Aims/hypothesis: Low-dose IL-2 (ld-IL2) selectively activates and expands regulatory T cells (Tregs) and thus has the potential to skew the regulatory/effector T (Treg/Teff) cell balance towards improved regulation. We investigated which low doses of IL-2 would more effectively and safely activate Tregs during a 1 year treatment in children with recently diagnosed type 1 diabetes. Methods: Dose Finding Study of IL-2 at Ultra-low Dose in Children With Recently Diagnosed Type 1 Diabetes (DF-IL2-Child) was a multicentre, double-blinded, placebo-controlled, dose-finding Phase I/II clinical trial conducted in four centres at university hospitals in France: 24 children (7-14 years old) with type 1 diabetes diagnosed within the previous 3 months were randomly assigned 1:1:1:1 to treatment by a centralised randomisation system, leading to a 7/5/6/6 patient distribution of placebo or IL-2 at doses of 0.125, 0.250 or 0.500 million international units (MIU)/m2, given daily for a 5 day course and then fortnightly for 1 year. A study number was attributed to patients by an investigator unaware of the randomisation list and all participants as well as investigators and staff involved in the study conduct and analyses were blinded to treatments. The primary outcome was change in Tregs, expressed as a percentage of CD4+ T cells at day 5. It pre-specified that a ≥60% increase in Tregs from baseline would identify Treg high responders. Results: There were no serious adverse events. Non-serious adverse events (NSAEs) were transient and mild to moderate. In treated patients vs placebo, the commonest NSAE was injection site reaction (37.9% vs 3.4%), whereas other NSAEs were at the same level (23.3% vs 19.2%). ld-IL2 induced a dose-dependent increase in the mean proportion of Tregs, from 23.9% (95% CI -11.8, 59.6) at the lowest to 77.2% (44.7, 109.8) at the highest dose, which was significantly different from placebo for all dose groups. However, the individual Treg responses to IL-2 were variable and fluctuated over time. Seven patients, all among those treated with the 0.250 and 0.500 MIU m-2 day-1 doses, were Treg high responders. At baseline, they had lower Treg proportions in CD4+ cells than Treg low responders, and serum soluble IL-2 receptor α (sIL-2RA) and vascular endothelial growth factor receptor 2 (VEGFR2) levels predicted the Treg response after the 5 day course. There was no significant change in glycaemic control in any of the dose groups compared with placebo. However, there was an improved maintenance of induced C-peptide production at 1 year in the seven Treg high responders as compared with low responders. Conclusions/interpretation: The safety profile at all doses, the dose-dependent effects on Tregs and the observed variability of the Treg response to ld-IL2 in children with newly diagnosed type 1 diabetes call for use of the highest dose in future developments. The better preservation of insulin production in Treg high responders supports the potential of Tregs in regulating autoimmunity in type 1 diabetes, and warrants pursuing the investigation of ld-IL2 for its treatment and prevention.