(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Defining cellular diversity at the swine maternal–fetal interface using spatial transcriptomics and organoids [1]

['Cole R. Mccutcheon', 'Department Of Integrative Immunobiology', 'Duke University School Of Medicine', 'Durham', 'North Carolina', 'United States Of America', 'Allyson Caldwell', 'Liheng Yang', 'Elisa Crisci', 'Department Of Population Health']

Date: 2025-09 

The placenta is a dynamic, embryo-derived organ essential for fetal growth and development. While all eutherian mammals have placentas composed of fetal-derived trophoblasts that mediate maternal–fetal exchange, their anatomical and histological structures vary across species due to evolutionary divergence. Despite the cellular heterogeneity of porcine trophoblasts in vivo, understanding the mechanisms driving porcine placental development has been limited by the lack of in vitro models replicating this heterogeneity. In this study, we derived swine trophoblast organoids (sTOs) from full-term porcine placentas, retaining key transcriptional signatures of in vivo trophoblasts. To identify conserved cell populations, we integrated Visium spatial transcriptomics from mid-gestation porcine placentas with single-cell transcriptomics from sTOs. Spatial transcriptomics revealed novel markers of the porcine uterus and placenta, enabling precise separation of histological structures at the maternal–fetal interface. The integration of tissue and sTO transcriptomics showed that sTOs spontaneously differentiate into distinct trophoblast populations, with conserved gene expression and cell communication programs. These findings demonstrate that sTOs recapitulate porcine placental trophoblast populations, offering a powerful model for advancing placentation research. Our work also provides a spatially resolved whole-transcriptome dataset of the porcine maternal–fetal interface, opening new avenues for discoveries in placental development, evolution, and health across mammals.

Funding: This work was supported by funds from Duke University (C.B.C.). Placental tissue samples were derived from work supported by the Foundational and Applied Science Program, project award no. 2023-67015-39338 (J.A.P.), from the U.S. Department of Agriculture’s National Institute of Food and Agriculture (NIFA), https://nifa.usda.gov . The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability: All code is publicly available on the Coyne Lab Github ( https://github.com/CoyneLabDuke/Swine-maternal-fetal-interface ) and archived via Zenodo ( https://doi.org/10.5281/zenodo.15747602 ). All raw and mapped data obtained via Bulk RNA-Sequencing, Single Cell, and Visium Spatial Transcriptomics data generated in this paper are deposited on the Gene Expression Omnibus (GSE296975 and GSE301436) and additionally are available via the Sequencing Read Archive (under bioproject number PRJNA1252046). Additionally, a searchable, user-friendly version of the Visium Spatial Transcriptomics dataset has been uploaded at https://coynelab.shinyapps.io/myshinyapp/ .

In human and murine systems, organoids have proven to be a valuable tool for studying complex biological systems in part because they can be cryopreserved, are scalable, and are often tractable [ 31 ]. Furthermore, as organoids spontaneously differentiate into various cell populations, they represent one of the most physiologically relevant in vitro models [ 31 ]. We and others have successfully derived trophoblast organoids from human and non-human primate (NHP) placentas, demonstrating that they recapitulate the cellular heterogeneity and gene expression patterns of the human and NHP placenta [ 32 – 37 ]. Because of the success of trophoblast organoid models, we sought to develop a similar model of the swine placenta. Here, we show the derivation of swine trophoblast organoids (sTOs) from full-term placental tissue. Our data demonstrate that sTO can be expanded, passaged, and cryopreserved, and recapitulate the expression of canonical markers of the swine placenta. Using Visium spatial transcriptomics of mid-gestation swine placentas, we identified the gene signatures of diverse trophoblast populations in vivo and identified novel maternal–fetal crosstalk pathways. By coupling spatial transcriptomics of whole placentas with single-cell transcriptomics from sTO, we found that organoids spontaneously differentiated into various trophoblast populations, accurately replicating their gene signatures. Additionally, we identified gene changes along a differentiation trajectory in sTOs, revealing signatures associated with the differentiation of distinct swine trophoblast subpopulations. Taken together, these data demonstrate that sTO provide an accurate and comprehensive model of the swine placenta, offering a powerful tool for advancing studies of porcine placentation and reproductive biology in vitro.

Another major hurdle to furthering our understanding of porcine placentation is the lack of suitable in vitro models. Limited access, coupled with high expense render in vivo and ex vivo models challenging. While porcine trophoblast cell lines, such as PTr2 [ 27 , 28 ], Jag-1 [ 29 ], and pTR cells [ 30 ] are available, these cells likely represent pre-implantation trophoblasts. Whether these cell lines recapitulate later stages of placental development, or the cellular heterogeneity observed in vivo remains unclear.

Unlike human trophoblasts, which have been heavily characterized and subtyped [ 12 , 18 ], the properties and subpopulations of swine trophoblasts are comparatively unknown, but several lines of evidence suggest that subpopulations exist. First, the point of contact between the fetal trophoblasts and uterine epithelium, termed the maternal–fetal interface, is extensively folded to enhance placental surface area [ 19 ]. Along these folds, it has been reported that aquaporins [ 20 ] and various transporters vary in expression [ 21 ], suggesting that trophoblast polarization along these folds occurs. Second, the porcine placenta contains areolae, a cluster of specialized, columnar trophoblast cells that sit atop uterine glands and facilitate the transfer of uterine gland secretions to the developing fetus [ 22 ]. It is known that areolar trophoblasts are primed to uptake uteroferrin compared to other trophoblasts [ 23 ] and express specific markers, such as cathepsin L1 (CTSL) [ 24 ], suggesting a unique pattern of differentiation. Lastly, recent studies in early gestation porcine placentas have revealed subpopulations of trophoblasts, which display unique expressional and putative functions [ 25 , 26 ]. While these data serve as a critical advancement, these studies are unable to spatially resolve these populations to known structures such as the interface or the areola, thereby limiting the interpretation of these data [ 25 , 26 ].

