We cloned a cDNA encoding a novel synGAP, syn-GAP-d (GenBank TM accession number AB016962), from a rat brain cDNA library. The clone consisted of 4801 nucleotides with a coding sequence of 3501 nucleotides, encoded a protein consisting of 1166 amino acids with >99% homology with 1092 amino acid overlaps to syn-GAP, and contained a 13-nucleotide insertion to the previously reported synGAP mRNAs, which suggested that the clone was a splice variant of synGAP. We also found that there are at least seven variants in the 3 portion of the synGAP mRNA and that they encoded five different protein isoforms. The coding sequence of these C-terminal variants were classified into ␣1, ␣2, 1, 2, 3, 4, and ␥, and synGAP-d was classified as the 1 form. The previously reported synGAPs (synGAP-a, -b, and -c and p135synGAP) can be classified as the ␣1 isoform. All isoforms were expressed specifically in the brain. Unexpectedly, the  isoform, which lacks a C-terminal PSD-95-binding motif ((S/T)XV), was more restricted to the postsynaptic density fraction than the motif-containing ␣1 isoform. The  isoform did not interact with PSD-95 but specifically interacted with a nonphosphorylated ␣ subunit of Ca 2؉