Collagen is present in Porifera, the lowest multicellular animals, but there is no information available on the primary structure of the collagen chains in this phylum. Developing fresh-water sponges have been used to extract total RNA in order to study in vitro translation products and to construct a cDNA library. Four translated proteins were collagenase-sensitive (200 kDa, 160 kDa, 81 kDa and 48 kDa). The cDNA library was screened with a human collagen probe and a clone, EmC4, covering 1.2 kb was isolated. Nucleotide sequencing of EmC4 revealed a conceptual open reading frame coding for 366 amino acids terminated by a stop codon TGA with 103 nucleotides downstream. The presumed translation product encoded contained several domains: a non-collagenous C-terminal domain of 156 amino acids with 9 cysteines, an uninterrupted collagenous domain of 171 amino acids, a non-collagenous domain of 16 amino acids with 3 cysteines and a probably incomplete N-terminal collagenous domain of 23 amino acids. Comparison with other sequences suggested that this collagen chain might belong to a non-fibrillar collagen family which evolved into several sub-families giving rise to nematode cuticular collagens, and type IV collagens.