It is becoming clear that one key characteristic of 5G communication systems will be heterogeneity. Heterogeneity is important to be able to reuse equipment that is available, extract maximum capacity from what is allowed to be used in given spectrum bands and locations due to regulatory limitations, and to match capabilities to the extremely wide range of use cases of 5G. Importantly, this heterogeneity will extend to the use of a range of different spectrum bands and spectrum types. Moreover, 5G must be supported extensively by lower-frequency options; millimeter-wave will be groundbreaking, but for propagation reasons can't be the only solution to serve 5G, e.g., in achieving five-nines reliability. A key to realising lower-frequency spectrum for broadband (including mobile) communications systems is spectrum sharing. This has been emphasised by various reports by bodies such as the President's Council of Advisors on Science and Technology in the US, and the European Commission in the EU. It is also noted that high-profile efforts such as TV White Space (TVWS) and the 3.5 GHz "Innovation Band" in the US imply geolocation database-assisted spectrum sharing as the solution to realising this extra capacity. This paper proposes the broad introduction of a geolocation database-based system to assist heterogeneous spectrum usage and spectrum sharing in 5G - both to enable 5G systems to use spectrum assigned to other services (e.g., TV spectrum, as in the TVWS example), and to enable 5G systems operating under different operators/owners to share spectrum with each other. This paper also argues for the geolocation databases to act as a management capability among secondary/opportunistic networks and devices, to better allocate resources, or combinations (aggregations) thereof, in the light of traffic requirements. The benefits of the approach are shown through reference to performances achieved by the authors within a major regulator-driven trial of TVWS in the UK. Further architectural observations are derived based on those performances. It is important to note that this paper only proposes the high-level nature of such an architecture. Extensive detail is currently impossible to provide given that the basic architectural assumptions for 5G systems are not yet known.