Optogenetic “low-theta” pacing of the septohippocampal circuit is sufficient for spatial goal finding and is influenced by behavioral state and cognitive demand

Abstract

Hippocampal theta oscillations show prominent changes in frequency and amplitude depending on behavioral state or cognitive demands. How these dynamic changes in theta oscillations contribute to the spatial and temporal organization of hippocampal cells, and ultimately behavior, remain unclear. We used low-theta frequency optogenetic stimulation to pace coordination of cellular and network activity between the medial septum (MS) and hippocampus during baseline and MS stimulation while rats were at rest or performing a spatial accuracy task with a visible or hidden goal zone. Hippocampal receptivity to pan-neuronal septal stimulation at low-theta frequency was primarily determined by speed and secondarily by task demands. Competition between artificial and endogenous field potentials at theta frequency attenuated hippocampal phase preference relative to local theta, but the spike-timing activity of hippocampal pyramidal cells was effectively driven by artificial septal output, particularly during the hidden goal task. Notwithstanding temporal reorganization by artificial theta stimulation, place field properties were unchanged and alterations to spatial behavior were limited to goal zone approximation. Our results indicate that even a low-theta frequency timing signal in the septohippocampal circuit is sufficient for spatial goal finding behavior. The results also advance a mechanistic understanding of how endogenous or artificial somatodendritic timing signals relate to displacement computations during navigation and spatial memory.

1 |. INTRODUCTION

A key feature of how neural activity represents movement and space is its dynamic organization in time (Buzáki, 2006; Buzsaki & Llinas, 2017; O’Keefe & Nadel, 1978). The temporal organization of field potentials and cell activity by theta oscillations (5–12 Hz) has been proposed to be critical for the self-motion based estimation of linear distance through the integration of speed, distance and time (Burgess, 2008; Buzsaki, 2005; Buzsaki & Llinas, 2017; Hasselmo & Brandon, 2008; Jacob et al., 2017; McNaughton et al., 1996). The phasic organization of cell activity by theta has also been proposed to underlie plasticity mechanisms for the segregation of memory encoding and recall (Douchamps, Jeewajee, Blundell, Burgess, & Lever, 2013; Hasselmo, Bodelon, & Wyble, 2002; Hasselmo & Stern, 2014) as well as working memory (Kaminski, Brzezicka, Mamelak, & Rutishauser, 2020). Phasic discoordination due to pharmacological or gene manipulation, or status epilepticus in early life, correlates with memory impairment (Bender, Luikart, & Lenck-Santini, 2016; Robbe & Buzsaki, 2009) and spatial deficits (Barry et al., 2016). These interdependent processes, memory and spatial navigation, are both accomplished by neural throughput within and between the hippocampus and neocortex (Mizuseki, Sirota, Pastalkova, & Buzsaki, 2009), as well as subcortical projections from the medial septum (MS) (Colom, Castaneda, Reyna, Hernandez, & Garrido-Sanabria, 2005; Justus et al., 2017; Manns, Mainville, & Jones, 2001; Salib et al., 2019; Unal et al., 2018).

The MS has been proposed to act as a pacemaker for theta oscillations in the hippocampus (Stewart & Fox, 1989; Stewart & Fox, 1990; Zutshi et al., 2018) by generating rhythmic disinhibition at the synapses of hippocampal pyramidal neurons via long-range axons to hippocampal GABAergic neurons (Freund & Antal, 1988; Fuhrmann et al., 2015; Gulyas, Gorcs, & Freund, 1990; Hangya, Borhegyi, Szilagyi, Freund, & Varga, 2009; Petsche, Stumpf, & Gogolak, 1962; Toth, Freund, & Miles, 1997). Pharmacological inactivation of the MS and subsequent loss of theta oscillations in the hippocampus alter spatially tuned cells in the hippocampal-entorhinal circuit, where grid cell spacing (Brandon et al., 2011; Koenig, Linder, Leutgeb, & Leutgeb, 2011) and the goal-directed increase of in-field place cell firing rates are both disrupted (Aoki, Igata, Ikegaya, & Sasaki, 2019). However, although septal inactivation does not affect the spatial firing properties of hippocampal place cells in familiar or novel environments (Aoki et al., 2019; Bolding, Ferbinteanu, Fox, & Muller, 2019; Brandon, Koenig, Leutgeb, & Leutgeb, 2014; Koenig et al., 2011), the loss of the MS and hippocampal theta significantly interferes with navigation to a hidden goal and spatial reference memory (Bolding et al., 2019; Hagan, Salamone, Simpson, Iversen, & Morris, 1988; Winson, 1978) as well as working memory (Wang, Romani, Lustig, Leonardo, & Pastalkova, 2015).

Several questions remain regarding how and when the MS contributes to the dynamic temporal organization of hippocampal cells by pacing oscillations throughout theta frequency range. This dynamism in theta-scale pacing of local field potentials (LFPs) and timing of cell activity not only occurs in relation to behavioral variables such as speed (Blumberg et al., 2016; Geisler, Robbe, Zugaro, Sirota, & Buzsaki, 2007; Huxter, Burgess, & O’Keefe, 2003; Justus et al., 2017; Lu & Bilkey, 2010; Richard et al., 2013) and novelty (Jeewajee, Lever, Burton, O’Keefe, & Burgess, 2008; Wells et al., 2013), but spatial and cognitive demands also correlate with phasic organization levels in hippocampal cells (Barry et al., 2016; Fernandez-Ruiz et al., 2017; Schomburg et al., 2014). From both basic science and clinical perspectives, the logical extension of these findings is that stimulating the MS to correct theta rhythmopathy could also improve associated cognitive deficits (Barry et al., 2016; Fenton, 2015; Reinhart & Nguyen, 2019; Shuman, Amendolara, & Golshani, 2017; Solomon et al., 2018). However, our poor understanding of how endogenous and artificial temporal signals might interact within intact circuits precludes a mechanistic understanding of neurostimulation as a clinical treatment option (Ezzyat et al., 2017). This challenge is exacerbated by the consistently changing throughput dynamics between the MS pattern generator and its downstream targets in accordance with behavioral state (Carpenter, Burgess, & Barry, 2017; Dragoi, Carpi, Recce, Csicsvari, & Buzsaki, 1999).

We therefore measured the baseline timing relationship between the MS and hippocampus at rest and during varied cognitive demand in a hidden or visible goal spatial accuracy task. We then applied static, low-frequency optogenetic artificial theta stimulation (ATS) in the MS to either supersede or compete with dynamic endogenous septohippocampal theta. By addressing how endogenous and artificial timing signals interact, according to behavior and corresponding septohippocampal circuit state, we were able to test theoretical predictions of how theta-scale temporal coordination underpins spatial memory and goal finding behavior.

The first experimental goal was to test for limits to pan-neuronal optogenetic entrainment of the hippocampus. Specifically, we tested whether behavioral state and cognitive demand affected CA1 theta rhythmicity, as well as pyramidal cell phase preference and burst-timing. The relationship between optogenetic entrainment efficacy and cognitive demand at a cell and network level has never been tested, either in wild-type or transgenic animals. In relation to this goal, we addressed two predictions: (a) as endogenous hippocampal theta amplitude and frequency scale linearly with movement speed (Hinman, Penley, Long, Escabi, & Chrobak, 2011; Richard et al., 2013; Vanderwolf, 1969), we countered the upward theta frequency shift at faster speeds by constraining the ATS frequency downward at 6 Hz. Prevalence of 6 Hz signal, even at faster speeds, would indicate MS ATS dominates endogenous theta; (b) increased memory (Fernandez-Ruiz et al., 2017; Schomburg et al., 2014) and spatial cognitive demands (Barry et al., 2016) correlate with the phasic organization of hippocampal pyramidal cells relative to local theta. We predict that cell responsiveness to either baseline theta or low-frequency artificial theta will increase with spatial cognitive demand.

The second goal was to test the effect of superseding or perturbing endogenous theta rhythmicity on active spatial cognition. Although endogenous timing of the septohippocampal circuit has been proposed to be necessary for spatial goal finding behavior (Aoki et al., 2019; Bolding et al., 2019; McNaughton, Ruan, & Woodnorth, 2006; Winson, 1978), it is unknown whether supplanting endogenous rhythmicity with low-frequency ATS would be sufficient or disruptive for spatial behavior. We address three theoretical predictions that low-frequency stimulation would disrupt spatial behavior in a theta dependent task (Bolding et al., 2019): (c) hippocampal theta shifts to lower frequencies in reaction to novelty (Jeewajee et al., 2008; Villette et al., 2010; Wells et al., 2013). This phenomenon is absent in a rodent Alzheimer’s disease model (Villette et al., 2010) and activation of septal cholinergic cells elicits hippocampal theta frequency reduction and novelty-responsive behavior (Carpenter et al., 2017). We therefore test the prediction that reducing theta frequency could elicit novelty-responsive behavior and disrupt spatial accuracy performance; (d) ATS could create a microcircuitry conflict in the endogenous theta timing drivers affecting dendritic integration and the burst-timing or phasic organization of cell outputs (Royer et al., 2012; Vaidya & Johnston, 2013). This could register as a discrepancy between endogenous versus artificial entrainment of the LFP (Blumberg et al., 2016) and the subsequent burst timing or phasic organization of single cells (Barry et al., 2016; Zutshi et al., 2018). If open-loop low-frequency stimulation disrupts plasticity regimes in the phasic organization of hippocampal cells associated with memory retrieval (Douchamps et al., 2013; Hasselmo et al., 2002; Hasselmo & Stern, 2014; Siegle & Wilson, 2014), we would predict deficits in goal location memory; (e) the MS has been proposed to be part of a mechanism for translating movements into displacement coordinates among place fields in the cognitive map (O’Keefe & Nadel, 1978) where resultant theta oscillations in the septohippocampal-entorhinal network support spatial navigation (Burgess, 2008; Hasselmo & Brandon, 2008). By separating “high-frequency theta” (> 7 Hz) during mobility from “low-frequency” theta (< 7 Hz) associated with immobility and novelty, 6 Hz entrainment could disrupt the theta frequency and speed relationship underpinning mobility-based displacement metric calculations. If the linear speed and theta frequency relationship is necessary for self-localization and navigation (Burgess, 2008; Hasselmo & Brandon, 2008; O’Keefe & Nadel, 1978), we predict that ATS should induced deficits in goal finding.

Septal ATS entrainment of hippocampal timing was primarily a function of movement speed and secondarily a function of spatial attention. Despite temporal reorganization, place field properties were unaffected, and rats were able to locate a hidden goal in the spatial accuracy task. A cohesive timing signal in the septohippocampal circuit may be both necessary and sufficient for spatial memory and goal finding, but a static low-frequency signal may disrupt theta mediated displacement metric calculations in path integration processes. Implications for physiological theories of spatial memory and navigation as well as the use of neuroprosthetics for correcting theta rhythmopathy are discussed.

2 |. METHODS

2.1 |. Subject details

Seven male Sprague–Dawley rats (obtained from Charles River, Montreal), 3–8 months old were subjects in this study. Animals were housed individually and maintained on a 12-hr light/dark cycle and 85% of baseline body weight and given free access to water. All procedures were approved by the University of Vermont’s Institutional Animal Care and Use Committee and conducted in accordance with guidelines from the National Institutes of Health.

2.2 |. Injection surgery

At approximately 3 months old, four animals underwent the surgical procedure for the intracranial injection. Two of these animals were common to both experiments. Rats were anesthetized with a mixture of 5% inhaled isoflurane in oxygen and placed in a stereotaxic frame. Anesthesia was maintained at 1.5% for the duration of the procedure. All stereotaxic coordinates were relative to bregma (Paxinos & Watson, 1998). The skull was exposed, and a burr hole was made in the skull. A Hamilton injection syringe was then used to deliver 1 μL bolus injections of the Adeno-Associated viral vector (AAV2-hSyn-hChR2[H134R]-EYFP; 5.7 × 10¹² virus molecules/ml; UNC Vector Core, Chapel Hill, NC) at a rate of 0.1 μl/min into the vertical limb of the diagonal band of Broca in the MS (Anterior/Posterior (AP) = 0.7 mm; ML = −1.4 mm) and angled 12° medially. The first injection of 0.15 μl was made at a depth of 7.1 mm from brain surface and approaching the midline. The syringe was retracted three times at 0.3-mm steps with injections of 0.2, 0.25, and 0.2 μl at each consecutive step. In the final step, the needle was raised 0.2 mm to a final depth of 6.0 mm and an ultimate injection of 0.2 μl was made. The wound was sutured, and the rats were returned to their home cages to recover.

