Brain-wide connectivity and novelty response of the dorsal endopiriform nucleus in mice

SUMMARY

The dorsal endopiriform nucleus (EPd) is a cortical subplate structure within the piriform cortex that shares similar developmental origins to those of the claustrum. Although implicated in epilepsy and olfaction, the EPd’s connectivity and function remain largely unclear due to the lack of specific molecular markers. Our recent mapping study identifies the oxytocin receptor (Oxtr) as highly enriched in the EPd. Immunohistochemical and spatial transcriptomic analyses confirm Oxtr enrichment and a distinct molecular profile of the EPd compared to the claustrum. Whole-brain input-output mapping of EPd-Oxtr neurons unveils extensive bidirectional connections with ventral brain regions, orchestrating circuits regulating olfaction, internal state, and emotion. Furthermore, in vivo miniscope recordings reveal that EPd-Oxtr neurons exhibit high baseline activity during exploration, with a sharp decrease in response to novel stimuli. Together, these findings suggest that EPd-Oxtr neurons integrate interoceptive and exteroceptive signals, contributing to internal state regulation and behavioral adaptation to novel environmental cues.

Graphical Abstract

In brief

Manjila et al. identify the dorsal endopiriform nucleus (EPd) as a distinct cortical subplate structure enriched with oxytocin receptors (Oxtrs). EPd-Oxtr neurons connect bidirectionally with olfactory and limbic regions. They exhibit high baseline activity that decreases with novel stimuli, suggesting a role in regulating internal states and processing environmental novelty.

INTRODUCTION

The dorsal endopiriform nucleus (EPd) is a thin, elongated structure along the anterior-posterior axis in the mouse brain that is situated medially to the piriform cortex (PIR).^1,2 Moreover, the EPd is located underneath the claustrum (CLA) in the dorsoventral axis and shares a similar developmental origin to that of the CLA.³ The EPd is a conserved structure in rodent brains and higher mammals and isa part of the CLA complex.⁴ While the CLA forms extensive connections with iso(neo)cortical areas,^5-9 previous studies suggest that the EPd exhibits widespread connectivity with the olfactory areas and limbic cortices.^2,10-12 Although the functions of the EPd are largely unknown, a few studies have demonstrated its potential involvement in multisensory (olfactory and gustatory [GU])¹² and limbic information processing² and conscious perception.¹³ Moreover, dysfunction of the CLA complex, including the EPd, has been implicated in many neurological conditions, such as epilepsy, autism, and neurodegenerative disorders.^14-17 Despite the potential significance of the EPd, very little information is known about this region, including its basic neuroanatomy, molecular and cellular characteristics, and function in regulating behavior.

One of the major challenges in examining the EPd is the lack of clear molecular markers for targeted studies in the area. While we investigated oxytocin receptor (Oxtr) expression throughout the whole mouse brain,¹⁸ we serendipitously found that Oxtr is densely expressed in the EPd. In this study, we utilized in situ hybridization and spatial transcriptomics and confirmed that Oxtr is indeed richly expressed in the EPd. This offers opportunities to perform initial groundwork to characterize the cellular architecture of the EPd, its connectivity, and its potential functions.

Detailed anatomical wiring diagrams have been crucial in gaining accurate anatomical connectivity of target areas, which helps to infer relevant functions. For instance, our Oxt wiring diagram helps to identify nine functional circuits where Oxt signals can modulate different functions.¹⁹ Here, we perform detailed anatomical studies to investigate the input-output connectivity of the EPd using Oxtr-Cre mice. Our findings unveil extensive bidirectional connections with ventral brain regions, particularly limbic and olfactory areas, in stark contrast to neighboring CLA connectivity patterns largely targeting the dorsal brain regions, including the isocortex.^5,6,9,20 Moreover, we found unidirectional monosynaptic input from specific thalamic, midbrain, and hindbrain areas that are linked to the limbic system and alertness. Lastly, our results show that EPd-Oxtr neurons exhibit high baseline neural activity during exploratory behavior and significantly decrease their activities when exposed to novel stimuli, including novel social exposure. Collectively, our findings imply that the EPd-Oxtr neurons are well positioned to regulate behavior driven by interoception vs. exteroception.

RESULTS

Molecular signature of the EPd in the mouse brain

Cell types with distinct gene expression patterns have been crucial in defining specific anatomical structures.^21-25 To identify shared and distinct gene expression signatures in the EPd against neighboring areas, we performed multiplexed error-robust fluorescence in situ hybridization (MERFISH)-based spatial transcriptomics using known cell-type- and area-specific markers such as Nr4a2 (also called Nurr1),^26-28 which help to identify the EPd, the CLA, and the neighboring cortical areas²⁹ (n = 3, 2 males and 1 female, 2–3 months old; Figures 1A and 1B) (see STAR Methods for details). We performed precise image registration to align the Allen Common Coordinate Framework v.3 (CCFv3) labels³⁰ to an individual section using striatal and cortical layer-specific markers (see STAR Methods and Figure S1 for details). Next, we performed unbiased gene clustering analysis with known cell-type markers and found 20 different cell-type clusters³¹ (Figure 1C). When anatomical areas are overlaid in the uniform manifold approximation and projection (UMAP), we found that the CLA and EPd formed distinct clusters (Glut_4 for CLA and Glut_10 and Glut_11 for EPd; Figures 1D and 1E; for statistics, see Tables S5 and S6). Visualizing these clusters in the raw data further showed that most neurons in Glut_4, Glut_10, and Glut_11 were located within the CLA and EPd (Figures 1F and 1G). Additionally, the neighboring regions—the PIR and the agranular insular cortex (AI)—exhibited enrichment in separate clusters, distinct from those of the CLA and EPd (Figures S2A and S2B). We compared the cell-type compositions of the CLA, EPd, AI, and PIR and found that the EPd consists of approximately equal proportions of neuronal and non-neuronal cell types, with ~45% excitatory neurons, ~5% inhibitory neurons, and the remaining 50% comprising non-neuronal cells (Figure S2C). Then, we examined genes differentially enriched in the CLA-EPd area. We found that genes like Cplx3, Nxph3, and Npsr1 are more significantly expressed in the EPd (Figures 1H and 1I; statistical analysis in Table S3), while Synpr, Gnb4, and Gng2 are enriched in the CLA (Figures S2D and S2G). The AI and PIR showed different sets of genes enriched compared to the CLA-EPd (Figures S2E, S2F, S2H, and S2I; Table S3), further supporting that the EPd has a distinct gene expression pattern compared to neighboring areas.

Figure 1. Molecular signature of the EPd with Oxtr enrichment in the mouse brain.

(A) Coronal cryosectioning from the anterior EPd (bregma: +0.5) for spatial transcriptomics.

(B) Enrichment of Nr4a2 (Nurr1), a cortical subplate marker, with elevated expression in the CLA and EPd.

(C) UMAP clustering of all cells from the agranular insular area (AI), CLA, EPd, and piriform area (PIR) (n = 3 anterior EPd sections from 3 mice, 2–3 months old).

(D and E) Cells from the EPd and CLA overlaid on the UMAP. The dotted line highlights clusters Glut_4 and Glut_10 and Glut_11, primarily composed of excitatory neurons from the CLA and the EPd, respectively.

(F) Cells from Glut_10 and Glut_11 clusters were overlaid on raw data, showing that the cells within these clusters label the EPd.

(G) Most of the cells from the Glut_4 cluster fall within the CLA borders.

(H) Representative images of genes enriched in the EPd compared to the CLA, revealed through spatial transcriptomics of the same coronal section.

(I) Dot plots showing genes that are differentially expressed between the EPd and CLA. p values and associated statistics can be found in Table S3 (n = 3 sections from 3 mice, one section per mouse).

(J) Oxtr is enriched in the EPd, expressed with another EPd marker, Cplx3 (n = 1 mouse).

(K) Independent validation by RNA in situ hybridization, demonstrating colocalization of Oxtr, vGlut1, and vGAT.

(L) Quantification of vGlut1+ and vGAT+ neurons in the EPd (n = 6 mice, with at least 4 sections per mouse). *p values were obtained by paired t test followed by Wilcoxon test (see Table S4 for details). Data are represented as mean ± SD.

(M) Quantification of the colocalization of Oxtr+ neurons with vGlut1 and VGAT (n = 6 mice, at least 4 sections each). *p values were obtained by paired t test followed by Wilcoxon test. Data are represented as mean ± SD.

(N) Oxtr-Cre mice were injected with AAV-Syn-FLEX-GCaMP6m, resulting in localized expression in the EPd.

(O) Anterior EPd sections from Oxtr-Venus heterozygous mice (n = 3) were analyzed after cholera toxin B subunit (CTB) injections into the retrosplenial cortex. CTB-labeled neurons were observed in the CLA, where Oxtr-Venus expression was absent. In contrast, Oxtr-Venus expression was confined to the EPd, with no labeling detected in the CLA.

