4. Hippocampal neural circuit activity in ASD model mice
One of the benefits of modeling genetic disorders in the mouse is the ability to understand at a circuit and systems level how changes in behavior are related to changes in the well-understood physiology of the hippocampal formation \cite{RN94,RN23}. The last fifty years of research has given us a rich template of physiological signatures or correlates of mnemonic processing in the hippocampus, including the formation and spatial coding of place cells \cite{RN95,RN13}, the temporally organized activity of ensembles of place cells, both during movement and sleep, as well as the oscillatory patterns which dominate the hippocampal local field potential, theta (4-12Hz) gamma (30-100Hz) and sharp wave ripples (SWR; 120-180Hz) \cite{RN97,RN96,RN98,RN99}, which can shed light on how information flow and processing are altered in a dynamic manner (Figure 2). Although these physiological measures are complex, they can serve as indicators of how changes in the balance between excitation and inhibition are manifest on the population level during specific behavioral or memory states.
Theta is most prominent oscillation in the hippocampus during locomotion and attention and has been tightly linked to memory function, with manipulations that decrease theta power linked to encoding deficits \cite{RN102,RN13,RN100,RN101,RN103}. Gamma oscillations are modulated by movement velocity, sensory processing, attention, and cognition and memory and in CA1, have been used as a proxy for the influence of CA3 (low gamma 30-60 Hz) and entorhinal cortical (high gamma, 60-100 Hz) inputs on circuit function \cite{RN104}. Moreover, the timing and amplitude of these distinct gamma bands can be modulated by the slower theta oscillation, a phenomenon termed cross-frequency coupling, thought to be important for the temporal organization of circuit activity \cite{RN97,RN94,RN105,RN106}. Finally, SWRs, triggered by input from CA3 and/or CA2 \cite{RN99,RN107,RN108} create short periods (~100-200 msec) of precise temporally organized neuronal spiking during slow-wave sleep and quiet wakefulness and have been implicated in memory consolidation, storage and recall \cite{RN99}. All these oscillations require not just external inputs to the CA1 region, but also precise interaction in local microcircuits containing both excitatory and inhibitory neurons \cite{RN99}. Thus, understanding how genetic models of ASD impact these rhythms and the temporal and sequential coordination of spiking in the structure they support is an important bridge between behavioral phenotypes and shifts in E/I balance on the level of circuits and synapses.
Both mouse \cite{RN40,RN109} and rat \cite{RN110,RN111} models of FXS have been generated and subject to in vivo electrophysiological analysis. Two studies recorded from the CA1 region of Fmr1-/- mice performing an active place avoidance task \cite{RN113,RN112} and observed changes in the temporal and spatial coordination of hippocampal oscillations in a cognitive state dependent manner, with alterations in the patterns of coupling between the theta and slow gamma rhythms, as well as an inflexibility of the spatial representations in the Fmr1-/+ mice. A third study from the same lab \cite{RN114} recorded CA1 neuronal activity and local field potential in Fmr1-KO mice in a fixed context and found that while place fields were relatively intact, on the network level pyramidal cells were less modulated by ongoing theta and gamma oscillations and place cells with overlapping fields showed a decrease in positively correlate firing, forming weaker cell assemblies. Another 2018 study from Arabab et al using the same model \cite{RN115}reported increases in theta power and local gamma coherence on the LFP level. On the network level they observed the pairs of interneurons in CA1, as well as interneuron-pyramidal cell pairs, showed significant decreases in spike count correlation, indicating a reduction in the correlated variance of these cell assemblies, and hypersynchrony between inhibitory neuron firing and the theta and gamma oscillations. A third group published a 2018 study recording in the CA1 region of Fmr1-KO mice \cite{RN116} finding a significant decrease in the frequency of REM sleep, increased firing of pyramidal cells during both wakefulness and rest, an increase in low gamma power and alterations in SWRs, with longer and slower oscillations coupled with a decrease in pyramidal cell firing across the events. While there are some disparities between these results, which could be related to the tasks and contexts employed for recording, there is consistent findings of dysregulation of oscillatory coupling of neurons with the local field potential and changes in network coordination.
More recently \cite{RN117} Asiminas et al reported the first in vivo recordings from the hippocampus of a rat model of FXS. They observed while place fields in a novel environment were similar between control and Fmr1-KO rats, the FXS-model animals failed to show experience-dependent improvements in spatial coding when returned to the context the following day. Further, consistent with results from mice, they observed a decrease in the modulation of pyramidal cell firing by the slow gamma oscillation and significant shifts in the preferred firing phase of pyramidal cells to both theta and slow gamma.
Hippocampal activity in the Mecp2-/+ mice has also been carefully studied at the single neuron and network levels. Lu et al \cite{RN50} used both 2-photon calcium imaging and in vivo electrophysiology to establish that while CA1 pyramidal cells in the KO mice are overall less active, they showed a significant increase in synchronous activity, which interestingly could be rescued by deep brain stimulation of the fornix. A second study employing high-density tetrode recording in the CA1 region \cite{RN118} found that, like what was reported in the Fmr1-KO rats \cite{RN117}, place fields in the KO mice failed to show experience-dependent improvements in the spatial coding. Interestingly, the authors also observed the hypersynchronous activity in these mice extended to the SWR events, perhaps occluding learning-dependent consolidation mechanisms necessary for place field refinement. Finally, a recent study \cite{RN119} employed 1-photon calcium imaging in the CA1 of Mecp2-/+ mice subject to contextual fear conditioning and observed that during memory recall the active CA1 neuronal ensembles were larger and more correlated, changes the authors attributed to a specific deficit in the function of the OLM class of CA1 interneurons.