2.3 |. Logic of pan-neuronal septal stimulation

We chose to use wild-type rats and inject a nonselective AAV in the MS. This is an explicit choice as we propose to test for any limitations in the efficacy of pan-neuronal entrainment of septohippocampal theta oscillations at both the cell and network levels in conjunction with titrated behavioral state and cognitive demand. Transgenic rat lines that are necessary for cell-type selective optogenetic stimulation are still difficult for many researchers to obtain. This is an obstacle for behavioral neuroscientists whose tasks might be limited to the use of rats as subjects (Iannaccone & Jacob, 2009). Furthermore, it is unknown if the ability to pace hippocampal oscillations by cell-type selective stimulation of the MS is comparable between rats and mice.

Selective stimulation of mouse septal GABAergic cells (Etter et al., 2019; Zutshi et al., 2018) and glutamatergic cells (Fuhrmann et al., 2015; Robinson et al., 2016) lead to excellent control of hippocampal oscillations at most frequencies within the theta band. However, stimulation of glutamatergic axon terminals in the hippocampus does not alter theta frequency (Robinson et al., 2016). This finding supports the notion of an intraseptal tuning mechanism where septal glutamatergic neurons initiate synchronization with GABAergic neurons, and to a lesser extent cholinergic neurons, to pace hippocampal theta oscillations (Robinson et al., 2016). Septal cholinergic neurons, due to their contrasting temporal dynamics, are less likely to fire at the same time as GABAergic and glutamatergic neurons in response to stimulation (Simon, Poindessous-Jazat, Dutar, Epelbaum, & Bassant, 2006) and are more involved with theta amplitude than theta frequency (Mamad, McNamara, Reilly, & Tsanov, 2015; Vandecasteele et al., 2014; Zhang, Lin, & Nicolelis, 2010). Ultimately, septal GABAergic interneuron activation alone is sufficient to pace hippocampal theta due to their unique connectivity with hippocampal GABAergic interneurons and tendency to phase-lock to theta (Freund & Antal, 1988; Hangya et al., 2009; Petsche, Gogolak, Stumpf, & Newkirk, 1964; Stewart & Fox, 1989; Stewart & Fox, 1990; Unal et al., 2018).

Experiments using pan-neuronal optogenetic septal stimulation, in rats or mice, to control both the frequency of hippocampal theta and burst-timing rhythmicity of hippocampal pyramidal cells are conspicuous by their absence in the literature. Given the inherent intraseptal tuning mechanism between GABAergic and glutamatergic neurons over hippocampal theta frequency, we propose that our experiments using pan-neuronal septal stimulation will fill a critical gap in our understanding of septohippocampal temporal coordination. They also provide important data for best practice in the manipulation of hippocampal rhythmicity in rat behavioral paradigms. We contend that the “additive” stimulation approach can be as informative as the selective “subtractive” stimulation approach.

Finally, pan-neuronal optogenetics still has some advantages over electrical stimulation. It is less likely to create artifacts that complicate simultaneous stimulation and recording, and it is spatially restricted and may be less likely to stimulate en passant fiber tracts (Dai, Brooks, & Sheinberg, 2014).

2.4 |. Optical fiber and hippocampal tetrode implants

Detailed descriptions of the optical fiber preparation can be found in a previous publication and are described in brief here (Blumberg et al., 2016). We used a 200-μm multimode optic fiber (Thorlabs, CFLC230–10; Montreal, QC, Canada) as part of our chronic septal implant. The optical fiber was then glued to a 230-μm ferrule (Thorlabs, CFLC230–10; TP01235931). The percentage transmittance of light through the fiber was tested using 100% blue light transmittance (Spectralynx light-emitting diode [LED] source) and measured by a light meter with a photodiode sensor (Thorlabs; Model PM100D). For testing purposes, a 50-μm patch cable was used and only fibers that allowed for at least 70% light transmittance at approximately 0.5 mm from the tip of the optical fiber were used. The MS implant included an optic fiber with an array of eight recording electrodes glued to the surface that extended 0.25–0.5 mm from the end of the optic fiber. The custom hippocampal implant consisted of eight separately drivable tetrodes arranged in a circular array with a 1.25 mm diameter and spaced 0.25 mm apart (32 channel electrode array; “BLAK drives”; Grasshopper Machine Works, Wolfeboro, NH). Each tetrode was composed of four twisted 25-μm diameter nichrome wires (California Fine Wire, Grover Beach, CA). Each electrode was connected to custom millimax pins (Neuralynx, Bozeman, MT).

2.5 |. Optical probe and electrode array chronic implantation surgery

Approximately 2 weeks of postinjection, a second surgery was done in order to chronically implant the custom microdrives into the left hippocampus (ML = +3.3 mm; AP = −4.5 mm; Dorsal/Ventral (DV) = −2.0 mm). The optical/recording ensemble was lowered into the MS along the same path previously taken by the Hamilton injection syringe, with the end of the optical fiber lowered to a final depth of ~6.5 mm below brain surface. An illustration of the experimental approach using hippocampal septal and hippocampal implants is shown in Figure 1a.

FIGURE 1.

Histology, immunohistochemistry and experimental approach. (a) Experimental approach using optogenetics and electrophysiology in the dorsal hippocampus and medial septum (MS) via custom devices for the implantation of an optical fiber and EEG electrodes in the MS and tetrode array in the hippocampus. (b) The titration of movement and spatial cognitive demand. Left: Protocol for Experiment 1 where electrophysiological recordings occurred during 6 min of 6-Hz MS stimulations at both control and experimental blue light stimulation while the rat is at rest in a tall, narrow flowerpot (absence of movement and cognitive demand). Right: Protocol for Experiment 2 where electrophysiological recordings are carried out during 30 min of 6-Hz MS stimulations at both control and blue wavelengths while the rat performs a visible (goal-directed movement and absence of hippocampal cognitive demand) or hidden goal spatial accuracy task (goal-directed movement and hippocampal cognitive demand). (c) Recording tetrode tracks through the neocortex and CA1 of the dorsal hippocampus (white arrows). Note heavy projection of transduced MS axons to the alveus and str. oriens layers above CA1. (d) Estimation of the volume of viral transduction in the MS for P2, P3, and P4. Estimated Green Fluoresent Protein (GFP) expression volumes were consistent with the estimated total volume of the adult rat MS (mean = 1.69 ± 0.3 mm³). The density of GFP expression was similar at the injection site in the MS across all rats. Scale bar on lower right of each section = 500 μm. Analysis of GFP expression volumes for each slice in each rat are shown in Table 1. (e) At top left: MS Glutamic Acid Decarboxylase (GAD) stained GABAergic cell bodies (red) colocalized with GFP (green) and 4′,6-diamidino-2-phenylindole (DAPI in blue) (white arrows); MS Chat stained cholinergic cell bodies (red) colocalized with GFP and DAPI (white arrows); Antibody for MS Vesicular Glutamate Transporter (VGLUT in red) targets vesicular transporter in axon terminals rather than soma of MS cells and therefore shows poor colocalization with GFP and DAPI at the soma (*). Bottom left: 10× resolution image of VGLUT (red) stained hippocampal section with septal projection axons detected via GFP and hippocampal neurons detected via DAPI. Scale bar = 500 μm. Right: Three-dimensional rendering of 11 images taken at 2-μm steps of the section on the lower left (white box). Yellow puncta indicate the colocalization of VGLUT from MS axonal terminals and GFP in both str. pyramidale and str. oriens. Scale bar = 50 μm. (f) Left: Histology of optical and recording probe sites in the MS. Virally transduced MS axons and neurons in green via Enhanced Yellow Flourescent Protein (EYFP), 4′, DAPI-stained cell nuclei are visible in blue while GABAergic neurons are indicated in red (GAD). Dark areas indicate tracks occupied by the optical probe and EEG wires (top) and the initial viral injection via the Hamilton syringe (bottom). These regions are in the center of the MS, flanked by the ventral region of the lateral septum nucleus and the anterior commissure, above the horizontal and vertical limbs of the diagonal band of Broca (hDB and vDB). Right: Hippocampal GAD expression (white arrows) in red reveals Oriens-Lacunosum Moleculare (OLM) interneurons in str. oriens and basket cells in str. pyramidale as well as virally transduced MS axons (EYFP). DAPI-stained cell nuclei are visible in blue. The Adeno-Associated Virus therefore transduces MS cells whose axonal projections can be traced to the hippocampus, illustrating the microcircuitry through which the MS paces hippocampal theta

Four skull screws (FHC Inc.) were inserted. Two were anterior to bregma while the two remaining screws were placed over the left and right of the cerebellum. Grounding was achieved via connection to the right cerebellar screw while a reference wire was placed through a small burr hole at brain surface over the cerebellum. Both implants were fixed to the skull via the skull screws and Grip Cement (Dentsply Inc.). The wound was sutured, and topical antibiotic applied. The interval between surgery and the beginning of electrophysiological recording was 1 week.

2.6 |. Stimulation and recording protocols

A 200-μm multimode optic fiber (Thor Laboratories, Budapest, Hungary) was used to connect the implanted optic fiber’s 1.25 ceramic ferrule via patch cable to our light source. Rather than using laser stimulation, we used Spectralynx, a computer controlled optical LED system (Neuralynx). We chose to use LED stimulation to avoid possible photovoltaic artifacts associated with laser stimulation (Mikulovic et al., 2016).

The pulse program (Neuralynx) was used to stimulate the rats with experimental B (wavelength 470 nm) and control Y (wavelength 590 nm). Maximum light intensity was set to 100% (1.8–2.2 mW) Light transmittance to the MS via the chronically implanted optical fiber ranged from 70–85%, at the fiber tip. Intensity of both blue and yellow light wavelengths were therefore the same. While tissue penetration may be greater in the blue than yellow spectrum, this variable is negated by the proximity of the probe to the MS (see implantation coordinates above).

Sinusoidal wave stimulatory patterns were generated using the Pulse program. The stimulation frequencies of artificial sinusoidal waves at 6 Hz (peak/trough = 83.3 ms; period = 166.6 ms) were generated by creating an ascending and descending light intensity gradient with a mean light intensity of 52%. The light stimulation intensity was divided into 23 epochs corresponding to 0–255 bits where each increment was a percentage of peak amplitude at 255 bits (i.e., ranging from 1.6 to 100%).

The 6-Hz stimulation frequency was central to our experimental design as it allows for the separation of “high-frequency theta” (>7 Hz) during mobility from “low-frequency” theta (<7 Hz) associated with immobility and novelty or arousal (Kelemen, Moron, & Fenton, 2005; Kramis, Vanderwolf, & Bland, 1975; O’Keefe & Nadel, 1978; Sainsbury, Heynen, & Montoya, 1987; Vanderwolf, 1969). This stimulation frequency is both a challenge to the entrainment efficacy of the septohippocampal circuit, that should shift to higher frequencies at faster speeds, as well as a test of the functional effects of low-frequency theta in terms of eliciting novelty responses (Carpenter et al., 2017; Jeewajee et al., 2008; Wells et al., 2013) or disrupting the frequency and speed relationship components of displacement metric calculations necessary for path integration and goal finding (Burgess, 2008; Hasselmo & Brandon, 2008; O’Keefe & Nadel, 1978).

Two experiments were conducted to control for the influence of movement and cognition on the septohippocampal circuit and to understand how these variables affect entrainment efficacy through pan-neuronal optogenetic MS stimulation (Figure 1b). Both experiments included optical stimulation in two wavelengths, yellow (590 nm) and blue (470 nm). The yellow (Y) wavelength served as control stimulation as this wavelength does not activate channelrhodopsin (ChR2) light-gated ion channels (Blumberg et al., 2016). The Blue (B) wavelength served as experimental stimulation as it causes ChR2-ion channels in the membrane of transduced cells to open.

In Experiment 1, stimulation and recording occurred in a narrow flowerpot that prevented the animals from walking. They were therefore only at rest in this condition and did not experience cognitive demand. In Experiment 2, animals performed two versions of a spatial accuracy task modeled after the cue and place navigation tasks in the Morris Water maze (Morris, Garrud, Rawlins, & O’Keefe, 1982). The cued version of the task was hippocampal independent, as it requires stimulus-response or beacon strategies requiring the dorsolateral striatum or other extrahippocampal structures (Packard & McGaugh, 1996; White & McDonald, 2002). The place navigation version, that required searching for the hidden underwater platform, was hippocampal dependent (Morris et al., 1982) as hippocampal lesions caused a profound and enduring place navigational impairment that could not be attributed to motor, motivational or reinforcement deficits. Similarly, the spatial accuracy task in our experiment had two versions. In the first version the goal zone was cued and the second version the goal zone was hidden and required the rat to navigate to the remembered goal location via distal room cues and a spatial stimulus on the arena wall. The cued version of the task therefore required goal-directed movement and minimal cognitive demand while the hidden version of the task required goal-directed movement and increased spatial cognitive demand. As the goal zone is consistently in the same location, it may also be considered a test of reference memory (Olton, 1979).

