Translation of mRNA lacking an in-frame stop codon leads to ribosome arrest at the 3′ end of the transcript. In bacteria, the tmRNA quality control system recycles these stalled ribosomes and tags the incomplete nascent chains for degradation. Although ubiquitous in eubacteria, the ssrA gene encoding tmRNA is not essential for the viability of Escherichia coli and other model bacterial species. ArfA (YhdL) is a mediator of tmRNA-independent ribosome rescue that is essential for the viability of E. coliΔssrA mutants. Here, we demonstrate that ArfA is synthesized from truncated mRNA and therefore regulated by tmRNA tagging activity. RNase III cleaves a hairpin structure within the arfA-coding sequence to produce transcripts that lack stop codons. In the absence of tmRNA tagging, truncated ArfA chains are released from the ribosome. The truncated ArfAΔ18 protein (which lacks 18 C-terminal residues) is functional in ribosome rescue and supports ΔssrA cell viability when expressed from the arfA locus. Other proteobacterial arfA genes also encode hairpins, and transcripts from Dickeya dadantii and Salmonella typhimurium are cleaved by RNase III when expressed in E. coli. Thus, synthesis of ArfA from truncated mRNA appears to be a general mechanism to regulate alternative ribosome rescue activity.
Not all ribosomes that initiate translation are able to complete synthesis of a full-length protein. Non-processive translation losses can be significant, particularly during the synthesis of large proteins such as β-galactosidase, in which 25–31% of initiating ribosomes fail to synthesize a full-length chain (Manley, 1978; Jorgensen and Kurland, 1990). Some incomplete synthesis is due to the translation of non-stop messages, which lack in-frame termination codons. Non-stop mRNAs are produced by premature transcription termination, non-enzymatic cleavage and nuclease activity (Hayes and Keiler, 2010). Ribosomes decode to the 3′ end of non-stop mRNA and then stall with an incomplete codon in the A-site. Non-stop mRNA is particularly problematic for bacteria, which have no quality control step between transcription and protein synthesis to ensure that only full-length transcripts are translated. The eubacteria address this molecular problem with the tmRNA•SmpB quality control system, which ‘rescues’ stalled ribosomes from non-stop mRNA (Moore and Sauer, 2007; Hayes and Keiler, 2010). The system's core is tmRNA, a stable RNA that functions as both transfer RNA and messenger RNA during ribosome rescue. tmRNA contains an alanine-charged tRNA-like domain and a small open reading frame (ORF) that encodes the SsrA peptide tag. During rescue, tmRNA enters the ribosome A-site and adds its alanine residue to the incompletely synthesized nascent chain (Keiler et al., 1996). The ribosome then releases the non-stop transcript and resumes protein synthesis using the tmRNA-encoded ORF. As a result, the C-terminus of the nascent chain is tagged with the SsrA peptide and the stalled ribosome is recycled into its constituent subunits. SmpB is a tmRNA-binding protein that is essential for both ribosome binding and SsrA peptide synthesis (Karzai et al., 1999; Sundermeier et al., 2005). The SsrA peptide targets tagged proteins for rapid degradation by ClpXP and other proteases (Keiler et al., 1996; Gottesman et al., 1998; Herman et al., 1998; Choy et al., 2007). Thus, tmRNA promotes the recycling of stalled ribosomes and the destruction of incomplete polypeptides.
Although tmRNA is thought to play a critical role bacterial physiology, the ssrA gene encoding tmRNA is not essential in several model bacteria including Escherichia coli, Bacillus subtilis and Caulobacter crescentus (Oh and Apirion, 1991; Wiegert and Schumann, 2001; Keiler and Shapiro, 2003). In contrast, ssrA is essential for the viability of other bacteria with small genomes, such as Neisseria gonorrhoeae, Haemophilus influenzae and Mycoplasma (Huang et al., 2000; Hayes and Keiler, 2010). These observations suggest that some bacteria possess tmRNA-independent rescue pathways to recycle stalled ribosomes. Indeed, ArfA (for alternative ribosome-rescue factor) has recently been identified as a possible mediator of tmRNA-independent ribosome rescue in E. coli (Chadani et al., 2010). Chadani et al. showed that ArfA induces nascent chain release from stalled ribosomes in S30 translation reactions. The arfA gene is not essential, but the deletion of both ssrA and arfA is synthetically lethal for E. coli. Thus, the emerging view is that ribosome rescue is indeed essential for bacteria, and that tmRNA-independent systems have evolved to deal with translational arrests in a more robust fashion.
In this report, we demonstrate that ArfA is synthesized from non-stop mRNA in E. coli. The arfA message is truncated immediately after the codon for residue Gly55, which induces tmRNA-mediated peptide tagging of ArfA for degradation. In the absence of tagging, truncated ArfA peptides are released from the ribosome. The major truncated species is ArfAΔ18, which lacks the C-terminal 18 amino acid residues. The ArfAΔ18 peptide is functional for tmRNA-independent ribosome rescue, and is capable of supporting the viability of ΔssrA cells. The 3′-coding region of arfA encodes a predicted hairpin that plays a critical role in generating non-stop mRNA. Disruption of the arfA hairpin structure by synonymous recoding prevents cleavage and allows production of full-length ArfA protein. Finally, we show that RNase III catalyses cleavage of arfA transcripts in vivo and in vitro. Together, these results strongly suggest that ArfA synthesis is regulated by tmRNA-mediated SsrA peptide tagging and subsequent proteolysis. Such regulation would provide a mechanism to increase ArfA-mediated ribosome rescue in response to decreased tmRNA activity.
