Gut vagal sensory signaling regulates hippocampus function through multi-order pathways

Novel object in context (NOIC) learning is a rodent model of contextual episodic memory. During day 1 of training, SDV and sham groups exhibited similar object discrimination indices (DIs; a measure of exploration time of both objects, Fig. 2a, left), indicating that baseline preference for objects A and B did not differ by group. On test day, SDV animals had impaired contextual episodic memory, demonstrated by a significantly reduced DI relative to sham animals (Fig. 2a, right; a DI above 0.50 means animals spent more time exploring the novel object for the test context). Repeated-measures analysis of variance (ANOVA) analyses across days revealed a significant day × group interaction (F[1,13] = 5.564, p = 0.0347), with Newman–Keuls’ post hoc analyses confirming a significant sham vs. SDV group difference on day 3 (p = 0.0047) but not on day 1.

Analogous to the SDV and sham group, CCK-SAP and SAP (controls) groups demonstrated similar DIs on day 1 of NOIC training (Fig. 2g, left). On test day, however, the CCK-SAP group had a significantly reduced DI relative to the sham group (Fig. 2g, right). Repeated-measures ANOVA analyses revealed a significant day × group interaction (F [1,15] = 6.496, p = 0.0223), with Newman–Keuls’ post hoc analyses confirming a significant SAP vs. CCK-SAP group interaction on day 3 (p = 0.0241) but not on day 1. Thus, hippocampal-dependent contextual episodic memory in rats requires intact GI vagal afferent signaling.

We developed a modified Barnes maze procedure to assess hippocampal-dependent spatial working memory in rats in which the spatial location of an escape hole is constant across two consecutive trials per day, but changes each subsequent training day. The index of learning on this task is the difference in the number of errors (exploration of escape holes that do not contain the escape box) from trial 2 to trial 1 for each individual training day. Results show that the SDV group was impaired in spatial working memory performance relative to shams (Fig. 2b, right). Repeated-measures ANOVA analyses of average difference in number of errors from trial 2 to trial 1 revealed significant group main effect across the 5 training days (F[1,19] = 6.8565, p = 0.0169). Individual training day analyses showed trends toward a group effect statistical significance on Day 2 (F[1,14] = 3.626, p = 0.0776, ANOVA), Day 4 (F[1,14] = 3.842, p = 0.0702, ANOVA), and Day 5 (F[1,14] = 3.555, p = 0.0803, ANOVA) compared with shams (Fig. 2b, left).

Similar to the SDV and sham groups, the CCK-SAP group was also impaired in this task relative to SAP controls, with repeated-measures ANOVA analyses of average error difference (T2–T1) revealing a significant group main effect across the 5 training days (F[1,13] = 8.66, p = 0.0114) (Fig. 2h, left). Individual training day analyses also indicated a significant group effect in error difference (T2–T1) on Day 2 (F[1,13] = 6.824, p = 0.0215, ANOVA) relative to sham (Fig. 2h, right). Overall, these findings indicate that spatial working memory in rats is impaired following GI vagal afferent ablation.

Deprivation intensity discrimination learning is a hippocampal-dependent procedure in which rats learned to use interoceptive energy status cues (0 vs. 24 h food restriction) as discriminative stimuli for a forthcoming food reinforcer⁶. Repeated-measures ANOVA over six 8-trial blocks of training (before SDV and sham surgeries) showed that both Group 0 + and Group 24 + learned to respond more (anticipatory food cue entries before food pellet delivery) during the last minute of test sessions under their reinforced compared to non-reinforced food restriction level (Fig. 2c), supported by a significant deprivation state × deprivation group interaction (F[1,22] = 135.54, p < 0.0001) and a significant block × deprivation state × deprivation group interaction (F[5,110] = 13.6535, p < 0.0001). When analyzing each block individually, Newman–Keuls’ post hoc analyses indicate that Group 0 + responded significantly more under 0 h compared with 24 h food deprivation during blocks 3–6 (all ps < 0.0017, Fig. 2c, left), whereas Group 24 + responded significantly more under 24 h compared with 0 h food deprivation during blocks 2–6 (all ps < 0.000178, Fig. 2c, right). Testing of deprivation intensity discrimination retention occurred following recovery from SDV and sham surgeries. Results confirmed that SDV lesions had no impact on retention of this type of interoceptive-based discrimination. In Fig. 2d, the effects of total SDV vs. sham surgery on food cup entry during the last minute of each test session performance when both Groups 0 + and 24 + were tested under 0 h and 24 h food deprivation are shown. Repeated-measures ANOVA revealed no significant surgery group effect (F[1,22] = 0.0123, p = 0.9126) or deprivation level × surgery group interaction (F[1,22] = 0.1665, p = 0.687). Animals in Group 0 + responded significantly more under 0 h than under 24 h food deprivation (Fig. 2d, left) and those in Group 24 + responded significantly more under 24 h than 0 h food deprivation (Fig. 2d, right), regardless of surgical group. These conclusions are supported by Newman–Keuls’ post hoc analyses revealing a significant 0 h vs. 24 h deprivation state interaction for Group 0 + SDV (p = 0.0028), Group 0 + sham (p = 0.004), Group 24 + SDV (p = 0.000589), and Group 24 + sham (p = 0.000894). These results suggest that in the absence of GI vagal signaling via SDV, non-vagal cues are sufficient to sustain the learned ability to use interoceptive energy status cues as discriminative stimuli for food reinforcement.

