The endo-beta-1,4-glucanase gene (celA) present in clone TC1 from Ruminococcus albus strain AR67 contained an open reading frame (ORF) of 414 codons, encoding a protein of calculated molecular weight 46 kD. The N-terminal 26 amino acids had features characteristic of bacterial signal peptides and a typical Shine-Dalgarno sequence was present 8 bp upstream from the translation start site. Analysis of RNA transcribed from the gene in E. coli and R. albus showed that transcription was initiated at different sites in the two bacteria. Primer extension analysis revealed four possible transcription initiation sites in E. coli, with typical promoter-like sequences situated upstream from all four sites. In contrast, mRNA from AR67 appeared to extend far enough upstream from celA to encode at least one other protein, indicating that celA was expressed as part of a polycistronic mRNA. Consensus promoter sequences recognized by E. coli in clone TC1 appeared nonfunctional in R. albus. The celA gene and its product (RaEND) showed 84% DNA homology, and 88% amino acid homology to the R. albus SY3 celA gene. However, the sequences differed markedly upstream from amino acid 58 of RaEND, although PCR analysis confirmed that no DNA rearrangements had occurred to cause the difference during cloning of TC1. This region contains the signal peptide in RaEND, a feature that is present in all members of the A4 sub-family of cellulases with the exception of the SY3 EGA.