All of the coronavirus defective interfering (DI) RNAs analyzed thus far contain an open reading frame (ORF) from which DI RNA-specific protein(s) are translated, although the function of the DI-specific protein and the significance of the ORF are not known. A complete cDNA clone of a mouse hepatitis virus (MHV) DI RNA, NE-1, containing a single nucleotide deletion in the 5' region of the ORF was obtained and analyzed. Due to this single nucleotide deletion, a DI-specific protein of 7.5-kDa was made from NE-1, in contrast to the 88-kDa protein made from the wild-type DI RNA. NE-1 RNA was efficiently replicated after transfection into MHV-infected cells. However, after one passage of NE-1 RNA-containing virus, the 88-kDa wild-type protein was synthesized, indicating that the large ORF was restored during NE-1 DI RNA replication. Sequence analysis of NE-1 DI RNA from infected cells demonstrated that in approximately half of the DI RNA population, the ORF was restored by RNA recombination between NE-1 DI RNA and helper virus genomic sequence. The sequences of other DI RNAs contained an additional nontemplated A at the five-A sequence nine nucleotides upstream of the deletion site, resulting in a stretch of six consecutive As. In these "edited"-type DI RNAs, the original nucleotide deletion was maintained and no RNA recombination was observed. This "editing" produced an ORF of the same size as the wild-type DI RNA. We conclude that the DI RNA with a large ORF has a selective advantage. There was no significant difference in replication efficiency among these RNAs when they replicated alone. However, cotransfection of two DI RNA species and time course experiments suggested that homologous interference and other mechanism(s) during the early stage of virus multiplication are responsible for the accumulation of DI RNAs containing the large ORF.