Reliable and reversible manipulation of hippocampal theta oscillations on a linear track.

A, Schematic of experimental approach. Rats implanted with electrodes targeted to hippocampal region CA1 and a tapered optic fiber targeted to the dorso-ventral extent of the medial septum. Optogenetic stimulation was applied only when the rats were locomoting on a behavioural arena (a linear track shown), using a real-time closed-loop system. B, Two types of optogenetic stimulation protocols, either rhythmic stimulation (defined by period and pulse length), or a theta phase-specific stimulation (90 degree targeting shown), was applied in each session (upper grey boxes). A day consisted of interleaved run and rest epochs of 15–20 minutes each. Run epochs included ‘test’ intervals (purple, “ stimulation-on”), where the optogenetic stimulation would be applied during locomotion, and “control” intervals (blue, “ stimulation-off “), where no optogenetic manipulation was applied during locomotion. C, Virus (AAV5 Ef1a DIO-hChR2(H134R), labeled with eYFP) targeted the medial septum (left) and long-range projecting axon fibers in the hippocampus (right). D, Individual examples (upper two) and epoch-averaged (lower) hippocampal CA1 LFP traces during 10 Hz rhythmic stimulation in one control animal (grey) and one targeted animal (orange) triggered by stimulation onset. Error bars show 95% confidence intervals (CI). E, Power spectral density of hippocampal LFP during 10Hz stimulations during stimulation-on and stimulation-off intervals in one control animal (upper) and one targeted animal (lower). F, Entrainment scores distributions across sessions for n = 3 control animals (grey) and n = 5 targeted animals (orange). Targeted animals show significant entrainment compared to controls (mixed effect linear model (animal, transfection state), effect of transfection; p = 0.0015, targeted n = 19816 interval pairs, control n = 8517 interval pairs). G, Individual examples (upper two) and epoch-averaged (lower) hippocampal CA1 LFP traces during theta phase-specific stimulation in one control animal (grey) and one targeted animal (orange) over time since stimulation onset. Error bars show 95% CI. H, Power spectral density of hippocampal LFP during theta-phase specific manipulation during stimulation-on and stimulation-off intervals in one control animal (upper) and one targeted animal (lower). I, Suppression score distributions across sessions for n = 3 control animals (grey) and n = 5 targeted animals (orange). Targeted animals show significant suppression compared to controls (mixed effect LM (animal, transfection state), effect of transfection p = 2.6 × 10−27, targeted n = 34268 intervals, control n = 14250 intervals). Note, one control animal (Rat I) lacked sufficient runs, and one targeted animal (Rat D) was not tested at the highest laser power during locomotion. Hence these animals could not be included in the entrainment and suppression score calculations.

Theta disruption impairs temporal coding of pyramidal cells, but preserves place code on linear track.

A, Raster plots of simultaneously recorded hippocampal neurons ordered by the peak of their place fields with MS optogenetic theta phase-specific stimulation pulses highlighted (purple) in one control (top) and one targeted (bottom) animal. Note that in control rats, endogenous theta sequences are apparent. B, Stimulation-triggered averages of simultaneously recorded hippocampal neurons show the absence of consistent stimulation-triggered organization in control animal (top) and entrainment of spiking by optogenetic activation in targeted animal (bottom). Vertical dashed lines indicate stimulation time. Firing rates (normalized) shown in heatmap. C, Stimulation-triggered MUA across all control (n = 499 units, 4815 pulses, top) and targeted (n = 583 units, 5143 pulses, bottom) animals shows distinct MUA spiking structure during stimulation. In control animals, the expected relationship is observed, with spiking probability decreasing during stimulation times that occur in the ascending theta phases, whereas in targeted animals spiking probability increases ∼10 ms after stimulation onset. Shaded regions are 95% CI of the mean. D, Pairwise temporal offsets between cross-correlograms computed in stimulation-on versus stimulation-off intervals in control (n = 3674 pairs, top) and targeted (n = 2375 pairs, bottom). Note, a strong diagonal indicates that the cross-correlation structure is maintained in control animals (as expected). Control vs. targeted p < 10−4 for pooled data; hierarchical bootstrap p = 0.15. E, Examples of three place cells’ spatial selectivity in targeted animals during stimulation-off (blue) versus stimulation-on (purple) intervals. Note similar spatial tuning patterns. F, Spatial stability across all identified place cells in one targeted animal during stimulation-off (left, blue box) and stimulation-on (right, purple box) periods in one targeted animal. Note similar spatial tuning patterns. G, Place field stability is measured for all units as the KL divergence (see Methods) of their spatial firing selectivity between stimulation-on and stimulation-off intervals (t-test of distributions of KL divergence between targeted (n = 35 units satisfying inclusion criteria across n = 3 animals) and control (n = 43 satisfying inclusion criteria across n = 4 animals) animals, p = 0.8).

