In Caenorhabditis elegans, the gene unc-89 is required for A-band organization of striated muscle. In mammals, a likely homolog of UNC-89, called obscurin, has been described and found to be localized at both the M-lines and Z-discs of striated muscle. Here, we show that the coding sequence for unc-89 is larger than originally thought, and that the gene encodes at least four major isoforms: UNC-89-A (original isoform, 732 kDa), UNC-89-B (potentially 900 kDa), and UNC-89-C and UNC-89-D (each 156 kDa). UNC-89-C and -D, except for unique N-terminal tails of eight and 11 residues, respectively, are co-linear with the C terminus of UNC-89-B. The unc-89 complex transcription unit contains at least three promoters: one directing UNC-89-A and -B primarily in body-wall and pharyngeal muscle, one internal promoter directing expression of UNC-89-C primarily in body-wall muscle, and one internal promoter directing expression of UNC-89-D primarily in a few muscle cells of the tail. Isoform-specific RNA interference resulted in a muscle structural phenotype similar to a typical unc-89 mutant, but with varying degrees of severity. Antibodies generated to the interkinase region shared by the UNC-89-B, -C and -D isoforms localize to the middle of A-bands, like previously-described UNC-89 antibodies, and detect proteins on immunoblots consistent with the proposed gene organization and additional isoforms. The three new UNC-89 isoforms contain two protein kinase domains, of the myosin light chain kinase (MLCK) family. UNC-89-B contains two complete protein kinase domains, designated PK1 and PK2. UNC-89-C and -D begin with partial kinase domains, PK1-C and PK1-D. Homology modeling suggests that PK2 is catalytically active, PK1 is inactive, and that PK1-C and PK1-D have similar structures at their N termini that may create sites for interaction with other proteins.