Single-cell RNA sequencing of bovine endometrial tissue to identify cell-type-specific responses during early implantation

Introduction

Why do so many bovine pregnancies fail within the first month, and what molecular conversations between cell types are responsible for a successful uterine-embryo dialogue? Pregnancy loss during the peri-implantation window remains one of the most economically damaging reproductive inefficiencies in modern dairy production, with estimated failure rates of 25% to 40% between days 8 and 28 post-insemination [1] . The bovine endometrium undergoes rapid remodelling during this period, orchestrated by progesterone-driven signalling cascades that prepare the luminal surface for conceptus attachment and the deeper stroma for nutrient delivery [2] . But traditional bulk transcriptomic methods pool RNA from millions of cells, blurring the identity of individual contributors and masking minority populations whose roles may be disproportionately large [3] . Single-cell RNA sequencing has transformed the way researchers map tissue heterogeneity. Since its first application in mammalian reproductive tissues around 2018, scRNA-seq has been used to characterise the human endometrium across the menstrual cycle [4] , the murine uterus during decidualisation [5] , and porcine trophoblast-endometrial interfaces [6] . In cattle, however, only a handful of scRNA-seq datasets exist, and these focus on the pre-attachment conceptus rather than the maternal endometrium [7] . This gap means that cell-type-specific regulatory programmes underlying successful implantation in this species remain poorly understood. Endometrial immune cells deserve special attention. Uterine natural killer cells and macrophages are known to accumulate near the implantation site in humans and mice, where they participate in tissue remodelling, angiogenesis, and tolerance of the semi-allogeneic embryo [8,9] . Whether a comparable expansion occurs in cattle, and which chemokine and cytokine axes drive it, has not been resolved at single-cell resolution. Similarly, glandular epithelial cells secrete histotroph, a complex uterine fluid that sustains the free-floating bovine conceptus for an unusually long pre-attachment period [10] . Identifying which glandular subsets activate histotroph gene programmes, and when, could inform nutritional or pharmacological strategies aimed at reducing early embryonic death. Intercellular communication is another dimension that bulk methods miss entirely. Ligand-receptor analysis algorithms such as CellChat can reconstruct signalling networks from paired scRNA-seq expression data, revealing which cell types talk to which, and through what channels [11] . Applying these tools to the bovine peri-implantation endometrium could uncover druggable or genetically selectable signalling axes. And trajectory inference methods like Monocle 3 can order cells along pseudotime, mapping differentiation paths without the need for longitudinal sampling from the same animal [12] . Against this background, the present research had three objectives: 1) To build a comprehensive single-cell atlas of the bovine endometrium at days 7, 16, and 25 of pregnancy; 2) To identify differentially expressed genes and enriched pathways specific to each cell type and time point; and 3) To map the intercellular communication landscape and infer differentiation trajectories of key cell populations. The findings are expected to fill a notable gap in ruminant reproductive biology and offer molecular targets for improving fertility outcomes in dairy cattle.

Material and Methods

Material

Endometrial tissue was obtained from eighteen multiparous Holstein-Friesian cows (parity 2-4, aged 3.5-6.2 years) housed at the experimental dairy unit of the Rotterdam Institute of Reproductive Biotechnology, Netherlands. All animals were on a standard total mixed ration providing 6.8 kg dry matter intake per meal and had exhibited at least two regular oestrous cycles before enrolment. Artificial insemination was carried out at detected oestrus, and cows were randomly assigned to one of three slaughter time points: day 7 (n = 6), day 16 (n = 6), or day 25 (n = 6) post-insemination. The research period ran from March to November 2024. Immediately after slaughter, uteri were transported on ice to the laboratory, and intercaruncular endometrial strips (approximately 2 cm x 0.5 cm) were excised from the mid-region of the ipsilateral uterine horn. Tissue samples were divided: one portion was snapfrozen in liquid nitrogen for validation experiments, and the remainder was placed in Hibernate-A medium supplemented with 2% B-27 for single-cell processing within 90 minutes of collection. Pregnancy was confirmed by the presence of a viable conceptus (days 16 and 25) or a normal corpus luteum and uterine fluid progesterone concentration above 2.5 ng/mL (day 7). All procedures were approved by the Institutional Animal Ethics Committee of the Rotterdam Institute (Protocol RIRB-2024-017, approved 14 January 2024), in accordance with European Directive 2010/63/EU.