Oxtr+ neurons are enriched in the EPd

A lack of clear molecular markers to label EPd neurons presents a major challenge to understanding the anatomical and functional organization of the EPd. In our previous Oxtr mapping in the whole mouse brain, we coincidentally identified enriched expression of Oxtr(+) neurons in the EPd.¹⁸ To ascertain the previous finding, we conducted spatial transcriptomics experiments employing Oxtr and other genes associated with anatomical markers, with a focus on the anterior CLA-EPd complex (~bregma: +0.7 mm) in an adult C57BL/6J mouse (n = 1 male, 3 months old; see STAR Methods for details). We confirmed the robust expression of Oxtr, primarily in the EPd, which was also marked by the high expression of the EPd marker Cplx3²⁸ (Figure 1J). Both Oxtr and Cplx3 were expressed in the EPd, with Oxtr showing a more lateral distribution and Cplx3 a more medial distribution within the EPd (Figure 1J, right). We also analyzed two spatial transcriptomics datasets from the Brain Image Library^32,33 corresponding to the anterior EPd (aEPd) alongside our processed sample to quantify Oxtr expression in the EPd. All three samples consistently showed enrichment of Oxtr expression in the EPd (Table S4). Additionally, we identified that Oxtr-expressing neurons in the EPd are predominantly glutamatergic (vesicular glutamate transporter 1 [vGlut1]+) (Figure S2J; Table S4). To quantify this, we employed RNA in situ hybridization (RNA scope) (Figures 1K and 1M). We found that most neurons in the EPd are excitatory, with about 85% expressing vGlut1, while the remaining 15% are inhibitory neurons expressing vesicular GABA transporter (vGAT) (Figure 1L; Table S4). Approximately 40% of all neurons (total of excitatory and inhibitory neurons) in the EPd were Oxtr+ (Table S4). Among the Oxtr+ neurons, about 90% were excitatory (vGlut1+), and the remaining 10% were inhibitory (vGAT+) (Figure 1M), conforming to the spatial transcriptome data (Figure S2I; Table S4) and a previous report.³⁴

Next, we tested whether Oxtr-Cre mice can be used as a transgenic animal tool to selectively label EPd neurons when combined with stereotaxic injection of conditional reporter virus (adeno-associated virus [AAV]-Syn-FLEX-GCaMP6m) into the EPd area (Figure 1N). Indeed, viral reporter gene expression was confined in the EPd (Figure 1N). To further confirm that Oxtr expression is primarily confined to the EPd with minimal expression in the CLA, we retrogradely labeled CLA neurons by injecting cholera toxin B into the retrosplenial cortex (RSP) of Oxtr-Venus mice that label Oxtr+ neurons faithfully^18,35 (Figure 1O). Our analysis confirmed that Oxtr-Venus expression was absent in the CLA, supporting the use of Oxtr as a specific marker for the EPd (Figure 1O). The CLA is also characterized by a strong parvalbumin (PV) neuronal plexus.³⁶ Hence, we stained for PV in Oxtr-Cre;Ai14 mice and found that the CLA, which is clearly delineated by the PV plexus, lacks Oxtr-expressing neurons (Figure S2K). Based on these findings, we selected Oxtr-Cre mice as a reliable tool to investigate the neuroanatomical and functional organization of the EPd.

EPd-Oxtr neurons mainly project to the ventral half of the brain

Elucidating long-range output connectivity can help infer anatomical areas under the control of the EPd for various functional roles.^37,38 Hence, we sought to create a detailed anterograde projection map originating from EPd neurons labeled by Oxtr-Cre mice and analyze their functional significance.

We injected a Cre-dependent AAV (AAV2-CAG-Flex-eGFP), covering different anterior-posterior areas of the EPd (n = 9 animals, 6 males and 3 females, one injection per animal; Figures 2A-2C). We used serial two-photon tomography (STPT) to image the whole brain at single-cell resolution.^19,39 We detected projection signals using machine learning algorithms (using ilastik) and registered individual samples onto the CCFv3 (using Elastix) (Figures 2D and 2E; Video S1). Lastly, maximum projection data from all samples were used to represent the efferent output from the EPd-Oxtr neurons (Figures 2E and 2F; Video S1). Our analysis revealed abundant projections from EPd-Oxtr neurons to the ventral half of the telencephalon, with sparse projections to the diencephalon and no projections to the midbrain and the hindbrain (Figures 2D-2F). To understand if there is any topographical organization within EPd-Oxtr neurons, we subdivided injections into the anterior and posterior parts of EPd (Figure S3A). We found that the aEPd-Oxtr neurons project more to the anterior brain regions, while the posterior EPd (pEPd)-Oxtr neurons project more to the caudal brain regions (Figure S3B). We grouped the areas with long-range projections based on their known function to understand downstream circuits under the control of the EPd.

Figure 2. EPd-Oxtr neurons primarily project to the ventral half of the brain.

(A) Schematic of AAV-CAG-Flex-GFP injections into the EPd of Oxtr-Cre mice.

(B) Locations of injections along the anterior-posterior axis of the EPd and the size of individual injections. (n = 9 mice).

(C) Top: representative image of injection site. Bottom: higher magnification of injection site showing cells labeled within the EPd.

(D) Examples of long-range projections (green) from EPd-Oxtr neurons. The bottom image shows high-magnification images of the white boxed areas in the top image.

(E) Maximum projection outputs (from all injections, n = 9 mice) from the EPd, registered to the CCFv3.

(F) Bar graph representing major output areas throughout the brain. The area-normalized projection represents the ratio of the area covered by the projection signal to the total area of the region of interest (ROI), shown as a percentage. The full names of abbreviations according to CCFv3 are listed in Table S1.

The major projection areas were olfactory regions, particularly the main olfactory epithelium pathway for processing volatile odor cues, including the main olfactory bulb (MOB) and its direct downstream areas (e.g., the anterior olfactory nucleus [AON], the PIR, and the olfactory tubercle [OT]).⁴⁰ In contrast, the vomeronasal pathway to process mechanical cues (e.g., pheromone) received no projections (e.g., the accessory olfactory bulb [AOB]) or sparse projections (e.g., cortical amygdala posterior-medial [COApm]) from the EPd-Oxtr neurons.⁴¹ Another area with prominent projections is the basal forebrain area (magnocellular area [MA], substantia innominata [SI], nucleus of the diagonal band [NDB], and medial septum [MS]), which has been implicated in attention and reward processing.⁴² Moreover, significant projections were observed in the limbic areas, including the central amygdala (CEA), basolateral amygdala (BLA), medial amygdala (MEA), bed nucleus of stria terminalis (BNST), infralimbic cortex (ILA), and ventral subiculum (SUBv), that regulate affective behavior.⁴³ In contrast, the majority of isocortical and dorsal hippocampal areas that regulate cognitive behavior receive very sparse or no projections. These output connectivity patterns suggest that EPd-Oxtr neurons may play a significant role in regulating olfactory and limbic information processing.

EPd-Oxtr neurons project selectively to deep layers of the lateral association and medial prefrontal cortices

Previous projection mapping of the CLA has revealed abundant isocortical projections to all layers.⁵ To compare how EPd projections differ from CLA projections, we analyzed layer-specific projection patterns to all the isocortical areas (Figure 3). Our analysis revealed that the majority of isocortical areas receive sparse projections, except the lateral association areas (AI, GU, perirhinal [PERI], visceral cortex [VISC], and ectorhinal [ECT] areas) and the medial prefrontal cortex (prelimbic [PL], ILA, and orbital [ORB]) (Figures 3A-3C). Notably, the EPd-Oxtr neurons send little to no projections to the RSP, to which the CLA heavily projects⁵ (Figure 3C). Additionally, the EPd-Oxtr neurons only project to deep layers, predominantly layer 6 (L6), of the target areas (Figures 3C and 3D), which contrasts sharply with the CLA, which projects to all layers of the isocortex.⁵ L6 contains neurons with cortical thalamic projections.^44,45 Hence, selective L6 innervation by EPd-Oxtr neurons may play a role in regulating corticothalamic output largely from limbic cortical areas.⁴⁶

Figure 3. EPd-Oxtr neurons selectively project to the deep layers of the lateral association and medial prefrontal cortices.

(A) Isocortical flatmap showing the CCFv3 anatomical regions and border lines.

(B) Isocortical flatmap depicting the projection density from the EPd (n = 9 mice).

(C) Heatmap of area-normalized projections (percentage of area with signal/total area) from the EPd to isocortical layers. The full names of abbreviations according to the CCFv3 are listed in Table S1.

(D) Examples of raw images showing long-range projections (green) from the EPd to the isocortex. The right image shows high-magnification images of the red boxed areas (isocortical layers 5 and 6) in the left image.

The EPd is topographically organized to project to downstream areas

EPd-Oxtr neurons can broadly project to downstream areas regardless of their location along the anterior-posterior axis, or subregions of the EPd may have preferential projections to a subset of downstream areas. To further understand the anatomical organization of the EPd, we crossed Oxtr-Cre mice with Ai65 mice (Cre- and Flp-dependent intersectional reporter)⁴⁷ and administered retro-AAV-EF1a-DIO-FLP-WPREhGH poly(A) injections into one of four major projection areas of the EPd: two anterior brain regions (AON and ILA) and two posterior brain regions (COA and PIR) (Figure 4A). We used STPT to examine labeled cells in the whole brain (Figure 4B). Subsequent analysis revealed that injections into the AON and ILA predominantly labeled Oxtr neurons in the anterior parts of the EPd, while injections into the COA and PIR resulted in labeling Oxtr neurons in the posterior parts (Figures 4B and 4C), indicating topologically distinct projection domains within the EPd. We also found that Oxtr neurons in the EPd and a small population of Oxtr+ neurons in the CLA project to the AON and ILA (Figures 4B and 4C). In contrast, the PIR and COA received input predominantly from the EPd-Oxtr neurons (Figures 4B and 4C).

Figure 4. The EPd-Oxtr neurons are topographically organized to project to downstream areas.

(A) Schematic illustrating the breeding and injection strategy.

