Dg–ca3 circuitry mediates hippocampal representations of latent information

ABSTRACT Survival in complex environments necessitates a flexible navigation system that incorporates memory of recent behavior and associations. Yet, how the hippocampal spatial circuit

represents latent information independent of sensory inputs and future goals has not been determined. To address this, we image the activity of large ensembles in subregion CA1 via

wide-field fluorescent microscopy during a novel behavioral paradigm. Our results demonstrate that latent information is represented through reliable firing rate changes during unconstrained

navigation. We then hypothesize that the representation of latent information in CA1 is mediated by pattern separation/completion processes instantiated upstream within the dentate gyrus

(DG) and CA3 subregions. Indeed, CA3 ensemble recordings reveal an analogous code for latent information. Moreover, selective chemogenetic inactivation of DG–CA3 circuitry completely and

reversibly abolishes the CA1 representation of latent information. These results reveal a causal and specific role of DG–CA3 circuitry in the maintenance of latent information within the

hippocampus. SIMILAR CONTENT BEING VIEWED BY OTHERS LINKING HIPPOCAMPAL MULTIPLEXED TUNING, HEBBIAN PLASTICITY AND NAVIGATION Article 20 October 2021 THE REPRESENTATION OF CONTEXT IN MOUSE

HIPPOCAMPUS IS PRESERVED DESPITE NEURAL DRIFT Article Open access 03 May 2022 REMAPPING REVISITED: HOW THE HIPPOCAMPUS REPRESENTS DIFFERENT SPACES Article 07 May 2024 INTRODUCTION

Hippocampal subregions CA1 and CA3 represent space through the coordinated activity of place cells—sparsely active cells tuned to different preferred locations that tile the navigable space,

collectively forming a hippocampal map specific to each context1. Changes to sensory cues or cognitive demands within a context can lead to widespread changes in the firing rates of place

cells2,3,4,5,6,7,8, a type of hippocampal pattern separation known as rate remapping5. Previous work has demonstrated that different sources of information and circuit mechanisms are capable

of driving rate remapping: sensory-driven remapping5 is mediated in part by both lateral entorhinal9 and trisynaptic10 circuits, while goal-oriented remapping4,7,8 is mediated by a

prefrontal-thalamic circuit6 which is distinct from the lateral entorhinal and trisynaptic circuits. However, a flexible navigational system should encode aspects of the current context

beyond immediate sensory input and future goals. Indeed, latent information, such as a memory of recent behavior or experiences independent of future goals, can be especially important for

discovering and representing relationships that extend beyond the capacity of immediate sensory information. Yet, whether and how the hippocampus represents latent information, as well as

the neural circuitry that maintains these representations in the absence of continuous sensory information or goal-directed behavior, has not been determined. Here we show that hippocampal

subregions CA1 and CA3 represent latent information through changes in firing rate during unconstrained 2D navigation in a novel behavioral paradigm in the absence of explicit task demands.

Next, we demonstrate that chemogenetic inactivation of DG–CA3 reversibly abolishes this representation. Together, these results demonstrate a causal and specific role of DG–CA3 circuitry in

the representation of latent information during unconstrained navigation. RESULTS PARADIGM AND NEURAL RECORDING We first developed a simple behavioral paradigm to characterize whether and

how latent information is represented during unconstrained 2D navigation. To this end, we designed a single-compartment arena accessible via two entryways connected by a short hallway (Fig. 

1a). Previous results have demonstrated that point of entry can modulate hippocampal representations across different compartments11,12,13. In our paradigm, entry to the single compartment

through either doorway is associated with a distinct recent history including the animal’s spatial location, direction of running within the hallway, and the turns made as the mouse enters

the larger 2D compartment. Once inside the compartment, however, the spatial position and sensory cues available to the animal are identical regardless of the prior point of entry; thus any

remapping within the room can be attributed to latent information. We imaged the calcium dynamics of large populations of CA1 neurons using a miniaturized head-mounted wide-field microscope

(miniscope.org) as mice freely explored this environment for 20 min14,15 (Fig. 1b; _n_ = 6 mice, one mouse excluded for persistent entryway behavioral bias, _n_ = 5 analyzed, 29 sessions;

Supplementary Fig. 1, Supplementary Table 1). Following motion correction16, cells were segmented and calcium traces were extracted via constrained nonnegative matrix factorization17,18

(Supplementary Figs. 2, 3). The likelihood of spiking events, a correlate of firing rate19, was inferred from deconvolution of the filtered calcium traces via a second-order autoregressive

model (Fig. 1c). All further analyses were conducted on the inferred likelihood of spiking events. Next, we extracted times when the mouse was in the compartment and partitioned these data

according to both the most recent entryway and session half. From these data, we identified place cells as cells which demonstrated reliable spatial correlations across session halves either

after entering from the same entryway or different entryways (correlations greater than the 95th percentile relative to a shuffled control, corrected for multiple comparisons, see “Methods”

section; Fig. 1d, Supplementary Figs. 4, 5). Importantly, because biases in spatial sampling can be correlated with the most recent entryway, we subsampled the data to match the spatial

sampling distributions across entryways and halves prior to all analyses (Supplementary Figs. 6, 7; see “Methods” section). THE CA1 RATE CODE REPRESENTS MOST RECENT ENTRYWAY If point of

entry drives remapping within an environment, then maps of the compartment should be more similar across session halves when the mouse entered from the same entryway as opposed to different

entryways. To quantify representational similarity, we computed the mean split-half correlation of the place cell population vector following entry through either the same entryways (PVs) or

different entryways (PVd). This analysis revealed that, at a population level, the CA1 compartment map was more similar when the mouse entered through the same entryway than when he entered

through different entryways (Fig. 1e, Supplementary Fig. 8). Further analysis revealed that these population-level differences reflected reliable differences in place cell firing rates

