Single-nucleus atlas provides insights into changes in the evolution of the pig hippocampus

Abstract

Background

The large-scale development of pig farming has introduced significant stressors that negatively affect pigs’ mental health, behavior, and production efficiency. The hippocampus, crucial for cognition and stress response regulation, plays a central role in these processes. However, the regulatory mechanisms underlying hippocampal function across pig breeds with different domestication statuses and their implications for behavior and breeding strategies remain unclear.

Results

We performed single-nucleus RNA sequencing (snRNA-seq) on hippocampal tissues from 22,342 cells across three pig breeds: Asian wild boar, Jinhua, and Duroc, representing different domestication statuses. We identified six major hippocampal cell types and annotated 108 breed-specific transcription factors, including GATA2, SPI1, and EBF1. Additionally, we characterized 83 co-expression modules and 50 significant ligand-receptor pairs, such as TGFβ, WNT, and SPP1, revealing complex intercellular communication networks. Oligodendrocyte expression patterns were conserved across all breeds. We identified 194 candidate genes linked to stress resilience, mental health, and feeding behavior, including MC4R, RYR2, PDE10A, and ABCG2. Alzheimer’s disease-related gene enrichment was lower in Duroc pigs, consistent with reduced APOE expression. We also developed the Pig Hippocampus Single-cell Atlas (PHiSA, http://alphaindex.zju.edu.cn:8503/), an open-access database allowing breed-specific hippocampal analyses and validation of gene expression at the single-nucleus level.

Conclusions

This study offers insights into hippocampal function regulation in pigs, focusing on stress resilience, behavior, and productivity. It highlights conserved and breed-specific molecular features of hippocampal cell types and their roles in adaptability and mental health. By integrating single-nucleus data, the research suggests that genetic strategies could be used to improve animal welfare, stress management, and production efficiency in pig breeding programs.

Graphical Abstract

We created a single-nucleus atlas of the hippocampus for pigs representing three domestication statuses: wild boars, a local domesticated breed, and an intensive breed with distinct behavioral traits. This comprehensive atlas revealed both cell type-specific and breed-specific patterns related to their biological functions. Additionally, we identified candidate genes associated with pig mental health across evolutionary stages. These findings strengthen the use of pig breeds as models for human mental disorders and deepen our understanding of hippocampal function at the single-nucleus level. This figure was created with BioRender.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12915-026-02560-4.

Background

The pig farming industry is advancing rapidly toward large-scale and intensive development. However, the intensive production management systems, such as group housing, weaning, transportation, and limited living space impose various stress responses (e.g., fear, anxiety, and depression) on pigs [1], adversely affecting their mental health. Behaviorally, stress in pigs can manifest in several ways, including increased defecation, aggression, infanticide by sows, lethargy, loss of appetite, and postpartum depression [2]. These behavioral changes can lead to a significant decline in pig production performance. For example, pigs under stress from crowding and group housing experience reductions in growth rates of 15.7% and 7.1%, and their feed conversion ratios decline by 15% and 10%, respectively [3], resulting in substantial economic losses for the pig farming industry. Consequently, stress arising from intensive pig farming systems impacts pig production efficiency by affecting the mental health of pigs. However, previous research on the mental health traits of pigs has been limited due to the challenges in measuring these traits, leaving these traits rarely selected in breeding and their genetic basis largely unexplored. Therefore, deciphering the genetic basis of pig mental health traits may unlock potential for genetic improvement, providing an effective new approach to enhance production efficiency and economic benefits in pig farming. This will also significantly improve the innovation and utilization capabilities of pig genetic resources [4].

The domestication of pigs can be traced back to artiodactyl ancestors approximately 40 million years ago. Eurasian wild boars, the direct ancestors of modern domestic pigs, were independently domesticated by humans around 9000 years ago across multiple geographic regions [5]. Prolonged domestication and artificial selection have led to pronounced morphological, physiological, and behavioral differentiation, giving rise to a variety of pig breeds with diverse phenotypic traits. With the development of intensive farming, the selection process has accelerated, further amplifying behavioral and phenotypic divergence among breeds [6]. These breeds not only reflect the profound impact of human activity on species evolution but also carry accumulated genetic variations in their genomes, offering key insights into the genetic basis of complex traits [7, 8]. Asian wild boars, as wild populations native to East Asia, retain ancestral genomic features that have remained largely untouched by domestication. Jinhua pigs, representing traditional indigenous breeds, are known for their superior meat quality and stress resistance. In contrast, Duroc pigs, a widely used lean-type commercial breed, have undergone high levels of genetic improvement through intensive selection, exemplifying the effects of strong artificial selection. Overall, these three breeds—spanning a continuum from wild ancestors to traditional indigenous breeds and modern commercial breeds—offer a coherent framework for studies on comparative genomics, phenotypic plasticity, and the mechanisms of domestication [9]. The distinct behavioral characteristics of these breeds also highlight how domestication and artificial selection have differentially shaped traits related to mental health. Their rich genetic and behavioral diversity offers a valuable model for exploring mental health traits in pigs, with potential implications for understanding similar traits in humans. Importantly, gene expression serves as a critical intermediate link between the genome and phenotypic outcomes, especially in the brain, where gene function is intricately linked to specific cell types.

