Although the CBFA1 gene encodes an osteoblast-specific transcription factor that regulates osteoblast differentiation, uncertainty exists about the organization of its 5' end and the relevance of a novel N-terminal sequence identified in the mouse Cbfa1/Osf2 isoform. We found the novel 5' Cbfa1/Osf2 sequence is encoded by a previously unrecognized upstream exon, designated exon -1, which is highly conserved in mouse, rat and human. In addition, two splice donor sites may be utilized to generate Cbfa1/Osf2 cDNAs containing different N-terminal sequences. The first ATG and splice donor site in exon -1 is predicted to transcribe a cDNA containing the unique Osf2 5' sequence, whereas a second donor splice site gives rise to cDNAs that contain sequences encoding an 11 amino acid insert. In the human CBFA1 gene, an additional 2-bp nucleotide insert shifts the reading frame and results in stop codons in the cDNA sequence derived from exon -1. The 5'-most exon of the human CBFA1 gene, therefore, contains the 5' non-coding region rather than a human OSF2 homolog. The absence of a homologous OSF2 coding sequence in the human CBFA1 cDNA suggests that the novel mouse N-terminal Osf2 sequence is not essential for functioning of the CBFA1 gene product. In addition, multiple transcripts derived from a single CBFA1/Cbfa1 gene were detected in osteoblasts by Northern analysis and RT-PCR, including additional Cbfa1/Osf2 isoforms containing deletions of exons 1 and 4. Thus, the alternative use of transcription start sites and splicing leads to the genesis of CBFA1/Cbfa1 isoforms with possible differences in transactivation potentials.