Current gene-transfer technologies display limitations in achieving effective gene delivery. Among these limitations are difficulties in stably integrating large corrective sequences into the genomes of long-lived progenitor-cell populations. Current larger-capacity viral vectors suffer from biosafety concerns, whereas plasmid-based approaches have poor efficiency of stable gene transfer. These barriers hinder genetic correction of many severe inherited human diseases, such as the blistering skin disorder recessive dystrophic epidermolysis bullosa (RDEB), caused by mutations in the large COL7A1 gene. To circumvent these barriers, we used the phi C31 bacteriophage integrase, which stably integrates large DNA sequences containing a specific 285-base-pair attB sequence into genomic 'pseudo-attP sites'. phi C31 integrase-based gene transfer stably integrated the COL7A1 cDNA into genomes of primary epidermal progenitor cells from four unrelated RDEB patients. Skin regenerated using these cells displayed stable correction of hallmark RDEB disease features, including Type VII collagen protein expression, anchoring fibril formation and dermal-epidermal cohesion. These findings establish a practical approach to nonviral genetic correction of severe human genetic disorders requiring stable genomic integration of large DNA sequences.