Control of Fur synthesis by the non-coding RNA RyhB and iron-responsive decoding

Abstract

The Fe²⁺-dependent Fur protein serves as a negative regulator of iron uptake in bacteria. As only metallo-Fur acts as an autogeneous repressor, Fe²⁺scarcity would direct fur expression when continued supply is not obviously required. We show that in Escherichia coli post-transcriptional regulatory mechanisms ensure that Fur synthesis remains steady in iron limitation. Our studies revealed that fur translation is coupled to that of an upstream open reading frame (uof), translation of which is downregulated by the non-coding RNA (ncRNA) RyhB. As RyhB transcription is negatively controlled by metallo-Fur, iron depletion creates a negative feedback loop. RyhB-mediated regulation of uof-fur provides the first example for indirect translational regulation by a trans-encoded ncRNA. In addition, we present evidence for an iron-responsive decoding mechanism of the uof-fur entity. It could serve as a backup mechanism of the RyhB circuitry, and represents the first link between iron availability and synthesis of an iron-containing protein.

Introduction

Iron is required as a cofactor for a large number of enzymes, and thus is pivotal for cellular metabolism. In contrast, under oxygen-rich conditions, iron is a source of toxic radicals, which can damage cellular components. Therefore, bacteria have evolved fine-tuned regulatory systems, which on one hand ensure sufficient iron uptake, and on the other hand minimize iron toxicity (Escolar et al, 1999; Hantke, 2001; Kadner, 2005). The ferric uptake regulator Fur is a key regulator of iron metabolism (Hantke, 1984). In the presence of Fe²⁺, metallo-Fur binds as a dimer to iron-responsive promoter regions termed Fur-boxes and blocks the access of RNA polymerase (Escolar et al, 1999). In contrast, gene induction by metallo-Fur has remained puzzling until Massé and Gottesman (2002) showed that transcription of the non-coding RNA (ncRNA) RyhB is negatively controlled by metallo-Fur, and that RyhB downregulates the expression of genes encoding iron-containing proteins. These authors proposed that RyhB blocks ribosome binding to the respective mRNA(s) through base-pairing. This hypothesis was supported by studies of the sodB gene encoding FeSOD, the iron-containing superoxide dismutase (Večerek et al, 2003; Geissmann and Touati, 2004). Moreover, blockage of sodB mRNA translation by RyhB was shown to result in rapid turnover of both the mRNA and the ncRNA (Massé et al, 2003; Afonyushkin et al, 2005; Morita et al, 2006).

Despite the central role of Fur in iron metabolism and virulence in several bacterial pathogens (Braun, 2005), the mechanisms governing fur expression are not fully understood. The primary regulators of the oxidative stress response, OxyR and SoxRS, stimulate Escherichia coli fur transcription from different promoters, which complies with the function of Fur in protecting cells against oxidative damage (Zheng et al, 1999). Autogeneous regulation by binding of metallo-Fur to its own promoter has been reported in several bacteria (Delany et al, 2002; Sala et al, 2003; Hernandez et al, 2006), and E. coli fur expression is known to be moderately auto-regulated (De Lorenzo et al, 1988). However, a remaining paradox in fur auto-repression is that only metallo-Fur is able to bind efficiently to Fur-boxes, that is, the transition from iron-replete to iron-deplete conditions would lead to transcription of fur under conditions where Fur requirement is not obvious.

In this study, we have tested whether fur gene expression is post-transcriptionally regulated in E. coli. We show that Fur synthesis is translationally coupled to that of an upstream open reading frame, which hereafter is termed uof (upstream of fur), and that the ncRNA RyhB downregulates translation of uof. Moreover, evidence is presented for iron-responsive decoding of uof. The key features of this novel mechanism involve decoding of the Ser codon UCA by tRNA_IV^Ser and an adjacent rare Arg codon.

