ΔFosB accumulation in hippocampal granule cells drives cFos pattern separation during spatial learning

Lamothe-Molina, Paul J.; Franzelin, Andreas; Beck, Lennart; Li, Dong; Auksutat, Lea; Fieblinger, Tim; Laprell, Laura; Alhbeck, Joachim; Gee, Christine E.; Kneussel, Matthias; Engel, Andreas K.; Hilgetag, Claus C.; Morellini, Fabio; Oertner, Thomas G.

doi:10.1038/s41467-022-33947-w

Download PDF

Article
Open access
Published: 26 October 2022

ΔFosB accumulation in hippocampal granule cells drives cFos pattern separation during spatial learning

Nature Communications volume 13, Article number: 6376 (2022) Cite this article

5956 Accesses
3 Citations
36 Altmetric
Metrics details

Subjects

Abstract

Mice display signs of fear when neurons that express cFos during fear conditioning are artificially reactivated. This finding gave rise to the notion that cFos marks neurons that encode specific memories. Here we show that cFos expression patterns in the mouse dentate gyrus (DG) change dramatically from day to day in a water maze spatial learning paradigm, regardless of training level. Optogenetic inhibition of neurons that expressed cFos on the first training day affected performance days later, suggesting that these neurons continue to be important for spatial memory recall. The mechanism preventing repeated cFos expression in DG granule cells involves accumulation of ΔFosB, a long-lived splice variant of FosB. CA1 neurons, in contrast, repeatedly expressed cFos. Thus, cFos-expressing granule cells may encode new features being added to the internal representation during the last training session. This form of timestamping is thought to be required for the formation of episodic memories.

Formation of memory assemblies through the DNA-sensing TLR9 pathway

Article Open access 27 March 2024

Vladimir Jovasevic, Elizabeth M. Wood, … Jelena Radulovic

Complex activity and short-term plasticity of human cerebral organoids reciprocally connected with axons

Article Open access 10 April 2024

Tatsuya Osaki, Tomoya Duenki, … Yoshiho Ikeuchi

Neuronal activity rapidly reprograms dendritic translation via eIF4G2:uORF binding

Article Open access 08 April 2024

Ezgi Hacisuleyman, Caryn R. Hale, … Robert B. Darnell

Introduction

Adaptive decision making requires an accurate memory of past events. For instance, a mouse should revisit locations where it has found food before and avoid dangerous places. The hippocampus is well-known for processing spatial information^1,2 but it also processes other types of sensory input and organizes sequential elements of experience into episodic memories^3,4,5. Our understanding of how time is represented in the hippocampus is incomplete. The expression of many genes affecting synaptic plasticity undergoes pronounced circadian oscillations, changing the rules of synaptic plasticity depending on the time of day⁶. Theoretical and empirical work suggests that the dentate gyrus (DG) actively reduces the overlap between activity patterns from the entorhinal cortex, a process dubbed pattern separation^7,8,9. In vivo electrophysiology and calcium imaging experiments revealed that the majority of the granule cells (GCs) in the DG are silent, and of the active cells, just a small fraction show spatial tuning (place cells)^10,11,12. Spatially tuned GCs provide a stable, albeit coarse representation of the global environment across time^10,13,14,15. These findings highlight an apparent design conflict: while a perfect ‘episode encoder’ should avoid using the same neurons on consecutive days, accurate place coding is thought to require stable place cells that signal the animal’s position in a given environment. In view of this conundrum, we set out to study the impact of time and space on GCs in the dorsal DG.

To monitor and manipulate neuronal activity in freely behaving animals, reporter mice have been developed that use the immediate-early gene cFos¹⁶ to drive the expression of fluorescent proteins and optogenetic actuators¹⁷. Fear conditioning experiments with an activity-dependent expression of optogenetic silencing tools suggest that recall of a fearful episode requires reactivation of the original encoding ensemble^18,19,20. Vice versa, artificial reactivation of the encoding ensemble has been shown to reinstate a fearful state^19,21,22, suggesting that memories can be activated by specific subsets of hippocampal neurons. When freezing is used as a proxy to assess the emotional state of the animal, it is difficult to determine whether, during the optogenetic reactivation, the animal recalls the fear-inducing episode (engram) or simply feels fear. The reactivated ensemble may form an internal representation of the external world (cognitive map theory^23,24) or trigger the pattern of cortical activity that was active in the previous experience (indexing theory²⁵).

We investigated the temporal stability of DG cFos ensembles and their relation to spatial learning and the formation of cognitive maps. We used cFos-dependent tagging to assess how cFos ensemble overlap changes over training days in the Morris water maze (WM). Even in expert mice, cFos expression patterns in DG changed from day to day. In spite of these changing expression patterns, optogenetic inhibition of cFos-tagged GCs impaired navigation 5 days after tagging, suggesting that the absence of cFos expression in these GCs did not imply the absence of activity. We show in vivo and in vitro that cFos⁺ GCs accumulate ΔFosB, a long-lived splice variant of FosB. As this splice variant inhibits cFos expression²⁶, it provides a potential mechanism for the daily shift of cFos ensembles that occurs even when mice are exposed to the same environment. The changing pattern of cFos expression in DG could be the basis of an episodic memory system that write-protects synapses on the most recently used subset of GCs during the following days.

Results

Spatial learning and cFos ensemble overlap in the DG: novice vs expert mice

Although cFos expression is considered a proxy of neuronal activity^27,28,29, the relationships between action potential firing, immediate-early gene expression, and synaptic long-term plasticity are not obvious³⁰. We chose a spatial learning paradigm, the Morris WM, to investigate cFos expression patterns in DG during learning. TetTag mice were trained to find a hidden platform in the WM (Fig. 1a–c). We labeled cFos ensembles from two consecutive days (Fig. 1d). The window for labeling the first cFos ensemble was opened by taking the mouse off doxycycline (Dox), allowing for long-term expression of the fluorescent protein mKate2 fused to an opsin for membrane labeling (cFos-tagged, Fig. 1d). A short half-life version of the green fluorescent protein (shEGFP) works as an in-built cFos reporter in the TetTag mouse to label the second cFos⁺ ensemble. In calibration experiments, we observed slightly faster temporal dynamics of the native cFos protein than of the shEGFP reporter, yet nearly all shEGFP-expressing neurons also expressed cFos (Supplementary Fig. 1). The highest correlation between native cFos and shEGFP was obtained 2–4 h after stimulation. We, therefore, chose 3 h after the last training trial as the time point to sacrifice mice and fix the brains. The first cFos-tagged ensemble was revealed by immunofluorescence (IF) staining against mKate2 and the second cFos-expressing ensemble was revealed by staining against the shEGFP (cFos⁺). The ensemble sizes (proportion of GCs cFos-tagged and cFos⁺, Supplementary Fig. 2) were used to calculate the amount of overlap (GCs that are both cFos-tagged and cFos⁺) expected due to chance for each sample and compared to the actual overlap (number of cFos-tagged cFos⁺ GCs).

**Fig. 1: Behavioral performance in a spatial memory task is not reflected by cFos ensemble overlap in the DG.**

During probe trials, mice early in training (ET, Fig. 1b, c, e, f) were less precise and accurate while searching for the target but within 5 days became experts (overtrained group, OT, Fig. 1b, c, g, h). Not only the time spent searching in an annulus around the hidden platform increased (Fig. 1f, h) but also the distance to the platform was significantly lower in the expert mice (Fig. 1i). Our expectation was that, if cFos expression is linked to the activity of the neurons that are important for solving the WM task, there would be a high degree of overlap between the cFos⁺ and the cFos-tagged neurons, which were labeled 1 day earlier. We also expected overlap to increase with increasing search accuracy in the probe trials. Instead, we observed that very few GCs expressed both mKate2 (cFos-tagged) and shEGFP (cFos⁺) and that both overlap (ET 9% and OT 7% Fig. 1e, g, j) and ensemble sizes were similar in novices and expert mice (Supplementary Fig. 2). In both ET and OT mice the observed overlap was, however, significantly higher than expected due to chance (Fig. 1j, k, Supplementary Fig. 2). Surprisingly, there was no difference between the novices and experts and, assuming that cFos indicates the active GCs, less than 90% of the neurons that expressed cFos the first day were reactivated the following day. Thus, in the DG cFos overlap in GCs on consecutive days of WM training is low and independent of spatial search accuracy.

Effect of novelty on cFos ensembles: reversal training vs novel environment

We next tested whether changing the position of the escape platform (reversal training) would decrease cFos ensemble overlap. The first cFos ensemble was labeled in expert mice on day 5 of WM training (OFF Dox Fig. 2a). The following day, these mice were trained (ON-Dox) with the platform on the opposite side of the tank (reversal training, RT). Reversal training with the new platform position had no effect on cFos overlap (9%, Fig. 2a; compare with ET and OT groups, Fig. 1e, g), although it dramatically decreased search accuracy during the day 6 probe trials (Fig. 2b). The mice now searched equally in both E and W annuli, further suggesting that cFos overlap in DG GCs is a poor indicator of performance.