(Fig. 1f, Supplementary Fig. 9) in the absence of substantial changes in preferred locations (Supplementary Fig. 10), consistent with rate remapping5. The orientation of the CA1 map was not

modulated by entryway (Supplementary Fig. 10). Remapping was robust when selecting place cells via other criteria (Supplementary Fig. 11), most pronounced in the place cell population

(Supplementary Fig. 12), and robust to other quantifications of calcium transients (Supplementary Fig. 13). The magnitude of entryway remapping was not predicted by behavioral or

signal-to-noise (SNR) covariates (Supplementary Fig. 14). Entryway remapping effects were also observed without matching sampling distributions prior to computing comparisons (Supplementary

Fig. 15). Similar results were observed in a larger environment in a separate cohort of mice (Supplementary Fig. 16). Together these results demonstrate coding of recent history through

remapping of the CA1 rate code during unconstrained 2D navigation. For completeness, the CA1 map of the hallway also remapped according to point of entry (Supplementary Fig. 17), as expected

given the typical observation of direction-selective coding in linear track environments20. THE CA3 RATE CODE REPRESENTS MOST RECENT ENTRYWAY Previous work has demonstrated that the

trisynaptic circuit partially mediates remapping of the CA1 rate code when immediate sensory cues are available10. We hypothesized that CA1 coding of latent variables may also be driven by

trisynaptic input. If so, similar remapping should also be observed in upstream CA3. To test this possibility, we repeated our experiment and analyses while recording place cells from right

CA3 (_n_ = 4 mice, 32 sessions; Fig. 2a, Supplementary Figs. 1–6; Supplementary Table 1). Computing [PVs − PVd] revealed that the CA3 compartment map was more similar when the mouse entered

through the same entryway than when he entered through a different entryway (Fig. 2b, Supplementary Fig. 8). Again, this difference reflected reliable differences in place cell firing rates

(Fig. 2c, Supplementary Fig. 9) without substantial changes in preferred locations (Supplementary Fig. 10). The orientation of the CA3 map was not modulated by entryway (Supplementary Fig. 

10). Remapping was robust when selecting place cells via other criteria (Supplementary Fig. 11), most pronounced in the place cell population (Supplementary Fig. 12), and robust to other

quantifications of calcium transients (Supplementary Fig. 13). The magnitude of entryway remapping was not predicted by behavioral or signal-to-noise (SNR) covariates (Supplementary Fig. 

14). Entryway remapping effects were also observed without matching sampling distributions prior to computing comparisons (Supplementary Fig. 15). Together these results indicate reliable

remapping of the CA3 rate code on the basis of the most recent point of entry, similar to that of the CA1 place code. For completeness, the CA3 map of the hallway also remapped according to

point of entry (Supplementary Fig. 17), as expected given the typical observation of direction-selective coding in linear track environments20. CODING IS ENTRY-SPECIFIC AND PERSISTS ACROSS

TIME AND SPACE Other latent variables, such as distant prior experience or planned future behavior, may also be represented via remapping of the CA1 and/or CA3 rate codes. Indeed, in

goal-oriented tasks both retrospective and prospective rate coding have been observed throughout the broader hippocampal formation7,8. We tested this possibility in our data by repeating our

analysis when segmenting the behavioral trajectory according to the second-most recent entryway into the compartment (Two-back) and the exitway (Fig. 3a). In both CA1 and CA3, only

segmentation by the most recent entryway yielded reliable representation. Together this suggests that recent prior experience, but not distant experience or planned future behavior, is coded

during navigation in the absence of task demands. If the representational differences we observe reflect the sustained influence of latent entryway information, then these differences

should persist across locations within the compartment and across behaviorally-relevant timescales. To test these possibilities, we first computed entryway [PVs − PVd] separately at each

location in the compartment (Fig. 3b). Reliable differences in map similarity were observed across the compartment in both the CA1 and CA3. The magnitude of remapping was negatively

correlated with the distance to the nearest entryway in CA1 (Pearson’s correlation, _r_ = −0.279, _p_ = 4.920e–4) but not CA3 (_r_ = 0.009, _p_ = 0.917). The magnitude of remapping was not

correlated with differences in median speed across locations (Pearson’s correlation, CA1: _r_ = 0.127, _p_ = 0.139; CA3: _r_ = 0.046, _p_ = 0.595). Next, we again computed [PVs − PVd] over

the entire compartment, while including data only from progressively later times since the mouse entered the compartment. This analysis revealed that reliable remapping was observed when

including only data recorded at least 5 s since entering the compartment in both CA1 and CA3 (Fig. 3c, Supplementary Fig. 6), well beyond the timescales of both neural activity and calcium

reporter dynamics. Together these results indicate that the rate remapping we observed in CA1 and CA3 reflects a representation of the most recent entryway which persists across both space

and time. INACTIVATION OF DG–CA3 ABOLISHES ENTRY CODING IN CA1 If CA1 rate remapping in this paradigm is driven by trisynaptic input, then disruption of the trisynaptic circuit should

eliminate such remapping. If, on the other hand, remapping of the CA1 and CA3 place codes is driven by common input originating outside of the hippocampus, then inhibition of the trisynaptic

circuit should spare this remapping. To causally adjudicate between these possibilities, we repeated our experiment while recording from right CA1 and simultaneously manipulating the

trisynaptic circuit in a new cohort of mice. To this end, we employed a Grik4-cre mouse line which expresses cre in both the dentate gyrus and CA3, but not CA121. These regions were

transfected bilaterally with the cre-mediated inhibitory designer receptor hm4di22 (_n_ = 3 mice; Fig. 4a, Supplementary Fig. 18). On interleaved trials, either clozapine-N-oxide (CNO), the

designer drug which activates hm4di, or saline, the vehicle control, was administered intraperitoneally one hour and fifteen minutes prior to recording (19 Saline sessions, 19 CNO sessions,

Supplementary Figs. 1–4; Supplementary Table 1). An additional control cohort not expressing hm4di was also recorded with the same paradigm (_n_ = 3 mice, 14 CNO sessions, 13 saline

sessions, Supplementary Figs. 1–4; Supplementary Table 1). Inhibition of the trisynaptic circuit eliminated remapping of the CA1 place code by entryway as assayed by multiple measures (Fig. 