The hippocampus is an essential structure located in the medial temporal lobe of the brain, playing a critical role in various cognitive functions, particularly in memory and learning. Damage to its function can lead to memory disorders, such as Alzheimer's disease and other types of dementia. Impaired hippocampal function in pigs may also affect their mental health status, which subsequently impacts their productivity [10, 11]. Moreover, the brain structure of pigs is similar to that of humans, making pigs valuable experimental models for neuroscience research. Understanding the functions of the pig hippocampus can provide a theoretical basis for constructing pig models of human mental diseases [12]. However, research on the biological functions of the porcine hippocampus, particularly at the single-nucleus level, remains limited. Increased research efforts on the porcine hippocampus could enhance our understanding of its role in cognitive functions and its potential applications in modeling human mental diseases and improving pig breeding programs.

In this study, we aimed to generate a comprehensive single-nucleus atlas of the pig hippocampus, encompassing three stages of pig breed domestication: wild, domesticated, and selectively bred under intensive farming systems. The selected breeds exhibit distinct behavioral characteristics: Asian wild boars are rough and untamed, Jinhua pigs are docile and calm, and Duroc pigs are well-adapted to intensive farming systems. For example, Duroc pigs have undergone long-term selective breeding under intensive farming systems, leading to genomic variations and differential gene expression associated with mental health traits [7, 8]. These genetic changes may affect hippocampal function, leading to behavioral and physiological differences among the breeds. Using comparative genomics analysis, we systematically characterized gene expression patterns across various cell types and identified breed-specific variations. Additionally, we identified candidate genes related to pig mental health and explored potential roles of specific cell types in human mental diseases. Our results provide novel biological insights into the cellular and molecular architecture of the pig hippocampus, offering valuable resources for exploring pigs as animal models for hippocampus-related human diseases. Additionally, our finding facilitates cross-species comparisons of hippocampal functions at the single-nucleus level, advancing our understanding of mental health in both pigs and humans.

Results

Construction of the porcine single-cell atlas of the hippocampus in multi-breeds

To construct a single-nucleus resolution atlas of the pig hippocampus, we isolated cells from hippocampal samples of three breeds: Asian wild boar (AWB), Jinhua pig (JH), and Duroc pig (DU). The single-nucleus atlas of hippocampus comprised 22,342 high-quality cells (AWB: n = 7495; JH: n = 7495; DU: n = 8463) and covered six major cell types: excitatory neurons (ExN), inhibitory neurons (InN), oligodendrocyte progenitor cells (OPC), oligodendrocytes (OLG), astrocytes (AST), and microglia (MG) (Additional file 1: Fig. S1A). All six cell types were robustly identified across the three breeds (Fig. 1A–D and Additional file 1: Fig. S1B), indicating consistency in cell types across breeds [13]. Despite the uniformity of cell types among breeds, we observed heterogeneity in cell composition across the breeds (Fig. 1E). For example, compared to AWB and DU, the JH showed a lower proportion of OLG (31.20%, Fig. 1E) and increased proportion of neurons, including ExN and InN (32.93% and 2.47%, Fig. 1E). This increased proportion of neurons in domestic pigs suggests an enhanced neurotransmitter transmission process [14]. Additionally, OLG (38.57%) and ExN (27.81%) (Additional file 2: Table 1) had the highest proportions among all cell types in all three breeds, consistent with previous studies [15].

Fig. 1.

The single-nucleus atlas of pig hippocampus. A T-distributed stochastic neighbor embedding (t-SNE) plots of 22,342 cells, grouped by expression similarity and colored by cluster. B Heatmap depicting the expression of top five differentially expressed genes across six identified cell types. The bar color indicates the corresponding cell types. C Matrix plot highlights the relative expression differences of marker genes across different cell types. The bar color indicates the corresponding cell types. D, E Bar chart showing the distribution of hippocampal cell types within each pig breed

Characterization of cellular functions of cell types in the hippocampus

We first identified cell type-specific expressed genes and conducted functional enrichment analyses to investigate the biological functions of each cell type. The genes specifically expressed in AST (256 genes), MG (359 genes), and OLG (411 genes) were enriched in biological processes related to synaptic and nervous system regulation, such as synapse organization, nervous system development, and axonogenesis (Fig. 2A, Additional file 1: Fig. S2A–B and Additional file 2: Table 2). These observations are consistent with their roles as glial cells [16–18]. Additionally, the ExN specific genes were enriched in GO terms related to signaling pathways and metabolism (e.g., regulation of apoptotic signaling pathway) and in KEGG pathways associated with human mental disorders (e.g., Parkinson’s disease and Huntington’s disease pathways, Additional file 1: Fig. S2A–B and Additional file 2: Table 3).

Fig. 2.

Characterization of cellular functions of pig hippocampus. A Diagram summarizing the main functions of the six hippocampus cell types. B Heatmap showing the distribution of transcription factors (left) and cell type-specific active transcription factors (right). C The snRNA-seq UMAP is colored by module eigengene (ME) to represent modules with specific expression in cell types. D Network of hub genes within the MG-specific MG-M5 module. E Violin plot of M3 module expression in OLG. F Gene co-expression network derived from OPC, represented as a two-dimensional UMAP embedding of the TOM. Nodes represent genes, colored by module assignment, and edges represent co-expression relationships between genes and module hub genes. The central panel displays functional enrichment results for the module hub gene

