A novel mechanism for translation initiation operates in haloarchaea

Authors


*E-mail soppa@bio.uni-frankfurt.de; Tel. (+49) 69 798 29564; Fax (+49) 69 798 29527.

Summary

Four different mechanisms for translation initiation are known, i.e. one prokaryotic mechanism involving a Shine–Dalgarno sequence, two eukaryotic mechanisms relying on ribosomal scanning or internal ribosomal entry sites, and one mechanism acting on leaderless transcripts. Recently it was reported that the majority of haloarchaeal transcripts is leaderless and that most leadered transcripts are devoid of a Shine–Dalgarno sequence, excluding the operation of a ‘bacterial-like’ initiation mechanism. Therefore, the current study concentrated on elucidating whether a ‘eukaryotic-like’ scanning mechanism might operate instead. GUG and UUG were efficiently used as start codons on leadered transcripts in vivo, in contrast to initiation on leaderless transcripts (and leadered eukaryotic transcripts). Deleted versions of the 5′-UTR initiated translation very inefficiently. Introduction of additional upstream AUGs did not influence the initiation efficiency at internal start codons. An additional in-frame AUG at the 5′-end led to the simultaneous usage of two start sites on the same message. A stable stem-loop structure at the 5′-end inhibited only initiation at the first AUG, but did not influence usage of the internal AUG. Taken together, operation of a scanning mechanism was excluded and the results indicate that a novel mechanism for translation initiation operates at least in haloarchaea.

Introduction

Translation of the open reading frames (ORFs) of mRNAs into proteins is a central step for gene expression. With very few exceptions (e.g. mitochondria) the genetic code is universally used, and ribosomal RNAs as well as ribosomal proteins are conserved in all three domains of life, i.e. archaea, bacteria and eukaryotes. The process of translation can be subdivided into the three phases of initiation, elongation and termination. In addition, recycling of ribosomes is possible at polycistronic transcripts in prokaryotes (‘translational coupling’) and at circular eukaryotic transcripts. Typically, initiation is rate-limiting, and therefore regulation of translational efficiencies most often occurs during the initiation phase. Initiation has the greatest evolutionary divergence of the translation steps, and the following four mechanisms for translation initiation are currently known.

(i) In bacteria, start codon recognition involves base-pairing of the so-called Shine–Dalgarno (SD) sequence in the mRNAs with the 3′ end of the 16S rRNA of the small ribosomal subunit. The SD sequence is optimally localized at a distance of 5–7 nt upstream of the start codon. After subsequent recruitment of the large ribosomal subunit, the start codon is placed in the P-site of the ribosome. Three initiation factors (IF1, IF2, IF3) are involved in different steps of the initiation process (Gualerzi and Pon, 1990; Marintchev and Wagner, 2004; Laursen et al., 2005). The direct recognition of the start codon via the SD sequence enables the existence of polycistronic transcripts containing several ORFs, which are common in bacteria.

(ii) In eukaryotes, translation initiation is more complex and a much larger number of initiation factors are involved (Pestova et al., 2001; Marintchev and Wagner, 2004; Lopez-Lastra et al., 2005). The factor eIF4E binds to the cap structure present at the 5′-end of eukaryotic transcripts. With the help of additional eIFs the small ribosomal subunit is recruited to the 5′-end of the transcript. Subsequently, it scans along the transcript until the first AUG codon is reached. The interaction of the anticodon in the initiator tRNA (bound by eIF2) with the start codon is involved in stopping the small ribosomal subunit at the start site. Subsequently the large subunit is recruited and the complete ribosome starts translation. The following results provide strong evidence for the linear scanning of the small subunit from the 5′-end to the first AUG: when natural or designed mutations lead to the introduction of an additional AUG upstream of the native start site, translation initiation is shifted to the new ‘first AUG’; when the native AUG is removed by natural or designed mutations, the first AUG 3′ of the native site is used for initiation; initiation can be inhibited by the introduction of stable stem-loop structures at the 5′-end or between the cap and the native start site (Kozak, 2002).

(iii) Recently it was discovered that a second mechanism operates in eukaryotes simultaneously, i.e. the usage of so-called ‘internal ribosomal entry sites’ (IRES). IRES are complex structures formed within the 5′-UTRs of eukaryotic transcripts. They are recognized by ‘IRES-binding transacting factors’ which are involved in recruitment of the small ribosomal subunit. IRES were first discovered to be essential for translation initiation on transcripts of picornaviruses. Initially they were thought to be confined to viral initiation and had evolved to allow the production of viral proteins while cap-dependent translation of host proteins had been inhibited by the destruction of initiation factor by viral proteases (Kean, 2003; Jackson, 2005; Martinez et al., 2008). However, ribosomal profiling of uninfected cells revealed that translation of a variety of transcripts persists under conditions when cap-dependent translation is shut down, e.g. under stress conditions or during apoptosis. Currently it is believed that translation on about 10% of all eukaryotic transcripts can be initiated via a cap-independent mechanism, probably involving IRES (Spriggs et al., 2005; 2008).