ArfA protein is truncated
To characterize tmRNA-independent ribosome rescue activity, we cloned the E. coli arfA gene into a T7 promoter expression plasmid for the overproduction and purification of ArfA. Two constructs were tested initially, one encoding an N-terminal His6 epitope (His6–ArfA) and the other encoding a C-terminal His6 tag (ArfA–His6). Although both constructs produced protein, only the N-terminal tagged His6–ArfA bound to Ni2+-affinity resin under denaturing conditions (Fig. 1A and data not shown), suggesting that the C-terminal His6 tag was cleaved or otherwise missing from ArfA–His6. In the initial report of ArfA activity, Chadani et al. used an expression construct lacking the last 12 codons of arfA because the encoded His6–ArfAΔ12 protein was produced at high levels (Chadani et al., 2010). We analysed purified His6–ArfAΔ12 by SDS-PAGE and found that it migrated as a larger protein than His6–ArfA expressed from the full-length construct (Fig. 1A). Mass spectrometry of purified His6–ArfA showed that the chains were not full-length, but rather were truncated after residues Ala53, Ser54 and Gly55 (Fig. 1B and Table 1). To examine whether the C-terminus of ArfA was cleaved by protease activity, we expressed the full-length his6–arfA construct in strains lacking specific proteases and found that cells deleted for either clpX or clpP produced a larger His6–ArfA peptide (Fig. 1A and data not shown). Mass spectrometry revealed that the larger His6–ArfA species produced in ΔclpP cells were actually incomplete chains tagged with the SsrA peptide after residues Glu52, Ala53 and Ser54 (Fig. 1B and Table 1). This finding was confirmed by western blot analysis of purified His6–ArfA using antibodies to the SsrA peptide (Fig. 1A). These results indicate that a fraction of His6–ArfA is tagged by the tmRNA system and the tagged chains accumulate in the absence of the ClpXP protease. In contrast, SsrA-tagged His6–ArfAΔ12 did not accumulate in ΔclpP cells (Fig. 1A). We also examined a his6–arfAΔ18 construct (which corresponds to the most abundant residue Ser54 truncation product) and found that the overproduced His6–ArfAΔ18 protein was tagged at a much lower level than His6–ArfA produced from the full-length construct (Fig. 1A). Together, these results suggest that truncated arfA mRNA is produced from the full-length construct.
Table 1. Mass spectrometry of purified His6–ArfA proteins.a
Observed mass (Da)
Predicted mass (Da)
His6–ArfA proteins were overproduced using T7 promoter overexpression constructs, and purified by Ni2+-affinity chromatography and reverse-phase HPLC as described in Experimental procedures.
Represents the C-terminal residue of full-length protein.
Deduced from the mass (18 501 Da) of disulphide-bonded dimeric protein.
To determine whether arfA mRNA is truncated, we examined transcripts expressed from the arfA, arfAΔ12 and arfAΔ18 constructs by northern blot using a probe to the 5′-untranslated region (UTR) (Fig. 2A). This analysis showed that the arfA message was indeed truncated and accumulated to very high levels in ΔssrA cells (Fig. 2B), in accord with previous work showing that non-stop mRNA is stabilized in the absence of tmRNA (Yamamoto et al., 2003; Richards et al., 2006). Truncated mRNA was also observed with the arfAΔ12 and arfAΔ18 constructs, but full-length transcripts were also produced at approximately the same levels (Fig. 2B). Presumably the difference between truncated arfA and truncated arfAΔ12/arfAΔ18 messages is significant because overproduced His6–ArfAΔ12 and His6–ArfΔ18 proteins are not SsrA-tagged to the same degree as the wild type.
We next mapped the 3′ end of the truncated arfA message using S1 nuclease protection. This analysis demonstrated that the arfA transcript was truncated immediately downstream of codon for residue Gly55 (Fig. 2C). An additional cleavage product was detected downstream of the Gly55 codon (Fig. 2C, marked with an asterisk), but could not be unambiguously identified in the S1 protection map. Therefore, we performed primer extension analysis and determined that the second cleavage occurred immediately downstream of the codon for Thr63 (Fig. 3A and B). These results demonstrate that the arfA message is truncated upstream of the stop codon, accounting for the production of both truncated and SsrA-tagged ArfA chains. However, given that nearly all arfA message was truncated, it was surprising that only a fraction of the ArfA chains were tagged in ssrA+ΔclpP cells (Fig. 1A). Perhaps overproduced ArfA protein competes with tmRNA for the release of other ArfA nascent chains, initiating a positive feedback that allows untagged ArfA to accumulate in ssrA+ cells.
The arfA message is cleaved by RNase III
T7 RNA polymerase (RNAP)-based transcription reactions produce only full-length arfA mRNA in vitro (Fig. 2B and C and data not shown), suggesting that truncated message is produced by an RNase in vivo. Analysis of the arfA sequence using mfold revealed that the truncation occurs within a hairpin structure that encompasses much of the 3′-coding region (Fig. 3A). The arfAΔ12 and arfAΔ18 constructs lack the descending limb of this predicted hairpin, suggesting that the structure is important for mRNA truncation. Therefore, we disrupted the hairpin through synonymous recoding of residues Glu52–Cys72, and found that the recoded message was not truncated to the same extent as wild-type arfA and also produced full-length transcript (Fig. 4A and B). The arfA hairpin resembles known RNase III substrates (Regnier and Grunberg-Manago, 1990), prompting us to test whether this endoribonuclease is responsible for cleavage. We deleted the rnc gene (encoding RNase III) and found that full-length arfA mRNA accumulated to higher levels compared with rnc+ cells (Fig. 4B). Moreover, northern blot analysis using a probe to the 3′-UTR detected a downstream arfA cleavage fragment in rnc+ cells, but not Δrnc cells (Fig. 4B, lower panel). Taken together, these data strongly suggest RNase III cleaves arfA transcripts within the hairpin. This was confirmed by in vitro digestion of full-length arfA message with RNase III. Purified RNase III cleaved immediately after the codons for Gly55 and Thr63, which are identical to the cleavage sites observed in vivo (Figs 2C, 3B and 4B).