As previous studies have shown that certain human mental disorders are linked to the dysregulation of transcription factors (TFs), particularly in the hippocampus [15], we investigated the cell type-specificity of TFs in this brain region. We identified 108 TFs that exhibit specific activity in cell types of the hippocampus, with the highest number found in microglia (MG) (66 TFs, e.g., IRF5, TAL1, SPI1, IRF8, NFATC2, Fig. 2B). MG, as primary immune cells in the central nervous system, are crucial for inflammatory damage and tissue repair, and they regulate neuronal activity and behavior [19]. Our analysis linked microglial TFs to mental disorders, including neurodegeneration, Alzheimer’s disease (AD), schizophrenia (Scz), Parkinson’s disease (PD), and bipolar disorder (BD) (https://www.grnpedia.org/trrust/), highlighting their importance in these diseases [20]. Notably, members of the NFATC family (NFATC1 and NFATC2) are highly expressed in MG (Fig. 2C), which may target neuroinflammation in PD [21]. Additionally, PFX7 from the RFX family and SPI1 also showed high activity in MG (Fig. 2C), with disruptions in these factors associated with autism, attention-deficit/hyperactivity disorder, intellectual disability, behavioral disorders, and AD [22–25]. These findings highlight the potential of pig as animal models for further research into microglial function in neuropsychiatric conditions. They support the adoption of cell type-specific strategies in both research and treatment, which could enhance the efficacy of therapeutic interventions for mental disorders.

We used the standard hdWGCNA workflow [26] to identify co-expression modules for each major cell type. Module eigengenes (MEs) were then calculated using the entire snRNA-seq dataset to examine the cell type specificity of the expression programs within these modules. This analysis identified 83 co-expression modules across the six major cell types (Additional file 1: Fig. S3). Differential module eigengene (DME) analysis revealed both shared and distinct module expression patterns among these cell types, highlighting specific modules for cell types (Fig. 2D). The expression of hub genes within these modules was found to be cell type-specificity, particularly within cell type-specific modules, which provide insights to cell-specific functions. For example, CX3CR1 gene was identified in the MG-M5 module of MG, and LHFPL3 gene was found in the OPC-M10 module of OPC (Fig. 2E and Additional file 1: Fig. S2C). These genes also serve as markers for their respective cell types [27]. Subsequent functional enrichment analysis of cell type-specific modules revealed that genes within the MG-specific module M5 were enriched in pathways related to synaptic pruning and immune cells (Additional file 1: Fig. S2D), consistent with their known functions [16]. Additionally, genes within modules specific to OPC and OLG—OPC-M10 and OLG-M3—were enriched in pathways related to synaptic signaling, glutamate receptor signaling, and OLG differentiation (Fig. 2F–G and Additional file 1: Fig. S2D), consistent with their reported functions [18]. The above results indicate that the cell type-specific expression modules and their hub genes can reflect key aspects of cell identity and function.

To explore ligand-receptor interaction networks among different cell types, we conducted a ligand-receptor interaction analysis, identified 50 significant ligand–receptor pairs, spanning a total of 982 signaling interactions (Additional file 2: Table 4) across six cell types. This included classical signaling pathways such as TGFβ, WNT, SPP1, PTN, PDGF, CXCL, and CCL (Additional file 1: Fig. S4A). Among these cell types, InN exhibited the highest number of signaling pathways (n = 224, 22.8%) despite their relatively low abundance (Additional file 1: Fig. S4B and Additional file 2: Table 4), suggesting their critical role in coordinating hippocampal functions. We also observed significant differences in pathway strength, particularly between OLG and ExN (Fig. 3A). Previous studies indicate that neuronal activity impacts OLG, potentially contributing to glioma development [28], while interactions with ExN are linked to AD [29]. Additionally, our analysis revealed that primary interactions in the TGFβ pathway occurred between MG and ExN (Fig. 3B). TGFβ is a pleiotropic cytokine associated with various neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS) [30], as well as psychotic disorders like Scz and BP [31]. Notably, key ligand-receptor pairs and TGFβ signaling genes were primarily expressed in MG and neurons, indicating a cell-specific role for TGFβ in regulating mental health (Fig. 3C, D).

Fig. 3.

Intercellular communication between hippocampus cell types. A Network diagram showing the strength of significant ligand-receptor pairs between any two cell populations. Edge width is proportional to the interaction strength. B Hierarchical plot showing the inferred intercellular communication network of TGFb signaling pathway. Circle sizes are proportional to the number of cells in each cell group, and edge width represents the communication probability. C Bar chart illustrating the relative contribution of each ligand-receptor pair to the overall TGFb signaling network. D Violin plot showing the expression distribution of signaling genes involved in the inferred TGFb signaling network

Breed-specific functions of cell type in the hippocampus

To investigate breed-specific expressed genes and their functions, we analyzed differentially expressed genes (DEGs) across breeds for each cell type. In total, 3596 unique DEGs were identified, with the majority enriched in OLG (800 genes) and ExN (1041 genes), particularly in comparisons between DU and the other two breeds (Additional file 1: Fig. S5A and Additional file 2: Table 5). This indicates that during the process of domestication and the evolution of production models, Duroc pigs have undergone more genetic changes. Notably, 1673 of these DEGs were unique to specific breed comparisons (Additional file 1: Fig. S5B and Additional file 2: Table 5). We then conducted functional enrichment analyses of DEGs between different breeds within the same cell type. In AST and MG, DEGs between breeds were similarly involved in the regulation of synaptic organization and neuronal projection development (Fig. 4A). In contrast, DEGs in MG were enriched in the glutamate receptor signaling pathway, which plays a critical role in excitatory neurotransmission within the central nervous system (Fig. 4A). Furthermore, functional enrichment analysis of breed-specific DEGs highlighted distinct functional roles across various cell types. For instance, in MG, DEGs between AWB and JH were primarily associated with cardiac chamber development and morphogenesis. Conversely, DEGs related to ExN between AWB and DU were primarily enriched in purine-containing compound metabolic processes, while DEGs in InN were associated with endoplasmic reticulum regulation (Additional file 1: Fig. S5C and Additional file 2: Table 6).