Social transmission of food preference (STFP) is a hippocampal-dependent procedure involving social-based learning using olfactory cues¹⁶. Percent paired flavor preference at testing was above 50% chance for both sham and SDV groups (Fig. 2e, left). Both sham and SDV group significantly preferred the paired flavored chow to the non-paired flavored chow when tested 24 h after the social interaction (Fig. 2e, right; 30 min cumulative food intake). One-way ANOVA analyses revealed no significant SDV vs. sham group effect for 30 min percent paired flavor preference (F[1,12] = 0.0225, p = 0.883). Paired Student’s t-test analysis indicated a significant preference for paired vs. non-paired flavored chow for both SDV (p = 0.014) and sham (p = 0.014) groups. Thus, GI vagal signaling has minimal impact on hippocampal-dependent social olfactory-based learning.

The Zero maze procedure is an established rodent model of anxiety-like behavior that is similar to the elevated plus maze procedure. ANOVA revealed no significant surgical group main effect between SDV and shams for time spent in the open section (F[1,19] = 0.0454, p = 0.833) (Fig. 2f; left) and number of open section entries (F[1,19] = 4.861, p = 0.731) (Fig. 2f; right). Similarly, ANOVA revealed no significant surgical group main effect between CCK-SAP vs. SAP for: time in open section (F[1, 15] = 0.0103, p = 0.92) and number of open section entries (F[1,15] = 0, p = 1.0; groups had equal means) in the zero maze test of innate anxiety (Supplementary Fig. 1a), as well as center zone distance (F[1,15] = 0.198, p = 0.663), number of center zone entries (F[1,15] = 0.6269, p = 0.441), and total distance (F[1,15] = 0.1784, p = 0.679) in the open field test (Supplementary Fig. 1b). Thus, observed contextual episodic and spatial working memory impairments observed in SDV and CCK-SAP rats are unlikely to be secondary to effects on anxiety-like behavior. Overall, there were three cohorts of SDV and sham animals, and terminal body weights did not differ between surgical groups in any cohort: cohort 1 underwent deprivation intensity discrimination task (sham: 413.54 ± 10.74, SDV: 396.78 ± 11.62; p = 0.329, ANOVA), cohort 2 tested in the Barnes task (sham: 391.19 ± 8.33, SDV: 365.09 ± 11.95; p = 0.095, ANOVA), and cohort 3 tested in NOIC and STFP tasks (sham: 371.69 ± 7.16, SDV: 354.47 ± 11.42; p = 0.199, ANOVA). Similarly, the cohort of SAP and CCK-SAP animals used in this study showed no significant group differences in terminal body weight (SAP: 394.16 ± 11.04, CCK-SAP: 383.28 ± 11.53; p = 0.508, ANOVA). Therefore, we conclude that the observed contextual episodic and spatial working memory impairments observed in SDV and CCK-SAP rats are unlikely to be secondary to surgical effects on body weight regulation.

Immunoblot analyses from dorsal HPC lysates revealed that total SDV reduced BDNF and DCX protein expression in the dorsal HPC relative to sham controls (Fig. 3a, b), with a significant main effect of surgical group observed for BDNF (F[1,14] = 4.609, p = 0.049, ANOVA) and DCX (F[1,14] = 5.5133, p = 0.034, ANOVA). Similar to SDV, the CCK-SAP significantly reduced levels of both proteins in the dorsal HPC relative to SAP controls (Fig. 3c, d). This conclusion is supported by one-way ANOVA analyses indicating a significant main effect of surgical group for both dorsal hippocampal BDNF (F[1,13] = 4.881, p = 0.0457) and DCX (F [1,14] = 5.494, p = 0.034) levels. On the other hand, immunoblot analyses from whole hypothalamic lysates revealed no significant group differences for both BDNF (F[1,13] = 0.26, p = 0.619, ANOVA) and DCX (F[1,14] = 0.4955, p = 0.493, ANOVA) levels in CCK-SAP vs. control SAP animals (Supplementary Fig. 2 a, b), indicating that GI vagal afferent ablation is unlikely to have systemic brain-wide impact of these neurotrophic (BDNF) and neurogenic (DCX) markers. Hypothalamic tissue was not collected in SDV and sham groups, and therefore analyses for these groups could not be included.

