It is widely accepted that translation of canonical prokaryotic mRNAs is achieved mainly through ribosomal recruitment signals present in the 5′-untranslated leader region, and molecular mechanisms governing their translational efficiency have been the subject of many reviews. In contrast, the mechanism(s) leading to translation of leaderless mRNA, in which the start codon is either preceded by only a few nucleotides or which starts directly with a 5′-terminal AUG, has remained elusive. In this review, we summarize the current knowledge on translation of leaderless mRNAs in Escherichia coli and discuss their possible biological implication.
Ribosomal recruitment signals on leaderless mRNAs downstream of the start codon: do they exist?
The finding of a Watson and Crick complementarity between the initiation–proximal coding region (downstream box, DB) of several phage and bacterial mRNAs and bases 1469–1483 within helix 44 of 16S rRNA (anti-DB, aDB) was at the origin of the hypothesis that, similarly to the Shine–Dalgarno (SD)–anti-SD interaction, a DB–aDB pairing could enhance the translational efficiency (Sprengart et al., 1990). Soon afterwards, it was suggested that the DB could compensate for the absence of a SD sequence in leaderless mRNAs (Shean and Gottesman, 1992). However, evidence in support of the DB–aDB interaction from biochemistry, mRNA–rRNA co-variation and rRNA mutagenesis is invariably lacking. Chemical probing studies have failed to show protection of the putative DB of the leaderless λcI mRNA bound in 70S initiation complexes (Resch et al., 1996). Moreover, the 3D crystal structure of the Thermus thermophilus 30S subunit (Wimberley et al., 2000) revealed that the whole shoulder of the body of the ribosomal particle is situated between the putative DB of the mRNA and that part of helix 44 of 16S rRNA comprising the proposed aDB (Moll et al., 2001). This renders any model in which the start codon is placed in the ribosomal P-site whereas the adjacent DB interacts with the aDB (Sprengart and Porter, 1997) or in which the DB basepairs synergistically with the SD–anti-SD interaction (Sprengart et al., 1996) untenable. Likewise, there is no experimental support for the speculation that a DB–aDB basepairing could contribute to the initial 30S–mRNA interaction before translation initiation complex formation (Sprengart and Porter, 1997). Neither natural λcI mRNA nor a derivative of this leaderless mRNA with an optimized basepairing potential with the aDB formed a binary complex with ribosomes (Moll et al., 2001). This result was consistent with previous kinetic toeprint experiments (Resch et al., 1996), as well as with the findings that the aDB is not accessible to a complementary DNA oligonucleotide, and that the aDB in situ is in a double-stranded conformation not amenable to basepairing (La Teana et al., 2000).
The occurrence of leaderless mRNAs in all domains of life suggested that the ability to translate this class of mRNAs is evolutionarily conserved (Janssen, 1993), and several studies demonstrated that leaderless mRNAs of different origin are faithfully translated in heterologous bacterial systems as well as in both archaeal and eukaryal in vitro translation systems (Wu and Janssen, 1996; 1997; Tedin et al., 1997; Grill et al., 2000; Moll et al., 2001). Helix 44 of 16S rRNA is a phylogenetically conserved element in ribosomes, but the primary sequence of the rRNA region corresponding to the putative E. coli aDB is variable in bacteria and archaea. In spite of this, leaderless mRNA translation does not display any species- or even kingdom-specificity. Furthermore, O´Connor et al. (1999) have reversed all 12 bp of the stem containing the putative aDB. The expression of this 16S rRNA aDB-flip mutant in an E. coli strain carrying deletions in all of the seven rrn operons had no effect on the translational efficiency of mRNAs with a putative DB. Taken together, there is neither biochemical nor genetic evidence supporting the proposed role of the DB–aDB interaction in ribosomal recruitment of leaderless mRNA initiation sites.
Nonetheless, several in vitro studies suggested that the nature of the coding region immediately downstream of the start codon can profoundly affect the translational efficiency of a given leaderless mRNA. This effect is not because of the presence of the DB sequence, but seems to be related to secondary structures in the mRNA region (P. Sairafi and U. Bläsi, unpublished). Some support for this notion comes from studies of Martin-Farmer and Janssen (1999) who showed that CA repeats down-stream of the start codon stimulated in vivo translation of leaderless and leadered mRNAs. These authors considered that the CA-rich sequence might provide for a lack of structure.
Translation initiation with 70S monosomes: an alternative pathway?
