Spatiotemporal information is differentially conveyed by hippocampal projections to the anterior olfactory nucleus during episodic-like odour memory

The hippocampus is essential for representing spatiotemporal context and associating it with the sensory details of daily life to form episodic memories. However, the neural circuit underlying this process remains poorly understood. We selectively inhibited hippocampal projections to the anterior olfactory nucleus (AON) during behavioural tests of contextually-cued odour recall. We found that inhibition of intermediate HPC (iHPC)-lateral AON (lAON) pathway impaired spatial odour memory while inhibition of ventral HPC (vHPC)-medial AON (mAON) pathway disrupted both spatial and temporal odour memory. Our results indicate that the spatial and temporal information of episodic-like odour memory is conveyed by topographically distinct hippocampal-AON pathways.

The hippocampus is essential for representing spatiotemporal context and associating it with the 19 sensory details of daily life to form episodic memories. However, the neural circuit underlying this 20 process remains poorly understood. We selectively inhibited hippocampal projections to the 21 anterior olfactory nucleus (AON) during behavioural tests of contextually-cued odour recall. We What happened, when, and where? The ability to readily integrate elements of a unique event into 28 a single representation is a fundamental property of episodic memory 1

. Lesion and recording studies in 29
humans and nonhuman animals have highlighted a central role of the hippocampus (HPC) in mediating 30 episodic memory 2,3 . The spatial and temporal context of an event are first encoded within the HPC as the 31 collective activity of place and time cells 4,5 . Contextual information later serves as a potent retrieval cue, 32 bringing about the rich multisensory details of the original experience 6,7 . An emerging theory holds that 33 the HPC conducts this retrieval process by reinstating patterns of cortical activity observed during 34 learning 8,9 . Yet, it is not known how hippocampal transmission of contextual information can reproduce 35 the sensory aspects of episodic memory. 36 Olfaction is considered the most evolutionarily ancient sense as evidenced by the direct 37 anatomical connections between the olfactory cortex and the limbic system 10 . In particular, hippocampal 38 projections to the olfactory cortex offer a unique experimental model for understanding the context-driven 39 recollection of previously encountered sensory stimuli. Recently, we revealed a dense and topographically 40 organized projection from the dorsoventral extent of the HPC to the anterior olfactory nucleus (AON) 41 ( Fig. 1a; Aqrabawi and Kim, 2017). The AON receives unidirectional, monosynaptic inputs from the 42 CA1, in contrast to other primary sensory areas that receive hippocampal inputs indirectly via adjacent 43 medial temporal lobe structures, underlining further the intimate relationship between olfaction and 44 memory 11 . 45 The AON is an ideal site of convergence for olfactory and contextual information given its 46 anatomical position as the initial recipient of input from the olfactory bulb and the largest source of 47 feedback projections within the olfactory cortex 12,13 . Consistently, it has been shown that hippocampal 48 inputs to the AON can alter olfactory perception and odour-guided behaviours 14 . However, the functional 49 role of the HPC-AON pathway remains unexplored, leaving open fundamental questions regarding 50 olfactory episodic memory. Here, we combine chemogenetic and optogenetic approaches to demonstrate 51 that information regarding the spatial and temporal context of odour memory is delivered by 52 topographically organized hippocampal inputs to the AON. 53 To manipulate activity in the HPC-AON pathway, we infused the retrogradely propagating canine 54 adenovirus encoding Cre recombinase (CAV2-Cre) into the AON, followed by HPC infusions of an AAV 55 vector carrying a Cre-dependent inhibitory hM4D-mCherry. This allowed us to selectively inhibit HPC 56 cells that project to the AON upon the administration of clozapine-N-oxide (CNO). Cre-mediated viral 57 mCherry expression was observed throughout the ventral two thirds of the hippocampal CA1 (Fig. 1b). 58 Three weeks after viral infusion, mice underwent behavioral tests to evaluate memory for the associations 59 between odours and the spatiotemporal context in which they occurred. Assessing the retrieval of episodic 60 odour information was made possible by capitalizing on the innate tendency of mice to preferentially 61 investigate novel stimuli 15 . Thus, mice can behaviourally express correct memory by investigating odours 62 paired with a novel position in space, or temporal sequence, more so than familiar configurations. 63 We first tested the ability to remember 'where' specific odours occurred in context. In two 64 encoding phases of a spatial odour memory test, mice were presented with two different odours (odour 1 65 and 2) placed on opposite ends of a distinct context (A) for 5 min (Fig. 1ci). Mice were then placed in a 66 separate context (B) where the same odours were positioned in reversed locations. Each exposure was 67 followed by a 15 min retention delay, the second of which was preceded by a CNO injection. In the 68 subsequent retrieval phase, animals were reintroduced to context A wherein two copies of either odour 1 69 or 2 were presented on both sides. In this paradigm, correct memory expression would drive mice to 70 investigate the odour found at the novel location within the context, as seen in control mice expressing 71 mCherry alone (Fig. 1cii). In contrast, CNO-treated hM4D mice investigated both, otherwise identical 72 odours for a similar proportion of time. These results cannot be explained by differences in total 73 investigation time or distance traveled as both groups displayed similar measures in each (Supplementary 74 Fig. 1). 