2.6.1 |. Flowerpot

Recording sessions were made while rats rested in a 40-cm high ceramic flowerpot that was 27 cm wide at the base and lined with home cage bedding. The size of the ceramic pot limited movements of the rat to rearing and minor head movements and therefore limited theta oscillations associated with active exploration. The experiment began without stimulation for 2 min, followed by 6 min of 6-Hz control yellow wavelength stimulation, then 6 min of 6-Hz B wavelength stimulation and ending with an additional 2 min without stimulation.

The flowerpot also served as the location for screening session following the lowering of microdrives. Electrodes were advanced until they contacted the pyramidal cell layer in the CA1 field of the dorsal hippocampus, as characterized by the observation of clustered units as well as 140–200 Hz ripple activity in the local field potentials (Barry et al., 2016). Experimental sessions were carried out in the spatial accuracy task when >20 hippocampal cells were isolated.

2.6.2 |. Visible and hidden goal spatial accuracy tasks

Spatial accuracy is an operant reward task modeled after Kubie et al. (2007) and Bolding et al. (2019). The grey arena was 75 cm in diameter with a polarizing white cue card on the Northern wall that covered approcximately 90° of arc. The rat’s location in the arena was sampled at 30 Hz (Tracker, Bio-signal Group Corp, Brooklyn, NY) using a firewire digital camera that detected an LED on the preamplifier attached to the rat’s implant. As in Bolding et al., food-restricted rats entered a visible or hidden target zone on the southern part of the arena for 1.2 s to trigger food reward. Previous studies with goal zones at similar distances from polarizing cue cards found that behavioral choices correlated with place field location (Lenck-Santini, Save, & Poucet, 2001). The distance between the target center and the wall in both our experiment and Bolding et al. was ~19 cm, or half the arena radius. However, we decreased the target diameter from 16 to 8 cm in diameter in order to increase task difficulty. In alternating sessions, the target zone was hidden or made visible by a white, 20-mm diameter bottle cap. Dwell-time in the target zone for 1.2 s elicited a +5-V Transistor-Transistor Logic (TTL) pulse via the Peripheral Component Interconnect (PCI) board to a custom overhead feeder, triggering release of a food pellet reward (Bioserv, 20-mg dustless precision pellets) to a random location in the arena. After triggering reward, a refractory reward period of 5 s was set to encourage the animal to leave the target area and forage, thereby spatially sampling the entire arena. If the rats left the goal zone before the 1.2 s threshold, they would need to wait another 1.2 s to trigger food reward. Ultimately the rat’s spatial behavior provided a proxy measure of position estimation relative to the goal zone location. Videos of electrophysiological recordings during control stimulation in cued and hidden goal spatial accuracy sessions can be found as supporting multimedia files (MultiMedia 1 and MultiMedia 2, respectively).

Each series of sessions began with open-loop 6-Hz MS control optogenetic stimulation for 30 min during the cued version of the task. The experimenter did not interrupt the rat’s behavior for this 30-min period. Between sessions, rats were placed in a flowerpot alongside the arena for approximately 5 min. During this time the floor panel was cleaned with soap and water and rotated in order to remove potential olfactory cues. The rats were then returned to the arena for another 30 min of 6-Hz control stimulation during the hidden goal version of the task. We note that it was not expected that the rat would “forget” the goal location in this period between the hidden and visible goal sessions, but that the absence of the bottle cap in the goal zone would force the rat to navigate to the goal position using available spatial cues rather than a hippocampal independent beacon strategy (O’Keefe & Nadel, 1978). After the hidden goal session, the rats were returned to their home cages. Following approximately 1 hr in the home cage, the rats were placed back in the arena for successive 30-min visible and hidden goal sessions with open-loop 6-Hz B stimulation in each session. For each rat, the series of 6-Hz stimulation sessions, Yellow Cued (CtrlVis), Yellow Uncued (CtrlHid), Blue Cued (StimVis), and Blue Uncued (StimHid) was completed 3–4 times. In each series the order of the StimVis and StimHid sessions was alternated.

The spatial accuracy task allows for several behavioral measures that reflect the animal’s ability to self-localize relative to the goal zone. Basic measures include the number of goal zone entrances and dwell-time ratios in the goal zone and each arena quadrant (Target, Clockwise, Counter-Clockwise, and Opposite). We also utilized the complementary measures of dwell-time and speed relative to the goal zone center to graphically depict the animal’s choice behavior. We also used speed to calculate the search distance from the goal zone center as a function of speed. This was done using convolution windows (Matlab, Mathworks Inc.) to find continuous bouts of movement ≤ 3 cm/s between a lower time limit of 1.2 s and an upper time limit of 20 s. These speed boundaries were selected so as to distinguish pauses in the goal vicinity from instances where the rat might be travelling through the target quadrant while searching for food pellets. The mean distance of each slow movement epoch from the goal center served as a representative measure of search distance from the goal.

2.6.3 |. Foraging

Three additional animals were used as nonstimulated controls. As in previous experiments (Barry & Muller, 2011), rats were food restricted to 80% of baseline body weight and exposed to the same arena used in the spatial accuracy experiments. Food pellets fell at regular 30-s intervals from the overhead feeder. Each exposure lasted an hour a day for approximately 5 days. By the end of the foraging training, rats sampled the entire surface area of the arena within 30 min. Screening and recording sessions began a week after implantation of the same tetrode array as animals in the other conditions. During screening, electrodes were advanced until they contacted the pyramidal cell layer in the CA1 field, as characterized by the observation of clustered units as well as 140–200 Hz ripple activity in the local field potentials (Barry et al., 2016). Recordings consisted of two consecutive 30 min foraging sessions. Between sessions rats were held in the flowerpot for approximately 5 min, in which time the arena floor was cleaned. We wished to determine how successive sessions in the recording environment would affect the properties of theta oscillations. A further question was how movement and the absence of cognitive demand would affect resultant theta phase population vectors. Phase preference results could then be compared between the foraging task and cued version of the spatial accuracy task, given that we assume neither task requires hippocampal-dependent cognition (Morris et al., 1982). Results of a previous study showed that foraging does not lead to tight temporal control of CA1 cell activity by local theta oscillations (Barry et al., 2016).

2.7 |. Position tracking

The rat’s location in the arena was sampled using a digital camera that detected a LED placed near the back of the animal’s head and attached to the preamplifier. This tracking information was filtered and recorded utilizing custom software that allowed for the synchronization of the rat’s position and speed with properties of the recorded cell and Electroencephalography (EEG) signals. The rat’s behavior and positional information in relation to the goal zone was monitored using the Bio-signal Tracker program (Bio-Signal Group Corp, Brooklyn, NY). This program detected the LED in the goal zone and triggered the release of food pellets.

2.8 |. Signal processing

Rats were tethered to an electrophysiology cable during recording sessions in all conditions. Signals were preamplified X 1 at the headstage and channeled through the tether cable to the signal amplifiers and computer interface. Sampling frequency of LFPs was at 30.3 kHz and filtered at 1–9,000 Hz (Neuralynx), subsampled offline at 3000 Hz. Both MS and hippocampal LFPs were referenced against a 50-μm diameter stainless steel wire (California Fine Wire, Grover Beach, CA) placed at brain surface over the cerebellum. LFPs were processed offline using custom software that utilized the Matlab signal processing toolbox “spectrogram” function that returns the time-dependent fast Fourier transform (FFT) for a sequence using a sliding window (window = 1 s, overlap = 0.5 s). This form of the Fourier transform, also known as the short-time Fourier transform, produces a matrix of frequency, time and signal amplitude in the theta band at 5–12 Hz. The absolute value of the squared FFT signal, or power density, was then converted to decibels. As the amplitude of the signal is therefore a power ratio, it is described in the results using arbitrary units (A.U.).

2.8.1 |. Theta amplitude

The mean and standard error of the spectral signal in the theta band was calculated for four sampled CA1 channels across animals during control and blue light stimulation epochs in Experiment 1 and during each stimulation condition in Experiment 2 (CtrlVis, CtrlHid, StimVis, and StimHid). Signal peaks were then compared across conditions testing for statistical interactions between stimulation condition and frequency (see Section 2.9.1.).

2.8.2 |. Speed/theta properties

As in previous work (Blumberg et al., 2016; Richard et al., 2013), we analyzed the linear relationship between animal speed and either theta band frequency or amplitude during each version of the spatial accuracy task. We also analyzed speed and theta frequency during foraging sessions in non-stimulated control animals.

Instantaneous running speed was computed every 10 position samples (333 ms). Speed data were then linearly interpolated to fit the spectral data time range. For each session, the Pearson correlation coefficients (r) were computed between speed and theta power or speed and theta frequency for all collected data points. The slope of speed-theta relationships was estimated by fitting a line in the data using a least squares linear regression (Polyfit.m function, Matlab, Mathworks).

2.8.3 |. Theta coherence

The magnitude-squared coherence estimate was applied to LFPs in the MS and the CA1 region of the hippocampus as a measure of the degree to which the two regions exhibit coherent oscillatory activity. Coherence has been proposed to represent the effective connectivity or the routing of information between neural circuits (Fries, 2005). Coherence is a function of frequency with values between 0 and 1, indicating how well x corresponds to y at each frequency in the theta band. The magnitude-squared coherence is a function of the power spectral densities, P_xx(f) and P_yy(f), and the cross power spectral density, P_xy(f), of x and y (Matlab, Mathworks Inc.). One-second Hanning windows were used for analysis over the duration of the recording sessions in each experiment.

2.8.4 |. Phase preference

The number of action potentials from each hippocampal cell relative to the phase of the theta band in the local hippocampal or septal field potential was analyzed via phase preference analysis as in previous studies (Barry et al., 2016). Several criteria were imposed on the selection reference channel that served as the representative LFP signal for each tetrode. LFPs were only considered if they exhibited 140–200 Hz ripple activity, when the rat was at rest, and the wire had at least one cell. The selected LFP was filtered between 4 and 14 Hz using Chebyshev type 2 filters, then the phase was extracted using the Hilbert transform. All the spikes recorded from this tetrode and from neighboring tetrodes were assigned a phase value from the ongoing theta phase from the referenced LFP. The trough of the theta cycle was assigned to 0°/360° while the peak of the theta cycle was assigned to 180°. The mean preferred phase of firing was calculated for each cell by averaging the circular phase angle of each spike divided by the total spike count. Cells were considered to exhibit a significant phase preference when the p value calculated with Rayleigh’s test for nonuniformity of circular data (Fisher, 1993) was determined at p ≤ .01.

2.8.5 |. Unit isolation and classification

Single units were sampled at 30.3 kHz and filtered at 300–6000 Hz (Neuralynx). All tetrode signals were referenced against a tetrode wire near the corpus callosum. The activity of individual units was separated offline into different clusters based on their waveform properties. Waveform properties were defined in three-dimensional feature space using custom spike-sorting software (Plexon, Dallas, TX). The primary quantitative measure of cluster quality provided by OFS (Offline Sorter) was J3, a nonparametric measure of the quality of cluster sorting (Wheeler, 1999). A measure of the average distance between unit clusters (J2) is first calculated: J2 = ΣN(u)E(m(u) − m). The summation is over units u, N(u) is the number of points in unit u, E(x) represents the Euclidean distance squared. m(u) is the cluster center for unit u, and m is the grand center of all points in all units. The final measurement, J3 is then calculated by dividing J2 by a third variable (J1): J1 = ΣΣE(f(u,i) − m(u)). This is a measure of the average distance in three-dimensional feature space between points in a cluster (f) from their center (m). E(x) represents the Euclidean distance squared ($\tilde{\mathbf{x}}$ * x). The summations are over units u, and over feature vectors (points) in each unit i. J3 therefore takes on a maximum value for compact, well-separated clusters.

Finally, spikes that had a refractory period of less than 1 ms and units that were not determined to be pyramidal cells or interneurons (Fox & Ranck, 1981; Robbins, Fox, Holmes, Scott, & Barry, 2013) were removed from analysis. Putative Oriens-Lacunosum Moleculare (OLM) and basket cell activity were distinguished via estimated recording depth, theta amplitude, phase preference (Klausberger & Somogyi, 2008) and phase offset between recording sites to distinguish between str. oriens and str. pyramidale. Pyramidal cells, interneurons, and axonal activity were distinguished from each other via activity and waveform properties as described previously (Barry, 2015; Fox & Ranck, 1981; Robbins et al., 2013), but are briefly described here. Interneurons discharge at high rates (>10 Hz), have short duration waveforms (peak to trough duration ~250 μs), never show complex spike bursts and tend not to exhibit circumscribed firing fields. Pyramidal cells discharge at lower rates (<9 Hz), have wider waveforms (peak to trough duration ~500 μs) and fire in complex-spike bursts. Pyramidal cells may act as “place cells” whose activity is typically confined to a small, cell-specific region called the “firing field” (Barry & Muller, 2011; O’Keefe & Nadel, 1978). Axonal activity typically exhibits a short duration waveform (peak to trough duration <250 μs) with a triphasic waveform shape (Barry, 2015; Robbins et al., 2013).