(iv) The fourth mechanism for translation initiation acts on leaderless transcripts that are devoid of a 5′-UTR. They occur in bacteria and in eukaryotes, but are thought to be very seldom exceptions. In Escherichia coli, the efficiency of initiation on leaderless transcripts is much lower than initiation via the ‘SD mechanism’ described above (Moll et al., 2002). Several lines of evidence prove that initiation on leaderless transcripts is distinct from the other mechanisms: initiation involves the undissociated ribosome and not the small subunit (O'Donnell and Janssen, 2002; Moll et al., 2004; Andreev et al., 2006); the initiator tRNA is required for complex formation between mRNA and ribosome; the antibiotic kasugamycin inhibits only canonical initiation and does not affect initiation on leaderless transcripts (Moll and Bläsi, 2002); variation of the intracellular ratio of the initiation factors IF2 and IF3 differentially influences efficiency initiation on leaderless versus leadered transcripts (Grill et al., 2000; 2001).

Compared with the other two domains, much less is known about translation initiation in archaea. Three studies with species of four distantly related archaeal genera indicate that leaderless transcripts are much more common in archaea than in bacteria and eukaryotes and that the leaderless mechanism might in fact represent the major pathway for initiation in archaea (Tolstrup et al., 2000; Slupska et al., 2001; Brenneis et al., 2007).

It was shown that initiation on leaderless mRNAs and the SD mechanism might operate simultaneously: in vivo and in vitro the mutation of a SD sequence resulted in an inhibition of translation initiation, while total removal of the 5′-UTR restored translation or even largely increased translational efficiency (Condo et al., 1999; Sartorius-Neef and Pfeifer, 2004). Analysis of 62 haloarchaeal transcripts revealed that about two-thirds were leaderless, while one-third contained 5′-UTRs of an average length of about 20 nt (Brenneis et al., 2007). Surprisingly, nearly all of the leadered transcripts were devoid of a SD sequence, and a bioinformatic analysis of the Halobacterium salinarum genome revealed the severe rareness of potential SD sequences upstream of haloarchaeal start codons. Using a reporter gene system it was verified for six natural and synthetic SD-less 5′-UTRs that they drive efficient translation in vivo (Brenneis et al., 2007). Archaea and eukaryotes share various features to the exclusion of the bacteria, e.g. in replication, DNA packaging, translation initiation and RNA modification (Sandman and Reeve, 2000; Soppa, 2001; Omer et al., 2003; Barry and Bell, 2006). Noteworthy, their genomes encode more than 10 translation initiation factors, highly outnumbering the three bacterial IFs (Londei, 2005). Therefore, it was tempting to speculate that a eukaryotic-like scanning mechanism could operate on leadered SD-less archaeal transcripts, in spite of the fact that archaea contain neither a cap structure nor the cap-binding protein eIF4E. As has been noted earlier, ‘it is a mistake to think that, because archaeal mRNAs lack a 5′ cap, translation in that system cannot occur via scanning’ (Kozak, 2002). It was also pointed out that ‘the essence of the scanning model is 5′ entry of ribosomes and position-dependent selection of the AUGSTART codon. Those key points hold with naturally uncapped mRNAs produced by some viruses’ (Kozak, 2002). Recently it was discovered that the archaeal initiation factor aIF2/5a binds to the 5′-end of transcripts in a triphosphate-dependent manner (Hasenöhrl et al., 2008). This observation would at least be consistent with the operation of a scanning mechanism for translation initiation.

Therefore, the present study concentrated on investigating whether or not a scanning mechanism is operating on leadered haloarchaeal transcripts devoid of a SD sequence. Surprisingly, evidence was found that excludes the operation of a scanning mechanism and thus shows that a fifth, novel mechanism of translation initiation operates on transcripts of that category at least in haloarchaea.

Results

Usage of alternative start codons on leadered and leaderless mRNAs

The first aim was to analyse whether start codon requirements are identical or different for translation initiation on leadered and on leaderless transcripts in Haloferax volcanii. The dhfr gene was chosen as a reporter gene for the in vivo analysis, because it had been used successfully in previous studies (Danner and Soppa, 1996; Brenneis et al., 2007). The native start codon AUG of the naturally leaderless transcript was changed to GUG and to UUG by in vitro mutagenesis. The translational efficiencies of the three transcripts in exponentially growing cultures were determined by quantifying the steady-state transcript levels using qRT-PCR and the resulting steady-state protein levels using an enzymatic assay. At least three biological replicates were performed for all experiments described in this contribution, and average values and standard deviations were calculated. The DHFR protein levels, dhfr transcript levels and translational efficiencies are summarized in Fig. 1A, and Fig. 1B gives an overview of the translational efficiencies normalized to the AUG-containing control transcript. As can be seen, only AUG served as a start codon and the two single-base mutations totally inhibited translation of the two mutant transcripts. Notably, the levels of all three transcripts were identical, showing that in this case the half-life of the transcript is not influenced by the frequency of ribosome binding.