In humans, trophoblasts consist of several subpopulations [ 10 – 12 ]. The three primary classes of trophoblasts include cytotrophoblasts (CTBs), extravillous trophoblasts (EVTs), and the multinucleated syncytiotrophoblast (STB). CTBs represent the mononuclear, proliferative, undifferentiated trophoblast population of the placenta [ 10 ] whereas EVTs are highly invasive, immunosuppressive cells, that primarily serve to remodel the maternal spiral arteries [ 13 – 15 ]. Conversely, the STB is a terminally differentiated multinuclear trophoblast population that covers the entire surface of the chorionic villi. They serve dual roles as both a barrier to infection and a critical site layer for nutrient and gas transfer [ 16 , 17 ].

Placental structure is commonly categorized by the tissue layers separating the maternal and fetal vasculature [ 7 , 8 ]. In the most invasive placental type, termed hemochorial, the placenta erodes the uterine epithelium and remodels the uterine vasculature resulting in the influx of maternal blood which directly bathes the placental epithelial cells, termed trophoblasts [ 7 ]. This type of placentation is observed in many species, including humans and rodents. Conversely, porcine placentae are considered the least invasive placental type, termed epitheliochorial. In this system, the uterine vasculature, stroma, and epithelium remain intact due to a lack of invasion by the trophoblast cells [ 7 ]. In this case, the primary contact point for the placenta is the uterine epithelium. Critically, regardless of the placental type or structure, all gas and nutrient transfer must occur across the trophoblast layer [ 9 ].

The defining feature of eutherian mammals is the presence of the fetal-derived placenta. The placenta is critical for fetal growth due to its numerous functions ranging from nutrient transfer [ 1 , 2 ], steroidogenesis [ 3 ], and immunological defense [ 4 , 5 ]. While these placental functions are largely maintained across species, the placenta has repeatedly evolved via a combination of convergent and divergent evolution resulting in variable placental morphologies across species [ 6 , 7 ].

Patterning analysis revealed that the senders and receivers, were conserved across several of these signaling pathways, suggesting concordance between sTO and tissue trophoblast signaling ( Fig 8C – 8F ). Some of these pathways displayed bidirectional crosstalk between interface trophoblasts and areolar trophoblasts, such as the ANGPTL and CDH pathways, which displayed extensive autocrine and paracrine signaling ( Fig 8G – 8N ). The ANGPTL pathway displayed a high degree of both autocrine and paracrine signaling between interface and areola trophoblasts, primarily due to the widespread expression of integrins, syndecans, and ANGPTLs ( Fig 8G – 8J ). While we similarly found that areolar trophoblasts display both autocrine and paracrine signaling in the CDH pathway, we found that interface trophoblasts do not display the same ability to signal in an autocrine manner, likely due to the lack of CDH4 expression ( Fig 8K – 8N ). In contrast, we found that some pathways were cell type specific, such as the AGRN pathway, which is restricted to interface trophoblasts ( Fig 8O and 8P ). While we detected high levels of DAG1 in both populations, we found that AGRN was restricted largely to interface trophoblasts ( Fig 8Q and 8R ). Importantly, we found that in many of these pathways, including AGRN, CDH, and ANGPTL, displayed a high degree of similarity in signaling patterns between these systems highlighting the utility of sTO in studying inter-trophoblast communication in the placenta ( Fig 8G – 8R ).

A) Venn Diagram showing the number of unique and shared trophoblast communication pathways between sTO and porcine trophoblasts in vivo. B) Heatmaps showing the number (top left panel) and strength (top right panel) of interactions between clusters within sTO. Similarly, the bottom panels show the number (bottom left panel) and strength (bottom right panel) of interactions within porcine trophoblasts from spatial transcriptomics data. In both cases Color indicates increased number (left) or strength (right) between the sending cluster and the receiving cluster. C–F) Heatmaps showing the signaling patterns (top panels) and the shared signaling pathways contributing to those patterns, for sTO and porcine trophoblasts from spatial transcriptomics. C, D) shows the outgoing patterns for C) porcine trophoblasts obtained via spatial transcriptomics and D) sTO. E, F) shows the incoming patterns for E) porcine trophoblasts obtained via spatial transcriptomics and F) sTO. For upper panels, coloring indicates relative cluster contribution to a pattern with red being high and blue being low contribution level. For lower panels, color indicates the relative contribution of a pathway to a given pattern. G, H) Chord diagrams showing ANGPTL signaling within both G) Visium trophoblasts and H) sTO. I, J) Dot plots showing the expression level of pathways within the ANGPTL pathway for I) Visium trophoblasts and J) sTO. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. K, L) Chord diagrams showing CDH signaling within both K) Visium trophoblasts and L) sTO. M, N) Dot plots showing the expression level of pathways within the CDH pathway for M) Visium trophoblasts and N) sTO. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. O, P) Chord diagrams showing AGRN signaling within both O) Visium trophoblasts and P) sTO. Q, R) Dot plots showing the expression level of pathways within the AGRN pathway for Q) Visium trophoblasts and R) sTO. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980980, GSM8980981, GSM8980982, GSM8980983, GSM8980992, GSM8980993, GSM8980994.

Crosstalk within the human placenta is well documented and controls processes such as fetal development and immunological tolerance. Our data indicates that the porcine placenta similarly displays a high level of intraplacental crosstalk, but additionally displays crosstalk between various trophoblast subpopulations, including crosstalk between the areolar trophoblasts and interface trophoblasts. Given that sTO contain both areolar and interface trophoblast populations, we wanted to better characterize these inter-trophoblast crosstalk events within the porcine placenta and assess how well these interactions are recapitulated in sTOs. To do this, we first subsetted both the Visium data and sTO datasets to only contain areolar trophoblast and interface trophoblast subpopulations. For clarity, areola trophoblast clusters were combined into a singular cluster for each dataset. We then used Cellchat to assess cellular communication pathways present between these populations. This analysis revealed extensive crosstalk in sTOs and in tissue derived trophoblasts with substantial overlap in the signaling pathways between these systems ( Fig 8A and 8B ). Of note we found 28 shared signaling pathways between sTO and tissue trophoblasts ( Fig 8A ). In both sTO and tissue trophoblasts, we observed that interface trophoblasts provide more substantial signaling than areolar trophoblasts ( Fig 8B ).