Theta disruption impairs learning of outbound trial structure in a novel W-track environment.

A, W-track schematic for Outbound trials (left). Learning curve for control animals (n = 4, middle, grey) and targeted animals (n = 4, right, red) on Outbound trials. B, Performance accuracy (left) and learning trial (right) for Outbound trials (n = 8 control animals, grey; n = 4 targeted animals, red; 4 control animals from this study, 4 control animals from Joshi et al., 2023). Performance accuracy t-test p = 0.010. C, W-track schematic for Inbound trials (left). Learning curve for control animals (n = 4, middle, grey) and targeted animals (n = 4, right, red) on Inbound trials. D, Performance accuracy (left) and learning trial (right) for Inbound trials (n = 8 control animals, grey; n = 4 targeted animals, red; n = 4 control animals from this study; n = 4 control animals from Joshi et al., 2023). Performance accuracy p = 0.126. E, Probability of “repeat” error trials (re-visiting a reward port on outbound trials) in control animals (grey solid line, left, n = 1160 trials) and targeted animals (right, red solid line, n = 1233 trials) compared to the 95% CI (grey shaded region) of an analytical chance distribution. Targeted animals’ choices are not different from the chance distribution. F, Left: Distribution of running speeds in control (grey) and targeted (red) animals on inbound and outbound trials (linear mixed effects model: effect of targeting, p = 0.61; interaction (targeting × trial type), p = 6.52 × 10−4) during time spent outside reward ports. Right: Distribution of occupancy times around the choice point (right) in control (grey) and targeted (red) animals on inbound and outbound trials (linear mixed effects model: effect of targeting, p = 0.94; interaction (targeting × trial type), p = 1.39 × 10−4).

Theta disruption suppresses fast timescale sequential position representations.

A, Examples of decoding position from population spiking on outbound trials in control animals. From top to bottom: actual linearized position (red) and estimated position posterior (greyscale); multiunit spike rate (spikes/s); ahead/behind distance of decoded position to animal’s actual position (cm); probability of a spatially continuous decoding state; and entropy of posterior over space. B, Power spectrum of the ahead/behind distance for stimulation-on and stimulation-off intervals across all trials in control animals, separated by trial type (top: outbound; bottom: inbound). Note the clear peak near 8 Hz reflecting the presence of robust theta sequences. C, Quantification of stimulus-triggered probability of the posterior to be continuous during stimulation-on trials for control animals. Note prominent theta-timescale structure reflecting more continuous representations in early phases of theta. D, Quantification of stimulus-triggered posterior entropy across all trials in control animals. Note increases in entropy toward later phases of theta corresponding to stimulation times. E, Examples of decoding position from population spiking on outbound trials in targeted animals; plot elements as in A. Note the disruption in theta-timescale structure. F, Power spectrum as in B for targeted animals. Note suppression of theta-timescale peak on outbound trials (top) and absence of obvious structure on inbound trials. G, Probability of continuous state as in C for targeted animals. Note altered distribution as compared to C. H, Entropy of posterior as in D for targeted animals. Note altered distribution as compared to D. I, Quantification of relative power in the power spectrum of the ahead/behind distance in the 8–12 Hz band relative to the mean of the 4–8 and 12–15 Hz bands. All comparisons were significant (all pooled data pairwise t-tests: p < 10−4; all hierarchical bootstrap tests: p < 0.05).