Following Shapiro–Wilk test of normality to confirm that the data were normally distributed [CCK-SAP vs. SAP controls: BDNF, W = 0.75377; DCX, W = 0.90026; DI, W = 0.96494; spatial working memory, W = 0.94624; for SDV vs. sham controls: BDNF, W = 0.88919; DCX, W = 0.92599; DI, W = 0.93305; spatial working memory, W = 0.91989], linear regression analyses were conducted to examine whether dorsal hippocampal BDNF and DCX protein expression were functionally correlated with HPC-dependent spatial working memory (Fig. 3e, f) and NOIC (Fig. 3g, h) task performance. For Barnes maze, analyses included all groups (sham, SDV, SAP, and CCK-SAP) and revealed that (1) dorsal HPC BDNF levels are negatively correlated with average error difference (T2–T1) (F[1,28] = 4.211, R2 = 0.1307, p = 0.0496) and (2) dorsal HPC DCX levels showed a trend toward significant correlation with Barnes performance (F[1,29] = 3.546, R2 = 0.1089, p = 0.0698). These results indicate that lower levels of BDNF and DCX are associated with poorer performance in the Barnes task. For NOIC, analyses of SAP and CCK-SAP groups only (as HPC lysates were not collected from SDV and sham rats that performed NOIC) reveal a significant positive correlation for both BDNF (F[1,13] = 5.277, R2 = 0.2887, p = 0.0389) and DCX (F[1,13] = 7.36, R2 = 0.3615, p = 0.0178) levels with DI, indicating that lower levels of BDNF and DCX are associated with poorer performance in contextual episodic memory.

Intraperitoneal (i.p.) injections of CCK-8 (8 μg/kg) increased the number of c-Fos protein immunoreactive cells expressed in the dorsal CA3 (dCA3; F[1,9] = 20.236, p = 0.001492, ANOVA) (Fig. 4a, c) and dentate gyrus (DG; F[1,9] = 37.917, p = 0.000167, ANOVA) (Fig. 4b, d) relative to i.p. saline treatment. In addition, i.p. CCK injections increased the number of labeled cells for c-Fos messenger RNA (fluorescent in situ hybridization, FISH) expressed in the dCA3 (F[1,9] = 53.093, p = 0.000046, ANOVA) (Fig. 5a) and DG (F[1,9] = 40.496, p = 0.000131, ANOVA) (Fig. 5b) relative to saline treatment, with 93.11% and 94.35% of c-Fos mRNA-positive cells in dCA3 (Fig. 5c) and DG (Fig. 5d) being VGLUT1 positive, respectively, and only 4.29% (Figs. 5e) and 8.44% (Fig. 5f) of c-Fos mRNA-positive cells being GAD2 mRNA positive, respectively.

The medial nucleus tractus solitarius (mNTS) is the first central nervous system (CNS) site to receive GI-derived vagal sensory input; however, the mNTS does not communicate monosynaptically with the HPC¹³. To identify regions of possible relay between the mNTS and HPC, a combination of retrograde and anterograde pathway tracing was used: unilateral iontophoretic injections of a retrograde pathway tracer targeted to the dCA3 (cholera toxin subunit B (CTB) AlexaFluor 488 (AF488) conjugated) (Fig. 6a, c) were combined with ipsilateral iontophoretic injections of an anterograde pathway tracer targeted to the mNTS (AAV1-hSyn-TurboRFP-WPRE-rBG) (Fig. 6b, d). Red fluorophore (red fluorescent protein (RFP) and Cy3 following immunohistochemistry (IHC)) anterogradely labeled axons originating from the mNTS were found in apposition to green fluorophore (AF488 following IHC) labeled cell bodies in the MS that were retrogradely labeled from dCA3 in each of the three animals that were confirmed as double hits in both injection sites (representative appositions in Fig. 6e–g), whereas such appositions were not observed in various control animals (n = 11) in which either (or both) injection site(s) were either undeterminable or adjacent to the intended target.