A) UMAPs showing Slingshot calculated trajectory of trophoblasts within sTO. Top panel shows the trajectory across labeled clusters, whereas bottom panel shows the trajectory across a pseudotime. B) Heatmap showing temporally regulated genes across the calculated pseudotime trajectory. Individual columns represent individual cells that are ordered along the calculated pseudotime. Identities are positions along the pseudotime are shown above the heatmap. C) Representative expression plots showing gene expression changes along the calculated pseudotime that are enriched in the proliferative stem cell population. D) Representative expression plots and spatial feature plots showing genes that are enriched during the transition to Interface-Trophoblasts and their expression pattern within our spatial transcriptomics dataset. E, F) Representative expression plots and spatial feature plots showing genes that are enriched during the transition to Areola-2 trophoblasts and their expression pattern within our spatial transcriptomics dataset. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980980, GSM8980981, GSM8980982, GSM8980983, GSM8980992, GSM8980993, GSM8980994.

Organoids provide a powerful model for defining differentiation trajectories, as they enable the study of cell fate transitions from progenitor cells to more differentiated cell states in a controlled environment. We applied Slingshot trajectory analysis to sTOs to trace the dynamic progression of trophoblast cells along distinct developmental pathways. To do this, we first subsetted the sTOs single-cell dataset to include major cell populations, which included proliferating trophoblasts (ProfTroph), interface trophoblasts (IntTroph), and Areola-2 trophoblasts. We then performed Slingshot trajectory analysis by first setting the starting point at the proliferating trophoblasts, which displayed the highest expression of known stemness markers, including MKI67 and TOP2A. We then allowed inference of downstream differentiation paths without further constraint. This analysis revealed a single differentiation trajectory, starting in ProfTroph, progressing through IntTroph, and terminating in Areola-2 ( Fig 7A ). To define gene expression changes along this trajectory, we applied the fitGAM function to model dynamic gene expression patterns to assess how specific gene expression profiles evolved across the different stages of trophoblast differentiation. Many genes were temporally regulated along the trajectory, with some showing distinct patterns of upregulation and downregulation, while others exhibited cyclical regulation across the different stages of trophoblast differentiation ( Figs 7B and S5 and S11 Table ). Notably, TOP2A, HMGB2, and PRC1 were strongly downregulated along the trajectory, suggesting a loss of a proliferative capacity as differentiation progresses ( Fig 7B and 7C ). In contrast, GRN and ITM2C, both of which are known to be expressed in the swine placenta, were markedly increased in expression in IntTroph, but then became downregulated in the Areola-2 differentiated state ( Fig 7D , left). Similarly, several genes, including NUS1, NUPRR1, SQSTM1, and ARF4, all increased in expression at the terminal areola-2 differentiation stage ( Fig 7E and 7F , left). To validate these findings from the trajectory analysis of sTOs, we compared the gene expression patterns to the Visium spatial transcriptomics data. This comparison revealed a high level of concordance between the temporally regulated markers identified in sTOs and their expression in corresponding tissue structures, further confirming that sTOs effectively recapitulate the differentiation processes observed in vivo ( Figs 7D – 7F and S5A – S5F ).

In vivo, interface trophoblasts and areola trophoblasts have distinct functions. Our data indicated that sTOs differentiated into both of these populations, yet whether these populations retained the functional capacity of their in vivo counterparts remained unclear. To this end, we examined the functional pathways active in both sTOs and tissues across trophoblast subpopulations. Given that IntTroph and Areola-2 clusters had the highest probability scores and shared markers between sTO and tissue, we chose to focus this analysis on these two clusters. We performed differential expression analysis between interface trophoblasts and areola-2 trophoblasts, finding that expressional patterns between these two trophoblast populations were starkly different in both the tissue and sTO ( S4A and S4B Fig ). Using GO-Term analysis, we identified enriched pathways for areola-2 and interface trophoblasts, several of which were enriched in both sTOs and tissue analysis ( S7 – S10 Tables ). Notably, we found that interface trophoblasts are enriched in GO terms such as extracellular matrix, focal adhesions, and angiogenesis ( S7 and S9 Tables ). Conversely, areola-2 clusters were enriched primarily in GO terms surrounding endosome formation ( S8 and S10 Tables ). Expression of various endosome-related genes demonstrated that areola-2 clusters in sTOs and tissue Visium datasets are highly enriched in endosomal markers, whereas interface trophoblasts distinctly lack endosomal gene expression ( S4C and S4D Fig ). Taken together these data demonstrate that sTOs retain at least some key functions of the corresponding tissue populations, highlighting their utility for studying swine placental biology.

We then identified shared markers between sTO clusters and their corresponding spatial transcriptomics clusters to determine the gene expression signatures that defined these mappings. We found that sTO cluster 0 cells expressed 59 markers that were conserved in interface-trophoblasts, which included AKR1B1, HAVCR1, C1QTNF4, and TGM2 ( Fig 6F and 6G and S3 Table ). Similarly, sTO cluster 1 expressed 63 shared markers with areola-2 cluster, including cathepsins A and D (CTSA and CTSD), SLC7A4 and NUPR1 ( Fig 6F and 6G and S4 Table ). sTO cluster 3 had 110 shared markers with areola-2 cluster, with both clusters highly expressing ISG15, FTL, MDK, and B2M ( Fig 6F and 6G and S5 Table ). Conversely, sTO cluster 4 expressed only 20 shared markers to areola-1 cluster, largely including COX1, COX3, ND4, and CYTB ( Fig 6F and 6G and S6 Table ). Using this approach, we were able to assign cellular identities and identify shared gene expression profiles for all trophoblast populations within sTOs, which were distributed at similar ratios across independent lines ( Fig 6H ). Taken together these data indicate that sTO are able to spontaneously differentiate into various populations of trophoblasts found at the porcine maternal–fetal interface.