Fig. 4.

Breed-specific functions in the hippocampus. A Radar map displaying the functional enrichment analysis of DEGs between breeds within the same cell type. The significance is indicated by the black line, representing a P-value < 0.05. B Scatter plot showing the scaled regulon activity of all regulons in each subcellular type of AST. C Violin plot of OLG-specific module expression across cell types for each breed. D Violin plot of DU-specific module expression within AST across three breeds. E The selected gene ontology (GO) terms enriched in the DU-specific co-expression module M4 in AST are displayed. The bar chart shows the log-scaled enrichment for each term. F Co-expression network of hub genes within DU-specific module M4 in AST. G Bar graph indicating the number of signaling pathways involved in cell-type interactions across the three breeds. H Network showing the strength of signaling pathways for the interaction between different cell types across the three breeds. I Chordal graph of multiple ligand-receptor pairs in the NRXN signaling pathways across breeds

To study breed-specific TFs, we analyzed each breed within the same cell type. Overall, MG still exhibited the highest number of enriched TFs. Our analysis also revealed cell type specificity and varying enrichment levels of TFs across breeds (Additional file 1: Fig. S6A). This indicates that the activity of TF regulation varies under different feeding conditions, subsequently affecting the expression of key genes and behavioral phenotypes. For instance, the TF RXRA shows higher activity in DU’s MG; the RXRA gene has been found to potentially act as a risk factor for AD by influencing brain cholesterol metabolism [32] and TFs of the E2F family (E2F7, E2F1, and E2F8) and MYBL1 exhibit higher activity in AWB’s OPC. This suggests that DU have undergone specific changes during the long-term process of intensive farming system. Additionally, the TF GATA2 in AST showed decreasing activity from AWB to JH to DU (Fig. 4B and Additional file 1: Fig. S6B). It is associated with glioma and known to form networks with other factors that regulate GABAergic neuron subtype development and sleep regulation [33, 34]. This may relate to the diurnal behavior of AWB. These findings highlight the role of TFs in regulating cell types of hippocampus and breed-specific functions.

To investigate the conservation and heterogeneity of gene co-expression modules among breeds, we performed co-expression network analysis for the major cell types of hippocampus separately for each breed. Significant differences in the expression patterns among breeds were evidenced by variations in the number of identified modules. Notably, OLG from different breeds exhibit a high degree of similarity in their expression patterns, including the number of co-expressed modules and the functionality of breed-specific modules. These modules, which include hub genes, were primarily associated with protein transport and amino acid transport (Fig. 4C and Additional file 1: Fig. S5D). This high level of conservation suggests that oligodendrocyte-specific expression patterns are largely preserved across breeds. Additionally, this suggests that despite extensive domestication, certain specific expression patterns have been retained. Further, we conducted gene co-expression network analyses within the same cell type across different breeds to identify breed-specific co-expression modules. Among all cell types examined, only AST exhibited a breed-specific co-expression module; this suggests that the changes during the domestication process may be concentrated in AST (Fig. 4D). Specifically, the M4 module displayed DU-specific expression (Fig. 4D), and the specifically expressed modules were predominantly enriched in pathways related to axon and neuron development (e.g., commissural neuron axon guidance, Fig. 4E). The hub genes of M4, including LSAMP and NRXN3 genes, are strongly associated with various mental disorders (Fig. 4F). LSAMP and NRXN3 genes have been linked to MDD and panic disorders [35], AD [36], Scz [37], and developmental delay with motor impairment [38]; this indicates that under high-intensity intensive farming systems, DU exhibit distinct changes in neural development and certain aspects of mental health. These findings suggest that during the long process of domestication and evolution, some genes that regulate fundamental functions have not undergone significant changes, while certain key genes related to behavior and mental health exhibit breed-specific expression. This also highlights that genetic factors play a significant role in influencing behavioral phenotypes, particularly those related to mental health. The presence of breed-specific modules in AST highlights the potential for genetic diversity to impact cognitive and psychological traits, underscoring the importance of considering breed-specific genetic backgrounds in studies of mental health and behavior.

To explore differences in ligand-receptor interaction networks among breeds, we analyzed these interactions across cell types for each breed. DU exhibited a higher number of signaling pathways compared to AWB and JH (Fig. 4G), with noticeable variability in signaling intensity across different cell types (Fig. 4H). This indicates that under changing production environments, DU is evolving toward closer intercellular communication. Pairwise breed comparisons revealed several breed-specific signaling pathways, including OPIOID and SOMATOSTATIN (SST) in AWB, IL17 in JH, and CDH1 and HSPG in DU (Additional file 1: Fig. S7). The NRXN-NLGN pathway, associated with suicide, autism, and Scz [39], showed heterogeneity, with AWB having weaker signaling intensity (Additional file 1: Fig. S8); this may suggest that AWB has a healthier mental state. This pathway’s cell–cell communication was richer in ExN across all breeds (Fig. 4H). Notably, AWB’s MG showed no involvement in this pathway, and OLG lacked NRXN3-mediated signaling communication. In JH, the NRXN3-NLGN1 pathway was observed only from ExN to MG (Fig. 4I). This indicates breed and cell type specificity in key disease-related pathways, as well as the changes in intercellular connections, both in type and strength, during the domestication process. These changes may also indirectly affect brain function and phenotypes.