Figure 1.

Start codon selectivities at leadered and leaderless haloarchaeal transcripts.
A and D. The transcripts are shown schematically and their start codons are listed. The specific DHFR activities, the dhfr transcript levels and the translational efficiencies in exponentially growing cells are tabulated.
B. The translational efficiencies of (A) are shown schematically after normalization to transcript 1.
C. The translational efficiencies of (D) are shown schematically after normalization to transcript 4.

In a previous study, six natural or synthetic 5′-UTRs, which were all devoid of a SD sequence, had been fused to the dhfr reporter transcript. All of them were translated in vivo. The efficiencies were lower or higher than that of the leaderless control, depending on the respective 5′-UTR (Brenneis et al., 2007). One construct including a synthetic SD-less 5′-UTR of 20 nt, which resulted in a high translational efficiency, was chosen (sequence of the 5′-UTR: TACCACATTTCAGGCAAGAT). The AUG start codon was changed to GUG and UUG, respectively, by in vitro mutagenesis. The resulting plasmids were used to transform H. volcanii, and the DHFR protein levels, dhfr transcript levels and translational efficiencies were determined using exponentially growing cultures. The results are summarized in Fig. 1C and D. Again, the levels of the three transcripts were identical. However, the GUG and UUG start codons led to a decrease of the protein amount to about 55% and 38%, respectively, compared with the native AUG start codon. Nevertheless, both alternative start codons were used for translation initiation, in contrast to their inability to drive initiation in leaderless transcripts. This discrepancy suggested that two distinct initiation mechanisms operate at leaderless and leadered transcripts in haloarchaes.

5′-UTR length and translation initiation efficiency

Four additional constructs were generated to determine the influence of the 5′-UTR length on translational efficiencies and to explore whether potential cis-acting elements could be discovered. The four constructs represent deletion variants of the 5′-UTR, which was shortened either at its 5′-end or at its 3′-end by 5 nt or by 10 nt respectively (compare overview in Fig. 2, No. 7–10). The translational efficiencies of the respective transcripts were compared with those of the leaderless and leadered control transcripts (No. 1 and 4). Unexpectedly, even a deletion of only 5 nt from either end resulted in a more than 10-fold reduction in translational efficiency (No. 7 and 9). The results for the deletion variants lacking 10 nt were very similar (No. 8 and 10). Therefore, at least in the context of this 5′-UTR more than 15 nt are needed for efficient translation initiation. The translational efficiencies of both transcripts with a 15 nt 5′-UTR and of one transcript with a 10 nt 5′-UTR were also significantly lower than that of the leaderless control transcript (No. 1). This indicates that a 15 nt 5′-UTR is already too long to efficiently use the pathway of leaderless transcripts. Taken together, the results underscore that initiation at leaderless and at leadered transcripts occurs via two different mechanisms.

Figure 2.

The influence of the length of the 5′-UTR on translational efficiency.
A. Four deletion versions of the 5′-UTR of the leadered transcript (construct No. 4) were generated and are shown schematically (constructs No. 7–10). The 5′-UTR has been schematically divided into four segments of 5 nt, labelled A to D, and the occurrence of these fragments in the 5′-UTRs of the mutated variants is indicated. 5′-UTR lengths, the specific DHFR activities, the dhfr transcript levels and the translational efficiencies in exponentially growing cells are tabulated.
B. The translational efficiencies of (A) are shown schematically after normalization to transcript 1.

Inhibition of translation initiation using a selected stable aptamer

Recently a tetracycline-binding aptamer had been selected that inhibited translation initiation in Saccharomyces cerevisiae in vivo in a tetracycline-dependent manner (Hanson et al., 2003). To test the applicability of the tetracycline-binding aptamer for inhibition of translation in H. volcanii in vivo it was fused to the leaderless dhfr transcript. The translational efficiencies of the resulting leadered transcript and the leaderless control transcript were determined in exponentially growing cultures in the absence and presence of tetracycline, and the results are summarized in Fig. 3A and B. In contrast to the results obtained in low salt solutions in vitro and in S. cerevisiae in vivo, no effect of tetracycline on the performance of the aptamer was detectable. However, the aptamer totally inhibited translation, strongly indicating that it forms a stable structure in the high-salt cytoplasm of H. volcanii even in the absence of tetracycline. In any case, it prevented ribosomes from binding the close-by start codon.

Figure 3.

Selective inhibition of translation initiation by a stable tetracycline-binding aptamer integrated near an internal start site or at the 5′-end.
A and D. The transcripts are shown schematically and the absence and presence of tetracycline are indicated. The specific DHFR activities, the dhfr transcript levels and the translational efficiencies in exponentially growing cells are tabulated.
B. The translational efficiencies of (A) are shown schematically after normalization to transcript 1.
C. The translational efficiencies of (D) are shown schematically after normalization to transcript 4.

