Two lines of evidence point to a hairpin DNA intermediate in V(D)J joining (V, variable; D, diversity; J, joining) [Lieber, M.R. (1991) FASEB J. 4, 2934-2944]. One is the presence of P nucleotide insertions (short inverted-repeat sequence) in V(D)J junctions [Lafaille, J. J., DeCloux, A., Bonneville, M., Takagaki, Y. & Tonegawa, S. (1989) Cell 59, 859-870]; a second is the detection of site-specifically broken DNA molecules with covalently closed (hairpin) termini in thymus DNA [Roth, D. B., Menetski, J. P., Nakajima, P., Bosma, M. J. & Gellert, M. (1989) Cell 70, 983-991]. However, P nucleotide insertions could be generated in ways not involving a hairpin structure, and because physical evidence for hairpin-ended DNA fragments has been obtained only with mutant mice, there is some uncertainty regarding the role of hairpin molecules in the normal V(D)J joining pathway. To determine whether mammalian cells are capable of metabolizing this odd type of DNA terminus and whether, in doing so, junctions with P insertions are in fact created, a linear DNA molecule with a hairpin closure at each end was transfected into several murine cell lines. The hairpin-ended molecules were recircularized, and the junctions exhibited P insertions at a high frequency. This result directly links the presence of P insertions to a hairpin precursor, providing strong evidence for the notion that a hairpin DNA intermediate exists in V(D)J recombination. A comparison of hairpin end joining in various cells, including those derived from mice with the severe combined immunodeficiency (scid) mutation, is presented.