Pig mental health candidate genes

To determine genes associated with pig behavior and mental health, we compared our identified DEGs with 1,549 human-to-pig homologous genes linked to mental disorders from our previous study [7, 8]. This comparison enabled us to connect transcriptional differences across breeds to mental health-related phenotypes. As a result, we identified 194 DEGs classified as candidate genes for pig mental health (Additional file 1: Fig. S9A-B and Additional file 2: Table 7). These genes are primarily distributed between AWB and DU; the number of differentially expressed genes between these two breeds is the highest, as anticipated. Functional enrichment analysis revealed that these candidate genes were significantly enriched in pathways related to synapses, AD, ALS, neurodegenerative diseases, and neuroactive ligand-receptor interactions (Fig. 5A). This enrichment further supports the reliability of our candidate gene set.

Fig. 5.

Pig mental health candidate genes. A The barplot displays the functional enrichment of candidate genes related to pig mental health (adjusted P-value < 0.05). The colors represent the adjusted P-values. B Dot plot illustrating the enrichment of pig traits (P-value < 0.05) from the Pigbiobank dataset. Point size indicates fold enrichment, and color represents the P-value. C, D Manhattan plots of statistically significant (P-value < 10⁻⁸) SNPs associated with porcine backfat thickness (DFI) trait (C) and loin muscle area (RFI_ADG) trait (D) on autosomes. Red markers denote SNPs related to the DFI and RFI_ADG traits within pig mental health candidate genes. Venn plots show the overlap between candidate genes associated with traits and pig mental health candidate genes. The left circle represents candidate genes at loci significantly associated with traits, the right circle represents all pig mental health candidate genes, and the overlapping region indicates genes linked to the marked SNPs. P-values were calculated using the hypergeometric distribution test

Next, due to the close relationship between pig mental health and production efficiency, further annotation and validation of the gene set are essential. We conducted enrichment analysis using the PigBiobank database [40, 41] to explore correlations between these genes and production traits. The results indicate that the mental health-related genes showed cell type-specific enrichment in economic traits such as average daily gain, backfat thickness, number of live piglets, and fatty acid traits (e.g., arachidonic acid and palmitic acid) (Fig. 5B). This suggests that these genes influence both mental state and production performance, leading us to speculate that differences in pig mental health status may subsequently impact their production efficiency.

Additionally, we believe that abnormalities in mental health may manifest in specific behaviors. Further analysis using GWAS data indicated the mental health-related genes were significantly enriched in loci for daily feed intake (DFI) and residual feed intake adjusted for average daily gain (RFI_ADG) (Additional file 2: Table 8); the candidate genes MC4R and RYR2 were associated with DFI (Fig. 5C), while PDE10A and ABCG2 were linked to RFI_ADG (Fig. 5D). The MC4R gene, which plays a critical role in appetite regulation [42], is upregulated in AWB’s ExN compared to the other two breeds (Additional file 1: Fig. S10A). The RYR2 gene, associated with neuronal hyperactivity and Alzheimer’s [43], was underexpressed in AWB’s MG (Additional file 1: Fig. S10B). The ABCG2 gene was upregulated in AWB’s OLG, indicating a distinct disease profile (Additional file 1: Fig. S10C), while the PDE10A gene, linked to hyperactivity [44], was upregulated in AWB’s AST (Additional file 1: Fig. S10D and Additional file 2: Table 8). These results indicate that different environments and feeding conditions can affect the expression of key genes, subsequently leading to variations in behavioral phenotypes and production status. Moreover, the identification of candidate genes related to mental health suggests a genetic basis for behavioral traits in pigs. This insight informs breeding strategies aimed at improving both welfare and productivity.

Cell type associations with human mental disease

To explore the potential of pigs as model organisms, we used the LDSC approach to evaluate the enrichment of human neurological diseases across different cell types. The major diseases analyzed included anorexia, AD, major depressive disorder (MDD), autism spectrum disorder (ASD), BD, and Scz. Our analysis revealed that AD was significantly enriched in MG, consistent with previous studies [45, 46]. Additionally, AD-related pathways were enriched in AST, with the degrees of enrichment varying among breeds. Importantly, DU exhibited lower enrichment for AD in AST compared to the other two breeds, suggesting a relatively weaker association between DU and AD. Furthermore, we found that genes associated with Scz, BD, and depression were primarily enriched in neuronal cells, including ExN and InN. These results indicate that different types of mental disorders exhibit varying degrees of enrichment across specific cell types, reflecting the distinct functions of those cells (Fig. 6A and Additional file 2: Table 9).

Fig. 6.

