The diversity and composition of local communities depends strongly on the pool of species that have been able to colonize that community from elsewhere. Typically this is thought to depend on a larger regional species pool that is subject to local environmental constraints that act as "filters." Often, however, colonists arrive from multiple sources that differ in habitat conditions and have therefore already experienced distinct "prefiltering." Consequently, it is the interaction of species from these distinct pools that determine the composition of local communities. This interaction is particularly important when certain colonist pools provide keystone species with disproportionate roles on community assembly. We propose to identify these key colonist pools and their interaction with local habitat filters by quantifying community-level responses to colonist pool manipulation. We tested this framework to assess the contribution of surface and burrow sediment bacteria to bacterial communities associated with the fiddler crab, Uca panacea. In a mesocosm experiment, we combined normal and autoclaved surface and burrow sediment in a factorial experimental design, and we evaluated the community-level responses of carapace and gut microbial assemblages to sediment treatments with next-generation sequencing of the 16S rRNA gene. Results from carapace bacterial communities indicate that burrow sediments contribute most recruits, but surface sediments provide a few key colonizers that become established in the carapace community. In contrast, the composition of gut-associated microbial communities responded only to surface bacteria manipulation, despite being highly dissimilar from the community composition in both the surface and burrow source pools. These results suggest that assembly in the gut depends primarily on colonization from the surface sediment and regulation by habitat filtering. For fiddler crab-associated bacteria, we can conclude that key colonist pools and habitat filters regulate the influence of multiple colonist pools. Incorporating and distinguishing the contribution of multiple sources of species, rather than a single regional species pool, may better explain community dynamics in many systems, especially those with weak habitat filters.