Clustering of the dynamics of milk lactose content throughout lactation and identification of variation factors

Interest in milk lactose content (LC) has grown due to its potential as an indicator of udder health and metabolic disorders in dairy cattle. However, the variability of LC dynamics during lactation remains poorly described, and a better characterization of these dynamics could clarify our understanding of LC variations among cows, and potentially those due to udder health and metabolism-related variations. The aim of this study was to identify distinct patterns of LC dynamics and assess their environmental and genetic determinants, as well as their phenotypic and genetic associations with milk yield, Na, K, SCC, and their phenotypic associations with fat-to-protein ratio (FPR) and BHB. A total of 1,980,693 test-day records were analyzed from 183,150 Holstein cows in 2,239 herds across France. At least 2 records in the first 90 d and 4 records between 7 and 300 DIM were available for each cow, averaging 7.3 records per cow. Functional principal component analysis was used to describe LC dynamics throughout lactation. This approach involved smoothing the LC curves for each cow and then summarizing their overall shape using 3 principal components: average LC throughout lactation, LC slope and LC at mid lactation. Dynamics of LC were grouped into 6 clusters. Three clusters (3, 4, and 5) represented 86% of the data and shared similar dynamics with a flat trend after an initial rise in LC at early lactation and different average LC levels (cluster 3: 4.68%, cluster 4: 5.10%, cluster 5: 4.86%). Environmental and intrinsic animal factors explained 69% of the variability in average LC among clusters 3, 4, and 5, with cow parity and LC EBV identified as the main intrinsic contributors. The remaining clusters (1, 2, and 6) showed LC levels similar to the mean of the dataset (LC: 4.88% ± 0.19%) until 150 DIM. Thereafter, 2 clusters displayed negative LC slopes (cluster 1: -0.13% and cluster 6: -0.07% per month of lactation) and one a positive slope (cluster 2: +0.05% per month). Fifteen percent of the variations in the LC slopes of clusters 1, 2, and 6 was explained by environmental factors, mainly calving season: winter calving was associated with clusters 1 and 6, and summer calving with cluster 2. The remaining unexplained negative slope variation from mid lactation onward appears to be related to distinct patterns characterized by higher FPR and milk BHB concentrations preceding the decline in LC, followed by increased SCC after 150 DIM. The lactose-based clusters also corresponded to distinct curves for milk sodium contents, and there were slight correlations for potassium levels, indicating different equilibria between the 3 main osmotic agents (lactose, sodium, and potassium), probably to maintain milk osmolarity. As a result, LC clustering uncovered meaningful physiological profiles: average LC levels were primarily driven by parity and genetics, whereas LC slope variations appeared to be more sensitive to environmental and health-related factors. These findings support the potential of LC dynamics, accessible via mid-infrared spectra, to serve as functional biomarkers for udder health and, potentially, of metabolic status.