**Fig. 2: Effects of task and environment on cFos ensemble overlap.**

Given that after reversal training, the overlap was still higher than expected (Fig. 2f), we tested whether overlap of cFos ensembles is specific to the WM. The overlap was reduced to 4% and no longer above chance when cFos-tagged ensemble on day 1 of WM training was compared with the cFos+ ensemble generated while mice explored an open field for 20 min (Fig. 2c, d, f, novel environment NE). Likewise, cFos overlap was only 2% if mice were kept in their home cages (Fig. 2e, f home cage HC) for the 24 h OFF Dox, as in the other groups, and mice were sacrificed the following day. The cFos overlap of all groups trained in the WM (the ET, OT, and RT mice) were not significantly different from each other but significantly higher than expected due to chance whereas in both NE and HC groups, cFos overlap was at chance level and significantly lower than the WM groups (Fig. 2f, Supplementary Fig. 2). Despite ensemble sizes being the same in all conditions except the home cage, the intensity of shEGFP was higher in all groups WM trained on the second day (Fig. 2g, Supplementary Fig. 2). Thus, DG cFos expression and overlap are strongly driven in the WM regardless of training level (ET vs OT) or difficulty of the task (OT vs RT), consistent with the concept that the hippocampus is highly engaged in spatial navigation tasks and that very different environments activate non-overlapping ensembles.

cFos⁺ neurons in DG participate in spatial memory recall days later

Lesion^31,32 and optogenetic manipulation^33,34,35,36 experiments indicate that the DG plays a role in spatial memory acquisition and recall. However, those experiments did not address if particular memories are encoded by a specific subset or ensemble of neurons in the DG (engram cells). We speculated that cFos⁺ cells in the DG may encode relevant information needed to solve the WM task. To test their importance during platform search, we employed a bidirectional optogenetic tool that can be used to inhibit or excite neurons with 473 and 594 nm light, respectively (BiPOLES³⁷). We expressed BiPOLES under the control of the non-leaky TRE3G promoter (Supplementary Fig. 3) and confirmed that 473 nm light prevented action potentials in BIPOLES-expressing GCs while 594 nm light pulses repeated at 20 Hz reliably elicited single action potentials (Supplementary Fig. 4). For in vivo experiments, mice were bilaterally injected in DG with AAV_PHP.eB-TRE3G-BiPOLES-mKate2 and implanted with a custom-made tapered fiber implant³⁸ (Fig. 3a). To perform optogenetic experiments in the WM, we had to ensure that the weight of the optical fibers was compensated when mice were tethered to them via their implants. We achieved this by attaching a helium balloon to the optical fibers (Fig. 3a). Implanted mice were trained without being tethered and successfully learned the WM task (Fig. 3b). Tethering to the weight-balanced optical fibers did not affect swimming speed (Fig. 3c), nor did 473 or 594 nm light (Fig. 3d).

**Fig. 3: Optogenetic silencing of cFos-tagged granule cells impairs spatial memory recall.**

To investigate the importance of day 1 cFos-tagged neurons on memory recall, we silenced BiPOLES-expressing neurons during probe trials on days 2, 3, and 5 (Fig. 3f). Mice were placed on an elevated Atlantis platform that submerged after 30 s, forcing the mice to swim³⁹. After 60 s, a second Atlantis platform (in quadrant E) was elevated, and mice were directed towards it if they did not find it themselves. We assessed memory performance in the first half of the probe trials (30 s). Optogenetic inhibition did not affect time spent in the target quadrant on days 2 and 3 when mice were novices and their spatial accuracy was still poor (Fig. 3g). On day 5, when mice were experts and spatial accuracy was high, optogenetic inhibition significantly decreased time spent in the target quadrant (Fig. 3h, Supplementary Video 1). The mean distance to the target platform confirmed the results of the quadrant analysis, detecting significant effects of optogenetic inhibition on days 3 and 5 (Fig. 3i). We confirmed that GCs still expressed BiPOLES even 9 days after tagging (Fig. 3j). These results suggest that activity in the small ensemble of cFos-tagged neurons from the first training day (~2% of GCs) is important for successful memory recall on subsequent days. BiPOLES can also be used to increase the activity of tagged neurons (Supplementary Fig. 4). However, artificial firing (20 Hz optogenetic stimulation) during day 6 probe trials had no significant effect on WM performance in expert mice (Supplementary Fig. 5).

To address whether the activity of cFos-tagged cells is of particular importance for WM navigation or if inhibition of a random group of GCs would suffice to decrease performance, we used a chemogenetic inhibition strategy that is not dependent on cFos (Supplementary Fig. 6). The proportion of hM4Di-mCherry expressing neurons in the DG determined whether chemogenetic silencing disrupted memory recall (R² = 0.67, p = 0.003, Supplementary Fig. 6). Silencing of ~30% of GCs was required to decrease WM performance, a much higher fraction compared to cFos-dependent optogenetic silencing, suggesting that GCs expressing cFos on training day 1 remain important for successful spatial navigation in the WM many days later. If cFos-tagged cells from day 1 are indeed highly active on consecutive WM days, this activity should generate intracellular calcium transients. To generate a lasting record of calcium concentrations during behavior, we tagged cFos⁺ GCs with calcium-dependent photoconvertible CaMPARI2⁴⁰ on training day 1 and applied photoconversion light via implanted optical fibers either during all WM training trials on day 2 (WM–WM) or during the same amount of time in a novel environment (WM–NE, Fig. 4a, b). In order to estimate the volume of tissue that can be photoconverted by 405 nm light, we tagged cFos+ GCs with CaMPARI2 together with an opsin that is activated by PC light, resulting in strong photoconversion of cells up to 300 μm from the tapered fiber tip (Supplementary Fig. 7). Based on these tests, we restricted our analysis to GCs within 200 μm of the fiber tip to assess calcium levels during behavior. During WM training on day 2, photoconversion occurred in 23% of the cFos-tagged GCs, indicating a high degree of functional reactivation (Fig. 4d). This reactivation was context-specific, as only 5% cFos-tagged GCs were photoconverted during exploration of a novel environment (Fig. 4e)⁴¹. We also analyzed cFos expression on day 2 in the cFos-tagged cells. As expected from our previous experiments (Figs. 1 and 2), cFos overlap was higher in the WM-WM group (7%, Fig. 4d) than in the WM-NE group (2%, Fig. 4e). Interestingly, even in the WM-WM group, 85% of photoconverted (and thus reactivated) GCs did not express cFos again, confirming our suspicion that cFos expression is not a good proxy for GC activity.

**Fig. 4: cFos-tagged dentate granule cells are reactivated during water maze training.**

Temporal dynamics of cFos ensembles

An engram cell, by definition, is a cell that is active during both the encoding event and during memory retrieval¹⁷. Many studies used cFos expression as a surrogate marker of neuronal activity, but our WM experiments suggest that the underlying assumption may be wrong. In our experiments, cFos overlap was about 9% when mice revisited the WM arena on consecutive days (Δt = 1 days). Would these GCs continue to activate cFos on the following days of WM training? To evaluate longer time intervals, we trained the mice exactly as the RT group (highest overlap) but tagged GCs on day 1 and sacrificed the mice on day 6 (Δt = 5 days, Fig. 2a). In contrast to the RT_Δt = 1, cFos overlap in the RT_Δt = 5 was significantly lower and not different from overlap expected by chance (Fig. 5a–c, Supplementary Fig. 2). Thus, the cFos ensemble tagged in DG on the first day of WM training is not stable over multiple training days, but the neurons remain relevant for spatial memory retrieval (Figs. 3g–i, 4e).

**Fig. 5: Temporal dynamics of cFos overlap in dentate gyrus.**

We next explored the dynamics of cFos patterns in untrained mice, tagging cFos⁺ neurons in DG and CA1 in the home cage. Two weeks after injecting both regions of TetTag mice with AAV₉-TRE-mKate2, mice were taken OFF Dox for a 24 h period and either immediately sacrificed, or given an i.p. Dox injection to rapidly close the tagging window and sacrificed after 6, 12, or 24 h (Fig. 5d). Overlap between cFos-tagged (mKate2⁺) and cFos⁺ (shEGFP⁺) neurons was highest at the end of the OFF Dox window in both DG (59%) and CA1 (56%, Fig. 5e, f). During the next 24 h, the overlap dropped to 2% in DG but only to 27% in CA1, indicating different temporal dynamics of cFos expression in the two areas (Fig. 5e, f). In the DG the overlap at Δ0h was 12-fold higher than expected due to chance and fell below-chance level within 24 h (Fig. 4g). CA1 cFos ensembles were larger (Supplementary Fig. 2), increasing the expected overlap due to chance. At Δ0h overlap in CA1 was significantly higher than expected by chance but within 6 h was at chance levels (Fig. 4g). The very linear drop of ensemble similarity in DG was initially surprising (Fig. 4g), but could be readily reproduced in a numerical simulation of a negative feedback loop, including the below-chance overlap at Δt = 24 h (Fig. 4h). As the shEGFP reporter is more stable than cFos itself, endogenous cFos expression might change somewhat faster. These results indicate that in a stable environment and without training (i.e., home cage), the default mechanism is for DG GCs to shift cFos expression to a completely new ensemble every day.

cFos-expressing GCs form spatial clusters that segregate over time

The sparse activity of GCs is controlled by robust inhibitory inputs from hilar interneurons⁴². We were interested in the spatial patterns of cFos-expressing GCs, which may not be randomly distributed, but could instead reflect the sphere of influence of individual interneurons. We used the maps generated in the home cage experiments (Fig. 5d) to analyze clusters of cFos⁺ neurons at the different time points (Fig. 5i, j). As a measure of cluster overlap I_c, we divided the overlapping area by the union of both areas (see methods). While at Δt = 0, clusters were partially overlapping (high I_c), cluster overlap gradually decreased with time to very low values at Δt = 24 h (Fig. 5k). This result was not simply a consequence of the decreasing number of double-positive neurons because randomizing the positions of single- and double-positive neurons within the GC layer yielded clusters with higher overlap and less segregation over time. We conclude that the process of temporal segregation we describe here not only plays out in individual GCs, but also prevents the re-use of DG regions that were highly active 24 h ago.