75 Next, we examined memory for 'when' specific odours were encountered. Mice were presented 76 with a sequence of odours in three successive encoding phases followed by a retrieval phase where two of 77 the previously encountered odours were reintroduced (Fig. 1di). Control mice preferentially investigated 78 the odour encountered earlier in the sequence, yet CNO-treated hM4D mice investigated both odours to a 79 largely equal extent (Fig. 1dii). Importantly, the spatial context was consistent throughout the test where 80 novelty was only conferred by temporal distance between the two odours. Lastly, to examine whether 81 hippocampal function can be extended to memory for odours regardless of context, we conducted a novel 82 odour recognition test where animals were presented with a familiar and previously unexplored odour 83 (Fig. 1ei). Both groups preferentially investigated the novel odour despite CNO administration (Fig. 1eii). 84 Together, these results indicate that AON-projecting HPC cells mediate the retrieval of odour memory 85 only when it is tied to spatiotemporal context. 86 Topographically organized HPC terminals at the AON may transmit distinct contextual cues 87 depending on where they arise within the HPC. We employed archaerhodopsin (ArchT) to 88 optogenetically inhibit intermediate or ventral HPC terminals that innervate separate AON subregions. using AAV expressing GFP only. Animals were tested in the same behavioural paradigms used for hM4D 95 experiments, except inhibition was mediated by light (532 nm, 12 mW) limited to the retrieval phase. 96 Illumination of the iHPC-lAON and vHPC-mAON pathways disrupted memory for 'where' 97 odours were encountered as both groups were unable to discriminate odours tied to a spatial location 98 within a specific context (Fig. 2b). In contrast, only vHPC-mAON, but not iHPC-lAON pathway 99 inhibition impaired memory for 'when' odours occurred in a temporal sequence (Fig. 2c). Importantly, all 100 groups showed similar levels of novelty preference in a context-independent novel odour recognition test 101 (Fig. 2d). The identified impairments in episodic odour memory retrieval suggest that representations of 102 space and time are differentially distributed across the AON whereby both iHPC and vHPC inputs deliver 103 spatial information, but information regarding the temporal context is supported only by vHPC inputs. 104 To investigate further the spatiotemporal contributions of hippocampal inputs to the AON, we 105 employed an episodic memory test where recollection of an odour, its spatial location, and temporal 106 occurrence (what-when-where) were tested simultaneously 16,17 . The test involved two encoding phases 107 and one retrieval phase, each separated by a one hour delay (Fig. 2ei). Light-mediated inhibition was 108 least. Strikingly, inhibition of the vHPC-mAON pathway produced an inverse pattern of investigation to 120 the control group. The odour presented in the FL/TR configuration was investigated most, while the 121 NL/TD odour was investigated least. A separate analysis delineating the spatial and temporal 122 contributions to memory revealed that iHPC-lAON pathway inhibition impaired spatial but not temporal 123 odour associations, while the vHPC-mAON pathway inhibition impaired both components to a similar 124 extent (Fig. 2eiii). Collectively, these findings confirm the differential roles of the iHPC-lAON and 125 vHPC-mAON pathways in conveying spatiotemporal information during olfactory episodic memory 126 retrieval. 127 The ability to explicitly recollect sensory information is a hallmark of episodic memory, 128 particularly when lacking a physical sensory cue 18 . Context alone can drive the activity of primary 129 olfactory regions to form internal representations of odours although the underlying neural circuit is 130 unknown 6,7 . To this end, we examined whether the HPC-AON pathway could support this function. Mice 131 were allowed to explore a rich spatial context in the presence of a pure odour emitted from a cotton swab 132 tip for 30 min per day over nine consecutive days. On day 10, mice were reintroduced to the context in 133 the absence of the applied odour (Fig. 3a). The mismatch between the odour-paired context and the lack 134 of an emitted scent drives an increase in the investigation time for the cotton swab (Supplementary Video 135 1). Indeed, control mice exhibited a marked increase in investigation of the cotton swab upon failure to 136 find the expected odour (Fig. 3b). In contrast, hM4D-and ArchT-mediated inhibition of the HPC-AON 137 pathway abolished this behaviour. All groups showed no increase in investigation time when similarly 138 trained but tested in a novel context ( Supplementary Fig. 4). Therefore, HPC-AON communication is 139 necessary for mediating the context-driven recollection of odours. 140 Ultimately our findings support a model of episodic odour memory whereby information 141 regarding odour quality and spatiotemporal context merge at the level of the AON and, as a natural 142 consequence of hebbian synaptic plasticity, produces cellular populations that represent previously 143 encountered odours within the context in which they occurred (Fig. 3c). Such a system maintains the 144 fidelity of the original odour memory and allows access for retrieval of the full trace via partial cues from 145 either olfactory or contextual inputs. 146