We next asked whether conditions that prevent arfA cleavage also allow synthesis of full-length ArfA protein. We expressed the his6–arfA construct in Δrnc cells and found that a larger protein was produced compared with the His6–ArfA produced in rnc+ cells (Fig. 5A). SDS-PAGE analysis showed that this larger protein species co-migrated with His6–ArfA produced from the recoded arfA construct (Fig. 5A). Mass spectrometry confirmed that these proteins were indeed full-length His6–ArfA (Table 1). Although full-length His6–ArfA was synthesized in Δrnc cells, we still detected SsrA tagging of the protein in Δrnc ΔclpP cells (Fig. 5A). His6–ArfA synthesized from the recoded construct was also tagged at low levels (Fig. 5A). These tagged peptides appeared to be distinct from the SsrA-tagged His6–ArfA produced in rnc+ cells, so we identified the tagged peptides using cells that express tmRNA(His6). The tmRNA(His6) variant encodes the SsrA(His6) tag, which is not degraded by ClpXP and can be used as an affinity epitope to specifically purify SsrA(His6)-tagged proteins for analysis (Roche and Sauer, 2001). We produced ArfA (without the His6 affinity tag) in Δrnc ssrA(his6) cells from the wild-type construct and from the recoded construct in rnc+ssrA(his6) cells, and purified the tagged chains by Ni2+-affinity for identification by mass spectrometry. This analysis revealed that ArfA synthesized in the Δrnc background was SsrA(His6)-tagged at several positions from residues Lys44–Gly65 (Table S1). ArfA synthesized from the recoded construct in rnc+ cells was tagged primarily after residues Leu68, Ser69 and Ala71 (Table S1). These results indicate that arfA contains other SsrA-tagging determinants that are independent of RNase III activity.
To ascertain the role of RNase III in arfA cleavage, we used a completely defined transcription–translation system to synthesize ArfA proteins in vitro. This highly purified system contains all the factors required for protein synthesis, but lacks tmRNA•SmpB and cellular RNases (Shimizu and Ueda, 2010). Reactions programmed with wild-type and recoded arfA plasmid constructs produced the same ArfA protein species, which migrated more slowly than either ArfAΔ12 or ArfAΔ18 in SDS-polyacrylamide gels (Fig. 5B). Remarkably, addition of purified RNase III to the wild-type arfA translation reactions produced a protein that co-migrated with ArfAΔ18 produced from the arfAΔ18 plasmid (Fig. 5B). Taken with the in vivo results, these data demonstrate that RNase III generates the non-stop mRNA from which truncated ArfA chains are synthesized.
Endogenous arfA mRNA is cleaved by RNase III
The preceding experiments used bacteriophage T7 RNAP to synthesize arfA mRNA. T7 RNAP transcribes more rapidly than the E. coli polymerase, which results in the uncoupling of transcription and translation (Iost et al., 1992). It is possible that this translational uncoupling allows adventitious cleavage of the arfA hairpin by RNase III. Moreover, the arfA hairpin may actually function as an intrinsic terminator of E. coli RNAP transcription. Therefore, we re-examined arfA messages synthesized by E. coli polymerase using a plasmid-borne arabinose-inducible PBAD promoter. Northern blot analysis again showed that the arfA message was less abundant in ssrA+ cells compared with the ΔssrA background, and that the transcript was truncated in rnc+ cells (Fig. 6A). In contrast, full-length arfA mRNA was detected in both ssrA+Δrnc and ΔssrA Δrnc cells (Fig. 6A), indicating that the arfA hairpin is not a potent terminator of E. coli RNAP-mediated transcription. However, significant levels of truncated arfA mRNA was still observed in Δrnc cells (Fig. 6A). These transcripts were truncated around the 3′ end of the hairpin structure (data not shown), and therefore could represent mRNA degradation intermediates that are stabilized by the hairpin in the absence of RNase III cleavage.
We next asked whether RNase III cleaves the endogenous arfA message produced from the chromosomal locus. To test the effects of the arfA hairpin, we generated strains in which the wild-type arfA gene was replaced with the arfAΔ18 and arfA(recode) alleles. Because these genetic manipulations introduced kanamycin resistance cassettes downstream of the arfA locus, we also generated a control arfA+ strain containing the resistance cassette inserted at the same position as the arfA(recode) strain. The resistance cassettes were then removed with FLP recombinase, leaving a single FRT ‘scar’ downstream of each arfA gene (Datsenko and Wanner, 2000). The endogenous arfA+(FRT) transcript was not detected in ssrA+rnc+ cells, but accumulated in the ΔssrA rnc+ background (Fig. 6B). Moreover, a significantly larger arfA+(FRT) transcript was observed when the rnc gene was deleted (Fig. 6B). This larger transcript co-migrated with the endogenous arfA(recode) message (Fig. 6B), strongly suggesting that the full-length arfA+(FRT) mRNA accumulates in the absence of RNase III. The FRT site is an inverted repeat that forms a stem-loop structure when transcribed into RNA. Because the FRT stem-loop is likely to stabilize the arfA+(FRT) message, we also examined samples from cells with unperturbed wild-type arfA loci to determine the effects of ssrA and rnc deletions (Fig. S1). Endogenous arfA message was only detected in ΔssrA cells and a slightly larger transcript was observed in the ΔssrA Δrnc background (Fig. S1). We also examined the effect of polynucleotide phosphorylase (PNPase) on endogenous arfA mRNA because deletion of the rnc gene significantly increases production of this exoribonuclease (Jarrige et al., 2001). Deletion of the pnp gene (encoding PNPase) had no discernable effect on endogenous arfA mRNA in ΔssrA cells, but altered the transcript pattern in the ΔssrA Δrnc background (Fig. S1). Taken together, these results indicate that RNase III cleaves the endogenous arfA message, but other nucleases can also produce non-stop arfA mRNA in the absence of RNase III.
tmRNA regulates the synthesis of ArfA
Because endogenous arfA mRNA was dramatically influenced by tmRNA and RNase III, we next sought to determine the effect of these factors on endogenous ArfA protein levels. We introduced a flag–arfA construct into the arfA chromosomal locus, but were unable to detect the encoded FLAG–ArfA protein by western blot analysis of either ssrA+ or ΔssrA cell lysates (data not shown). Therefore, we expressed the same flag–arfA construct from a plasmid-borne PBAD promoter, but were unable to detect FLAG–ArfA protein in either ssrA+rnc+ or ssrA+Δrnc cells (Fig. 7A). In contrast, FLAG–ArfA accumulated readily in the ΔssrA background. In ΔssrA rnc+ cells, the flag–arfA construct produced a protein that co-migrated with FLAG–ArfAΔ18, and a slightly larger protein accumulated in ΔssrA Δrnc cells (Fig. 7A). To determine the precise ArfA species produced under these conditions, we expressed his6–arfA from the same PBAD plasmid and purified His6–ArfA chains for mass spectrometry. This analysis showed that His6–ArfA protein was truncated primarily after residues Ala53 and Ser54 in ΔssrA rnc+ cells, whereas larger peptides truncated after residues Asn59 and His60 were produced in ΔssrA Δrnc cells (Table S1). In contrast, tmRNA had little effect on FLAG–ArfAΔ18 synthesis from the flag–arfAΔ18 construct, and we were unable to detect FLAG–ArfA protein in either ssrA+ or ΔssrA cells expressing the flag–arfA(recode) construct (Fig. 7A).
