel conductance has a lasting effect on the refractory state of pyramidal cells. The timescale in which potassium conductance recovers to baseline levels (~ 80-320 ms) is consistent with the theta frequency band (3-12 Hz). The amount that potassium-channel conductance increases as a result of depolarization and spiking during autonomous oscillations is unlikely large enough to 53 interrupt spiking in pyramidal cells in the same manner as depolarization block. Instead, the large amount of feedback inhibition likely shunts recurrent excitation in pyramidal cells to prevent runaway spiking that could otherwise affect oscillation frequency. An increase in network excitatory drive would be balanced with commensurate feedback inhibition, lowering membrane resistance which reduces the ability of synaptic inputs to affect subthreshold voltage. Because pyramidal cells exhibit spike-frequency adaptation (Fernandez & White, 2010; Fernandez et al., 2011), in which a constant rate of firing would require ever-increasing excitatory drive, this could provide a network mechanism to limit pyramidal cell spiking to a specific rate. 54 Figure 3.1 Bath applications of gabazine unveils the recurrent excitatory network in pyramidal cells that drives intrinsic hippocampal oscillations. A. LFP and membrane voltage of 3 example pyramidal cells after bath application of gabazine, which takes effect at ~40 seconds. Coherent oscillations transform into random fluctuations marked by longer cycles and greater depolarizations. B. Spectogram analysis of the LFP and membrane voltage of the 3 pyramidal cells, during the 5-minute time period after bath application of gabazine. C. Cross-correlation between the LFP and membrane voltage of all pyramidal cells (n = 6) before (top) and after gabazine has taken effect. Loss of rhythmicity in the LFP and membrane voltage is apparent; however, cross correlation at short timescales remains high. 55 56 Figure 3.2 Gabazine transforms rhythmic activity in LFP and membrane voltage of pyramidal cells into random and more intense fluctuations. A. Averaged membrane voltage as a function of LFP phase from pyramidal cells (n = 6) before and after gabazine effect. Ai. Membrane voltage is significantly more depolarized during bouts of activity post-gabazine. Gabazine also lowered the variance in peak angle during depolarizations but not to a significant degree. B. Averaged LFP as a function of LFP phase concurrently recorded during whole-cell patch of pyramidal cells (n = 6). Bi. Peak amplitude in the LFP tended to increase post-gabazine but not significantly. Peak angle was not significantly changed by gabazine. 57 Figure 3.3 Subsequent bath application of DNQX after gabazine blocks recurrent excitation in pyramidal cells. A. LFP and membrane voltage of 3 example pyramidal cells after bath application of DNQX, which similarly to gabazine, takes effect at ~40 seconds. Cross-correlated LFP and membrane voltage fluctuations are eliminated, as well as spiking in pyramidal cells. B. Spectrogram analysis of the LFP and membrane voltage of the 3 pyramidal cells, during the 2-minute time period after bath application of DNQX. C. Distribution of membrane voltage skew (top) and standard deviation (bottom) from all pyramidal cells (n = 6) before and after bath application of gabazine. Pyramidal cells previously driven to depolarization block are rendered quiescent. 58 59 Figure 3.4 Injection of conductance to elicit spiking in quiescent pyramidal cells alters subsequent spiking response to injected conductance. A. Injection of paired conductance steps (black) into an example pyramidal cell with varying amounts of delay. If the delay is too short (left), refractory dynamics will prevent the second conductance step from eliciting an action potential in the membrane voltage (blue). Spiking returns after a sufficiently long delay between the two conductance steps (right). B. Data from all pyramidal cells (n = 12) indicate that spiking during the second conductance step returns when the delay is around the timescale of autonomous oscillations in the whole hippocampal preparation. C. Membrane voltage traces from two example pyramidal cells during the second conductance step at multiple delays. These pyramidal cells exhibited the least (top) and most (bottom) amount of sag potential, indicative of Ih current. The small amount of sag did not appear to shift spiking preference for delays outside the timescale of theta oscillations. 