A) Spatial Dim-Plot showing facetted clusters from two representative maternal–fetal interfaces. As shown Areola-1, Areola-2, and Interface-Trophoblasts were included in this analysis. B) UMAP of sTOs clusters obtained via single-cell RNA-sequencing (n = 3 lines). C) Stacked bar-plot showing relative proportions of various clusters across individual maternal–fetal interfaces. D, E) Spatial feature plot showing prediction scores for each spot in the spatial dataset subsetted by sTO cluster. Color shown indicates probability that the classified sTO cluster is localized to a given position within the spatial dataset. F, G) Dot-plot showing the localization of shared markers between the sTO clusters (F) and maternal–fetal interface clusters (G) Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. H) Stacked Bar Plot showing the relative proportion and identity of sTO cell populations. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980980, GSM8980981, GSM8980982, GSM8980983, GSM8980992, GSM8980993, GSM8980994.

To define trophoblast populations within sTOs, we performed cell type deconvolution by integrating the single-cell transcriptomics dataset from sTOs with the spatially resolved transcriptome of the swine maternal–fetal interface. To do this, we filtered the Visium dataset to only include trophoblast populations, specifically clusters 5, 8, and 9 ( Fig 6A – 6C ). Next, we integrated the single-cell RNA-sequencing data from sTO with the Visium spatial transcriptomics dataset. For cell type prediction, we applied a previously established probabilistic modeling algorithm, which leverages the likelihood of gene expression patterns across datasets to assign cluster identities in a spatial context. This analysis enabled us to predict the identities of sTO clusters with high confidence. We found that sTO cluster 0 very strongly mapped to interface trophoblasts, whereas sTO clusters 1, 3, and 4 map to areola-2 and areola-1 trophoblast clusters ( Figs 6D , 6E , S3A and S3B ). In contrast, cluster 2 did not exhibit strong mapping to any structure ( Figs 6D , 6E , S3A and S3B ). However, this cluster expressed high levels of the proliferation-associated genes TOP2A and MKI67, suggesting that it represented the stem/progenitor cell population, which is likely present at very low abundance in tissue at the stage of gestation when we were collecting the tissues ( Fig 6F ).

A) Chord diagrams showing the number (Left panel) and strength (right panel) of calculated inter-cluster interactions within the porcine maternal–fetal interface. B) Heatmap showing the number (left panel) and strength (right panel) of interactions between clusters within the porcine maternal–fetal interface. Color indicates increased number (left) or strength (right) between the sending cluster and the receiving cluster. C) Heatmaps showing the outgoing (upper left panel) and Incoming (upper right panel) signaling patterns by cluster for Visium spatial transcriptomics. Coloring indicates relative cluster contribution to a pattern with red being high and blue being low contribution level. Lower panels similarly indicate the contribution a specific signaling pathway to the corresponding pattern designation with the lower right panel indicating outgoing signaling patterns and the lower left panel indicating incoming signaling patterns. D, E) Plots showing HSPG signaling within the porcine maternal–fetal interface, with D) being a chord diagram and E) showing the spatial localization of these interactions. F) Spatial feature plot showing the localization of various genes associated with the HSPG pathway. The right panel shows the localization of HSPG2, whereas the left panel shows the localization of DAG1. G, H) Plots showing MK signaling within the porcine maternal–fetal interface, with G) being a chord diagram and H) showing the spatial localization of these interactions. I) Spatial feature plot showing the localization of various genes associated with the MK pathway. The right panel shows the localization of MDK, whereas the left panel shows the localization of SDC2. J, K) Plots showing IL6 signaling within the porcine maternal–fetal interface, with J) being a chord diagram and K) showing the spatial localization of these interactions. L) Spatial feature plot showing the localization of various genes associated with the IL6 pathway. The right panel shows the localization of IL6, whereas the left panel shows the localization of IL6R. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980980, GSM8980981, GSM8980982, and GSM8980983.

Maternal–fetal crosstalk has been widely studied in human pregnancy, where it plays crucial roles in nutrient exchange, immunological tolerance, and successful placentation [ 42 – 44 ]. Recent evidence suggests that similar communication processes take place in the peri-implantation porcine embryo and the maternal endometrium, highlighting the importance of defining these interactions in the swine placenta [ 26 , 45 ]. To delineate communication networks at the swine maternal–fetal interface, we used CellChat to define cell-cell signaling pathways. CellChat is based on a comprehensive dataset of known ligand-receptor pairs, using their expression to infer communication between cells. In spatially resolved data, CellChat also incorporates the physical distance between tissue sites, filtering out interactions that are unlikely to occur due to the spatial separation between these sites. Using this approach we found that the myometrium (Myo, which was merged into a single cluster for this analysis) and the vasculature (Vasc) were responsible for the most robust communication, while the interstitium (Ist), endometrium (Endo), Areola (merged into a single cluster), and IntTroph showed the weakest number of communication networks ( Fig 5A and 5B ). Notably, all maternal clusters were able to crosstalk with the fetal trophoblast populations except the myometrium and vasculature. Using CellChat, we then classified both outgoing and incoming signaling patterns. For outgoing signaling, pattern 1 and pattern 2 were largely derived from the maternal compartment, with pattern 1 including the endometrium and blood, whereas pattern 2 is primarily derived from the myometrium and vasculature. Conversely, pattern 3 was shared among fetal-derived populations, including the Ist, IntTroph, and Areola ( Fig 5C ). We similarly observed that incoming patterns were similarly segregated by fetal and maternal compartments, with pattern 1 exhibiting the strongest signaling for Areola, Ist, and IntTroph, and pattern 2 being composed of Myometrium, Vasculature, and Blood ( Fig 5C ). Each pattern for the outgoing and incoming signaling networks corresponded to distinct communication networks. For example, outgoing pattern 1 was dominated by OSTN, VEGF, and BMP signaling, whereas pattern 3 was dominated by HGF, EGF, and CDH signaling. Similar observations were made for incoming signaling patterns, with pattern 1 corresponding to high levels of EGF, ANGPT, and GRN signaling, and pattern 2 being characterized by high levels of NPR1, KIT, and CXCL signaling ( Fig 5C ). Notably, this compartmentalization of the maternal and fetal signaling suggests that there is extensive maternal–fetal crosstalk occurring in the porcine placenta. Some of these signals were unidirectional, with all outgoing signals deriving from the maternal compartment, such as the HSPG signaling pathway, which is largely derived from the myometrium and vasculature. Interestingly, this HSPG2 signal, which is important for angiogenesis, is capable of being sensed by both the maternal system, and the interface trophoblasts, largely due to the restriction of the receptor DAG1. In contrast, many crosstalk pathways were enriched in various placenta trophoblast subpopulations. For example, the MK pathway was restricted primarily to areola trophoblasts, whereas the IL6 pathway was largely restricted to the interface trophoblasts. In both cases, the outgoing signal, MDK and IL6, were localized to the areola and interface trophoblasts, respectively. While the MDK receptor SDC2 was broadly expressed in both the maternal and fetal compartment, IL6R was largely restricted to the endometrium, with some lower-level expression in all other tissues. Interestingly, in both the MK and IL6 pathways we observed some degree of inter-trophoblast communication, with interface trophoblast and areolar trophoblasts displaying some degree of crosstalk. Together, these findings highlight the extensive, bidirectional maternal–fetal crosstalk occurring at the porcine maternal–fetal interface.