After the inhibitory effect of the aptamer in H. volcanii was proven, it was fused to the leadered transcript at two different positions, i.e. (i) at the 3′-end of the leader directly upstream of the start codon (construct 12), and (ii) at the 5′-end of the leader, resulting in a 20 nt spacer between aptamer and start codon (construct 13). The translational efficiencies of the two aptamer-containing transcripts were compared with that of the leadered control transcript using exponentially growing cells, both in the absence and in the presence of tetracycline. The results are summarized in Fig. 3C and D. Again, the results were identical irrespective of the addition of tetracycline. In all cases the transcript levels were identical. Translation was totally inhibited when the aptamer was positioned directly upstream of the start codon. However, when the aptamer was positioned at the 5′-end of the transcript, translation was initiated as efficiently as on the control transcript lacking the aptamer. Taken together, the results shown in Fig. 3A–D strongly suggest that the first step of initiation on leadered transcripts is not the binding of the ribosome to the 5′-end of the transcript followed by scanning to the internal start codon and the results indicate that an internal recognition mechanism exists for start codons of leadered transcripts.

Translation initiation on transcripts with additional in-frame and out-of-frame start codons

A typical feature of the eukaryotic scanning mechanism is that the introduction of an additional start codon upstream of the native start codon shifts the initiation site and inhibits initiation at the native site (Kozak, 2002). Therefore, it was tested whether this is also true for the initiation on leadered haloarchaeal transcripts. First, 18 nt were added between the start codon and the second codon of the dhfr reporter gene. The extra sequence encoded a tetrahistidine tag in the +1 reading frame, compared with the dhfr reading frame. The addition decreased the DHFR-specific activity in exponentially growing cells by a factor of about 10, most probably because the second amino acid was changed and the stability of the fusion protein was reduced (compare No. 14 in Fig. 4A with No. 4 in Fig. 1D). The next step was the addition of an in-frame start codon at the 5′-end of the transcript (construct No. 15 in Fig. 4), thereby generating a transcript that has a leaderless start codon and an internal start codon simultaneously. The alteration increased the total specific DHFR activity by more than threefold (Fig. 4B). A Western blot analysis revealed that the transcript was translated into two different proteins, i.e. the additional AUG at the 5′-end also served as a start codon and led to the simultaneous generation of a larger fusion protein (Fig. 4C). A densidometric quantification of the signals revealed that the efficiency of translation initiation at the internal start codon (band marked with an arrow in Fig. 4C) was not reduced by the presence of the additional start codon (normalized to the transcript level: 100% for construct 14 and 130% construct 15).

Figure 4.

The effects of additional in-frame and out-of-frame start codons.
A. The transcripts are shown schematically. The sequence added between the first and second dhfr codon (encoding a tetrahistidine tag in the +1 frame) is indicated as a hatched box. The second, internal start codon of transcript 15 is indicated by a vertical line. The peptides encoded in the +1 frame are indicated by rectangles shaded in grey. The specific DHFR activities, the dhfr transcript levels and the translational efficiencies in exponentially growing cells are tabulated.
B. The translational efficiencies are shown schematically after normalization to transcript 14.
C. A Western blot analysis of the proteins translated from transcripts of plasmids 14, 15, and a control plasmid (c) lacking the dhfr. The size of the translation product initiated at the internal AUG is indicated by an arrow.

Next, it was tested whether the introduction of an additional out-of-frame start codon upstream of the internal start codon would influence usage of the latter site. Initiation at the added AUG would lead to the production of a protein of 73 amino acids including a tetrahistidine tag (construct No. 16 in Fig. 4). Quantification of DHFR protein and dhfr transcript levels in exponentially growing cultures revealed that the efficiency of initiation at the internal start codon was not influenced by the addition of an upstream out-of-frame start codon. The 73-amino-acid protein encoded by the additional ORF of construct 12 could not be detected by a Western blot analysis with a commercial anti-tetrahistidine antibody and thus this AUG does not seem to be used as a start codon (data not shown). In a control experiment the 73-amino-acid protein was produced in E. coli and it was verified that it could be detected with the anti-tetrahistidine antibody (data not shown). It should be noted that despite this control it cannot be totally excluded that the added AUG is in fact used as a start codon, the 73-amino-acid protein is produced, but is unstable and can therefore not be detected by Western blot analysis. However, even if this unlikely explanation would be true, it would be an as strong argument against the operation of a scanning mechanism for initiation, because the efficiency of initiation at the native AUG (a ‘downstream AUG’ in the mutant) is not reduced in construct 16 compared with 14.

Taken together, the lack of influence of the addition of in-frame and out-of-frame start codons on the efficiency of initiation at the native start codon provides additional proof that a eukaryotic-like scanning mechanism does not operate in haloarchaea.