60 Figure 3.5 Injection of either current or conductance to elicit spiking in quiescient pyramidal cells has a similar effect on refractory dynamics. A. Paired steps of initial conductance (left) or current (right) to elicit a single action followed by a current step in both cases to elicit continuous spiking. During the second step, the first action potential is analyzed for spiking characteristics impacted by refractory dynamics as a function of delay between the two steps. B. Unlike the spike dynamics during the delayed second step, spiking during the initial step, driven by conductance (blue), had a higher spike threshold compared to spikes elicited by an initial current (green) step. C. Spiking characteristics of the first action potential during the second current step, after an initial current (green) or conductance (blue) step. Median and interquartile (25 - 75%) values are presented (red = conductance, grey = current) for the rate of action potential rise (top) and action potential height (bottom) as a function of delay between the two steps, presented on a logarithmic scale. D. Mean and interquartile values as function of delay, presented on a linear scale. These spiking characteristics, that are more sodium-channel dependent, recover on a timescale well outside of theta (80 320 ms). 61 Figure 3.6 Spiking characteristics that are determined by potassium-channel conductances recover within the timescale of the theta frequency band. A. Spiking characteristics of the first action potential during the second current step, after an initial current (green) or conductance (blue) step. Median and interquartile (25-75%) values are presented (red = conductance, grey = current) for the rate of action potential fall, threshold, half-width, interspike interval, and time to first spike, as a function of delay between the two steps, presented on a logarithmic scale. B. Mean and interquartile values as function of delay, presented on a linear scale. These spiking characteristics, which are more potassium-channel dependent, recover on a timescale within the theta range (80 - 320 ms). 62 CHAPTER 4 CONCLUSION Major Findings Our studies in Chapter 2 and 3 provide evidence for an alternative mechanism in generating hippocampal theta oscillations in the isolated hippocampus. In Chapter 2, utilizing electrophysiological and imaging techniques along with pharmacological manipulation, we gathered intracellular and spiking data to describe the phase-dependent activity of both excitatory and inhibitory neurons during autonomous hippocampal oscillations. By organizing and averaging the collected data as a function of the LFP phase, this allowed us to relate neuronal activity across different animals and cell types during an oscillation cycle. In this manner, we discovered that during the rising phase of an oscillation cycle, pyramidal cells experience a simultaneous increase of excitatory and inhibitory conductance, at a ratio that approximates balanced synaptic currents. This results in depolarizations that produce 1-2 spikes in pyramidal cells at the peak of net synaptic conductance, in spite of the large inhibitory conductance. After spiking at the peak of network oscillations, pyramidal cells experience a simultaneous decrease in excitatory and inhibitory conductance during the falling phase, and as suggested by our data in Chapter 3, likely enter a refractory period until the next cycle. 64 Fast-firing PV+ interneurons also experience an increase in excitatory conductance during the rising phase of the oscillation cycle, but very little inhibitory conductance. Similar to pyramidal cells, interneurons fire at the peak of net synaptic conductance, at spike rates only slightly higher than pyramidal cells. Subsequently, spiking subsides, and the large excitatory conductance decreases to baseline levels during the falling phase. Overall, our data from Chapter 2 point to a recurrent excitatory network of pyramidal cells in the subiculum that recruit proportional amounts of feedback inhibition from fast-firing interneurons. This study is the first to show experimentally that balanced network properties are capable of generating oscillations near the theta frequency range. Other studies of balanced networks generate oscillations at either higher or lower frequencies compared to theta (Shu et al., 2003; Haider et al., 2006; Atallah & Scanziani, 2009; Tahvildari et al., 2012; Gan et al., 2017). Although most of our data were collected during the slower frequencies of the theta band (~3 Hz), oscillations at higher theta frequencies (5-7 Hz) did not seem