(B) Examples of raw images showing Oxtr+ neurons across the anterior to posterior extent of the CLA and EPd, projecting to the AON, ILA, PIR, and COA (AON: n = 4 mice, ILA: n = 3 mice, PIR: n = 2 mice, COA: n = 2 mice, anterior EPd: n = 3 mice, and intermediate EPd: n = 1 mouse; see STAR Methods for more sample details).

(C) Summary figure depicting CLA and EPd-Oxtr neurons projecting to the AON, ILA, PIR, and COA. Larger color-filled circles indicate the injection area, while smaller circles of the same color represent Oxtr+ cells in the CLA and EPd that project to the respective areas.

(D) Topographic organization of projections within the EPd. Anterior EPd injections show stronger projections in the anterior regions of the CLA and EPd, whereas posterior injections show more projections within the posterior regions of the CLA and EPd. aEPd, anterior EPd; iEPd, intermediate EPd; pEPd, posterior EPd. Whole-brain registration to the CCFv3 was performed using Elastix software.

CLA neurons are known to have extensive inter-connectivity rostrocaudally within the structure.⁴⁸ Hence, we hypothesized that EPd-Oxtr neurons would also exhibit uniform rostrocaudal connectivity within the structure. To test this, we injected rAAV-EF1a-DIO-FLP-WPRE-hGH poly(A) into either the aEPd or pEPd of the Oxtr-Cre;Ai65 mice, followed by STPT imaging. The results revealed that the aEPd exhibited stronger connections to anterior regions of the complex, whereas the pEPd injections showed stronger connections to posterior regions (Figure 4D). Altogether, these findings demonstrate an anatomical topology in the connections of the EPd along its rostrocaudal axis, as well as their projections to downstream areas.

The EPd-Oxtr neurons receive monosynaptic inputs mainly from the limbic and olfactory processing areas

To understand the immediate upstream inputs of the EPd-Oxtr neurons, we performed brain-wide monosynaptic input tracing using conditional AAV expressing avian tumor virus receptor A (TVA) and optimized G protein with mCherry reporter injected into the EPd of Oxtr-Cre mice, followed by G-deficient avian rabies viruses with GFP expression in the same area (Figure 5A)¹⁹ (see STAR Methods for more details). Ten mice (7 males and 3 females, 2–3 months old) were used, with different injection sites covering the entire anteroposterior extent of the EPd (Figure 5B). We used tissue clearing that preserves endogenous fluorescent protein,⁴⁹ followed by light sheet fluorescence microscopy (LSFM) imaging with cellular resolution. We utilized our automated cell counting methods to quantify fluorescently labeled cells and registered individual samples onto the CCFv3 using Elastix⁵⁰ (Figure 5C). We confirmed the localization of starter cells in the EPd and samples with little to no starter cell contaminations from neighboring areas were used in the analysis (Figure 5D). We also confirmed no leaky (TVA-independent) labeling from the rabies virus (Figure S4).

Figure 5. The EPd-Oxtr neurons receive monosynaptic inputs primarily from limbic and olfactory processing areas.

(A) Schematic of conditional monosynaptic tracing using rabies virus injections into the EPd of Oxtr-Cre mice.

(B) Locations of injections along the anterior-posterior axis of the EPd and the size of individual injections.

(C) Brain-wide inputs into the EPd (green, n = 10 mice) targeting Oxtr neurons. Maximum signals from all samples were overlaid on the reference brain. The registration of individual brain samples to the CCFv3 was performed using Elastix software.

(D) Starter cells showing colocalization of nucleus localization signal-mCherry and rabies-GFP in the EPd.

(E) Representative monosynaptic inputs in different coronal planes (top) with high-magnification images from the white boxed areas (bottom).

(F) Bar graph representing major brain regions providing monosynaptic inputs to the EPd. The full names of abbreviations according to the CCFv3 are listed in Table S1.

Overall, EPd-Oxtr neurons receive strong input from the limbic and olfactory areas, similar to the projection mapping, indicating very strong reciprocal connectivity (Figures 5E and 5F; Video S2). In addition, we found strong input from many thalamic areas, such as the medial dorsal thalamus (MED) and the intralaminar thalamus (ILM), linked with limbic, pain, emotional, and autonomic functions (Figures 5E and 5F; Video S2).^51-55 Notably, EPd-Oxtr neurons receive strong input from the basal forebrain, particularly the MA, which contains cholinergic neurons broadly projecting to the olfactory areas and ventral hippocampus.^56-58 In the hypothalamus, the parasubthalamic nucleus (PSTN), which regulates emotion and autonomic functions,^59,60 provides strong input to the EPd-Oxtr neurons. In the midbrain, the substantia nigra reticular (SNr) dorsal part and the superior collicular (SC) lateral part, which receive input from limbic cortices,^61,62 and the dorsal raphe (DR), for emotional control,^63,64 provide input to the EPd-Oxtr neurons (Figures 5E and 5F; Video S2). In the hindbrain, the locus coeruleus (LC), which provides norepinephrine projections for attention,^65,66 sends input to the EPd-Oxtr neurons (Figures 5E and 5F; Video S2).

Lastly, we closely examine monosynaptic input from the isocortical areas using our isocortical flatmap (Figures 6A and 6B). Overall, there were relatively sparse inputs from the isocortex, with the highest inputs from the deep layers (L5 and L6) of lateral association areas (Figures 6A-6D). This suggests that EPd-Oxtr neurons mainly receive limbic and internal (GU and VISC) information from the isocortex.

Figure 6. The EPd receives monosynaptic inputs selectively from the limbic and lateral cortices.

(A) Isocortical flatmap showing the CCFv3 anatomical regions and border lines.

(B) Isocortical flatmap depicting monosynaptic inputs to the EPd from the isocortex.

(C) Heatmap showing the density of monosynaptic input cells from isocortical layers to the EPd (n = 10 mice). The full names of abbreviations are listed in Table S1.

(D) Examples of raw images showing monosynaptic inputs from the isocortex to the EPd. The right image shows high-magnification images of the red boxed areas (isocortical layers 5 and 6) in the left image.

Collectively, these data suggest that EPd-Oxtr neurons receive monosynaptic input from brain areas processing the internal state, including interoception.

EPd-Oxtr neurons show downregulated activity upon novel cue exposure

The EPd is a part of the CLA complex that has been previously implicated in salience processing.⁶⁷ Monosynaptic input data also showed strong input from the salience network (e.g., amygdala and anterior insular).^68,69 Moreover, it is well established that Oxt signaling has been implicated in regulating environmental novelty for both social and non-social stimuli.^70-73 Based on this evidence, we hypothesized that Oxtr+ EPd neurons will be activated upon social cues.

To examine the neural activity patterns of the EPd-Oxtr neurons to social cues, we injected AAV-Syn-FLEX-GCaMP6m into the aEPd and implanted a graded index (GRIN) lens with a miniscope (Figure 7A; see STAR Methods for more details). We exposed our target mice to a novel social cue (an opposite-sex stranger mouse) in a home cage while recording GCaMP6m signals (Figures 7B-7E). In contrast to our original prediction, we observed high baseline neural activity during exploratory behaviors in the home cage followed by significantly decreased neural activity upon encountering a stranger mouse (Figures 7B-7E). The neural activity returned to high baseline levels about 50 s after, which were maintained during subsequent no-stimulus periods (Figures 7C and 7D). Further, to determine if this decrease in neuronal activity was dependent on social or non-social novelty, we introduced a novel object. Similar to the social novelty cue, there was a significant decrease in activity upon introducing a novel object (Figures 7D and 7E).

Figure 7. EPd-Oxtr neurons exhibit downregulated activity upon exposure to novel cues.

(A) Top: GCaMP6m virus injection into the EPd of Oxtr-Cre mice. Bottom: GRIN lens implantation in the target area and an example field of view for in vivo imaging.

(B) Behavioral paradigm. Each experiment consists of a 6 min epoch alternating between freely moving and novelty phases. The freely moving phases consist of continuous recordings, while the stranger and novel object epochs include 6 min of continuous recording, with either the stranger mouse or the novel object introduced 1 min after the start of the epoch. For consistency, the 1 min mark from the start of the experiment was used as the reference point in every epoch for analysis.

(C) Heatmap showing all recorded neurons (399 neurons, n = 8 mice, one session per mouse) from 40 s before to 70 s after the 1 min mark in each epoch.

(D) Average $\Delta \mathsf{F}∕\mathsf{F}$ plot of all recorded neurons from 40 s before to 70 s after the 1 min mark in each epoch, showing high and sustained neuronal activity $(\Delta \mathsf{F}∕\mathsf{F})$ in the home cage (without external cues) and significantly decreased activity upon exposure to novel cues. Data are represented as mean ± SEM.

(E) Total number of calcium peaks between the 40 s preceding the 1 min mark and the 40 s following the 1 min mark in each 6 min epoch, analyzed using paired t tests (n = 8 behavioral sessions from 8 mice). Bar graphs represent the mean values with individual data points overlaid. Paired data points are connected across the two bars with a line.

(F and G) Repeated exposure to the same stranger mouse (F) did not cause a significant decrease in EPd neural activity $(\Delta \mathsf{F}∕\mathsf{F})$ (all recorded neurons from n = 6 mice, one session per mouse). Data are represented as mean ± SEM in (G).

(H) Number of calcium peaks before and after 40 s of introducing either a stranger or familiar mouse, Wilcoxon matched-pairs signed-rank test (n = 6 mice, one session per mouse). Bar graphs represent the mean values with individual data points overlaid. Paired data points are connected across the two bars with a line.