Differential inhibition of initiation on a transcript with two start sites

In a further approach it was attempted to inhibit translation initiation specifically on one of two initiation sites, which were simultaneously present on one transcript. As a control, a construct was generated that contained the native dhfr ORF without an extension as well as the leader with an added in-frame AUG at its 5′-end (No. 17 in Fig. 5). Again, both start sites were recognized and two proteins of different lengths were synthesized. Next, the 5′-end was extended by a putative stem-loop, which was predicted to be present in the 5′-UTR of a H. volcanii transcript (gene HVO_0721). It had reduced translational efficiency embedded in its native 5′-UTR (Brenneis et al., 2007); however, it had little or no influence on translation initiation at both start sites (No. 18, Fig. 5). Therefore, the stem-loop was stabilized and the overhanging ‘A’ at the 5′-end was integrated into the stem (No. 19, Fig. 5C). The stabilized stem-loop totally inhibited translation initiation on the AUG at the 5′-end of the transcript, while initiation on the internal start site was hardly effected (Fig. 5B). These data underscore the results presented above and exclude that ribosome binding at the 5′-end of the transcript followed by scanning is involved in initiation at the internal start codon.

Figure 5.

Selective inhibition of initiation at one of two start sites which are simultaneously present on one transcript.
A. The transcripts are shown schematically and the presence and nature of a stem-loop are indicated. The specific DHFR activities, the dhfr transcript levels and the translational efficiencies in exponentially growing cells are tabulated.
B. Western blot analysis of the DHFR amounts translated from the transcripts 17, 18 and 19. The size of the translation product initiated at the internal AUG is indicated by an arrow.
C. In silico predicted structures of the 5′-UTRs of transcripts 18 and 19.

Discussion

Usage of alternative start codons on leadered and leaderless mRNAs

Determination of start codon specificity revealed that AUG, GUG and UUG all serve as start codons in the context of leadered transcripts in H. volcanii in vivo, resulting in normalized translational efficiencies of 100%, 58% and 38% respectively (constructs 4–6). Start codon selectivity on leadered transcripts has also been measured in E. coli using different reporter gene systems (Van Etten and Janssen, 1998; O'Donnell and Janssen, 2002). A quantitative comparison is not possible because the transcript levels were not quantified and therefore translational efficiencies could not be calculated. Northern blot and primer extension analysis revealed that the levels of the AUG-, GUG- and UUG-containing transcripts varied considerably and thus a comparison of the reporter enzyme activities is not very meaningful. Nevertheless, qualitatively the results were comparable, i.e. all three codons were used as start codons in E. coli in vivo in the order AUG > GUG > UUG. The requirement for AUG as a start codon is much stricter in mammalian cells in vivo. In COS cells it was found that even in the most favourable context for initiation the second best codon, GUG, resulted only in 3–5% of the protein level compared with AUG, while five other codons did not lead to the production of measurable amounts of protein (Kozak, 1989). Also in CV1 monkey cells AUG was a much better start codon than all nine codons that can be derived from AUG by single-base mutations. The order of efficiency was somewhat different from that in the COS cell study. ACG was the second best codon (leading to about 10% of the protein level) and GUG did not lead to measurable amounts of protein (Peabody, 1989). Nevertheless, both studies revealed that the requirement for AUG as a start codon on leadered transcripts is much higher in eukaryotes than in bacteria and archaea.

In stark contrast to leadered transcripts, solely AUG was found to act as a start codon on leaderless transcripts in H. volcanii in vivo, and single-point mutations to GUG or UUG abolished translation initiation completely (constructs 1–3). The same was found to be true for translation initiation on leaderless transcripts in E. coli. Furthermore, for E. coli it was revealed that codon–anticodon base-pairing is not sufficient for translation initiation on leaderless transcripts, but that the presence of an AUG start codon is required (Van Etten and Janssen, 1998). Start codon selectivity on leaderless transcripts has not yet been studied in eukaryotes.

Both in bacteria and in eukaryotes it was shown that initiation on leaderless transcripts requires the presence of complete, undissociated ribosomes and the initiator tRNA, in contrast to all other mechanisms, which rely on the initial interaction of the transcripts with the small subunit of the ribosome (Moll et al., 2004; Udagawa et al., 2004; Andreev et al., 2006). Therefore, unravelling of the divergent start codon specificities on leadered and leaderless haloarchaeal transcripts is in congruence with the results obtained with other species.