2.8.6 |. Single-cell temporal autocorrelation and frequency

As in Zutshi et al. (2018), for each cell, spike times were binned at a sampling rate of 500 Hz. The temporal autocorrelation between spike times was then calculated from this resulting vector. The power spectrum of the temporal autocorrelation was computed via the Chronux function mtspectrumumc() (http://chronux.org) using a padding factor equal to six powers of 2 over the sample size (Mitra & Bokil, 2008). The single-cell frequency was then taken as the frequency with maximum power in the 4–14 Hz range.

2.8.7 |. Place cell firing field properties

Several firing field properties were analyzed to determine the effect of task and stimulation condition on the spatial tuning of hippocampal cells as well as to verify whether or not each unit met place cell criteria (Barry & Muller, 2011).

Firing rate

The number of action potentials fired by the cell in the firing field divided by dwell-time in the field. The mean overall firing rate was also determined by dividing the total number of action potentials by the length of the recording session.

Firing field coherence

Coherence was calculated as previously described (Barry & Muller, 2011) and estimates the regularity of the firing rate transition from pixel to pixel in relation to the spatial firing distribution. Cells with field coherence values > 0.2 were considered to be place cells.

Field size

The size of firing fields considered for analysis was set to a lower limit of nine contiguous pixels and a maximum of 700 pixels (approximately 60% of the arena surface).

Firing field stability

Positional firing stability was estimated from the similarity of the place cell’s spatial firing pattern between each stimulation and task condition during spatial accuracy performance. In order to test for the influence of task and stimulation, direct comparisons were made between control and blue light stimulation sessions during the hidden goal task, control and blue light stimulation sessions during the visible task and between control stimulation sessions during the hidden and visible goal tasks. Similarity is defined as the Fisher z transform of the product-moment correlation between the firing rates in corresponding pixels for both intervals.

2.9 |. Statistical analyses

2.9.1 |. General estimating equations

As the behavioral, single unit and EEG data sets contain data from multiple sessions in a repeated measures design in single animals, the assumptions of independence of observations are invalid. The observations for each of these measures within single animals are likely to be correlated and these data can be represented as a cluster. In this case, the existence of a relationship between each measure of interest within an individual animal may then be assumed (Ziegler, Kastner, & Blettner, 1998). We therefore used general estimating equations (GEE; SPSS, Armonk, NY), a class of regression marginal model, for exploring multivariable relationships between clustered signal property data or behavioral measures for individual animals sorted by stimulation condition. Unless otherwise indicated, analyses are by rat. The model was adjusted according to the distribution of each analyzed variable, that is, gamma with log link models were used for non-normally distributed data while poisson loglinear distributions were used for count data. Correlation analyses that produced negative values were transformed into small positive values so that gamma with log link models could be fit to the correlation data for GEE analysis. In both Experiment 1 and Experiment 2 we used a repeated measures design using the same animals between stimulation conditions to compare the effects of yellow control or blue light stimulation in each task condition on electrophysiological and behavioral measures. In Experiment 2, the principal condition used as a reference for comparison across stimulation conditions was the StimHid session. When warranted, the CtrlVis baseline stimulation condition was also used as a reference.

2.9.2 |. Circular statistics and phase preference

Rayleigh’s test for nonuniformity of circular data was calculated as (RBAR = n*r) (Fisher, 1993), where n is the sum of the number of incidences in cases of binned angle data and r is the resultant vector length of the distribution. The angle (Dir) and length of the resultant phase vector (RBAR) for the individual cell and the population of cells for each rat were then calculated and plotted for each stimulation condition in both experiments. As a compliment to RBAR, we also measured population dispersion levels (δ= 1 − T₂ / [2 * RBAR²]) (Fisher, 1993) and measured the variability of the population vector relative to theta cycle. If the population RBAR levels are low, dispersion levels should be high. We used the F-test for homogeneity of variance to test for differences in dispersion levels between stimulation conditions. To test the changes in the circular distribution of preferred firing phase between stimulation conditions, we used the parametric Watson–Williams multisample test for equal means, used as a one-way analysis of variance for circular data (Berens, 2009).

2.10 |. Histological procedures

Rats were placed in a chamber with 5% isoflurane until heavily anesthetized and were perfused intracardially with saline followed by 4% formaldehyde. Brains were extracted and stored in 4% formaldehyde. Frozen coronal sections (30 μm) were cut and stained with cresyl violet. Electrode positions in the hippocampus (Figure 1b), optical probe and electrode positions in the MS (Figure 1c) were assessed in all rats. The volume of the Enhanced Green Fluorescent Protein (EGFP) labeled viral expression area in the MS was calculated using previously described methods (Zutshi et al., 2018). Coronal brain slices (30 μm thick) containing the MS area for each rat was stained with 4′,6-diamidino-2-phenylindole (DAPI) and tiled images were taken on a Nikon C2 Confocal Microscopy System (Nikon Corporation, Tokyo, Japan) using a 10× (0.45 NA) objective. All images were obtained using identical parameters and dimensions for consistency in the analysis. Images of the MS were obtained along the AP axis close to the beginning, center, and terminus of the EGFP expression area to calculate the volume of expression. Using ImageJ (FIJI), tile images were stitched together (Preibisch, Saalfeld, & Tomancak, 2009) and a region of interest (ROI) was drawn around the MS and the diagonal band of Broca (ROI area = 1.47 mm²); the same ROI was used across animals. For each stitched image (three per animal), EGFP pixel area was analyzed within the ROI. The final viral expression volume was obtained by summing the EGFP area of each stitched imaged and multiplying by the distance between imaged slices.

Fixed forebrain sections were dried for 30 min then rehydrated in 150 mM Tris, pH 9.0. Antigen unmasking was performed by incubating sections at 80°C in 150 mM Tris, pH 9.0 for 2 min, twice. Sections were washed in phosphate-buffered saline (PBS) for 3 min and then permeabilized in PBS containing 0.4% Triton X-100 for 30 min at room temperature. Following permeabilization, the brain sections were blocked in 10% normal donkey serum for 1 hr at room temperature. Primary antibodies (1:500 dilutions) were then incubated on the brain sections for 16 hr at 4°C. The sections were then washed in PBS three times for 5 min each time. The tissue sections were incubated with anti-choline acetyltransferase (polyclonal rabbit anti-ChAT antibody, Synaptic Systems), anti-glutamate decarboxylase 2 (polyclonal guinea pig anti-GAD 2, Synaptic Systems), and anti-vesicular glutamate transporter 2 (polyclonal chicken anti-VGLUT2, Synaptic Systems) at 1:500 overnight at 4°C. Sections were then washed three times in PBS, for 5 min each time. The tissue sections were then incubated in secondary antibody (Donkey anti-rabbit Cy3, or Donkey anti-guinea pig Alexa 647 both at 1:500 dilutions) for 3 hr at room temperature in the dark. Following three 5-min washes in PBS, the sections were then coverslipped with fluromount.

All imaging was performed in SPOT 5.0 advanced imaging software (SPOT Imaging Solutions, Sterling Heights, MI) using a Nikon DS-Fi1 color digital camera on a Nikon E400 microscope with Fluorescein Isothiocyanate (FITC), Tetramethylrhodamine (TRITC), and DAPI filter set. Two 20× images were taken from septal sections of each animal stained for GABAergic, cholinergic and glutamatergic neurons. An additional 40× image was taken of a hippocampal section stained for glutamatergic axon terminals projecting from the MS that expressed VGLUT. Cells were considered colocalized when GAD and ChAT overlapped with GFP. Only cells with distinct nuclei as determined by the presence of DAPI were considered to be representative examples. For VGLUT expression, MS axon projection terminals that expressed GFP and VGLUT in hippocampus were considered to be co-localized.

3 |. RESULTS

We examined the role of behavior and task demands in determining basal septohippocampal circuit state and the efficacy of optogenetic entrainment in two experiments. Experiment 1 (N = 3) consisted of 6 Hz control yellow and blue MS stimulation while animals were at rest in a flowerpot, predicting less competition between endogenous hippocampal and artificial theta without movement or cognitive demand. In Experiment 2 (N = 3), animals performed a spatial accuracy task where the goal zone was either visible or hidden. Each task was repeated to allow for comparison between continuous open-loop 6 Hz control and blue light MS stimulation during each task condition (control stimulation with visible goal = CtrlVis, control stimulation with hidden goal = CtrlHid, stimulation with visible goal = StimVis, stimulation with hidden goal = StimHid) and whether task demands would affect circuit entrainment efficacy and whether subsequent perturbation would affect spatial accuracy behavior. Finally, to compare hippocampal temporal phenomena during movement, but in the absence of cognitive demand, we also compared spatial accuracy data with electrophysiological recordings from an additional three animals while foraging for food pellets.

3.1 |. Confirmation of MS probe and electrode placement and viral transduction volume

Examples of immunohistochemistry of ChR2 transduction in the MS, including coexpression of ChR2 in cholinergic and GABAergic cells, can be seen in Figure 1e. Transduction was estimated to occur in 75% of GABAergic cells and 45% of cholinergic cells. As our chosen VLGLUT antibody targets vesicular transporters expressed in axon terminals rather than the soma, we found sparse expression in the somata of MS cells (Barrows, McCabe, Chen, Swann, & Weston, 2017). We therefore show colocalization of VGLUT and GFP at the axon terminals of septal projections in str. pyramidale and str. oriens in CA1 (right, Figure 1e). An example of GFP and GAD expression in conjunction with septal probe and electrode array placement in the MS can be found in Figure 1f (left). These examples serve to demonstrate that the AAV-transduced ChR2 pan-neuronally in cell types of the MS. The axonal projections from transduced septal cells to GAD-positive interneurons in str. oriens and str. pyramidale layers of CA1 in hippocampus are also shown in Figure 1f (right). To interpret the effect of optogenetic stimulation on electrophysiological and behavioral results we estimated the volume of viral transduction in the MS. Estimated GFP expression volumes were consistent with the estimated total volume of the adult MS (mean = 1.69 ± 0.3 mm³; Table 1).

TABLE 1.

Analysis of GFP expression volumes for each slice in each rat

Animal ID	Area (mm2) anterior slice	Area (mm2) middle slice	Area (mm2) posterior slice	Sum (mm2)	AP distance (mm)	Volume (mm3)
P2	0.175659	0.835616	0.665063	1.676338	1.2	2.011605
P3	0.671386	0.376186	0.286113	1.333685	1.2	1.600422
P4	0.336196	0.807112	0.071398	1.214706	1.2	1.457647
Mean = 1.69 mm3
SD = 0.30 mm3

4 |. EXPERIMENT 1

4.1 |. MS optical stimulation paces hippocampal theta band oscillations

As illustrated for ~20-s epochs in Figure 2a, 6-Hz control stimulation has no effect on hippocampal or MS signals while 6-Hz B stimulation entrains oscillations in the MS and hippocampus at 6 Hz. Corresponding spectrograms for the entire 6 min of stimulation are shown in Figure S1a. Artificial timing of interneuron types at opposite phases matches previous descriptions (Freund & Buzsaki, 1996). The examples in Figure 2a show that during baseline control stimulation, when the animal is at rest, stationary or immobility hippocampal theta oscillations are slower and smaller in amplitude than theta during active exploration (Kelemen et al., 2005; Kramis et al., 1975; Sainsbury et al., 1987; Vanderwolf, 1969). At rest, there is also little endogenous MS theta. Artificial MS theta generated by 6-Hz blue light stimulation entrains both MS and hippocampal LFPs.

FIGURE 2.