4b, Supplementary Fig. 19). Computing [PVs − PVd] revealed that the CA1 compartment map was more similar when the mouse entered through the same entryway than when he entered through a

different entryway in all conditions except for the CNO with hm4di condition (Fig. 4c, Supplementary Fig. 8). Furthermore, [PVs − PVd] of the CNO with hm4di condition was reliably decreased

relative to all other conditions and was not significantly larger than zero. Importantly, PVs remained equally high in all conditions (Wilcoxon rank-sum tests: Zs < 1.0920, ps > 

0.274), suggesting that the lack of remapping during CNO with hm4di sessions was not driven by a disruption of the CA1 place code more generally (Fig. 4c). Consistent with this, other coding

properties such as mean and peak firing rates, spatial information content, and overall split-half reliability did not reliably differ during Saline versus CNO sessions for with hm4di mice

(Supplementary Fig. 20). At an individual cell level, mean firing rates were reliably more similar when entering through same entryway than different entryways in all conditions except for

the CNO with hm4di condition (Fig. 4d, Supplementary Fig. 9). These effects were qualitatively robust when selecting place cells via other criteria (Supplementary Fig. 11), most pronounced

in the place cell population (Supplementary Fig. 12), and robust to other quantifications of calcium transients (Supplementary Fig. 13). The magnitude of these effects was not predicted by

behavioral or signal-to-noise (SNR) covariates (Supplementary Fig. 14), and similar results were observed without matching sampling distributions prior to computing comparisons

(Supplementary Fig. 15). Moreover, the elimination of entryway remapping in mice with hm4di under CNO persisted across time and sessions (Supplementary Fig. 21). Remapping of the hallway

place code persisted across all conditions (Supplementary Fig. 22). DISCUSSION Characterizing the neural basis of navigation has been a major focus of research since the discovery of

hippocampal place cells; yet a mechanistic explanation of how hippocampal spatial circuits represent information beyond sensory inputs and goal-oriented behavior has been limited, likely

hindered by difficulties assaying latent information isolated from a behavioral task. Here we used a novel behavioral paradigm aimed to examine how the hippocampus encodes latent information

in a 2D environment. We first report that latent information, in the form of the most recent entryway in a multientry compartment, is robustly encoded by rate changes in the CA1 place cell

representation in the absence of explicit task demands. Building on proposals that CA3 recurrent connections are theoretically capable of maintaining information in the absence of continuous

input via attractor dynamics, and that the DG is particularly well-suited to disambiguate similar input representations via pattern separation, we hypothesized the DG–CA3 circuit mediated

the representation of latent information in this paradigm23,24,25. Recordings in CA3 revealed that this circuit also maintained an analogous code for latent information, indicating that such

a role was plausible. Finally, to causally test this hypothesis, we used a transgenic mouse line and chemogenetics to reversibly inactivate the DG–CA3 circuit while recording from region

CA1. These experiments revealed that the representation of latent information in CA1 is completely dependent on CA3–DG activity in this paradigm. Importantly, the spatial tuning of the CA1

representation remained intact during DG–CA3 inactivation, offering support to the hypothesis that the key role of the trisynaptic circuit (EC–DG–CA3–CA1) may not be spatial coding per se

(similar to the results of ref. 26), but rather the encoding of additional information beyond a navigator’s immediate spatial location. Our observation of entryway coding in the absence of

task demands prompts further questions about the specific source and ethological relevance of such information. One possibility is that the remapping we observe in our paradigm may be driven

by the stereotyped behavior of entering from the hallway, where navigation is largely constrained. Prior work has demonstrated that direction of entry is a powerful disambiguating cue

capable of modulating the similarity of hippocampal maps across repeated connected compartments11,12,13. Moreover, we observed strong remapping within the hallway, and the strength of this

remapping was generally correlated with the extent of entryway remapping in the compartment across experiments. Thus, each entry was preceded by a unique neural state space. This suggests

that entryway remapping may follow from reinforcement learning models, where transitions in hippocampal state space are governed by repeated experience with the navigational structure of the

environment27, and that DG–CA3 circuitry may be critical to maintaining this information in the absence of immediate sensory cues. At an ethological level, entryway remapping, and

representation of latent variables during unconstrained navigation more generally, may be crucial for flexibly discovering and representing contingencies during navigation. For example, one

could imagine that trajectories originating from one burrow through an open space may be associated with potential predators, while entry from an alternate borrow may subsume less risk.

Finally, we note that repeated experience may also influence the degree of entryway remapping. In our data, entryway remapping persevered for up to 21 sessions of recording and following

repeated exposure to a fixed map in the case of mice with hm4di under CNO. Nevertheless, experience on longer timescales may serve to stabilize the compartment map. At a theoretical level,

our results contribute to a growing literature demonstrating that hippocampal rate remapping can be driven by multiple sources, potentially mediated by distinct circuit mechanisms. Previous

studies have demonstrated that rate remapping can be driven by sensory inputs5 and that such representation partially depends on both the lateral entorhinal cortex9 and NDMA receptors in the