GCs possess an intrinsic cFos blocking mechanism

The decorrelation of cFos ensembles in DG could reflect changes in synaptic input (i.e. network-level effects), membrane excitability⁴³, or a cFos shut-off mechanism in individual GCs, e.g. by transcriptional or translational repression of cFos. A candidate suppressor of cFos transcription is ΔFosB, a long-lived splice variant of FosB that accumulates in active neurons^26,44,45,46. To find out whether ΔFosB accumulates during WM training, we analyzed the DG of mice that were trained for 7 days in the WM, using a pan-FosB antibody that recognizes both FosB and ΔFosB (pan-FosB, Fig. 6a). GCs had either high cFos expression or high pan-FosB, but rarely both, resulting in an inverse correlation that was highly significant in data pooled from 3 mice and in individual mice (Fig. 6b, Supplementary Fig. 8). When considering only GCs that were cFos-tagged on day 1 of WM training (magenta) we observed high levels of pan-FosB on day 7 but low cFos (Fig. 6c, d). Using the same overlap analysis as in previous experiments, cFos overlap was significantly below chance in this ∆6-day cohort (Fig. 6e) whereas overlap between day 1 cFos-tagged and day 7 pan-FosB⁺ cells was 8-fold higher than expected by chance (Fig. 6f). Thus, cFos expression in GCs that previously expressed cFos in the WM appears to be inhibited, possibly by the FosB variant ∆FosB. The high levels of pan-FosB staining suggest that indeed the cFos-tagged neurons were repeatedly activated during the subsequent WM training, giving ample time for ΔFosB to accumulate in this specific population.

**Fig. 6: Mechanism of cFos suppression in WM-trained mice.**

To test whether cFos suppression is a cell-autonomous process or whether it reflects a lack of synaptic input, we used a chemogenetic approach. We expressed the Gq-DREADD to directly activate neurons and trigger cFos expression in rat hippocampal slice cultures^47,48. In DG, the first clozapine-N-oxide (CNO) application induced cFos in 59% of the DREADD-expressing GCs (Fig. 7a, Supplementary Fig. 9). A second CNO application 24 h later induced cFos in a much smaller fraction (17%). In CA1 pyramidal cells, conversely, the first and the second CNO application were equally effective in inducing cFos (Fig. 7b). To our surprise, we detected high levels of pan-FosB in CA1 neurons on the second day, which apparently did not interfere with their ability to express cFos. The pan-FosB AB detects both FosB and ΔFosB, but only the latter splice variant is thought to suppress cFos⁴⁴. Thus, while we could recapitulate GC-specific cFos inhibition in these slice experiments, we failed to specifically label the inhibiting agent.

**Fig. 7: Repeated chemogenetic induction of cFos expression in slice culture.**

To visualize ΔFosB protein levels in individual neurons, we subtracted from the pan-FosB fluorescence the fluorescence signal from a second antibody that recognizes the C-terminus of FosB, which is absent in ΔFosB (Supplementary Figure 10). In hippocampal slice cultures, we detected significantly more ΔFosB in DG neurons than in CA1 neurons, in both stimulated and non-stimulated cultures. This area difference was even more pronounced in mice after 7 days of WM training, where ΔFosB accumulated in many GCs, but not in CA1 pyramidal cells (Supplementary Figure 10).

Spatial learning increases the number of GCs in a cFos-repressed state

If accumulated ΔFosB inhibits cFos in GCs, concentrations of the two transcription factors should be inversely correlated. To test this, we compared home-caged mice with mice that were WM trained on two (spaced training) or on seven days (daily training). As anticipated, WM training increased the number of ΔFosB⁺ GC (Fig. 8a) as well as the average intensity of ΔFosB⁺ expression (Fig. 8b). In individual GCs, cFos expression was inversely correlated with ΔFosB, and the strength of the correlation in individual mice increased with WM training (Fig. 8c). We generated kernel density estimates (KDE) of the actual data (Fig. 8d) and 100 KDEs of scrambled data to simulate the null hypothesis (no interaction between ΔFosB and cFos). Compared to the expectation of the null hypothesis, GCs were overrepresented in two regions: Low cFos with high ΔFosB intensity, which we termed the repressed state (Fig. 8e, upper left quadrant), and high cFos with low ΔFosB, which we denote the permissive state (lower right quadrant). In home-cage mice, only 20% of analyzed GC were in the repressed state (Fig. 8f). The repressed fraction increased to 32% after spaced WM training and to 44% after daily training. Thus, training increased the number of cFos-repressed GCs, which was consistent across animals (see Supplementary Fig. 11 for more examples). Importantly, only GCs that expressed cFos or ΔFosB were included in this analysis. GCs that expressed neither marker were presumably not very active in the WM. In the CA1 region of the same animals we rarely detected neurons with ΔFosB signal, even after daily WM training (Supplementary Fig. 12). This suggests that regardless of training level, almost all CA1 pyramidal cells remained in the permissive state.

**Fig. 8: Water maze training leads to accumulation of ΔFosB and cFos repression in DG.**

Discussion

How time is represented in the brain is still a matter of debate. Using a multi-day training paradigm, we discovered a time-dependent shift in cFos expression patterns that seems to be specific to GCs of the DG. We observed that in individual GCs, accumulation of ∆FosB during water maze training was negatively correlated with cFos expression. Suppression of cFos transcription by ∆FosB has previously been described under pathological conditions (AD models, seizures)⁴⁴, and interfering with ∆FosB in hippocampal neurons is known to affect learning and memory²⁶. Here we show that even in the home cage, ∆FosB suppression of cFos ensures that the cFos⁺ ensemble of GCs changes from day to day. Given the extremely high stability of place cells in DG¹⁰, we expected to find stable cFos⁺ ensembles during multi-day training in a consistent environment, which was not the case. Two questions arise: What does cFos expression tell us about neuronal activity, and second, are cFos⁺ GCs place cells?

While cFos is known to be driven by high-frequency spiking^27,49,50, the reverse is not true in the DG: The absence of cFos is not evidence for the absence of activity. This became clear when we inhibited the activity of the original (day 1) cFos-tagged ensemble on consecutive days: WM performance was compromised in trials with optogenetic inhibition, even though a very small fraction of GCs in dorsal DG (~1%) was silenced (Fig. 3). The inhibition of cFos-tagged neurons was specific, as inhibition of random GCs was much less effective: we observed behavioral effects only in animals where >30% of GCs were inhibited. Thus, GCs that expressed cFos on the first training day remained particularly relevant for successful navigation in that environment, even though they rarely re-expressed cFos on the successive training days. We confirmed this by showing that cFos-tagged GCs are functionally reactivated when mice revisited the WM, but were not reactivated in a novel environment (Fig. 4). Therefore, low cFos overlap in DG is not in contradiction with the very high stability of place cells in DG¹⁰. Our findings do, however, indicate a problem when using cFos to identify engram cells. A defining criterion of engram cells stipulates that reactivation—presumably electric—in the same context must occur^17,19,51. As cFos-tagged neurons in DG remain cFos^- when electrically reactivated on subsequent days (Figs. 4 and 5), using cFos as a proxy for electrical activity is problematic. Defining only a few GCs that express cFos twice as the engram cells would strongly underestimate the number of neurons relevant for memory recall. The DG cFos⁺ map produced on a given day will mostly contain GCs that have been activated for the first time in the WM. These cFos⁺ GCs may encode new features being added to the internal representation, possibly acting as a timestamp on the most recently acquired memory. Our optogenetic inhibition experiments suggest that the full set of GCs required to recall the platform position includes many cFos-negative cells.

Why does electrical reactivation of GCs in a behavioral task rarely trigger renewed cFos expression? In slice cultures, we could reproduce this cFos refractory period by repeated chemogenetic stimulation of GCs. CA1 pyramidal cells, in contrast, readily express cFos on successive days. In search of the repressive mechanism, we turned to the FosB splice variant ΔFosB, which is known to be expressed in GCs and to inhibit the expression of cFos in these neurons^26,44. We quantified the amount of ΔFosB in individual neurons, revealing the effect of WM training on the GC population: With repeated training, more and more GCs shifted from a low ΔFosB state that is permissive for cFos expression to a high ΔFosB state, in which cFos is repressed. Behavioral training did not have the same effect on CA1 neurons, the large majority of which remained in the permissive state even after 7 days of daily training. Interestingly, the few ΔFosB-positive CA1 neurons we found were all cFos-negative, suggesting that the suppressive mechanism is functional in pyramidal cells, but little ΔFosB is produced during behavior. In vivo imaging experiments have indeed shown that in 60% of the cFos-expressing pyramidal neurons in CA1⁴⁷ and 80% in barrel cortex³⁰, cFos expression is not transient, but sustained for several days. We likewise observed that CA1 neurons express cFos rather persistently, with 27% of homcage-tagged neurons continuing to express cFos 24 h after the mice were back on doxycycline, compared to just 2% of GCs. Active suppression also explains the perfectly linear drop in cFos levels over time we observed in GCs (Fig. 5g) which could not be fitted by simple (exponential) decay functions.