A recent study examined the impact of the deletion of NGL3 on hippocampal physiology \cite{RN120}, focusing on the dorsal CA2 and CA3 regions of the circuit, as they have been implicated in social memory \cite{RN121,RN122}. These mice, which have an impairment in social behavior, demonstrated CA2 specific alterations in the entrainment of pyramidal cell spiking by slow oscillations, as well as decrease in gamma power in both the CA2 and CA3 regions. Ex vivo recordings found a shift in the E/I balance towards excitation, with CA2 pyramidal cells in the KO mice showing an increase in spontaneous excitatory input with a concomitant decrease in spontaneous inhibitory input, suggesting this shift of the local network excitation could connect the impairments in oscillatory activity and temporal coordination of spiking to deficits in social behavior.
Cntnap2 KO mice capture ASD-like behavioral phenotypes, including social impairments, reduced vocalization repetitive behaviors and impaired cognition \cite{RN123}, and in the CA1 region have an E/I balance shifted towards excitation, with reduced perisomatic inhibition of pyramidal cells \cite{RN60,RN59}. In vivo hippocampal recordings in behaving mice revealed that during movement there was an overall decrease in theta power, as well as impaired phase-amplitude coupling between fast gamma, thought to reflect inputs from the entorhinal cortex (EC), and theta oscillations in the KO mice. During rest, the occurrence of ripples decreased, as did the amplitude of the oscillations themselves, with no change in the size of the sharp wave, suggesting CA3 input was unchanged. This is consistent with the observation of a decrease in PV density and a decrease in inhibitory input to pyramidal cells, suggesting this is a result of the shift in the local E/I balance that dampens the ability of the circuit to generate oscillations capable of entraining pyramidal cell activity \cite{RN59}.
Three studies have examined the impact of the loss of function of Shank3 on hippocampal in vivo physiology, making it one of the best characterized models in terms of circuit function. Dhamne et al. \cite{RN124} conducted long-term EEG recordings in control and Shank3B-/- mice, both under baseline conditions and following chemical induction of seizure. Interestingly, these mutants were resistant to PTZ induced seizure, suggesting an E/I balance shifted to increased inhibition and/or reduced excitation, consistent with ex vivo recording data \cite{RN124}. Further, under baseline conditions Shank3B-/- mice demonstrated increased power in the gamma band. Cope et al. \cite{RN125} recorded from the ventral CA1 region of the same Shank3B-/- mice during social behavior and observed no change in theta or gamma power in the mutants but did observe that chemogenetic activation of the CA2 region increased CA1 theta power specifically in the KO mice and led to a concomitant rescue of social behavioral deficits. Interestingly, a recent study from Tao et al. \cite{RN126} conducted a similar study, recording in vCA1 of control and Shank3-/- mice during a social discrimination task and found a decrease in the fraction of cells encoding social information during the task, as well as a decrease in the power of SWRs and a decrease in the correlation between cell sequences observed during behavior and SWRs. These results suggested an impairment in the reactivation of social sequences in these animals. Mechanistically, the phase locking of interneurons to the SWR oscillation was impaired in the mutants relative to controls, although there were no differences in the entrainment of PC firing, suggesting the shift in the activity of the inhibitory neurons may underlie the impairments in the re-expression of sequential activity.
In vivo hippocampal activity has also been examined in SHANK2 deficient mice. Sato et al. \cite{RN127} employed longitudinal 2-photon calcium imaging of CA1 pyramidal cell activity in head-fixed mice performing a virtual reality-based goal localization task. While the pyramidal cells in control mice stably overrepresented both the locations of landmarks and rewards following learning, Shank2 mutant mice, engineered to mimic a human ASD-linked microdeletion, demonstrated overrepresentation of the reward sites, but not at the location of salient landmarks.
Turning to genes that impact neuronal excitability, mice that are haploinsufficent for Scn1a (SCN1a+/-) exhibit impaired social and spatial cognition and stereotypic behaviors and in vitro recordings in CA1 revealed a profound shift in the E/I balance, with reduced spontaneous inhibitory currents and enhanced spontaneous excitatory currents \cite{RN61}. Recently, in vivo recording was performed in the hippocampus of rats with injected with a short-hairpin virus to knockdown expression of the Nav1.1 channel \cite{RN128}. They observed a specific decrease in the firing rate of inhibitory neurons, consistent with the ex-vivo data, and a shift in the E/I balance towards increased excitation. On the level of oscillations this resulted in weaker phase-amplitude coupling between theta and gamma and impairments in the theta-modulated spiking of PCs. There was a shift of the preferred phase to the descending phase of theta and significantly weaker theta phase precession. Further, on the sequence level, the relationship between the spiking of pairs of place cells with nearby place fields, typically observed in normal rats, was impaired, with a breakdown of the expected relationship between distance of the fields and phase offset.
Finally, recordings in the CA1 region of freely behaving SCN2a+/- mice found no changes in theta or gamma oscillations during exploration, nor in the firing or spatial coding of the pyramidal cells. However, during SWRs there was a significant reduction in the reactivation of cell assemblies, and on the level of sequences, the replay of behavioral place-cell sequences was significantly shorter, attributed to shift in the E/I balance towards increased inhibition \cite{RN129}. Although the loss of a single copy of Nav1.1 and Nav1.2 led to distinct physiological phenotypes, they both resulted in a decrease in the ability of the hippocampus to accurately represent longer trajectories through space, consistent with dysfunction in the local CA1 circuits.