(I and J) In contrast, introducing two different stranger mice (I) led to continued downregulation of Oxtr neuronal activity $(\Delta \mathsf{F}∕\mathsf{F})$. Data are represented as mean ± SEM in (J).

(K) Number of calcium peaks before and after 40 s of introducing two stranger mice, analyzed using a paired t test (n = 6 mice, one session per mouse). Bar graphs represent the mean values with individual data points overlaid. Paired data points are connected across the two bars with a line.

To test whether novelty drives the downregulation of EPd-Oxtr neuronal activity, we repeatedly exposed the mice to the same stranger mouse (Figures 7F-7H) or two different stranger mice (Figures 7I-7K). Repeated exposure to the same mouse abolished the significant decrease in EPd activity. In contrast, exposure to a different stranger mouse again significantly reduced activity, confirming that the novelty of the social stimuli causes the downregulation of the EPd activity (Figures 7I-7K). Similar to the results from repeated social cue introductions, repeated exposure to the non-social cue (toy mouse) abolished the downregulation of neuronal activity (Figure S5), indicating that novel stimuli (social or non-social) downregulated the activity of Oxtr+ EPd neurons.

DISCUSSION

We present a comprehensive cellular characterization and anatomical connectivity map of Oxtr+ neurons within the EPd, a highly enigmatic brain area. We found strong connections from the EPd to the ventral half of the brain, including limbic and olfactory areas. Our in vivo recording shows selective downregulation of EPd neuronal activity upon novel stimuli from high baseline activity, suggesting its potential role in mediating interoceptive states, which is attenuated by external cues. Collectively, our study provides an initial anatomical and functional characterization of the EPd to elucidate its role in the brain.

In rodents, the CLA and the EPd are identified as separate structures, whereas in primates, they appear as a single, continuous structure.⁷⁴ This has led to ongoing debates over the anatomical boundaries of the CLA complex in rodents as well as inconsistent nomenclature,^9,75 complicating the selection of a specific marker exclusive to the EPd. Using spatial transcriptomics and FISH, we identified distinct gene expression signatures of the EPd and achieved a clear delineation of anatomical boundaries between the EPd and the adjacent structures (the CLA, AI, and PIR). A recent study in macaques identified distinct gene sets enriched in the CLA and EPd similar to the findings in this study, reinforcing the notion that these are separate structures with molecular signatures conserved across species, including higher-order primates.⁷⁶ Notably, we found Oxtr as an additional marker for the EPd in mouse brains.^18,34 Previously, Oxtr expression has been reported in the CLA complex across species, including rodents^34,77 and primates such as rhesus macaques.⁷⁸ Moreover, our current study identified that EPd-Oxtr neurons are largely glutamatergic neurons, consistent with the previous report.³⁴

We found that Oxtr-Cre mice can serve as a genetic tool to selectively target EPd neurons. Combined with viral tools and 3D mapping methods, we established brain-wide input-output connectivity maps of the EPd. While the input-output architecture of the CLA^5,9 is widely studied, there is a lack of comprehensive studies that explain the connectivity of the EPd across the whole mouse brain.⁷⁹ Both the CLA and the EPd originate from the same part of the pallium based on Nr4a2 expression during development, with neurons that migrate to the ventrolateral dorsal pallium forming the CLA and those that migrate further ventrally forming the EPd.³ Due to their common origin, these areas were thought to have a similar topographical organization.¹ However, recent studies suggest regional differences in the input-output architecture of the CLA complex, with the CLA exhibiting a dorsoventral organization in its output projections.⁸⁰ In this study, we identified extensive projections of EPd neurons in the ventral half of the brain, broadly covering the olfactory and limbic areas, which contrasts strongly with the CLA output covering the dorsal half of the brain, including the entire isocortical areas.^5,9 Moreover, our projection mapping from the EPd neurons shows no projections to the RSP, which is a defining feature of CLA output.²⁸ For the monosynaptic input, the EPd receives strong input from limbic and olfactory areas, suggesting a reciprocal connection between the areas. In contrast, thalamic, midbrain, and hindbrain areas that regulate attentive state (MED and LC) and process emotion/affective/pain/visceral information (subparafascicular nucleus, PSTN, and DR) provide monosynaptic input without receiving reciprocal projection from the EPd. Another noticeable area is the basal forebrain (e.g., SI and MA), which contains cholinergic neurons to mediate attention.^42,81 Collectively, anatomical connectivity data strongly suggest that the EPd is positioned to modulate limbic and olfactory areas based on input from emotional and visceral circuits, serving as a parallel circuit to the CLA for motor sensory processing in response to external stimuli.²

Previous studies suggest that the EPd is implicated in recognition memory,⁸² epilepsy,^13,83 and higher-order olfactory processing.⁸⁴ Our in vivo recording data showed a high level of neural activity at the baseline exploratory stage and persistent downregulation of the EPd activity upon novel stimuli (both social and non-social). Hence, our data suggest that the EPd may regulate baseline exploratory behavior while inhibition of the EPd allows attentive neural circuits to engage in response to novel stimuli. Previous studies showed that long-range projections from the CLA excitatory neurons preferentially target local interneurons in cortical areas, resulting in net inhibition of the downstream circuit.^85-87 Moreover, inhibition of the CLA impaired performance selectively in high-order decision-making.^88,89 Similarly, the EPd may provide top-down inhibition to the olfactory and limbic circuit, and inhibition of the EPd can disinhibit the downstream circuit to express initial cautious and anxious approach behavior upon novel stimuli while considering internal state information. Another interesting observation is the high baseline neural activity of the EPd, which suggests the EPd’s involvement in the interoceptive state. Indeed, a prior study identified a very high level of EPd neural activity during slow-wave sleep (or deep sleep) compared to awake or rapid eye movement (REM) sleep.⁹⁰ This evidence suggests that the EPd may actively participate in maintaining the baseline interoceptive state and potentially even deep sleep, serving in opposition to neural circuits that mediate exteroception. The ability to shift between interoceptive and exteroceptive states is vital for assessing internal states and determining appropriate responses to potential environmental threats or opportunities. Impairment of such function can result in pathological conditions such as anxiety or hypervigilance. Indeed, neuroimaging data from patients with schizophrenia show abnormal activity within the CLA complex, which corresponds to the EPd.⁹¹

Limitations of the study

Our study is not without limitations. Our anatomical characterization was based on multiple stereotaxic viral injections, each with inherent technical limitations. Although we have excluded samples with clear contamination, minor injection site spillover may have resulted in a small number of labeled cells in adjacent regions, such as the CLA and PIR, in both anterograde and retrograde viral tracing experiments. To achieve a more precise understanding of the input-output patterns of the EPd, future studies should consider using EPd-specific Cre driver lines along with iontophoretic stereotaxic tracer injections.⁹² This approach will help minimize potential off-target effects and improve the accuracy of connectivity mapping. Furthermore, the EPd contains additional cell types beyond Oxtr+ neurons, each with distinct local and long-range connectivity. Investigating their synaptic connections and elucidating their functional roles will enhance our understanding of EPd circuits. Furthermore, exploring how Oxt signaling influences EPd circuits and how the EPd is altered in various pathological conditions will be critical for future research.

We envision that our anatomical and functional studies provide a foundation for these future investigations.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

All animal care and experimental procedures were approved by the Penn State University Institutional Animal Care and Use Committee (IACUC). The Oxtr-Cre line, originally established by Hidema et al.,⁹⁵ was imported to Penn State University via mouse rederivation. Oxtr-Cre mice, which have a 129 × C57BL/6J mixed genetic background, are not commercially available. Oxtr-Cre mice were then crossed with Ai14 (Jax: 007914, C57Bl/6 J background) to generate Oxtr-Cre:Ai14 mice. Oxtr-Venus mice, originally established by Yoshida et al., were imported to Penn State University (Kim lab) and are not commercially available. Only Oxtr-Venus heterozygote mice were used in this study. C57BL/6J mice were utilized for RNA in situ hybridization and spatial transcriptomics analysis. Both male and female mice were used in all experiments, with exact numbers provided in the corresponding sections below. Mice were provided food and water ad libitum and were housed under controlled temperature and light conditions, with a 12/12-h light/dark cycle.

METHOD DETAILS

Spatial transcriptomics

For spatial transcriptomics, initial experiments were conducted using C57BL/6J mice (n = 3; 2 males, 1 female; 2–3 months old; one anterior EPd coronal section per mouse) to delineate the CLA from the EPd and neighboring areas. A custom 500 gene panel was used, consisting of genes enriched in the CLA, EPd, PIR, and AI (refer to Table S2 for the gene panel) that constitute 265 genes. Hierarchical clustering and differential gene expression analyses were performed on these three datasets (Figures 1A-1G, S1 and S2A-S2I). Moreover, we conducted an additional experiment using a different gene panel from Vizgen (Table S2), which included Oxtr (n = 1, C57BL/6J, male, ~3 months old), confirming that Oxtr transcripts are enriched in the EPd (Figures 1H and S2J). Furthermore, we analyzed two publicly available datasets from the Allen Institute using the Brain Image Library that correspond to the anterior EPd co-ordinates^32,33 (https://download.brainimagelibrary.org/aa/79/aa79b8ba5b3add56/609882/1198980035/, https://download.brainimagelibrary.org/aa/79/aa79b8ba5b3add56/609889/1198980584/) and confirmed Oxtr enriched expression in the EPd (Table S4).