Hippocampal theta oscillations, phase preference and burst-timing rhythmicity of hippocampal pyramidal cells in Experiment 1. (a) Twenty-second epochs of raw EEG and corresponding spectrograms from CA1 (top) and medial septum (MS) (bottom) during 6 min of 6 Hz control (left) and blue (right) MS stimulation while a rat is in a narrow flowerpot. (b) Mean and SE of theta amplitude during 6 Hz baseline and MS stimulation. Theta peak shifts to 6 Hz during B stimulation but peak amplitude is unchanged. Inset shows a significant increase in MS-CA1 signal coherence between control and blue stimulation at 6 Hz. (c) Phase preference circular statistics for hippocampal CA1 pyramidal cells from 3 animals (n = 64) during control (left) and blue (right) MS stimulation relative to local hippocampal theta. Each circle represents the mean angle and magnitude of phase preference for an individual pyramidal cell. The arrow indicates the angle and significance of the resultant population phase vector. RBAR levels for the circular plot are shown in red and mean vector RBAR is presented at the tip of each arrow. Red asterisk denotes vector significance. During control stimulation (left), hippocampal cells have a robust mean resultant vector (arrow) at 66° relative to endogenous theta oscillations. During B MS stimulation, the phase vector angle shifts toward the artificial theta trough at 0°. (d) B stimulation changes the angle of the hippocampal phase vector but does not significantly change RBAR. (e) Referenced to septal theta, during control stimulation there is a weak phase vector at theta trough (left). During B stimulation (right), the phase vector becomes more robust and the phase angle shifts to 28°. (f) Relative to artificial MS theta, B MS stimulation significantly increases phase vector RBAR. (g) Histograms illustrating the distribution of spike timing frequencies for hippocampal pyramidal cells as determined by autocorrelation analysis during control (top) and B (bottom) MS stimulation. Distribution curves, mean, and SE of the cell population burst frequency in each condition are shown. During MS stimulation, pyramidal cell burst timing shifts to within 1 Hz of the MS input frequency at 6 Hz, significantly lowering the mean frequency in comparison to baseline. (h) Example autocorrelations from two cells illustrating weak modulation during control stimulation (top) and robust modulation near 6 Hz during B stimulation (bottom). Each pair of plots illustrates the modulation of the number of action potentials over time (left) and corresponding power calculated from the spike-time autocorrelation (right)

Analysis of the mean hippocampal theta signal across animals shows a universal shift in peak immobility theta frequency from 6.4 ± 0.001 Hz during baseline control stimulation to 6 ± 0.094 Hz during blue stimulation (Figure 2b). The peak amplitude at these frequencies in each stimulation condition was not statistically different between control (92.91 ± 0.73 A.U.) and blue (94.48 ± 1.34 A.U.) MS stimulation (Wald test = 0.498; p = .116). The MS exhibited little theta when the animals were at rest. During control stimulation the mean peak theta frequency was close to 1/f at 5.3 ± 0.21 Hz but shifted to 6.0 ± 0.0001 Hz during blue light stimulation. Unlike the hippocampus, the MS peak theta amplitude increases significantly (Wald test = 94.94; p < .0001) between control (85.51 ± 2.33 A.U.) and blue stimulation (107.84 ± 1.15 A.U.). In correspondence with spectral results in each region, coherence at 6 Hz increases significantly (Wald test = 11.79; p = .001) between control (0.088 ± 0.0390 S.U.) and blue MS stimulation (0.453 ± 0.162 S.U.). This indicates that stimulation initiates effective connectivity between the MS and CA1 at 6 Hz (Inset Figure 2b). Low MS-CA1 coherence values at baseline also imply that immobility theta is largely MS independent. That the mean coherence levels post stimulation are lower than 1 suggests that the artificial hippocampal theta is not a result of volume conduction.

Although the MS generates little endogenous theta while the rat is at rest, optogenetic septal ATS generates hippocampal theta oscillations at matching frequency that are similar in amplitude to endogenous theta. We now analyze how ATS affects the phase preference and burst rhythmicity of CA1 pyramidal cell action potentials.

4.2 |. MS optical stimulation aligns septal and local theta phase preference of hippocampal neurons

Stimulation effects on CA1 pyramidal cell phase preference (n = 64) relative to MS and local theta, pooled for all three animals (n = 33, 14 and 17), are shown in Figure 2c. Significantly different mean resultant vector lengths (RBAR) and phase angles were found between control and blue light stimulation (control RBAR = 0.679 at 66°, dispersion = 0.81; blue RBAR = 0.761 at 14° dispersion = 0.52; Watson Williams test, F = 37.70, p < .0001) where the phase angle moves significantly closer to theta trough. A paired t test comparing RBAR values (Figure 2d) indicates blue stimulation had no effect on the length of the resultant hippocampal phase vector (t = 1.59, p = .115). Relative to septal theta, significantly different vector lengths and phase angles (Figure 2e) were also found between control and blue stimulation (control RBAR = 0.355 at 357°, dispersion = 4.04; blue RBAR = 0.539 at 28°, dispersion = 1.27; Watson Williams test, F = 5.01, p = .027). Unlike hippocampal theta, paired t-tests comparing RBAR values in each condition (Figure 2f) show that MS stimulation significantly increased vector lengths of hippocampal pyramidal cells relative to septal theta (t = 5.47, p < .0001).

Although the septal component of hippocampal theta phase organization was weak during baseline control stimulation, it was significantly stronger during ATS. During blue light stimulation, the pyramidal cell phase angles relative to hippocampal and septal theta are aligned toward theta trough. While the rats were at rest, artificial septal theta was not in conflict with other theta generators and dictated preferred theta phase angle of hippocampal cells. A microcircuit-level experimental model demonstrating how MS stimulation paces hippocampal oscillations and phase preference via putative basket cell and OLM interneurons is provided in Figure S1b,c.

4.3 |. CA1 pyramidal cell burst rhythmicity matches septal ATS

To examine whether pan-neuronal septal stimulation also paced the rhythmic spiking frequency of hippocampal pyramidal cells we analyzed autocorrelations for each neuron (n = 64) and used a FFT to calculate the dominant frequency (Zutshi et al., 2018). As shown in Figure 2g, the distribution of baseline frequencies was spread out across the theta band (mean = 8.89 ± 0.322 Hz) but clustered at 6 Hz during blue light stimulation (mean = 7.30 ± 0.262 Hz). The mean spike-timing rhythmicity was significantly lower during blue light stimulation (Wald test = 14.45, p < .0001).

As autocorrelation analyses can be significantly altered by firing rate, we note there was no significant change in mean firing rate between baseline (mean = 2.95 ± 0.334 Hz) and MS stimulation (mean = 3.09 ± 0.330 Hz; Wald test = 0.10, p = .75). Two examples of cellular autocorrelations and their corresponding power analyses are shown in Figure 2h and illustrate robust spike-timing modulation during MS stimulation. Figure S1d shows two examples of CA1 str. pyramidale interneurons whose spike-timing rhythmicity are significantly modulated by MS stimulation.

5 |. EXPERIMENT 2

Experiment 1 showed that optogenetic MS stimulation generates an artificial theta signal that entrains hippocampal LFP frequency and neuronal burst-timing patterns and shifts the preferred phase preference of these cells, relative to local and artificial septal theta, into alignment. In the absence of large amplitude endogenous hippocampal theta signals, as typically seen during movement and active exploration, there was no conflict between artificial and endogenous theta oscillations. In Experiment 2 we address the hypothesis that ATS and endogenous theta oscillations may compete during active navigation and that this competition may be influenced by spatial task demands. In addition, we also ask if subsequent perturbation of the septohippocampal circuit by ATS will interfere with spatial memory or navigation in a goal finding task. We therefore tested the effect of stimulation on septo-hippocampal entrainment of cellular and LFP phenomena at different movement speeds as well as effects on spatial behavior while rats searched for a visible or hidden goal zone during 6 Hz control and blue light stimulation (stimulation conditions = CtrlVis, CtrlHid, StimVis, and StimHid).

5.1 |. MS entrainment of hippocampal LFP theta band oscillations is influenced by task and speed

Animal speed has an important effect on theta band properties (Richard et al., 2013; Zutshi et al., 2018) and prior studies have described optogenetic lateral septum stimulation effects on movement speed (Bender et al., 2015). In our experiment, GEE analysis revealed no main effect of stimulation condition on mean animal speed (CtrlVis = 11.73 ± 0.645 cm/s; CtrlHid = 10.75 ± 0.586 cm/s; StimVis = 11.78 ± 0.639 cm/s; StimHid = 9.97 ± 0.539 cm/s; Wald test = 7.11, p = 0.070). The mean speed for each subject in each stimulation condition as well as the overall mean speed for each rat across multiple sessions is shown in Figure S2a. The mean speed across rats in each stimulation condition is shown in Figure 2b. Post-hoc comparisons show no difference in speed between CtrlVis and StimVis sessions (Wald test = 0.019, p = .891) or between CtrlHid and StimHid (Wald test = 1.17, p = .280). There are trends for decreased mean speed in the hidden goal tasks in comparison to the visible goal tasks, as well as decreased mean speed in StimHid compared to CtrlHid. We contend that these trends reflect increased difficulty in the hidden goal task, and that MS stimulation did not significantly reduce speed in comparison to baseline conditions. If MS stimulation reduced speed, it would occur in both task versions. It is possible that increased subject power would make the speed analysis more sensitive to effects of stimulation condition, yet our results are consistent with MS optogenetic stimulation not having an effect of animal speed using 6-Hz MS stimulation during open field foraging (Blumberg et al., 2016) or during 10-Hz MS PV-selective stimulation in the home cage (Zutshi et al., 2018).

Figure 3a shows an example of the relationship between speed and theta band properties in each stimulation condition for one EEG channel from one rat. Control stimulation does not perturb the linear relationship between speed and theta amplitude or frequency. In the visible goal condition, stimulation at movement speeds ≥ 5 cm/s resulted in a separation of the theta band into two spectral peaks at 6 and ~9 Hz. We propose that this faster frequency is a competitive endogenous response to 6-Hz MS stimulation. Like the stimulation result in Experiment 1, when animals were at rest, the septal ATS frequency at 6 Hz was more prevalent in hippocampal theta oscillations during slower movement speeds at ≤ 2 cm/s. At rest or moving slowly, there was no secondary peak at ~9 Hz. Plotting frequency of theta cycles across all speeds during StimVis (top right Figure 3a) suggests that the reactive component of the endogenous signal still increases linearly with movement speed but is shifted approximately 1.5 Hz faster at each speed increment in comparison to CtrlVis. Supplementary material includes an additional example of this competitive dynamic during StimVis, illustrating real-time epochs of slow and fast movement (Figure S3a) and speed-sorted extremes (Figure S3b). When the ~9 Hz signal is present, it is larger in amplitude during faster movements than slower movements. During StimHid, the 6-Hz artificial theta signal dominated the bandwidth regardless of movement speed (bottom right Figure 3a). Speed-sorted spectrograms for the entirety of stimulation conditions in Figure 3a are shown in Figure 3b.

FIGURE 3.

Hippocampus speed/theta property dynamics relative to control and B stimulation during visible and hidden spatial accuracy conditions. (a) In each condition mean theta amplitude by frequency during slow (≤ 2 cm/s = blue lines) and fast (≥ 5 cm/s = black lines) velocities (v) during control and blue light medial septum (MS) stimulation is shown. Beneath each mean amplitude plot, speed and frequency or amplitude of each theta cycle (blue dots) for each condition is illustrated while the line of best fit (black line) and the corresponding correlation coefficients and p values are shown above each plot. The linear relationships between theta frequency and amplitude during control stimulation are perturbed during StimVis and StimHid, where stimulation has a complex relationship with speed. When the rat is moving less than 2 cm/s, theta peaks are more likely to occur at the MS input frequency at 6 Hz. At faster speeds the signal is split between 6 Hz and an endogenous response to stimulation at ~9 Hz. The 6 Hz component of the signal is larger in amplitude during StimHid than StimVis. These dynamics alter the relationship between speed and theta frequency and amplitude dynamics. (b) Speed-sorted spectrogram data in A during visible and hidden goal spatial accuracy tasks. Theta amplitude and frequency increase linearly with speed during control stimulation. Black arrowheads indicate 200-s epochs of the speed-sorted spectrograms expanded beneath each plot. During StimVis, endogenous signals at approximately 9 Hz compete with the septal input and during StimHid theta frequency predominately matches septal input at 6 Hz. (c) Peak theta amplitude at slow speeds (i) during StimVis and StimHid is the same and therefore independent of task demands. Peak theta amplitude during both B stimulation sessions is larger than baseline conditions. At fast speeds (ii), peak theta amplitude during StimVis and StimHid is split between two peaks. While the 6 Hz peaks during StimVis and StimHid are not statistically different from each other, the StimHid 6 Hz peak is larger than its secondary ~9 Hz peak but equal to baseline peaks. Artificial hippocampal theta during locomotion in StimHid is therefore similar in amplitude to endogenous theta during locomotion in either CtrlVis or CtrlHid. In contrast, the StimVis 6 Hz peak is equal to its secondary ~9 Hz peak and smaller than baseline peaks. This suggests a greater distribution of theta signal amplitude between 6 and ~ 9 Hz during StimVis than StimHid. (d) Mean and SE of the correlations between movement speed and the frequency (i) or amplitude (ii) of each theta cycle. Task demands significantly improve the fidelity between baseline theta frequency and speed. The relationship between theta properties and speed is significantly disrupted during StimHid

At speeds ≤ 2 cm/s, GEE statistical analysis of peak spectral amplitude (Figure 3c,i) found a significant interaction between stimulation condition and frequency (Wald test = 1,198.18, p < .0001). Peak theta amplitude was the same in StimVis and StimHid conditions (Wald test = 0.144, p = .704), but was larger in StimHid than both CtrlVis (Wald test = 13.25, p < .0001) and CtrlHid (Wald test = 10.36, p = .001) baseline sessions (Table 2a). Unlike Experiment 1, at slow speeds and at rest, stimulation makes peak theta amplitude larger than baseline. This is likely due to increased arousal in the context of spatial accuracy performance as opposed to resting quietly in the flowerpot (Green & Arduini, 1954; Kramis et al., 1975; Sainsbury et al., 1987).