It has been reported that 70S ribosomal monomers have a preference over 30S subunits for in vitro formation of an initiation complex with the 5′-terminal AUG of leaderless phage λcI mRNA (Balakin et al., 1992). In the same article, the authors have discussed the possibility that translation initiation of leaderless mRNAs may be accomplished by 70S ribosomes, and suggested that 70S initiation could confer a competitive advantage to leaderless mRNAs under adverse physiological conditions, which would suppress the translation of canonical mRNAs. In vitro studies performed in our laboratory have demonstrated that 70S monosomes, in spite of the accessibility of their anti-SD sequence, are not capable of forming translation initiation complexes at internal translation initiation regions (TIRs). However, the same monosomes displayed a preference for 5′-terminal start codons of several natural leaderless mRNAs as well as of an artificial mRNA bearing both a 5′-AUG and an internal canonical initiation site (Moll, 2000). The ability of 70S ribosomes to form initiation complexes at 5′-terminal start codons, but not at internal initiation sites, might be explained by topological constraints of the mRNA track that is situated at the subunit interface. In light of the crystal structure of the ribosome, it appears that entry of the mRNA into the template channel of the 70S monomers requires its 5′-end to be pulled through the channel between both subunits. Thus, although it seems conceivable that a 5′-terminal AUG could find its way through the tunnel until it is lodged in the ribosomal P-site, the same seems to be less probable for an internal AUG. Some texture in favour of this idea comes from studies showing that the 5′-proximal position of the start codon is important for the translational efficiency of leaderless mRNAs (Jones et al., 1992).
Several studies have revealed that in the presence of initiator-tRNA, 70S ribosomes have an approximately 10-fold higher affinity for 5′-AUGs than 30S subunits (Moll, 2000). An explanation for this finding could be the greater stability of 70S initiation complexes compared with 30S initiation complexes with 85% of the initiator-tRNA being in contact with the 70S ribosome (Hüttenhofer and Noller, 1992). Nevertheless, it should be noted that although there seem to be no means to obtain an unambiguous in vivo confirmation for this premise, all available evidence in favour of the 70S initiation pathway for leaderless mRNA translation is rather indirect and comes exclusively from in vitro studies. Perhaps only a kinetic analysis of leaderless initiation complex formation and/or the use of mutant ribosomes forming particularly tight 70S couples could help to test whether this alternative translation initiation pathway really exists, or if dissociation of ribosomes is a mandatory step also for translation initiation of leaderless mRNAs.
Leaderless mRNA recognition by a 30S-IF2-initiator-tRNA complex
In textbooks, translation initiation in bacteria is usually shown to proceed via an 30S-mRNA intermediate (Fig. 1A) before the 30S initiation complex is formed. An increase of the concentration of IF2, a factor known for its capacity to stimulate fMet-tRNAfMet (initiator-tRNA) binding to the ribosomal P-site (Gualerzi et al., 2000), was found to enhance selectively λcI mRNA translation in vivo and in vitro. This increased efficiency of leaderless mRNA translation was also reflected in an increased capacity of this mRNA to compete with a canonical mRNA for ribosomes (Grill et al., 2000). As λcI mRNA did not form a binary complex with 30S subunits, it has been suggested that ribosomal recognition of leaderless mRNAs is accomplished by a 30S-IF2-initiator-tRNA complex (Fig. 1B), an intermediate equivalent to that formed during translation initiation in eukaryotes (Fig. 1C). This 30S-initiator-tRNA pathway, which apparently does not require any additional rRNA–mRNA contacts, has, hitherto, remained unappreciated in the textbook view of translation in prokaryotes. It should be noted, however, that the existence of the 30S-initiator-tRNA pathway in vivo, even for prokaryotic mRNAs with a consensus ribosome binding site, has also been suggested by the use of mutant tRNAs and specialized ribosomes (Wu and RajBhandary, 1996).
The consensus ribosomal recruitment signals are mechanistically different in pro- and eukaryotes. Whereas ribosomal recruitment to canonical prokaryotic mRNAs seems to rely primarily on rRNA–mRNA interactions, binding of eukaryotic ribosomes to mRNA is mainly mediated by protein–protein interactions, regardless of whether it is cap-dependent, or whether ribosomes are recruited to internal ribosome entry sites (Pestova et al., 2001). Nevertheless, it has been reported that an eukaryotic leaderless mRNA is translated in the absence of both a 5′ cap structure and a consensus sequence around the start codon (Hughes and Andrews, 1997). In addition, the faithful translation of the leaderless λcI mRNA in both an archaeal and a reticulocyte lysate in vitro translation system (Grill et al., 2000; Moll et al., 2001) lends support to the hypothesis that a ribosome-initiator tRNA complex is sufficient for recognition of leaderless mRNAs in all kingdoms.