Sample collection

C57BL/6J mice, aged 2–3 months, were euthanized via cervical dislocation to collect fresh frozen whole brain samples for MERFISH experiments. After excision, brains were placed in cryomolds (Ted-Pella, 27110) partially filled with Optimal Cutting Temperature (O.C.T.) compound (Tissue-Tek 4853). The molds were then filled completely and rapidly frozen using 2-methylbutane cooled by dry ice. Once the O.C.T. compound had solidified, the samples were stored at −80°C for long-term preservation. Dissection tools were disinfected with RNaseZap (Invitrogen, AM9780) before and after each use to ensure RNA integrity.

MERFISH 500 gene panel selection

A 500-gene panel was selected for the MERFISH experiments. This panel included 265 genes, comprising cell-type markers and genes specifically enriched in regions such as the claustrum (CLA), dorsal endopiriform nucleus (EPd), piriform cortex (PIR) and agranular insular cortex (AI) that was used for clustering and differential expression analysis. The panel also comprised cortical layer-specific markers (refer to Table S2 for gene list).

MERFISH imaging experiment

MERFISH experiments were performed using the Vizgen MERSCOPE platform. Fresh frozen mouse brain samples were cryosectioned at a thickness of 10 μm using a cryostat maintained at −20°C. Sections were collected with approximate co-ordinates anteroposterior (AP) from the bregma: +0.7 mm. Sections were mounted on MERSCOPE slides (Vizgen 2040000) and fixed in 4% paraformaldehyde (PFA) in 1× PBS for 15 min. Following fixation, sections were washed three times with 1× PBS and permeabilized in 70% ethanol at 4°C overnight. Probe hybridization was carried out with a custom 500-gene MERSCOPE Gene Panel Mix (Vizgen 20300008) or a standard 140 gene panel (Table S2) in a 37°C incubator for 36–48 h. Post-hybridization, sections underwent two 30-min washes with Formamide Wash Buffer at 47°C. Samples were then embedded in a hydrogel using the Gel Embedding Premix (Vizgen 20300004), ammonium persulfate (Sigma 09913-100G), and TEMED (Sigma T7024-25ML) from the MERSCOPE Sample Prep Kit (10400012). Once the hydrogel solidified (~1.5 h), the samples were cleared overnight at 37°C with a Clearing Solution containing Proteinase K (NEB P8107S) and Clearing Premix (Vizgen 20300003). After clearing, sections were stained with DAPI and Poly T Reagent (Vizgen 20300021) for 15 min at room temperature, followed by a 10-min wash in Formamide Wash Buffer. Imaging was then performed using the MERSCOPE system (Vizgen 10000001). Comprehensive protocols for sample preparation and instrument usage can be found at Vizgen Resources and Vizgen Instrumentation.

Registration to the CCFv3

The Allen Common Coordinate Framework (CCF, version 3, 2017 version of the delineations)³⁰ was used to register all brain slices that underwent MERFISH experiments. Initially, datasets were loaded individually into the MERFISH Visualizer, where images with DAPI and MOBP were exported, along with a metadata file detailing image size and center coordinates in microns. Each brain slice was then registered to the CCFv3 template using QuickNII⁹⁶ and VisuAlign⁹⁷ for linear and non-linear registration, respectively. The position and exact cutting plane for each section was determined using QuickNII (Figure S1A). Deviations from the coronal plane were identified by inspecting the relative positions of prominent landmarks such as the genu of the corpus callosum, olfactory tubercle, and the dorsal and ventral parts of the striatum. The registration was then non-linearly refined using VisuAlign, which allows for placing and moving markers manually until the registration is deemed satisfactory (Figure S1A). The final registration was further verified using additional markers from the MERFISH panel that show strong and selective expression in different brain regions. This was done using markers for multiple major brain regions (e.g., slc32a1 in the striatum; Figure S1B) as well as for the CLA and EPd (Figure S1C). Inspection of the delineations on these marker images confirmed the accuracy of the registration. The.flat file exported from VisuAlign was converted to a 16-bit label image for each MERFISH slice using custom Python code. These custom atlas maps, where each pixel is represented by an ID representing its assigned region, were used for further analysis of the MERFISH data.

Re-segmentation

To improve the initial Watershed-based segmentation, we re-segmented our data using Cellpose (1.0.2),¹⁰² which increased average cell volume and transcript counts but misclassified spherical imaging artifacts as cells. To address this, we retrained a Cellpose2¹⁰⁰ model with manually annotated cell boundaries to exclude artifacts and tested its generalizability. We validated segmentation by randomly selecting 100 × 100 μm patches per experiment, comparing Watershed, default Cellpose, and custom Cellpose2 models using precision, recall, F1 score, and cell tracking consistency. Our custom Cellpose2 model achieved an F1 score of ~90%, outperforming Cellpose (82%) and Watershed (79%). Based on these results, we used the Cellpose2-resegmented datasets for further analysis.

Preprocessing and quality control

Preprocessing was conducted using the Scanpy framework in Python³¹ along with a custom-built spatial transcriptomics module based on Scanpy. Data from three replicates were pooled, and doublets were identified and removed using Scrublet.¹⁰³ Quality control included both quantitative and visual assessments, ensuring comparable sample-level metrics such as cell counts, average volume, transcripts per cell, and unique genes per cell. Volume filtering was applied using the 0.1–0.9 quantile range, with a minimum of 48 μm³ and a maximum of 3,583 μm³. The dataset’s median volume was 1,984 μm³. Additional filtering removed cells expressing fewer than 5 genes, fewer than 30 total transcripts, and genes detected in fewer than 3 cells. After filtering, the median total transcript count per cell was 576, with a median of 155 unique genes per cell.

Hierarchical clustering

We normalized total counts per cell to 500, applied a natural logarithm transformation, and scaled data to unit variance with a mean of zero (clipping values at 10). The gene set was refined to 266 genes related to key brain region and cell type markers. Cells from the CLA, EPd, AI and PIR were selected, yielding 8,260 cells across three samples. We extracted the top 50 marker genes for major cell types (astrocytes, choroid plexus cells, vasculature-associated cells, immune cells, neurons, oligodendrocytes, and OPCs) from,¹⁰⁴ selecting those present in our custom panel and supplementing with well-documented markers. Neuron subtype markers were curated from,¹⁰⁵ aggregated by neuronal subclass (e.g., IT-ET Glut, CTX-MGE GABA) and cortical layer-specific markers. The list was manually refined to remove duplicates and lowly expressed genes. Next, we applied a two-stage clustering approach: an initial round using a grid search to optimize PCA components, neighbors, and Leiden resolution, followed by hierarchical clustering to refine subclusters. Clustering performance was assessed using a composite score based on silhouette score and marker purity. Hierarchical clustering was guided by a marker-based directed acyclic graph, allowing parent and child cell types to be linked. Neurons were re-clustered separately to identify subtypes. To assess regional enrichment of cell type clusters, we performed a permutation-based test, shuffling region labels within each sample while preserving cell type counts. For each cluster-region pair, the observed proportion was compared to a null distribution from 10,000 permutations, yielding empirical one-tailed p-values. Multiple testing was controlled using Benjamini-Hochberg correction, with adjusted p ≤ 0.05 considered significant.

Differential gene expression analysis was performed on raw expression counts using ALDEx2¹⁰⁶ in R software, which applies Dirichlet-multinomial Monte Carlo sampling to estimate gene abundances while accounting for uncertainty in total expression levels. To allow for variability in abundance estimation without assuming an implicit scale shift between conditions, we modeled gamma as = $2I\gamma$, where $I\gamma$ was drawn from a normal distribution (μ = 0,σ = 0.2) across Monte Carlo replicates. For differential expression testing, we applied a mixed-effects model to each Monte Carlo instance, with multiple testing corrections using the Benjamini-Hochberg method to control the false discovery rate (FDR) as discussed in,¹⁰⁷ considering adjusted p-values ≤0.05 as significant.

To identify genes differentially expressed across the four brain regions of interest, we applied a mixed-effects model with brain region as a fixed effect and sample as a random effect to account for biological and technical variability. The EPd was set as the reference level. Statistical significance was determined using empirical Bayesian estimation based on posterior distributions from the Monte Carlo simulations.

To test differential expression at the neuronal cluster level, we applied a similar mixed-effects model, using cluster identity as a fixed effect and sample as a random effect. The cluster with the highest average expression across neuron subtype marker genes was set as the reference level. Genes were considered significantly enriched in a cluster if they had an adjusted p-value ≤0.05 and a positive estimate relative to the reference cluster. To refine cluster-specific gene sets, we further filtered results to include only genes with the highest raw expression in the corresponding cluster.

Oxtr dataset analysis

This analysis was performed on three datasets: (1) our MERFISH experiment using a 140-gene panel (Table S2) and (2) two spatial transcriptomics datasets from the Allen Institute^32,33 (both using the same 500-gene panel).

Initially, we pooled all three datasets but found that clustering performance was poor when limited to the 53 overlapping genes between panels. To retain full gene information, we instead clustered the 140-gene dataset and the Allen datasets (merged) separately, applying the same preprocessing steps. Doublets were removed with Scrublet, and volume and transcript-based filtering were applied using comparable thresholds. We then identified marker genes for major cell types and neuron subtypes and performed hierarchical clustering to classify neurons as excitatory or inhibitory. Following clustering, we subset EPd neurons and analyzed the datasets separately. To determine a dataset-independent Oxtr expression threshold, we examined Oxtr expression relative to sequencing depth, using average transcripts per gene (total transcripts/unique genes) as a normalization metric. Since Oxtr expression in the 140-gene dataset was ~2× higher than in the Allen datasets—matching differences in sequencing depth—we used Oxtr/average transcripts per gene to define a consistent threshold. Based on histograms, we selected 0.157 as the cutoff, roughly corresponding to ≥2 Oxtr transcripts in raw counts. Applying this threshold, we identified and quantified Oxtr-expressing excitatory and inhibitory neurons in the EPd across all datasets.

Single-molecule mRNA fluorescence in situ hybridization

C57BL/6J mice (2–3 months old, n = 6 mice- 3 males and 3 females) were euthanized by cervical dislocation, and their brains were immediately dissected and immersed in Optimal Cutting Temperature (OCT) media (Tissue-Tek, catalog #4853). The immersed brains were rapidly frozen using dry ice-chilled 2-methylbutane and stored at −80°C until needed. Coronal brain sections, 10 μm thick, were collected using a cryostat and stored at −80°C. In situ hybridization was conducted within one month of sectioning. At least 4 technical replicates per sample were used for mRNA fluorescence in-situ hybridization and analysis.

The RNA Scope Multiplex Fluorescent Reagent Kit v2 (ACDBio) was employed to detect and quantify target mRNA at single-molecule resolution, following the manufacturer’s protocols for fresh frozen samples. To quantify the colocalization of Oxtr-positive cells with markers for excitatory and inhibitory neurons, Probe-mm-Oxtr (454011-C2), Probe-mm-slc17a7 (ACDBio, 416631), and Probemm-VGAT (319191-C3) were used to detect mRNA expression of Oxtr, vGlut1, and vGAT.

Vibratome sectioning of mouse brain samples

Vibratome sectioning was performed on the following samples. 1. Oxtr Cre mice injected with pAAV.Syn.Flex.GCaMP6m.WPRE. SV40 (AAV1) virus (titer: 2.1 × 10∧13 GC/mL), purchased from Addgene and generously provided by Douglas Kim &the GENIE Project (RRID: Addgene_100838) into the anterior part of the EPd (AP: 1.1 mm; ML: 2.45 mm; DV: −4.13 mm), (n = 3 mice). 2. Oxtr-Venus heterozygous mice (n = 3) injected with Cholera Toxin B subunit to the Retro splenial cortex. 3. Oxtr Cre-Ai14C mice (n = 3). All the abovementioned mice were transcardially perfused with 4% paraformaldehyde (PFA). Brain samples were dissected and stored in 4% PFA overnight. Afterward, the samples were washed four times with 0.05M phosphate-buffered saline, followed by embedding in 4% agarose solution. The embedded samples were sectioned using a vibratome at a thickness of 60 μm.

The anterior EPd sections from Oxtr Cre mice injected with pAAV.Syn.Flex.GCaMP6m.WPRE.SV40 containing the signal were mounted on glass slides with FluorSave reagent and coverslipped. These brain sections were then imaged using a fluorescence microscope.

The sections containing anterior EPd from Oxtr-Venus heterozygous mice and Oxtr Cre-Ai14C mice underwent antibody staining. Anterior EPd sections from Oxtr-Venus heterozygous mice, containing CTB signals in the CLA, were stained for GFP. Anterior EPd sections from Oxtr Cre-Ai14C mice were stained for parvalbumin.

Immunohistochemistry

Sections containing the anterior EPd from Oxtr-Venus and Oxtr-Ai14C mice underwent a series of processing steps. First, they were rinsed three times with 1× PBS, followed by a one-hour blocking step at room temperature using 1% donkey serum diluted in PBST (1× PBS +0.03% Triton X-). After blocking, the slices were incubated overnight at 4°C in a primary antibody solution prepared in the blocking buffer with gentle rotation. The next day, the slices were washed three times in 1× PBS and then incubated with a secondary antibody for one hour at room temperature. Before mounting, the slices were washed three additional times in 1× PBS and mounted using Vectashield mounting medium containing DAPI (Vector Laboratories, Cat# H-1500, RRID: AB_2336788).

The following primary and secondary antibodies were used. For Oxtr-Venus sections injected with CTB: Anti-Chicken Green Fluorescent Protein (Aves Labs, Cat# GFP1202, 1:800 dilution) as the primary antibody and Alexa Fluor 488 AffiniPure Donkey Anti-Chicken IgY (IgG) (H + L) (Jackson Immunoresearch, Cat# 703-545-155, 1:500 dilution) as the secondary antibody. For Oxtr-Ai14C sections: Anti-parvalbumin primary antibody (Sigma, Cat# P3088, 1:1000 dilution) and Alexa Fluor 488 Donkey Anti-Mouse IgG (H + L) Highly Cross-Adsorbed Secondary Antibody (Invitrogen, Cat# A-21202, 1:500 dilution).

Fluorescence microscopy imaging

Microscopic imaging was performed using a BZ-X700 fluorescence microscope (Keyence). A low magnification objective lens (4×) provided a broad view to define brain AP locations from bregma, while a higher magnification objective lens (20×) was used to image the EPd area in each brain section. Images were manually delineated based on the Allen atlas, and colocalization analysis was conducted manually using the Cell Counter plugin in FIJI (ImageJ, NIH).

Stereotaxic surgery and virus injections

For anterograde tracing, 50–500 nL of AAV2-CAG-Flex-EGFP virus (titer 3.7 × 10¹² vg/mL, purchased from the UNC Vector Core) was injected into the EPd as previously described.² Oxtr-Cre mice (2–4 months old, 6 males and 3 females) were anesthetized with isoflurane (administered using Somnosuite, Kent Scientific) and mounted on a stereotaxic instrument (Angle Two, Leica) with a heating pad underneath. All injections were performed with pulled micropipettes (VWR, catalog #53432-706). The virus was delivered through the small opening of the micropipette at a rate of 75–100 nL/min. The speed and volume of injection were monitored using the calibration marks on the micropipette (1 mm = 100 nL). To target the anteroposterior extent of the EPd, two major coordinates were used, and post hoc analysis was performed to confirm the actual injection site. Coordinates for anterior EPd injections were anteroposterior (AP) from the bregma: 1.1 mm; mediolateral (ML): 2.45 mm; dorsoventral (DV): −4.13 mm. For intermediate EPd injections, the coordinates were: AP: −0.7 mm, ML: 3 mm, and DV: −4.3 mm. After 3 weeks, the mice were deeply anesthetized with a ketamine-xylazine mixture (100 mg/kg ketamine, 10 mg/kg xylazine, i.p.) and perfused. Brain samples were then collected by cardiac perfusion with 4% paraformaldehyde (PFA). Dissected brain samples were then postfixed in 4% PFA overnight at 4⁰C. The fixed brains were then stored in 0.05M phosphate buffer (PB) at 4°C until serial two-photon tomography imaging.

To study the anatomical topology of EPd neuronal projections, we employed a transgenic approach combined with viral injections. Oxtr-Cre mice were crossed with Ai65 mice (Jax; Stock No: 021875). Adult mice (2–3 months old) were injected with AAV2/retro rAAV-EF1alpha-DIO-FLP-WPRE-hGH polyA (BioHippo, catalog #BHV12400173-12, 200 nL each) into six major target locations: the anterior olfactory nucleus (AON, 2 males and 2 females; coordinates AP: 2.22 mm, ML: 0.7 mm, DV: −4.0 mm), the infralimbic area (ILA, 1 male and 2 females; coordinates AP: 1.70 mm, ML: 0.46 mm, DV: −2.70 mm), the piriform cortex (PIR, 2 females; coordinates AP: −2.56 mm, ML: 3.35 mm, DV: −5.20 mm), the cortical amygdaloid nucleus (COA, 1 male and 1 female; coordinates AP: −1.58 mm, ML: 2.50 mm, DV: −5.61 mm), the anterior EPd (3 males; coordinates AP: 1.10 mm, ML: 2.45 mm, DV: −4.13 mm), and the posterior EPd (1 male; coordinates AP: −0.7 mm, ML: 3 mm, DV: −4.3 mm). Three weeks post-injection, mice were deeply anesthetized with a ketamine (100 mg/kg) and xylazine (10 mg/kg) mixture and then transcardially perfused with 4% paraformaldehyde (PFA). Dissected brain samples were then postfixed in 4% PFA overnight at 4°C. The fixed brains were then stored in 0.05M phosphate buffer (PB) at 4°C until serial two-photon tomography imaging.

For monosynaptic input tracing, AAV5-CAG(del)>TCIT(-ATG)-Flex(loxP)-SV40 (1:8 dilution, titer: 1.6E+12 gc/mL, ref) was co-injected with AAV5CAG(del)>nC2oG-Flex(loxP)¹⁰⁸ (from Dr. Todd Anthony at Harvard University, titer: 2.8E+12 gc/mL) was injected in Oxtr-Cre mice (n = 10 mice- 7 males and 3 females, 2–3 months old) for the whole length of the EPd (150 nL per brain, same coordinates as anterograde mapping) as previously described.¹⁹ After 14d, the same volume of EnvA G-deleted Rabies-EGFP virus (titer: 4.89E+09 TU/mL, purchased from Salk Institute viral vector core, a gift from Edward Callaway; RRID:Addgene_32635¹⁰⁹) was injected into the same location with a 5° tilted angle. The mice were euthanized 7 d later for brain collection, tissue clearing and imaging.

For labeling CLA neurons, Cholera Toxin Subunit B CF Dye Conjugates (Biotium, cat#00070, 400nL of 1mg/mL solution) was injected to the retrosplenial cortex (coordinates AP: −1.34 mm, ML: 0.41 mm, DV: −0.6 mm) of Oxtr-Venus (n = 3 mice) heterozygous mice. Two weeks post injection, mice were deeply anesthetized with a ketamine (100 mg/kg) and xylazine (10 mg/kg) mixture and then transcardially perfused with 4% paraformaldehyde (PFA). Brain samples were collected for vibratome sectioning followed by staining with anti-GFP antibody.