A new picture of time-controlled cFos regulation in the DG begins to emerge. Only 19% of cFos-tagged GCs in the DG expressed cFos 12 h later, and after 24 h, the overlap dropped below-chance level. When neurons express cFos they are hyperexcitable^29,52,53, whereas overexpression of ΔFosB reduces excitability⁵⁴. Hyperexcitability increases the likelihood of burst firing, creating ideal conditions for the potentiation of incoming synapses⁵⁵. Burst firing also maximizes the impact of GCs on postsynaptic CA3 neurons, as the GC-CA3 mossy fiber synapses display extremely strong short-term facilitation^56,57. During the high cFos period, the probability of long-term potentiation may therefore be increased at both GC input and output synapses, i.e., onto CA3 pyramidal cells and interneurons targeted by the current set of cFos⁺ GCs⁵⁸. In contrast, synaptic plasticity may be reduced during the refractory period when ΔFosB is high, cFos expression is inhibited and excitability is reduced, possibly write-protecting synapses on high ΔFosB neurons from further modification. In nucleus accumbens, this regulatory mechanism has been proposed to underlie the switch from casual drug use to cocaine addiction⁵⁹. In the amygdala, oscillations of CREB/ICER have been suggested as a mechanism to group memories in time⁶⁰.

The assumption that cFos is an indicator of highly active (engram) neurons may hold for naive mice housed under standardized (deprived) conditions and exposed to fear conditioning or other forms of one-trial learning^61,62. Multi-day training, which may be slightly closer to the daily challenges faced by mice in the wild, leads to the accumulation of ΔFosB in many GCs and potentially other brain areas⁶³. Our results suggest that under these conditions, cFos patterns in DG depend not only on input from the entorhinal cortex, but are also gated by the activity history of each GC. Investigation of memory systems “under load” during complex behaviors⁶⁴ may thus change our view of hippocampal processing and engram formation.

Methods

Experimental animals

B6.Cg-Tg^{(Fos-tTA,Fos-EGFP*)1Mmay/J} (TetTag) mice were obtained from the Jackson Laboratory (Strain #018306) and bred to wildtype (noncarrier) C57BL6/J mice from our colony. Mice were group-housed with littermates until 2 weeks before rAAV injections, then were single-caged. Mice had access to food and water ad libitum and were kept in an animal facility next to the behavioral rooms on a reversed light-dark cycle (dark 7 am–7 pm) at 20–23 °C with 45–65% humidity. All behavioral experiments were done during the dark phase of the cycle. Due to the requirement to swim with optical fibers, only male mice between 20–40 weeks (>28 g by the time of surgery) were used for optogenetic WM experiments (Figs. 3–5). Both male and female mice were included in the cFos ensemble overlap experiments (Figs. 1, 2 and 5). All experiments were conducted in accordance with German law and European Union directives on the protection of animals used for scientific purposes and were approved by the local authorities of the City of Hamburg (Behörde für Justiz und Verbraucherschutz, Lebensmittelsicherheit und Veterinärwesen, N 100/15 and N 046/2021).

Viral constructs

To label the membrane of cFos⁺ neurons, we used a red fluorescent protein fused to an opsin⁶⁵ (iChloC-linker-mKate2). This construct was inserted into a pAAV-TREtight backbone using MluI and EcoRI restriction enzymes to produce pAAV-TREtight-iChloC-mKate2. To drive spiking in cFos neurons, iChloC was replaced with CheRiff⁶⁶ to create pAAV-TREtight-CheRiff-mKate2. To inhibit cFos⁺ neurons, we used pAAV-TRE3G-BiPOLES-mKate2 (Addgene # 192579). Constructs were packaged into AAV₉ (iChloC-mKate2) or into AAV_PHP.eB (BiPOLES, CheRiff) by the UKE Vector Facility. To photoconvert cFos⁺ cells with high calcium levels, iChloC-mKate was replaced with CaMPARI2⁴⁰. AAVs for the DREADD experiments were obtained from Addgene (AAV_AAV-CaMKIIa-h4MD(Gi)-mCherry, Addgene # 50476-AAV₉; AAV₉-CaMKIIa-hM3Dq-mCherry, Addgene # 50477-AAV₉).

Stereotactic injection and fiber implant

TetTag mice were virus-injected under analgesia and anesthesia using a stereotaxic drill and injection robot (Neurostar). Mice were fixed to the frame under isoflurane anesthesia (1.5% mixed in O₂), skin and connective tissue was removed, and two craniotomies were performed using an automated drill on the desired coordinates. AAVs were injected at 10¹² vg/ml concentration, except for iChloC-mKate2 experiments where it was 10¹³ vg/ml. AAVs were delivered bilaterally into the dorsal hippocampus using a glass micropipette attached to a 5 μl syringe (Hamilton). A single injection per site was performed using stereotaxic coordinates for DG (−2.2 AP, ± 1.37 ML, −1.9 DV) with a volume of 500 nl on each side (injection speed: 100 nl/min). CA1 was also injected in a set of experiments (Fig. 5) by moving to −1.4 DV and injecting 400 nl after the DG injection. After the last injection, the pipette was retracted 200 µm and left for at least 5 min to minimize efflux of virus during withdrawal. After the injections, the bone surface was cleaned with 0.9% NaCl solution and the skin was stitched. To avoid hypothermia, a heated pad was placed under the animal during surgery and under its cage for 1 h until full recovery. We provided post-surgery analgesia with Meloxicam mixed with softened Dox-food (see below) for 3 days after surgery. Animals recovered at least 2 weeks before behavioral experiments. Mice for optogenetic experiments were implanted with a custom-made bilateral tapered tip optic fiber implant (Doric) targeting the DG sulcus (−2.2 AP, ±1.37 ML, −1.7 DV) right after AAV injection. Implants were attached to the skull using dental cement (C&B Metabond). A protective cap made of an Eppendorf tube was secured by applying acrylic resin (Pattern Resin LS, GC America) to the exposed skull around the implant.

Doxycycline treatment

Animals were given doxycycline-containing food (Altromin-Dox, 50 mg per kg of body weight, red pellets). To tag the first cFos ensemble, animals were changed to doxycycline-free food (Altromin, light-brown pellets) 24 h before exposure to the task. To ensure that the animals did not eat any Dox-food crumbles that had fallen into their cages, they were moved into new cages with fresh bedding. The old nesting material was transferred to the new cage to decrease novelty. Dox food was resupplied exactly 24 h after removal, right before the behavioral task. For cFos ensemble temporal shift experiments (Fig. 4), doxycycline (50 µg/g body weight) was injected I.P. after the end of the 24 h OFF Dox period.

Behavioral experiments

For ensemble overlap experiments, mice were injected in batches and randomly assigned to an experimental group. Optogenetic experiments had a crossover design where all mice had probe trials with and without light, to minimize sampling errors and reduce the number of animals required to reach statistical power (paired statistics). Mice that did not have adequate viral transduction (see image analysis section), had off-target implants, or performed poorly (floaters or implant intolerance) were excluded from the analysis. All experiments were recorded on digital video. Ethovision XT 11.5 was used for automated tracking.

Water Maze

Animals were handled for 1 week before the start of the pre-training sessions to reduce stress during behavioral tasks. Pre-training: Mice were pre-trained for 2 days before their first exposure to the WM arena. Sessions (3-4 trials of max. 60 s each on two days) were done in a small rectangular water tank in the dark, in the same room where the WM task was performed. Water level was 1 cm above the 14 cm diameter escape platform. The position of the platform was alternated between the left and right side of the tank between trials, keeping a distance of 5 cm from the walls to avoid thigmotaxis. Once the animals found the platform, a grid was presented until the animals climbed onto it and were returned to their home cages in the waiting area of the behavioral room. Training: The WM consisted of a circular tank (1.45 m diameter) with visual asymmetrical landmarks, filled with water with (non-toxic) white paint. A platform (submerged by 1 cm) was placed in the center of the east quadrant during regular training and switched to the opposite (west) quadrant for reversal training (max. swim time: 90 s). To test spatial reference memory, a probe trial (PT) without a platform was performed on each day. Mice underwent six trials every day (4 training trials (TTs, 90 s) + 1 PT (60 s) + 1 TT), inter-trial interval (ITI) was 8–10 s. For the TTs, mice were lowered into the tank facing the wall in different, pseudo-randomized positions (avoiding the target quadrant). In PTs, mice were lowered in the center of the tank. An opaque cup-sized chamber attached to a pole was used to transfer the mice from their home cage to the drop position and a plastic grid attached to a pole was used to pick up the mice. Mice were picked up 10 s after they found the platform and were returned to their home cage. Mice that did not find the platform during the TT were guided to it using the grid and were picked up after a 10 s on-platform waiting period. For both pre-training and WM, water temperature was 19–21 °C and a heat lamp was placed over the waiting area to prevent hypothermia.

Home cage and open field control experiments

To evaluate the temporal stability of cFos ensembles, mice were left unperturbed in their home cage (HC) during the OFF Dox period and sacrificed after 0, 6, 12, or 24 h ON-Dox. For all the cFos overlap experiments, both cFos⁺ tagging events were designed to have the same amount of trials and training length. For the open field experiments, mice were placed inside a square arena (50 × 50 cm, 50 lux) for 20 min and stayed in the behavioral room at the same time as their WM counterparts.

Optogenetic silencing in the water maze

Mice were connected to two thin optic fibers (200 μm core diameter, 2 m, Doric Lenses, Canada) using ceramic ferrules (Doric) and put back in their home cage for at least 5 min before any behavioral task. In pre-training trials (2 days, three trials per day), mice were connected and disconnected before every trial to habituate them to the tether. To avoid stress-related cFos tagging, mice were never tethered during the Dox-OFF period. In all trials with tether, the weight of the fibers was compensated with a white helium balloon (0.08 N pull force), attached to a light fiber ~30 cm above the mouse with a transparent thin plastic tubing, long enough to keep the balloon out of the field of view of the video camera. Optogenetic silencing occurred only during memory recall (PTs, no target platform). Blue (473 nm) or yellow (594 nm) light was delivered using a laser combiner (LightHUB, Omicron) connected to a commutator (FRJ_1x2i_FC-2FC_0.22 Doric Lenses, Canada) to split the output into two fibers. The commutator was placed close to the ceiling, at the same height as the camera. Blue light delivery was started when mice were still in their HC (473 nm light pulsed at 33 Hz, 10 ms pulses, 8–10 mW). Immediately after light on, mice were placed on a starting platform facing the wall on the south quadrant. The starting platform was submerged with a hydraulic mechanism (Atlantis platform) after 30 s. After 60 s of swim time, the target platform (east quadrant) was raised to just below the water level. Mice that found the target platform was transferred to their HC 10 s later. In cases where mice did not find the platform they were guided to it and then transferred to their HC. Blue light was on for a total of 2 min, covering the entire WM trial. “No light” probe trials were done exactly the same, but with the blue laser off.

Training protocol

Two batches of mice were used for optogenetic manipulation in the water maze (Fig. 3). On training day 1, the protocol consisted of six trials: four TTs of max. 90 s with the platform in quadrant E, followed by a PT (60 s, without platform) and one more TT. On day 2, mice were placed in the starting Atlantis position for 90 s (visual recall trial). Afterwards, they performed two PTs while blue light was being delivered in the same crossover design, followed by three TTs. Day 3 started with two TTs, followed by two PTs (crossover design, blue light pulses), and two more TTs. On day 4 mice underwent the standard protocol (four TTs, one PT without light, one TT). Day 5 training, mice had two TTs followed by two PTs (blue light, crossover design). We repeated this protocol on day 6 with yellow light pulses during PTs. This batch was perfused for immunohistochemistry on day 9. The second batch was trained similarly. After tagging (training day 1) mice went through the same training protocol from day 2 to day 4, adding a yellow light trial before the first TT. However, we found that yellow light had no effect on behavior, cFos expression, or cFos overlap. Consequently, we pooled data from “yellow light” and “no yellow light” animals to increase statistical power (Fig. 3h–j). On day 5, mice had two TTs, followed by two PTs, where half of the mice received blue light for the duration of the trial. On day 6, mice had two more TTs and a PT. This batch of mice was perfused for immunohistochemistry on day 7, 90 min after the last PT.

Chemogenetic silencing in the WM

Wild-type mice were bilaterally injected in DG with AAV9-h4MD(Gi)-mCherry. After at least 2 weeks of recovery, mice were handled and pre-trained like the previous groups. On day 1, mice were trained for 6 trials. On day 2, mice had two PTs. 40 min before the 1st PT, they received an intraperitoneal (I.P.) injection of 0.9% NaCl solution (vehicle). Mice were then injected with clozapine-N-oxide (CNO, 5 mg kg⁻¹) in 0.9% NaCl solution. After 40 min, their performance was tested again (2nd PT). From days 3–7, mice were trained every day with 4–6 TTs and 1 PT. On day 8, mice were injected again with CNO 40 min before the PT.

CaMPARI2 conversion experiments

TetTag mice were injected bilaterally with AAV_PHP.eB-TRE-CaMPARI2 in DG (1.2 × 10¹²). Tapered fiber stubs were implanted bilaterally like in the optogenetic experiments. All mice were trained Off-Dox in the WM for 1 day. The following day (On-Dox), one group of mice were again trained in the WM. Photocoverting (PC) light (405 nm) was delivered throughout each WM trial in the ET group and the total illumination time was measured. For each WM-WM animal, we prepared a yoked control animal which, on the second day, received the same dose of PC light (same intensity, same duration) while exploring a novel environment (open field). 90 min after the last trial, mice were sacrificed for immunostaining (cFos-GFP, red CaMPARI2).

Ex vivo brain processing and IF staining

For cFos overlap analysis, mice were perfused 3 h after the last trial. In optogenetic silencing experiments, mice were perfused 1.5 h after the last trial. Mice were injected with ketamine/xylazine (100/10 mg kg⁻¹) intraperitoneal and intracardially perfused with 1x phosphate-buffered saline (PBS, Sigma) followed by 4% paraformaldehyde (PFA, Roth). Brains were extracted and stored in 4% PFA for at least 24 h at 4 °C. Before sectioning, brains were washed with 1× PBS for 20 min at room temperature (RT). The dorsal hippocampal region (−1.2 to −2.3 AP from bregma) was cut into 40–50 µm coronal sections using a vibratome (Leica VT100S) and collected in PBS. From each series, six sections were selected from Bregma −1.7 to −2.3 mm and incubated in blocking buffer (1× PBS, 0,3% TritonX, 5% goat serum) for 2 h at RT. Next, sections were placed in the primary antibody carrier solution (1× PBS, 0,3% TritonX, 1% goat serum, 1% BSA) and incubated overnight at 4 °C. After 3× washing with 1× PBS for 5 min, the sections were incubated for 2 h at RT in the secondary antibody carrier solution (1× PBS, 0.3% TritonX, 5% goat serum). Sections were washed 3 × 10 min in 1× PBS, stained with 4′,6-diamidino-2-phenylindole (DAPI, 1:1000) for 5 min and mounted on coverslips using Immu-Mount (Shandon). Complete antibody information in Supplementary Table 2.

Confocal imaging

Mounted brain slices were labeled with a code to blind the analyst. For cFos overlap experiments, immunostained slices were imaged with a confocal microscope (Olympus Fluoview FV 1000) using an oil immersion objective (UPLSAPO ×20/0.85). From each of the six slices per brain (left and right DG), 15 µm were imaged (stack of 10 images at 1024 × 1024 resolution, Z-step: 1.5 µm). Slices from BiPOLES-expressing animals were imaged with a Zeiss LSM 900 (Plan-Apochromat 20×/0.8). From each section, 18 µm were imaged (stack of six images at 1024 × 1024 resolution, Z-step: 3 µm). Excitation/emission filters were selected using the dye selection function of the Fluoview/Zen3.5 software (Alexa 405 (DAPI), Alexa 488 (shEGFP) and Alexa 568 (mKate2), Alexa 647 (cFos or FosB). The image acquisition settings were optimized once and kept constant for all images within an experimental data set. Images were not deconvolved or filtered for quantitative analysis. Exemplary images presented in the figures are median-filtered (2 × 2 kernel) and cropped. Each color channel was linearly adjusted (Image J).

Ensemble size calculation and overlap analysis

Confocal image stacks (pseudo-colored cyan: shEGFP; pseudo-colored magenta: mKate2; pseudo-colored gray: DAPI) were analyzed with Imaris (Oxford Instruments). The volume of the upper and lower blade of DG (granule cell layer, GCL) based on the DAPI channel was used to estimate the total GC number based on published cell densities⁶⁷ and to mask the cyan and the magenta channel to restrict analysis to the GCL. Automatic spot detection was used to identify shEGFP⁺ cells in the cyan channel. The quality filter (round nuclei with a diameter of 8 µm) was adjusted once and then applied to every image stack of the same experiment. False positive spots (e.g., staining artifacts) were manually removed. It was not possible to detect mKate2-expressing GCs automatically as only the plasma membrane was labeled. To count mKate2-positive cells, spots were placed manually (spot size 12 µm) using the pseudo-colored magenta channel only. Double-positive cells (distance between shEGFP⁺ and mKate2⁺ spots <5 µm) were identified using a Matlab script. They were then manually inspected to check for artifacts and, if necessary, corrected. From the Imaris analysis, we calculated the following quantities:

$${(1)}\quad{{{{{\rm{Number}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{granule}}}}}}\,{{{{{\rm{cells}}}}}}\,=\frac{{{{{{\rm{DAPI}}}}}}\,{{{{{\rm{surface}}}}}}\,{{{{{\rm{volume}}}}}}}{{{{{{\rm{reported}}}}}}\,{{{{{\rm{GC}}}}}}\,{{{{{\rm{density}}}}}}}$$

$${(2)}\quad{{{{{\rm{Fraction}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{shEGFP}}}}}}\,{{{{{\rm{tagged}}}}}}\,{{{{{\rm{cells}}}}}}\,=\frac{{{{{{\rm{number}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{shEGFP}}}}}}+\,{{{{{\rm{cells}}}}}}}{{{{{{\rm{number}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{GCs}}}}}}}$$