Results

RyhB affects the abundance and stability of fur mRNA

Overexpression of the hfq RNA chaperone gene led to an increase in RyhB levels and to a concomitant decrease in the levels of fur mRNA (Večerek et al, 2003), indicating a possible role of RyhB in the post-transcriptional control of fur. We first verified this observation by determining the steady-state levels of fur mRNA in an E. coli ryhB− strain and in the parental strain upon addition of the iron chelator 2,2′-dipyridyl (dip), which is known to induce ryhB transcription by alleviating the negative control of Fur (Massé and Gottesman, 2002). The levels of fur mRNA increased during a time course of 16 min after addition of dip in the ryhB− mutant, whereas they did not significantly change in the ryhB+ strain (Figure 1A and B). One possible explanation for this observation was that the absence of RyhB results in higher levels of iron-containing proteins (Massé et al, 2005), which would decrease the cellular Fe²⁺ pool, leading to a reduced Fur activity, which in turn could result in derepression of fur transcription. Another explanation was that RyhB negatively regulates fur expression by directly affecting the stability of the transcript. The stability of fur mRNA was determined in the presence and absence of RyhB. The strains were grown to an OD₆₀₀ of 0.5, treated with rifampicin, and the fur mRNA levels were determined several times thereafter. As shown in Figure 1C and D, when compared with the parental strain (t_1/2∼1.6 min) the stability of fur mRNA was significantly increased in the ryhB− strain (t_1/2 ∼2.5 min). These experiments suggested that RyhB somehow counterbalances the increasing fur transcript levels, which under conditions of iron-depletion, results from a lack of metallo-Fur.

Figure 1.

RyhB negatively affects the steady-state levels and stability of fur mRNA. (A) Steady-state levels of fur mRNA upon addition of the iron chelator 2,2′-dipyridyl (dip) to exponentially growing cultures of E. coli DJ480 (ryhB+) and the isogenic ΔryhB (ryhB−) strain. Total RNA was extracted before (time 0) and several times after addition of dip as indicated, and the steady-state levels of fur mRNA and 5S rRNA (internal control) were determined by primer extension. (B) Graphical representation of the fur mRNA levels at different times after addition of dip in the ryhB+ (▪) and the ryhB− (▴) strain. The signals shown in the autoradiograph (A) were quantified by the ImageQuant software and the fur-specific signals were normalized to that of 5S rRNA. The signal intensities obtained at time 0 were set to 100%. (C) Stability of fur mRNA in the presence and absence of RyhB. Total RNA was extracted from exponentially growing ryhB+ and ryhB− strains before (time 0) and several times after addition of rifampicin. The amount of fur mRNA and 5S rRNA at each time was determined as described in Materials and methods. (D) The fur and 5S rRNA-specific signals (C) were quantified and the fur-specific signals were normalized to 5S rRNA. The relative amount of fur mRNA remaining at each time point in the ryhB+ (▪) and ryhB− (▴) strains is plotted as a function of time. The signal intensities obtained at time 0 were set to 100%. The experiments were performed in duplicate. The relevant sections of representative autoradiographs are shown in (A) and (C). Error bars (B, D) represent s.d.

Features of the 5′-untranslated region suggest a mechanism for indirect post-transcriptional control of fur by RyhB

As ncRNAs can prevent ribosome loading, which is followed by degradation of both the mRNA and the ncRNA (Massé et al, 2003), we searched for a possible RyhB target site in the vicinity of the fur ribosome binding site (rbs). However, an obvious complementarity between RyhB and the fur translation initiation region was not apparent. When transcribed from its authentic promoter, the E. coli fur transcript starts at nt −183 with regard to the A (nt +1) of the fur mRNA start codon (Zheng et al, 1999). We noticed that the fur transcript comprises an open reading frame (uof), which consists of 28 codons. It is located immediately upstream of and overlaps with the 5′-coding region of the fur gene (Figure 2). An extended complementarity between RyhB and the putative translation initiation region of uof was revealed by bioinformatics (Figure 2). In addition, we noted that the Shine and Dalgarno (SD) sequence of fur mRNA is masked by a putative stem–loop structure (Figure 2). Given that uof translation was predicted to be terminated at either of the two consecutive stop codons in the immediate coding region of fur mRNA, we next addressed the questions whether translation of fur is positively coupled to that of uof, and whether negative regulation of fur by RyhB could be exerted through inhibition of uof translation.

