We report the isolation and characterization of a cDNA clone encoding HSP47, a transformation-sensitive heat shock protein that binds to collagen. A cDNA library was prepared from total RNA isolated from heat-shocked chicken embryo fibroblasts and screened by using oligonucleotide mixtures prepared on the basis of the N-terminal amino acid sequence of biochemically purified HSP47. The cDNA insert contained 3,278 bp, which encoded a 15-amino-acid signal peptide and a mature protein coding region consisting of 390 amino acid residues; it also included part of the 5' noncoding region and a long 3' noncoding region. The deduced amino acid sequence revealed an RDEL sequence at the C terminus, which is a variant of the KDEL retention signal for retention of proteins in the endoplasmic reticulum. Northern (RNA) blot analyses and nuclear run-on assays established that the induction of HSP47 by heat shock and its suppression after transformation of chicken embryo fibroblasts by Rous sarcoma virus are regulated at the transcriptional level. A homology search revealed that this protein belongs to the serpin family, the superfamily of plasma serine protease inhibitors. Although structurally homologous to the serpins, HSP47 lacks the active site thought to be essential for the inhibition of proteases and does not appear to bind to intracellular proteases. HSP47 is the first heat shock protein found to be a member of the serpin superfamily. Conversely, it is the first serpin family member that is not secreted from cells, which could be explained by acquisition of the RDEL retention signal during evolution.