$${(3)}\quad{{{{{\rm{Fraction}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{mKate2}}}}}}\,{{{{{\rm{tagged}}}}}}\,{{{{{\rm{cells}}}}}}\,=\frac{{{{{{\rm{number}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{mKate2}}}}}}+\,{{{{{\rm{cells}}}}}}}{{{{{{\rm{number}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{GCs}}}}}}}$$

$${(4)}\quad{{{{{\rm{Fraction}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{double}}}}}}\,{{{{{\rm{positive}}}}}}\,{{{{{\rm{cells}}}}}}\,=\frac{{{{{{\rm{number}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{double}}}}}}\,{{{{{\rm{positive}}}}}}\,{{{{{\rm{cells}}}}}}}{{{{{{\rm{number}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{GCs}}}}}}}$$

$${(5)}\quad{{{{{\rm{expected}}}}}}\,{{{{{\rm{overlap}}}}}}\,=\,{{{{{\rm{fraction}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{shEGFP}}}}}}\,{{{{{\rm{cells}}}}}}*\,{{{{{\rm{fraction}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{mKate2}}}}}}\,{{{{{\rm{cells}}}}}}$$

$${(6)}\quad{{{{{\rm{Overlap}}}}}}/{{{{{\rm{chance}}}}}}\,=\frac{{{{{{\rm{fraction}}}}}}\,{{{{{\rm{of}}}}}}\,{{{{{\rm{double}}}}}}\,{{{{{\rm{positive}}}}}}\,{{{{{\rm{cells}}}}}}}{{{{{{\rm{expected}}}}}}\,{{{{{\rm{overlap}}}}}}}$$

Cluster analysis

Neurons were clustered by the Agglomerative Clustering algorithm implemented in the scikit-learn python library. For each image, we first computed the connection graph by using kneighbors_graph, which is also implemented in scikit-learn. The number of neighbors was fixed as one-tenth of the total number of neurons. First (mKate⁺) and second set (shEGFP⁺) of cFos-positive neurons were clustered separately (linkage criterion: “ward”). Clustering was performed for 16 different numbers of clusters ${n}_{c}\in [{{{{\mathrm{5,\,20}}}}}]$. To determine the optimal number of clusters (${n}_{c}$) for each image, we generated 500 sets of artificial data from each image by randomly distributing the cFos⁺ neurons within the cell body layer of DG. Double-positive neurons were treated as a third type, keeping their numbers identical to the experimental data. The cell body layer area was defined by the DAPI channel, and the size of each neuron was standardized (16*16 pixels). Each set of artificial data was clustered with the same procedure as the empirical data. For each clustering result, a clustering index

$${(7)}\quad{C}_{c}={\left\langle \frac{{{{{{{{\rm{median}}}}}}}({D}_{{ij}})}_{j\in {{{{{\rm{cluster}}}}}}}}{{{{\min }}({D}_{{ij}})}_{j\notin {{{{{\rm{cluster}}}}}}}}\right\rangle}_{i\in U}$$

was calculated, where ${D}_{{ij}}$ is the distance between neurons ${i}$ and j. Smaller ${c}_{c}$ indicates stronger clustering. For each ${n}_{c}$, we calculated the proportion (of the 500 artificial sets) where ${c}_{c}$ was smaller in the artificial data than in the empirical data. The optimal ${n}_{c}$ was defined as the one with the smallest proportion, i.e., the strongest clustering compared to randomly distributed cells. In the case of multiple ${n}_{c}$ with the same smallest proportion, the largest ${n}_{c}$was selected. The optimal ${n}_{c}$ was determined for mKate2⁺ and shEGFP⁺ clusters separately (always including the double-positive cells).

The sphere of influence of a cluster was defined as the convex hull of all its neurons, calculated by using ConvexHull implemented in SciPy python library. The overlap index between mKate2⁺ clusters ${M}_{i}$ and shEGFP⁺ clusters ${S}_{j}$ was defined as

$${(8)}\quad {I}_{c}=({M}_{i}\cap {S}_{j})/({M}_{i}\cup {S}_{j}).$$

In case of overlap between multiple clusters, the largest overlap was scored. Another 200 sets of artificial data were generated from each image and analyzed in an identical fashion to create cumulative distributions of ${I}_{c}$ values expected for randomly distributed cFos⁺ cells (Fig. 5h).

Analysis of CaMPARI2 photoconversion

The antibody combinations (primary/secondary) used were Anti-CaMPARI2-Red (4F61)/rabbit-568 and cFos/rat-647. Due to the high scattering of 405 nm light by the brain tissue, only slices where the implant site was visible were analyzed (200–250 µm from the implant side, Supplementary Fig. 7). The DAPI channel was used to create a surface of the GCL to estimate the total number of GCs. 12 µm spots (Imaris) were manually placed on converted (red) and non-converted (green) CaMPARI2⁺ GCs. Spots were exported to Matlab, the fluorescence intensities of all three channels were extracted, calcium-dependent conversion of each CaMPARI2⁺ cell was calculated (R/R+G) and correlated with its cFos intensity.

Hippocampal slice culture

TetTag mice (male and female) or Wistar Unilever rats (Female, Envigo, HsdCpb:WU) were prepared at postnatal day 4–8⁶⁸. Briefly, animals were anesthetized with 80% CO₂ 20% O₂ and decapitated. Hippocampi were dissected in cold dissection medium containing (in mM): 248 sucrose, 26 NaHCO₃, 10 glucose, 4 KCl, 5 MgCl₂, 1 CaCl₂, 2 kynurenic acid, 0.001% phenol red (310–320 mOsm kg⁻¹, saturated with 95% O₂, 5% CO₂, pH 7.4). Tissue was cut into 400 μM thick sections on a tissue chopper and cultured on membranes (Millipore PICMORG50) at 37 °C in 5% CO₂. No antibiotics were added to the slice culture medium which was partially exchanged (60–70%) twice per week and contained 394 ml Minimal Essential Medium (for 500 ml), 20% horse serum, 1 mM L-glutamine, 0.01 mg ml⁻¹ insulin, 1.45 ml 5 M NaCl, 2 mM MgSO₄, 1.44 mM CaCl₂, 0.00125% ascorbic acid, 13 mM D-glucose.

cFos expression induced by chemogenetic stimulation

Organotypic rat slice cultures were injected with AAV9-CaMKIIa-hM3Dq-mCherry at DIV 14 using a Picospritzer III (Parker Hannafin). Virus was injected into DG, CA3, and CA1 (1.8 bar pulses, 50 ms duration). After 4–5 days of expression, slices were stimulated either once or twice by applying a 10 µl drop of CNO (1 µM, Tocris) in buffer (in mM: 145 NaCl, 10 HEPES, 25 D-glucose, 1.25 NaH₂PO₄, 2.5 KCl, 1 MgCl, 2 CaCl₂, 0.01 TTX, pH 7.4) on top of the slice 24 h apart. Culture inserts were transferred to a new well containing only the buffer for 3 min to wash off the CNO and then returned to the incubator in culture medium. In twice-stimulated cultures, the first stimulation was done without TTX to avoid homeostatic plasticity. As controls, we included slices that were not treated (NT group) or treated with TTX-containing buffer without CNO (vehicle). All slices were fixed 70 min after the last treatment and stained against cFos and pan-FosB. Analysis was performed blind. DREADD-expressing cells were randomly selected based on their mCherry signal (20 spots for DG, 50 spots for CA1 per slice), nuclear cFos and pan-FosB fluorescence intensity was extracted.

Acute slice preparation and electrophysiology

Mice were decapitated under CO₂ anesthesia and the brains were dissected. Acute coronal slices (300 µm) were cut on a vibratome (Leica VT1000 S) in cold (4 °C) cutting solution containing (in mM): choline-chloride (110), KCl (2.5), NaH₂PO₄ (1.25), NaHCO₃ (25), MgCl₂ (7), CaCl₂ (0.5), glucose (25), sodium ascorbate (11.6), sodium pyruvate (3.1). Oxygenation and pH (7.4.) were maintained by bubbling (95% O₂, 5% CO₂). Slices were kept in a holding chamber with artificial cerebrospinal fluid (aCSF) containing (in mM): NaCl (125), KCl (2.5), NaH₂PO₄ (1.25), NaHCO₃ (26), MgCl₂ (1), CaCl₂ (2), glucose (10), saturated with 95% O₂, 5% CO₂. Brain slices recovered at 34 °C for at least 45 min before the start of whole-cell patch-clamp recordings (30−31 °C). Patch pipettes were pulled from borosilicate glass and had a resistance of 3–5 MΩ when filled with internal solution containing (in mM): K-gluconate (135), HEPES (10), MgCl₂ (4), Na₂-ATP (4), Na-GTP (0.4), Na₂-phosphocreatine (10), L-ascorbic acid (3), EGTA (0.2). Internal solution had pH 7.2 and 295 mOsm/L. The mKate2-positive cells in the DG were patched and signals acquired through an Axopatch 200B or Multiclamp 700B (Axon Instruments, Inc.), National Instruments A/D boards, and Matlab running Ephus software⁶⁹. Action potential firing was electrically evoked by somatic current injection. Optogenetic stimulation was given through the objective (Olympus, ×60, 1.0 NA) at 473 or 594 nm (CoolLED, pE-4000). Data were analyzed with Matlab or Clampfit 10.7 (Molecular Devices).

