The hippocampus provides the brain's memory system with a subset of neurons holding a map-like representation of each environment experienced. We found in mice that optogenetic silencing those neurons active in an environment unmasked a subset of quiet neurons, enabling the emergence of an alternative map. When applied in a cocaine-paired environment, this intervention neutralized an otherwise long-lasting drug-place preference, showing that recoding a spatial memory engram can alleviate associated maladaptive behavior.
At a glance
- Nat. Rev. Neurosci. 8, 657–661 (2007). , &
- Brain Res. 34, 171–175 (1971). &
- Science 261, 1055–1058 (1993). &
- Science 309, 619–623 (2005). et al.
- Neuron 68, 362–385 (2010).
- Ann. NY Acad. Sci. 1141, 1–21 (2008). , &
- J. Neurosci. 7, 1951–1968 (1987). &
- Hippocampus 1, 193–205 (1991). , &
- Science 305, 1295–1298 (2004). , , , &
- PLoS Biol. 8, e1000403 (2010). &
- Science 308, 873–876 (2005). , , , &
- Nature 430, 456–459 (2004). , , &
- J. Neurosci. 9, 2382–2390 (1989). &
- Neuron 70, 109–120 (2011). , &
- J. Neurosci. 28, 14271–14281 (2008). &
- Cell Rep. 4, 1010–1021 (2013). &
- Science 337, 849–853 (2012). , &
- Proc. Natl. Acad. Sci. USA 98, 9386–9390 (2001). , , , &
- Science 341, 387–391 (2013). et al.
- Nature 513, 426–430 (2014). et al.
- Front. Mol. Neurosci. 7, 82 (2014). , , &
- Science 317, 1230–1233 (2007). , , &
- Front. Syst. Neurosci. 5, 18 (2011). et al.
- Nat. Neurosci. 17, 1658–1660 (2014). , , , &
- J. Neurosci. 19, 274–287 (1999). , , , &
- J. Neurophysiol. 84, 401–414 (2000). , , , &
- Nat. Neurosci. 15, 1265–1271 (2012). et al.
- Neuron 82, 872–886 (2014). et al.
- eLife 3, 04006 (2014). et al.
- Nat. Neurosci. 13, 995–1002 (2010). , , &
- J. Neurosci. 9, 4101–4110 (1989). &
- Nat. Neurosci. 11, 209–215 (2008). , , , &
- J. Neurosci. Methods 220, 149–166 (2013). , &
- Nature 424, 552–556 (2003). , , , &
- Math USSR-Sbornik 1, 457–483 (1967). &
- PLoS One 6, e20996 (2011). , , , &
- J. Comput. Neurosci. 29, 309–325 (2010). , , , &
- Nat. Neurosci. 12, 919–926 (2009). , , , &
- J. Neurosci. Methods 94, 141–154 (1999). , &
- IEEE Trans. Neural Netw. 10, 626–634 (1999).
- http://cran.r-project.org/package=fastICA (2013). , &
- Neuron 78, 166–180 (2013). , &
- Stat. Med. 26, 3997–4008 (2007). , , &
- Neuroreport 14, 2127–2131 (2003). , &
- Behav. Pharmacol. 20, 720–730 (2009). , , &
- Brain Res. Rev. 59, 253–277 (2009). , &
- Phil. Trans. R. Soc. Lond. B 369, 20120518 (2014). , , , &
- Nat. Neurosci. 16, 1802–1811 (2013). et al.