to affect the spike phase-locking profiles of either cell type, and pyramidal cells still exhibited large conductances with ratios indicative of balanced synaptic currents. Because our results suggest autonomous oscillations are generated by balanced network mechanisms not commonly associated with theta rhythmogenesis, our data are in opposition to previous work in area CA1 of the whole hippocampal preparation (Huh et al., 2016), and in vivo (Klausberger & Somogyi, 2008; Stark et al., 2013). For these studies, fast-firing interneurons spiked prior to pyramidal cell firing, indicative of a mechanism that involves post-inhibitory rebound-spiking. The cellular property that 65 enables this mechanism is considered to be the Ih current, which imparts theta resonance and rebound-spiking dynamics to neurons (Cobb et al., 1995; Dickson et al., 2000; Hu et al. 2002; Haas et al., 2007; Zemankovics et al., 2010). We did not find evidence of post-inhibitory rebound spiking in subicular pyramidal cells during autonomous oscillations. Instead, pyramidal cells fired in-phase with interneurons, similar to spiking behavior during movement-related (type I) theta oscillations (Fox et al., 1986; Skaggs et al., 1996; Csicsvari et al., 1999). Furthermore, the large synaptic conductance in pyramidal cells during autonomous oscillations likely attenuates theta resonance (Fernandez & White, 2008). Therefore, our results suggest that the Ih current is unlikely to set the timescale for autonomous oscillations in the subiculum. Instead, we provide evidence in Chapter 3 that the timescale for oscillations could be dependent on spiking behavior of pyramidal cells and its effect on refractory dynamics, similar to lower frequency oscillations (Sanchez-Vives & McCormick, 2000; Shu et al., 2003; Haider et al., 2006; Tahvildari et al., 2012). By blocking GABAa receptors during autonomous oscillations, we show that without the large inhibitory conductance normally present, the recurrent network of excitatory pyramidal cells (Harris et al., 2001) enters a runaway spiking regime which extends the LFP cycle that ends when the population of pyramidal cells encounter depolarization block. This suggests that during autonomous oscillations, feedback inhibition, which is recruited proportionally to the amount of network excitation, limits the spike rate of the pyramidal cell population. Furthermore, we show in Chapter 3 that the rate at which pyramidal cells fire during autonomous oscillations puts pyramidal cells into a refractory state that recovers on a timescale matching the theta frequency. During 66 this refractory period, a step of artificial conductance that previously elicited spiking in pyramidal cells, is not sufficient to elicit further spiking until the refractory period is over. Our results in Chapter 3 implicate the buildup of potassium-channel conductances during pyramidal cell spiking as a likely candidate in setting the refractory period. An increase in potassium conductivity or decrease in sodium conductivity reduces neuronal excitability which likely underlies the refractory state. We measured changes in conductivity indirectly by analyzing the difference in spike characteristics relative to baseline levels at multiple points within the refractory period. Certain spike characteristics are indicative of specific voltage-gated conductances underlying the generation of action potentials. For example, the negative derivative and half-width of the spike is a marker for the magnitude of potassium-channel conductances, whereas the positive derivative and spike height are more indicative of sodium-channel conductivity. With this method, we found that sodium-channel conductivity recovered on a timescale substantially longer than the theta frequency, whereas the recovery of potassium-channel conductivity matched the theta timescale and recovery of spiking behavior. Future Directions In Chapter 3, we indirectly quantified the changes in voltage-gated conductances that could lead to a decrease in pyramidal cell excitability during the refractory period and, importantly, the timescale of their recovery. Additional experiments utilizing voltage-clamp and pharmacology would be useful to determine directly the changes in sodium and potassium-channel conductivity and their timescale of recovery. Applying a 67 depolarizing step that mimics low-rate spiking behavior in pyramidal cells, then quickly switching to voltage-clamp at multiple holding voltages, would reveal the magnitude change of these voltage-gated