Résumé : Aims/hypothesis: Alginate-encapsulated human islet cell grafts have not been able to correct diabetes in humans, whereas free grafts have. This study examined in immunodeficient mice whether alginate-encapsulated graft function was inferior to that of free grafts of the same size and composition. Methods: Cultured human islet cells were equally distributed over free and alginate-encapsulated grafts before implantation in, respectively, the kidney capsule and the peritoneal cavity of non-obese diabetic mice with severe combined immunodeficiency and alloxan-induced diabetes. Implants were followed for in vivo function and retrieved for analysis of cellular composition (all) and insulin secretory responsiveness (capsules). Results: Free implants with low beta cell purity (19 ± 1%) were non-functional and underwent 90% beta cell loss. At medium purity (50 ± 1%), they were functional at post-transplant week 1, evolving to normoglycaemia (4/8) or to C-peptide negativity (4/8) depending on the degree of beta cell-specific losses. Encapsulated implants immediately and sustainably corrected diabetes, irrespective of beta cell purity (16/16). Most capsules were retrievable as single units, enriched in endocrine cells that exhibited rapid secretory responses to glucose and glucagon. Single capsules with similar properties were also retrieved from a type 1 diabetic recipient at post-transplant month 3. However, the vast majority were clustered and contained debris, explaining the poor rise in plasma C-peptide. Conclusions/interpretation: In immunodeficient mice, i.p. implanted alginate-encapsulated human islet cells exhibited a better outcome than free implants under the kidney capsule. They did not show primary non-function at low beta cell purity and avoided beta cell-specific losses by rapidly establishing normoglycaemia. Retrieved capsules presented secretory responses to glucose, which was also observed in a type 1 diabetic recipient. Trial registration: ClinicalTrials.gov NCT01379729 Funding: This study was supported by grants from the JDRF (centre grant 4-2005-1327), the Research Foundation Flanders (G.0801.10), the 6th and 7th Framework Program of the European Commission (numbers 512145 and 241883), and the Agency for Innovation by Science and Technology in Flanders (IWT-TBM7 090884). © 2013 Springer-Verlag Berlin Heidelberg.