PS12-10. Bulk RNA-Seq Deconvolution Identifies Peripheral Blood Cell Signatures of Fertility in Beef Heifers.

Abstract Reproductive inefficiency remains a challenge in the cattle industry. Molecular-based approaches are being explored to understand why heifers with similar phenotypes and reproductive management fail to become pregnant at the end of the breeding season. Multi-omics analyses provide insight into the biological mechanisms underlying fertility, but high-resolution approaches, such as single-cell RNA sequencing, are often cost-prohibitive for studies requiring large sample sizes. Deconvolution of bulk RNA sequencing data offers a cost-effective alternative by estimating underlying cell-type composition and associated gene expression profiles. Herein, we characterized the cellular heterogeneity in peripheral white blood cells (pWBCs) and granulosa cells from beef heifers with differing fertility outcomes. To this end, 123 Angus–Simmental crossbred heifers were subjected to a 7-day CO-Synch + CIDR estrus synchronization protocol followed by fixed-time artificial insemination (AI). Heifers were then exposed to a fertile bull for 60 days beginning 14 days post-insemination. Pregnancy status was assessed 120 days after AI, and heifers were classified as fertile (pregnant) or subfertile (not pregnant). Following the classification, all pregnancies in the fertile group were terminated. pWBC and granulosa samples were collected at the time of harvest (fertile, n = 7; subfertile, n = 5). Total RNA was extracted, and libraries were sequenced on the NovaSeq platform. After quality control, reads were aligned to the ARS-UCD 2.0.114 Bos taurus reference genome using STAR (v2.7.5). Bulk RNA-seq data from both pWBCs and granulosa cells were subsequently analyzed using CAM3 to infer cell population structure and identify transcriptional signatures associated with fertility status. The raw counts were deconvoluted under multiple model configurations to estimate between 2 to 8 putative cell populations (K). After dimensionality reduction and filtering, genes were clustered with K = 3 for both the datasets representing the putative cell-populations underlying the difference in fertility status. In pWBCs, all subfertile heifers clustered within one group (Cluster 1; 2,154 genes), while most fertile heifers (5 out of 7) clustered separately (Cluster 3; 1,883 genes). Genes enriched in Cluster 1 were associated with folate transport and metabolism, steroid hormone biosynthesis, hormone signaling, and cAMP signaling pathways. In contrast, genes from Cluster 3 were over-represented for adrenergic signaling in cardiomyocytes, cAMP signaling, and neuroactive ligand–receptor interaction pathways. Granulosa cell samples did not exhibit a clear clustering pattern associated with fertility status. These findings demonstrate that deconvolution can reveal cell population - level transcriptional differences associated with fertility; however, deconvolution of bulk RNA-seq provides only inferred cell populations, limiting resolution of true cell identities and distinguishing changes in cell proportions. Thus, scRNA-seq reference is needed to directly resolve cellular heterogeneity and validate fertility-associated transcriptional signatures.