5′-UTR length and translation initiation

The results obtained with four deletion variants of the 20 nt leader underscored that translation on leadered and leaderless transcripts is initiated via two different mechanisms (constructs 1, 4, 7–10). Two transcripts with 15 nt leaders were translated very inefficiently, indicating that they could use neither of the two pathways. The analysis of 62 transcripts from H. salinarum and H. volcanii had revealed that only one of them had a 5′-UTR with a length between 5 and 14 nt (and might be regarded exceptional), while 61 5′-UTRs were either longer or shorter (Brenneis et al., 2007). Six native transcripts had 5′-UTR lengths below 20 nt. Therefore, the lower limit of 5′-UTRs to direct a transcript to the pathway for leadered transcripts seems to be between 14 and 20 nt and may be influenced by its sequence or structure. The upper limit for a 5′-UTR to allow usage of the leaderless pathway seems to be around 5 nt, because 40 haloarchaeal transcripts with 0–5 nt 5′-UTRs had been characterized, five of them with a 5 nt 5′-UTR (Brenneis et al., 2007). In exceptional cases also transcripts with a slightly larger 5′-UTR might be able to use the leaderless pathway, e.g. the arcA transcript of H. salinarum with a 7 nt 5′-UTR and possibly construct No. 10 with a 10 nt 5′-UTR.

These results are in congruence with a variety of studies that have shown that in bacteria the mechanism for initiation on leaderless transcripts is very different from the mechanism operating on leadered transcripts with a SD sequence, and transcripts use either one or the other of the two pathways (Van Etten and Janssen, 1998; Grill et al., 2000; 2001; 2002; Moll and Bläsi, 2002; Moll et al., 2002; 2004; O'Donnel and Janssen, 2002; Udagawa et al., 2004; Andreev et al., 2006; Schluenzen et al., 2006).

Translation on SD-less leadered haloarchaeal transcripts is initiated via a novel mechanism

The characteristic features of the eukaryotic scanning mechanism have been summarized and discussed very thoroughly by Kozak (2002). An important point is the position effect, i.e. utilization of the first AUG of the transcripts. Many examples are given of natural or designed mutations that create an AUG upstream of the native AUG and result in a shift of start site utilization, including human mutations leading to diseases (e.g. Liu et al., 1999; Kozak, 2002). The first AUG codon was shown to be the exclusive site of initiation even when the second AUG was positioned just a few bases downstream from and in the same optimal context as the first (Kozak, 1995). In addition, 35 examples of vertebrate transcripts are given (a ‘Partial list’) in which the second, natively downstream AUG became the start site after the original start-AUG had been mutated (table 2 in Kozak, 2002). A further important point is that stem-loop structures at the 5′-end or within the 5′-UTR inhibit translation initiation by blocking ribosome recruitment or scanning. It was noted that ‘a stem-and-loop structure . . . is most inhibitory when its proximity to the 5′-end blocks ribosome binding’ (Kozak, 2002).

Using exactly these defining points for the scanning mechanism of translation initiation we designed several experiments to clarify whether such a mechanism operates on leadered SD-less haloarchaeal transcripts. The following results clearly excluded that a scanning mechanism is used:

  • (i) A stable aptamer inhibited translation only when localized directly adjacent to the start codon, but did not influence the efficiency of initiation when situated at the 5′-end of the transcript (construct 13).
  • (ii) The introduction of in-frame and out-of-frame upstream AUGs did not influence translation initiation at the native start site. Instead, an additional AUG at the 5′-end of the transcript led to the simultaneous production of a second, larger protein initiated at the leaderless start site (constructs 15, 16).
  • (iii) A stem-loop at the 5′-end of a transcript with two start sites inhibited only initiation at the 5′-end, but did not influence initiation at the internal AUG on the same molecule (construct 19).

The currently known mechanisms of translation initiation are summarized in Table 1. The mechanism for initiation on leadered SD-less haloarchaeal transcripts deviates from all four previously known pathways and thus constitutes a novel mechanism. Therefore, haloarchaea use three different mechanisms simultaneously, the largest fraction of transcripts is leaderless, about one-third of the transcripts have a 5′-leader devoid of a SD sequence and can be assumed to use the novel pathway, and a very small fraction of transcripts include SD sequences. A mutational study revealed that the SD sequence preceding the start codon of a gas vesicle gene is essential for translation initiation and it is thus functional in vivo (Sartorius-Neef and Pfeifer, 2004). Therefore, three different mechanisms were shown experimentally to operate in haloarchaea in vivo. As mentioned in Introduction, haloarchaea are not the only organisms that use more than one mechanism for translation initiation simultaneously, e.g. eukaryotes use IRES-dependent initiation in addition to the scanning mechanism. In addition, leaderless transcripts can be translated by eukaryotic in vitro translation systems.