Statistical analysis

Numerical data from individual experiments were collected and ensemble sizes were calculated in MS Excel (version 16.0). Statistical tests were conducted in GraphPad Prism (version 8) and assumed an alpha level of 0.05. To analyze differences, we used one and two-way analysis of variance and the Šídák method to correct for multiple comparisons. In some cases, we used two-sided t tests. A detailed description of experimental groups and statistical tests is provided in Supplementary Table 1. For all figures, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

Source data are provided as a Source Data file. Other data are available upon request from the corresponding authors. Source data are provided with this paper.

Code availability

Custom code to generate kernel density estimates in Matlab and to simulate cFos suppression is provided via GitHub (https://github.com/toertner/Kernel-density-estimate).

References

Bird, C. M. & Burgess, N. The hippocampus and memory: insights from spatial processing. Nat. Rev. Neurosci. 9, 182–194 (2008).
Article CAS PubMed Google Scholar
Morellini, F. Spatial memory tasks in rodents: what do they model? Cell Tissue Res. 354, 273–286 (2013).
Article PubMed Google Scholar
Aronov, D., Nevers, R. & Tank, D. W. Mapping of a non-spatial dimension by the hippocampal–entorhinal circuit. Nature 543, 719–722 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Eichenbaum, H. Time cells in the hippocampus: a new dimension for mapping memories. Nat. Rev. Neurosci. 15, 732–744 (2014).
Article CAS PubMed PubMed Central Google Scholar
Mau, W. et al. The same hippocampal CA1 population simultaneously codes temporal information over multiple timescales. Curr. Biol. 28, 1499–1508.e4 (2018).
Article CAS PubMed PubMed Central Google Scholar
McCauley, J. P. et al. Circadian modulation of neurons and astrocytes controls synaptic plasticity in hippocampal area CA1. Cell Rep. 33, 108255 (2020).
Article CAS PubMed PubMed Central Google Scholar
Deng, W., Mayford, M. & Gage, F. H. Selection of distinct populations of dentate granule cells in response to inputs as a mechanism for pattern separation in mice. Elife 2, e00312 (2013).
Article PubMed PubMed Central Google Scholar
Leutgeb, J. K., Leutgeb, S., Moser, M.-B. & Moser, E. I. Pattern separation in the dentate gyrus and CA3 of the hippocampus. Science 315, 961–966 (2007).
Article ADS CAS PubMed Google Scholar
Chavlis, S. & Poirazi, P. Pattern separation in the hippocampus through the eyes of computational modeling. Synapse 71, e21972 (2017).
Hainmueller, T. & Bartos, M. Parallel emergence of stable and dynamic memory engrams in the hippocampus. Nature 558, 292–296 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Danielson, N. B. et al. In vivo imaging of dentate gyrus mossy cells in behaving mice. Neuron 93, 552–559.e4 (2017).
Article CAS PubMed PubMed Central Google Scholar
Pofahl, M. et al. Synchronous activity patterns in the dentate gyrus during immobility. Elife 10, e65786 (2021).
GoodSmith, D. et al. Spatial representations of granule cells and mossy cells of the dentate gyrus. Neuron 93, 677–690.e5 (2017).
Article CAS PubMed PubMed Central Google Scholar
Senzai, Y. & Buzsáki, G. Physiological properties and behavioral correlates of hippocampal granule cells and mossy cells. Neuron 93, 691–704.e5 (2017).
Article CAS PubMed PubMed Central Google Scholar
van Dijk, M. T. & Fenton, A. A. On how the dentate gyrus contributes to memory discrimination. Neuron 98, 832–845.e5 (2018).
Article PubMed PubMed Central Google Scholar
Reijmers, L. G., Perkins, B. L., Matsuo, N. & Mayford, M. Localization of a stable neural correlate of associative memory. Science 317, 1230–1233 (2007).
Article ADS CAS PubMed Google Scholar
Josselyn, S. A. & Tonegawa, S. Memory engrams: recalling the past and imagining the future. Science 367, eaaw4325 (2020).
Denny, C. A. et al. Hippocampal memory traces are differentially modulated by experience, time, and adult neurogenesis. Neuron 83, 189–201 (2014).
Article CAS PubMed PubMed Central Google Scholar
Lacagnina, A. F. et al. Distinct hippocampal engrams control extinction and relapse of fear memory. Nat. Neurosci. 22, 753–761 (2019).
Article CAS PubMed PubMed Central Google Scholar
Trouche, S. et al. Recoding a cocaine-place memory engram to a neutral engram in the hippocampus. Nat. Neurosci. 19, 564–567 (2016).
Article CAS PubMed PubMed Central Google Scholar
Liu, X. et al. Optogenetic stimulation of a hippocampal engram activates fear memory recall. Nature 484, 381–385 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Perusini, J. N. et al. Optogenetic stimulation of dentate gyrus engrams restores memory in Alzheimer’s disease mice. Hippocampus 27, 1110–1122 (2017).
Article CAS PubMed PubMed Central Google Scholar
Moser, E. I., Moser, M.-B. & McNaughton, B. L. Spatial representation in the hippocampal formation: a history. Nat. Neurosci. 20, 1448–1464 (2017).
Article CAS PubMed Google Scholar
O’keefe, J. & Nadel, L. The hippocampus as a cognitive map. (Oxford university press, 1978).
Tanaka, K. Z. & McHugh, T. J. The hippocampal engram as a memory index. J. Exp. Neurosci. 12, 1179069518815942 (2018).
Article PubMed PubMed Central Google Scholar
Eagle, A. L. et al. Experience-dependent induction of hippocampal ΔFosB controls learning. J. Neurosci. 35, 13773–13783 (2015).
Article CAS PubMed PubMed Central Google Scholar
Sheng, H. Z., Fields, R. D. & Nelson, P. G. Specific regulation of immediate early genes by patterned neuronal activity. J. Neurosci. Res. 35, 459–467 (1993).
Article CAS PubMed Google Scholar
Trojanowski, N. F., Bottorff, J. & Turrigiano, G. G. Activity labeling in vivo using CaMPARI2 reveals intrinsic and synaptic differences between neurons with high and low firing rate set points. Neuron 109, 663–676.e5 (2021).
Article CAS PubMed Google Scholar
Yassin, L. et al. An embedded subnetwork of highly active neurons in the neocortex. Neuron 68, 1043–1050 (2010).
Article CAS PubMed PubMed Central Google Scholar
Lee, J. et al. FosGFP expression does not capture a sensory learning-related engram in superficial layers of mouse barrel cortex. Proc. Natl. Acad. Sci. USA 118, e2112212118 (2021).
Lee, I. & Kesner, R. P. Encoding versus retrieval of spatial memory: double dissociation between the dentate gyrus and the perforant path inputs into CA3 in the dorsal hippocampus. Hippocampus 14, 66–76 (2004).
Article PubMed Google Scholar
Clark, R. E., Broadbent, N. J. & Squire, L. R. Impaired remote spatial memory after hippocampal lesions despite extensive training beginning early in life. Hippocampus 15, 340–346 (2005).
Article PubMed PubMed Central Google Scholar
Qin, H. et al. A visual-cue-dependent memory circuit for place navigation. Neuron 99, 47–55.e4 (2018).
Article CAS PubMed PubMed Central Google Scholar
Masachs, N. et al. The temporal origin of dentate granule neurons dictates their role in spatial memory. Mol. Psychiatry 26, 7130–7140 (2021).
Gu, Y. et al. Optical controlling reveals time-dependent roles for adult-born dentate granule cells. Nat. Neurosci. 15, 1700–1706 (2012).
Article CAS PubMed PubMed Central Google Scholar
Andrews-Zwilling, Y. et al. Hilar GABAergic interneuron activity controls spatial learning and memory retrieval. PLoS One 7, e40555 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Vierock, J. et al. BiPOLES is an optogenetic tool developed for bidirectional dual-color control of neurons. Nat. Commun. 12, 4527 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Vogt, N. Tailoring optogenetic illumination through tapered fibers. Nat. Methods 14, 763–763 (2017).
Article CAS Google Scholar
Spooner, R. I., Thomson, A., Hall, J., Morris, R. G. & Salter, S. H. The Atlantis platform: a new design and further developments of Buresova’s on-demand platform for the water maze. Learn. Mem. 1, 203–211 (1994).
Article CAS PubMed Google Scholar
Moeyaert, B. et al. Improved methods for marking active neuron populations. Nat. Commun. 9, 4440 (2018).
Article ADS PubMed PubMed Central Google Scholar
Tanaka, K. Z. et al. The hippocampal engram maps experience but not place. Science 361, 392–397 (2018).
Article ADS CAS PubMed Google Scholar
Yuan, M. et al. Somatostatin-positive interneurons in the dentate gyrus of mice provide local- and long-range septal synaptic inhibition. Elife 6, e21105 (2017).
Shen, Y. et al. CCR5 closes the temporal window for memory linking. Nature 606, 146–152 (2022).
Article ADS CAS PubMed Google Scholar
Corbett, B. F. et al. ΔFosB regulates gene expression and cognitive dysfunction in a mouse model of Alzheimer’s disease. Cell Rep. 20, 344–355 (2017).
Article CAS PubMed PubMed Central Google Scholar
Nestler, E. J. Transcriptional mechanisms of drug addiction. Clin. Psychopharmacol. Neurosci. 10, 136–143 (2012).
Article CAS PubMed PubMed Central Google Scholar
Gajewski, P. A. et al. Epigenetic regulation of hippocampal fosb expression controls behavioral responses to cocaine. J. Neurosci. 39, 8305–8314 (2019).
Article CAS PubMed PubMed Central Google Scholar
Poll, S. et al. Memory trace interference impairs recall in a mouse model of Alzheimer’s disease. Nat. Neurosci. (2020) https://doi.org/10.1038/s41593-020-0652-4.
Mizuno, N. & Itoh, H. Functions and regulatory mechanisms of Gq-signaling pathways. Neurosignals 17, 42–54 (2009).
Article CAS PubMed Google Scholar
Tyssowski, K. M. et al. Different neuronal activity patterns induce different gene expression programs. Neuron 98, 530–546.e11 (2018).
Article CAS PubMed PubMed Central Google Scholar
Schoenenberger, P., Gerosa, D. & Oertner, T. G. Temporal control of immediate early gene induction by light. PLoS One 4, e8185 (2009).
Article ADS PubMed PubMed Central Google Scholar
Khalaf, O. et al. Reactivation of recall-induced neurons contributes to remote fear memory attenuation. Science 360, 1239–1242 (2018).
Article ADS CAS PubMed Google Scholar
Pignatelli, M. et al. Engram cell excitability state determines the efficacy of memory retrieval. Neuron 101, 274–284.e5 (2019).
Article CAS PubMed Google Scholar
Sun, X. et al. Functionally distinct neuronal ensembles within the memory engram. Cell 181, 410–423 (2020).
Article CAS PubMed PubMed Central Google Scholar
Eagle, A. L., Williams, E. S., Beatty, J. A., Cox, C. L. & Robison, A. J. ΔFosb decreases excitability of dorsal hippocampal CA1 Neurons. eNeuro 5, ENEURO.0104-18.2018 (2018).
Wittenberg, G. M. & Wang, S. S.-H. Malleability of spike-timing-dependent plasticity at the CA3-CA1 synapse. J. Neurosci. 26, 6610–6617 (2006).
Article CAS PubMed PubMed Central Google Scholar
Vyleta, N. P., Borges-Merjane, C. & Jonas, P. Plasticity-dependent, full detonation at hippocampal mossy fiber-CA3 pyramidal neuron synapses. Elife 5, e17977 (2016).
Pernía-Andrade, A. J. & Jonas, P. Theta-gamma-modulated synaptic currents in hippocampal granule cells in vivo define a mechanism for network oscillations. Neuron 81, 140–152 (2014).
Article PubMed PubMed Central Google Scholar
Guo, N. et al. Dentate granule cell recruitment of feedforward inhibition governs engram maintenance and remote memory generalization. Nat. Med. 24, 438–449 (2018).
Article CAS PubMed PubMed Central Google Scholar
Cates, H. M., Lardner, C. K., Bagot, R. C., Neve, R. L. & Nestler, E. J. Fosb induction in nucleus accumbens by cocaine is regulated by E2F3a. eNeuro 6 (2019).
Rashid, A. J. et al. Competition between engrams influences fear memory formation and recall. Science 353, 383–387 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Cai, D. J. et al. A shared neural ensemble links distinct contextual memories encoded close in time. Nature 534, 115–118 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Tonegawa, S., Morrissey, M. D. & Kitamura, T. The role of engram cells in the systems consolidation of memory. Nat. Rev. Neurosci. 19, 485–498 (2018).
Article CAS PubMed Google Scholar
Eagle, A. L. et al. Circuit-specific hippocampal ΔFosB underlies resilience to stress-induced social avoidance. Nat. Commun. 11, 4484 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Rosenberg, M., Zhang, T., Perona, P. & Meister, M. Mice in a labyrinth show rapid learning, sudden insight, and efficient exploration. Elife 10, e66175 (2021).
Wietek, J. et al. An improved chloride-conducting channelrhodopsin for light-induced inhibition of neuronal activity in vivo. Sci. Rep. 5, 14807 (2015).
Article ADS CAS PubMed PubMed Central Google Scholar
Hochbaum, D. R. et al. All-optical electrophysiology in mammalian neurons using engineered microbial rhodopsins. Nat. Methods 11, 825–833 (2014).
Article CAS PubMed PubMed Central Google Scholar
Jinno, S. & Kosaka, T. Stereological estimation of numerical densities of glutamatergic principal neurons in the mouse hippocampus. Hippocampus 20, 829–840 (2010).
PubMed Google Scholar
Gee, C. E., Ohmert, I., Wiegert, J. S. & Oertner, T. G. Preparation of Slice Cultures from Rodent Hippocampus. Cold Spring Harb. Protoc. 2017 (2017).
Suter, B. A. et al. Ephus: multipurpose data acquisition software for neuroscience experiments. Front. Neurosci. 4, 53 (2010). G.
Google Scholar
Reference Atlas:: Allen Brain Atlas: Mouse Brain. https://mouse.brain-map.org/static/atlas. Accessed 30 July 2020.