In the ribosome-initiator-tRNA model of translation initiation, the 5′-terminal AUG is the only necessary and constant element of a leaderless mRNA which is recognized by the translational machinery. Support for this premise comes from studies which have demonstrated that the translational efficiency of mRNAs deprived of their 5′-UTR (including the SD-sequence) depended exclusively on the presence of a 5′-AUG start codon (van Etten and Janssen, 1998). In fact, substitution of the 5′-AUG with the otherwise canonical initiation triplets GUG or UUG resulted in a complete loss or at least in a dramatic decrease of translation. As the ribosome-initiator-tRNA model relies entirely on basepairing between the anticodon of initiator-tRNA and the initiation triplet of leaderless mRNA, the above observation could be explained by an intrinsic instability of ternary initiation complexes built in the absence of ribosome-mRNA contacts at the 5′-side of a wobbling codon–anticodon interaction. This explanation agrees with a recently published study wherein the translational efficiency of a λcI–lacZ leaderless reporter construct begining with AUG, CUG, GUG or UUG was determined (O´Donnel and Janssen, 2001). Only the construct with the AUG start codon supported a high level expression, whereas a low level of expression was obtained with GUG, and no expression at all with the constructs beginning with UUG or CUG. Nevertheless, the finding that after conversion of its AUG start codon to an amber codon, a leaderless λcI–lacZ construct was not expressed even though an amber suppressor initiator-tRNA was co-expressed, has been interpreted as showing that it is not the codon–anticodon–complementarity but, instead, the AUG codon per se which is required for translation initiation of leaderless mRNAs (van Etten and Janssen, 1998). The lack of expression of the λcI–lacZ mRNA starting with the amber codon is not incompatible with our explanation. In fact, whereas the amber suppressing tRNA is a reasonable substrate for glutaminyl-tRNA synthetase (Varshney et al., 1991), formylation is substantially reduced when the initiator tRNA is charged with glutamine (Lee et al., 1991). As IF2 does not display any affinity for non-formylated initiator-tRNA (Gualerzi et al., 2000), it seems reasonable to suggest that the reduced concentration of formylated amber-suppressor initiator-tRNA may account for the findings of van Etten and Janssen (1998). Finally, the stimulation of leaderless mRNA translation seems to be exerted by IF2 not only through the promotion of 30S-initiator-tRNA complex formation, but also by conferring an increased stability to the initiation complex (Grill et al., 2001). Most probably, this effect of IF2 is mediated through interactions of the factor with the ribosome and/or with fMet-tRNAfMet on the ribosome (Gualerzi et al., 2000).
Modulation of leaderless mRNA translation
Translation initiation factor IF3 acts as a fidelity factor by promoting the dissociation of aminoacyl-tRNAs other than initiator fMet-tRNA from the ribosomal P-site, and by discriminating against non-canonical start codons, such as AUU (Gualerzi et al., 2000). Recent data (Petrelli et al., 2001) clearly indicate that this discrimination is not based on a direct ‘inspection’ of the codon–anti–codon interaction by IF3 (Hartz et al., 1990), but is indirectly caused by the effect of this factor on the conformational dynamics of the 30S ribosomal subunit. Previous studies have shown that IF3 is a negative effector of translation initiation complex formation on 5′-terminal start codons and that an artificial increase of the intracellular IF3 concentration inversely correlates with the translational efficiency of a leaderless reporter construct (Tedin et al., 1999). The IF3 discrimination against the initiation complexes formed at the 5′-AUGs is also difficult to reconcile with a direct contact–inspection mechanism and supports, instead, the ribosome-mediated mechanism of action of IF3, whereby the C-terminal domain of IF3 induces a conformational change around the cleft between the platform and the head of the 30S subunit which affects the P–site interaction of the tRNA (Sette et al., 1999; Petrelli et al., 2001). As mentioned above, we consider the codon–anticodon interaction at 5′-terminal start codons intrinsically un-stable and, therefore, unable to withstand the destabilization induced by IF3. This hypothesis is supported by studies showing that different E. coli infC mutants, defective in the IF3 ‘destabilization activity’, allowed for an enhanced translation of leaderless mRNAs (Tedin et al., 1999; Moll, 2000).
As expected from the described effects of both IF2 and IF3 on translation initiation of leaderless mRNAs, the IF2:IF3 molar ratio plays a crucial role in determining the translational efficiency of leaderless mRNAs (Grill et al., 2001). As an increase in the growth rate results in a transient increase of the ribosome: IF3 ratio, it is tempting to speculate that a relative IF3 deficiency occurring at high growth rates could favour an increased translational rate of leaderless mRNAs and a more efficient competition of these mRNAs with the canonical leadered mRNAs. A similar effect has also been observed by a transient elevation of the IF2 level (Grill et al., 2000; 2001), as it is known to occur when the E. coli cells are subjected to cold shock (Jones et al., 1987). Thus, it is conceivable that changes in environmental conditions and different types of stress can give rise to transient oscillations of the relative levels of initiation factors thereby modulating the translational efficiency of leaderless mRNAs in bacteria.
