Investigating the variance components pattern and the genetic basis of repeated Fourier-transform infrared spectra in sheep milk

Abstract

The variability of milk spectra within and between dairy animals reflects the complex biological processes underlying milk production. Studying this variability may offer insights into an animal's physiological or health status and help elucidate relationships between specific genes and milk's chemical structure. Over 4 yr, biweekly Fourier-transform infrared spectra were recorded on 41,075 sheep milk samples collected from 1,256 Sarda ewes at morning and evening milkings. Individual variance components for absorbance at wavenumbers between 5,000 and 925 cm^-1 were estimated using a REML procedure with a repeated measures mixed model. Fixed effects included the milking time, the combined effect of test date and lactation stage, and the combined effect of age and parity. Random effects included ewe within date, within lactation, and across lactations. Genetic random effects were modeled using a genomic relationship matrix based on 43,390 SNP. Moreover, a GWAS was conducted for each wavenumber. The most important component of individual variation across the spectrum was the additive genetic, on average, 17% of total phenotypic variance, with peaks in ranges (1,142-1,273 cm^-1; 1,443-1,470 cm^-1; 2,839-2,978 cm^-1) associated with protein and fat contents. The across-lactation component was generally higher than the within-lactation component, suggesting that individual nongenetic effects permanently affect differences between animals. The ewe within-date component explained a smaller proportion of variance but showed high values at specific wavenumbers (e.g., 2,975 cm^-1, 1,555-1,524 cm^-1), potentially linked to short-term physiological events, such as stress or metabolic disorders. Repeatability estimates ranged from 0 to 0.75. A total of 471 significant SNP × wavenumber combinations were identified, corresponding to 18 SNP on 8 chromosomes and 381 wavenumbers. Significant SNP included rs399070200 in LALBA and rs410141696 close to the casein gene cluster. This study provides insights into the origin of individual variation in milk Fourier-transform infrared spectra, which could support breeders in assessing the suitability of existing simplified large-scale protocols for spectral traits of potential interest for selection and in exploring innovative recording strategies. Moreover, estimates of the genetic components and GWAS results can enable the design of selective breeding programs targeting milk composition, animal well-being, and health traits in dairy sheep.

Keywords: GWAS; dairy sheep; mid-infrared spectra; variance components.