conductances relative to baseline, as well as kinetics of potassium-channel deactivation or sodium-channel de-inactivation, when using sodiumor potassium-channel blockers, respectively. Furthermore, repeated depolarizations at theta frequencies could elucidate longterm changes in voltage-gated conductances that may explain the slow decrease in LFP oscillation frequency over time in the whole hippocampal preparation. Since our results in Chapter 3 indicate that potassium-channels recover to baseline values within the theta timescale but sodium-channels take much longer, repeated depolarizations at theta frequencies would likely not allow for full recovery of sodium conductivity leading to a decrease in neuronal excitability over the long-term, potentially extending the refractory period of pyramidal cells and slowing oscillations. Our experiments in Chapter 2 are performed in an in vitro preparation that does not include structures normally associated with theta oscillations, such as the medial septum “pacemaker” and entorhinal cortex. However, our results in Chapter 2 provide evidence for a balanced network architecture in subiculum that seems particularly tuned for the lower end of the theta-frequency band. In cortical networks, balanced networks are tuned for slower oscillations, exhibiting longer periods of pyramidal cell spiking. The mechanisms for slow cortical oscillations were first uncovered in slice preparations (Sanchez-Vives & McCormick, 2000; Shu et al., 2003), and later confirmed in vivo (Haider et al., 2006). Similarly, the ideal experimental study would be to confirm our results in vivo with measurements of synaptic conductance and spiking output relative to 68 the phase of LFP theta oscillations. These experiments are technically challenging but would help confirm our results that the subiculum has the necessary properties to function as the pacemaker of theta oscillations. Lastly, our experimental studies could be followed by a computational modeling study that recreates the balanced network and refractory dynamics we described. This model would have to be conductance-based rather than using integrate-and-fire neurons, as we determined that the large inhibitory conductance was imperative to setting the spike rate of pyramidal cells through changes in subthreshold response and spikefrequency adaptation. REFERENCES Amaral DG, Dolorfo C, Alvarez-Royo P (1991) Organization of CA1 projections to the subiculum: a PHA-L analysis in the rat. Hippocampus 1:415–435. Amilhon B, Huh CYL, Manseau F, Ducharme G, Nichol H, Adamantidis A, Williams S (2015) Parvalbumin interneurons of hippocampus tune population activity at theta frequency. Neuron 86:1277–1289. Anderson JS, Carandini M, Ferster D (2000) Orientation tuning of input conductance, excitation, and inhibition in cat primary visual cortex. J Neurophysiol 84:909– 926. Atallah BV, Scanziani M (2009) Instantaneous modulation of gamma oscillation frequency by balancing excitation with inhibition. Neuron 62:566–577. Bartos, M., Vida, I., Frotscher, M., Meyer, A., Monyer, H., Geiger, J. R. P., & Jonas, P. (2002). Fast synaptic inhibition promotes synchronized gamma oscillations in hippocampal interneuron networks. Proc Natl Acad Sci USA 99:13222–13227. Böhm, C., Peng, Y., Maier, N., Winterer, J., Poulet, J. F. A., Geiger, J. R. P., & Schmitz, D. (2015). Functional diversity of subicular principal cells during hippocampal ripples. J Neurosci 35:13608–13618. Boyce R, Glasgow SD, Williams S, Adamantidis A (2016) Causal evidence for the role of REM sleep theta rhythm in contextual memory consolidation. Science 352:812–816. Burgess N, Maguire EA, O’Keefe J (2002) The human hippocampus and spatial and episodic memory. Neuron 35:625–641. Cei A, Girardeau G, Drieu C, Kanbi KE, Zugaro M (2014) Reversed theta sequences of hippocampal cell assemblies during backward travel. Nature Neuroscience 17:719–724. Chance FS, Abbott LF, Reyes AD (2002) Gain modulation from background synaptic input. Neuron 35:773–782. Cobb SR, Buhl EH, Halasy K, Paulsen O, Somogyi P (1995) Synchronization of neuronal activity in hippocampus by individual GABAergic interneurons. Nature 378:75– 78. 