DG10, though on average these studies report a 20% and 50% reduction in rate remapping, respectively. Importantly, such paradigms are not designed to rule out a latent/mnemonic contribution

to rate remapping; instead, these paradigms may plausibly assay contributions of both sensory inputs and latent information. In the current study in which we rule out the possibility of

sensory-driven remapping, inactivation of DG–CA3 led to an average 86% reduction in remapping, with residual remapping failing to significantly differ from chance. Moreover, in our data

remapping within the hallway persisted even in experimental mice with hm4di under CNO, suggesting that hallway remapping and remapping by entryway within the compartment are driven by

dissociable mechanisms. In the hallway, movement is constrained such that external cues are consistently experienced in a set pattern. Thus within the hallway, remapping may be driven by

repeated experience with immediately-available sensory cues, as is typically seen on linear tracks28. Once within the compartment, however, behavior is unconstrained and immediate external

cues therefore do not disambiguate the most recent entryway. Taken together, it is possible that the lateral entorhinal cortex and the DG–CA3 circuit mediate distinct sensory- and latent-

contributions to rate remapping. In a similar vein, previous studies demonstrating rate remapping on the basis of goal-oriented behavior have not disambiguated a contribution of latent

information from a contribution of planned behavior4,7,8, and have demonstrated that remapping reflecting planned behavior is primarily driven by the nucleus reuniens, which bypasses the

DG–CA3 circuit6. Consistent with this distinction we observed no coding of future exitway, a component which has been observed in goal-oriented tasks6,7,8, suggesting that the

DG–CA3-dependent remapping we observed in this paradigm did not reflect planned behavior. We anticipate that a similar experimental approach will help to address several lingering questions,

including the relationship between latent, sensory, and goal-oriented rate remapping, the specificity of trisynaptic, lateral entorhinal, and prefrontal-thalamic circuit contributions to

these forms of rate remapping, and the contribution of the trisynaptic circuit to other forms of hippocampal remapping5,29,30. METHODS SUBJECTS Naive male mice (Grik4-cre transgenic mouse

line C57BL/6-Tg(Grik4-cre)G32-4Stl/J, The Jackson Laboratory; all other mice C57Bl/6, Charles River) were housed individually on a 12-h light/dark cycle at 22 °C and 40% humidity with food

and water ad libitum. All experiments were carried out during the light portion of the light/dark cycle, and in accordance with McGill University and Douglas Hospital Research Centre Animal

Use and Care Committee (protocol #20157725) and in accordance with Canadian Institutes of Health Research guidelines. One mouse was excluded from the initial CA1 recording experiments prior

to analysis due to an extreme entryway bias (>5:1) which persisted across all sessions. SURGERIES During all surgeries, mice were anesthetized via inhalation of a combination of oxygen

and 5% Isoflurane before being transferred to the stereotaxic frame (David Kopf Instruments), where anesthesia was maintained via inhalation of oxygen and 0.5–2.5% Isoflurane for the

duration of the surgery. Body temperature was maintained with a heating pad and eyes were hydrated with gel (Optixcare). Carprofen (10 ml kg−1) and saline (0.5 ml) were administered

subcutaneously at the beginning of each surgery. Preparation for recordings involved three surgeries per mouse. First, at the age of six to ten weeks, each mouse was transfected with a 400 

nl injection of the calcium reporter GCaMP6f according to the specific viral construct and injection coordinates described in Supplementary Table 2. The original titre of the

AAV9.syn.GCaMP6f.WPRE.SV40 construct, sourced from University of Pennsylvania Vector Core, was 3.26e14 GC-ml and was diluted in sterile PBS (1 part virus to 1 parts PBS) before surgical

microinjection. The original titre of the AAV5.CaMKII.GCaMP6f.WPRE.SV40 construct, sourced from Addgene, was 2.3e13 GC-ml and was diluted in sterile PBS (1 part virus to 2 parts PBS) before

surgical microinjection. For chemogenetic experiments, mice expressing Grik4-Cre were also injected bilaterally with the cre-mediated Designer Receptor Exclusively Activated by Designer

Drugs (DREADDs) viral construct AAV8.hsyn.DIO.hm4di.mcherry according to the injection schedule described in Supplementary Table 3 during this surgery. The original titre of this virus,

sourced from Addgene, was 4.3e12 GC-ml and was diluted in sterile PBS (5 parts virus to 6 parts PBS) before surgical microinjection. All injections were administered via glass pipettes

connected to a Nanoject II (Drummond Scientific) injector at a flow rate of 23 nl s−1. One week post-injection, either a 1.8 mm or 0.5 mm diameter gradient refractive index (GRIN) lens

(Go!Foton) was implanted above either dorsal CA1 or CA3 as indicated in Supplementary Table 2. Implantation of the 1.8 mm diameter GRIN lens required aspiration of intervening cortical

tissue, while no aspiration was required for implantation of the 0.5 mm diameter GRIN lens. Results observed using 1.8- or 0.5-mm diameter grin lenses were similar. In addition to the GRIN

lens, two stainless steel screws were threaded into the skull above the contralateral hippocampus and prefrontal cortex to stabilize the implant. Dental cement (C&B Metabond) was applied

to secure the GRIN lens and anchor screws to the skull. A silicone adhesive (Kwik-Sil, World Precision Instruments) was applied to protect the top surface of the GRIN lens until the next

surgery. One to three weeks after lens implantation, an aluminum baseplate was affixed via dental cement (C&B Metabond) to the skull of the mouse, which would later secure the

miniaturized fluorescent endoscope (miniscope) in place during recording. The miniscope/baseplate was mounted to a stereotaxic arm for lowering above the implanted GRIN lens until the field

of view contained visible cell segments and dental cement was applied to affix the baseplate to the skull. A polyoxymethylene cap with a metal nut weighing ~3 g was affixed to the baseplate

when the mice were not being recorded, to protect the baseplate and lens, as well as to simulate the weight of the miniscope. After surgery, animals were continuously monitored until they

recovered. For the initial three days after surgery mice were provided with a soft diet supplemented with Carprofen for pain management (MediGel CPF, ~5 mg kg−1 each day). Screening and

habituation to recording in a rectangular 20 × 40 cm semi-transparent plastic home cage environment began 3–7 days following the baseplate surgery and continued for at least 2 days until the

quality and reliability of the calcium data were confirmed. DATA ACQUISITION In vivo calcium videos were recorded with a miniscope (v3; miniscope.org) containing a monochrome CMOS imaging

sensor (MT9V032C12STM, ON Semiconductor) connected to a custom data acquisition (DAQ) box (miniscope.org) with a lightweight, flexible coaxial cable14. The DAQ was connected to a PC with a