Disease traits association with cell types. A Heatmap of the association of curated common human traits and diseases with the cell types across three breeds, based on LSDC analysis. B Expression of AD risk genes in different cell types across the three pig breeds

To investigate breed heterogeneity associated with AD, we analyzed 23 AD risk genes [47] using AUCell to calculate their activity scores across cell types. The activity scores for AST and MG showed breed variability, with DU showing weaker enrichment compared to the other breeds, consistent with the above research. Further examination of these AD risk genes across various cell types revealed both cell type-specific and breed-specific expression patterns. For instance, SCIMP, TREM2, and CASS4 genes were enriched in MG, while ADAM10, PICALM, and BIN1 genes were enriched in OLG. CLU and SLC24A4 genes were primarily expressed in AST of AWB and JH (Fig. 6B). Additionally, APOE, a major genetic risk factor for AD [48], was enriched in AST and OLG of AWB and JH but was expressed at low levels in DU (Fig. 6B and Additional file 2: Table 10). These findings highlight the association of AD with specific hippocampal cell types and reveal breed-specific differences in gene expression. The reduced enrichment of AD-related genes in DU suggests that certain pig breeds may be more suitable models for studying human mental disorders, particularly those related to AD. The changes observed in DU, compared to other breeds, are attributed to the intensification of production models.

PHiSA

To facilitate the exploration of hippocampal atlases for pigs, we developed a user-friendly website to provide a query service using the PHiSA (Pig Hippocampus Single-cell Atlas, http://alphaindex.zju.edu.cn:8502/) for researchers, which provides the query of gene expression in cells (Additional file 1: Fig. S11).

Discussion

The hippocampus is essential for memory, learning, and spatial navigation, and its dysfunction is an early indicator of various mental health disorders, including AD [11, 49]. Despite its critical role, the cellular regulatory mechanisms within the pig hippocampus and their conservation with humans remain poorly understood. Recent advancements in snRNA-seq have facilitated the identification of major cell types, marker genes, and cross-species comparisons in the hippocampi of humans, monkeys, and mice. However, studies exploring evolutionary changes in pig brain function across different breeds and their relationship with productivity traits are limited. To address this gap, we selected three representative pig breeds from distinct evolutionary periods—AWB, JH, and DU—to construct a single-nucleus atlas of the hippocampus, aiming to elucidate alterations in mental health and productivity traits during evolution on a large scale. We thus developed a comprehensive single-nucleus atlas of the hippocampus, comprising 22,342 cells and six conserved cell types, from three distinct pig breeds: AWB, JH, and DU. These breeds represent different stages of lineage and industry evolution, respectively. Moreover, we developed a user-friendly website to provide a search service for investigating gene expression of users’ genes of interest.

Our systematic analysis of the biological functions of hippocampal cell types revealed distinct functional preferences. Glial cells—AST, MG, and OLG—were enriched in pathways related to synapses and neural system regulation. In contrast, neuronal cells—ExN and InN—showed enrichment in metabolic pathways and pathways associated with human mental health disorders, such as PD and Huntington’s diseases. Moreover, TFs exhibited cell type specificity in their functions. TFs regulating neuronal activity, synaptic interactions, behavior, and neurological diseases like AD, Scz, PD, and BD were primarily concentrated in MG. For example, TFs from the NFATC family [21], RFX family [25], and SPI1 were specifically enriched in MG and associated with conditions such as autism, intellectual disability, and AD [22]. Furthermore, TF regulatory activity demonstrated breed specificity. The TF GATA2 in AST showed a decreasing activity gradient from AWB to JH to DU. GATA2 and its network regulate the development of GABAergic neuron subtypes and sleep modulation characteristics [33]. Additionally, the RXRA gene, which exhibited higher expression activity in DU’s MG cells, may act as a risk factor for AD by influencing brain cholesterol metabolism [32]. Co-expression module analysis identified genes specifically expressed in DU-associated AST modules, such as genes LSAMP and NRXN3, which are linked to MDD, AD, and Scz [35, 37]. These findings suggest that the changes during the evolutionary process are greatly reflected in mental health traits, while genetic factors significantly influence behavioral phenotypes and mental health, with breed-specific regulatory mechanisms playing a crucial role.

In our investigation of cell communication, we discovered that certain pathways exhibit cell-specific communication intensities. For example, the TGFβ pathway, associated with numerous mental health diseases, primarily mediates interactions between MG and EN, with pathway-related genes mainly expressed in these cell types. This indicates that TGFβ regulation in mental health is cell-specific [31]. Additionally, the NRXN pathway demonstrated breed heterogeneity; unlike in JH and DU, AWB’s MG lacked interaction with this pathway, which has been previously linked to suicide, autism, and Scz [39]. These observations suggest that key disease-related pathways exhibit both breed and cell type specificity, highlighting the complexity of genetic regulation in mental health.

During the process of pig domestication, in addition to genetic changes, alterations in gene expression also occur. To investigate the DEGs associated with mental health during domestication, we compared these DEGs with previously identified human-pig mental health reference genes. This comparison resulted in a candidate gene set of 194 genes linked to mental health pathways, which are also associated with productivity traits and economic characteristics. We hypothesized that behavioral and habitual differences among these breeds would be reflected in corresponding changes in brain function. Consistently, we identified 3596 DEGs in the hippocampus across breeds, primarily enriched in oligodendrocytes (800 genes) and excitatory neurons (1041 genes), particularly between Duroc breed and the other two breeds. Additionally, four genes—MC4R, RYR2, PDE10A, and ABCG2—are related to feeding traits and have been implicated in neurological disorders and behavior. For instance, the MC4R gene plays a crucial role in regulating appetite, energy homeostasis, and body weight in the central nervous system, making it a primary target for anti-obesity drug development. Studies have shown significant upregulation of the MC4R gene in the hippocampus of AWB, potentially contributing to their more robust physique and lower fat content [42, 50]. These findings highlight the importance of the identified gene set for future breeding analyses and model organism research, demonstrating that the pig hippocampus influences both mental health and productivity traits.

