We study flow optimization in Delay Tolerant Networks (DTNs), which we model using Capacity Region Evolving Graphs (CREGs). CREGs consist of different instances (called replicas) of the network graph in cascade; each replica is associated with a distinct time period (called epoch) and its own Capacity Region. Although CREGs can model any DTN, they are particularly well suited for the study of wireless ones. We define a single-commodity utility maximization problem in a CREG of T replicas that contains as special cases various interesting flow maximization problems. Using dual decomposition, we cast the maximization as a dual problem that can be solved iteratively and where in each iteration a set of T problems, T times smaller than the original, are solved, potentially (if multiple processors are available) in parallel. In addition, we propose two suboptimal utility maximization heuristics that operate on an epoch-by-epoch basis and we discuss a multi-commodity extension to the problem.