Figure 2.

Organization of fur mRNA and base-pairing between uof and RyhB. The uof and fur reading frames are depicted by white and black arrows, respectively. The sequence comprising nt −110 to +12 is enlarged. The putative SD sequences and start codons of uof and fur are boxed. The -GG- to -GGAGG- mutation in the putative SD of uof is indicated (see Figure 3C). The uof codons UCA₆ and AGA₇ involved in iron-responsive decoding (see text) and the two consecutive stop codons of uof in the proximal coding region of the fur gene are underlined. The stem–loop structures I and II were revealed by enzymatic probing (see Supplementary Figure S1). Bases protected from RNase T1 are indicated by filled arrowheads. The potential base-pairing between uof-fur mRNA and RyhB RNA is depicted. Analysis of complementarity was performed using the SeqMan software (Lasergene, DNA Star Inc.). The deletions made in the RyhB variants RyhBΔ2 and RyhBΔ3 are indicated. The mutations introduced in uof (pUuof_AUGfur-inv) and RyhB (RyhB-inv) are denoted by arrows.

fur translation is coupled to that of uof

As a canonical SD sequence was not apparent for uof (Figure 2), we first tested by in vitro toeprinting whether ribosome binding occurs at its putative start codon. As shown in Figure 3A, lane 3, a 30S ternary complex formed over the predicted start codon of uof (AUG_uof) and fur (AUG_fur), respectively. Two primer extension inhibition signals were obtained as a result of ribosome binding at AUG_fur, which was indicative of sub-optimal 30S ribosome binding (Bläsi et al, 1989). Therefore, the structure of uof-fur mRNA comprising the region upstream of AUG_fur was mapped by enzymatic probing. This analysis supported the idea that the fur SD sequence is occluded by a stem–loop structure (Figure 2 and Supplementary Figure S1), which in turn underlined the possibility that fur translation is coupled to that of uof.

Figure 3.

Coupled translation of uof and fur. (A) Toeprinting of fur242 mRNA. Lane 1: primer extension (EX) in the absence of 30S ribosomal subunits and tRNA_f^Met, lane 2: primer extension in the presence of 30S subunits (binary complex, BC), lane 3: toeprinting (TP) in the presence of 30S subunits and tRNA_f^Met, lanes 4–7: sequencing reactions. Arrows denote the primer extension inhibition signals as a result of ternary complex formation at the start codon of uof (AUG_uof) and the fur gene (AUG_fur). The dots in the autoradiograph depict the positions of the uof and fur start codons. (B) The full-length ‘wild-type' and mutant fur mRNAs were transcribed in vitro from the corresponding plasmids, as shown schematically below the autoradiograph. The uof and fur genes are depicted by white and gray bars, respectively. Mutations created in the translation initiation signals, the coding sequence and the stop codons of uof at the mRNA level are underlined. Equimolar concentrations (400 nM) of the respective fur mRNAs were used to program the in vitro translation reactions. The ¹⁴C-labeled translation products were resolved by SDS–polyacrylamide gel electrophoresis. Lanes 1–4: in vitro translation of fur mRNAs derived from plasmids pUuof_AUG-fur (1), pUuof_UA−CUG-fur (2), pUuof₅-fur (3) and pUuof₇₃-fur (4), respectively. The positions of the ∼17-kDa Fur protein and the uof-specific 73-aa product (P73) are indicated by arrows on the right. The positions of marker proteins and their molecular masses in kDa are shown by arrows on the left. (C) The E. coli MC4100 harboring plasmid pRuof-lacZ (1), wherein the first 24 codons of uof are fused to the 8th codon of the lacZ gene, or plasmids pRuof_AUGfur-lacZ (2), pRuof_CUGfur-lacZ (3), pRuof_UAfur-lacZ (4), pRuof_UA-CUGfur-lacZ (5) or pRuof_GGAGGfur-lacZ (6), wherein the first 36 codons of the fur gene are fused to the 8th codon of the lacZ gene, respectively, was grown in M9 medium to an OD₆₀₀ of 0.5. Then, 1 ml aliquots of the respective cultures were pulse labeled with [³⁵S]methionine for 2 min. Lane 1: UofΦLacZ protein synthesized in strain MC4100(pRuof-lacZ), lanes 2–6: FurΦLacZ proteins synthesized in strains MC4100(pRuof_AUGfur-lacZ; wt uof-fur entity), MC4100(pRuof_CUGfur-lacZ; uof_{AUG → CUG}), MC4100(pRuof_UAfur-lacZ: putative SD of uof was changed from -GG- to -UA-), MC4100(pRuof_UA-CUGfur-lacZ; uof_{AUG → CUG} and putative SD of uof was changed from -GG- to -UA-) and MC4100(pRuof_GGAGGfur-lacZ; putative SD of uof was changed from -GG- to -GGAGG-), respectively, lane 7: lack of β-galactosidase synthesis in strain MC4100 carrying the parental plasmid pRB381 (control), lane 8: molecular weight marker β-galactosidase. Only the relevant part of the autoradiograph is shown. The position of the FurΦLacZ fusion proteins and that of LacZ are denoted by arrows. The corresponding plasmids (1–6) are schematically shown below the autoradiograph. The uof, fur and lacZ gene(s) are depicted by white, gray and black bars, respectively. Mutations created in the translation initiation signals of uof at the mRNA level are underlined.