It has been shown that discrimination of 5′-terminal start codons by IF3 requires the presence of ribosomal protein S1 (Moll et al., 1998). Apparently protein S1, which is not required for in vitro 30S-initiation complex formation on leaderless mRNA (Tedin et al., 1997), contributes to the destabilizing effect exerted by IF3 during translation initiation. Protein S1 is bound to the ribosome via its N-terminus, whereas its elongated C-terminal domain ex-tends into the cytoplasm. In canonical translation initiation complexes, the C-terminus of S1 seems to contact the upstream region of canonical mRNAs thereby facilitating the formation of translation initiation complexes and increasing their stability (Boni et al., 1991). In contrast, in the case of leaderless mRNAs the protruding C-terminal domain of S1 apparently contributes to the IF3-dependent destabilization of translation initiation complexes formed at 5′-AUGs.
It has been reported previously that translation of the leaderless λcI mRNA is stimulated in an E. coli rpsB mutant deficient for ribosomal protein S2 (Shean and Gottesman, 1992). We have analysed this phenomenon at the molecular level by making use of an E. coli rpsBts mutant. The analysis of the ribosomes isolated under the non-permissive conditions revealed that, in addition to ribosomal protein S2, ribosomal protein S1 was absent (I. Moll and U. Bläsi, unpublished), which demonstrated that S2 is essential for binding of S1 to the 30S subunit. Therefore, the simplest explanation for the enhanced translation of a leaderless mRNA in the S2-deficient strain (Shean and Gottesman, 1992) and its selective translation at the non-permissive temperature in the rpsBts mutant (I. Moll and U. Bläsi, unpublished) is the dispensability of S1 for translation of transcripts with a 5′-AUG. A possibility, also in light of the fact that a functional homologue of S1 appears not to be at hand in all biological systems in which leaderless mRNAs are present, is that the ribosomal population might be heterogeneous with regard to the presence of S1, and that this may contribute to translational control of leaderless mRNAs.
Translation of leaderless mRNAs has been studied in vitro over a temperature range from 25°C to 42°C (Grill, 2000). Translation competition assays with E. coli extracts revealed that leaderless mRNAs can efficiently compete at low (25°C) temperature for ribosomes with an mRNA containing an internal canonical TIR, whereas at 42°C canonical mRNA translation completely surpassed that of leaderless mRNA. The inverse correlation of the translatability of the two classes of mRNA as a function of temperature has again been attributed to ribosomal protein S1, whose activity required for translation initiation at internal canonical TIRs seems to be cold-sensitive (Grill, 2000). Thus, downregulation of canonical mRNA translation caused by the reduced S1 function at low temperatures seems to account for an increased translation of leaderless mRNAs vis-a-vis a reduced competition by canonical mRNAs.
Leaderless mRNAs in different kingdoms: occurrence and biological implication
A fairly large number of bacterial genomes has been sequenced in the last decade yet only about 35–40 leaderless mRNAs have been identified with certainty in bacteria. However, considering that the transcriptional start sites are largely unknown for most open reading frames (ORFs) deduced from DNA sequences, the identification of leaderless transcripts represents a difficult task for bioinfomatics. Thus, it is easy to predict that the number of leaderless mRNAs will increase considerably in the future.
Leaderless mRNAs appear to be rather infrequent in Gram-negative bacteria. These include the λcI mRNA (Walz et al., 1976) and the Tn1721 tetR mRNA (Baumeister et al., 1991) derived from accessory genetic elements in E. coli, two leaderless mRNAs identified in Caulobacter crescentus (Winzeler and Shapiro, 1997), one mRNA of Thermus thermophilus (Sanchez et al., 2000) and several leaderless transcripts of Mycoplasma pneumoniae (Weiner et al., 2000). Their number in different Gram-positive genera such as Streptococci, Lactococci, Streptomyces and Corynebacterium (Janssen, 1993, and references therein) by far outnumbers that identified in Gram-negative species. The relative abundance of leaderless mRNAs in some eubacterial species seems to correlate with the absence of a functional homologue of ribosomal protein S1, but, clearly, it remains to be demonstrated whether this is of relevance for their translational control.
