Since publication of the above article, the Editorial Office and the authors have become aware that data provided by one of the co-authors could not be traced back to the original experiments. The authors have, therefore, removed the following figure panels: Fig. 5C (all bar charts) and Fig. 5D (the black and striped bars); Fig. 6B (the black and striped bars). In the Supplementary material, Supplementary Fig. 2 and the related legend and methodology (HMGB1 analysis by electrospray ionization liquid chromatography mass spectrometry) have been removed. The fundamental conclusion of the article, that the predictive biomarker value of total blood HMGB1 levels for treatment response to antioxidant drugs, remains unaltered after removing the isoform-related data, although it can no longer be stated that the acetylated, reduced and disulphide isoforms of HMGB1 are changing in the brain and blood of animals and whether they predict any of the pathological outcomes. The pharmacological effects of the treatments on disease outcomes are unaffected by the removal of the isoform data. Therefore, the disease modification effects of the antioxidant treatments on seizures, neurodegeneration and cognitive deficit remains unaltered. These data and the name of the co-author have been removed from the manuscript and the Editor is satisfied that the fundamental conclusion of the article is unaffected by these changes. The original manuscript has therefore been retracted and replaced with the corrected version (doi:10.1093/brain/awz130).