After all the above-mentioned stereotaxic surgeries, the mice were injected intraperitoneally with 0.2mL of 1mg/mL Carprofen for 3–4 consecutive days to minimize pain.

STPT imaging and related data analysis

Both anterograde tracing (n = 9 mice) and the anatomic topology of EPd (AON: n = 4 mice, ILA: n = 3 mice, PIR: n = 2 mice, COA: n = 2 mice, anterior EPd: n = 3 mice and posterior EPd: n = 1 mouse) projections were determined using serial two-photon tomography (STPT) imaging and analysis, as previously described.^19,101,110 Briefly, dissected brains were postfixed in 4% paraformaldehyde (PFA) overnight at 4°C. The fixed brains were then stored in 0.05 M phosphate buffer (PB) at 4°C until imaging. To image the entire brain, STPT was performed using the TissueCyte 1000 system (TissueVision) following established protocols.^111,112 The brains were embedded in 4% oxidized agarose and cross-linked with a 0.2% sodium borohydride solution. Imaging was conducted with a resolution of 1 × 1 μm² in the x and y dimensions, across 12 × 16 × 280 tiles, with 50-μm intervals in the z-dimension. A wavelength of 910 nm was used for two-photon excitation to simultaneously excite both green (e.g., eGFP) and red (e.g., tdTomato) fluorescent signals. The signals were separated using a 560-nm dichroic mirror and two band-pass filters (607/70-25 for red and 520/35-25 for green). The imaging tiles for each channel were stitched together using custom-built software.^19,111,113

For quantitative projection data analysis, we used our previously published pipeline with minor modifications.¹⁹ Briefly, both the signal and background channels were z-normalized. The background channel images were then subtracted from the signal channel images to increase the signal-to-noise ratio. Next, we employed ilastik⁹⁹ for signal detection. By integrating ilastik into the automatic workflow, our algorithms performed parallel computations to detect each pixel with the maximum likelihood of it belonging to a projection signal, cell body, brain tissue, or empty space. Projection strength for each area was calculated by summing all projection signal pixels within an anatomically defined region. Autofluorescence of the brains was used to register each brain to the CCFv3³⁰ using Elastix,⁵⁰ after which the projection signals were transformed to the reference brain. We then used the maximum projection of registered long-range output datasets (N= 10) from each area to create a representative projection dataset for further quantitative analysis. “Area normalized projection” represents the ratio of the area covered by the projection signal to the total area of the region of interest (ROI), depicted as a percentage. For example, if the total pixel count for one ROI was 20,000 and the area covered by the projection signal for that ROI was 2,000, the area normalized projection would be (2,000/20,000) × 100 = 10%. For EPd anatomical topology distinction, we used the cell body classifier from the ilastik to count the number of cells, rather than using the projection pixels. We then quantified the cell density per ROI by dividing the total number of cells within one ROI by the total volume of that ROI, as previously described.¹⁰¹

Tissue clearing, light sheet imaging and related data analysis

Details of tissue clearing, light sheet imaging, and data analysis are as previously described.¹¹⁴ We used these methods for monosynaptic rabies tracing of EPd (n = 10, Oxtr-Cre mice- 7 males and 3 females, 2–3 months old), primarily employing SHIELD (Stabilization under Harsh conditions via Intramolecular Epoxide Linkages to prevent Degradation) tissue clearing to ensure minimal tissue volume changes while preserving endogenous fluorescence signals.⁴⁹ SHIELD reagents and protocols were obtained from LifeCanvas Technologies (https://lifecanvastech.com/). For P56 brains, PFA-fixed samples were incubated in SHIELD OFF solution for 4 days at 4°C, followed by SHIELD ON buffer for 24 h at 37°C. Tissues were then incubated in delipidation buffer at 37°C for 10 days, washed overnight in PBS, and optically cleared by sequential incubation in 75% EasyIndex +25% water for 24 h followed by 100% EasyIndex solution at 37°C for 24 h. For LSFM imaging, samples were embedded in 2% low-melting agarose (Millipore Sigma, cat. no.: A6013, CAS Number: 9012-36-6) in EasyIndex using a custom holder. They were incubated in EasyIndex at room temperature (20°C–22°C) for at least 12 h before imaging with the SmartSPIM light sheet fluorescence microscope (LifeCanvas). During imaging, the holder arm with the sample was immersed in 100% EasyIndex. Our setup included a 3.6× objective lens (LifeCanvas, 0.2 NA, 12 mm working distance, 1.8 μm lateral resolution), lasers at 488 nm, 560 nm, and 642 nm wavelengths, and a 5 μm z-step size. After imaging, samples were stored in 100% EasyIndex at room temperature (20°C–22°C). We developed a parallelized stitching algorithm for 3D reconstruction, inspired by Wobbly Stitcher,¹¹⁵ aimed at conserving hard drive space and minimizing memory usage. The algorithm first captured 10% outer edges of each image tile and generated a maximum intensity projection (MIP) in the z-direction for every set of 32 slices in the stack. It then aligned the z coordinates of MIP images across columns, followed by x and y coordinate alignment. Within each MIP, adjustments to 32 slices were made using curve fitting to finalize tile coordinates.

We enhanced our cell density mapping workflow by integrating the ilastik-based cell detection algorithm with our established methods.^111,112,116 Signals smaller than the cellular diameter 3.6 μm were filtered out. Centroids of individual cells were then identified. Subsequently, we performed image registration to align cell detection results with the CCFv3³⁰ template using Elastix.⁵⁰ Centroids were counted in each brain region to conduct 3D cell counting. To determine the anatomical volume of each sample, we initially registered the CCFv3 to individual samples with Elastix, adjusting anatomical labels accordingly based on registration parameters. We then quantified the number of voxels associated with specific anatomical IDs to estimate the 3D volume of each anatomical area. The density of 3D cell counting per anatomical regional volume (mm³) provided a quantitative measure of cell distribution across brain regions.

Ca²⁺ imaging with a head-mounted fluorescent microscope

For miniscope-based recordings, adult Oxtr-Cre mice (approximately 9–11 week old) underwent two stereotaxic surgeries: one for virus injection and a second for lens implantation. For Figures 7B-7E, n = 8 sessions from 8 mice (4M and 4F) were included for analysis. For Figures 7F-7K, n = 6 sessions from 6 mice (3M and 3F, one session per mouse) were included for analysis. For Figure S5, n = 3 sessions from 3 mice (all male mice, one session per mouse) were included in analysis. Mice were anesthetized with isoflurane, administered viaa Somnosuite (Kent Scientific), and positioned in a stereotaxic frame (Angle Two, Leica) with a heating pad underneath.

The pAAV.Syn.Flex.GCaMP6m.WPRE.SV40 (AAV1) virus (titer: 2.1 × 10∧13 GC/mL), purchased from Addgene and generously provided by Douglas Kim & the GENIE Project (RRID: Addgene_100838), was diluted in phosphate-buffered saline (1:3) based on titration experiments to achieve optimal cytoplasmic, non-nuclear localization of GCaMP6m in fixed tissue slices from injected animals. A volume of 200nL was injected into the anterior part of the EPd (AP: 1.1 mm; ML: 2.45 mm; DV: −4.13 mm).

Post-surgery, mice were housed in fresh autoclaved cages with ad libitum food and water. After a recovery period of two weeks, a second surgery was conducted for GRIN lens implantation. An incision was made to expose the skull, and stereotaxic alignment was performed using the inferior cerebral vein and bregma as vertical references. The same coordinates used for GCaMP injection (AP: 1.1 mm; ML: 2.45 mm; DV: −4.13 mm) were employed for lens implantation. The GRIN lens (Inscopix: ProView Integrated Lens, 0.6 mm × 7.3 mm; 1050–004413) was lowered into place at a rate of 200 μm per minute. The lens, attached to a baseplate, was secured to the skull with a small drop of dental cement (C&B Metabond Adhesive Cement system, #S380). After curing, thin layers of dental cement were applied to cover the exposed skull, followed by multiple layers of dental acrylic (Ortho-Jet BCA, Lang Dental) to stabilize the lens and baseplate. Once the dental acrylic had cured, mice were removed from isoflurane and allowed to recover in clean, autoclaved cages to minimize infection risk. Mice with GRIN lens implants were housed singly to prevent damage to the implanted lens. After both the stereotaxic surgeries for virus injection and GRIN lens implantation, the mice were injected intraperitoneally with 0.2mL of 1mg/mL carprofen for 3–4 consecutive days to minimize pain.

After an additional two-week recovery period, the mice were moved to a behavior room on a 12/12-h light/dark cycle, with lights turning off at noon and experiments conducted in the afternoon (after 2 p.m.).

