In the Model-Based Product Line Engineering (MBPLE) context, modularization and separation of concerns have been introduced to master the inherent complexity of current developments. With the aim to exploit e ciently the variabilities and commonalities in MBPLs, the challenge of management of dependencies becomes essential (e.g. hierarchical and variability decomposition, inter-dependencies between models). However, one may observe that, in existing approaches, relational information (i) is mixed with other concerns, and (ii) lacks semantics and abstraction level identi cation. To tackle this issue, we make explicit the relationships and their semantics, and separate the relational concern into a Domain Speci c Modeling Language (DSML) called PLiMoS. Relationships are treated as rst-class entities and quali ed by operational semantics properties, organized into viewpoints to address distinct objectives, e.g. product derivation, variability consistency management, archi- tectural organization. This paper provides a description of the PLiMoS relationships de nition and its implementation in a model-based product line process using two variability languages: Feature Model and OVM. The independence with variability and core assets modeling languages provides bene ts to cope with the product line maintenance.