Theta disruption does not impact sharp-wave ripples (SWRs) or replay.

A, Examples of decoding position from population spiking in control (left) and targeted (right) animals during SWR events. Bottom row shows events occurring during the first five minutes of experience on the track (a stimulation-on period). Each panel: Top: actual linearized position (red) and estimated position posterior (greyscale). Middle: multiunit spike rate (spikes/s). Bottom: bandpass-filtered ripple events. SWR events highlighted in light pink. B, SWR rates (top) and length (bottom) do not differ significantly between stimulation-on and stimulation-off intervals in targeted (shades of orange) and control (shades of grey) animals (paired t-test for each animal, p values reported in the figure; comparison of groups not significant, mixed effects linear model p = 0.929 for rates, p = 0.426 for lengths). C, Fractions of non-local SWRs that are spatially continuous, fragmented, or a mix of continuous and fragmented states, in targeted (orange) and control (grey) animals. Note overlapping distributions for each ripple content category. D, Average movement speed of the peak of the posterior distribution for continuous decodes (duration > 50 ms, probability of the event being continuous > 90)%. Distributions of speeds were very similar across control and targeted animals (mixed effects linear model, p = 0.253). Note, Rat B had a 32Ch linear probe and hence could not be used for SWR-associated content analysis. Rat L did not have enough continuous replay qualifying the inclusion criteria, but relaxing the inclusion criteria yielded similar results. We noted that there were some very high speed values. A visual inspection of those events suggested that they arose from cases where the posterior distribution was bimodal, and the peak of the distribution transitioned from one mode to the other over a single timestep, thus suggesting high speed movement even though each mode of the distribution was evolving much more slowly.

Entrainment of theta at different stimulation frequencies.

A, Entrainment scores distributions across sessions for control (n = 3, grey) and targetd (n = 3, grey) animals at 6 Hz entrainment frequency. Targeted animals show significant entrainment compared to controls (mixed effects linear model (animal, targeted state), effect of targeting; p = 3.3 × 10−7, control n = 17492 interval pairs, targeted n = 21040 interval pairs). B, Entrainment scores distributions across sessions for control (n = 3, grey) and targeted (n = 5, orange) animals at 8 Hz entrainment frequency (control n = 8012 interval pairs, targeted n = 34184 interval pairs). C, Entrainment scores distributions across sessions for targeted (n = 5, orange) and control (n = 3, grey) animals at 12.5 Hz entrainment frequency. Targeted animals show significant entrainment compared to controls (mixed effects linear model (animal, targeted state), effect of targeting; p = 1.4 × 10−6, control n = 20152 interval pairs, targeted n = 30655 interval pairs). D, Entrainment scores distributions across sessions for control (n = 3, grey) and targeted (n = 2, orange) animals at 20 Hz entrainment frequency. Targeted animals show significant entrainment compared to controls (mixed effects linear model (animal, targeted state), effect of targeting; p = 4.3 × 10−84, control n = 10315 interval pairs, targeted n = 12863 interval pairs).

Accuracy and effects of theta phase–specific stimulation.

A, Stimulus-triggered phase histogram for all linear track targeted animals (mean vector length = 0.4; p < 0.005; target phase = 90◦; pulse length > 20 ms; number of pulses = 9275). B, Normalized spike rate during optical stimulation in targeted (red, n = 307 neurons) and control (grey, n = 295 neurons) animals during theta phase–specific stimulation (theta phase = 90◦). For each neuron, spikes were collected within each stimulus event and binned by distance of the animal from the neuron’s place field peak defined from whole-epoch data. Shaded regions are the 95% CI with neuron resampling (see Methods). Note that extra spikes from stimulation are strongly localized to the place field center in targeted animals.

Disruption of short-timescale sequential spiking with preserved place fields in targeted animals.