The inability to detect full-length FLAG–ArfA was somewhat surprising given that full-length His6–ArfA was produced by the T7 expression system in Δrnc cells (Fig. 5A and Table 1). Because ArfA is small, we reasoned the peptide could be proteolytically labile and therefore difficult to detect when expressed from the weaker PBAD promoter. To stabilize ArfA, we fused the last 38 codons of arfA (Gly35–Cys72) downstream of a flag–trxA module. Thioredoxin (TrxA) is routinely used to make soluble fusion proteins and is not normally tagged by the tmRNA system (Hayes et al., 2002a; Ruhe and Hayes, 2010). We were unable to detect the resulting fusion protein in ssrA+ cells, but FLAG–TrxA–ArfA(38) readily accumulated in ΔssrA cells (Fig. 7B). Moreover, a larger FLAG-tagged species was produced in ΔssrAΔrnc cells (Fig. 7B), consistent with the lack of hairpin cleavage. However, these data indicate that ArfA levels are effectively suppressed by tmRNA activity even in the absence of RNase III.
To determine whether the arfA hairpin is sufficient to subject a protein to RNase III/tmRNA regulation, we generated another fusion construct in which the hairpin sequence (encoding residues Gly49–Ser69) was fused in-frame between two trxA gene modules (Fig. 7C). Cleavage in the arfA hairpin should produce a truncated FLAG–TrxA protein that is easily differentiated from the full-length FLAG–TrxA-(hairpin)–TrxA fusion protein by western blot analysis. Indeed, truncated FLAG–TrxA was detected in ΔssrA rnc+ cells but not ssrA+rnc+ cells (Fig. 7C), presumably because the truncated protein was tagged and degraded the ssrA+ background. A slightly larger truncated protein was produced in ΔssrA Δrnc cells (Fig. 7C). The size of this truncated protein was consistent with the truncated arfA transcripts and SsrA(His6)-tagged ArfA chains observed in Δrnc cells (Figs 4B and 6A and Table S1). No truncated proteins were produced from a control construct in which the synonymously recoded arfA sequence was fused between the trxA modules (Fig. 7C). Taken together, these results demonstrate that tmRNA activity regulates ArfA synthesis. However, it appears that there are at least two distinct mechanisms producing truncated arfA mRNA: RNase III-mediated cleavage within the hairpin, and nucleolytic degradation to the 3′ end of the hairpin structure.
ArfAΔ18 and full-length ArfA are functional
The preceding results suggest that full-length ArfA is not synthesized at high levels in E. coli cells. If true, then ArfAΔ18 and similar truncated peptides are probably the most abundant arfA gene products and should be functional in tmRNA-independent ribosome rescue. We examined function by testing the ability of ArfA to release untagged nascent chains in cells expressing tmRNA(DD). Plasmid pPW500 expresses the N-terminal domain of phage λ cI repressor (λN-trpAt) from a well-characterized non-stop mRNA (Karzai et al., 1999; Roche and Sauer, 1999). This non-stop mRNA produces two distinct proteins in cells expressing tmRNA(DD): approximately 70% of the λN-trpAt chains are tagged with the SsrA(DD) peptide and the remaining chains are untagged (Fig. 8A, upper panel). In contrast, only untagged chains accumulate in ΔssrA cells (Fig. 8A, upper panel). ArfA is presumably required for the release of untagged chains because these species are not observed in ΔarfA cells (Fig. 8A). We expressed both wild-type arfA and arfAΔ18 constructs from plasmid-borne PBAD promoters and found that both promoted the release of untagged λN-trpAt chains (Fig. 8A, upper and middle panels). As a negative control, we expressed wild-type arfA and arfAΔ18 constructs carrying the Ala18Thr mutation, which abolishes ArfA ribosome rescue function (Chadani et al., 2010). Interestingly, the inactive ArfAA18T proteins appeared to exert a dominant-negative effect on endogenous ArfA activity (Fig. 8A). Deletion of arfA and ssrA is synthetically lethal in E. coli (Chadani et al., 2010), presumably because ArfA and tmRNA mediate parallel ribosome rescue pathways. To unequivocally demonstrate that ArfAΔ18 is functional, we deleted the last 18 codons of the chromosomal arfA gene and introduced the resulting arfAΔ18 allele into ΔssrA cells. The arfAΔ18 ΔssrA strain grew at essentially the same rate as arfA+ΔssrA cells (Fig. 8B, middle panel), indicating that ArfAΔ18 is fully function in tmRNA-independent ribosome rescue.
We also examined the function of ArfA produced from the recoded construct. ArfA expressed from a plasmid-borne arfA(recode) construct was just as efficient in the release of λN-trpAt chains as ArfA produced from the wild-type construct (Fig. 8A, lower panel). Additionally, ΔssrA cells carrying a chromosomal arfA(recode) allele were viable, although they exhibited a reproducible growth defect compared with the ΔssrA arfA+ and ΔssrA arfAΔ18 strains (Fig. 8B, middle panel). A related phenotype was observed in the Δrnc ΔssrA background, in which cells with the wild-type arfA gene grew more slowly than those carrying the arfAΔ18 allele (Fig. 8B, lower panel). Although it is unclear how much full-length ArfA is produced by these cells, the data suggest that truncated ArfA is more active than full-length protein when expressed from the chromosomal arfA locus.