A) Spatial Feature Plots showing the localization of known or novel pan-trophoblast markers, interface-trophoblast markers, areola-2 and areola-1 specific markers. Two representative maternal–fetal interfaces are shown. The reference localization of these structures are shown on the far-left panels. B) Dot Plot showing expanded markers of various trophoblast populations within the maternal–fetal interface. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. C) Dot Plot showing the cluster-specific expression of known and novel trophoblast markers from (B) in sTOs. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980980, GSM8980981, GSM8980982, and GSM8980983.

Using the spatial transcriptomics dataset described above, we analyzed the cellular and molecular landscape of the swine maternal–fetal interface to identify distinct markers associated with specific cell types and regions. This approach enabled us to map the spatial distribution of key genes, including those enriched in select trophoblast populations. To identify trophoblast specific markers, we performed differential expression analysis between clusters containing trophoblasts, (clusters 5, 8, and 9) and all other clusters. This analysis revealed both pan-trophoblast markers, and markers of specific trophoblast populations ( Figs 4A , 4B and S2A – S2E ). Specifically, the genes encoding cathepsin Z (CTSZ) and cathepsin B (CTSB) were highly expressed in both interface trophoblasts as well as areolae trophoblasts, but only weakly expressed elsewhere ( Figs 4A , 4B and S2A ). Similarly, the genes encoding carbonic anhydrase 4 (CA4) and pregnancy-associated glycoprotein L4 (PAG2L4) demarcated interface trophoblasts ( Figs 4A , 4B and S2B ). Areolar trophoblasts universally expressed CTSL ( Figs 3H and S2C ); however, areola-1 trophoblasts specifically expressed coiled coil domain containing 162 (CCDC162) and serine peptidase inhibitor kazal type 4 (SPINK4) ( Figs 4A , 4B and S2E ). Conversely, areola-2 trophoblasts were marked by an upregulation of matrix metalloprotease 7 (MMP7) and anterior gradient protein 2 homolog (AGR2) ( Figs 4A , 4B and S2D ). Notably, all of these highly specific markers were also detected within sTO, with varied expression levels across clusters, suggesting some degree of differentiation across sTO cell populations ( Fig 4C ). Although these markers partially resolved the clusters, their leaky expression across multiple clusters limited our ability to accurately assign cellular identities.

A) Schematic of spatial transcriptomics workflow of gestational day 66 (GD66) maternal–fetal interfaces (n = 4; 2 male and 2 female). Created in BioRender. Mccutcheon, C. (2025) https://BioRender.com/hus6ktj B) UMAP showing 10 distinct populations obtained via spatial transcriptomics. C) Dot-plot of top markers for each cluster obtained via spatial transcriptomics. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. D) H&E staining of individual porcine maternal–fetal interface sections (top). Spatial Dim-Plot showing the histologic localization of UMAP cluster populations (bottom). Each image represents a separate maternal–fetal interface. E) Spatial Dim-Plot showing separation of maternal and fetal compartments. Representative maternal–fetal Interface shown. F) Spatial Dim-Plot showing the histologic localization of UMAP individual cluster populations. Representative maternal–fetal Interface shown (MFI094). G) Stacked bar-plot showing the relative proportion of each cluster across individual maternal–fetal interfaces. H) Spatial Dim-Plot (Left panels) and Spatial Feature plots (Right panels) showing the localization of areola-1 and areola-2 structures (Left panels) and the expression of CTSL (right panels) within representative maternal–fetal interfaces. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980980, GSM8980981, GSM8980982, and GSM8980983.

A) UMAP from single-cell transcriptomics of 3 independent lines of sTOs. B) Stacked bar-plot showing the relative proportions of various cell populations across three sTO lines. C) Violin plot showing the expression of canonical trophoblast markers in sTO single-cell data. D) Violin plot showing the expression of canonical fibroblast markers in sTO single-cell data. E) Dot-plot showing the top enrichment markers for various cell populations in sTOs. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. F) Dot-plot of canonical markers of various porcine trophoblast populations. Color of dots denotes average expression, whereas size represents percentage of cells in a cluster expressing the gene of interest. G) Feature plot showing widespread expression of pan-trophoblast markers in sTOs. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980992, GSM8980993, and GSM8980994.