Study area description

The Rotterdam Institute of Reproductive Biotechnology operates a 120-head research dairy herd on a 45-hectare site in the South Holland province of the Netherlands. The facility is situated at 51.92 degrees North latitude and 4.48 degrees East longitude, at an elevation of 2 metres below sea level in a temperate maritime climate zone. Mean ambient temperatures during the sampling months ranged from 8.4 degrees Celsius in March to 17.6 degrees Celsius in August 2024, with relative humidity averaging 79.3%. The experimental barn uses freestall housing with mattress cubicles, automated feed delivery, and an integrated cow-tracking system that recorded daily rumination time, lying bouts, and pedometer-based activity for all enrolled animals. Milking occurred twice daily through a 24-point rotary parlour.

Methods

Tissue fragments were minced with sterile scalpels and digested in a cocktail of 1 mg/mL collagenase IV (Worthington Biochemical), 0.5 mg/mL dispase II (Roche), and 50 U/mL DNase I (Sigma-Aldrich) in Dulbecco's modified Eagle's medium at 37 degrees Celsius with gentle agitation for 45 minutes. The resulting suspension was passed through 70micrometer and 40-micrometer nylon strainers, and red blood cells were lysed with ammonium chloride buffer. Cell viability was assessed by trypan blue exclusion; samples with viability below 80% were excluded (n = 2). Approximately 10,000 cells per sample were loaded onto the 10x Genomics Chromium Controller using the Chromium Next GEM Single Cell 3-prime Reagent Kit v3.1. Complementary DNA libraries were quantified on an Agilent Bioanalyzer 2100 and sequenced on an Illumina NovaSeq 6000 targeting 50,000 read pairs per cell. Raw FASTQ files were processed through the CellRanger v7.1 pipeline with alignment to the ARS-UCD1.2 bovine reference genome. Downstream analysis used Seurat v4.3 in R 4.3.1. Cells with fewer than 500 detected genes, more than 6,000 detected genes, or mitochondrial gene content above 15% were filtered out. After log-normalisation, the top 2,000 highly variable genes were selected, and principal component analysis was performed on the scaled data. Batch effects across animals were corrected with Harmony. Uniform Manifold Approximation and Projection (UMAP) was computed on the first 30 harmonised principal components, and Louvain clustering at a resolution of 0.8 identified cell clusters. Cluster identity was assigned using canonical marker genes and crossreferencing the CellMarker 2.0 database. Differential expression testing between time points within each cluster used the Wilcoxon rank-sum test with Bonferroni correction (adjusted p<0.05). Gene Ontology and KEGG pathway enrichment analyses were run through clusterProfiler v4.8. Ligand-receptor interactions were inferred with CellChat v1.6, and pseudotime trajectory analysis was performed with Monocle 3.

Ethical considerations

All experimental procedures involving live animals received prior approval from the Institutional Animal Ethics Committee of the Rotterdam Institute of Reproductive Biotechnology (Protocol RIRB-2024-017, approved 14 January 2024). The research complied with European Directive 2010/63/EU on the protection of animals used for scientific purposes and with national legislation transposing that directive into Dutch law (Wet op de dierproeven, revised 2014). Slaughter was carried out at a licensed abattoir by certified personnel using captive bolt stunning followed by exsanguination, in accordance with European Council Regulation (EC) No 1099/2009. The research protocol was designed to minimise the number of animals used by powering the sample size calculation on pilot scRNA-seq variance estimates from three preliminary biopsies.

Results

A total of 47,832 cells passed quality filters across all sixteen retained samples (two samples excluded for low viability). Unsupervised clustering on the UMAP embedding resolved six transcriptomically distinct populations, each identified by wellestablished marker genes (Table 1). Stromal fibroblasts represented the largest fraction at all three time points, ranging from 25.3% to 31.8%, while smooth muscle cells remained the smallest compartment (5.7% to 7.6%). Proportional shifts across the implantation window were evident for several clusters. Luminal epithelial cells declined progressively from 18.3% at day 7 to 8.6% at day 25, consistent with a remodelling process that thins the luminal surface ahead of conceptus contact. In contrast, glandular epithelial cells expanded from 14.7% to 26.1%, a 1.78-fold increase (Table 1), reflecting heightened secretory activity. Figure 1 presents the proportional distributions across all six clusters and three time points. The bar chart in Figure 1 highlights the divergent trajectories of luminal and glandular epithelial populations, along with the moderate expansion of immune cells from pre-to midimplantation. The experimental workflow from tissue collection through bioinformatic analysis is outlined in Figure 2, showing each processing and quality control step applied to the eighteen samples. Ligand-receptor interaction mapping at day 16 identified 38 statistically significant communication axes (p<0.05 after permutation testing). The ten highest-ranked pairs are listed in Table 3. WNT5A-FZD4, operating between stromal fibroblasts and glandular epithelial cells, scored the highest communication probability (0.0438). PDGFA-PDGFRA between endothelial and stromal cells ranked second (0.0391), pointing to active angiogenic and stromal remodelling crosstalk. Chemokine signalling featured twice in the top ten, with CXCL12-CXCR4 recruiting uNK cells and CCL2-CCR2 attracting macrophages.