ArfA homologues in other bacteria are synthesized from non-stop mRNA
The results presented thus far provide the basis for a regulatory pathway in which tmRNA controls the cellular levels of ArfA through SsrA peptide tagging and proteolysis. Recognizable homologues of ArfA are only found in a subset of γ- and β-proteobacteria. We found that many if not all arfA genes encode potential hairpin structures (Figs 9A and S2). To determine whether other arfA transcripts are also cleaved by RNase III to generate non-stop mRNA, we cloned the arfA genes from Salmonella typhimurium LT2 and Dickeya dadantii 3937 for the overproduction of His6–ArfA proteins in E. coli. Northern blot analysis revealed that each transcript was cleaved in an RNase III-dependent manner (Fig. 9B). Moreover, both His6–ArfA proteins were truncated when produced in rnc+ cells, whereas full-length proteins were synthesized in Δrnc cells (Fig. 9C and Table 1). Thus, RNase III and tmRNA probably act together to regulate ArfA synthesis in other bacterial species.
The results presented here demonstrate that RNase III cleaves arfA mRNA within a hairpin structure to produce non-stop transcripts. Translation of such non-stop mRNA leads to ribosome arrest, tmRNA-mediated ribosome rescue, and ultimately degradation of the ArfA nascent chain. This pathway probably serves to maintain low levels of ArfA when there is sufficient tmRNA activity, thereby preventing competition between the ribosome rescue systems. However, when tmRNA activity becomes limiting, untagged ArfA chains will be released and increase the capacity for tmRNA-independent ribosome rescue. Thus, ArfA acts as a fail-safe system that is only deployed when tmRNA is overwhelmed or incapacitated. This model predicts that ArfA levels are significantly higher in ΔssrA cells compared with ssrA+ cells, which is consistent with the results showing FLAG–ArfA expression is much higher in the ΔssrA background. Additionally, ArfA should also be more abundant in cells expressing tmRNA variants that encode protease-resistant SsrA peptides. Although the non-degradable SsrA peptide tags could conceivably interfere with ArfA activity, our results suggest that SsrA(DD)-tagged ArfA retains ribosome rescue function based on the ArfA/tmRNA(DD) competition experiments. Intriguingly, ArfA may be required for its own release from the ribosome, which could explain why most ArfA is not SsrA-tagged when expressed by the T7 RNAP system. As overproduced ArfA accumulates, it would compete more effectively with tmRNA for the release of additional ArfA nascent chains. If ArfA is indeed required for its own release, then there must be a mechanism to maintain low levels of the factor so that cells have the ability to release ArfA nascent chains in response to transient lapses in tmRNA activity.
RNase III and tmRNA are ubiquitous in eubacteria, and therefore ArfA synthesis could be regulated by SsrA-peptide tagging in other species as well. RNase III from E. coli efficiently cleaves the arfA transcripts from D. dadantii 3937 and S. typhimurium LT2, resulting in the production of truncated ArfA proteins. The arfA genes from other γ- and β-proteobacteria also encode predicted hairpin structures (Fig. S2), suggesting that these species also regulate ArfA synthesis in the same manner as E. coli. Although all arfA messages contain predicted hairpins, the primary sequence of this structure is not well conserved. This is reflected by the encoded C-terminal peptide sequences, which diverge significantly from one another. Multiple sequence alignment of ArfA proteins shows that homology only extends to E. coli residue His45 (Fig. S3), suggesting the C-terminus is not critical for function. Instead, maintenance of hairpin secondary structure appears to be the primary evolutionary constraint. For example, ArfA homologues from E. coli, Haemophilus influenzae and Mannheimia haemolytica are unrelated from residue Glu52 (E. coli numbering) to their C-termini, but each of the corresponding mRNA sequences forms an extensive hairpin structure (Figs S2 and S3). A hairpin structure is also found in arfA from Neisseria gonorrhoeae, whose ArfA homologue has diverged from the γ-proteobacterial proteins (Fig. S3). Despite the presence of arfA, tmRNA-mediated ribosome rescue function is essential in N. gonorrhoeae (Huang et al., 2000), perhaps indicating that ArfA and tmRNA are not strictly parallel pathways in this bacterium. Ribosome rescue function is probably required for the viability of all bacteria, yet arfA homologues are absent from B. subtilis and C. crescentus: two species in which the ssrA gene can be deleted (Wiegert and Schumann, 2001; Keiler and Shapiro, 2003). We suspect that these bacteria also possess tmRNA-independent rescue systems and predict that the corresponding genes will encode non-stop mRNA to facilitate regulation by tmRNA.
RNase III is known to cleave a variety of cellular and bacteriophage mRNAs. In most instances, RNase III cleaves within the 5′-UTR (Barry et al., 1980; Bardwell et al., 1989; Regnier and Grunberg-Manago, 1990; Jarrige et al., 2001), but cleavage in the coding sequence of some messages has also been reported (Shen et al., 1982; Darfeuille et al., 2007; Sim et al., 2010; Stead et al., 2010). Cleavage in the 5′-UTR typically modulates transcript stability, whereas RNase III activity in the coding region serves to inactivate messages. The regulatory mechanism described here appears to be distinct from these previously characterized RNase III cleavages. RNase III cleavage is likely to destabilize the arfA transcript, but it appears the primary purpose of this activity is to facilitate co-translational control of ArfA synthesis by tmRNA. Although the biological rationale for such regulation appears straightforward, the advantage of using RNase III to generate truncated transcripts is unclear. Non-stop mRNA can be produced by genes containing ρ-independent transcription terminators within the ORF (Keiler et al., 1996; Roche and Sauer, 1999). Such systems are efficient and obviate the need for a site-specific RNase. Because RNase III is itself regulated at the post-transcriptional and post-translational levels (Bardwell et al., 1989; Kim et al., 2008; Sim et al., 2010), cleavage could be influenced by additional regulatory inputs. However, our data show that non-stop arfA mRNA is still produced in Δrnc cells, so RNase III is not absolutely required to subject ArfA to regulation by tmRNA. Perhaps the advantages of the proposed regulatory pathway have led to redundant RNase III-dependent and -independent mechanisms of arfA mRNA cleavage. The selective pressure to produce truncated ArfA is supported by the slow growth of arfA(recode) ΔssrA cells compared with arfA+ΔssrA cells, as well as the more subtle growth phenotype of the wild-type arfA+ gene compared with arfAΔ18 allele in the ΔssrA Δrnc background. Taken together, it appears that truncated ArfA is optimal for cell growth in the absence of tmRNA.