To determine whether sTOs differentiated into multiple trophoblast lineages, we next performed single-cell RNA-sequencing on three unique lines of sTOs. This analysis revealed five cell clusters ( Fig 2A ), with similar proportions across lines ( Fig 2B ). Canonical trophoblast markers, including KRT7, KRT18, KRT8, CDH1, GATA2, HAND1, TP63, and GATA3, were highly expressed in sTOs ( Fig 2C ) with low expression of the fibroblast markers VIM, PDGFRA, and COL1A1 ( Fig 2D ). Similarly, we observed very low expression of the uterine gland markers FOXA2, MUC16, MUC1, and CD36 ( Fig 2D ). Taken together, these data indicate that sTO are a pure population of trophoblasts that do not contain fibroblasts or glandular epithelial cells. Cluster gene expression enrichment analysis revealed that trophoblast clusters were discernable by cluster-specific gene expression ( Fig 2E and S1 Table ). For example, cluster 0 was enriched with transcripts such as SOX5 and WNT6, cluster 1 had enriched expression of tight junction-associated genes OCLN and CLDN1, cluster 2 was marked by cell proliferation genes like TOP2A, cluster 3 showed high expression of inflammatory genes including CXCL8 and interferon-stimulated genes (ISGs), and cluster 4 exhibited elevated levels of COX enzymes ( Fig 2E ). Previous studies using in situ hybridization, antibody staining, and single-cell transcriptomics have identified markers of both swine interface trophoblasts and areolar trophoblasts [ 20 , 24 , 25 , 40 , 41 ]. While we observed expression of these markers within sTO cell populations, they did not fully segregate the trophoblast populations, ultimately limiting the utility of these markers to define the cellular identity of sTO-derived clusters ( Fig 2F ). Of note, we did observe that known pan-trophoblast markers such as CTSB and KRT18 were uniformly expressed in sTOs suggesting that these various sTO cell populations are distinctive lineages rather than contaminating cell populations ( Fig 2G ). While this analysis identified several trophoblast populations in sTOs, the lack of a comprehensive marker list specific to the swine maternal–fetal interface made it difficult to definitively assign cellular identities to these clusters.

To assess the ability of sTOs to recapitulate tissue-specific gene expression profiles, we performed bulk RNA sequencing on sTOs. Using a previously established protocol for porcine sex determination which relies on the expression of key Y-linked genes, we determined that sTO were derived from both male and female placentae ( Fig 1J ) [ 39 ]. Having confirmed that sTO were indeed fetal derived, we then compared their expressional profile to publicly available expressional datasets of full-term porcine placental tissue, porcine uterine tissue, and PTr2 cells, a commonly used in vitro model of swine trophoblasts. This approach allowed us to compare the global transcriptional landscapes of organoids and their corresponding tissues, identifying key expressional patterns to indicate how closely sTOs mimic the native tissue profile. This analysis revealed that sTOs also express high levels of canonical swine trophoblast markers, at similar levels to porcine placental tissue., whereas PTr2 cells exhibited very low levels of the same markers ( Fig 1K ). Conversely, we found that while uterine tissues expressed high levels of known uterine gland and fibroblast markers, sTO displayed an absence of expression of these markers. While term placentas did exhibit some expression of fibroblast markers, they similarly expressed an absence of uterine gland markers ( Fig 1K ). Together, these data indicate that sTO are composed of fetal-derived swine trophoblasts and display higher expression of these markers than conventional models such as PTr2 cells.

A limitation of hTOs is their inverted polarity when grown in Matrigel “domes”, resulting in an inward-facing STB that does not accurately mimic the cellular polarity of the human placenta in vivo [ 32 ]. To define the cellular orientation of sTOs, we performed immunostaining with the tight junction marker ZO-1, which localizes to the apical surface of swine trophoblasts in vivo [ 38 ]. We found that sTOs expressed ZO-1, which localized to the inner surface of organoids when grown in Matrigel ( Fig 1G ). We showed previously that hTOs cultured in suspension grow with more physiologic polarity, with the STB on the outer surface [ 32 ]. Using a similar approach, we found that sTOs also reversed their polarity when grown in suspension for approximately 48 h and that ZO-1 relocalized to the organoid outer surface ( Fig 1H ). sTOs grown in suspension also displayed a more cystic morphology ( Fig 1I ). This suggests that the polarity of sTOs is easily reversed, allowing for flexibility in experimental design.

Having successfully derived sTO, we then sought to confirm their epithelial origin. To do this, we performed immunofluorescence staining for various known markers of the porcine maternal–fetal interface. Notably, sTOs exhibited strong expression of the proliferation marker MKI67 ( Fig 1E ), with heterogeneous staining intensity across cells. This pattern suggests a transition from a highly proliferative state toward differentiation. Similarly, we found that sTO were positive for the epithelial marker cytokeratin 18 ( Fig 1F ) and the tight junction marker ZO-1 ( Fig 1G ), both of which are known to be expressed by porcine trophoblasts. These data suggest that sTO contain a highly proliferative population of epithelial cells derived from the maternal–fetal interface.

A) Schematic of sTO derivation workflow. Created in BioRender. McCutcheon, C. (2025) https://BioRender.com/ohwh6rx . B) Representative low magnification brightfield image from a single well of sTO. C) High magnification brightfield image of sTO grown in Matrigel. D) Representative Hematoxylin and Eosin staining of a sectioned sTOs. E, F) Immunofluorescence microscopy image of sTO grown in Matrigel stained with E) mKi67 and F) cytokeratin 18. G) Immunofluorescence microscopy image of sTO grown in Matrigel “domes”. Apical marker ZO-1 is in green, DAPI in gray, and Actin in red. H) Immunofluorescence microscopy image of sTOs grown in suspension for 48 hours. Apical marker ZO-1 is in green, DAPI in gray, and Actin in red. I) Representative brightfield image of sTO grown in Matrigel “dome” (right) and suspension (left). J ) Bar-plot showing the average summation of RPKM for nine Y-linked genes across 3 individual sTO lines. Black dots indicate replicates. sTO lines are denoted on the x-axis. K) Heatmap of known pig placentae and uterus markers expressed across Uterus, PTr2 cells, sTOs, and term pig placentae. Individual columns indicate replicates. Clustering of columns is performed using hierarchical clustering, which groups columns with similar expressional patterns. The underlying data for this figure can be found at the Gene Expression Omnibus via accession numbers GSM8980984, GSM8980985, GSM8980986, GSM8980987, GSM8980988, GSM8980989, GSM8980990, GSM8980991, GSM9083242, and GSM9083243.

To derive sTOs, we collected full-term swine placenta tissue and applied a similar two-step enzymatic digestion protocol used previously for the derivation of human TOs (hTOs) [ 33 ] ( Fig 1A ). Isolated stem/progenitor cells were resuspended in Matrigel and allowed to self-organize over a period of 7–10 days until spherical structures, reminiscent of hTOs, were observed via brightfield microscopy ( Fig 1B and 1C ). Organoids were isolated from initial wells by picking out single organoids, dissociated them, and then passaging using a combination of enzymatic and mechanical dissociation. sTO growth was robust and scalable, with an average propagation interval of 5–7 days. Unlike hTOs, which develop with “mini-cavities” at their centers [ 32 ], sTOs exhibited a denser structure with no visible central cavities, which may reflect species-specific differences in tissue architecture and cellular organization ( Fig 1C and 1D ). Similar to hTOs, sTOs could be continuously passaged with no apparent phenotypic changes (e.g., organoid size, morphology, growth kinetics) and can be cryopreserved and revived. We successfully derived sTOs from five unique swine placentas with 100% success across all tissues collected.

