Misled by sequence complementarity: does the DB–anti-DB interaction withstand scientific scrutiny?

Sir,

The downstream box (DB) has been identified as a sequence element capable of enhancing translation of some bacterial and phage mRNAs and has been suggested to serve as an mRNA recognition element for the ribosome. The partial complementarity of this mRNA sequence to bases 1469–1483 in the penultimate stem of 16S rRNA first led Sprengart et al. (1990, Nucleic Acids Res18: 1719–1723) to propose that the DB enhanced translation by basepairing to the complementary sequence in rRNA (the anti-downstream box, anti-DB) in much the same way as the Shine–Dalgarno (SD) sequence basepairs to the 3′ end of 16S rRNA. Since the Sprengart proposal, several reports have appeared purporting to provide evidence for this interaction. The supporting evidence for the basepairing interaction consists entirely of manipulations of DB-containing mRNAs and the observations that, in general, increases in complementarity to 16S rRNA led to increased expression and decreases in rRNA–mRNA complementarity had corresponding downward effects on expression. However, biochemical evidence in support of this interaction is lacking. Here, we review the experiments performed in our and other laboratories that not only question the DB–anti-DB basepairing model, but ultimately lead us to conclude that the proposed DB–anti-DB interaction does not exist.

The proposed mRNA–rRNA interaction imposes considerable constraints on the mRNA path through the ribosome. In the light of our current knowledge concerning the location of the ribosomal P-site and the penultimate stem of 16S rRNA in the 30S subunit, it is unlikely that the initiation codon can be placed in the ribosomal P-site while the adjacent DB can interact simultaneously with bases 1469–1483. The putative anti-DB has been located in the body of the 30S subunit (McCarthy and Brimacombe, 1994, Trends Genet6: 78–85), one and a half helix turns away from the P-site. It is difficult to reconcile how the anti-DB region can be brought into alignment with the mRNA track as shown in a model proposed by Sprengart et al. (1996, EMBO J15: 665–674).

The proposed DB–anti-DB interaction is also not supported by chemical footprinting or cross-linking studies. Attempts to cross-link the anti-DB to mRNA have failed (R. Brimacombe, unpublished). Chemical protection studies on λcI mRNA −70S initiation complexes (Resch et al., 1996, EMBO J15: 4740–4748) and on phage T4 gene 32 mRNA −30S and −70S complexes (Hüttenhofer and Noller, 1994, EMBO J13: 3892–3901) have likewise failed to show protection of the putative DB. As expected, in one of these experiments, the SD region of the T4 gene 32 message was protected from chemical modification in a manner consistent with rRNA–mRNA basepairing.

The DB does not appear to increase the affinity of mRNA for 30S subunits. We have created a leaderless λcI mRNA that contained a perfect match with the putative anti-DB. Filter-binding studies with 30S ribosomes revealed that the retention rate obtained with this mRNA was indistinguishable from that obtained with cI wild-type mRNA (M. Huber and U. Bläsi, unpublished). In another approach, attempts have been made to anneal a DB oligonucleotide to 30S and 70S ribosomes prepared at 37°C as well as under cold shock conditions. In both cases, the labelled DB oligonucleotide failed to bind to the ribosomes (C. O. Gualerzi, unpublished). These experiments make it very unlikely that the DB can serve as a transient binding domain for the 30S subunits before the ternary complex is formed, as speculated recently by Sprengart and Porter (1997, Mol Microbiol24: 19–28).

The proposed anti-DB is located in a phylogenetically conserved helical element in the penultimate stem of 16S rRNA. One of the most disconcerting aspects of the basepairing model is that it requires extensive unwinding of this helix and, presumably, reformation of the helix after initiation. Shean and Gottesman (1992, Cell70: 513–522) have suggested an alternative basepairing of the 16S rRNA sequence complementary to the anti-DB in mutants deficient in ribosomal protein S2. This alternative basepairing would free the anti-DB for pairing with the DB on mRNA. However, the observations made by Shean and Gottesman (1992) cannot be explained by a lack of S2, as the protein has subsequently been shown to bind to a region different from that proposed in their model (Powers and Noller, 1995, RNA1: 194–209). Mutational studies indicate that alterations on either side of the anti-DB-containing helix that disrupt helical continuity have deleterious effects on ribosome function and that these effects can be reversed upon introduction of compensatory mutations that restored basepairing within the helix (Firpo and Dahlberg, 1998, Nucleic Acids Res26: 2156–2160). These studies clearly indicate that a stable helix rather than a particular primary sequence is important for ribosome function. More recently, one of our laboratories has reversed all 12 basepairs of the stem containing the putative anti-DB, thereby creating a mutant 16S rRNA with the same stability as wild type, but with a radically altered mRNA basepairing potential. This mutant 16S rRNA (anti-DB flip) has been expressed in an Escherichia coli strain in which all of the seven rrn operons have been deleted (Asai et al., 1999, Proc Natl Acad Sci USA96: 1971–76). The expression of several previously described, DB-containing reporter constructs, including λcI, rpoH, lysU, glnS and vph, was found to be indistinguishable in both the anti-DB flip mutant and in the isogenic wild-type strain, thus showing definitively that any DB-associated enhancer activity does not involve basepairing between the DB and proposed anti-DB (M. O'Connor, T. Asai, C. L. Squires and A. E. Dahlberg, unpublished).

Finally, it has been shown that the leaderless λcI mRNA is expressed with nearly the same efficiency in a B. stearothermophilus in vitro translation system as in E. coli (Tedin et al., 1997, Mol Microbiol25: 189–199) or in vivo in B. subtilis (I. Moll and U. Bläsi, unpublished), although the putative anti-DBs of the Gram-positive organisms and E. coli are completely dissimilar.

Depending on the mRNA under study, the DB–anti-DB has been proposed to be operational in the absence of a SD sequence (Shean and Gottesman, 1992), only in the presence of the SD (see accompanying MicroCorrespondence by Etchegaray and Inouye) as well as in the presence and absence of a SD (Sprengart et al., EMBO J15: 665–674). The data in support of the requirement for a DB are at variance with that of Resch et al. (1996), who showed that deletion of the DB in leaderless mRNAs had no effect on ribosome binding. Sprengart and Porter (1997) argued that the deletion of the putative DB in λcI mRNA might have created alternative DB stretches. However, the suggested alternative DB (see accompanying MicroCorrespondence by Etchegaray and Inouye) had been deleted in an earlier study and was shown to have no effect on expression (Resch et al., 1995, FEMS Microbiol Rev17: 151–157). In addition, Odjakova et al. (1998, Microbiol Res153: 173–178) performed a deletion study on E. coli cat mRNA and found no evidence for an involvement of the putative DB in translation either in the presence or in the absence of the SD sequence.

While the mutational studies on DB-containing mRNAs have been interpreted in terms of a DB–anti-DB basepairing model, other explanations are equally consistent with the observed effects on gene expression. These effects could derive from an altered codon usage (Faxen et al., 1991, Nucleic Acids Res19: 5247–5251), from an alteration of an inhibitory secondary structure (Nagai et al., 1991, Proc Natl Acad Sci USA88: 10515–10519) or from effects on mRNA or protein stability.

Taken together, the experimental data discussed above lead us to conclude that the DB–anti-DB interaction does not exist and that other explanations must be sought to explain the observed effects of the DB sequence on gene expression. Moreover, we would like to argue that, in the absence of supporting biochemical or genetic evidence, novel mRNA–16S rRNA interactions based primarily on mRNA–rRNA sequence complementarities should not be proposed.