To further support this multi-order pathway, we utilized a novel dual-synaptic virus-based pathway tracing approach to examine whether mNTS neurons synpatically communicate to the HPC via a MS relay pathway¹⁸. The AAV2/1-hSyn-Cre drives Cre expression in first-order neurons infected at the injection site, as well as in second-order (but not third-order) neurons based on virion release from first-order axon terminals¹⁸. A unilateral iontophoretic co-injection of AAV2/1-hSyn-Cre (and CTB-488 to confirm injection site) targeted to the mNTS (at the level of the area postrema; Fig. 7a) was followed by a pressure injection of AAV1-CAG-FLEX-TdTomato (a Cre-dependent anterograde tracer) targeted to the MS (Fig. 7b, c). Results revealed robust axon terminal fields in the dCA3 (Fig. 7d, e) and DG (Fig. 7h, i) for each of the three animals that were confirmed as double hits in both injection sites. A schematic of the rostral-caudal distribution of axon terminal fields in the dCA3 (Fig. 7f, g) and DG (Fig. 7j, k) from a representative double hit animal are displayed. Axonal labeling in the dHPC was not observed for rats (n = 13) in which either the mNTS or MS injection was absent or adjacent to the targeted region. Overall, these results indicate the presence of synaptic connections to the dHPC from MS neurons that receive direct input from the mNTS.

Male Sprague–Dawley rats (Envigo; 320–450 g on arrival) were individually housed with ad libitum access (except where noted) to water and chow (LabDiet 5001, LabDiet, St. Louis, MO) on 12 h:12 h light/dark cycle (lights on at 08:00 h). All procedures involving animals were approved by the University of Southern California Institute of Animal Care and Use Committee.

Rats were habituated to liquid diet (Research Diets; AIN76A) for five days before surgery. Following a 24 h fast and under ketamine (90 mg/kg), xylazine (2.7 mg/kg), and acepromazine (0.64 mg/kg) anesthesia and analgesia (Metacam 2 mg/kg), the trunks of the subdiaphragmatic vagus nerve were transected as described previously⁵⁵. A midline abdominal incision was made and then the stomach was retracted caudally and the liver was retracted cranially to expose the esophagus. The dorsal and ventral branches of the vagus were then dissected from the esophagus. Each vagal branch was ligated twice with a surgical thread at an interval of 1–2 cm, and then cauterized between the ligatures. In sham surgeries, the trunks were exposed but the vagus nerve was not ligated or cauterized. The incision was then closed with running sutures along the abdominal wall and stop sutures along the skin. Rats were allowed to recover until liquid diet intake stabilized (at least 1 week) and were then returned to a standard chow diet. After behavioral testing, SDV was verified functionally with i.p. CCK-induced food intake reduction as described^56,57. Briefly, the functional verification consists of analysis of food intake following i.p. CCK-8 (2 μg/kg; Bachem) or saline injections (treatments given counterbalanced on separate days) after an overnight fast. SDV rats were included in the statistical analysis if CCK treatment resulted in a < 30% reduction of their food intake, as described^57,58. Of the three separate cohorts of rats that underwent SDV surgery and subsequent behavioral testing, four animals were removed from deprivation intensity discrimination analyses and one animal was removed from NOIC analyses based on these criteria.

CCK-SAP targets CCK receptor expressing cells, which are localized on vagal afferent neurons that specifically innervate the upper GI tract^59,60. As recently confirmed, CCK-SAP injection in the nodose ganglia (vagal afferent cell bodies) selectively eliminates ~ 80% of GI-derived vagal afferent signaling, while preserving colonic and supradiaphragmatic vagal sensory pathways, as well as all vagal efferents¹⁷. Twenty-four hours before surgery, rats were given 15 ml of condensed milk in addition to their normal ad libitum access to chow and water, and were fasted before lights went off (18:00 h). Twenty minutes before surgery, rats received an i.p. injection of atropine sulfate (0.05 mg/kg) and carprofen (5.0 mg/kg; Henry Schein), and then anesthetized with a ketamine (90 mg/kg), xylazine (2.7 mg/kg), and acepromazine (0.64 mg/kg) cocktail. A midline incision was made along the length of the neck. The vagus nerve was separated from the carotid artery with Graefe forceps until the nodose ganglion was visible and accessible. A glass capillary (20 μm tip, beveled 30° angle) attached to a micromanipulator was used to position and puncture the nodose ganglion and 1 µl volume of CCK-SAP (250 ng/µl) or SAP (250 ng/µl) was injected with a Picospritzer III injector (Parker Hannifin) at two sites: 0.5 µl rostral and 0.5 µl caudal to the laryngeal nerve branch. The same procedure was performed for both nodose ganglia on either side before the skin was closed with sterile and stop sutures along the skin. Rats were returned to their home cage and deprived of water for 6 h and food overnight. The post-op care was optimized to avoid excessive weight loss post surgery and increase survival rate, as previously described⁶¹. The schedule was as follows: Day 1 post-op rats received carprofen (5.0 mg/kg; SQ) and were given ad libitum access to condensed milk; day 2 post-op rats received mash (10 g powdered chow mixed with 20 ml condensed milk diluted as described above); day 3 post-op rats received mash and solid chow pellets; day 4 and onwards, rats were given ad libitum access to chow. After behavioral testing, CCK-SAP was verified functionally with i.p. CCK-induced food intake reduction. The functional verification modeled that of the SDV approach described above and that published for CCK-SAP in¹⁷. Based on this verification test, four CCK-SAP rats were removed from all analyses.