Animals 148
Male C57BL/6 mice (Charles River Laboratories) were used in all behavioural tests. All mice were 8-10 149 weeks old at the time of surgery and 12-14 weeks old at the time behavioural testing began. Prior to 150 surgery, mice were group-housed in a temperature-controlled room on a 12 hour light/dark cycle with ad 151 libitum access to food and water. Following surgery, mice were individually-housed. A total of 60 mice 152 were distributed into two groups for hM4D experiments (hM4D-mCherry: n=10, mCherry control: n=8) 153 and three groups for ArchT experiments (GFP control: n=14, iHPC-lAON: n=12, vHPC-mAON: n=16). 154 All procedures were performed in accordance with the guidelines of the Canadian Council on Animal 155 Care (CCAC) and the University of Toronto Animal Care Committee. minute interval was allotted after each infusion to limit the viral spread. The atlas of the mouse brain by 178 Paxinos and Franklin (2007)  Odours were presented mixed with woodchip bedding in 3 cm wide, 1 cm high aluminum cups. Multiple 195 identical odour cups were used such that an animal never investigated the same cup twice. The odours 196 used included nutmeg, vanillin, coriander, banana, garlic, cinnamon, thyme, almond, onion, curry, ginger, 197 savory, cumin, dill, jasmine, coffee, oregano, sage, and rosemary. The odours presented and the order of 198 their presentation between animals was pseudorandomized. For habituation, mice were given 15 minutes 199 of exploration time for each unique context prior to initial exposure. Each exposure was 5 minutes in 200 length and inter-trial intervals were 15 minutes. All tests were video-recorded at 60 fps using a NIKON 201 D5200 equipped with a 30 mm lens. An additional overhead video was recorded using a Logitech 202 webcam. All videos were subsequently scored blind to the treatment groups. Exploration was strictly 203 defined as head up sniffing, directed towards and within 1 cm of the odour source. This definition 204 excludes the use of the odour cup for sitting or as support during rearing.

206
Behavioural protocols 207 Olfactory Spatial Memory Test 208 In this paradigm adapted from Eacott and Norman (2004), mice were tested for memory of odour location 209 in context. The test chamber was altered to produce two distinct contexts. Zebra-patterned paper was used 210 to line the walls of context A whereas context B had transparent walls surrounded by red plastic cups and 211 bedding on the floor. Each animal underwent two encoding phases and one retrieval phase. During the 212 first encoding phase, mice explored context A where two highly distinct odours were placed at opposite 213 ends of the chamber (odour 1 on the left and odour 2 on the right). Next, mice were removed from the 214 chamber and placed in a holding cage. The mice were then returned to the chamber, except it was now 215 configured as context B and contained both odours in opposite positions (odour 2 on the left and odour 1 216 on the right). The animals were allowed to explore both odours in their new positions before being placed 217 back into the holding cage. For the retrieval phase, the chamber was reconfigured as context A but now 218 two copies of one odour were presented on both sides of the chamber. Time spent investigating the odour 219 cups was measured. The novel configuration consists of the familiar odour in a novel position within the 220 original context. The initial context and left/right position of the odour cups were pseudorandomized.

222 Olfactory Temporal Order Memory Test 223
This paradigm is based on similar tests used previously to measure memory for the temporal order of 224 objects 20,21 . Mice were tested in a transparent chamber with spatial cues kept constant throughout the 225 session. Each animal underwent three encoding phases and one retrieval phase. In the first exposure, mice 226 were placed in the chamber with two copies of one odour presented on opposite sides of the arena. After 227 exploring both copies, the animal was removed from the chamber and placed in a holding cage. This 228 process was repeated two more times using different odours each time. During the retrieval phase the 229 animal was returned to the chamber, but this time earlier and recently explored odours were presented on 230 opposite sides. In this case, the odour explored earlier is 'more novel' given its temporal distance 231 compared to the recently explored odour. The left-right positions of the first and last odours during the 232 test were pseudorandomized. Time spent investigating both odours was measured. 233

Novel Odour Recognition Test 234
This test was given 72 hours after examining performance on temporal order memory and followed a 235 similar paradigm. The animals underwent three encoding phases where two copies of a unique odour were 236 presented in each. On the retrieval phase, mice were presented with the initially encountered odour and a 237 previously unexplored odour on opposite sides of the chamber. Time spent investigating each odour was 238 measured.