USB 3.0 SuperSpeed cable and controlled with Miniscope custom acquisition software (miniscope.org). The outgoing excitation LED was set to between 3–6%, depending on the mouse to maximize

signal quality with the minimum possible excitation light to mitigate the risk of photobleaching. Gain was adjusted to match the dynamic range of the recorded video to the fluctuations of

the calcium signal for each recording to avoid saturation. Behavioral video data were recorded by a webcam mounted above the environment. Behavioral video recording parameters were adjusted

such that only the red LED on the CMOS of the miniscope was visible. The DAQ simultaneously acquired behavioral and cellular imaging streams at 30 Hz as uncompressed avi files and all

recorded frames were timestamped for post-hoc alignment. All recording environments were constructed of a gray Lego base and black Lego bricks (Lego, Inc) according to the dimensions

specified in the main text and Supplementary Figures. All external walls had a height of 22 cm; all internal walls had a height of 15 cm. All hallways were 5 cm wide; due to the width mice

typically ran the length of the hallway rather than turning around in the hallway. During recording, the environment was dimly lit by a nearby computer screen, which served as directional

cue. A white-noise generator was placed above the environment to mask uncontrolled ambient sounds. Each recording session lasted 20 min, and only one session was recorded per day to avoid

photobleaching. The mouse was always placed in the corner of the hallway at the start of the session and was allowed to explore the environment for 15–30 s prior to data acquisition.

Following each recording the environment was cleaned with disinfectant (Prevail). For CA3 inactivation experiments, 5 mg kg−1 of clozapine-N-oxide (CNO + 0.7% DMSO) was injected

Intraperitoneally 1 h and 15 min prior to recording. Mice were returned to their home cage between the injection and the start of the recording session. We conducted two separate control

experiments to rule out the possibilities that our results could be explained by injection procedure, expression of hm4di-mcherry, or to non-specific effects of CNO itself. As our first

within-animal control, we injected sterile saline instead of CNO and repeated the recording and analysis procedures (Fig. 4b–d). The order of Saline and CNO recordings was interleaved within

mouse, and whether the first recording session for a mouse followed a Saline or CNO injection was randomized. As a second control, to ensure that any differences between Saline and CNO

sessions were attributable to the interaction between hm4di and CNO, and not an effect of CNO or its metabolites alone, CNO injection experiments were repeated in a second across-mouse

control group which did not express hm4di (Fig. 4c). DATA PREPROCESSING Calcium imaging data were preprocessed prior to analyses via a pipeline of open source MATLAB (MathWorks; version

R2015a) functions to correct for motion artifacts16, segment cells and extract transients17,18, and infer the likelihood of spiking events via deconvolution of the transient trace through a

second-order autoregressive model19. The motion-corrected calcium imaging data were manually inspected to ensure that motion correction was effective and did not introduce additional

artifacts. Following this preprocessing pipeline, the spatial footprints of all cells were manually verified to remove lens artifacts. Position data were inferred from this LED offline

following recording using a custom written MATLAB (MathWorks) script and were manually corrected if needed. The experimenter manually segmented data recorded in the compartment and hallway,

as well as the most recent entryway based on the recorded position data prior to all further analyses. In one supplemental analysis, we characterized calcium transients by eschewing firing

rate estimation via autoregressive deconvolution and instead binarized each trace as follows. dF/F traces were first _Z_-scored relative to their baseline variance, estimated from the

half-normal distribution as when computing SNR. Next, all periods such that _Z_ > 3 and dZ/dt > 0 were assigned a value of one. All other periods were assigned a value of zero. This

binarization procedure thus captures the rise times of high-confidence calcium transients. DATA ANALYSIS All analyses were conducted using the vector of inferred likelihood of spiking events

(ILSE), treating this vector as if it were the firing rate of the cell19, except for one supplemental analysis, as described above. Rate maps of activity in the compartment were constructed

by first binning the position data into pixels corresponding to a 2.5 cm × 2.5 cm grid of locations. To construct a rate map, the mean ILSE was computed for each pixel and then smoothed

with a 4 cm standard deviation isometric Gaussian kernel. Rate maps of activity in the hallway were created by first collapsing the position data onto a line drawn through the center of the

hallway. Next, the position data on this line were binned into 17 equally spaced pixels. Next, the mean ILSE was computed for each pixel. This map was then smoothed with a 2 pixel standard

deviation Gaussian kernel. To identify place cells coding for locations within the compartment in a manner to avoid a bias for or against the observation of remapping, we applied the

following procedures. First, we partitioned the data within the compartment according to the most recent entryway, and according to the first and second halves of the recording and created

rate maps for each partition. Next, we computed the Pearson correlation between each pair of cross-half rate map comparisons (e.g., Entry A half 1 to Entry A half 2; Entry A half 1 to Entry

B half 2; Entry B half 1 to Entry A half 2; Entry B half 1 to Entry B half 2), while subsampling the data to match the spatial sampling distributions (described below). Next, for each cell,

we created a null distribution of chance correlation values by circularly shifting the ILSE vector in time a random amount at least 60 s away from its proper alignment and recomputing all

four correlations 500 times. The maximum correlation value of each iteration was then compared to the maximum value of the cell’s actual correlations. A cell was categorized as a place cell

if the maximum of its actual correlations exceeded the 95th percentile of this shuffled control. This selection procedure thus inherently corrects for multiple comparisons and weighs

correlations within and across both most recent entryways equally. This procedure identified as place cells an average of 29.3% ± 1.8% (SEM) with split-half reliable compartment rate maps in

the initial CA1 recordings, 39.2% ± 1.9% in the CA3 recordings, 27.7% ± 1.4% in the recordings of experimental mice with hm4di (26.9% ± 1.9% under Saline; 28.5% ± 2.0% under CNO), and 28.5%

