The production of H(2)O(2) by isolated mitochondria is frequently used as a measure of mitochondrial superoxide formation. Matrix superoxide dismutase quantitatively converts matrix superoxide to H(2)O(2). However, matrix enzymes such as the glutathione peroxidases can consume H(2)O(2) and compete with efflux of H(2)O(2), causing an underestimation of superoxide production. To assess this underestimate, we depleted matrix glutathione in rat skeletal muscle mitochondria by more than 90% as a consequence of pretreatment with 1-chloro-2,4-dintrobenzene (CDNB). The pretreatment protocol strongly diminished the mitochondrial capacity to consume exogenous H(2)O(2), consistent with decreased peroxidase capacity, but avoided direct stimulation of superoxide production from complex I. It elevated the observed rates of H(2)O(2) formation from matrix-directed superoxide by up to two-fold from several sites of production, as defined by substrates and electron transport inhibitors, over a wide range of control rates, from 0.2-2.5 nmol H(2)O(2) min(-1) mg protein(-1). Similar results were obtained when glutathione was depleted using monochlorobimane or when soluble matrix peroxidase activity was removed by preparation of submitochondrial particles. The data indicate that the increased H(2)O(2) efflux observed with CDNB pretreatment was a result of glutathione depletion and compromised peroxidase activity. A hyperbolic correction curve was constructed, making H(2)O(2) efflux a more quantitative measure of matrix superoxide production. For rat muscle mitochondria, the correction equation was: CDNB-pretreated rate = control rate + [1.43 x (control rate)]/(0.55 + control rate). These results have significant ramifications for the rates and topology of superoxide production by isolated mitochondria.