Folding of the human coxsackie and adenovirus receptor immunoglobulin (Ig) variable-type domain (CAR D1) during overexpression in the Escherichia coli cytoplasm was shown previously to be partially rescued by fusion to a 22-residue C-terminal peptide. Here, peptide sequence features required for solubilization and folding of CAR D1 and similar Ig variable-type domains from two other human membrane proteins were investigated. Peptide extensions with net negative charge > -6 fully solubilized CAR D1, and approximately half of the peptide-solubilized protein was correctly folded. The Ig variable-type domains from human A33 antigen and myelin P-zero proteins were only partially solubilized by peptide extensions with net charge of -12, however, and only the solubilized P-zero domain appeared to fold correctly whereas the A33 domain formed soluble microaggregates of misfolded protein. Our results suggest a model where the large net charge of peptide extensions increases electrostatic repulsion between nascent polypeptides. The resulting decrease in aggregation rate can enable some polypeptides to fold spontaneously into their native protein conformations. Analysis of the solubility and folding status of sets of structurally homologous proteins, such as the Ig variable-type domains described here, during overexpression could provide insights into how amino acid and gene sequences influence the efficiency of spontaneous protein folding. (C) 2004 Elsevier Inc. All rights reserved.