 ± 1.8% in the recordings of control mice without hm4di (24.6% ± 2.3% under Saline; 32.1% ± 2.5% under CNO). Signal-to-noise (SNR) for each dF/F trace was computed as a measure of data

quality prior to spike estimation via deconvolution as described in previously31. Briefly, because calcium transients around the baseline can only be positive, for each trace the variance of

a normal noise distribution was estimated by scaling the standard deviation of trace values below baseline via a half-normal distribution, such that: $${\mathrm{NOISE}} =

\frac{{{\mathrm{std}}\left( {{\mathbf{t}}\left( {{\mathbf{t}} {\,}< {\,}0} \right)} \right)}}{{\sqrt {1 - \frac{2}{{\it{\uppi }}}} }}$$ (1) where _t_ is the detrended filtered calcium

trace and T(T < 0) are all filtered trace values below baseline. Next, we _z_-score the trace T such that _z_(T) = T/NOISE. These z-scored values are next transformed into probabilities

_p_(T) = _φ_(−_z_(T)), where _φ_ denotes the cumulative distribution function of the standard normal distribution. Then we compute the lowest probability event over a time window of 0.4 s

(_N_ = 12 frames), the duration around which a GCamp6F transient is typically at its maximum, such that: $${\it{p}}_{{\mathrm{min}}} = \mathop {{{\mathrm{min}}}}\limits_{\it{i}} \left(

{\mathop {\prod}\limits_{{\it{j}} = 0}^{{\it{N}} - 1} {\it{p}} \left( {{\mathbf{t}}\left( {{\it{i}} + {\it{j}}} \right)} \right)} \right)^{1/{\it{N}}}$$ (2) where \(\mathop

{{{\mathrm{min}}}}\limits_{\it{i}}\) is the minimum value across all timepoints _i_ from _i_  = 1 to _i_  = the length of T − _N_ + 1. Finally, we define the average peak SNR as:

$${\mathrm{SNR}} = - \varphi ^{ - 1}\left( {{\it{p}}_{min}} \right)$$ (3) where _φ_−1 is the quantile function of the standard normal distribution (logit function). Spatial information

content (SIC) was computed from the whole-session rate maps of each cell as described previously32 via the equation: $${\mathrm{SIC}} = \mathop {\sum}\limits_{\it{i}}

{{\mathbf{s}}_{\it{i}}{\mathbf{r}}_{\it{i}}} \,{\mathrm{log}}\left( {\frac{{{\mathbf{r}}_{\it{i}}}}{{{\bar{\boldsymbol{r}}}}}} \right)$$ (4) where _i_ is the rate map pixel index, S_i_ is

the probability of sampling pixel _i_, R_i_ is the mean firing rate at pixel _i_, and \({\bar{\mathbf{r}}}\) is the mean firing rate across all pixels. To correct for biases in sampled

spatial locations, we subsampled our data during all comparisons to match the spatial sampling distributions across all comparisons (Supplementary Fig. 6). To do so, we computed the minimum

number of samples recorded at each pixel location across all comparisons. Next for each comparison, we included a random subset of the data recorded at that location to match that minimum

number of samples. Because this random subsampling introduces variability, we repeated all comparisons 100 times and took the mean value as the final measure of that comparison. The primary

measure of interest in most comparisons was population vector similarity assessed via a population vector correlation. This measure was computed by first concatenating the pixel values of

all rate maps and all cells into a large vector for each comparison. Next, the Pearson correlation between these two vectors was taken. Similar results were observed for all analyses when

cosine distance was instead used to assess population similarity (not shown). To quantify changes in the relative firing rates at an individual cell level, we computed the absolute percent

change in firing rate within the compartment across both halves of the recording when entering through either the same entryway or different entryways. Mean firing rates were defined as the

firing rate averaged over the entire time spent within the compartment following each entryway for each half independent of position (after downsampling to correct for differences in the

spatial sampling distribution). Firing rates in each half were normalized to the maximum firing rate for that half. For example, the absolute percent change in firing rate between entering

from entryway A in the first half and entryway B in the second half Δ_F_A,B was computed as: $$\Delta {\it{F}}^{A,B} = \left| {\frac{{{\it{F}}_1^{\mathrm{A}}}}{{{\mathrm{max}}\left(

{{\it{F}}_1^{\mathrm{A}},{\it{F}}_1^B} \right)}} - \frac{{{\it{F}}_2^{\mathrm{B}}}}{{{\mathrm{max}}\left( {{\it{F}}_2^{\mathrm{A}},{\it{F}}_2^B} \right)}}} \right|$$ (5) where

\({\it{F}}_1^{\mathrm{A}}\) denotes the firing rate when entering from entryway A in the first half of the recording, \({\it{F}}_2^{\mathrm{B}}\) denotes the firing rate when entering from

entryway B in the second half of the recording, etc. HISTOLOGICAL VALIDATION OF EXPRESSION AND RECORDING TARGETS Once mice had completed all behavioral experiments, animals were perfused to

verify GRIN lens placement and virus expression. Mice were deeply anesthetized and intracardially perfused with 4% paraformaldehyde in PBS. Brains were dissected and post-fixed with the same

fixative. Coronal sections (40 μm) of the entire hippocampus were cut using a vibratome and sections were mounted directly on glass slides. Sections were split and half of all sections were

stained for cresyl violet to localize GRIN lens placement, the other half of sections were stained for DAPI and mounted with Fluromount-G (Southern Biotechnology) to evaluate virus

expression. To evaluate the hm4di-mCherry fluorescence for each mouse, pyramidal cells were quantified across eight coronal slices representing the injection sites along the septo-temporal

axis of the hippocampus, with three randomly selected images taken across three subregion (CA1, CA3, dentate gyrus) for a total of 72 images analyzed per animal. To evaluate the level of

expression in each subregion, the number of transfected pyramidal cells was counted using imageJ (version 1.51j8) and virus expression was imaged with an AxioObserver.Z1 microscope (Carl