Discussion

US agricultural commodities from swine account for approximately $62 billion in added economic value and is heavily reliant on successful animal husbandry [46]. In all eutherian mammals, including swine, successful placentation is critical for fetal growth and survival [7]. Therefore, understanding non-human placental function, development, and defenses is critical for ensuring continued reproductive success. While there is growing interest in studying non-human placentation, the lack of appropriate models remains a major hurdle. We and others have demonstrated that trophoblast organoids derived from human placentae are a critical tool for the study of human placentation due to their ability to be passaged, cryopreserved, and to recapitulate human trophoblast heterogeneity [32,34,47]. Here, we report the establishment of sTOs, which recapitulate key aspects of the swine placenta.

We show that sTOs can be derived from full-term swine placentae, enabling model establishment without the need to sacrifice animals. Like hTOs, sTO can be expanded, cryopreserved, and successfully recovered, making them an ideal in vitro model. Current models, such as PTr2 cells, contain a unique expressional profile to swine placental tissue, and express low levels of several trophoblast markers. We show here that sTOs express high levels of the trophoblast markers KRT18, ELF3, and GATA3, and an expressional signature more consistent with swine placentae. Importantly, sTOs cultured within Matrigel spontaneously form an interiorly localized apical surface that is reversible by culturing in suspension, thereby allowing for customization of this model. Together, our data suggest that sTOs better recapitulate the transcriptional signature of swine placental tissue than currently available in vitro models.

As discussed previously, the swine placenta is distinct from the human placenta at both a structural and cellular level [7]. In humans, placental trophoblasts fall into several distinct subpopulations, including CTBs, EVTs, and the STB, all of which have distinct roles in the formation of the hemochorial placenta [7]. In the epitheliochorial placenta, such as those found in swine, there is no direct contact between maternal blood and fetal trophoblasts, which theoretically could limit the efficiency and molecular breadth of nutrient transfer to the developing fetus [7,8]. To facilitate this exchange, the highly folded interface expresses various nutrient transporters and aquaporins [20,21]. Further, the porcine placenta contains a specialized structure known as an areola, which is a dome like cluster of trophoblasts that sit atop uterine glands [22]. These structures, which have a peak density of approximately 2,500 per placenta [48], collect and transfer maternally-derived histotroph to the fetus as a nutrient source. Indeed, previous studies have indicated that areola surface area is directly correlated with fetal growth, further supporting their critical role in nutrient acquisition [48]. Interestingly, areola structures have been noted in many ungulates, including hippopotamuses [49], dolphins [50], and llamas [51], suggesting that areola are critical for fetal growth across species.

Several studies have indicated that these porcine trophoblast subpopulations have some degree of cellular heterogeneity in vivo. It is known that along placental folds of the interface, expression of various transporters varies. For example, GLUT1 and GLUT8 have opposing expression profiles, with GLUT1 being located at the base of the folds and GLUT8 being localized to the tip of the folds [40]. Similar spatially localized gene expression has been observed for aquaporins [20]. Additionally, trophoblasts localized to the areola have increased expression of CTSL, AQP5, and ITGB3 relative to other trophoblast populations [20,24,41]. Furthermore, how these various populations differentiate and develop is largely unknown, primarily due to the lack of an in vitro model that provides this cellular heterogeneity without the need to consistently access fresh tissue.

We and others have previously demonstrated that in human trophoblast organoids, there is extensive cellular heterogeneity, indicating that trophoblast organoids spontaneously differentiate into the various trophoblast populations of the placenta [32–34,52]. Single-cell RNA-sequencing of sTO revealed similar cellular heterogeneity within this organoid model. While previous studies have identified markers of various trophoblast subpopulations in swine placentae [20,24,40], these markers were insufficient to segregate sTO cell populations. Using Visium spatial transcriptomics of mid-gestation swine maternal–fetal interfaces, we defined the transcriptomic landscape of swine trophoblast populations in situ. By coupling a single-cell RNASeq dataset from sTOs and Visium spatial transcriptomics dataset from swine placentae, we were able to assign identities to sTOs populations using an unbiased probabilistic modeling approach. Simultaneously, we identified a shared gene expression profile between sTO cell populations and their corresponding placental cell populations. Using this approach, we demonstrate that sTOs contain a proliferative stem cell population, an interface trophoblast population, and three areola trophoblast populations. Critically, this approach allowed for robust identification of cell populations in an unbiased manner.

Using trajectory analysis, we defined how sTOs gain their heterogeneity. Using Slingshot, we demonstrate that proliferative stem cells differentiate into interface trophoblasts, which can then differentiate into areola-2 trophoblasts. This analysis uncovered numerous expression changes throughout the differentiation process, including novel targets like ITM2C and NUPR1, which could be investigated in future mechanistic studies to enhance our understanding of swine placental development. Having defined the expressional changes along this trajectory, we then confirmed their localization in tissue finding high concordance between sTOs and tissue-level expression, suggesting a shared trajectory between these two systems. By comparing global gene expression changes in sTO-derived interface trophoblasts and areola-2 trophoblasts, we go on to show that these two populations display starkly different functional niches, with interface trophoblasts being primed to form focal adhesions and promote angiogenesis, whereas areola-2 trophoblasts were enriched in programs for the formation of endosomes. It is noted within the literature that areola trophoblasts contain a high number of intracellular vesicles, presumably to facilitate transport of histotroph across this epithelial barrier [22]. As expected, when we examined the expression of key endosome markers within the tissue and sTO, we observed similar patterns of expression, with endosome markers expressed at a higher level in areola-2 cells compared to interface trophoblasts. Together, these data demonstrate that sTOs recapitulate the developmental and functional differences in trophoblast populations observed in the porcine placenta, highlighting their utility in studying porcine placental development.

