ABSTRACT: Human glucose 6-phosphate dehydrogenase, purified after overexpression in E. coli, was shown to contain one molecule/subunit of acid-extractable "structural" NADP+ and no NADPH. This tightly bound NADP+ was reduced by G6P, presumably following migration to the catalytic site. Gel-filtration yielded apoenzyme, devoid of bound NADP+ but, surprisingly, still fully active. Mr of the main component of "stripped" enzyme by gel filtration was approximately 100,000, suggesting a dimeric apoenzyme (subunit Mr = 59,000). Holoenzyme also contained tetramer molecules and, at high protein concentration, a dynamic equilibrium gave an apparent intermediate Mr of 150 kDa. Fluorescence titration of the stripped enzyme gave the K d for structural NADP+ as 37 nM, 200-fold lower than for "catalytic" NADP+. Structural NADP+ quenches 91% of protein fluorescence. At 37 degrees C, stripped enzyme, much less stable than holoenzyme, inactivated irreversibly within 2 d. Inactivation at 4 degrees C was partially reversed at room temperature, especially with added NADP+. Apoenzyme was immediately active, without any visible lag, in rapid-reaction studies. Human G6PD thus forms active dimer without structural NADP+. Apparently, the true role of the second, tightly bound NADP+ is to secure long-term stability. This fits the clinical pattern, G6PD deficiency affecting the long-lived non-nucleate erythrocyte. The Kd values for two class I mutants, G488S and G488V, were 273 nM and 480 nM, respectively (seven- and 13-fold elevated), matching the structural prediction of weakened structural NADP+ binding, which would explain decreased stability and consequent disease. Preparation of native apoenzyme and measurement of Kd constant for structural NADP+ will now allow quantitative assessment of this defect in clinical G6PD mutations.