Zeiss). STATISTICS AND REPRODUCIBILITY All statistical tests are noted where the corresponding results are reported throughout the main text and supplement. All tests were uncorrected

2-tailed tests unless otherwise noted. _Z_-values for nonparametric Wilcoxon tests were not estimated or reported for comparisons with fewer than 15 datapoints. In all figures, the mean

served as the measure of central tendency, and error bars reflected ±1 standard error of the mean, unless otherwise noted. REPORTING SUMMARY Further information on research design is

available in the Nature Research Reporting Summary linked to this article. DATA AVAILABILITY The complete dataset for all experiments are publicly available at

https://doi.org/10.5061/dryad.crjdfn31g or via request to the corresponding authors. Source data for all main text and Supplementary Figures can be found in the accompanying Source data

Excel spreadsheet as noted in the figure captions. Source data are provided with this paper. CODE AVAILABILITY All custom code written for reported analyses are publicly available at

https://github.com/akeinath/EntrywayRemapping or via request to the corresponding authors. REFERENCES * O’Keefe, J. & Dostrovsky, J. The hippocampus as a spatial map. Preliminary

evidence from unit activity in the freely-moving rat. _Brain Res._ 34, 171–175 (1971). Article  Google Scholar  * Anderson, M. I. & Jeffery, K. J. Heterogeneous modulation of place cell

firing by changes in context. _J. Neurosci._ 23, 8827–8835 (2003). Article  CAS  Google Scholar  * Skaggs, W. E. & McNaughton, B. L. Spatial firing properties of hippocampal CA1

populations in an environment containing two visually identical regions. _J. Neurosci._ 18, 8455–8466 (1998). Article  CAS  Google Scholar  * Wood, E. R., Dudchenko, P. A., Robitsek, R. J.

& Eichenbaum, H. Hippocampal neurons encode information about different types of memory episodes occurring in the same location. _Neuron_ 27, 623–633 (2000). Article  CAS  Google Scholar

  * Leutgeb, S. et al. Independent codes for spatial and episodic memory in hippocampal neuronal ensembles. _Science_ 309, 619–623 (2005). Article  ADS  CAS  Google Scholar  * Ito, H. T.,

Zhang, S.-J., Witter, M. P., Moser, E. I. & Moser, M.-B. A prefrontal-thalamo-hippocampal circuit for goal-directed spatial navigation. _Nature_ 522, 50–55 (2015). Article  ADS  CAS 

Google Scholar  * Frank, L. M., Brown, E. N. & Wilson, M. Trajectory encoding in the hippocampus and entorhinal cortex. _Neuron_ 27, 169–178 (2000). Article  CAS  Google Scholar  *

Ferbinteanu, J. & Shapiro, M. L. Prospective and retrospective memory coding in the hippocampus. _Neuron_ 40, 1227–1239 (2003). Article  CAS  Google Scholar  * Lu, L. et al. Impaired

hippocampal rate coding after lesions of the lateral entorhinal cortex. _Nat. Neurosci._ 16, 1085–1093 (2013). Article  CAS  Google Scholar  * McHugh, T. J. et al. Dentate gyrus NMDA

receptors mediate rapid pattern separation in the hippocampal network. _Science_ 317, 94–99 (2007). Article  ADS  CAS  Google Scholar  * Grieves, R. M., Jenkins, B. W., Harland, B. C., Wood,

E. R. & Dudchenko, P. A. Place field repetition and spatial learning in a multicompartment environment. _Hippocampus_ 26, 118–134 (2016). Article  Google Scholar  * Spiers, H. J.,

Hayman, R. M. A., Jovalekic, A., Marozzi, E. & Jeffery, K. J. Place field repetition and purely local remapping in a multicompartment environment. _Cereb. Cortex_ 25, 10–25 (2015).

Article  Google Scholar  * Fuhs, M. C., Vanrhoads, S. R., Casale, A. E., McNaughton, B. & Touretzky, D. S. Influence of path integration versus environmental orientation on place cell

remapping between visually identical environments. _J. Neurophysiol._ 94, 2603–2616 (2005). Article  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  Google Scholar  * Ghosh, K. K. et al. Miniaturized integration of a fluorescence microscope. _Nat. Methods_

8, 871–878 (2011). Article  CAS  Google Scholar  * Pnevmatikakis, E. A. & Giovannucci, A. NoRMCorre: An online algorithm for piecewise rigid motion correction of calcium imaging data.

_J. Neurosci. Methods_ 291, 83–94 (2017). Article  CAS  Google Scholar  * Pnevmatikakis, E. A. et al. Simultaneous denoising, deconvolution, and demixing of calcium imaging data. _Neuron_

89, 285–299 (2016). Article  CAS  Google Scholar  * Zhou, P. et al. Efficient and accurate extraction of in vivo calcium signals from microendoscopic video data. _Elife_ 7, e28728 (2018). *

Friedrich, J., Zhou, P. & Paninski, L. Fast online deconvolution of calcium imaging data. _PLoS Comput. Biol._ 13, e1005423 (2017). Article  ADS  Google Scholar  * McNaughton, B. L.,

Barnes, C. A. & O’Keefe, J. The contributions of position, direction, and velocity to single unit activity in the hippocampus of freely-moving rats. _Exp. Brain Res._ 52, 41–49 (1983).