TABLE 2.

General estimating equation (GEE) results for hippocampal local field potential (LFP) comparisons across task conditions that compare slow speed mean signal peaks at 6 Hz during StimHid to peaks during StimVis, CtrlVis and CtrlHid (a), or compare fast speed mean signal peaks at 6 Hz during StimHid to each peak during StimVis, CtrlVis and CtrlHid (b); analyses in B are repeated using StimVis as a reference (c).

Speed	Stimulation condition * frequency interaction = p < .001
(a) Speed ≤ 2 cm/s
Reference = BUC peak 6 Hz
StimVis peak = 6 Hz	p = .704 ns
CtrlHid peak = 6.8 Hz	p = .001*
CtrlVis peak = 6.8 Hz	p < .001*
(b) Speed ≥ 5 cm/s
Reference = StimHid peak 6 Hz
StimVis Peak = 6 Hz	p = .141 ns
StimHid second peak = 8.8 Hz	p < .001*
StimVis second peak = 8.8 Hz	p = .004*
CtrlHid peak = 7.6 Hz	p = .367 ns
CtrlVis peak = 7.6 Hz	p = .334 ns
(c) Speed ≥ 5 cm/s
Reference = StimVis Peak 6 Hz
StimVis second peak = 8.8 Hz	p = .139 ns
CtrlHid peak = 7.6 Hz	p = .014*
CtrlVis peak = 7.6 Hz	p = .011*

Note: Significance of p < .05 for comparisons against the reference session is indicated by an asterisk (*).

As suggested by examples in Figure 3a, at speeds ≥ 5 cm/s peak spectral amplitude during stimulation was often split between 6 and ~9 Hz (Figure 3c,ii). Statistical analysis found a significant interaction between stimulation condition and frequency (Wald test = 570.14, p < .0001). Peak 6 Hz amplitudes were similar between StimVis and StimHid (Wald test = 2.17, p = .141). However, the StimHid peak at 6 Hz was larger than both its secondary peak at ~9 Hz (Wald test = 22.30, p < .0001) and the StimVis secondary peak at ~9 Hz (Wald test = 8.34, p = .004). Whereas the StimHid 6 Hz peak was equal to baseline peaks in CtrlVis (Wald test = 0.935, p = .334) and CtrlHid (Wald test = 0.813, p = .367) (Table 2b), the StimVis 6 Hz peak was significantly smaller than baseline peaks in CtrlVis (Wald test = 6.44, p = .011) and CtrlHid (Wald test = 6.10, p = .014) (Table 2c). This is likely because the signal amplitude in StimVis was evenly distributed between 6 and ~ 9 Hz (Wald test = 2.17, p = .139). Entrainment of hippocampal LFPs during spatial task performance is therefore primarily a function of speed and secondarily a function of task demands. The hippocampus is more receptive to 6 Hz artificial septal output at slow speeds and during increased spatial processing in the hidden goal task.

We also analyzed the linear relationship between speed and theta frequency or theta amplitude (Figure 3d). The speed/theta frequency relationship (Figure 3d) significantly increased from CtrlVis (mean = 0.230 ± 0.0126) to CtrlHid (mean = 0.292 ± 0.0148; Wald test = 23.69, p < .0001), reflecting background network level changes associated with task demands (Richard et al., 2013). Speed/theta frequency was maintained during StimVis (mean = 0.269 ± 0.0175) and was on par with CtrlHid, possibly due to the competition between endogenous and artificial theta frequencies. In contrast, the speed/frequency correlation was significantly lower in StimHid than CtrlHid (Wald test = 80.59, p < .0001) or StimVis (Wald test = 11.47, p = .001). This may have been due to the increased influence of septal ATS over hippocampal frequency, even at increased speed. In contrast to a previous report using selective optogenetic stimulation of PV interneurons in transgenic mice (Zutshi et al., 2018), pan-neuronal stimulation did not spare the speed/theta amplitude relationship. In comparison to CtrHid (mean = 0.173 ± 0.0320), the linear correlation between speed and theta amplitude was significantly lower in StimVis (mean = 0.040 ± 0.029; Wald test = 5.51, p = .019) and abolished during StimHid (mean = 0.00031 ± 0.0002; Wald test = 42.51, p < .0001). As the rats slowed down, theta amplitude increased. This inversion of the speed/theta amplitude relationship may have been caused by increased hippocampal acetylcholine and the optogenetic activation of septal cholinergic neurons.

Finally, hippocampal LFPs were recorded from an additional three control animals that foraged for pellets in the open field (Figure S4a, b). Speed/theta frequency correlations were on par with the CtrlVis condition (Table S1) in the first (mean = 0.225 ± 0.045) and second (mean = 0.215 ± 0.0428) forage sessions. There were no significant differences in speed/theta frequency (Wald test = 0.027, p = .87) or other theta property variables between two different exposures to the open field, implying that significant increases between CtrlVis and CtrlHid were not an effect of re-introduction to the arena, but a consequence of task demands.

5.2 |. MS stimulation overrides endogenous theta phase preference of hippocampal neurons

After demonstrating that ATS entrainment of hippocampal LFPs was affected by task demands, we then asked how the temporal organization of CA1 pyramidal cells, as measured by phase preference relative to hippocampal and MS signals, might have been affected in each stimulation condition. We calculated the resultant vector lengths (RBAR) from the mean angle and phase preference magnitude of CA1 pyramidal cell action potentials relative to endogenous or artificial hippocampal (Figure 4a top and bottom), and endogenous or artificial septal theta (Figure 4b, top and bottom) for cells pooled across all three animals (n = 22, 57, and 87, total = 166). Circular statistics for each condition are shown on the top of each plot in Figure 4 (p value, RBAR, dispersion, and mean phase angle). Table 3 shows the results of circular statistics (Watson–Williams multisample test comparisons) across conditions. As a further exploration of the influence of spatial demand on phase preference, we recorded cells from three additional animals (n = 62, 29, and 67; total = 158) that simply foraged for food pellets in the absence of operant conditions (Figure S4c).

FIGURE 4.

(a,b) Circular statistics for phase preference of CA1 pyramidal cells referenced to hippocampal (a) and MS (b) theta per task condition. Each circle represents the angle and magnitude of the resultant theta phase vector (RBAR) for an individual pyramidal cell. The arrow indicates the angle and magnitude of the resultant population vector. Discontinuous arrowheads indicate resultant vectors > 0.2. RBAR levels for the circular plot are shown in red. The p values, population RBAR values, levels of population dispersion and mean vector angle (MeanDir) are shown above each plot. Significant (*) and non-significant (ns) vectors are noted. Between CtrlVis and CtrlHid, the hippocampal theta phase vector angle is stable, and the dispersion level significantly decreases. Phase preference is significantly perturbed during StimVis and abolished during StimHid. Relative to the septal phase vector, between CtrlVis and CtrlHid phase angle shifts closer to the trough at 0° and dispersion level significantly decreases. In contrast to the hippocampal reference, septal phase preference vector during B stimulation remains significant but phase angle shifts toward theta peak. (c) Histograms illustrating the distribution of spike timing frequencies for hippocampal pyramidal cells as determined by autocorrelation analysis during each task condition. Distribution curves as well as the mean and standard error of spike-timing modulation frequency for cell populations in each condition are shown. During baseline conditions, the majority of cells exhibit spike-timing intervals at ~8.5 Hz. During B stimulation, spike-timing shifts to ~7 Hz. A subpopulation more responsive to ~9 Hz during StimVis is absent during StimHid. (d) Example autocorrelations from two cells illustrating robust modulation during control stimulation at 7.5–8 Hz. Cell 1 is part of the subpopulation that is more modulated at ~9 Hz during StimVis. Both cells are modulated at ~6 Hz during StimHid

TABLE 3.

The results of Watson–Williams multisample test comparisons across conditions for pooled pyramidal cell phase preference

Hippocampus			Medial septum (MS)
	F	p		F	p
CtrlVis-CtrlHid	0.929	.336 ns	CtrlVis-CtrlHid	9.82	.00193
CtrlVis-StimVis	6.31	.0125	CtrlVis-StimVis	82.76	<.0001
CtrlHid-StimHid	16.68	<.0001	CtrlHid-StimHid	133.84	<.0001
StimVis-StimHid	9.28	.0026	StimVis-StimHid	0.959	.329 ns

Relative to hippocampal theta (top, Figure 4a), phase vector angles were stable between CtrlVis and CtrlHid toward the peak of theta (180°). Watson–Williams circular test found no significant difference between the resultant phase angles (Table 3, p = .336). However, an F-test showed that dispersion levels in CtrlHid were down to about a third of CtrlVis (F = 8.89; p < .00001). Relative to endogenous septal theta (top, Figure 4b), the same cells exhibited phase vector angles significantly shifted (Table 3, p = .00193) toward theta trough (0°). An F-test revealed that dispersion levels were also significantly decreased (F = 5.88; p < .00001). Finally, we compared CtrlVis and CtrlHid hippocampal theta phase vectors with a foraging session (Figure S4c). While dispersion levels were equal between CtrlVis and Foraging (F = 1.04; p = .78), they were significantly lower in CtrlHid than Foraging (F = 8.48; p < .0001). Increased spatial demands therefore alter the state of the septo-hippocampal circuit and lead to increased temporal coordination of cell activity within CA1 relative to both hippocampal theta (Barry et al., 2016) and endogenous septal theta. Cognitive demands therefore affect temporal organization of CA1 cell populations (Fenton et al., 2010), which may be more receptive to both local theta and septal output. We now ask if cognitive demand also effects phase preference relative to artificially generated theta.

Hippocampal theta phase preference was non-significant during both StimVis and StimHid (bottom, Figure 4a), significantly differing from their respective baseline sessions in CtrlVis (Table 3, p = .0125) and CtrlHid (Table 3, p < .0001). During StimHid, the resultant phase vector length was even weaker and had higher dispersion levels than StimVis (F = .05; p < .0001). The negation of local hippocampal theta phase preference during stimulation is likely a result of competition between septal and extra-hippocampal theta generators, even when the artificial septal signal is more dominant in the hippocampal field potential. This suggests a disconnect between the phasic timing of hippocampal pyramidal cells and local theta oscillations. Stimulation should therefore perturb phase precession relative to local theta (O’Keefe & Recce, 1993; Zutshi et al., 2018).

In order to determine which of the competing synaptic inputs had more influence over the phasic organization of pyramidal cells, we also carried out phasic spike-timing analyses relative to artificial MS theta. Circular statistics revealed that hippocampal pyramidal cells were in fact phasically organized by artificial septal theta. Significant and stable phase vectors were maintained relative to ATS between StimVis and StimHid (bottom, Figure 4b). There was no significant difference in phase angle (Table 3, p = .329) or dispersion levels between both conditions (F = 1.05; p = .80). However, in comparison to baseline conditions the phase angle during stimulation was significantly shifted toward theta peak (Table 3: CtrlVis-StimVis p value <.001, CtrlHid-StimHid p value <.0001).

In correspondence with phase preference, we also measured MS-CA1 coherence in each condition. Despite the effect of spatial cognitive demand on phase preference, peak coherence levels at 7.5 Hz were similar between CtrlVis and CtrlHid (Figure S5). This suggests that the population changes relative to vector dispersion during increased cognitive demand are due to local neuromodulatory effects modifying receptivity to theta inputs. During stimulation, coherence levels during StimVis (0.617 ± 0.183 S.U.) and StimHid (0.622 ± 0.166 S.U.) both peaked at 6 Hz and were significantly greater than baseline measures at 6 Hz during CtrlVis (0.0721 ± 0.0263 S.U.; Wald test = 14.00, p < .0001) and CtrlHid (0.0465 ± 0.0107 S.U.; Wald test = 39.20, p < .0001). ATS therefore shifts the effective theta-scale connectivity between CA1 and the MS from 7.5 to 6.0 Hz. As in Experiment 1, that coherence levels during stimulation are significantly less than 1 implies that CA1-MS connectivity is not a result of volume conduction.

Given these results, we also asked if the timing of CA1 pyramidal cell somatic outputs were ultimately more influenced by the septal ATS frequency or the endogenous hippocampal response to stimulation. To this end we measured the frequency of burst rhythmicity for all recorded cells in each stimulation condition.

5.3 |. Spatial cognitive demands influence MS entrainment of pyramidal cell spiking rhythmicity

As stated previously, changes in cell firing rate caused by MS stimulation might alter spiking rhythmicity. Our analyses found that there was no main effect of stimulation condition on mean firing rate (CtrlVis = 1.52 ± 0.117 Hz, CtrlHid = 1.39 ± 0.115 Hz, StimVis = 1.45 ± 0.129 Hz, StimHid = 1.44 ± 0.137 Hz; p = .874). Mean firing rate was constant throughout the recording conditions.

