As systems are more and more complex and heterogeneous, Domain Specific Languages (DSLs) can be used to model their many concerns at a high abstraction level. Early Validation and Verification (V&V) of such models is an important and difficult task. There are numerous possible V&V techniques, none being completely satisfying regarding specific requirements (behavioral coverage, scalability, required expertise, etc.) or depending on the context. The objective of my PhD work is to find ways to make an efficient coordinated use of different complementary V&V techniques. In particular, I focus on the need to handle gathered evidence with corresponding property-behavior coverage, and on the semantic gap between domain languages and V&V languages. The proposed approach include the designing of a language to encode V&V evidence, and on the use of model types to both capitalize V&V manipulations and fill the gap between DSLs and V&V languages.