Several plasmids were constructed to test this hypothesis in vitro and in vivo. In all constructs, the lacpo (pR series) and lacpo or T7-promoter (pU series)-driven transcript encompasses uof-fur mRNA starting at nt −183 (Figure 2). The plasmid pair pUuof_AUGfur/pUuof_UA-CUGfur (Figure 3B) is identical except that the AUG start codon of uof was changed to a CUG and the -GG- bases, comprising the putative SD sequence of uof (Figure 2), were changed to -UA-. This was carried out with the reasoning that abrogation of uof translation should either drastically reduce or prevent translation of the fur gene. The removal of the translation initiation signals of uof abolished concurrent in vitro translation of the fur gene (Figure 3B, lane 2). Similarly, Fur synthesis did not occur upon in vitro translation of a fur mRNA variant bearing a UGA stop codon at position 6 in uof (Figure 3B, lane 3), that is, the premature stop codon in uof appeared to prevent translational coupling at the uof-fur boundary. The stop codons of uof are situated two nucleotides downstream of the fur start codon (see Figure 2). To test whether this putative ‘termination-restart motif' (Pavlov et al, 1997) is required for translational coupling at the uof-fur junction, the two stop codons of uof were replaced by sense codons (Figure 3B), and the fur mRNA variant was translated in vitro. The replacement of the uof stop codons drastically reduced fur translation and resulted in the synthesis of an uof-specific polypeptide (Figure 3B, lane 4; P73), which terminated at an in-frame stop codon located 73 codons downstream of the uof start codon. This experiment indicated that termination at the uof stop codon(s) is required for translation initiation of fur.

In contrast to the E. coli MC4100 harboring plasmid pRuof_AUGfur-lacZ (wild-type (wt) uof-fur entity; Figure 3C, lane 2), synthesis of the FurΦLacZ fusion protein was not observed in vivo in strains bearing plasmids pRuof_CUGfur-lacZ (Figure 3C, lane 3) and pRuof_UA−CUGfur-lacZ (Figure 3C, lane 5), in which the uof start codon was replaced by CUG and both the putative SD sequence and the start codon were altered, respectively. Similarly, when only the putative SD sequence -GG- of uof was changed to -UA-, synthesis of the UofΦLacZ protein was greatly diminished (Figure 3C, lane 4). Again, this set of in vivo experiments showed that translation of uof is required for concomitant translation of fur. To verify these observations, we also tested whether altering the weak SD sequence, -GG-, of uof to the canonical SD sequence -GGAGG- would result in enhanced translation of the fur reading frame. When compared with the authentic -GG- sequence, this modification increased FurΦLacZ synthesis ∼3.5-fold (Figure 3C, lane 6), which clearly supported the hypothesis that fur translation is coupled to that of uof. Moreover, the fusion proteins UofΦLacZ and FurΦLacZ, encoded by plasmids pRuof-lacZ and pRuof_AUGfur-lacZ, respectively, were synthesized at a comparable level (Figure 3C, lanes 1 and 2), suggesting that a translational termination event in uof results in a translational restart at the fur start codon. In summary, the data presented in Figure 3 revealed that translation of the distal fur gene depends on translation of the proximal uof reading frame.