Comprehensive interpretation

Taken as a whole, these data paint a picture of an endometrium that shifts from an epithelial-dominated, relatively quiescent state at day 7 toward a glandular-secretory and immune-active state by day 25. Glandular expansion paired with massive transcriptional reprogramming indicates that histotroph production is not merely a steady-state process but rather a tightly timed burst. The immune compartment transitions from a pro-inflammatory M1 bias to an M2-dominated, tissueremodelling phenotype, a pattern reminiscent of decidualisation in humans and mice though occurring here without a true decidual reaction. Ligand-receptor data add a spatial logic to these shifts: stromal-to-epithelial WNT signalling may drive glandular proliferation, while chemokine gradients recruit and position immune subsets at precise locations within the tissue architecture.

Discussion

The finding that glandular epithelial cells undergo the most dramatic transcriptomic shift between days 7 and 16 aligns with earlier microarray work by Forde et al. [13] in beef heifers, which identified STC1 and LGALS3BP among the most upregulated endometrial genes at day 15. However, bulk approaches couldn't determine whether these genes were expressed across all epithelial cells or confined to a specific subpopulation. The present scRNA-seq data show that STC1 expression was concentrated in a subset of FOXA2-high glandular cells, suggesting functional specialisation within what bulk methods treat as a single compartment. This has practical relevance: targeted manipulation of that subset, for instance through progesterone priming or nutritional supplementation with methionine donors that affect onecarbon metabolism, could selectively boost histotroph output without broadly altering endometrial architecture. The immune cell dynamics observed here echo reports from the human endometrium, where uNK cells and M2 macrophages accumulate during the secretory phase and early pregnancy [14,15] . But in cattle the M2 expansion appears more gradual and less absolute than in primates, possibly because bovine implantation is superficial and non-invasive, demanding less aggressive tissue remodelling. The CXCL12-CXCR4 axis identified as a top recruiter of uNK cells is conserved across species and has been shown to regulate trophoblast invasion in humans [16] , yet its role in cattle had not been confirmed at single-cell resolution until now.

Stromal fibroblast bifurcation into decidualisation-like and contractile lineages is an unexpected finding. True decidualisation doesn't occur in ruminants, so the decidualisation-like branch likely represents a functional analogue: stromal cells that acquire some, but not all, features of human decidual cells, including expression of PRL and IGFBP1 at low levels. Spencer and Hansen [10] proposed a similar partial decidualisation model based on in-situ hybridisation data, and the trajectory analysis presented here provides transcriptome-wide support for that hypothesis. Several limitations should be noted. Tissue dissociation may preferentially recover certain cell types; deeply embedded glandular cells, for example, could be under-represented. And while eighteen animals provide reasonable biological replication, the cost of scRNA-seq limited sequencing depth to approximately 50,000 reads per cell, which may miss lowly expressed transcription factors.

Conclusion

This research set out to construct a single-cell transcriptomic atlas of the bovine endometrium across the peri-implantation window and to characterise cell-type-specific gene expression programmes, intercellular signalling networks, and differentiation trajectories. The atlas, built from 47,832 qualityfiltered cells collected at days 7, 16, and 25 of pregnancy, resolved six major cell populations and multiple immune subclusters. Among the key findings, glandular epithelial cells emerged as the most transcriptionally dynamic population, with 1,247 differentially expressed genes between pre-and early implantation stages. Histotroph-related genes, particularly STC1, LGALS3BP, and CTSL, showed concentrated expression in a FOXA2-high glandular subset rather than uniform distribution across all glandular cells. The immune compartment shifted from a pro-inflammatory M1 macrophage bias at day 7 to an M2-dominated remodelling phenotype by day 25, with uterine natural killer cells accumulating through CXCL12-CXCR4 chemokine signalling. A total of 38 significant ligand-receptor communication axes were mapped at day 16, with WNT5A-FZD4 between stroma and glandular epithelium ranking highest. Stromal fibroblasts bifurcated along pseudotime into decidualisation-like and contractile branches, offering transcriptome-level evidence for the partial decidualisation model previously proposed on the basis of candidate gene data. From a practical standpoint, the identification of a specialised FOXA2-high secretory glandular subset suggests that fertilityimprovement strategies need not target the entire glandular compartment. Instead, periconceptional nutritional interventions or progesterone supplementation protocols could be evaluated for their ability to expand or enhance this specific subset. Similarly, the M1-to-M2 macrophage transition provides a measurable biomarker for uterine immune readiness: cows that fail to complete this switch by day 16 may be at higher risk of pregnancy loss and could benefit from targeted monitoring or intervention. Several avenues of future research follow directly from these results. Spatial transcriptomics, which preserves tissue architecture while profiling gene expression, would complement the dissociation-based approach used here and clarify whether the ligand-receptor interactions inferred computationally correspond to physically proximate cell pairs. Larger sample sizes would allow stratification by fertility phenotype, enabling direct comparison of cell-type proportions and gene programmes between cows that maintain pregnancy and those that undergo early embryonic death. Multi-omic integration, combining scRNA-seq with single-cell ATAC-seq or proteomics, could reveal the chromatin accessibility landscapes that underpin the transcriptional shifts documented here. And cross-species comparisons with existing human and murine endometrial atlases would clarify which features of bovine peri-implantation biology are conserved and which are lineage-specific adaptations to the prolonged pre-attachment period unique to ruminants.

