Nature 519, 92–96 (2015); doi:10.1038/nature14232

Some clarifications are provided to this Letter; these do not alter any of the central conclusions but, rather, are provided in the interests of transparency and reproducibility. Our Letter indicated that experiments were performed on 4-week-old mice (unless stated otherwise). In fact, for several experiments, mice ranged from 5 to 7 weeks as follows: Fig. 4a–h, Extended Data Fig. 9g–w, b′–t′, z: 5 weeks old; Figs 1, 2, 3a–d, 4i–o, Extended Data Figs 1a–d, 2, 4, 5s–v, 6, 7h–k, 8l–s, 9a–f, x, y, 10: 6 weeks old; Fig. 3e–l, Extended Data Fig. 1e–l, 5q, r, 7a–g, l–h′, 8a–k, t–o′, 9a′: 7 weeks old. The weight gain versus time curves are affected by mouse age and hence explain why the kinetics of weight gain differ among control mice when comparing between different experiments.

For each experiment, we listed the average n value for all the conditions within each panel, which differed from the exact n for each experimental condition within each figure, as shown in Supplementary Table 1 of this Corrigendum.

Furthermore, our Letter reported relative changes in body mass and absolute mass of fat pads, thus not permitting assessment of absolute weight changes nor fat pad mass relative to total body mass. Hence, we provide measures of absolute and relative body and fat pad mass in a side-by-side manner in the Supplementary Data to this Corrigendum.