Table 1.  Comparison of translation initiation mechanisms.
No.MechanismDomainCharacteristic featuresReference (review)
1Shine–DalgarnoBacteria
Archaea
Base-pairing of SD sequence with 3′-end of 16S rRNA; fixed distance between SD and start codon; polycistronic mRNAsLaursen et al. (2005)
2ScanningEukaryotes5′-cap recognition by eIFs and recruitment of 40S subunit; linear scanning of 5′-UTR until the first AUG is reachedKozak (2002)
3IRESEukaryotesComplex RNA structure within the 5′-UTR; recognition by ITAFs that recruit 40S subunitLopez-Lastra et al. (2005)
4LeaderlessArchaea
(Bacteria, Eukaryotes)
Complete ribosome + initiator tRNA bind the 5′-AUG; restricted to AUG; differential IF-dependence + antibiotic sensitivityLondei (2005)
5NovelArchaeaSD and IRES are absent from 5′-leader; not inhibited by mutations that block scanning (upstream AUG, aptamer, stem-loop); efficiency depends on sequence of 5′-UTRThis study

The molecular details of the novel mechanism remain to be uncovered. The minimal length of a 5′-UTR is about 15 nt, indicating that the ribosome (or an initiation factor) makes contact to this region. On the one hand the sequence and/or structure of the 5′-UTR has a great influence, since the translational efficiency of six otherwise identical transcripts, which differed in their 5′-UTRs, varied by more than 10-fold (Brenneis et al., 2007). On the other hand it seems clear that base-pairing of a cis-acting element is not necessary for efficient initiation, in contrast to the SD mechanism, since the comparison of many native SD-less haloarchaeal 5′-UTRs has not led to the identification of conserved motifs. Therefore, sequence and structural requirements of this region, which result in efficient initiation, will be determined experimentally using a randomized library, using the same principle that was applied to a promoter analysis earlier (Danner and Soppa, 1996).

It seems that alternatives to the SD sequence-dependent mechanism are much more widespread in bacteria than currently anticipated. A bioinformatic analysis of many genomes revealed that only 50% of all bacterial genes are preceded by a SD sequence. The fraction of SD-led genes is very diverse in different bacterial groups and ranges from about 15% in Bacteroides to more than 90% in Firmicutes (Chang et al., 2006). Most species have fractions of SD-led genes below 70% and thus are predicted to contain considerable fractions of transcripts on which translation is not initiated the ‘bacterial way’, but by a simultaneously operating second mechanism (e.g. the ‘leaderless mechanism’ or the ‘novel mechanism’).

Taken together, it has been recently discovered that translation initiation is not confined to two major mechanisms, i.e. scanning in eukaryotes and the SD sequence-dependent mechanism in prokaryotes, and very few additional exceptions. In contrast, eukaryotes have been shown to apply IRES-dependent initiation regularly and initiation on leaderless transcripts seems to be the major pathway in archaea. The novel pathway described above adds to this biodiversity. Future work will concentrate on elucidating the molecular mechanism of this novel pathway. It will be interesting to reveal which of the more than 10 translation initiation factors are essential for the different pathways and to uncover the sequence or structure requirements of the transcripts which could not be deduced from native and synthetic transcripts thus far. And, last but not least, it will be interesting to reveal how widespread SD-less leadered transcripts are in other archaea and in bacteria.

Experimental procedures

Microorganisms, media and growth conditions

Haloferax volcanii WR 340 was obtained from Moshe Mevarech (Tel Aviv University, Tel Aviv, Israel) and E. coli XL1 Blue MRF′ was purchased from Stratagene (Amsterdam, Netherlands). H. volcanii was grown aerobically in complex medium containing 2.9 M NaCl, 150 mM MgSO4, 60 mM KCl, 4 mM CaCl2, 0.275% (w/v) yeast extract, 0.45% (w/v) tryptone and 50 mM Tris-HCl, pH 7.2, at 42°C (Cline et al., 1989). E. coli XL1 Blue MRF′ was grown in SOB Medium at 37°C (Sambrook et al., 1989).

Construction of plasmids

All plasmids used in this study are listed in Table S1 and the oligonucleotides used for plasmid construction are summarized in Table S2. Three approaches were used for plasmid construction. (i) For the introduction of point mutations the parent plasmid (Table S1) was cut with the restriction enzymes XhoI and KpnI, generating a fragment containing the dhfr reporter gene, the 5′-UTR (if applicable) and the promoter region. The fragment was cloned into the vector pSKII+ (Stratagene, Amsterdam, the Netherlands). The mutations were introduced using the ‘site-directed mutagenesis kit’ (Stratagene, Amsterdam, the Netherlands) and the primers included in Table S1. After mutagenesis, the XhoI–KpnI fragment was reisolated and cloned into the shuttle vector pSD1/M2-18 (Danner and Soppa, 1996). (ii) For the construction of each of the four 5′-UTR deletion variants (constructs 7–10) two overlapping PCR fragments were generated using the plasmid pMB1 as template. The primers for the first PCR product were pSD1_Prom_for and OHX_rev (with X = 18–21), for the second PCR product the primers OHX_for and the primer pSD1_rev were used. The overlapping PCR fragments were purified and fused via a subsequent PCR reaction using the primers pSD1_Prom_for and pSD1_rev. The resulting fragment was purified, cut with XhoI and KpnI, and cloned into the plasmid pSD1/M2-18 cut with the same enzymes. (iii) For the introduction of stem-loop structures and the tetrahistidine tag two different PCR fragments were generated, which represented the regions upstream and downstream of the site of integration. Both PCR fragments contained the extra sequences at their 3′- or 5′-end respectively. The two PCR fragments were purified and fused into one fragment via a third PCR reaction, making use of the overlapping region. The PCR fragment was purified, cut with XhoI and KpnI, and cloned into the vector pSD1/M2-18 (Danner and Soppa, 1996). (vi) For the integration of the aptamer a first PCR fragment was generated using the plasmid pWHE601-AN32SH (Hanson et al., 2003) as a template. The PCR fragment consisted of the aptamer fused to the sequences upstream and downstream of the site of integration. Two other PCR fragments were generated consisting of the regions upstream and downstream of the site of integration respectively. The overlapping regions were used to fuse the three PCR fragments into one fragment via two consecutive PCR reactions. The fragment was purified, cut with XhoI and KpnI, and cloned into the vector pSD1/M1-18 (Danner and Soppa, 1996). The relevant parts of all newly generated plasmids were sequenced. Then they were used to transform H. volcanii as described previously (Cline et al., 1989).