70 Compte A, Sanchez-Vives MV, McCormick DA, Wang X-J (2003) Cellular and network mechanisms of slow oscillatory activity (<1 Hz) and wave propagations in a cortical network model. J Neurophysiol 89:2707–2725. Connors BW, Prince DA (1982) Effects of local anesthetic QX-314 on the membrane properties of hippocampal pyramidal neurons. J Pharmacol Exp Ther 220:476– 481. Csicsvari J, Hirase H, Czurkó A, Mamiya A, Buzsáki G (1999) Oscillatory coupling of hippocampal pyramidal cells and interneurons in the behaving Rat. J Neurosci 19:274–287. Denève S, Machens CK (2016) Efficient codes and balanced networks. Nat Neurosci 19:375–382. Deuchars J, Thomson AM (1996) CA1 pyramid-pyramid connections in rat hippocampus in vitro: dual intracellular recordings with biocytin filling. Neuroscience 74:1009– 1018. Drexel M, Romanov RA, Wood J, Weger S, Heilbronn R, Wulff P, Tasan RO, Harkany T, Sperk G (2017) Selective silencing of hippocampal parvalbumin interneurons induces development of recurrent spontaneous limbic seizures in mice. J Neurosci 37:8166–8179. English DF, Peyrache A, Stark E, Roux L, Vallentin D, Long MA, Buzsáki G (2014) Excitation and inhibition compete to control spiking during hippocampal ripples: intracellular study in behaving mice. J Neurosci 34:16509–16517. Fellous JM, Sejnowski TJ (2000) Cholinergic induction of oscillations in the hippocampal slice in the slow (0.5-2 Hz), theta (5-12 Hz), and gamma (35-70 Hz) bands. Hippocampus 10:187–197. Fernandez FR, Broicher T, Truong A, White JA (2011) Membrane voltage fluctuations reduce spike frequency adaptation and preserve output gain in CA1 pyramidal neurons in a high-conductance state. J Neurosci 31:3880–3893. Fernandez FR, White JA (2008) Artificial synaptic conductances reduce subthreshold oscillations and periodic firing in stellate cells of the entorhinal cortex. J Neurosci 28:3790–3803. Fernandez FR, White JA (2010) Gain control in CA1 pyramidal cells using changes in somatic conductance. J Neurosci 30:230–241. Fox SE, Wolfson S, Ranck JB (1986) Hippocampal theta rhythm and the firing of neurons in walking and urethane anesthetized rats. Exp Brain Res 62:495–508. Freund TF, Antal M (1988) GABA-containing neurons in the septum control inhibitory interneurons in the hippocampus. Nature 336:170–173. 71 Gabernet L, Jadhav SP, Feldman DE, Carandini M, Scanziani M (2005) Somatosensory integration controlled by dynamic thalamocortical feed-forward inhibition. Neuron 48:315–327. Gan J, Weng S, Pernía-Andrade AJ, Csicsvari J, Jonas P (2017) Phase-locked inhibition, but not excitation, underlies hippocampal ripple oscillations in awake mice in vivo. Neuron 93:308–314. Goutagny R, Jackson J, Williams S (2009) Self-generated theta oscillations in the hippocampus. Nat Neurosci 12:1491–1493. Haider B, Duque A, Hasenstaub AR, McCormick DA (2006) Neocortical network activity in vivo is generated through a dynamic balance of excitation and inhibition. J Neurosci 26:4535–4545. Harris E, Witter MP, Weinstein G, Stewart M (2001) Intrinsic connectivity of the rat subiculum: I. Dendritic morphology and patterns of axonal arborization by pyramidal neurons. J Comp Neurol 435:490–505. Harvey CD, Collman F, Dombeck DA, Tank DW (2009) Intracellular dynamics of hippocampal place cells during virtual navigation. Nature 461:941–946. Hasselmo ME, Fehlau BP (2001) Differences in time course of ACh and GABA modulation of excitatory synaptic potentials in slices of rat hippocampus. J Neurophysiol 86:1792–1802. Higley MJ, Contreras D (2006) Balanced excitation and inhibition determine spike timing during frequency adaptation. J Neurosci 26:448–457. Hippenmeyer S, Vrieseling E, Sigrist M, Portmann T, Laengle C, Ladle DR, Arber S (2005) A developmental switch in the response of DRG neurons to ETS transcription factor signaling. PLoS Biol 3:e159. Hu H, Vervaeke K, Storm JF (2002) Two forms of electrical resonance at theta frequencies, generated by M-current, h-current and persistent Na+ current in rat hippocampal pyramidal cells. J Physiol (Lond) 545:783–805. Huh CYL, Amilhon B, Ferguson KA, Manseau F, Torres-Platas SG, Peach JP, Scodras S, Mechawar N, Skinner FK, Williams S (2016) Excitatory inputs determine phaselocking strength and spike-timing of CA1 stratum oriens/alveus parvalbumin and somatostatin interneurons during intrinsically generated hippocampal theta rhythm. J Neurosci 36:6605–6622. Jackson J, Amilhon B, Goutagny R, Bott J-B, Manseau F, Kortleven C, Bressler SL, Williams S (2014) Reversal of theta rhythm flow through intact hippocampal circuits. Nat Neurosci 17:1362–1370. Jackson J, Goutagny R, Williams S (2011) Fast and slow γ rhythms are intrinsically and 72 independently generated in the subiculum. J Neurosci 31:12104–12117. Klausberger T, Somogyi P (2008) Neuronal diversity and temporal dynamics: the unity of hippocampal circuit operations. Science 321:53–57. Kramis R, Vanderwolf