Transcriptomic and phenomic data integration to identify markers of subclinical mastitis in Holstein cattle
Résumé
In this study, we integrated the RNA-Seq profiling of milk somatic cells upon healthy and naturally affected by subclinical intramam- mary infection (sIMI) cows with host phenotypic data, in order to gain more insights into the complexity of the molecular mecha- nisms driving the animals’ response to infection, and identify hub variables for early detection and prediction. Bacteriological screen- ing was performed on 188 Holstein cows reared in one herd in order to identify the experimental groups, which were defined as follows: (i) animals with no history of mastitis and negative at the bacteriological examination (NEG; n = 9) and, (ii) animals positive at the bacteriological examination for different pathogens (sIMI; n = 22). Host-related phenotypic traits included milk yield and com- position traits (i.e. protein, casein, fat, urea), udder health traits (lactose, somatic cell count- SCC, pH and conductivity) and immune cells population obtained with flow cytometry. The Data Integration Analysis for Biomarker discovery using Latent Components (DIABLO) approach was used for identifying highly correlated transcriptomic and phenotypic data capable of discriminating between NEG and sIMI animals. Then, interactions among vari- ables with r ≥ 0.90 were used for the network construction, which was explored using the cytoHubba plug-in within Cytoscape for the hub variables identification. Genes belonging to the selected hub variables were then submitted to a receiver operating characteristic (ROC) in order to test their predictive ability for the discrimination of NEG/sIMI animals. This integration analysis confirmed a strong correlation between the transcriptome and the leucocyte popula- tions (r2 = 0.72), and udder health traits (r2 = 0.64), strengthening the need to systematically include them for the screening of sIMI at herd level. The good predictive performances (Sensitivity >0.89, Specificity >0.81, Accuracy >0.87 and Precision >0.69) obtained with the ROC analysis on the selected hub genes might suggest their inclusion as putative biomarker of sIMI. Among these genes, the major histocompatibility complex (MHC), class II, DO alpha (BoLA-DOA) and class II transactivator (CIITA) could have a key role as regulators of the animals’ response to sIMI and therefore should be further valuated in future studies.