Three separate cohorts of rats underwent SDV (or sham) surgery and subsequent behavioral testing (described below), beginning 7 days post surgery. Cohort 1 underwent deprivation intensity discrimination training (described below) in which rats were assigned to one of two groups (group assignment matched based on body weight): Group 0 + (n = 16) or Group 24 + (n = 11). After asymptotic discrimination was reached, animals were then subdivided (matched based on body weight and performance over the last 8-trial block of training) into two additional groups to receive SDV (Group 0 + , n = 7; Group 24 + , n = 5) or sham (Group 0 + , n = 8; Group 24 + , n = 6) surgery 4 days after the last training day. After 7 days of post-surgery recovery, the animals were tested on deprivation discrimination performance for one 8-trial block (see Supplementary Table 1). Seven days after deprivation discrimination testing, animals in Cohort 1 were tested in the zero maze task. Cohort 2 (SDV n = 8; sham n = 8) was tested in the spatial working memory task (5 days, described below) 7 days post surgery. Five to 6 days later, dHPC tissue was harvested from Cohort 2 for immunblot analyses of BDNF and DCX (SDV, n = 8; sham, n = 8) (see Supplementary Methods). Cohort 3 (SDV n = 6, sham n = 9) was tested in the NOIC (5 days) task 14 days post surgery, followed by STFP 21 days post surgery (STFP; 3 days). For STFP (described below), the SDV and sham groups were observers, whereas demonstrator rats (non-operated, n = 8) were housed in a separate room from the observers. Based on results from our SDV experiments, CCK-SAP (n = 9) and SAP (control, n = 8)] rats were tested in the Barnes task (beginning 7 days post surgery; matching the timeline of post surgery SDV Barnes testing) followed by the NOIC task (beginning 14 days post surgery; matching the timeline of post surgery SDV NOIC testing). Seven days after NOIC testing, animals were tested in the zero maze task followed by the open field task 3 days later. Five to 6 days later, dHPC tissue was harvested for immunblot analyses of BDNF (CCK-SAP, n = 8; SAP, n = 7) and DCX (CCK-SAP, n = 8; SAP, n = 8) (see Supplementary Methods; see Supplementary Fig. 3 for uncropped scan of CCK-SAP vs. SAP BDNF blot). Groups were assigned matched according to body weight at the beginning of each experiment. For all video analyses for behavioral variables, experimenters were blinded to group assignments of the animals.

NOIC incorporates object recognition within a specific context, a test of contextual episodic memory. Procedures followed those previously described^15,62. Rats undergo 2 days of habituation: half of the rats (counterbalanced within surgical groups) are able to freely explore Context 1, a semi-transparent box (15 in W × 24 in L × 12 in H) with orange stripes, for 5 min, whereas the other half are habituated to Context 2, a gray opaque box (17 in W × 17 in L × 16 in H). The following day, groups are switched and habituated to the other context under the same conditions. Twenty-four hours later, on day 1 of NOIC, each animal is exposed to two distinct objects: a 500 ml jar filled with blue water (object A) and a square glass container (object B) in Context 1. The next day, the animals are placed in Context 2 with duplicates of object A. Twenty-four hours later the animals are placed in the previous day’s location (Context 2) with objects A & B and investigation of both objects is measured. Each day consists of 5 min sessions and contexts are cleaned with 10% ETOH between each animal. Exploration is defined as sniffing or touching the object with the nose or forepaws. The task scoring involves recording of the time spent exploring an object novel to the context vs. time spent exploring an object familiar to the context, and calculation of a novelty or DI is based on these measurements and calculated as [Exploration of object B/(Exploration of object A + object B)]. Rats normally will preferentially investigate object B given that object B is a familiar object that is now presented in a novel context. NOIC is hippocampal dependent, as performance is impaired in rats with reversible muscimol-induced inactivation of the dorsal HPC CA1¹⁵.