We also tested whether the lack of uof translation affects the stability of uof-fur mRNA in strains H1941Δfur(pUuof_AUGfur) and H1941Δfur(pUuof_UA-CUGfur). When compared with uof_AUGfur mRNA (t_1/2 ∼1.9 min), removal of the translation initiation signals in uof_UA-CUGfur mRNA resulted in greatly reduced steady-state levels as well as in a faster decay of the uof_UA-CUGfur mRNA (t_1/2 ∼0.6 min) in vivo (Supplementary Figure S2).

RyhB downregulates the translation of uof and thereby Fur synthesis

Next, we asked whether post-transcriptional regulation of fur entails translational repression of uof by RyhB, and consequently that of the fur gene. First, we employed plasmid pRuof-lacZ (see Figure 3C), encoding an uof-lacZ fusion to monitor uof translation in the ryhB+ and the ryhB− strain. When compared with the ryhB− mutant, a decreased synthesis of the UofΦLacZ protein was seen in the ryhB+ strain, which became more apparent when the cultures entered the stationary phase (Figure 4A), that is, concomitant with the reported increase in RyhB synthesis (Argaman et al, 2001). As RyhB is known to require the RNA chaperone Hfq for function, we also tested whether the observed decrease in synthesis of the UofΦLacZ protein in the presence of RyhB depends on Hfq. When compared with the wt strain, the absence of Hfq led to a comparable increase in translation of the uof-lacZ gene as observed in the absence of RyhB (Figure 4A), suggesting that RyhB-mediated regulation of uof depends on Hfq.

Figure 4.

RyhB inhibits translation of uof and concomitantly of fur. (A) Translation of the uof-lacZ fusion gene in the E. coli ryhB+ (▪), ryhB− (▴) and hfq− (•) strains. All strains harbored plasmid pRuof-lacZ (Figure 3C), wherein the first 24 codons of uof are fused to the 8th codon of the lacZ gene. The strains were cultivated in M9 medium and triplicate aliquots were taken at different optical densities during growth of the ryhB+ (□), ryhB− (Δ) and hfq− (○) strains. The experiment was performed in duplicate. Error bars represent s.d. (B) The coupled in vitro transcription/translation reactions were programmed with 4 μg of plasmid pRuof-lacZ (Figure 3C) and carried out in the absence (lane 1) or presence of 20, 40 and 80 pmol (lanes 2–4) of RyhB RNA (upper panel) or DsrA RNA (lower panel). Only the relevant parts of the autoradiographs are presented. (C) In vitro translation of full-length uof-fur (wt) mRNA derived from plasmid pUuof_AUGfur (Figure 3B) in the presence of RyhB and deletion variants thereof. Uof-fur mRNA (400 nM) was in vitro translated in the absence (lane 1) and in the presence of 2 μM (lane 2), 4 μM RyhB (lane 3), 4 μM RyhBΔ2 (lane 4) and 4 μM RyhBΔ3 (lane 5). The molar ratios of uof-fur:RyhB are indicated on top of the autoradiograph. (D) Lanes 1–3: in vitro translation of uof-fur mRNA (400 nM) derived from plasmid pUuof_AUGfur (Figure 3B) in the absence (−)and in the presence of 4 μM RyhB and 4 μM RyhB-inv, lanes 4–6: in vitro translation of uof-fur-inv mRNA (400 nM) derived from plasmid pUuof_AUGfur-inv (see Supplementary data) in the absence (−)and in the presence of 4 μM RyhB and 4 μM RyhB-inv, respectively.