The precise biochemical function of ArfA is still unclear. Chadani et al. have shown that ArfA binds ribosomes and that addition of purified His6–ArfAΔ12 to S30 extracts induces the release of nascent chains from stalled ribosomes (Chadani et al., 2010). However, S30 lysates contain hundreds of proteins, so these experiments do not demonstrate that ArfA itself possesses peptidyl-tRNA hydrolase activity. Although the E. coli arfA gene encodes a 72-residue protein, our data demonstrate that a significantly smaller 53-residue peptide (after post-translational removal of the initiating Met residue) is active and sufficient to support the viability of ΔssrA mutants. It seems unlikely that such a small peptide has both ribosome binding and hydrolase activities. Perhaps ArfA binds to another protein or RNA and the resulting complex catalyses peptidyl-tRNA hydrolysis. The E. coli YaeJ and PrfH proteins could conceivable serve this role. Both proteins are related to release factors, but lack the domain responsible for stop codon recognition (Hayes and Keiler, 2010). Although YaeJ was recently shown to release nascent chains from stalled ribosomes in E. coli, it appears to work independently of ArfA (Handa et al., 2010; Chadani et al., 2011). Another intriguing possibility is that ArfA mimics an A-site stop codon, thereby allowing release factors to bind stalled ribosomes and release nascent chains. This model appears plausible because release factors can mediate translation termination at sense codons under some circumstances (Zaher and Green, 2009), demonstrating that the stringency of release factor binding to the ribosome can be modulated significantly. Clearly, biochemical approaches with defined in vitro translation systems will be required to elucidate the mechanism of ArfA activity.
All E. coli strains were derivatives of X90 [F′lacIq lac′ pro′/araΔ(lac-pro) nal1 argE(amb) rifr thi-1] and are presented in Table 2. The ΔarfA::kan gene deletion was obtained from the Keio collection (Baba et al., 2006), and the Δrnc-38::kan mutation was provided by Dr Sydney Kushner (University of Georgia) (Babitzke et al., 1993). The last 18 codons of the arfA gene were deleted from the chromosomal locus using λ Red-mediated recombination as described (Hayes et al., 2002a; Datta et al., 2006). Recombination constructs were made by directionally cloning two homology regions into plasmid pKAN (Hayes et al., 2002a). The upstream homology region containing the arfAΔ18 gene was amplified using oligonucleotides arfA–Sac (5′-CGC GAG CTC CTG TCG ATC CGC ATC G) and arfAΔ18–Bam (5′-TTA GGA TCC TTT AAC TGG CCT CCC AGT TCC CCC G). The downstream homology region was amplified using arfA–Eco (5′-TCT GAA TTC AGG CGC TTG TTA AGA GCG G) and arfA–Kpn (5′-GCG CGG TAC CGT GGT GGA TTT GGC G). The resulting plasmid was digested with SacI/KpnI and the fragment containing arfA homology regions and the Kanr cassette was gel purified and electroporated into Red-expressing cells. The same strategy was used to introduce the recoded version of arfA into the chromosome. The upstream homology region containing recoded arfA was generated by sequential PCR. The arfA locus was first amplified using oligonucleotides, arfA–Sac2 (5′-GAG AGC TCG TGG AGT ATG TAT GTA TCG) and arfA–recode(1) (5′-CCT GTT GTG AAG AAA TGA TTT ACC TTC TTT CCT GAC GCT TCC CAG TTC CC). The resulting product was used as a template for a second PCR using arfA–Sac2 and arfA–recode(2) (5′-GTG CTC GAG TTA ACA TGC TCC ACT CAA CAA CAA CCC TGT TGT GAA GAA ATG). The second PCR product was used as a template for the final amplification using arfA–Sac2 and arfA–recode(3)–Bam (5′-TCG GGA TCC GCT GAA AGA GCA GAA TAA CCG CTC TTA ACA TGC TCC ACT CAA CAA C). The final product was digested with SacI/BamHI and ligated to plasmid pKAN. As a control, the Kanr cassette was also inserted downstream of the wild-type arfA gene. The upstream homology region for this construct was amplified using oligonucleotides arfA–Sac and arfA–Bam (5′-TCG GGA TCC GCT GAA AGA GCA GAA TAA CC). All mutations were transferred into strains CH12, CH113 and CH2385 by bacteriophage P1-mediated transduction. Chromosomal kanamycin resistance cassettes were removed with the FLP recombinase as described (Datsenko and Wanner, 2000).
Arabinose-induced ArfA expression under PBAD promoter, Tetr
Arabinose-induced ArfA expression under PBAD promoter, Tetr
Arabinose-induced ArfA expression under PBAD promoter, Tetr
Arabinose-induced FLAG–ArfA expression under PBAD promoter, Tetr
Arabinose-induced FLAG–ArfAΔ18 (Ser2–Ser54) expression under PBAD promoter, Tetr
Arabinose-induced expression of recoded FLAG–ArfA under PBAD promoter, Tetr
Arabinose-induced His6–ArfA expression under PBAD promoter, Tetr
Arabinose-induced expression of recoded His6–ArfA under PBAD promoter, Tetr
Arabinose-induced expression of FLAG–TrxA fused to the last 38 codons (Gly35–Cys72) of arfA, Tetr
Ptrc expression of FLAG-tagged fusion with the arfA hairpin (codons Gly49 –Ser69) inserted between two trxA ORFs, Ampr
Ptrc expression of FLAG-tagged fusion with recoded arfA codons Gly49–Ser69 inserted between two trxA ORFs, Ampr
The growth of all strains was determined in LB media at 37°C with aeration. Cells were first grown to mid-log phase in LB (OD600 ∼ 0.3–0.5), then diluted to OD600 = 0.05 in fresh pre-warmed LB media and growth monitored. All growth curves were determined by two independent experiments with identical results.