Judging from the complete genomes of some archaea, it appears that leaderless mRNAs are quite common in this kingdom. For instance, an analysis of 144 genes of the crenarchaeal Sulfolobus solfataricus revealed that genes encoded by monocistronic transcripts, as well as genes located at the 5′-proximal end of operons, lack a Shine–Dalgarno sequence and generally have little or no sequence upstream of the translational starts (Tolstrup et al., 2000). A likewise organization has been recently reported in the crenarchaeal hyperthermophile Pyrobaculum aerophilum. Transcript mapping of 10 unrelated genes and a whole genome computational analysis of the predicted translational start sites suggested that translation of single genes or genes that are first in operons proceeds mostly through leaderless transcripts, whereas genes internal to an operon are preceeded by a SD sequence (Slupska et al., 2001).
At this junction it seems reasonable to ask whether the leaderless mRNAs identified in bacteria and archaea encode functionally related genes. As the answer is no, we can at best attempt to group the known leaderless mRNAs into functional classes. At least three leaderless mRNAs of accessory genetic elements, two in E. coli (Walz et al., 1976; Baumeister et al., 1991) and one in Lactococcus lactis (Nauta et al., 1996), encode a regulatory protein. To ensure a latent state of these genetic elements, the corresponding repressors have to be continuously synthesized albeit at a low level. It might, therefore, be tempting to speculate that being leaderless provides a means for low level translation within a range of different physiological conditions. In different Streptomyces spp., nine leaderless mRNAs have been described that encode proteins which confer antibiotic resistance, and a handful encode functions that are related to secondary metabolism (Janssen, 1993; August et al., 1994). In the archaeon Halobacterium salinarium, three leaderless mRNAs (Janssen, 1993) specify functions involved in light-mediated ATP synthesis. Clearly, it is open to experiment whether this reflects in the different organisms particular physiological conditions that favour translation of leaderless mRNAs. On the other hand, the mapped leaderless transcripts of the crenarchaeal P. aerophilum (Slupska et al., 2000) encode genes with unrelated function. Again, whether this reflects pecularities of the translational apparatus of archaea, or perhaps that of crenarchaeota in particular, requires further analysis.
Although bacteria, archaea and eukarya translate mRNAs using a different set of factors, homology searches have revealed that translation factors from all kingdoms share a significant degree of homology leading to the conclusion that the rudiments of the present translational mechanisms and components, including the initiation factors, might have been present at the universal ancestor stage (Kyrpides and Woese, 1998a; 1998b). Among the conserved mechanisms, the translation initiation factor IF2 and initiator tRNA-dependent selection of the initiation site seems to be present in all translation systems. The fact that heterologous leaderless mRNA derived from a bacterium can be faithfully translated in both archaeal and eukaryal systems justifies the speculation that ‘today’s’ leaderless mRNAs might be remnants of ancestral mRNAs and, that in the universal ancestor cell, mRNA recognition may have been achieved by a ribosome–initiator tRNA complex, through a mechanism independent of additional ribosomal recruitment signals. The presence of eight leaderless transcripts in the small genome of human mitochondria (Janssen, 1993) supports this hypothesis.
The role of accessory genetic elements such as phage and transposons in horizontal gene transfer is widely acknowledged. Given that the best studied leaderless mRNAs in E. coli, λcI and tetR, are encoded by these genetic elements together with the universal translatability of these mRNAs in all kingdoms of life, it is finally tempting to suggest a possible role for leaderless mRNAs as mediators of horizontal gene transfer at the translational level.
Here, we have summarized recent insights into the translation of leaderless mRNAs in E. coli. These studies showed that the translational efficiency of leaderless mRNAs is, when compared with canonical bacterial mRNAs, not intrinsically determined but can be altered depending on the availability of components of the translational machinery. In contrast to Gram-negative bacteria, leaderless mRNAs seem to be more prevalent in Gram-positive bacteria and in archaea. It remains to be established whether translational control of leaderless mRNAs in these organisms is governed by the same mechanism(s) as in E. coli. If so, we need to know a good deal more about environmental conditions that affect the relative levels of initiation factors. Very little is currently known about the mechanism of translation initiation in archaea, some of which, like S. solfataricus, possess homologues of both eubacterial and eukaryal initiation factors. Studies on archaeal translation might not only elucidate the mechanistics of the process in this kingdom, they could also provide new insights on the role of the diverse initiation factors in the differential regulation of translation of leaderless and canonical mRNAs. In any case, the apparent abundance of leaderless transcripts in crenarchaeota, and the steadily increasing number of them found in eubacteria, indicate that leaderless mRNAs are not a rare phenomenon like in E. coli, and that further studies of this topic will be valuable for our understanding of the evolution of protein biosynthesis.