Finally, our analysis confirmed that AD is significantly enriched in MG and AST, aligning with human studies [45, 46]. Pigs have emerged as prominent model animals for human disease research. Beyond studying human developmental processes, congenital diseases, and pathogen response mechanisms, pigs also hold significant potential in the field of mental health [51]. Studies have indicated that the anatomy of the pig brain bears certain similarities to the human brain [12], and there is notable similarity at the single-nucleus transcriptome level [52]. Furthermore, the proportions of different cell types between humans and pigs are highly similar, with a conserved expression of cell type-specific genes across these species [15]. By examining a set of AD risk genes, we found that several genes displayed both cell type-specific and breed-specific expression patterns in the hippocampus [47]. Differential gene expression was predominantly concentrated in glial cells. Notably, the APOE gene, the largest genetic risk factor for AD [48], was enriched in AST and OLG of AWB and JH breed but showed relatively low expression in DU. This breed-specific expression pattern underscores the association of AD with particular cell types and suggests that certain pig breeds may be more suitable as models for studying mental health disorders, particularly AD.

One limitation of our study is the small sample size due to the high cost of snRNA-seq, which may partially affect the precision of proportion assessments and the representativeness of the findings. Previous high-quality research has acknowledged that multiple-sample RNA-Seq datasets can provide more information and reduce errors compared to single-sample datasets [53]. However, our single-cell analysis achieves a high level of specificity at the cell type level with a large number of cells, making it less susceptible to the drawbacks of low sample size. This approach enables us to capture detailed information about individual cell populations and their functional states, addressing some concerns associated with sample size limitations. Overall, despite this limitation, our comprehensive cell atlas offers valuable insights into the cellular heterogeneity and biological function evolution in the hippocampus during pig domestication.

In summary, our research provides a detailed single-nucleus atlas of the pig hippocampus across three domestication statuses, uncovering conserved and breed-specific gene expression patterns related to mental health and productivity traits. The identified candidate genes and modules offer valuable insights for genetic enhancement and the selection of pig models for human mental health studies. This comprehensive resource facilitates future investigations into the genetic basis of cognitive functions and mental health disorders, enhancing breeding strategies and the use of pigs as models for human neurological diseases.

Conclusions

In conclusion, we developed a single-nucleus atlas of the hippocampus for three domestication status pig breeds with distinct behavioral traits, revealing both cell type-specific and breed-specific patterns related to their biological functions. Our analysis identified genes linked to mental health across evolutionary stages and highlighted key targets for genetic improvement. These findings enhance the use of pig breeds as models for human mental disorders and advance our understanding of hippocampal function at the single-nucleus level. This resource supports cross-species comparisons and the selection of optimal animal models for human neurological disease research.

Methods

Pig hippocampus tissue collection and ethics statement

Hippocampus tissue samples were collected from three healthy pigs, including an Asian wild boar, a Jinhua pig, and a Duroc pig at an age of adulthood (Additional file 2: Table 1). Tissues of the hippocampus were obtained within 30 min post-mortem and rapidly frozen by liquid nitrogen until further processing. All experiments involving pigs were approved by the Institutional Animal Care and Use Committee of Zhejiang University (ZJU20220262).

Sample preparation and single-nucleus RNA sequencing

Frozen tissue was homogenized in 3 ml of lysis buffer using a glass dounce homogenizer. The tissue was dounced 10 times with a loose pestle, followed by an additional 5 strokes with a tight pestle to ensure thorough homogenization. The homogenate was allowed to lyse in a total volume of 5 ml of buffer for 5 min. Subsequently, 5 ml of wash buffer was added to the sample. The mixture was passed through a 30-μm cell strainer to remove debris and then centrifuged for 5 min at 500 × g. After pelleting, the nuclei were resuspended in 5–10 ml of wash buffer by pipetting up and down 8–10 times to ensure complete resuspension. Following three wash steps, the nuclei were resuspended in 1 ml of wash buffer and mixed with 25% Optiprep. This mixture was carefully layered onto a 29% Optiprep cushion and centrifuged for 30 min at 10,000 × g to purify the nuclei. Finally, the purified nuclei were resuspended in wash buffer and subjected to three additional washes to remove any residual Optiprep and contaminants. Following the manufacturer’s instructions, the snRNA-Seq libraries were prepared using 10X Genomics Chromium Single-Cell 3.1’ kit (V3.1), followed by sequencing by Illumina NovaSeq 6000 in Novogene.

Single-nucleus RNA analysis and cell type annotation

We first used fastp [54] (v0.23.1) with default parameters to process raw snRNA-seq reads, obtaining clean reads. These clean reads were aligned to the Sus scrofa 11.1 assembly [55] to generate a single-nucleus transcript expression matrix (measured as read counts per gene) using CellRanger (v7.0.0) [56]. Next, we used Scanpy (v1.9.1) [57] with specific parameters for quality control, removing cells with doublets, fewer than 200 or more than 7500 genes, or a mitochondrial read fraction greater than 50%, resulting in 22,342 high-quality cells (Additional file 1: Fig. S12). The final single-cell gene expression matrix was generated by normalizing and scaling the gene counts matrix using Scanpy (v1.9.1) [57]. For cell clustering and cell type annotation, we first performed batch corrections on three libraries using the Harmony method “sc.external.pp.harmony_integrate”. We then used the top 2,000 highly variable genes for dimensionality reduction and Leiden clustering (resolution 0.5–2) with default parameters. All cell clusters were annotated into cell types based on marker genes collected from the literature (Additional file 2: Table 11).

