Changes in milk fat composition influence its nutritional quality as well as the technological properties of butter. The impact of feed on fat composition is well known; however, limited information is available on the genetic variability of fatty acids in bovine milk. The overall aim of this PhD thesis was to study the genetic variability of fatty acid profile in bovine milk and fat. This type of research needs a large amount of data. Expensive reference analysis is used to measure the fatty acid contents in fat. The first objective of this thesis was to develop an alternative method that could be faster and cheaper than traditional methods. Calibration equations predicting the contents of fatty acid from mid-infrared spectrum were established. The contents of saturated and monounsaturated fatty acids, omega-9, fatty acids with short, medium and long chain were the best predicted. Thanks to the implementation of this method in the Walloon routine milk recording, more than 20,000 milk samples were analyzed. This database permitted to model the variation of fatty acid contents in milk and fat. From these models, the genetic variability of fatty acid profile was shown. The complexity of models increased throughout this project due to the increase of new available data. Differences across 7 dairy breeds were estimated using single and multi-trait mixed models. Milk fat and delta-9 desaturase activity of Jersey and dual purpose Belgian Blue differed significantly from Holsteins. Therefore, the choice of a given breed could modify the fat composition. Heritability values obtained for studied fatty acids with multi-trait mixed model ranged from 0.05 to 0.42. Higher values were observed for saturated compared to unsaturated fatty acids. Moderate heritability estimates were observed for the activity of delta-9 desaturase (0.20) and the hardness of butter (0.27). These two traits were estimated by specific fatty acid ratios. The heritability observed using a multi-trait random regressions test day model for the content of saturated fatty acids (0.42) was similar to the one observed for the percentage of fat (0.37). Considering the impact of selection on fat content, the selection could have a great impact on fat composition. High genetic correlations were observed between some fatty acids having similarities in their synthesis. Heritability and correlations varied through the duration of the lactation. Due to the large number of fatty acids, the estimation of an index, which includes the proper fatty acid profile, could be interesting for a future selection program. This PhD thesis provides the background required by future studies to estimate the impact of animal selection on milk fat composition.

Hedi HAMMAMI (2009). Genotype by Environment Interaction for Production Traits of Holsteins Using Two Countries as Model: Luxembourg and Tunisia (Doctoral thesis). Gembloux, Belgium, Gembloux Agricultural University, 170 p., 30 tabl., 16 fig. Summary. Under globalization, breeding organizations are selecting animals and exchanging germplasm across various environments. Ignoring genotype by environment interaction (G x E) may affect the efficiency of breeding strategies and limit outcomes from cooperation between breeding programs. Quantifying the effectiveness of indirect selection and effects of G x E for different breeds is therefore necessary. The objective of this thesis was to evaluate the magnitude of G x E for milk yield using Luxembourg and Tunisian Holstein populations. In fact, these two countries rely considerably on importation of superior genes from diverse origins for their breeding programs. This study needed records on both the genotype and the environment. In the first part of this thesis, genetic ties between the two populations were studied. Additive relationships and genetic similarity were important and genetic links have been strengthened with time which allowed the analysis of the phenotypic expression of daughters of common sires under each of these tow production environments. In the second part, genetic parameters for production traits of Tunisian Holsteins were estimated by a test-day random regression model (RRTD). Heritability estimates for 305-d milk, fat and protein yields were low to moderate (0.12 to 0.18) suspecting difficulties of high-producing cows to express their potential under limiting production conditions. In the third part, G x E for milk yield and persistency were investigated using character state models, where milk yield in each country was considered as a separate trait, and where the country border delimitation was designed as an environmental character state. A RRTD sire model was applied and was extended to a RRTD animal model. Significant G x E was detected for milk yield and persistency by both models. Large differences in genetic and permanent environmental variances between the two countries were observed. Genetic correlations for 305-d milk yield and persistency between Luxembourg and Tunisian Holsteins were 0.50 and 0.43 (sire model) and 0.60 and 0.36 (animal model). Moreover, low rank correlations obtained between estimated breeding values of common sires translate a significant re-ranking between the two environments. At the end of this thesis, a herd management (HM) parameter reflecting feeding and management intensity was defined. Three HM levels were identified in each country and G x E was investigated within- and across-environments. Significant G x E was detected between the Tunisian HM levels, whereas, only heterogeneous genetic variance for milk yield with limited re-ranking of sires across the three Luxembourg environments was observed. Overall, this thesis shows that under constraining environmental effects, selection for adaptive traits among economically valuable traits under their specific conditions is needed for low-input systems. When satisfactory feeding resources, management and husbandry practices are available, high degree environmental sensitivity is desired and the use of a high yielding breed may be encouraged.