Classic in vitro studies show that the Hsp70 chaperone system from Escherichia coli (DnaK-DnaJ-GrpE, the DnaK system) can bind to proteins, prevent aggregation, and promote the correct refolding of chaperone-bound polypeptides into native proteins. However, little is known about how the DnaK system handles proteins that have already aggregated. In this study, glucose-6-phosphate dehydrogenase was used as a model system to generate stable populations of protein aggregates comprising controlled ranges of particle sizes. The DnaK system recognized the glucose-6-phosphate dehydrogenase aggregates as authentic substrates and specifically solubilized and refolded the protein into a native enzyme. The efficiency of disaggregation by the DnaK system was high with small aggregates, but the efficiency decreased as the size of the aggregates increased. High folding efficiency was restored by either excess DnaK or substoichiometric amounts of the chaperone ClpB. We suggest a mechanism whereby the DnaK system can readily solubilize small aggregates and refold them into active proteins. With large aggregates, however, the binding sites for the DnaK system had to be dynamically exposed with excess DnaK or the catalytic action of ClpB and ATP. Disaggregation by the DnaK machinery in the cell can solubilize early aggregates that formed accidentally during chaperone-assisted protein folding or that escaped the protection of "holding" chaperones during stress.