Differential synaptic responses of mPFC neurons to hippocampal and reuniens stimulation

A. Schematic of experimental setup: intracellular recordings of the mPFC were obtained in vivo during targeted stimulation of the hippocampus and reuniens (top). Representative segment of intracellular recordings of the mPFC and proximal prefrontal LFP recorded in vivo during anesthesia (bottom).

B. Electrolytic lesions at stimulation sites in the ventral hippocampus (top) and midline thalamus (bottom) in Nissl-stained coronal sections.

C. mPFC synaptic responses to hippocampal stimulation (top) and reuniens stimulation (bottom), recorded from same neuron, 5.2 mm ventral from the cortical surface. Note, variable, complex responses to hippocampal stimulation (top) and rapid, consistent responses to reuniens stimulation (bottom).

D. Location of recorded cells, estimated according to the dorsoventral position of the recording pipette with respect to the cortical surface.

E. Heterogeneous postsynaptic potential responses of a single mPFC neuron to consecutive hippocampal stimulation (left): early double depolarizing response (top), early single response (middle), and no response (bottom). Reversal of early response with progressive depolarization (right), indicates involvement of GABAA-mediated inhibition (n=5 pulses at each voltage).

F. Intracellular responses to single pulse stimulation of the reuniens nucleus were orthodromic (left, n=17 neurons), antidromic (middle, n=2 neurons) or both orthodromic and antidromic (right, n=1 neuron), consistent with bidirectional connectivity between the reuniens and the mPFC. Jitter in response latency characterized orthodromic spikes (bottom, left), compared to the consistent timing of antidromic spikes (bottom, middle) that were further confirmed with collision tests (not shown).

G. Population data (n=49 neurons from 15 animals, mean ± SD), depicting the response properties of mPFC neurons to hippocampal and reuniens stimulation: response latency (6.05±3.6 ms and 9.95±4.6 ms for reuniens and hippocampal stimulation, respectively, t-test, p<0.01), response amplitude (3.1±2.0 mV and 8.4±4.1 mV for reuniens and hippocampal stimulation, respectively, t-test, p<0.01), slope of the early postsynaptic potential (1.9±0.9 V/s and 0.56±0.27 V/s for reuniens and hippocampal stimulation, respectively, t-test, p<0.01) and duration of the late, hyperpolarizing component (p=0.2, t-test).

H. Schematic of experimental setup illustrating targeted stimulation of the hippocampus and reuniens and field responses recorded from the mPFC at 1 mm intervals along the medial wall of the pericruciate, prefrontal cortex (top). Evidence for antidromic field responses elicited by reuniens stimulation: short-latency, depth-negative events (bottom, left) compared to orthodromic responses lacking this event (bottom, right).

I. Topography of mPFC responses to hippocampal and reuniens stimulation.

Stimulation of the reuniens evoked field responses in a wider prefrontal territory than hippocampal stimulation. Traces show the average mPFC field response to 10 pulses (0.2 ms, 0.3-1.5 mA at 1 Hz). Regions where hippocampal stimulation elicited responses overlapped with regions where reuniens stimulation elicited antidromic responses (grey shading, stars indicate traces expanded in Figure 1H, bottom).

Data are presented as mean ± SD; *p < 0.05, **p <0.01, ***p<0.001 (paired t-tests). Reu - reuniens, RH – rhomboid, CE – central, pars medialis, fht – hypothalamic-tegmental fasciculus

Reuniens activity and hippocampal SWRs precede the onset of mPFC spindles

A. Schematic of experimental setup illustrating LFP recordings of the hippocampus and mPFC and targeted spike/LFP recordings of the reuniens during natural sleep.

B. A segment of LFP recordings of the mPFC, reuniens and hippocampus during wake, NREM and REM states. For simplicity, we have chosen a segment with wake followed by unusually short NREM, followed by REM. Sleep/wake states were detected according to mPFC delta (1-4 Hz) power, mPFC signal amplitude and nuchal muscle activity (EMG).

C. The average power spectrum of LFP recordings in detected states, showing high gamma power in the mPFC during wake, high delta, and spindle power in the mPFC during NREM and high theta power in the hippocampus during REM (n=52 recording sessions from 4 animals).

D. Muscle activity in detected wake, NREM and REM states. The red line in boxplot indicates the median, box limits represent the first and third quartile and whiskers indicate data range. ***p<0.001; post-hoc Tukey test after one-way ANOVA, n=52 recording sessions.

E. Representative segment of LFP recordings during NREM with corresponding wavelet transforms of the signal, showing the correlated activity of mPFC slow waves, spindles, hippocampal SWRs (top and bottom) and reuniens multi-unit activity (middle). The frequency axes for wavelet transforms are plotted in logarithmic scale.

F. Representative example of overlapping reuniens and mPFC spindles. Blue lines indicate automatically detected spindles in reuniens and mPFC LFPs.

G. The relative onset of reuniens spindles with respect to the onset of mPFC spindles in 100 ms bins (top). The black arrow indicates the median delay time (median ± MAD = - 0.147±0.556s, Wilcoxon signed rank test, p = 6.13 x 10-15, n=346). The distribution of median delay times between each reuniens-mPFC spindle pair for each recording (n=49) in 20 ms bins (bottom, median = -0.083 s, n=49 sessions).

H. Reuniens single-unit activity relative to the phase of mPFC spindle cycles, aligned to mPFC spindle peaks. The average prefrontal LFP during spindles, filtered in the spindle frequency range (top left). Spike raster and the associated peri-event histogram of a reuniens single-unit during the above mPFC spindle (bottom left, 10 ms bins). Phase-locking of the same reuniens single-unit to mPFC spindle cycles (top right, n=351 phases, p = 9.120 x 10-29, Rayleigh test) and the mean phase preference of all significantly modulated reuniens single-units (bottom right, n= 24/49 single-units, mean ± SD = - 80.87 ± 41.64°, p = 2.942 x 10-7, Rayleigh test).

I. Reuniens single-unit activity relative to the phase of reuniens spindle cycles. Same analysis as in H but referenced to reuniens spindles in one session (n=451 phases, p = 8.048 x 10-18, Rayleigh test) and population data (n= 29/49 single-units, mean ± SD = - 118.25 ± 47.72°, p = 1.047 x 10-6, Rayleigh test). Polar histograms indicate the normalized count of LFP phases at spike time. The black line indicates the mean phase angle of the associated histogram. Spikes occurred at earlier phases referenced to reuniens spindles compared to mPFC spindles (p = 0.010, Watson-Williams test).

J. The amplitude of spindles detected in the mPFC versus the amplitude of the nearest reuniens spindle in one recording session (top, n=343 spindles, r=0.54, p = 3.22 x10-27, Pearson correlation) and the value of coefficients (r) in mPFC-reuniens amplitude correlations for all sessions (bottom, n=48 sessions with at least 50 spindles).

K. The amplitude of spindles detected in the reuniens versus the amplitude of the nearest mPFC spindle in one recording session (top, n=428 spindles, r=0.47, p = 8.61 x10-25, Pearson correlation) and the value of coefficients (r) for reuniens-mPFC amplitude correlations for all sessions (n= 48 sessions with at least 50 spindles).

L. The average change in LFP power during mPFC spindles (n=3162) relative to baseline, referenced to spindle onset, showing increased hippocampal ripple power, and increased reuniens spindle power prior to the onset of spindles in the mPFC (top). Histogram of SWR rate in the same time window, referenced to the onset of the same mPFC spindles as above (middle). Compared to shuffled data, SWR rate was significantly higher in the 250 ms window prior to the onset of mPFC spindles (t-test, p<0.05). Raster of reuniens multi-unit activity for a single recording session (n=86 spindles) and the population peri-event histogram (n=3162 spindles), referenced to the onset of mPFC spindles.

See also Figure 2 -figure supplement 1 for unit detection.

Slow wave-spindle and slow wave-SWR coupling in the mPFC-reuniens-hippocampal network

A. A single trial of a slow wave-spindle event, showing raw and filtered LFP traces of the mPFC (top) and reuniens (bottom).

B. Phase-amplitude coupling of mPFC spindles and mPFC slow waves, shown as the average time-frequency representation of mPFC amplitudes in the spindle frequency range, locked to the peak of the mPFC slow wave (percentage change from pre-event baseline, n = 626 slow waves).

C. Phase-amplitude coupling of reuniens spindles and mPFC slow waves, shown as the average time-frequency representations of reuniens amplitudes, locked to the peak of the same mPFC slow waves as in B (percentage change from pre-event baseline, n = 626 slow waves). Black curves in B and C show the average of the mPFC LFP, filtered in the slow wave frequency range (0.5-2Hz).

D. Histogram of slow wave–spindle coupling strengths for mPFC slow waves-mPFC spindles (blue) and mPFC slow wave-reuniens spindles (green). Vertical lines indicate the median of the distributions (median ± MAD, SI strength = 0.18 ± 0.09 and 0.24 ± 0.11 for mPFC and reuniens spindles, respectively, p = 2.126 x 10-17, Wilcoxon signed rank test).

E. Slow wave-spindle coupling strength of mPFC-mPFC pairs versus mPFC-reuniens pairs (n = 49 sessions, p = 3.39 x 10-2, Wilcoxon signed rank test). Each circle represents one session.

F. Population data showing the relative preference of spindles to the phase of the mPFC slow wave, shown as a histogram of SI angle means for mPFC slow wave-mPFC spindle coupling and mPFC slow wave-reuniens spindle coupling (mean ± SD = 27.92 ± 37.59°, p = 4.661 x 10-9, n=26 and mean ± SD = -108.48 ± 21.17°, p = 1.278 x 10-20, n = 36, for mPFC and reuniens respectively, Rayleigh test).

G. A single trial of an overlapping slow waves, showing raw and filtered LFP traces of the mPFC (top) and reuniens (bottom). Circles indicate slow wave peaks.

H. Single session data, showing the distribution of delays between reuniens and mPFC slow waves (vertical dashed line). Reuniens slow waves lagged mPFC slow waves (median ± MAD = 0.126 ± 0.043 s, p = 3.66 x 10-15, n = 82 Wilcoxon signed rank test).

I. Population data of median delay times between each reuniens-mPFC slow wave pair (n=49 sessions, median of the distribution = 0.136 s).

J. Single-trial of raw and filtered mPFC and hippocampal LFPs during one overlapping slow wave-SWR event. Inset shows an expanded view of the detected SWR.

K. Circular histogram of mPFC slow wave phases at the time of SWRs (mean phase ± SD = 162.7 ± 29.37°, p = 5.02 x 10-6, Rayleigh test).

*p < 0.05, **p <0.01, ***p<0.001.

See also Figure 3 – figure supplement 1 for spindle-SWR coupling.

Spindle-SWR coupling, p-values per session.

Hippocampal theta oscillations and SWRs modulate reuniens activity

A. Representative example of LFP recordings of the hippocampus, mPFC and reuniens during REM sleep (top three traces) and the bandpassed (300-8000 Hz) trace of the reuniens signal showing multiunit activity. Wavelet spectrogram of the hippocampal trace, showing prominent theta-band activity during REM. The frequency axis is logarithmic.

B. Expanded view of the recording epoch in the dotted-line box in panel A (left), further expanded in time (right), showing occasional locking of reuniens multi-unit activity to hippocampal theta cycles.

C. Peri-event histogram of a reuniens single-unit, referenced to the peak of hippocampal theta cycles (top). The raster is ordered according to the peak-to-peak latency, with short periods at the bottom and long periods at the top. Dark black dots indicate the timing of the next peak in the trial. The light blue trace is the average waveform of the hippocampal LFP.

D. Circular histogram of theta phase at the time of reuniens spikes, showing significant phase-locking of reuniens single-unit activity to hippocampal theta cycles (n = 5207 spikes, p<10-20, Rayleigh test).

E. Population data, showing peri-event histograms referenced to the peak of hippocampal theta cycles as in B (n=36 single units). Units above the white line were significantly modulated by hippocampal theta cycles.

F. Population data, showing the circular distribution of reuniens firing relative to theta cycles for all modulated single units (n = 21/36 single-units, mean ± SD = -124.5 ± 59.8 °, p = 0.011, Rayleigh test).

G. Single session data, showing power spectra of mPFC, reuniens and hippocampal signals during REM with prominent peaks in the theta range.

H. Single session data, showing signal coherence between mPFC-hippocampus, reuniens-hippocampus and mPFC-reuniens recording pairs during REM.

I. Single session data, showing cross-correlations and relative time lags of reuniens-hippocampus and mPFC-hippocampus recording pairs during REM.

J. mPFC signals lagged hippocampal theta signal by longer delay than reuniens signals lagged hippocampus (top, mean lag ± SD = 87.8 ± 55.5 ms and 109.1 ± 41.52 for reuniens and mPFC, respectively, n=20 sessions, p=0.002, t-test). Peak cross-correlation coefficients were significantly higher for reuniens-Hippocampus recording pairs than for reuniens-mPFC (bottom, mean peak value ± SD = 0.44 ± 0.13 and 0.22 ± 0.07 for reuniens and mPFC respectively=20 sessions, t-test, p<10-20).

K. Representative LFP recordings of the hippocampus and of reuniens multiunit activity during NREM sleep. Light blue traces show the bandpass filtered hippocampal recording in the ripple band (150-200 Hz). Light green shows multiunit activity in reuniens during SWR. Expanded view of a single SWR from the recording epoch in the dotted-line box and multiunit activity in reuniens.

L. Peri-event analysis of reuniens spiking activity around SWRs in one recording session.

The time frequency representation shows the mean wavelet transform of 458 detected SWRs, estimated by Morlet wavelets (top). Overlayed traces show the grand average LFP in the hippocampus (white) and reuniens (green), referenced to the onset of the SWR. Raster of reuniens spikes (every 10th sweep) and associated peri-event spike histogram, referenced to the onset of the SWR (bottom).

M. Proportion of reuniens single-units modulated by SWRs (10/49 activated, 16/49 suppressed).

*p < 0.05, **p <0.01, ***p<0.001.

Multiregional temporal interactions during slow wave sleep events in a hippocampal-thalamocortical model

A. The model is composed of two cortical networks (N 1 & 2, representing mPFC layers 5 and 6 respectively), two hippocampal (CA1-CA3) networks and three thalamic nuclei. The three thalamic networks represent the nucleus reuniens (Reu), mediodorsal thalamic nucleus (MD) and thalamic reticular nucleus (TRN). The cortical and hippocampal networks are composed of one group of identical excitatory (E) and one group of identical inhibitory (I) neurons. Excitatory synaptic connections are shown by arrows and inhibitory connections are indicated as lines ending in a dot.

B. From top to bottom: broadband and filtered (7-15 Hz, 0.5-2 Hz) traces of the membrane potential of excitatory neurons of cortical network 1 (light blue) and 2 (dark blue) with detected spindles (black horizontal lines) and slow waves (circles); broadband and filtered (7-15 Hz, 0.5-2 Hz) traces of the membrane potential of reuniens neurons (blue-green) with detected spindles and slow waves; broadband trace of the membrane potential and firing rate of MD (light green) and reticular neurons (dark green); broadband and filtered (150-200 Hz) traces of the membrane potential of excitatory neurons of the CA1 network (light blue) with detected SWRs (stars).

C. Distribution of thalamic slow wave troughs (down states) locked to the cortical slow wave trough (vertical dashed line at zero) occurring within ±0.5 s plotted in 20 ms bins in the model. The arrow indicates the median of the distribution (n = 277, median ± MAD = 0.113 ± 0.019 s). Thalamic slow waves occur significantly later than cortical slow waves (p < 10−20, Wilcoxon signed rank test).

D. Distribution of thalamic spindle onsets locked to the cortical spindle onset (vertical dashed line at zero) occurring within ±2 s plotted in 100 ms bins for one session. The arrow indicates the median of the distribution (n = 211, median ± MAD = -0.102 ± 0.387 s). Thalamic spindles occur significantly earlier than cortical spindles (p = 1.30 × 10−7, Wilcoxon signed rank test).

E. Phase-amplitude coupling of slow waves and spindles, calculated from the average time-frequency representation of amplitudes in the spindle frequency range, locked to the trough of the mPFC slow wave. Left, average of cortical slow wave trough-locked time-frequency representations (n =391) of the membrane potential of mPFC excitatory neurons (top) and reuniens neurons (bottom). Black curves represent grand average filtered membrane potential of cortical excitatory neurons in the slow wave frequency range (0.5-2 Hz) aligned to the cortical slow wave trough (time 0). Right, circular histogram of synchronization index (SI) angles of cortical (top) and reuniens (bottom) spindles relative to the cortical slow waves. SI angles were nonuniformly distributed (mean ± SD = 148.7 ± 23.9°, p < 10−20 and mean ± SD = 27.4 ± 20.4°, p < 10−20, for mPFC and reuniens respectively, Rayleigh test, n=391 slow waves).

F. Histogram of the slow wave–spindle coupling strengths (the absolute value of the synchronization index) for mPFC (blue) and reuniens (green) spindles. Vertical lines show the median of the distribution (p = 8.45× 10−9, Wilcoxon signed rank test, median ± MAD SI strength = 0.58 ± 0.11 and 0.66 ± 0.15 for mPFC and reuniens, respectively).

G. Average filtered membrane potential of cortical, reuniens and CA1 excitatory neurons, in the slow wave frequency range (0.5-2 Hz), aligned to the cortical slow wave trough (time 0).

H. Distribution of slow wave phases at the time of the SWRs. Zero phase shows the peak (up state) of slow waves (n = 714, mean ± SD = -24.97 ± 62.01°, p < 10−20, Rayleigh test).

I. Average membrane potential of the mPFC, reuniens and CA1 excitatory neurons aligned to the SWR peaks.

J. Average membrane potential of the mPFC (top) and reuniens (middle) neurons and histogram of SWR incidence (bottom), aligned to the onset of cortical spindles.

*p < 0.05, **p <0.01, ***p<0.001.

Bidirectional projections between the reuniens and CA1 control multiregional interactions

A. Left, correlation between reuniens and CA1 membrane potentials during SWRs (maximum value of the covariance between reuniens and CA1 membrane potentials ± 0.5 s around the SWR peak) versus the CA1-reuniens and reuniens-CA1 connection strengths ( show the factor multiplied to and representing the strength of the input from CA1 to reuniens and from reuniens to excitatory neurons of CA1 respectively. corresponds to the network shown in Figure 5). Right, average traces of the reuniens and CA1 membrane potentials aligned to the SWR peaks for 3 different values of CA1-reuniens connection strengths.

B. Left, SWR rate preceding the cortical spindle onset (peak-to-trough of the histogram of SWR incidence in a window of -1 s before the spindle onset) versus the CA1-reuniens and reuniens-CA1 connection strengths. Right, histogram of SWR incidence time-locked to the onset of cortical spindles for 3 different values of CA1-reuniens projection strengths.

C. Left, peak-to-trough of the CA1 deflections around cortical slow wave trough (± 1 s) versus the CA1-reuniens and reuniens-CA1 connection strengths. Right, average filtered membrane potential of cortical, and CA1 excitatory neurons, in the slow wave frequency range (0.5-2 Hz) aligned to the cortical slow wave trough (time 0) for 3 different values of CA1-reuniens connection strengths.

D. Left, mean slow wave phase at the time of SWRs versus the CA1-reuniens and reuniens-CA1 connection strengths. Mean phases are shown only for bins with nonuniformly distributed of phases (p < 0.05, Rayleigh test) and omitted for uniformly distributed phase (white boxes). Right, histogram of slow wave phases at the time of SWRs for 3 different values of CA1-reuniens connection strengths, showing decreasing mean phase with increasing CA1-reuniens connections strength (p = 4.28 x 10-8, p = 7.53 x 10-15 and p = 2.53 x 10-39 for , respectively).

Single-unit detection method from tetrode recordings

A. Illustrative example of reuniens LFPs with spiking activity, recorded by 4 closely-spaced electrodes in tetrode configuration. Recordings show the bandpass (300-8000 Hz) filtered traces of from 4 channels and detected spikes.

B. Average waveforms of detected spikes, sorted according to clustering of waveform features extracted from the each of the 4 channels.

C. Example of feature clustering for the same recording session as above, showing the principal components 1 versus principal component 2. Clusters were determined by a Gaussian mixture model and further refined by removing spikes with large Mahalanobis distance. Separated clusters indicate three classes of spike waveforms, presumed to have been generated by three distinct neurons (purple, green and pink).

D. Cluster separation for trough values for the same session as above.

E. Cluster separation for the peak-to-trough values for the same session as above.

Spindle-SWR phase locking

A. Assessment of SWR phase-locking to reuniens spindle cycles for sessions with significant (top) and not significant (bottom) phase-locking. The TFR shows the average hippocampal LFP in the ripple band, as percent change from pre-event baseline, referenced to the peak of the reuniens spindle (time 0) . Green curve shows grand average of the reuniens LFP, filtered in the spindle frequency range (7-15 Hz).

B. Circular histogram of SI angles of SWR amplitude relative to reuniens spindle phase for the same two sessions as in A, showing significant (p = 0.009, Rayleigh test) and not significant phase-locking (p = 0.893, Rayleigh test)

C. Proportion of recording sessions with significant SWR-reuniens spindle coupling phase.

D. Population data of sessions with significant SWR-reuniens spindle coupling phase, showing the phase histogram of SWR amplitude with respect to spindle cycles and the mean phase preference (black line, n = 10).

E. The same as D for SWR-mPFC spindle coupling (n = 3).