A systematic mutagenesis of the SV40 enhancer indicates that it spans -100 bp and is composed of at least two distinct DNA domains which exhibit very little enhancing activity on their own. Their association results in a 400-fold enhancement of transcription, virtually irrespective of their relative orientation and, to some extent, of the distance between them. Enhancer activity can also be generated by duplication of either domain. We show also that the activity of each domain is due to the presence of several specific sequence motifs. These motifs are found assorted in different combinations in other viral and cellular enhancers. Key words: transcription/site-directed mutagenesis/promoter/ RNA polymerase B(ll)/simian virus 40 1983;Lusky et al., 1983;Hearing and Shenk, 1983;Veldman et al., 1985). However, such enhancer consensus sequences occur in DNA segments without enhancer properties, and enhancer activity can be generated by duplicating DNA sequences without enhancer activity on their own (Weber et al., 1984;Swimmer and Shenk, 1984), thus questioning the real functional significance of these consensus sequences.Therefore, we decided to determine, at the nucleotide level, the DNA sequences essential for the activity of the prototype SV40 enhancer. Systematic deletions and point mutations have been constructed throughout the enhancer region and their effect on SV40 early transcription investigated in vivo, using a transient expression assay in HeLa cells. We show here that the SV40 enhancer encompasses a large DNA segment of -100 nucleotides containing the 72-bp sequence, but also extending further upstream. Furthermore, the present study reveals that the enhancer is composed of at least two distinct DNA domains which exhibit very little enhancing activity on their own. However, their association results in a dramatic 400-fold enhancement of transcription, virtually irrespective of their relative orientation and, to some extent, of the distance between them. Enhancer activity can also be generated by duplication of either domain. In addition, we show that the activity of each domain is due to the presence of several specific sequence motifs. Various assortments of these motifs are found in other viral and cellular enhancers, suggesting that enhancers are mosaics of a limited number of basic evolutionary related sequence motifs.