Download references

Acknowledgements

We would like to thank Iris Ohmert, Jan Schröder, Katryn Sauter, Sabine Graf, Chantal-Joy Agbo, Tanja Stößner, and Karen Kesseler for excellent technical assistance. The pAAV-TRE3G-BiPOLES-mKate2 construct (Addgene #192579) was a gift from J. Simon Wiegert. Ingke Braren from the UKE Vector Facility produced rAAV. We thank Mary Muhia, Irm Hermans-Borgmeyer, Silvia Rodriguez-Rozada, Brenna Fearey, and Sabine Hoffmeister-Ullerich for their discussions and helpful advice. This study was supported by the Deutsche Forschungsgemeinschaft (DFG) through Collaborative Research Center CRC 936 (#178316478) Projects B7 (F.M., T.G.O.), A2 (A.K.E.) and Z3 (C.C.H.), SFB TRR169-A2 (C.C.H., D.L.), Research Unit FOR 2419 (#278170285, MK; #282803474, CEG, TGO), and by Consejo Nacional para la Ciencia y Tecnología (CONACYT, Mexico, P.L.M).

Funding

Open Access funding enabled and organized by Projekt DEAL.

Author information

These authors contributed equally: Paul J. Lamothe-Molina, Andreas Franzelin.
These authors jointly supervised this work: Fabio Morellini, Thomas G. Oertner.

Authors and Affiliations

Institute for Synaptic Physiology, Center for Molecular Neurobiology (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany
Paul J. Lamothe-Molina, Andreas Franzelin, Lennart Beck, Tim Fieblinger, Laura Laprell, Christine E. Gee & Thomas G. Oertner
Institute of Computational Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
Dong Li & Claus C. Hilgetag
Research Group Behavioral Biology, Center for Molecular Neurobiology (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany
Lea Auksutat & Fabio Morellini
Department of Neurophysiology and Pathophysiology, Center for Experimental Medicine (ZEM), University Medical Center Hamburg-Eppendorf, Hamburg, Germany
Joachim Alhbeck & Andreas K. Engel
Institute for Molecular Neurogenetics, Center for Molecular Neurobiology (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany
Matthias Kneussel

Authors

Paul J. Lamothe-Molina
View author publications
You can also search for this author in PubMed Google Scholar
Andreas Franzelin
View author publications
You can also search for this author in PubMed Google Scholar
Lennart Beck
View author publications
You can also search for this author in PubMed Google Scholar
Dong Li
View author publications
You can also search for this author in PubMed Google Scholar
Lea Auksutat
View author publications
You can also search for this author in PubMed Google Scholar
Tim Fieblinger
View author publications
You can also search for this author in PubMed Google Scholar
Laura Laprell
View author publications
You can also search for this author in PubMed Google Scholar
Joachim Alhbeck
View author publications
You can also search for this author in PubMed Google Scholar
Christine E. Gee
View author publications
You can also search for this author in PubMed Google Scholar
Matthias Kneussel
View author publications
You can also search for this author in PubMed Google Scholar
Andreas K. Engel
View author publications
You can also search for this author in PubMed Google Scholar
Claus C. Hilgetag
View author publications
You can also search for this author in PubMed Google Scholar
Fabio Morellini
View author publications
You can also search for this author in PubMed Google Scholar
Thomas G. Oertner
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

P.L.M., A.F., A.K.E., F.M., and T.G.O. conceived and designed experiments. P.L.M., A.F., L.B., and L.A. performed behavioral experiments. A.F., T.F., J.A., L.L., and L.B. performed electrophysiological experiments. D.L. and C.C.H. applied cluster analysis. A.F. and C.E.G. performed slice culture experiments. P.L.M., L.B., L.A., and F.M. analyzed behavioral data. Confocal imaging, immunohistochemistry, and cFos scoring were done by A.F. and L.B. P.L.M., M.K., A.K.E., C.C.H., F.M., and G.O. obtained funding and provided infrastructure. P.L.M., C.E.G., F.M., and T.G.O. wrote the paper. All authors edited and approved the manuscript.

Corresponding authors

Correspondence to Paul J. Lamothe-Molina or Thomas G. Oertner.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Steve Ramirez and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Movie 1

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Lamothe-Molina, P.J., Franzelin, A., Beck, L. et al. ΔFosB accumulation in hippocampal granule cells drives cFos pattern separation during spatial learning. Nat Commun 13, 6376 (2022). https://doi.org/10.1038/s41467-022-33947-w

Download citation

Received: 09 September 2020
Accepted: 07 October 2022
Published: 26 October 2022
DOI: https://doi.org/10.1038/s41467-022-33947-w

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.