Determination of dhfr transcript levels

RNA was isolated from exponentially growing cultures (4 × 108 cells ml−1) as described by Chomczynski and Sacchi (1987). DNase treatment, reverse transcription and real-time PCR analysis were performed as described previously (Brenneis et al., 2007). The results of the real-time PCR were analysed using the ΔΔCt method, which includes two subsequent normalization steps, i.e. to an unregulated internal control and to one of the samples (Livak and Schmittgen, 2001). As an unregulated internal control, the hpyA transcript levels were determined with the primer pair hpyA-RT_f and hpyA-RT_r. The Ct values of the control transcript hpyA were used to normalize the Ct levels of the dhfr reporter transcripts. The dhfr level of the chromosomal gene copy was determined using a strain carrying a plasmid without a dhfr gene (pNP10) (Patenge and Soppa, 1999). The value was subtracted from all samples to quantify the transcript level of the plasmid-encoded dhfr gene. Expression of the chromosomal gene copy was very low and the chromosomally derived transcript level was maximally 20% of the total dhfr transcript level. In each experiment one control was included, which was set to 1 and used for normalization of the dhfr levels derived from the other plasmids.

Determination of DHFR activities and of translational efficiencies

For the determination of DHFR-specific activities the same cultures were used as for the determination of the transcript levels. Harvesting and washing of cells, generation of a cytoplasmic extract and the enzymatic assay were performed as described previously (Brenneis et al., 2007). In short, the reduction of NADPH was determined at 340 nm and 25°C using the following concentrations in the assay mixture: 3 M KCl, 25 mM potassium phosphate, 25 mM sodium citrate pH 6.0, 0.05 mM dihydrofolic acid and 0.08 mM NADPH. The assay was performed in 1 ml (cuvette) or 0.25 ml (microtitre plate) volume. Protein concentrations were determined using the BCA assay according to the instruction of the manufacturer (Pierce) with BSA as a standard. Specific DHFR activities were calculated as described (Brenneis et al., 2007). The DHFR level encoded by the chromosomal dhfr copy was determined using a strain carrying the plasmid pNP10, which lacks the dhfr gene (Patenge and Soppa, 1999). It was subtracted to quantify the plasmid-encoded DHFR level. The chromosomally encoded DHFR-specific activity was much lower than the plasmid-encoded DHFR-specific activity and constituted less than 20% of the total value.

The translational efficiencies were calculated by dividing the specific DHRF activities with the transcript levels. At least three independent experiments were performed, and average values and standard deviations were calculated.

Western blot analysis

Protein samples for immunoblotting were prepared as described before (Brenneis et al., 2007). Approximately 80 μg of total protein per lane was separated on 14 cm × 16 cm sodium dodecyl sulphate (SDS)-polyacrylamide gels. Subsequently, the proteins were transferred onto nitrocellulose membranes (Protran BA 83; Whatman, Schleicher and Schüll, Dassel, Germany) by semi-dry blotting. The blots were probed with a specific DHFR antibody diluted 1:4000 in blocking solution. Horseradish peroxidase-conjugated goat anti-rabbit antibody (Sigma, Steinheim, Germany) was used at concentrations recommended by the manufacturer. Immunoreactive bands were visualized by chemiluminescence (ECL substrate). For quantification the films were scanned and the pictures were analysed with the software ‘ImageJ’ (http://rsb.info.nih.gov/ij/index.html). The background was determined locally for each band and subtracted from the signal. At least three independent experiments were performed, and average values and standard deviations were calculated.

Prediction of RNA secondary structures

The program ‘Mfold 3.2’ (Mathews et al., 1999; Zuker, 2003) was used for the prediction of putative secondary structures (http://mfold.bioinfo.rpi.edu/).

Acknowledgements

This work was supported by the Deutsche Forschungsgemeinschaft through Grant B6 in the framework of the special research programme SFB579 ‘RNA ligand interactions’.

Ancillary