Behavior

Once GRIN lens implantation was completed, mice were individually housed in a behavior room. Prior to recording, test mice underwent a habituation phase where they were connected to a dummy microscope to acclimate to its weight for 10 min daily over three consecutive days in their home cages. On the recording day, mice first experienced a brief 10-min exploration period in their home cage to minimize any initial artifacts. Subsequently, recordings were conducted over several phases: a 6-min freely moving phase, followed by a 6-min social interaction session with an opposite-sex mouse, and another 6-min freely moving phase. Finally, a 6-min session involving a novel object was introduced. During the social interaction and novel object sessions, the respective stimuli (stranger mouse or object) were introduced into the home cage at 1 min from the start of recording and remained for the subsequent 5 min. Between each recording session, there was at least a 10-min interval. For repeated introductions of social or non-social cues, recordings were continuous for 8 min as detailed in the figures. Social or non-social cues, whether novel or familiar, were introduced at either 1 min or6 min into the recording session and removed 1 min after their introduction. This setup allowed for the assessment of neural responses across multiple presentations of stimuli within a single recording session. To align behavior and calcium videos, a TTL pulse triggered calcium recordings through Anymaze software (Stoelting) at the start of each trial along with a behavior video recording. All Ca²⁺ signals were recorded using a miniature microscope and nVista DAQ system (Inscopix) at 20 frames per second (fps).

Image processing and analysis

All calcium ion (Ca²⁺ imaging movies obtained with the nVista system underwent initial preprocessing using the Inscopix Data Processing Software (IDPS; Inscopix). The GCaMP6m emission signals were captured continuously at a frame rate of 20 Hz under blue LED light (455 ± 8 nm, with power ranging from 10% to 60% and analog gain set to 1). Prior to motion correction, the spatial resolution of the videos was reduced by downsampling (4×). Following motion correction, the video data were transformed into [fluorescence (F) - background fluorescence $({\mathsf{F}}_0)]∕{\mathsf{F}}_0(\Delta \mathsf{F}∕{\mathsf{F}}_0)$ values, utilizing the mean projection images of the entire movie as ${\mathsf{F}}_0$. Individual calcium signals originating from specific regions of interest (ROIs, i.e., cells) were identified using principal component analysis (PCA) and independent component analysis (ICA), as previously described.¹¹⁷ The acquired timeseries of Ca²⁺ signals $(\Delta \mathsf{F}∕{\mathsf{F}}_0)$ (IDPS; Inscopix) were further analyzed and visualized using Python 3.9, NumPy, SciPy, Pandas, Matplotlib, and Seaborn.

The peaks/troughs of $\Delta \mathsf{F}∕{\mathsf{F}}_0$ were detected as follows. First, the baseline fluctuation of $\Delta \mathsf{F}∕{\mathsf{F}}_0$ (bl_fluc) or the noise level was calculated within the baseline time window (0-180sec) by averaging 5 smallest peak-trough values. This approach allowed us to exclude biological Ca²⁺ spikes in the baseline. We defined standardized $\Delta \mathsf{F}∕{\mathsf{F}}_0$, $\mathsf{z}\_\Delta \mathsf{F}∕{\mathsf{F}}_0$, as $(\Delta \mathsf{F}∕{\mathsf{F}}_0\text{-}\mathsf{mean}(\Delta \mathsf{F}∕{\mathsf{F}}_0))∕\mathsf{bl}\_\mathsf{fluc}$. $\mathsf{z}\_\Delta \mathsf{F}∕{\mathsf{F}}_0$ was smoothed with Savitzky-Golay filter and peaks and troughs were detected using the functions in Python SciPy package (e.g., savgol_filter & find_peaks). $\mid \mathsf{z}\_\mathsf{dF}∕\mathsf{F}\mid \ge 3$ were considered as peak/trough events. The times of peaks/troughs within 0.66 s (1/sampling rate*10) were combined and considered as a single peak/trough.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis for spatial transcriptomics datasets (Figure 1) are available in Tables S3-S5. Statistical analysis of RNA in-situ hybridization in Figures 1L and 1M was performed in Prism (GraphPad) and data are represented as mean ± SD. For the anterograde projections (STPT: Figures 2, 3 and 4) and monosynaptic input tracing (LSFM: Figures 5 and 6) datasets, we used MATLAB (MathWorks) and/or Prism (GraphPad) for plotting the maximum values for each anatomical area (ROI), combining data from all injections. Data was organized in Prism (GraphPad), with each individual mouse’s data included in separate columns. Maximum values for each anatomical subregion (ROI) were then computed and plotted as bar graphs. Flatmaps were generated using MATLAB, while all other graphs were created using Prism version 9.

For in-vivo miniscope recordings, heatmaps showing $\Delta \mathsf{F}∕{\mathsf{F}}_0$ for individual cells were plotted using MATLAB. $\Delta \mathsf{F}∕{\mathsf{F}}_0$ plots were presented as the mean ± standard error of the mean (SEM) using Prism version 9. Bar graphs showing the number of peaks were plotted using Prism after collecting data from the analysis performed using Python (as explained previously for image processing and analysis). The datasets were assessed for normality and homogeneity of variance to ensure the assumptions for parametric tests were met. For single introduction experiments (Figures 7B-7E), all datasets passed the Kolmogorov Smirnov normality test and were analyzed using the paired-t tests for significance in Prism. Datasets in Figure 7H did not pass the Kolmogorov Smirnov normality test. Hence analyzed using Wilcoxon matched pairs signed rank test. For Figure 7K, the datasets passed the Kolmogorov Smirnov normality test. Hence, analyzed with a paired t test using Prism. All bar graphs in Figure 7 represent the mean values with individual data points overlaid.

Software accessibility

All custom-built codes and flatmaps used in the current study will be freely available on request and can be used without any restriction.

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.115827.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Chicken Green Fluorescent Protein Antibody	Aves Labs	Cat# GFP1202, RRID:AB_2734732
Mouse Anti-Parvalbumin Antibody	Sigma-Aldrich	Cat# P3088, RRID:AB_477329
Bacterial and virus strains
AAV-CAG-FLEX-GFP (AAV2)	UNC Vector Core	N/A
AAV5-CAG(del)>TCIT(-ATG)-Flex(loxP)-SV40	Dr. Todd Anthony, Boston Children’s Hospital Viral Core	N/A
CAG(del)>nc2oG-Flex(loxP)	Dr. Todd Anthony, Boston Children’s Hospital Viral Core	N/A
EnvA G-deleted Rabies-EGFP virus	UNC Vector Core	N/A
AAV2/retroAAV-EF1alpha-DIO-FLP-WPRE-hGH polyA	BioHippo	BHV12400173-12.
pAAV.Syn.Flex.GCaMP6m.WPRE.SV40 (AAV1)	Addgene	Addgene_100838
Chemicals, peptides, and recombinant proteins
Cholera Toxin Subunit B CF® Dye Conjugates	Biotium	Cat #00070
Critical commercial assays
MERSCOPE	Vizgen	RRID:SCR_026274
RNAScope Fluorescent Multiplex Assay V2	Advanced Cell Diagnostics (ACD Bio)	RRID:SCR_012481
Deposited data
Web visualization of anterograde and retrograde tracing	This paper	https://kimlab.io/home/projects/EPd_connectivity/
Experimental models: Organisms/strains
C57bl/6J mice	Jackson Laboratory	Strain #: 000664; RRID:IMSR_JAX:000664
Ai14 (B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J)	Jackson Laboratory	Strain #: 007914; RRID:IMSR_JAX:007914
Oxtr-Venus (heterozygotes) mice	Nishimori lab35	N/A
Oxtr-Cre mice	Nishimori lab95	N/A
Oligonucleotides
RNAscope® Probe - Mm-Oxtr-O1-C2 - Mus musculus oxytocin receptor (Oxtr) mRNA	Advanced Cell Diagnostics (ACD Bio)	Cat# 454011-C2
RNAscope® Probe - Mm-Slc17a7 - Mus musculus solute carrier family 17 (sodium-dependent inorganic phosphate cotransporter) member 7 (Slc17a7) mRNA	Advanced Cell Diagnostics (ACD Bio)	Cat# 416631
RNAscope® Probe - Mm-Slc32a1-C3 - Mus musculus solute carrier family 32 (GABA vesicular transporter) member 1 (Slc32a1) mRNA	Advanced Cell Diagnostics (ACD Bio)	Cat# 319191-C3
Software and algorithms
QuickNII	Puchades et al.96	RRID:SCR_016854
VisuAlign	Gurdon et al.97	RRID:SCR_017978
ImageJ	Schneider et al.98	https://imagej.nih.gov/ij/RRID:SCR_003070
Elastix	Klein et al.50	https://elastix.lumc.nl/RRID:SCR_009619
ilastik	Berg et al.99	RRID:SCR_015246
Python	Open-Source	https://www.python.org/; RRID:SCR_008394
Adobe Illustrator	Adobe	RRID:SCR_010279
Inscopix nVista	Inscopix	RRID:SCR_017407
Merscope Visualizer	Vizgen	RRID:SCR_026274
Biorender	Biorender	RRID:SCR_018361
Cellpose2	Pachitariu and Stringer100	N/A
R Studio	Open-source	https://posit.co RRID:SCR_000432
Prism 10	GraphPad Software Inc	RRID:SCR_002798
Machine learning based cell and pixel quantification	This paper and Liwang et al.101	Zenodo data: https://doi.org/10.5281/zenodo.15238592
Quantifications of in-vivo recordings	This paper	Zenodo data: https://doi.org/10.5281/zenodo.15238751

Highlights.

• Oxtr is highly enriched in the EPd, distinguishing it from the neighboring claustrum

• EPd-Oxtr neurons connect bidirectionally with olfactory and limbic brain areas

• EPd-Oxtr neurons decrease activity in response to novel stimuli

• EPd-Oxtr neurons likely regulate interoceptive states and adaptation to external cues

Data sharing plan

Web visualization of anterograde projectome and monosynaptic input mapping is available at https://kimlab.io/home/projects/EPd_connectivity/. Other experimental data will be freely available on request.