Spatial working memory is a component of short-term memory involving encoding and remembering information about one’s environment and spatial orientation. The spatial working memory task used in the present study involves an elevated white circular Barnes maze (Diameter: 122 cm, Height: 140 cm) with 18 holes evenly spaced around the outer edge of the table’s circumference. A hidden black escape box (38.73 cm × 11.43 cm × 7.62 cm) is placed under one of the holes. There are four sets of visuospatial cues (black and white stripes, a white circle, a red triangle, and an assortment of irregular shapes) placed on each of the walls surrounding the table. One day after the habituation session, spatial working memory training begins. Rats utilize visuospatial cues to learn the location of the escape box, while being exposed to mildly aversive stimuli (120 W bright overhead light, 75 db white noise), which motivates them to find the escape box to avoid these stimuli⁶³. Each rat receives two trials per day for five training days. The two trials are separated by 2 min (during which the maze is cleaned and rotated). Importantly, the escape box is placed in the same location for both trials that occur on each individual training day, but is placed in a new location at the beginning of the first trial for each subsequent training day. Errors are measured as investigation of holes that do not contain the escape box, and spatial working memory is measured as the difference in the number of errors from trial 2 to trial 1 on each separate training day (which is then averaged over the five training days). Control rats (Controls, Fig. 2b below) show improved performance (fewer errors) on the second trial of each training day compared to the 1^st, indicating that the spatial location of the escape hole can be integrated into working memory capacity in control rats.

Deprivation intensity discrimination involves learning to use different levels of food restriction as interoceptive discriminative stimuli for sucrose reinforcement. The behavioral paradigm follows that from previous publications^6,31,64. Training: Rats are divided into Group 0 + or Group 24 + and food deprivation levels alternate each day between 0 and 24 h for the entire experiment. Group 0 + receives a reinforcement of five sucrose pellets (45 mg, Product F06233↗, Bio-Serv) at the end of 4 min training sessions that take place under 0 h food deprivation, and receive no pellets during training sessions that take place under 24 h deprivation. Group 24 + receives the opposite contingency between food deprivation level and pellet delivery. During sessions in which rats were trained under their non-rewarded deprivation condition, the feeders operated but no pellets were delivered. The rats were then given 2 min to consume the pellets before being removed from the conditioning chambers and returned back to their home cages. Training sessions always occurred at the same time of day (10:00 h), but not every day to avoid a single-alternating schedule of pellet delivery. The index of learning in the deprivation intensity discrimination task is food-anticipatory responding (head pokes in sucrose pellet delivery location; detected with photobeam interruptions/magazine entries) during the 1 minute prior to pellet delivery. More specifically, the mean of the percentage of the three 20 s intervals during the last minute of the session in which the magazine photobeam was interrupted was calculated as the dependent variable (as we’ve previously described⁶). Rats with HPC lesions are impaired in learning and retention of this discrimination problem^6,65, indicating that the HPC is critical for this type of learning process. Training consisted of six eight-trial blocks (eight trials for each deprivation state per training block), and testing (same procedures as training) consisted of one eight-trial block 7 days after surgical recovery.

STFP learning involves the utilization of olfactory cues experienced during a social encounter to guide subsequent preference for a scented food. STFP task procedures were adapted from¹⁶. Briefly, demonstrators (non-treated rats) and observers (surgicated rats, controls and experimental) are first habituated to an unscented powdered rodent chow [LabDiet 5001 (ground pellets)] overnight. Twenty-four hours later, observers are then assigned to demonstrators (1 demonstrator assigned for 2 observers, counterbalanced by observer group) and are habituated to social interaction by being placed in a social interaction arena (23.5 cm W × 44.45 cm L × 27 cm H; clear plastic bin with Sani-chip bedding covering the bottom) to interact with each other for 30 min. The next day following a 23 h period of food restriction for all rats, demonstrators are given access to one of two flavors of powdered chow (flavored with 2.5% marjoram or 0.5% thyme; counterbalanced) for 30 min in a separate room from observers. Consistent with previous published work⁶⁶, our pilot studies demonstrated that animals equally prefer these flavors of chow. The demonstrator is then immediately placed in the social interaction arena with one of their two assigned observer rats and allowed to socially interact for 30 min, followed immediately by a 30 min interaction with the second assigned observer rat (observer rat order counterbalanced across experimental groups). Immediately after social interactions, observers are then returned to their home cages and allowed to eat ad libitum for 1 h. Twenty-four hours later, the 23 h food-restricted observers are tested for STFP by placing two jars in the home cage: one containing paired flavor (flavored chow that was given to the demonstrator) and the other with non-paired flavor (novel flavored chow not given to the demonstrator animal). Rats are then allowed to eat for 1 h and % preference for the paired flavor is calculated by: 100*Demonstrator-paired flavored chow intake/Demonstrator + Novel flavored chow intake. Cumulative food intake for paired and non-paired flavors are calculated. Untreated control animals learn to prefer the demonstrator-paired flavor based on social interaction and olfactory food cues from the breath of the demonstrator rat; however, lesioning the dorsal HPC impairs consolidation of this preference learning.