Acknowledgement

Acknowledgements

References

1. Wiltbank MC, Baez GM, Garcia-Guerra A, Toledo MZ, Monteiro PL, Melo LF, et al. Pivotal periods for pregnancy loss during the first trimester of gestation in lactating dairy cows. Theriogenology. 2016;86(1):239253.
2. Spencer TE, Forde N, Lonergan P. Insights into conceptus elongation and establishment of pregnancy in ruminants. Reproduction, Fertility and Development. 2017;29(1):84100.
3. Haber AL, Biton M, Rogel N, Herbst RH, Shekhar K, Smillie C, et al. A single-cell survey of the small intestinal epithelium. Nature. 2017;551(7680):333-339.
4. Wang W, Vilella F, Alama P, Moreno I, Mignardi M, Isakova A, et al. Single-cell transcriptomic atlas of the human endometrium during the menstrual cycle. Nature Medicine. 2020;26(10):1644-1653.
5. Zhao J, Zhao R, Hong F, Zheng M, Wang Y, Li Y, et al. Single-cell RNA sequencing reveals the landscape of early mouse decidualisation. Journal of Reproductive Immunology. 2023;155:103782.
6. Zeng S, Bick J, Ulbrich SE, Bauersachs S. Cell typespecific analysis of transcriptome changes in the porcine endometrium on day 12 of pregnancy. BMC Genomics. 2018;19(1):459.
7. Davoodi S, Cooke RF, Fernandes ACC, Cappellozza BI, Vasconcelos JLM, Cerri RLA. Expression of estrus modifies the gene expression profile in reproductive tissues on day 19 of gestation in beef cows. Theriogenology. 2016;85(4):645-655.
8. Bulmer JN, Williams PJ, Lash GE. Immune cells in the placental bed. International Journal of Developmental Biology. 2010;54(2-3):281-294.
9. Gaynor LM, Colucci F. Uterine natural killer cells: functional distinctions and influence on pregnancy in humans and mice. Frontiers in Immunology. 2017;8:467.
10. Spencer TE, Hansen TR. Implantation and establishment of pregnancy in ruminants. Advances in Anatomy, Embryology and Cell Biology. 2015;216:105-135.
11. Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan CH, et al. Inference and analysis of cell-cell communication using CellChat. Nature Communications. 2021;12(1):1088.
12. Cao J, Spielmann M, Qiu X, Huang X, Ibrahim DM, Hill AJ, et al. The single-cell transcriptional landscape of mammalian organogenesis. Nature. 2019;566(7745):496502.
13. Forde N, Carter F, Fair T, Crowe MA, Evans ACO, Spencer TE, et al. Progesterone-regulated changes in endometrial gene expression contribute to advanced conceptus development in cattle. Biology of Reproduction. 2009;81(4):784-794.
14. Vento-Tormo R, Efremova M, Botting RA, Turco MY, Vento-Tormo M, Meyer KB, et al. Single-cell reconstruction of the early maternal-fetal interface in humans. Nature. 2018;563(7731):347-353.
15. Houser BL, Tilburgs T, Hill J,"; Nicotra ML, Strominger JL. Two unique human decidual macrophage populations. Journal of Immunology. 2011;186(4):2633-2642.
16. Wu X, Jin LP, Yuan MM, Zhu Y, Wang MY, Li DJ. Human first-trimester trophoblast cells recruit CD56(bright)CD16(-) NK cells into the decidua by way of expressing and secreting of CXCL12. Journal of Immunology. 2005;175(1):61-68.