Combining supervised machine learning with statistics reveals differential gene expression patterns related to energy metabolism in the jejuna of chickens divergently selected for antibody response to sheep red blood cells

Abstract

Since the 1970s, 2 lines of White Leghorn chickens, HAS and LAS, have been continuously divergently selected for 5-day postinjection antibody titer to injection with sheep red blood cells (SRBC). Antibody response is a complex genetic trait and characterizing differences in gene expression could facilitate better understanding of physiological changes due to selection and antigen exposure. At 41 d of age, randomly selected HAS and LAS chickens, which had been coraised from hatch, were either injected with SRBC (HASI and LASI) or kept as the noninjected cohort (HASN and LASN). Five days later, all were euthanized, and samples collected from the jejunum for RNA isolation and sequencing. Resulting gene expression data were analyzed combining traditional statistics with machine learning to obtain signature gene lists for functional analysis. Differences in ATP production and cellular processes were observed in the jejunum between lines and following SRBC injection. HASN vs. LASN exhibited upregulation of ATP production, immune cell motility, and inflammation. LASI exhibits upregulation of ATP production and protein synthesis vs. LASN, reflective of what was observed in HASN vs. LASN. In contrast, no corresponding upregulation of ATP production was observed in HASI vs. HASN, and most other cellular processes appear inhibited. Without exposure to SRBC, gene expression in the jejunum indicates HAS generates more ATP than LAS, suggesting HAS maintains a "primed" system; and gene expression of HASI vs. HASN further suggests this basal ATP production is sufficient for robust antibody responses. Conversely, LASI vs. LASN jejunal gene expression implies a physiological need for increased ATP production with only minimal correlating antibody production. The results of this experiment provide insight into energetic resource needs and allocations in the jejunum in response to genetic selection and antigen exposure in HAS and LAS which may help explain phenotypic differences observed in antibody response.

Keywords: Virginia Tech HAS and LAS; antibody response; gene expression; genetic selection; machine learning.