The E. coli arfA gene was PCR-amplified using oligonucleotides, arfA–Eco/Nde (5′-CTG GAA TTC ATA TGA GTC GAT ATC AGC) and arfA–Xho (5′-AAA CTC GAG ATT TGC TGA AAG AGC AG). The resulting product digested with NdeI and XhoI and ligated to plasmid pET21b (Novagen – EMD) to generate plasmid pET–arfA. The arfAΔ12 and arfAΔ18 constructs were generated by PCR using pET–Sph/Pst (5′-CAA GGA ATG GTG CAT GCC TGC AGA TGG CGC CC) in conjunction with arfA(Δ12)–Xho (5′-TTT CTC GAG TTA GTG ATT TAC TTT CTT GCC) and arfA(Δ18)–Xho (5′-TTA CTC GAG TTT AAC TGG CCT CCC AGT TCC CCC G) respectively. The resulting PCR products were digested with SphI and XhoI and ligated to pET21b to generate pET–arfAΔ12 and pET–arfAΔ18. The arfA PCR product was also digested with EcoRI and XhoI and ligated to plasmid pCH450 (Hayes and Sauer, 2003) to generate an l-arabinose inducible expression construct. ArfA expression constructs encoding N-terminal His6 epitope tags were generated by amplification of each construct with arfA(His6)–Nco/Spe-for (5′-ATA CCA TGG CAA AAA GTC ATC ATC ATC ATC ACC ACA CTA GTC GAT ATC AGC ATA CTA AAG GG), in conjunction with the appropriate reverse primer. These PCR products were digested with NcoI and XhoI and ligated to plasmid pFG21b (Garza-Sánchez et al., 2006). The resulting constructs (pET–his6–arfA, pET–his6–arfAΔ12 and pET–his6–arfAΔ18) append an N-terminal Met–Ala–Lys–Ser–His6–Thr sequence onto the ArfA peptides.
The arfA(recode) and his6–arfA(recode) expression constructs lacking the hairpin structure was generated by sequential PCR. Plasmids pET–arfA and pET–his6–arfA were used as templates for PCR with pET–Sph/Pst and arfA–recode(1). The resulting products were used as templates for second PCRs using pET–Sph/Pst and arfA–recode(2). The second PCR products were used as templates for the final amplification using pET–Sph/Pst and arfA–recode(3)–Xho (5′-AAA CTC GAG ATT TGC TGA AAG AGC AGA ATA ACC GCT CTT AAC ATG CTC CAC TCA ACA AC). The final PCR products were digested with SphI and XhoI and ligated to plasmid pET21b to generate plasmids pET–arfA(recode) and pET–his6–arfA(recode). The arfA genes from S. typhimurium LT2 and D. dadantii 3937 were amplified with the following primers: Styph–arfA–Spe (5′-ATT ACT AGT CGT TAT CAA CAT AAG AAA GGG) and Styph–arfA–Xho (5′-AGA CTC GAG CGT AAA AAA GCA GAA GAG CC); and 3937–arfA–Spe (5′-CGA TTT ACT AGT CAC TAT CGC CAT ACC AAA GGG) and 3937–arfA–Xho (5′-AGA CTC GAG ATT TAC TCA AAC AAC AGG C). The resulting PCR products were digested with SpeI and Xho and ligated to vector backbone of SpeI/XhoI-digested plasmid pET–his6–arfA.
The FLAG-tagged ArfA expression plasmids were generated by PCR using oligonucleotide FLAG–arfA–Nco/Spe (5′-TTT CCA TGG ATT ATA AAG ATG ATG ATG ACA AGA CTA GTC GAT ATC AGC ATA CTA AAG) in conjunction with either arfA–Xho or arfA(Δ18)–Xho. The resulting PCR products were digested with NcoI and XhoI and ligated to plasmid pCH450. Plasmid pCH450–flag–trxA–arfA(38) was constructed in several steps. The E. coli trxA gene was first amplified using oligonucleotides trxA–FLAG–Nco (5′-AAA CCA TGG ACT ACA AGG ACG ACG ATG ACA AGA TGG CCG ATA AAA TTA TTC ACC TGA CTG AC) and trxA–Pst (5′-TTT CTG CAG ACG CCA GGT TAG CGT CGA GG). This product was digested with NcoI and PstI and ligated to pTrc99A (GE Healthcare) to generate pTrc–flag–trxA. The 3′ end of arfA was then amplified with arfA(G35)–Pst (5′-TTT CTG CAG GGA AAG GCA GTT ACA TGC) and arfA–Xho/Hind (5′-AAA AAG CTT CTC GAG AAC GTG GTG GAT TTG GCG GTG G), and ligated to pTrc–flag–trxA using the PstI and HindIII restriction sites. Finally, the fused flag–trxA–arfA(38) construct was subcloned into pCH450 using NcoI and XhoI. Plasmids pFG501–trxA–(arfA)–trxA and pFG501–trxA–(recoded)–trxA are derivatives of plasmid pTrxA–SecM′–TrxA (Garza-Sánchez et al., 2006). The NcoI/XhoI fragment containing the flag–trxA–secM′–trxA fusion construct was first subcloned into plasmid pFG501 (Holberger and Hayes, 2009). The arfA hairpin and recoded arfA fusions to the downstream trxA module were generated by sequential PCR. The first amplification reactions used either arfA–trxA-for1 (5′-GAA AGT AAA TCA CTT TTT TAC CAC TGG TCT TCT GCT TTC AAA TTC CGA TAA AAT TCA CC) or recode–trxA-for1 (5′-GGT AAA TCA TTT CTT CAC AAC AGG GTT GTT GTT GAG TAA TTC CGA TAA AAT TCA CC) in conjunction with trxA–Xho (5′-GAA CTC GAG ATT CCC TTA CGC CAG GTT AGC GTC G). The first PCR products were then re-amplified using arfA–trxA-for2 (5′-CGG GGA TCC AAG GGA ACT GGG AGG CCA GTG GCA AGA AAG TAA ATC ACT TTT TTA CC) or recode–trxA-for2 (5′-CGG GGA TCC AAG GGA ACT GGG AAG CGT CAG GAA AGA AGG TAA ATC ATT TCT TCA CAA C) with trxA–Xho. The final PCR products were digested with BamHI and XhoI and ligated to BamHI/XhoI-digested plasmid pFG501–trxA–secM′–trxA.