Seven days following memory test procedures, all groups of rats (SDV, sham, CCK-SAP, and SAP) were tested in the zero maze task to examine anxiety-like behavior. The zero maze is an elevated circular track, divided into four equal length sections. Two sections were open with 3 cm high curbs, whereas the two other closed sections contained 17.5 cm high walls. Animals were placed in the maze for 5 min before being returned to the home cage. After the trial, the maze was cleaned with 10% ethanol. Innate anxiety was measured as the number of open section entries and total time spent in open sections, defined as the head and front two paws in open sections. Previous studies have demonstrated that animals with high levels of innate anxiety (based on anxiolytic and anxiogenic experimental manipulations) spend less time in the open sections compared with animals with low levels of innate anxiety⁶⁷.

Three days following zero maze procedures, the CCK-SAP rats were then tested in the open field task, another behavioral paradigm used in rodents to evaluate innate anxiety-like behavior⁶⁷. Open field procedures are derived from⁴². The apparatus consists of a gray Plexiglas arena (60 cm × 56 cm), and a designated center zone within the arena (19 cm × 17.5 cm), placed under diffused lighting (30 lux). Animals are placed in one of the bin’s four corners and allowed to freely explore for 10 minutes. The apparatus is cleaned with 10% ethanol between rats. Innate anxiety was measured as the number of entries into the center zone, distance moved in center zone, and total distance moved in the entire arena. Previous work has shown that animals with innate anxiety-like behavior will spend less time in the center zone and more time in close proximity to the walls compared with animals with low levels of innate anxiety⁶⁷.

Nonsurgicated rats (n = 11) received i.p. injections of either saline (n = 5) or CCK (n = 6) (CCK-8, 8 μg/kg; Bachem), 90 min before perfusion and tissue was collected and processed as described above for IHC and FISH processing. IHC detection of c-Fos was performed using rabbit anti-c-Fos primary antibody (1:500, Cell Signaling, Catalog number: 2250 s) followed by a donkey anti-rabbit IgG-AlexaFluor AF488 secondary antibody (1:500, Jackson Immunoresearch, RRID: AB_2340619). In the same animals, mRNA detection followed FISH procedures (see Supplementary Methods) and used c-Fos (ACD, Catalog number: 403591), VGLUT1 (ACD, Catalog number: 317001), or GAD2 (ACD, Catalog number: 435801) probes. Representative images and quantification for c-Fos protein and mRNA expression in dCA3 and DG obtained from both i.p. saline and CCK injected animals were confined to Swanson Atlas level 28–30⁶⁸.

Neural pathway tracing experiments utilized two tracing techniques: co-injection monosynaptic neural tracing to identify second-order relay connections (COIN)⁴⁹ and Cre-mediated dual-synaptic anterograde tracing (below)¹⁸. Iontophoresis was performed using a precision current source (Digital Midgard Precision Current Source, Stoelting) as described previously⁶⁹. Rats (n = 14) received unilateral iontophoretic injections of AAV1-TurboRFP (AAV1-hSyn-TurboRFP-WPRE-rBG; Penn Vector Core, Catalog number: V5574L) targeting the mNTS and, in the same animal, CTB-488 (CTB, AlexaFluor 488 conjugate; ThermoFisher, Catalog number: C22841↗) targeting the dCA3. Following a 12-day survival period, animals were fixation-perfused and tissue was collected and processed as described in Supplementary Methods. The native fluorescent signal was amplified using a cocktail of mouse anti-CTB (1:500, Abcam, Catalog number: ab62429) and rabbit anti-RFP (1:2000, Rockland, RRID: AB_2209751) primary antibodies followed by a cocktail of donkey anti-mouse IgG-AF488 secondary antibody (1:500, Jackson Immunoresearch, RRID: AB_2341099) and donkey anti-rabbit IgG-Cy3 (1:500, Jackson Immunoresearch, RRID: AB_2307443). Detection of AAV1 and CTB was by visualization of red (TurboRFP and Cy3) and green (AlexaFluor 488) fluorescence. In 3 of the 14 animals, injection sites were confined to the mNTS (AAV1) and dCA3 (CTB) in the same animal. Similar neuroanatomical labeling was observed in each of these three animals and representative images were obtained from one of these animals. Experimental controls (n = 11) were included when neither anterograde nor retrograde labeling was observed in the MS following injection sites adjacent to (respectively) the mNTS or dCA3 that did not include these regions.

Rats (n = 16) received a unilateral iontophoretic co-injection of AAV-Cre (4:5; AAV1-hSyn-Cre-WPRE-hGH; Penn Vector Core, Catalog number: CS1087) and, to determine injection site, CTB-488 (1:5; ThermoFisher, Catalog number: C22841↗) in the mNTS. Next, the rats received a 200 nl pressure injection of AAV-Flex-tdTomato (diluted 1:2 in 0.1 M sodium phosphate-buffered saline, pH 7.4, AAV1-CAG-Flex-tdTomato-WPRE-rBG; Penn Vector Core) in the MS. Following a 3-week survival period, animals were perfused and tissue was collected and processed as described in Supplementary Methods. AAV1-Flex-TdTomato and CTB-488 was detected based on a combination of native fluorescence of RFP and green fluorescent protein, respectively, and amplification of signal using the same cocktail of primary and secondary antibodies as described in the co-injection neural tracing section. Representative images were obtained from one of the three animals with both CTB-488 and AAV1-Flex-tdTomato injection sites that were predominantly confined to the mNTS (CTB) and MS (AAV1), respectively (similar labeling was observed in the other two animals that were double hits). The specificity of this connection was further supported by the absence of axon labeling in dCA3 and DG following MS injections of AAV-FLEX-TdTomato alone (n = 3; w/AAV2/1-hSyn-Cre injections adjacent to mNTS), mNTS injections of AAV2/1-hSyn-Cre alone (n = 4; w/AAV-FLEX-TdTomato injections adjacent to the MS), or in rats in which either injection site was off target (misses, n = 6).

Terminal body weights and zero maze performance data were analyzed using one-way ANOVA, with surgery group as a between-subjects factor for both SDV and CCK-SAP groups. NOIC was analyzed using repeated-measures ANOVA with surgical group (between-subjects) and testing day (within-subjects) as factors for both SDV and CCK-SAP experiments, with Newman–Keuls post hoc comparisons to evaluate group differences on each separate day. Spatial working memory data (Barnes maze) were analyzed using repeated-measures ANOVA across training days with surgical group as a between-subjects variable and training day as a within-subjects variable. The deprivation intensity discrimination test data were analyzed using repeated-measures ANOVA with surgical and deprivation (0 + , 24 + ) assignment group as between-subjects factors, with deprivation state as a within-subjects factor, and Newman–Keuls post hoc tests for individual group × deprivation state comparisons. The STFP percent paired flavor preference data was analyzed using one-way ANOVA with surgical group as a within-subjects factor. Postmortem immunoblot for both SDV and CCK-SAP groups were analyzed using ANOVA with surgical group as a between-subjects factor. Linear regressions were used to calculate p-values, R² goodness-of-fit, 95% confidence bands of the best-fit line, and linear equations between NOIC and Barnes data with protein level data by surgical group. All groups (SDV, CCK-SAP, and their controls) were analyzed for linear regression of HPC protein expression vs. Barnes performance, whereas only CCK-SAP and their controls were analyzed for HPC protein expression vs. NOIC performance (as immunoblot HPC tissue was not extracted from SDV and sham/controls that underwent NOIC testing). Normality was confirmed for all analyses using Shapiro–Wilk’s test. c-Fos protein and mRNA expression levels in dCA3 and DG were analyzed using one-way ANOVA with drug treatment group as a between-subject factor. All statistical analyses were performed using the statistical software Statistica (Version 7; Statsoft) and linear regressions analyses were performed using Prism 7 (GraphPad) statistical software. Sample size was chosen based on a priori power analyses (conducted in Statistica V7) to ensure sufficient power to detect a pre-specified effect size. Pre-established exclusion criteria used was the Grubbs test for outliers (conducted in Prism 7). ANOVAs were followed by Newman–Keuls post hoc comparisons when significant main effects or interactions were obtained. Results are presented as mean ± SE. Statistical significance was set at p< 0.05.

The data that support the findings of this study are available in the Open Science Framework Repository [10.17605/OSF.IO/UYMBQ]⁷⁰.

The authors declare no competing interests.

The data that support the findings of this study are available in the Open Science Framework Repository [10.17605/OSF.IO/UYMBQ]⁷⁰.