Article  CAS  Google Scholar  * Nakazawa, K. et al. Requirement for hippocampal CA3 NMDA receptors in associative memory recall. _Science_ 297, 211–218 (2002). Article  ADS  CAS  Google

Scholar  * Armbruster, B. N., Li, X., Pausch, M. H., Herlitze, S. & Roth, B. L. Evolving the lock to fit the key to create a family of G protein-coupled receptors potently activated by

an inert ligand. _Proc. Natl Acad. Sci. USA_ 104, 5163–5168 (2007). Article  ADS  Google Scholar  * Marr, D. Simple memory: a theory for archicortex. _Philos. Trans. R. Soc. Lond. B. Biol.

Sci._ 262, 23–81 (1971). Article  ADS  CAS  Google Scholar  * O’Reilly, R. C. & McClelland, J. L. Hippocampal conjunctive encoding, storage, and recall: avoiding a trade-off.

_Hippocampus_ 4, 661–682 (1994). Article  Google Scholar  * Knierim, J. J. & Neunuebel, J. P. Tracking the flow of hippocampal computation: Pattern separation, pattern completion, and

attractor dynamics. _Neurobiol. Learn. Mem._ 129, 38–49 (2016). Article  CAS  Google Scholar  * Brun, V. H. et al. Place cells and place recognition maintained by direct

entorhinal-hippocampal circuitry. _Science_ 296, 2243–2246 (2002). Article  ADS  CAS  Google Scholar  * Stachenfeld, K. L., Botvinick, M. M. & Gershman, S. J. The hippocampus as a

predictive map. _Nat. Neurosci._ 20, 1643–1653 (2017). Article  CAS  Google Scholar  * Navratilova, Z., Hoang, L. T., Schwindel, C. D., Tatsuno, M. & McNaughton, B. L.

Experience-dependent firing rate remapping generates directional selectivity in hippocampal place cells. _Front. Neural Circuits_ 6, 6 (2012). Article  Google Scholar  * Bostock, E., Muller,

R. U. & Kubie, J. L. Experience-dependent modifications of hippocampal place cell firing. _Hippocampus_ 1, 193–205 (1991). Article  CAS  Google Scholar  * Mizumori, S. J., Ragozzino, K.

E., Cooper, B. G. & Leutgeb, S. Hippocampal representational organization and spatial context. _Hippocampus_ 9, 444–451 (1999). Article  CAS  Google Scholar  * Giovannucci, A. et al.

CaImAn an open source tool for scalable calcium imaging data analysis. _Elife_ 8, e38173 (2019). * Skaggs, W. E., McNaughton, B. L. & Gothard, K. M. An information-theoretic approach to

deciphering the hippocampal code. in _Advances in neural information processing systems_ 1030–1037 (1993). Download references ACKNOWLEDGEMENTS We kindly thank S. Kim and P. Mannarino for

technical assistance and J. Hinman, A. Alexander, S. Chekhov, R. Rozeske, J. Ying, C. Kramar, and M. Hasselmo for comments of previous versions of this paper. We especially thank D. Aharoni

for extensive guidance in using the UCLA miniscope. This work was supported by a McGill University Healthy Brains for Healthy Lives CFREF postdoctoral fellowship to A.T.K., Fonds de

Recherche du Québec—Santé (FRQS) postdoctoral fellowship to A.N.P., and funding from the Canadian Institutes for Health Research grants (#367017 and #377074), the Natural Sciences and

Engineering Council (Discovery grant #74105), and the Canada Research Chairs Program to M.P.B. AUTHOR INFORMATION AUTHORS AND AFFILIATIONS * Department of Psychiatry, Douglas Hospital

Research Centre, McGill University, 6875 Boulevard LaSalle, Verdun, QC, H4H 1R3, Canada Alexandra T. Keinath, Andrés Nieto-Posadas & Mark P. Brandon * Department of Psychological and

Brain Sciences, Rajen Kilachand Center for Integrated Life Sciences and Engineering, Boston University, 610 Commonwealth Avenue, Boston, MA, 02215, USA Jennifer C. Robinson * Integrated

Program in Neuroscience, Montreal Neurological Institute, McGill University, 3801 University Street, Montreal, QC, H3A 2B4, Canada Mark P. Brandon Authors * Alexandra T. Keinath View author

publications You can also search for this author inPubMed Google Scholar * Andrés Nieto-Posadas View author publications You can also search for this author inPubMed Google Scholar *

Jennifer C. Robinson View author publications You can also search for this author inPubMed Google Scholar * Mark P. Brandon View author publications You can also search for this author

inPubMed Google Scholar CONTRIBUTIONS A.T.K. contributed to experimental design, surgeries, recordings, analysis of data, as well as drafting and revising the paper. A.N.P. contributed to

surgeries and recordings. J.C.R. contributed to histological verification of recording sites and quantifications. M.P.B. contributed to experimental design, analysis of data, as well as

drafting and revising the paper. CORRESPONDING AUTHORS Correspondence to Alexandra T. Keinath or Mark P. Brandon. ETHICS DECLARATIONS COMPETING INTERESTS The authors declare no competing

interests. ADDITIONAL INFORMATION PEER REVIEW INFORMATION _Nature Communications_ thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports

are available 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 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 Keinath, A.T., Nieto-Posadas, A., Robinson, J.C. _et al._ DG–CA3 circuitry mediates

hippocampal representations of latent information. _Nat Commun_ 11, 3026 (2020). https://doi.org/10.1038/s41467-020-16825-1 Download citation * Received: 16 October 2019 * Accepted: 21 May

2020 * Published: 15 June 2020 * DOI: https://doi.org/10.1038/s41467-020-16825-1 SHARE THIS ARTICLE Anyone you share the following link with will be able to read this content: Get shareable

link Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative