In gene expression studies, promoters are often fused to a protein-encoding reporter gene, the expression of which is then taken as an indirect measure of their strength. Here, we advocate the use of a tRNA reporter for the direct quantification of promoter strength. Using this method, we have studied the bacteriophage T7 gene 10 promoter in an E.coli strain that produces saturating amounts of T7 RNA polymerase. At 370C in aminoacid-glycerol medium, we show that this promoter ranks amongst the strongest known, directing ca 1.1 transcription events per second, 2.2-fold more than the promoters for rRNA operons, or 15-fold more than the induced lac promoter. Surprisingly, compared to the lac promoter, the T7 promoter is far less efficient in driving the expression of protein-encoding genes such as cat, neo or lacZ. Therefore, the polypeptide yield per transcript is lower when the T7 RNA polymerase is used instead of the E.coli RNA polymerase. The former enzyme travels faster than the translating ribosomes, and we suggest that this desynchronization lowers the polypeptide yield per transcript. INTRODUCTIONBacteriophages T3, T7 and SP6 encode closely related RNA polymerases which transcribe the viral genome from highly specific promoters, during the late phase of infection (1). The prototype of the group, the T7 RNA polymerase, can be stably expressed in E. coli, and genes that are fused to a T7 late promoter can be expressed to high levels when introduced in such hosts (2, 3). In spite of this practical bearing, relatively little is known about the in vivo behaviour of these RNA polymerases. Indirect evidence suggests that the T7 RNA polymerase interacts very efficiently with its promoter in E. coli (4), whereas in vitro the interaction appears relatively weak (1,5). The enzyme has a very high turnover for the polymerisation of ribonucleotides, both in vitro (6) and in vivo (7). As a result, in E. coli it runs far ahead of the ribosomes which translate its nascent transcript, a situation

Eubacterial messenger RNAs are synthesized and translated simultaneously; moreover the speed of ribosomes usually matches that of RNA polymerase. We report here that when in Escherichia coli the host RNA polymerase is replaced by the eightfold faster bacteriophage T7 enzyme for the transcription of the lacZ gene, the beta-galactosidase yield per transcript is depressed 100-fold. But the overexpression of DEAD-box proteins greatly improves this low yield by stabilizing the corresponding transcripts. More generally, it stabilizes inefficiently translated E. coli mRNAs. Ribosome-free mRNA regions, such as those lying behind the fast T7 enzyme or between successive ribosomes on inefficiently translated transcripts, are often unstable and we propose that DEAD-box proteins protect them from endonucleases. These results pinpoint the importance of transcription-translation synchronization for mRNA stability, and reveal an undocumented property of DEAD-box RNA helicases. These proteins have been implicated in a variety of processes involving RNA but not mRNA stability.