Transcriptomic analyses of the oviduct at different stages of early pregnancy. A) Histological analysis of different oviductal regions (ampulla, isthmus, and near the uterotubal junction (UTJ)) in mice at different stages of pregnancy (0.5, 1.5, and 2.5 dpc) and pseudopregnancy (0.5, 1.5, and 2.5 dpp) using H&E staining (scale bars = 132 μm, n=3 mice/timepoint/region). B) Principal Component Analysis (PCA) of top 2500 DEGs identified from bulk-RNA seq of the infundibulum+ampulla (IA) and isthmus+UTJ (IU) regions of the oviduct collected at 0.5, 1.5, 2.5, and 3.5 dpc. Heatmap plots of unsupervised hierarchical clustering of top 2500 DEGs identified from bulk-RNA seq in the oviduct during pregnancy (0.5, 1.5, 2.5, and 3.5 dpc) of C. IA and D. IU regions. E-F) scRNA-seq analysis of the oviduct from superovulated (SO) estrus, SO 0.5 dpc, SO 1.5 dpc, and SO 2.5 dpc. Uniform Manifold Approximation and Projection (UMAP) of E) cell clusters identified from the oviduct F) at different regions (IA and IU) and G) at different timepoints (n=3-4 mice/timepoint/region). H and I. GOBPs dot plots of scRNA-seq analysis when compared between upregulated DEGs from H. secretory epithelial cells and I. ciliated epithelial cells at SO 0.5 dpc compared to SO Estrus from both IA and IU regions.

Analyses of protein abundance in the oviduct luminal fluid at different stages of pregnancy compared to Estrus. A) Manhattan hierarchical complete clustering dendrogram of natural (Estrus, 0.5 dpc, 1.5 dpc, and 2.5 dpc) and superovulated (SO Estrus, SO 0.5 dpc, SO 1.5 dpc, and SO 2.5 dpc) datasets (n= pooled of 3 biological samples/timepoint). B) PCA plot of all datasets generated utilizing Perseus software after integration of the Gaussian transformation. C) Correlation-based hierarchal clustering of all protein abundance. D-F) Volcano plots of significantly different protein abundances when compared between D) Natural fertilization, E) SO fertilization, and F) Natural fertilization vs. SO fertilization. Numbers of significant proteins were listed above the volcano plots. G and H) Gaussian transformed Perseus two-tail t-tests of differentially abundant proteins in oviductal fluid at different stages during G) Natural fertilization and H) SO fertilization. Differentially abundant proteins shared between Estrus and 0.5 dpc (100) or SO Estrus and SO 0.5 dpc (105) were underlined. I and J) Enricher Reactome pathway analysis of differentially abundant proteins shared at I) 0.5 and H) SO 0.5 dpc.

In vivo validation of RNA and proteins identified from bulk RNA- and scRNA-seq analysis. A and B) Expression of Tlr2, Ly6g, and Ptprc in the isthmus and UTJ regions at 0.5 dpc, 1.5 dpc, 0.5 dpp, and 1.5 dpp. 200× magnification. B) Quantification of fluorescent signal from images in A using FIJI software. Graph represent mean±SEM, n=3 mice/timepoint/region. C) Immunofluorescent staining of NFκB in the isthmus regions of the oviducts at 0.5 dpc, 1.5 dpc, 0.5 dpp, and 1.5 dpp. 200× magnification. D) Quantification of fluorescent signal from images in C using FIJI software. Violin plots represent all measurements, n=3 mice/timepoint/region, ****p<0.001 compared to 0.5 dpc, unpaired t-test. E) Immunoblotting of phosphorylated p38 and total p38 in the whole oviduct collected at 0.5 dpc, 0.5 dpp, 1.5 dpc, and 1.5 dpp. F) Violin plots of the quantification of band intensity represented as phosphor-p38/total p38 ratio (n=3 mice/timepoint/region). *p<0.05 compared to 0.5 dpc, unpaired t-test. G) IL1β ELISA of protein from the whole oviduct at 0.5 dpc, 1.5 dpc, 0.5 dpp, and 1.5 dpp (n=3 mice/timepoint/region).

Overall architecture of the transformer-based model to predict proteomic abundance from bulk RNA-seq data of natural fertilization of oviduct. A) Preprocessing steps using bulk RNA-seq count per million (cpm) normalization to calculate expression values. The transformer model is equipped with a single-layer transformer encoder featuring a single-head (1-Head) self-attention mechanism to predict the abundancy of proteins (abundant or not) from the input RNA-seq data. “Head” refers to blocks, modules, or connections that perform specific tasks in neural networks. A specific threshold of 0.6/0.8 was defined to label proteins as high abundance or low abundance. The Multi-Layer Perceptron (MLP) Head refers to the output layer, which is designed to perform a classification task. In this model, The MLP layer uses a multi-layer perceptron or linear layer as the backbone to divide high abundance and low abundance based on the importance or attention weights given by the previous transformer layer. B) The visual representation of a method to extract the top 25 TFs for differential significant proteins. DGE; Differential gene expression, DPA; Differential protein abundance; TF, Transcription factors