As in Experiment 1, we analyzed autocorrelations for each neuron and used a FFT to calculate the dominant frequency. Figure 4c shows the distributions, means and SE of frequencies in each stimulation condition. The distributions were similar between CtrlVis (mean = 8.09 ± 0.078 Hz) and CtrlHid (mean = 8.06 ± 0.095 Hz). Task demands therefore did not overtly influence mean rhythmic spike-timing frequency. In both StimVis (mean = 7.69 ± 0.168 Hz) and StimHid (mean = 7.16 ± 0.166 Hz) the timing of a clear majority of cells was entrained by ATS at 6 Hz. During StimVis, a subpopulation of cells was more responsive to ~9 Hz but shifted rhythmicity toward 6 Hz during StimHid. The mean frequency was the closest to 6 Hz in StimHid, lower than either CtrlVis (Wald test = 23.21, p < .0001), CtrlHid (Wald test = 20.36, p < .0001) or StimVis (Wald test = 4.81, p = .028). As in previous work, the burst-timing frequency in StimHid of most cells tended to be within 1 Hz of the MS stimulation frequency (Zutshi et al., 2018). Examples of autocorrelations and corresponding power analyses for each condition are shown for two cells in Figure 4d.

5.4 |. Stimulation and temporal reorganization does not alter place cell firing fields

As we found significant stimulation induced alterations in the temporal organization of hippocampal pyramidal cells, we also analyzed the spatial firing properties of place cell firing fields. A total of 103 cells met criteria for place cell analyses from three rats (n = 18, 44, and 41). Examples of simultaneously recorded place cell firing fields across stimulation conditions are shown in Figure 5a. As summarized in Figure 5b, there was no significant main effect of stimulation condition for mean field coherence, mean overall firing rate, mean firing field stability, mean field size, mean in-field firing rate, and mean out-field firing rate. We therefore conclude that as in other studies, perturbation of the septohippocampal circuit does not interfere with the spatial firing of hippocampal place cells (Aoki et al., 2019; Bolding et al., 2019; Brandon et al., 2014; Koenig et al., 2011; Robbe & Buzsaki, 2009). Although the temporal organization of cell activity was significantly altered by MS stimulation, the spatial code remained intact.

FIGURE 5.

Despite temporal reorganization during B stimulation, spatial coding remains constant. (a) Examples of simultaneously recorded place cell firing fields during each task condition. (b) Results of field property analysis between conditions. Task conditions did not elicit significant changes to place cell firing field properties (mean field coherence, mean overall rate, mean firing field stability, mean firing field size, mean in-field firing rate, or mean out-field firing rate)

The maintenance of stable place fields is not sufficient for spatial navigation if septohippocampal theta circuitry has been disrupted (Bolding et al., 2019). We therefore ask if open-loop low-frequency pacing of the septohippocampal circuit in the StimHid condition is sufficient for the spatial accuracy task, or if the downward frequency shift will cause memory or navigation errors.

5.5 |. Artificial theta stimulation spares spatial memory in the hidden goal accuracy task

For each subject, the series of stimulation conditions (CtrlVis, CtrlHid, StimVis, ad StimHid) was repeated 3–4 times with each series occurring on a different day. We found no order effect (StimVis then StimHid/StimHid then StimVis) for either of the behavioral measures (p > .05). An example of behavioral performance for one animal via a speed map and a dwell-time map in each stimulation condition is shown in Figure 6a. Additional examples from the remaining animals are shown in Figure S6a,b. We analyzed several measures of spatial accuracy performance: mean dwell-time ratio in each arena quadrant, mean dwell-time ratio in the goal zone, mean search distance from the goal zone and mean number of entrances (Figure 6b–e). Individual animal data for each of these measures is shown in Figure S6c–f. Despite reorganization of the septohippocampal circuit by an open-loop artificial timing signal, we found that MS stimulation did not improve or worsen spatial memory across multiple sessions in each of the animals. However, we found that search distance from the goal zone was marginally worse in StimHid than CtrlHid. These results are discussed below.

FIGURE 6.

Perturbation of the septohippocampal circuit via artificial theta generation in the MS does not affect spatial memory and but affects search distance from the goal zone. (a) Examples of visible (top) and hidden (bottom) spatial accuracy behavior during control (left) and B (right) stimulation conditions as analyzed by speed and dwell-time maps. The dashed circles in the speed-maps represent speed as a function of distance from the goal zone center where the rat slows its approach closer to the zone in the visible rather than the hidden conditions. The rat also spends more time in the goal quadrant and goal zone during the visible condition. (b) Regardless of task condition, rats spend significantly more time in the target quadrant than adjacent (clockwise [CW], counterclockwise [CCW]) or opposite (O) quadrants. (c) Rats spend more time in the goal zone during the visible than the hidden version of the task, indicating that the hidden goal task is more difficult. Compared to control stimulation, B stimulation did not affect the amount of time spent in the goal zone in either condition. (d) Rats slowed their inbound approach further from the goal zone center during StimHid than CtrlHid. (e) Task conditions did not significantly affect the number of goal zone entrances

5.5.1 |. Quadrant dwell-time

There was no main effect of stimulation condition on the distribution of dwell-time in the Target (T; Wald test = 2.46, p = .483), counter-clockwise (CCW; Wald test = 5.84, p = .12), or clockwise (CW; Wald test = 1.41, p = .703) quadrants. Stimulation had no effect on the distribution of quadrant dwell-time (Figure 6b).

5.5.2 |. Goal zone dwell-time ratio

There was a main effect of the proportion of time spent in the goal zone (Figure 6c) between conditions (Wald test = 25.62, p < .0001). Relative to StimHid (mean = 0.193 ± 0.029), the proportion of time spent in the goal zone was significantly higher in CtrlVis (mean = 0.325 ± 0.026, p = .017) and StimVis (mean = 0.297 ± 0.009, p = .002). This finding reflects the increased difficulty of the hidden goal condition, as there was no difference between StimHid and CtrlHid (mean = 0.210 ± 0.017, p = .454). This suggests that ATS was sufficient for the support of spatial memory and did not prevent recognition of the hidden goal zone location.

5.5.3 |. Search distance from goal zone

Search distance was measured as the distance from the goal zone center during continuous inbound slow speed movements. We found a main effect of search distance from the goal zone by condition (Wald test = 40.37, p < .0001). Search distance during StimHid (mean = 8.02 ± 0.559) was significantly further from the goal than CtrlVis (mean = 5.42 ± 0.358 cm, p < .0001), CtrlHid (mean = 6.32 ± 0.571 cm², p = .046) and StimVis (mean = 5.59 ± 0.444 cm², p < .0001). Although stimulation did not interfere with spatial memory when the goal zone is hidden, the mean search distance was marginally greater by approximately 2 cm (Figure 6d). This result suggests slow-theta stimulation may subtly affect the rat’s estimation of linear distance during approaches to the goal zone.

5.5.4 |. Number of entrances

There was no main effect of stimulation condition on the number of entrances (Wald test = 0.02, p = .989). The mean number of entrances was similar in CtrlVis (mean = 158.4 ± 14.55), CtrlHid (mean = 158 ± 10.48), StimVis (mean = 151.2 ± 11.35), and StimHid (mean = 155 ± 32.12). Increased SE during StimHid was driven by rat P3, who tended to make more entrances in this condition (right, Figure S4f). This may serve as another example of how stimulation might subtly affect goal zone approach.

6 |. DISCUSSION

Our experiments aimed to accomplish two primary goals: (a) test for behavioral or cognitive constraints to entrainment efficacy of hippocampal timing signals by low-theta frequency pan-neuronal optogenetic MS stimulation; and (b) test whether low-theta frequency entrainment would be sufficient or disruptive for hidden goal spatial accuracy behavior. In achieving these goals, our experiments provide a novel conceptual framework for how behavioral and cognitive states influence septohippocampal circuit timing. This framework focuses on both baseline endogenous and optogenetically generated ATS, while considering the relationship between dendritic inputs and somatic outputs in relation to navigation and spatial memory.

6.1 |. Behavior and cognitive demand influence theta entrainment of hippocampal timing signals

CA1 theta oscillations are primarily an aggregate of ongoing dendritic current along the somato-dendritic dipole (Buzsaki, 2002; Buzsaki, Anastassiou, & Koch, 2012; Stewart & Fox, 1990) resulting from local and distal synapses modulated by several interneuron types (Figure S7). Low-theta frequency entrainment results in some conflict with endogenous theta generators, but at the level of the LFP and cell burst-timing this competition is significantly affected by behavioral state and spatial task demands.

Entrainment efficacy of the hippocampal LFP improved when animals were at rest or moving slowly (Blumberg et al., 2016). At speeds greater than 2 cm/s, artificial theta conflicted with endogenous theta generators, as determined by the presence of a second peak oscillation at ~9 Hz. The resulting split in theta spectral amplitude between 6 and ~9 Hz was also influenced by task demands. The balance of signal amplitude at each frequency was equal during the visible goal task but significantly favored 6 Hz during the hidden goal task. Further experimentation will be necessary to identify the origin of theta competition in relation to speed, but one possibility is that it stems from septal glutamatergic cells that are more active during movement and at the initiation of movement (Fuhrmann et al., 2015; Justus et al., 2017; Robinson et al., 2016). The resulting balance of excitation and inhibition at hippocampal synapses may therefore dictate the level of influence pan-neuronal MS stimulation has over hippocampal theta in relation to speed.

Despite theta competition at the level of somato-dendritic current rhythmicity, the burst-timing of somatic outputs was predominantly driven by 6 Hz ATS frequency in both task conditions. However, in accordance with the LFP result, a subpopulation of pyramidal cells matched the endogenous response to stimulation at ~9 Hz in the visible goal condition. This sub-population was absent during stimulation in the hidden goal condition, shifting the mean burst-timing rhythmicity even closer to 6 Hz.

Phase preference relative to local hippocampal theta was lost during MS stimulation in Experiment 2. This is likely a consequence of endogenous and artificial theta competition, which we did not find when the animals were at rest in Experiment 1. Given robust burst-timing entrainment in both experiments, it is possible there was a dissociation between ATS influence over dendritic and somatic rhythmicity during movement in Experiment 2. In order to measure the phasic control of the 6 Hz frequency over hippocampal cells, we made phase preference measurements directly from the sinusoidal ATS signal in the MS. This analysis revealed that ATS effectively superseded endogenous CA1 phase organization. We found a robust and stable ATS phase vector in hippocampal cells during both hidden and visible task versions, albeit at an opposite phase from baseline.

Differences in cell phase preference relative to local or septal signals questions the fundamental relationship between somatodendritic current oscillations, perisomatic inhibition and the temporal organization of spike outputs. Septal influence over somatodendritic current rhythmicity is mediated by projections to OLM interneurons while its influence over phase preference of somatic action potentials is mediated by projections to basket cells (Figure S7). Emergent feedback inhibition with pyramidal cells may cause the rhythmic burst-timing frequency of somatic action potentials to be within 1 Hz of the dominant local field potential frequency (Magee, 2001; O’Keefe & Recce, 1993; Zutshi et al., 2018). In accordance with robust phase preference relative to ATS signal in the MS, autocorrelation analyses revealed that the rhythmic frequency in the majority of hippocampal cells was entrained within 1 Hz of the ATS frequency. This was particularly evident during the hidden goal condition. The increased amplitude of both the 6 Hz ATS signal in the LFP at all speeds, and the dominance of 6 Hz cellular burst-timing in the hidden goal task, provide further evidence that cell rhythmicity represents the temporal features of the predominant frequency band in extracellular oscillations (Constantinou et al., 2016; Takahashi & Magee, 2009; Zutshi et al., 2018). Although it is unlikely that LFPs and burst-timing rhythmicity are dissociable phenomena, at least with regard to the baseline integration of dendrite to soma input and output processes (Vaidya & Johnston, 2013), future work will need to study the mechanistic relationship between these timing levels during ATS. There could be biophysical limitations to this relationship, particularly at high-theta frequencies (Zutshi et al., 2018). It is unclear what the consequences of dissociation between cell and LFP timing would be on spatial cognition and memory.