A, Quantification of normalized multi-unit activity triggered by onset of stimulation pulses shows an increase in spiking ∼10 ms after stimulation onset on the W-track (as in the linear track) in targeted animals. As expected, control animals show a decrease in spiking (stimulation is phase-specific to the ascending phase of theta). Shaded areas are the 95% CI of the mean normalized firing rates in each condition. B, Pairwise temporal offsets between cross-correlograms of putative pyramidal neurons computed in stimulation-on versus stimulation-off intervals in control (n = 9582 pairs, left) and targeted animals (n = 17285 pairs, right). Note, a strong diagonal indicates that the cross-correlation structure is maintained in control animals (as expected). This structure is much less apparent in transfected animals during both stimulation-on and stimulation-off periods. C, Hierarchical bootstrap comparison of R2 values of linear fit, p = 8.6 × 10−4. D, Density plots of short-vs. long-timescale peaks for stimulation-on and stimulation-off periods for all cell pairs pooled across animals and run epochs. Top row: control animals; bottom row: targeted animals. The larger number of pairs for stimulation-on periods in targeted animals reflects a longer experimental period and stimulation driving larger numbers of short-timescale coincident firing events in cell pairs where too few events were detected for inclusion in the stimulation-off periods. E, Hierarchical bootstrap comparisons of R2 values across conditions demonstrating significant differences during both stimulation-on and stimulation-off periods. F, Place field stability for all units across control and targeted animals, measured as the KL divergence of their spatial firing selectivity between stimulation-on and stimulation-off intervals (t-test of distributions of KL divergence between targeted and control animals, p = 0.5). G, Quantification of the spatial extent of non-local representations at the choice point (maximum distance of the posterior ±65 ms around the stimulus) during inbound (IN) and outbound (OUT) task phases on the W-track for targeted (red) and control (grey) animals. Note that both targeted and control animals have a similar extent of non-local spatial representations.

Effects of rhythmic vs. phase-specific stimulation on decoded representations.

A, Power spectrum of the decode-to-animal distance trace for control (left) and targeted (right) animals in stimulation-on (purple) versus stimulation-off (blue) trials during rhythmic stimulation condition on the linear track. Note that the posterior rhythmically represents current and non-local positions at the targeted frequency of 6 Hz. Control animals continue to represent positions at 8 Hz despite the stimulation. This rules out that light flickering has an impact on the hippocampal representation. B, Power spectrum of the decode-to-animal distance trace for control (left) and targeted (right) animals in stimulation-on (purple) versus stimulation-off (blue) trials during rhythmic stimulation condition on the linear track. Note that the posterior rhythmically represents current and non-local positions at the targeted frequency of 8 Hz. C, Power spectrum of the decode-to-animal distance trace for control (left) and targeted (right) animals in stimulation-on (purple) versus stimulation-off (blue) trials during rhythmic stimulation condition on the linear track. Note that the posterior rhythmically represents current and non-local positions at the targeted frequency of 10 Hz. D, Power spectrum of the decode-to-animal distance trace for control (left) and targeted (right) animals in stimulation-on (purple) versus stimulation-off (blue) trials during rhythmic stimulation condition on the linear track. Note that the posterior rhythmically represents current and non-local positions at the targeted frequency of 12.5 Hz. E, Power spectrum of the decode-to-animal distance trace for control (left) and targeted (right) animals in stimulation-on (purple) versus stimulation-off (blue) trials during theta-phase–specific stimulation targeted at the ascending phase of theta (90◦) on the linear track. Note that the posterior rhythmicity is markedly reduced in targeted animals compared to control animals. F, Power spectrum of the decode-to-animal distance trace for control (left) and targeted (right) animals in stimulation-on (purple) versus stimulation-off (blue) trials during theta-phase–specific stimulation targeted at the ascending phase of theta (90◦) on the W-track. Note that the posterior rhythmicity is markedly reduced compared to control animals. We also note that the posterior rhythmicity does not recover during control periods on the W-track.

Laser-power dependence of LFP entrainment.

A, Stimulus-triggered LFP in one targeted animal shows a laser power–dependent response during REST periods at 5, 25, 50, and 77 mW laser powers and 1 ms pulse width. B, Stimulus-triggered LFP in one targeted animal shows a laser power–dependent response during RUN periods at 5, 25, 50, and 77 mW laser powers and 1 ms pulse width. C, Stimulus-triggered LFP in one targeted animal at 77 mW laser power and 40 ms pulse width shows complete entrainment during RUN periods.