240
Olfactory Episodic Memory Test 241 Animals were first habituated to the apparatus which consisted of a 50 cm x 50 cm x 20 cm transparent 242 plexiglass open field for a 30 minute period. The animals were then exposed to two encoding phases and 243 one retrieval phase each separated by a one hour delay. In the first encoding phase, animals were given 10 244 minutes to explore two different odours located at two adjacent corners of the arena. In the second 245 encoding phase, the animals were given an additional 10 minutes to explore another set of unique odours 246 presented on the opposite adjacent corners. During the retrieval phase, all four odours were presented 247 with the spatial position of one odour from each set exchanged. This presentation results in each odour where) memory results in a pattern of exploration such that the odour with the NL/TD configuration is 252 preferentially investigated the most while the FL/TR configuration is investigated the least. Time spent 253 investigating all four odours was measured for five minutes in the retrieval phase. Overhead videos were 254 analyzed using the ANY-maze software to produce average heat maps of each treatment group's position 255 within the arena.

257
Context-Driven Odour Recall Test 258 In this paradigm adapted from Mandairon et al. (2014), mice were trained to associate a visually distinct 259 context with an odour and subsequently tested for recollection of the odour when exposed to the context 260 alone. The testing apparatus consists of a 50 cm x 30 cm x 20 cm plexiglas cage with colourful visual 261 patterns pasted on the outside of the walls. A wooden applicator with a cotton swab tip was positioned 3 262 cm from the floor and 5 cm from one end of the chamber. Before introducing mice into the chamber, 100 263 μL of a pure odourant was applied to the cotton tip. Each mouse was randomly assigned a monomolecular 264 odourant to be trained with among limonene, isoamyl acetate, nonane, and 1-pentanol. Mice were allowed 265 to explore the context and the odorized cotton swab for 30 minutes per day for 9 consecutive days. On 266 Day 10, mice were once again placed into the cage, however no odour was added to the cotton swab. 267 Investigation time of the cotton swab was measured on day 9 and 10 for the first 5 minutes of their 268 exposure to the context. Upon failure to detect the expected odour, mice behaviourally expressed memory 269 by spending a greater amount of time investigating the cotton swab compared to their investigation time 270 when the odour was present on Day 9.

272
Histology 273 After behavioural testing, mice were transcardially perfused with Phosphate-Buffered Saline (PBS, pH 274 7.4), followed by 4% paraformaldehyde in phosphate buffer. Brain tissue was extracted and postfixed 275 overnight at 4°C. The brains were then cryoprotected using a 30% sucrose in PBS solution. Coronal 40 276 µm thick sections were collected using a cryostat (Leica, Germany). The sections were slide-mounted, 277 counterstained with 4′,6-diamidino-2-phenylindole 135 (DAPI) for five minutes, and subsequently 278 coverslipped with Aquamount (Polysciences Inc, Warrington, PA). Wide-field fluorescent images were 279 captured using a 4X objective lens on a fluorescent microscope (Olympus, Japan). Confocal images were 280 captured using a 20X and 60X objective through a Quorum spinning disk confocal microscope (Zeiss). 281 Adobe Photoshop CS6 (Adobe Systems Incorporated, San Jose, CA) was used to adjust the brightness 282 and contrast of representative sections. 283

Calculations and Statistical Analysis 284
The discrimination ratios were derived from the exploration time of odor-context pairings during the 285 retrieval phase of each test. For all tests based on the spontaneous novelty preference paradigm, the 286 discrimination ratio was calculated as the difference between the times spent exploring the novel and and multiplied by 100%: For analyzing spatial memory, the difference in the amount of time spent investigating odours with novel 298 and familiar positions was divided by the total investigation time: For temporal memory, the difference in the time spent investigating odours experienced earlier and later 302 was divided by the total investigation time: Performance in novelty preference-based paradigms was compared using t-test in hM4D experiments and 306 a one-way ANOVA with testing group as the factor in optogenetic experiments. Percent investigation 307 time data collected in the episodic memory test was analyzed using a two-way ANOVA with testing 308 group and odour spatiotemporal configuration as factors. A two-way ANOVA was used to analyze data in 309 the contextually-cued odour recall test for hM4D and optogenetic experiments, respectively, using ChR2-YFP and -mCherry injections in the hippocampus (i) and the resulting innervation pattern at the 360 AON (ii). Coordinates mark anteroposterior position from bregma.  investigation time between Day 9 and 10. Yet, following hM4D-(i) or ArchT-mediated (ii) inhibition of 400 hippocampal terminals at the AON, mice fail to exhibit this behaviour (hM4D experiment-hM4D-401