- Supplementary Figure 1: Expression of ArchT-GFP fusion protein under the control of the TRE3G promoter in cultured cells and the mouse hippocampus. (1,004 KB)
(a) Adherent cultured HeLa cells were transfected in the presence or absence of 1µM doxycycline (+/– Dox) in equimolar ratio with two plasmid DNA vectors: pAAV-TRE3G-ArchT-GFP (encoding ArchT-GFP fusion protein under the control of the TRE3G promoter) and pTet-DualOFF (encoding tetracycline-controlled Tet-OFF transcriptional activator). The original pTet-DualOFF (Clontech) vector was modified to remove ZsGreen open reading frame. The expression of ArchT-GFP is blocked in the presence of doxycycline (top; +Dox) but allowed in the absence of doxycycline (bottom; –Dox). Scale bars: 100µm. (b) Left: hippocampal coronal sections from a c-fos::ArchT tagged mouse perfused on Day 6 showing ArchT-GFP expression (native fluorescence) in the dorsal CA1 hippocampus. Dash lines outline the hippocampal principal cell layers. Middle: Dapi staining of the corresponding sections. Scale bars: 500µm. Right: schematics of hippocampal coronal sections (adapted from the Paxinos and Franklin's the Mouse Brain in Stereotaxic Coordinates) showing the histologically verified ArchT-GFP expression in 6 c-fos::ArchT tagged mice (n=4 mice perfused on Day 3 are marked with “ο” and n=2 mice perfused on Day 6 are marked with “x”; one color per mouse). Each colored symbol (“o” and “x”) refers to the localization of the virus expression for each mouse. Numbers in each panel indicate the antero-posterior coordinates from Bregma in mm.
- Supplementary Figure 2: Effect of CA1 light-delivery on the firing activity of CA1 principal cells in c-fos::ArchT mice. (428 KB)
(a) Schematic of the experimental protocol. Two groups of c-fos::ArchT mice (control and tagged) were recorded to assess the effect of ArchT photo-silencing on CA1 neuronal activity during the exploration of the circle. In these mice, the tTA can only interact with the tetracycline-responsive element (TRE3G) during a time window controlled by the doxycycline (Dox) system. Tagged, but not control mice, were transiently taken off Dox to tag CA1 neurons active during the exploration of the circle. The tagged neurons thus expressed ArchT-GFP (Figure 1c) and light was subsequently delivered in the CA1 of both groups during re-exposure to that circle. (b) Distribution of light-driven changes of CA1 principal cells firing rate, measured as a rate score (i.e., difference between the firing rate in light-ON epochs and light-OFF epochs divided by their sum). The distribution of firing rate scores obtained by shuffling the spike trains of these neurons is shown for comparison. Note that in the tagged mice a subset of principal neurons was drastically photo-silenced (putatively tagged neurons) while another subset of neurons exhibited an increased light-ON firing (alternative neurons). (c-f) Activity of CA1 principal cells recorded from tagged mice. (c) ArchT photo-silencing caused an overall increase of the mean firing rate of the CA1 principal cell population (i.e., tagged, alternative and unchanged neurons all included; P=2.2e-16). This was not the case in control mice (0.98±0.06Hz versus 1.03±0.06Hz, P=0.58). (d) Baseline (light-OFF) firing rate of tagged, alternative and unchanged neurons during the first two post-tagging days (Days 1–2) in the circle and the (non-tagged) square environment (mean±s.e.m). Note the rate difference between the three neuronal subsets in the tagged circle (P=4.4e-11) but not the square (P=0.83). However the principal cell population mean rate was similar across the two enclosures (0.87±0.05Hz and 0.89±0.06Hz for circle and square, respectively; P=0.78). Similarly, the spatial tuning difference initially found in the circle between tagged and alternative neurons (see Figure 2 and Supplementary Figure 5) was not observed in the square (spatial coherence: 0.56±0.03 and 0.56±0.01 for tagged and alternative neurons, respectively; P=0.96). (e) Cumulative frequency histogram of principal neuron light-OFF firing rates in the circle and the square (Days 1–2). Note that in the circle there was a higher proportion of low-firing alternative neurons than of tagged neurons (24.54% and zero respectively, using a 0.25Hz rate threshold; D=0.64, P<0.0001, Kolmogorov-Smirnov test). This was not the case in the square (D=0.20, P=0.16, Kolmogorov-Smirnov test). (f) Firing rate of tagged neurons relative to light onsets (Days 1–6; mean±s.e.m; with 15ms time bins). The mean rate score of the tagged neurons (-0.82; see Online Methods) was used as a threshold to isolate two subsets of tagged neurons and test whether the difference in their photo-silencing rate score was due to a difference in their response to light-delivery. We found that these two tagged neuronal subsets (n=83 and 50 for subset 1 and subset 2, respectively) exhibited a similar latency and firing rate-suppression to light-delivery (first light-ON bin: 0.37±0.08Hz for subset 1 versus 0.20±0.07Hz for subset 2, P=0.16; second light-ON bin: 0.03±0.01Hz for subset 1 versus 0.05±0.02Hz for subset 2, P=0.52; second light-ON bin versus remaining light-ON bins: both Ps>0.42). However, the tagged subset with the strongest photo-silencing (rate score<-0.82) exhibited a higher baseline light-OFF firing compared to the other tagged subset (1.19±0.09Hz for subset 1 versus 0.53±0.05Hz for subset 2, P<2.80e-6); this light-OFF rate difference was related to the decreased firing rate of the tagged neurons obtained over days of repetitive photo-silencing (see Figure 2a). **P<0.01, ***P<0.001.
- Supplementary Figure 3: Effect of CA1 light-delivery on the firing activity of CA1 interneurons in c-fos::ArchT mice. (466 KB)
(a) Schematic of the experimental protocol. (b) Distribution of light-driven firing rate score of interneurons, measured as the difference between the firing rate in light-ON epochs and light-OFF epochs divided by their sum. Note that some interneurons in tagged mice exhibited either a light-driven decreased or an increased firing rate. (c-e) Activity of CA1 interneurons recorded from tagged mice. (c) ArchT photo-silencing did not significantly alter the mean firing rate of the CA1 interneuron population (all interneurons included; P=0.07, paired t-test). (d,e) Example firing light-response of six individual interneurons (one interneuron per row). The theta phase distribution of firing probabilities and the firing probability relative to the peak power of the sharp wave/ripple events (normalized by the baseline off-SWR firing rate) were used to classify these cells within two putative interneuron types, namely the basket cells (d) and the bistratified cells (e) as previously described (see Online Methods). Note that within each putative interneuron type, the firing rate of individual cells either increased or decreased during light-delivery, or remained unaffected. This indicates that the switch between tagged and alternative neuronal activities caused by ArchT photo-silencing (see Figure 1e and Supplementary Figure 2b) was associated with a redistribution of interneuron firing activities without a global change in inhibitory tone.
- Supplementary Figure 4: ArchT photo-silencing in the circle, but not context re-exposure per se, caused the long-lasting change in firing rate of tagged and alternative neurons. (314 KB)
(a) Schematic of the experimental protocol. In order to test whether context re-exposure to the circle alone altered the activity of tagged neurons, two groups of c-fos::ArchT mice were transiently taken off Dox to tag CA1 neurons in the circle (tagged mice), with CA1 light delivered either during the subsequent exploration of the circle (light-IN group, n=5) or the non-tagged square (light-OUT group, n=3). The purpose of the ArchT photo-silencing in tagged light-OUT mice was to allow the identification of tagged and alternative neurons without affecting their activity during re-exposure to the circle. (b) Baseline (light-OFF) firing rate and spatial coherence (mean±s.e.m) of light-unresponsive neurons (“unchanged” neurons) recorded from tagged mice light-IN. In contrast to tagged and alternative neurons (see Figure 2), both firing activities were not significantly altered over the six post-tagging days (both Ps>0.34). (c) Example raster plots of two light-modulated CA1 principal cells from a tagged light-OUT mouse. The top cell was photo-silenced (0.93Hz versus 0.03Hz) whereas the bottom cell was excited (1.18Hz versus 3.22Hz) during light-ON epochs and were therefore classified as putative tagged and alternative neuron, respectively. (d) Baseline (light-OFF) firing rate of tagged and alternative neurons recorded in the second block of days (Days 4–6) from tagged mice during re-exposure to the circle. Note that the firing rate of tagged neurons (n=55) recorded from light-OUT mice in the circle was still higher than that of alternative neurons (n=75), despite the successive re-exposure to the circle (light-IN: P=1.6e-5; light-OUT: P=9.7e-10). ***P<0.001.
- Supplementary Figure 5: ArchT photo-silencing altered CA1 place maps in c-fos::ArchT tagged mice. (1,701 KB)
(a) Schematic of the experimental protocol. (b,c) A set of simultaneously recorded CA1 neurons with raster plots, color-coded place rate maps and individual spike locations superimposed on the animal's path in the circle on Day 1 (b) and Day 2 (c) (see Supplementary Fig. 6a for Day 6). Distinct neurons were recorded across days. For each cell, both rate maps and raw spike data are depicted side by side for light-OFF and ON epochs. The top right number of each spatial rate map is the peak firing rate (Hz). The color code of each map is scaled to the peak firing rate or 1Hz for peak rate of less than 1Hz (low-active cells). Warm colors (red) correspond to high firing rate region (i.e., place field) of the cell. Red dots mark locations where the cell fired action potentials while gray traces show animal's path. Note that the clear place field exhibited by neurons #1,4,6,13,18 and 20 on Day 1 and #21,25,29,34,35 and 39 on Day 2 were photo-silenced while the low-firing cells #2,3,7,8,10,11,14,15,17 and 19 on Day 1 and #22,23,26,27,30-32,36,37 and 40 on Day 2 expressed a light-evoked spatially-tuned firing activity. These cells were classified as putative tagged and alternative neurons, respectively. Cells #5,9,12,16 (Day 1) and #24,28,33,38 (Day 2) are examples of neurons with no light-modulated firing activity (“unchanged neurons”).
- Supplementary Figure 6: ArchT photo-silencing altered CA1 place maps in c-fos::ArchT tagged mice. (1,193 KB)
(a) A set of simultaneously recorded CA1 neurons with raster plots, color-coded place rate maps and individual spike locations superimposed on the animal's path in the circle on Day 6. For each cell, both rate maps and raw spike data are depicted side by side for light-OFF and ON epochs. The top right number of each spatial rate map is the peak firing rate (Hz). The color code of each map is scaled to the peak firing rate or 1Hz for peak rate of less than 1Hz (low-active cells). Warm colors (red) correspond to high firing rate region (i.e., place field) of the cell. Red dots mark locations where the cell fired action potentials while gray traces show animal's path. Note that the tagged cells #43,48,49,54,55 and 59 were still photo-silenced on Day 6 although no longer showed well-defined spatial tuning in the baseline light-OFF condition (see Supplementary Fig. 5b,c for comparison). This contrasted with the photo-excited cells #41,42,44,46,50,52,53,57,58 and 60 that expressed a place field in the light-OFF condition. Cells #45,47,51,56 are examples of neurons with no light-modulated firing activity (“unchanged neurons”). (b) For each cell represented in Supplementary Fig. 5b,c and Supplementary Fig. 6a, the light-driven change in firing rate measured from the original spike train as a score (i.e., difference between the firing rate in light-ON epochs and light-OFF epochs divided by their sum; each score represented as a vertical red bar) is compared to the distribution of rate scores obtained from 500 shuffling of the same spike train (in gray) and is represented together with the place rate maps scaled to the peak firing rate for both light-OFF and ON epochs.
- Supplementary Figure 7: Light-delivery did not alter CA1 place maps in c-fos::ArchT control mice. (983 KB)
(a) Schematic of the experimental protocol. (b) A set of simultaneously recorded CA1 principal cells with raster plots, color-coded place rate maps and individual spike locations superimposed on the animal's path in the circle on Day 3. For each cell, both rate maps and raw spike data are depicted side by side for light-OFF and ON epochs. Spatial rate maps are scaled to the peak firing rate (Hz; top right of each map) or 1Hz for peak rate of less than 1Hz (low-active cells). Warm colors (red) correspond to high firing rate region (i.e., place field) of the cell. Red dots mark locations where the cell fired action potentials while gray traces show animal's path. Note that light-delivery did not induce a remapping in these control mice.
- Supplementary Figure 8: The photo-evoked switch between CA1 hippocampal maps in c-fos::ArchT tagged mice was initially associated with an increased locomotor reactivity to the circle. (274 KB)
(a) Schematic of the experimental protocol. (b) Locomotor reactivity of tagged mice during re-exposure to the circle and exploration of the square over the successive post-tagging days. Locomotor reactivity was assessed by the animals’ locomotion speed during the first 5 minutes of light-OFF and light-ON epochs of each exploration. Note that the square was made novel each day and was associated with a higher locomotion. (c) Mean locomotor reactivity of c-fos::ArchT tagged and control mice during the first two post-tagging days. Note that the light-delivery initially caused an increased locomotion speed of tagged mice. This suggests that in the early stage of ArchT photo-silencing the switch from the original to the distinct alternative hippocampal map caused mice to treat the circle as a novel enclosure before it was treated as familiar in later days. *P<0.05 compared to OFF. (d) The population map similarity was established from the population of CA1 place cells simultaneously recorded in the circle to measure in a pairwise fashion the extent to which cells that fired in similar regions of space (i.e., overlapping place fields) in one epoch fired in similar regions in the other epoch. The population map similarity of the spatial firing maps expressed within the light-OFF epochs (first 10min versus second 10min) was used as a baseline. Note that in the early days of light-delivery the hippocampal maps expressed in light-ON epochs were weakly correlated to those expressed in light-OFF epochs in tagged, but not control, c-fos::ArchT mice (Tagged mice: n=1957, 886 and 1071 cell pairs; Control mice: n=1672, 530 and 820 cell pairs; for Baseline, Days 1-2 and Days 5-6, respectively.). ***P<0.001 compared to Baseline.
- Supplementary Figure 9: Detecting principal cell temporal assembly-patterns. (375 KB)
The different steps to identify neuronal co-activation patterns are illustrated for a recording-day with 40 simultaneously recorded principal cells. (a,b) Raster-plot of the spike-trains (a; one cell per row) with the corresponding 25ms-binned and z-scored spike-count matrix (b). For clarity, only a few second sample is shown. Rasters are color-coded relative to each neuron’s membership to the detected assembly-patterns (shown in e). (c) Correlation matrix of the binned (and z-scored) spike-count matrix. The main diagonal entries (which are all equal to 1) are black-colored for visualization. (d) The distribution of the eigenvalues obtained from the principal component analysis (PCA) is plotted with the Marčenko-Pastur threshold λmax as an analytical threshold to evaluate the number of significant patterns NA present in the data set. Note that eight eigenvalues exceed λmax. (e) The assembly-patterns extracted by the independent component analysis (ICA) based framework. For visualization purposes, the vector of each assembly-pattern is shown with those neurons whose weight exceeds two standard-deviations from the mean color-coded and the neurons ordered according to this color-code. The contribution of a neuron to the expression of an assembly-pattern is taken as its squared weight. (f) Single-neuron contribution (mean±s.e.m) of tagged and alternative neurons to assembly-patterns extracted during the first and the second three-day blocks using theta-band (4–12Hz) oscillatory waves as time windows (instead of 25ms bins; Figure 2f). *P<0.05, **P<0.01. (g) To track the activation-strength of each assembly-pattern k over time, a projector matrix Pk was constructed by taking the outer product matrix of its weight-vector and setting the diagonal to zero. The instantaneous assembly-pattern activation-strength Rk(t) was then taken as the quadratic form of the projector matrix with the convolved and z-scored spike-trains z(t). Assembly-activations were defined as peaks exceeding RTHRES = 5. (h) Average activation-strength of assembly-activations during the first and the second three-day blocks (n=75 and 86 patterns detected using 25ms time-bins, respectively; P=0.31).
- Supplementary Figure 10: Cocaine increased the subsequent re-activation of tagged CA1 pyramidal cells during drug-free context re-exposure. (378 KB)
(a,b) Schematics of the multi-transgene approach for activity-dependent neuronal tagging (a) and experimental procedure (b). Double transgenic TetTag mice (c-fos-tTA-LacZ) express the tetracycline transcriptional activator (tTA) under the control of the activity-regulated c-fos promoter. The transient removal of Dox from the diet allows neuronal activation to activate the transcriptional feedback loop (tTA*) and induce the lasting expression of the tau-LacZ protein. The c-fos-tTA-LacZ mice were injected with either saline (+S, n=4 mice) or cocaine (+C, n=7 mice) while transiently off Dox and allowed to explore an open field enclosure (“tagging procedure”; Day 0). All mice were put back on Dox to block further neuronal tagging after the initial exploration. Three days later drug-free mice were re-exposed to the same enclosure and perfused 90min later. (c) Example of cytoplasmic expression of tau-LacZ (in cyan) marking CA1 neurons that were recruited and tagged during the initial exploration of the enclosure on Day 0. The expression of the nuclear marker Fos (in magenta) was used to assess neuronal activity associated with context re-exposure on Day 3. The percentage of LacZ-positive tagged neurons that were Fos-positive (arrows), but not those that were Fos-negative (arrowheads), was used to evaluate the extent to which CA1 neurons that were initially active in one enclosure (Day 0) were subsequently active again during (drug-free) re-exposure to that enclosure (Day 3). Stratum oriens (or.), pyramidale (pyr.), radiatum (rad.). Cell nuclei stained with Dapi (in grey). Scale bar: 20µm. (d) Levels of CA1 neuronal activation during the initial exploration of an enclosure under saline or cocaine on Day 0 assessed by the percentage of LacZ-expressing CA1 neurons among the number of Dapi-positive cells. Cocaine did not significantly alter levels of initial neuronal activation compared to those related to the exploration of a saline-paired enclosure (P=0.45). (e) However, cocaine administration on Day 0 promoted the subsequent re-activation of those neurons that were initially activated during cocaine conditioning, as measured by a higher percentage of LacZ-tagged neurons that co-expressed the transient marker of neuronal activation Fos during the drug-free context re-exposure on Day 3 (P=0.04). These levels of neuronal re-activation were not correlated to those of neuronal activation during context re-exposure, as measured by the percentage of Fos-expressing neurons among Dapi cells (r=-0.04, P=0.88, Pearson’s correlation). * P<0.05.
- Supplementary Figure 11: Editing hippocampal neuronal activity prevented drug-primed reinstatement of cocaine place preference without altering spatial novelty detection. (173 KB)
(a) Mean firing rate of CA1 principal cells (n=226) recorded from mice (n=4) successively exploring one open-field paired with saline (+S) and another with cocaine (+C). Note that cocaine did not alter the mean firing rate of the CA1 principal cell population when compared to that associated with the saline-paired enclosure (P=0.77). (b) Experimental paradigm to assess both drug-primed reinstatement of cocaine place preference and spatial novelty detection. (c) Baseline (light-OFF) firing rate of tagged (n=23) and alternative (n=65) CA1 neurons from tagged c-fos::ArchT mice during re-exposure to the cocaine-paired enclosure with CA1 light-delivery (Days 1-3) and during Test 2 (Day 4). Note the shift in neuronal activities, similar to that in the experiment using a neutral enclosure (Figure 2a). This shift was however faster, which might relate to the shorter light-OFF epochs used in the CPP experiment during which the original hippocampal maps could be expressed (see Online Methods). * P<0.05, ** P<0.01: compared to Day 1. (d) Cocaine place preference was reinstated once primed by cocaine administration in control mice (n=6), but not tagged mice (n=6). Control light-OFF wild-type littermates (n=4) and control light-ON c-fos::ArchT mice (n=2) were grouped together, as their CPP scores for Pre-test, Test 1 and Test 2 were similar (P=0.50, ANOVA). Both tagged and control mice were a subset of animals used in the CPP experiment shown in Figure 3. The CPP of control mice was extinguished after Test 2 using a trial-to-criterion approach (see Online Methods). The drug-priming test was performed 24h after the extinction (“e”) for the control mice, and 24h after Test 2 for the tagged mice. Within-group comparison: *** P<0.001 for Test 1 and Test 2 compared to Pre-test, and Drug-priming compared to Extinction. Tagged compared to Controls: ### P<0.001. (e) Novel-place preference was preserved in mice with cocaine experience, including c-fos::ArchT mice which underwent CA1 hippocampal tagging and optogenetic manipulation (P=0.58). (f) Control experiment to assess the ability of mice without cocaine experience to exhibit novel-place preference on the CPP apparatus. c-fos-tTA mice (n=6) were allowed to explore the entire apparatus during three sessions (Pre-test, Test 1 and Test 2) to calculate their place preference for either the circle or the square compartment as in the cocaine CPP task. Between these three test sessions, mice were also re-exposed to both compartments separately. For Test 2, the least preferred compartment (i.e., the one which would have been paired with cocaine in the cocaine CPP task) was replaced by a novel compartment. Note that during Test 2 all mice expressed a clear spatial preference towards the compartment that was never experienced before, as a behavioral readout of their ability to detect spatial novelty on the CPP apparatus. Within-subject comparison: *** P<0.001.