RNA isolation, in vitro transcription and northern blot analyses were conducted as described previously (Hayes and Sauer, 2003). Northern blot analysis was conducted with 10 µg of total RNA for overexpressed arfA mRNA and with 40 µg of RNA for detection of endogenous arfA mRNA. In vitro transcription templates were generated by PCR using pET–Sph/Pst with arfA(A53) trunc (5′-GGC CTC CCA GTT CCC CCG), arfA(H60) trunc (5′-GTG ATT TAC TTT CTT GCC) and T7term rev (5′-AAA AAA CCC CTC AAG ACC CG). Oligonucleotides pET–rbs (5′-GTA TAT CTC CTT CTT AAA GTT AAA C) and pET-3′ (5′-TCA GCT TCC TTT CGG GCT TTG TTA GCA GCC) were used as 5′ and 3′ probes (respectively) for northern blot hybridization of pET plasmid produced arfA transcripts. Oligonucleotides arfA–Xho and arfA–probe (5′-CCT TTA TCT GCC CTT TAG TAT GCT GAT ATC G) were used as 3′ and 5′ northern blot probes (respectively) for endogenous arfA message and arfA transcripts expressed from the plasmid-borne PBAD promoter.
S1 nuclease protection analysis was conducted essentially as described (Hayes and Sauer, 2003) using arfA S1 probe (5′-GCA GAA GAC CAG TGG TAA AAA AGT GAT TTA CTT TCT TGC CAC TGG CCT CCC AGT TCC CCC GAT TGC CAT GTT TTC CTT TTC GC) as a probe to map the 3′ ends of arfA transcript cleavage products. Oligonucleotides arfA(E52) 5′-marker (5′-CTC CCA GTT CCC CCG ATT GCC ATG TTT TCC TTT TCG C), arfA(A53) 5′-marker (5′-GGC CTC CCA GTT CCC CCG ATT GCC ATG TTT TCC TTT TCG C) and arfA(G55) 5′-marker (5′-GCC ACT GGC CTC CCA GTT CCC CCG ATT GCC ATG TTT TCC TTT TCG C) were used as migration standards in the S1 nuclease protection mapping experiment. The marker oligonucleotides were first phosphorylated with T4 polynucleotide kinase, then radiolabelling using terminal transferase and [α-32P]-cordycepin triphosphate (3′-deoxyadenosine 5′-triphosphate, Perkin-Elmer).
Primer extension analysis was conducted as described (Diner and Hayes, 2009). RNA samples were hybridized to 5′-radiolabelled oligonucleotide arfA–RT (5′-TAA CCG CTC TTA ACA AGC GCC TGA AAG C) and then incubated with dNTPs and Superscript III reverse transcriptase (Invitrogen). Reactions were quenched with gel loading buffer and resolved on denaturing 10% polyacrylamide gels. Oligonucleotides arfA(F62) 3′-marker (5′-TAA CCG CTC TTA ACA AGC GCC TGA AAG CAG AAG ACC AGT GGT) and arfA(T63) 3′-marker (5′-TAA CCG CTC TTA ACA AGC GCC TGA AAG CAG AAG ACC AGT) were 5′-radiolabelled and used a gel migration standards. The full-length arfA in vitro transcript (10 pmol) was digested with 1.3 units of RNase III (New England Biolabs) in reaction buffer [150 mM NaCl–10 mM MgCl2–0.1 mM EDTA–0.1 mM DTT–25 mM Tris-HCl (pH 8.0)] for 10 min at 37°C. Reactions were quenched with an equal volume of 20 mM EDTA (pH 8.0) and analysed by northern blot, S1 nuclease protection and primer extension analyses.
Escherichia coli strains were grown overnight at 37°C in LB medium supplemented with the appropriate antibiotics (150 µg ml−1 ampicillin or 25 µg ml−1 tetracycline). The following day, cells were resuspended to an optical density at 600 nm (OD600) of 0.05 in fresh media and grown at 37°C with aeration. Once cultures reached OD600 ∼ 0.6, protein production was induced by addition of isopropyl β-d-1-thiogalactopyranoside (IPTG, 1.5 mM final concentration) or 0.4% (w/v) l-arabinose. After further incubation for 90 min, cells were collected by centrifugation and frozen at −80°C. Protein was extracted from frozen cells in urea lysis buffer [8 M urea–150 mM NaCl–10 mM Tris-HCl (pH 8.0)]. His6-tagged proteins were purified by Ni2+-nitrilotriacetic acid (Ni2+-NTA) affinity chromatography as described previously (Hayes et al., 2002a). Western blot analysis was performed as described (Hayes et al., 2002b). For mass spectrometry analysis, proteins were extracted with guanidine lysis buffer [6 M guanidinium-HCl–10 mM Tris-HCl (pH 8.0)] and purified by Ni2+-NTA affinity. His6-tagged ArfA proteins were further purified by reverse-phase high-pressure liquid chromatography as described (Garza-Sánchez et al., 2006). Fractions were then dried by speed-vac and reconstituted in aqueous 50% acetonitrile–1% formic acid and injected directly into a Waters Q-Tof II™ mass spectrometer. Mass data were processed using MassLynx analytical software.
ArfA activity was assessed by measuring the release of untagged λN-trpAt chains from stalled ribosomes. Plasmids pCH450–arfA, pCH450–arfAΔ18 and pCH450–arfA(recode) were co-transformed with plasmid pPW500 (Roche and Sauer, 1999) into ssrA(DD) or ssrA(DD) ΔarfA cells. The resulting strains were grown in LB media supplemented with 150 µg ml−1 ampicillin, 10 µg ml−1 tetracycline and 0.2% l-arabinose to mid-log phase. IPTG was then added (2 mM final concentration) to induce synthesis of λN-trpAt from plasmid pPW500, and the cultures incubated for a further 1.5 h. The λN-trpAt chains were then purified by Ni2+-affinity chromatography and equal amounts of protein analysed by SDS-PAGE. The in vitro translation experiments were conducted using the PURExpress® system (New England Biolabs) supplemented with [35S]-l-methionine (Perkin-Elmer) according to manufacturer's instructions. Reactions were precipitated with 20 volumes of 95% ethanol and analysed by SDS-PAGE and autoradiography.
This work was supported by the National Institutes of Health through Grant GM078634. Mass spectrometry was performed in the UC Santa Barbara Department of Chemistry and Biochemistry Mass Spectrometry Facility. We thank Dr Shaorong Chong (New England Biolabs) for providing purified transcription–translation reactions for preliminary experiments.