Cell type-specific genes and differentially expressed genes (DEGs) identification

The Wilcoxon rank-sum test was applied to identify cell type-specific genes and differentially expressed genes (DEGs) with the “scanpy.tl.rank_genes_groups” function in Scanpy (v1.9.1) [58]. Cell type-specific genes were defined by adjusted P-values (Benjamini–Hochberg method, BH) lower than 0.05. DEGs were defined by adjusted P-values (BH) lower than 0.05 and |log₂FC| greater than 1.5.

Functional enrichment analysis

Functional enrichment analysis was conducted using the R package ClusterProfiler (v4.10.1) [59]. The BH method was employed to correct for multiple testing. GO and KEGG terms with adjusted P-values less than 0.05 were considered significantly enriched.

High-dimensional weighted correlation network analysis (hdWGCNA)

We used the R package hdWGNCA (v0.2.19) [60] to identify intercellular co-expression modules. We selected a minimum soft threshold of greater than or equal to 0.8 for fitting the scale-free topology model to construct a co-expression network comprising genes expressed in at least 5% of cells. Co-expression modules associated with cell types or breeds were identified by calculating the Pearson correlation coefficient between co-expression modules and the relevant traits. Additionally, we utilized Seurat’s AddModuleScore function to calculate the expression activity (module score) of hub genes for each module within every cell.

Transcription factor-related gene regulatory network inference

We used pySCENIC (v0.12.1) [61] to infer the gene regulatory network. The TF genes were collected from the pySCENIC’s GitHub repository (https://github.com/aertslab/pySCENIC). The “—grn” and “—ctx” functions were used to generate co-expression modules and extract these modules using human TF motif information, respectively. The activities of the discovered regulons in all cells were evaluated with the “—aucell” function.

Ligand-receptor cell communication analysis

We employed CellChat (v0.0.1) [62] to predict ligand-receptor interactions and cell–cell communication networks. Receptor–ligand pairs collected from the human library (https://github.com/ventolab/cellphonedb-data) and scaled single-nucleus gene expression were used as input data to determine cell–cell interactions. Cell–cell interactions with a significance P-value less than 0.05 were kept for determining the final interaction pairs of cell types.

Association of human GWAS and genetic disease data with pig hippocampus cell types

To test the enrichment of cell type-specific genes related to human mental diseases, we applied linkage disequilibrium score regression analysis (LDSC) (https://github.com/bulik/ldsc) [63, 64]. We selected the top 500 specific to each tested cell type. Genome coordinates were converted to hg19 reference build using an orthologous gene list downloaded from Ensembl. Summary statistics for each trait were collected from the UK Biobank database or published studies (https://data.broadinstitute.org/alkesgroup/UKBB). After annotating the genes and calculating East Asian LD Scores, we performed regression analysis using the original LDSC parameters. This process enabled us to derive cell type-specific LD Scores, providing insights into the genetic architecture of human traits across different cell types.

Abbreviations

snRNA-seq

Single-nucleus RNA sequencing

DEGs

Differentially expressed genes

AWB

Asian wild boar

JH

Jinhua pig

DU

Duroc pig

ExN

Excitatory neurons

InN

Inhibitory neurons

OPC

Oligodendrocyte progenitor cells

OLG

Oligodendrocytes

AST

Astrocytes

MG

Microglia

TFs

Transcription factors

AD

Alzheimer’s disease

Scz

Schizophrenia

PD

Parkinson’s disease

BD

Bipolar disorder

MEs

Module eigengenes

DME

Differential module eigengene

ALS

Amyotrophic lateral sclerosis

DFI

Daily feed intake

RFI_ADG

Residual feed intake adjusted for average daily gain

MDD

Major depressive disorder

ASD

Autism spectrum disorder

Authors’ contributions

Z.W. conceived and designed this study; LY.X. and Z.W. wrote the manuscript; Tissue samples were collected by LY.X., QQ.X., XY.W., S.L., and JY.J.; LY.X. performed the bioinformatics analysis; PF.Y., CY.C., JM.G., and JB.S. optimized the technical approach; Z.Z., QS.W., and YC.P. provided useful suggestions for the computational framework; XY.W. provided data background for the research; ZY.Z., and JY.J. provided background research for this study; W.Z., QY.W., LY.X., YZ.Z., and XY.W. provided useful input for the analyses. All the authors have read and approved the final manuscript.

Funding

This work was supported by the National Key Research and Development Program of China (grant no. 2022YFF1000500 and 2021YFD1200802), the National Natural Science Foundation of China (grant nos. 32172691), the Hainan Province Science and Technology Special Fund (ZDYF2025XDNY068), and the Key Research and Development Program of Yangzhou (grant nos. YZ2024047).

Data availability

All single-cell datasets were available for interactive exploration and download in our interactive web portal (http://alphaindex.zju.edu.cn:8502/). In addition, the raw single-cell transcriptome data in this article has been uploaded to the National Genomics Data Center (NGDC), with the BioSample ID PRJCA052357 [65]. The adapted code used in the Methods section, along with certain visualization code, can be obtained from the authors via email upon request.

Declarations

Ethics approval and consent to participate

All experiments involving pigs were approved by the Institutional Animal Care and Use Committee of Zhejiang University (ZJU20220262).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.