Low-theta frequency stimulation in our study partially controlled theta oscillations but robustly entrained burst-timing rhythmicity. In contrast, high-theta stimulation at 10 or 12 Hz entrained theta oscillations at all speeds but had partial control over burst-timing (Zutshi et al., 2018). One possibility for the difference in ATS control over burst-firing rhythmicity between low-theta and high-theta frequency stimulation is the balance in the response level of OLM interneurons in hippocampus str. oriens versus basket cells in the pyramidal cell layer (Figure S5). It has been suggested that dendritic inhibition via OLM cells effectively controls the burst-firing of pyramidal cells, enhancing regulation of dendritic plateau generation (Buzsaki, Penttonen, Nadasdy, & Bragin, 1996; Miles, Toth, Gulyas, Hajos, & Freund, 1996; Royer et al., 2012). Dendritic inhibition may therefore be critical for filtering which frequency of CA1 inputs will be reflected in the burst-timing rhythmicity of somatic outputs to downstream synapses (Takahashi & Magee, 2009). In contrast, peri-somatic inhibition by CA1 basket cells is considered to be critical for fine control of firing rate and theta phase preference (Royer et al., 2012). The improved control over somatic burst rhythmicity in response to low theta may then be due to the balance of recruitment between OLM and basket cell interneurons. This discrepancy may also explain our stimulation results while rats were at rest in the flowerpot. During control stimulation, hippocampal pyramidal cells exhibited a local phase preference relative to immobility theta in the flowerpot but did not demonstrate a cohesive burst-firing rhythmicity. Although MS stimulation did not alter the magnitude of the resultant phase vector, it did significantly shift the burst-rhythmicity of the majority of hippocampal neurons within 1 Hz of the ATS frequency. This may have been due to the recruitment of OLM interneurons during stimulation (Royer et al., 2012).

Both low- and high-theta stimulation approaches significantly perturbed the linear relationship between speed and theta frequency, but only high-theta frequency stimulation spared the relationship between speed and theta amplitude (Zutshi et al., 2018). Although sparing of this speed and amplitude relationship could depend on high-theta frequency stimulation, the difference could also be due to selective PV+ MS stimulation in the Zutshi et al. study. Pan-neuronal stimulation in our study is likely to recruit septal cholinergic neurons that could affect amplitude gain control in relation to speed (Vandecasteele et al., 2014). However, increased amplitude of 6-Hz signal at slow speeds, and the inevitable inversion of the speed and theta amplitude relationship, would be the simplest explanation for its disruption.

The most conceptually novel aspect of our experiments is the incorporation of two versions of a spatial task that vary cognitive demand. This approach unmasked a potential hierarchy of hippocampal theta generators. The hippocampal preference for MS ATS frequency during the hidden goal version indicates background hippocampal network changes as a consequence of increased spatial task demands (Fernandez-Ruiz et al., 2017). The behavioral and electrophysiological results between visible and hidden goal conditions during baseline lend further support to this interpretation. Behaviorally, the rats spend significantly less time in the goal zone during the hidden goal task than the visible task. This strongly suggests that the hidden goal version of the task, largely involving goal navigation with available spatial cues (Morris et al., 1982), was more difficult than the visible goal version that simply required a beacon strategy. With increased cognitive demand, the baseline theta frequency and speed relationship and theta phase preference dispersion levels significantly improved. The improved relationship between theta frequency and speed during increased memory demand has been shown previously (Richard et al., 2013), but demonstration of this effect in a spatial navigation task is novel. Although the resultant phase angle was stable between both task versions, phase preference dispersion levels improved relative to both hippocampal and septal theta in the hidden goal task. The effect of increased attention and spatial navigation on the phasic organization of hippocampal cells relative to septal theta is also a novel finding. This suggests attention-like mechanisms governing the reliablity of cell signalling within and between brain regions in relation to goal-directed spatial demands (Aoki et al., 2019; Barry et al., 2016; Fenton et al., 2010). A possible mechanism for this increased coordination by theta phase is the improved coupling between septohippocampal GABAergic pacemakers and septal cholinergic cells (Dannenberg et al., 2015). Apart from intraseptal changes, increased spatial attention and the engagement of OLM interneurons through elevated acetylcholine levels (Leao et al., 2012) may increase inhibitory tone at the neocortical inputs to the hippocampus. This could play a mitigating role in amplifying the amplitude relationship between septal output frequency to the hippocampus and local theta oscillations. However, the reciprocal influence of the hippocampus to the MS via the lateral septum cannot be excluded (Tingley, Alexander, Quinn, Chiba, & Nitz, 2018).

6.2 |. Low-theta frequency is sufficient for goal finding but alters goal approach behavior

The most salient advantage of our approach is that it allowed for testing of whether low-theta frequency entrainment of the septo-hippocampal circuit would affect the generation of a stable spatial code and the performance of a spatial task (Barry et al., 2016; Bender et al., 2016; Bolding et al., 2019; Robbe & Buzsaki, 2009; Schomburg et al., 2014; Wang et al., 2015). Despite conflict between theta generators at synaptic inputs, low-theta frequency ATS significantly supported pyramidal cell phase preference, dominated burst-timing rhythmicity frequency, and ultimately did not affect most aspects of spatial accuracy behavior. We address three theories that predicted disruption of memory or navigation processes by low-theta frequency entrainment.

Based on frequency and function, two types of theta have been characterized in rodents. Movement related or Type 1 theta is typically >7 Hz where the frequency range increases with speed and translates displacement relative to coordinates in a cognitive spatial map. Alert immobility related or Type 2 theta is typically <7 Hz and linked to arousal, anxiety or novelty at slow speeds or at rest (Jeewajee et al., 2008; Kelemen et al., 2005; Kramis et al., 1975; O’Keefe & Nadel, 1978; Sainsbury et al., 1987; Vanderwolf, 1969). A previous study found that activation of cholinergic MS cells decreased theta frequency while animals were foraging in a familiar environment and elicited novelty-responsive behavior such as thigmotaxis (Carpenter et al., 2017). Optogenetically shifting the hippocampal LFP to low-frequency theta in and of itself did not initiate novelty or anxiety behavior as the animals spent the same amount of time in the goal zone during stimulation as in baseline. This implies that frequency decrease in response to novelty (Jeewajee et al., 2008) and its associated behavioral responses are a consequence of changes in cholinergic tone rather than the cause. Alternatively, engagement in a spatial operant task may also counteract the elicitation of novelty behavior by low-theta frequency.

Optogenetic perturbation of the MS does not alter place cell firing field properties. This finding is in agreement with pharmacological inactivation studies that propose endogenous MS signals are not necessary for the generation of hippocampal place fields (Aoki et al., 2019; Bolding et al., 2019; Brandon et al., 2014; Koenig et al., 2011). Yet, the endogenous theta frequency range in relation to movement speed is proposed to provide an egocentric metric frame that translates synaptic strengths to perceived spatial displacement within the coordinate system provided by place cells and grid cells (Burgess, 2008; Buzsaki, 2005; Hasselmo & Brandon, 2008; O’Keefe & Nadel, 1978; Welday, Shlifer, Bloom, Zhang, & Blair, 2011). Inactivating the MS causes errors in the estimation of linear distance (Jacob et al., 2017) while hippocampal stimulation at irregular theta frequencies causes spatial deficits in a water maze task (McNaughton et al., 2006). Shifting the predominant theta frequency <7 Hz in the hidden goal task also negated the relationship between speed and theta frequency underpinning the metric of self-motion, but this did not cause spatial memory deficits. However, during stimulation the rats approached the hidden goal zone center from further away than during baseline. Although this effect is subtle, it suggests the task was either more difficult or the perceived linear distance between the rat’s location and goal zone was altered. This result is similar to the dissociation in spatial processing described by Whishaw between “knowing where” and “getting there” (Whishaw, Cassel, & Jarrad, 1995). Whishaw et al. found that after having learned the location of a hidden platform in a Morris water maze, rats that then received fimbria-fornix lesions were still able to find the platform. However, they did not quickly shorten their swims or develop approach patterns that brought them directly to the platform. Whishaw realized that spatial processing involves both place learning, defined as learning the layout of specific distal cues, and the development of navigational strategies to exploit this learned information. Whishaw suggested that path integration mechanisms involved in navigation, rather than spatial memory per se, were more affected by the fimbria-fornix lesion. Future work will expand on the distinction between memory and navigation by carrying out the stimulation experiment in larger arenas in order to test how the magnitude of the goal approach effect correlates with displacement distance (Jacob et al., 2017) and path integration (Izadi et al., 2019). Future work may also study the effects of ATS on the spatial firing properties and spacing of grid cells and the representation of goals in the entorhinal cortex (Boccara, Nardin, Stella, O’Neill, & Csicsvari, 2019; Brandon et al., 2011; Butler, Hardcastle, & Giocomo, 2019; Koenig et al., 2011).

Low-theta frequency stimulation also superseded local phase preference, causing a phase reversal of the cell ensembles relative to endogenous MS theta. Previous work has shown that perturbation of hippocampal theta phase organization, despite the maintenance of place fields, interfered with spatial working memory (Robbe & Buzsaki, 2009). This phasic organization has been hypothesized to govern the plasticity regimes responsible for memory encoding and retrieval by enhancing communication between CA1 cells and the entorhinal cortex or CA3, respectively (Douchamps et al., 2013; Hasselmo et al., 2002; Hasselmo & Stern, 2014; Manns, Zilli, Ong, Hasselmo, & Eichenbaum, 2007; Siegle & Wilson, 2014). Although theta is not necessary for building (Brandon et al., 2014) or maintaining place cell firing fields, theta oscillations may be necessary for “reading” the place cell cognitive map and retrieving spatial memories (Bolding et al., 2019). Despite phase reversal in the hidden goal task, we did not detect a memory deficit with respect to dwell time in the goal zone quadrant or the goal zone itself. Similarly, using closed-loop stimulation to elicit phase reversal in CA1 cell activity during memory retrieval did not affect performance in a spatial memory task (Siegle & Wilson, 2014). This could mean that spatial memory is not as susceptible to phase disruption as working memory tasks (Robbe & Buzsaki, 2009) or that phasic activity has to be more distributed over the theta cycle in order to have a detrimental effect. Future experiments will examine the effect of phase reversal on the encoding of novel goal zone locations in different quadrants.

The data demonstrate that even a static low-theta frequency ATS signal in the septohippocampal circuit is sufficient for spatial memory, with the caveat that goal approach behavior may be altered by the perturbation of the speed/theta frequency relationship. Again, this suggests a dissociation between “knowing where” and “getting there” (Whishaw et al., 1995). Ultimately, a coherent and rhythmic theta timescale hippocampal output, that coordinates activity within and between the hippocampus and its post-synaptic projections, may be both necessary and sufficient for “reading” the cognitive map. This result is similar to a previous study in which a static 7.5-Hz theta signal that bypassed an inactivated MS was sufficient for spatial learning in the initial training trials of a Morris water maze task (McNaughton et al., 2006). How the spatial code is adapted to spatial behavior under ATS control and how this control affects path integration requires further study.

6.3 |. Conclusions and implications for neuroprosthetics

If the goal of neurostimulation is to enhance or replace endogenous signals it is important to understand how artificial signals might interact with intact endogenous circuits that generate theta oscillations underpinning memory processes (Kaminski et al., 2020). What faculties may be lost or undermined in the stimulation process? Our results indicate there is less competition with endogenous signals when the stimulated circuit is engaged in a cognitive task and that rhythmic entrainment occurs without overall changes in excitation. Although we did not detect a spatial memory impairment, the estimation of linear distance may have been affected. This begs the question as to whether there is a difference in artificial theta’s role in memory retrieval, or the prospective planning between current location and the goal, versus the spatial translation metric in the actual goal trajectory (Gupta, van der Meer, Touretzky, & Redish, 2012). Given the role of theta phase in disambiguating current position and trajectory in the early and late stages of the theta cycle (Huxter, Senior, Allen, & Csicsvari, 2008), it is likely that both would be affected by low-frequency ATS. Perturbation of the frequency/speed relationship and path integration mechanisms in the processing of current and future position may explain why rats began their inbound goal trajectories further from the goal during stimulation in the hidden goal version of the task. Perturbations to theta oscillation frequency do not inhibit recognition of the goal zone (i.e., specific local view of cue card at a certain distance). Our data indicate that endogenous hippocampal theta oscillations may be more important for the more difficult aspect of the task, calculating a path to the goal.

If neuro-stimulation is to occur under encephalopathic conditions, it is also critically important to know if alterations to dendritic morphology and microcircuitry (Patterson et al., 2017) might affect the efficacy of ATS. It is therefore necessary to understand the functional microcircuit dynamics between control and encephalopathic animals in response to optogenetic stimulation (see Figure S1c).

One of the most important insights from these experiments is that we stand to gain a greater understanding of hippocampal microcircuitry, as well as how to best influence the emergent temporal coordination of somatic outputs, by taking advantage of intrinsic septohippocampal circuitry rather than direct stimulation of the hippocampus. Optogenetic ATS has the potential to unmask changing dynamics in the septohippocampal circuit that underpin spatial cognition. Defining this reciprocal relationship between internal brain states (Shanechi, 2019) and entrainment efficacy provides a mechanistic understanding of neurostimulation techniques and how they can best be applied to ameliorate temporal discoordination associated with cognitive deficits (Barry et al., 2016; Fenton, 2015; Izadi et al., 2019; Reinhart & Nguyen, 2019; Shuman et al., 2017; Solomon et al., 2018).