Nature 422, 164–169 (2003).

Owing to mislabelling by the mouse provider, the strain of mice used as a control in the experiment shown in Fig. 4 was not C3H/HeJ but C3H/HeOu. C3H/HeJ have a defect in TLR4-mediated signalling that the other C3H strains do not have. This correction does not affect our conclusions.