CH, Bland BH (1975) Two types of hippocampal rhythmical slow activity in both the rabbit and the rat: relations to behavior and effects of atropine, diethyl ether, urethane, and pentobarbital. Exp Neurol 49:58–85. Lee MG, Chrobak JJ, Sik A, Wiley RG, Buzsáki G (1994) Hippocampal theta activity following selective lesion of the septal cholinergic system. Neuroscience 62:1033–1047. Lynch MA (2004) Long-term potentiation and memory. Physiol Rev 84:87–136. Maccaferri G, McBain CJ (1996) The hyperpolarization-activated current (Ih) and its contribution to pacemaker activity in rat CA1 hippocampal stratum oriens-alveus interneurones. J Physiol (Lond) 497 (Pt 1):119–130. Madisen L, Zwingman TA, Sunkin SM, Oh SW, Zariwala HA, Gu H, Ng LL, Palmiter RD, Hawrylycz MJ, Jones AR, Lein ES, Zeng H (2010) A robust and highthroughput Cre reporting and characterization system for the whole mouse brain. Nat Neurosci 13:133–140. Maren S, Phan KL, Liberzon I (2013) The contextual brain: implications for fear conditioning, extinction and psychopathology. Nat Rev Neurosci 14:417–428. O’Keefe J, Recce ML (1993) Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus 3:317–330. Okun M, Lampl I (2008) Instantaneous correlation of excitation and inhibition during ongoing and sensory-evoked activities. Nat Neurosci 11:535–537. Peng, Y., Barreda Tomás, F. J., Klisch, C., Vida, I., & Geiger, J. R. P. (2017). Layerspecific organization of local excitatory and inhibitory synaptic connectivity in the rat presubiculum. Cerebral Cortex, 27: 2435–2452 Pike FG, Goddard RS, Suckling JM, Ganter P, Kasthuri N, Paulsen O (2000) Distinct frequency preferences of different types of rat hippocampal neurones in response to oscillatory input currents. J Physiol (Lond) 529 (Pt 1):205–213. Prescott SA, De Koninck Y (2003) Gain control of firing rate by shunting inhibition: roles of synaptic noise and dendritic saturation. Proc Natl Acad Sci USA 100:2076–2081. Ranganath C (2010) Binding items and contexts: the cognitive neuroscience of episodic memory. Curr Dir Psychol Sci 19:131–137. 73 Royzen F, Williams S, Fernandez FR, White JA (2019) Balanced synaptic currents underlie low-frequency oscillations in the subiculum. Hippocampus 29:1178-1189 Rudolph M, Pospischil M, Timofeev I, Destexhe A (2007) Inhibition determines membrane potential dynamics and controls action potential generation in awake and sleeping cat cortex. J Neurosci 27:5280–5290. Sanchez-Vives MV, McCormick DA (2000) Cellular and network mechanisms of rhythmic recurrent activity in neocortex. Nat Neurosci 3:1027–1034. Shu Y, Hasenstaub A, McCormick DA (2003) Turning on and off recurrent balanced cortical activity. Nature 423:288–293. Skaggs WE, McNaughton BL, Wilson MA, Barnes CA (1996) Theta phase precession in hippocampal neuronal populations and the compression of temporal sequences. Hippocampus 6:149–172. Stark E, Eichler R, Roux L, Fujisawa S, Rotstein HG, Buzsáki G (2013) Inhibitioninduced theta resonance in cortical circuits. Neuron 80:1263–1276. Tahvildari B, Wölfel M, Duque A, McCormick DA (2012) Selective functional interactions between excitatory and inhibitory cortical neurons and differential contribution to persistent activity of the slow oscillation. J Neurosci 32:12165– 12179. Tóth K, Freund TF, Miles R (1997) Disinhibition of rat hippocampal pyramidal cells by GABAergic afferents from the septum. J Physiol (Lond) 500 (Pt 2):463–474. Tsien JZ, Chen DF, Gerber D, Tom C, Mercer EH, Anderson DJ, Mayford M, Kandel ER, Tonegawa S (1996) Subregion- and cell type-restricted gene knockout in mouse brain. Cell 87:1317–1326. Wehr M, Zador AM (2003) Balanced inhibition underlies tuning and sharpens spike timing in auditory cortex. Nature 426:442–446. Wilent WB, Contreras D (2005) Dynamics of excitation and inhibition underlying stimulus selectivity in rat somatosensory cortex. Nat Neurosci 8:1364–1370. Witter MP (2007) Intrinsic and extrinsic wiring of CA3: indications for connectional heterogeneity. Learn Mem 14:705–713. Yang S, Yang S, Moreira T, Hoffman G, Carlson GC, Bender KJ, Alger BE, Tang C-M (2014) Interlamellar CA1 network in the hippocampus. PNAS 111:12919–12924. Yoder RM, Pang KCH (2005) Involvement of GABAergic and cholinergic medial septal neurons in hippocampal theta rhythm. Hippocampus 15:381–392. Zemankovics R, Káli S, Paulsen O, Freund TF, Hájos N (2010) Differences in 74 subthreshold resonance of hippocampal pyramidal cells and interneurons: the role of h-current and passive membrane characteristics. J Physiol (Lond) 588:2109– 2132.