Plastid symbioses between heterotrophic hosts and algae are widespread and abundant in surface oceans. They are critically important both for extant ecological systems and for understanding the evolution of plastids. Kleptoplastidy, where the plastids of prey are temporarily retained and continuously re-acquired, provides opportunities to study the transitional states of plastid establishment. Here, we investigated the poorly studied marine centrohelid Meringosphaera and its previously unidentified symbionts using culture-independent methods from environmental samples. Investigations of the 18S rDNA from single-cell assembled genomes (SAGs) revealed uncharacterized genetic diversity within Meringosphaera that likely represents multiple species. We found that Meringosphaera harbors plastids of Dictyochophyceae origin (stramenopiles), for which we recovered six full plastid genomes and found evidence of two distinct subgroups that are congruent with host identity. Environmental monitoring by qPCR and catalyzed reporter deposition-fluorescence in situ hybridization (CARD-FISH) revealed seasonal dynamics of both host and plastid. In particular, we did not detect the plastids for 6 months of the year, which, combined with the lack of plastids in some SAGs, suggests that the plastids are temporary and the relationship is kleptoplastidic. Importantly, we found evidence of genetic integration of the kleptoplasts as we identified host-encoded plastid-associated genes, with evolutionary origins likely from the plastid source as well as from other alga sources. This is only the second case where host-encoded kleptoplast-targeted genes have been predicted in an ancestrally plastid-lacking group. Our results provide evidence for gene transfers and protein re-targeting as relatively early events in the evolution of plastid symbioses.