While combining sTO and spatial datasets allowed for robust cell cluster identification, parallel analyses enabled us to further explore the functional similarities between sTOs and placental tissue. Using CellChat, we first identified dozens of signaling pathways occurring at the porcine maternal–fetal interface. Interestingly, we found several instances of inter-trophoblast signaling within the porcine placenta. By applying CellChat to the shared trophoblast populations found within our sTO and placental tissue datasets, we revealed several key signaling pathways are conserved between these systems. For example, we that the angiopoietin-like proteins (ANGPTL), Agrin (AGRN), and Cadherin (CDH) pathways are highly conserved within sTO and tissue trophoblasts. These pathways have been shown to be active in the placenta and play important roles in regulating epithelial to mesenchymal transitions, basement membrane function, and nutrient metabolism [53–56]. Importantly, sTO recapitulated autocrine and paracrine signaling patterns observed in placental tissue, highlighting their utility as a physiologic model. Importantly, these analyses were based on known human receptor-ligand interactions and rely on shared annotations between the human and porcine genome. It remains possible that some receptor–ligand pairs in the porcine genome are not annotated, which would preclude them from being included as an interaction in our results. Therefore, this represents a conservative estimate of cell-cell communication in the porcine placenta. We believe that these data provide a robust framework for hypothesis generation for understanding signaling within the porcine placenta.

Previous studies have used single-cell RNA-sequencing to characterize the porcine uterus and embryos [25,26,45,57]. However, these studies rely on existing markers or GO-term-based functions to assign cellular identity, lack histological structural context, do not provide spatial segregation of gene expression signatures, and were not performed at a gestational stage where critical structures, such as the areola, are present. Our data robustly define the expression profiles of various histological structures at the swine maternal–fetal interface. By mapping gene expressions directly to corresponding histological structures, we confirmed cellular identities without relying on conventional markers, making this assignment unbiased. Moreover, using mid-gestation swine maternal–fetal interfaces, we were able to characterize the expression profiles of key structures, including the areola and the interface, which have previously eluded detailed analyses. This dataset provides new insights into placental biology that were not achievable with earlier approaches.

Demonstrating the utility of this spatially resolved dataset, we identified two transcriptionally and structurally distinct areola populations within the porcine placenta: Areola-2, which appears continuous with the interface trophoblasts, and Areola-1, which is localized primarily to the placental interstitium without evidence of direct contact with the interface. This finding emphasizes the power of spatial transcriptomics in revealing previously unrecognized cellular diversity within structurally complex tissues such as the placenta. While both clusters expressed high levels of the known areola marker CTSL [24], these clusters had divergent gene expression profiles with Areola-2 expressing high levels of MDK, CTSD, PAG2L2, and Areola-1 expressing high levels of CCDC162, SPINK4, and MMP25. While it remains unclear why these two areola populations exist, previous results [24] have shown that histologically similar CTSL positive structures exist within the porcine placenta. It is possible that these represent areola at different stages of development, with Areola-1 representing a more mature population due to its higher expression of CTSL and its increased size. Together, these findings highlight the power and utility of the spatial transcriptomic dataset in advancing our understanding of the fundamental biology of the porcine placenta.

Our work offers a comprehensive characterization of spatial gene expression in the porcine placenta and successfully establishes an organoid model that captures its cellular heterogeneity. However, limitations include the resolution of Visium spatial transcriptomics, which does not allow for single-cell level definition of histological populations. For example, the identified Visium clusters may represent a mixture of gene expression profiles but nonetheless captures the dominant signals of key structures and cell populations. For instance, although we observed some overlap in marker expression between uterine epithelium and trophoblast populations, these clusters still separate along the interface. Moreover, the populations are strongly represented by known markers, indicating that despite resolution constraints, this technique effectively segregates distinct populations and provides robust biological insights. Another limitation of this study is that we focused on a single mid-gestation time point. While this provides valuable insights into the biology of the porcine placenta during this stage, it limits our ability to assess dynamic changes in cellular composition and signaling across different stages of gestation. While previous studies have highlighted that human trophoblast organoids represent first trimester tissue [33,35], our bulk RNA-seq data demonstrated that sTOs have similar expression to porcine placentae regardless of gestation. It remains possible that early or later gestation placentae would have mapped with higher probabilities to sTO populations enabling more robust identification. Future studies aimed at capturing multiple gestational timepoints in the porcine placenta are therefore warranted. While the data presented here demonstrate the utility and strength of spatial transcriptomics, they represent only a portion of the insights generated in this study. The thousands of spatially resolved genes and pathways identified provide a valuable resource for advancing porcine reproductive biology. To facilitate broader access, we have developed a user-friendly web-based app, enabling researchers without advanced computational resources to explore and analyze the data (see Data availability statement for further details)

We believe this study serves as a foundational study for studying the porcine placenta, via the presentation of a physiologic model and a spatially resolved roadmap of the maternal–fetal interface. These critical findings allow for future studies to be proposed. For example, future studies aimed at studying the function of the signaling pathways identified here may discover new avenues to reduce reproductive failure, including failure caused by vertical transmission of pathogens or defective placentation. Additionally, these data may inform studies aimed at identifying the triggers of areola development, which may now be possible in vitro. We hope that these data will spark new interest in studying porcine reproductive biology.

Taken together, our study describes the derivation of an organoid model that recapitulates the cellular heterogeneity and functionality of the swine placenta. Additionally, we provide a spatially resolved expression landscape of the swine maternal–fetal interface, creating a valuable resource for the study of non-human placentation. The ability to map gene expression directly to histological structures offers unprecedented insights into the cellular dynamics of the porcine placenta, setting a new benchmark for comparative placental biology research. These findings have the potential to significantly advance our understanding of placental biology across eutherians, offering a critical framework for studying key aspects of maternal–fetal communication and placental development. Moreover, the organoid model and spatial transcriptomic data will serve as foundational tools for exploring mechanisms of placental function and dysfunction, with broader implications for reproductive biology and veterinary medicine.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3003302

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/