Second, to further test whether RyhB inhibits uof translation, the uof-lacZ gene encoded by the pRuof-lacZ plasmid was expressed in an E. coli in vitro transcription/translation system in the presence of increasing concentrations of RyhB. As shown in Figure 4B, increasing concentrations of RyhB either reduced or inhibited uof-lacZ expression, whereas increasing amounts of the nonspecific ncRNA DsrA did not.

Third, we examined whether translational silencing of uof by RyhB inhibits translational coupling at the uof-fur boundary, and thereby Fur synthesis. Full-length in vitro-transcribed uof-fur mRNA derived from plasmid pUuof_AUGfur (Figure 3B) was translated in vitro in the presence of increasing concentrations of RyhB RNA. As shown in Figure 4C, lanes 2 and 3, increasing amounts of RyhB blocked fur mRNA translation. It is worth noting that the decay rates of fur mRNA in the in vitro translation system were comparable in the presence and absence of RyhB (Supplementary Figure S3), which indicated that inhibition of Fur synthesis results from a RyhB-mediated block of uof translation. For verification, two RyhB deletion constructs, RyhBΔ2 and RyhBΔ3 (Figure 2), were used. When compared with RyhB, neither of the RyhB variants with a reduced complementarity to uof exerted a notable effect on fur in vitro translation (Figure 4C, lanes 4 and 5). To verify the RyhB–uof interaction, mutations were introduced in RyhB RNA at nt positions 43 and 44 and in uof-fur mRNA at positions −57 and −58 (see Figure 2; RyhB-inv and uof-fur-inv). As shown in Figure 4D, lanes 3 and 5, the RyhB-inv and RyhB wt RNAs did not inhibit synthesis of Fur from uof-fur wt mRNA and uof-fur-inv mutant mRNA, respectively. In contrast, synthesis of Fur from the uof-fur-inv mRNA was blocked by RyhB-inv RNA (Figure 4D, lane 6) bearing the compensatory mutations. Taken together, these results showed that RyhB interacts with the uof sequence of the uof-fur mRNA, thereby blocking translation uof, and consequently that of fur.

Evidence for regulation of uof by iron-responsive decoding

To test whether the observed effects of RyhB on fur mRNA levels (Figure 1A) are mirrored in the de novo synthesis of Fur protein in vivo pulse-labeling experiments were performed. As shown in Figure 5A, upon addition of dip, the de novo synthesis of Fur was hardly affected in the ryhB+ strain during a time course of 20 min, as expected from the observed insignificant changes in the steady-state levels of fur mRNA (Figure 1A). In contrast, upon addition of the chelator, Fur synthesis increased up to 8 min in the ryhB− strain and then decreased to comparable levels, as observed in the ryhB+ strain (Figure 5A) even though the fur mRNA levels increased under these conditions (see Figure 1A and B).

Figure 5.

Iron-responsive decoding of uof. (A) De novo synthesis of Fur protein in E. coli DJ480 (upper panel) and in the isogenic ΔryhB strain (lower panel) upon addition of the iron chelator dip. The cells were grown in M9 medium to an OD₆₀₀ of 0.5. Aliquots (1 ml) of both cultures were pulse labeled with [³⁵S]methionine (10 mCi/ml) for 2 min before (time 0), 4, 8, 16 and 20 min after addition of dip. The samples were processed as described in Materials and methods. Only the relevant part of the autoradiograph of the immunoblot showing the synthesized Fur protein (arrow) is presented. (B) Graphical representation of the de novo synthesis of UofΦLacZ protein upon addition of dip in the E. coli DJ480ΔryhB harboring plasmids pRuof-lacZ (▪), pRuof_UCU-lacZ (▴) and pRuof_CGU-lacZ (◊). The experiment was performed in duplicate. Error bars represent s.d. Insert: DJ480ΔryhB harboring plasmids pRuof-lacZ (▪), pRuof_UCU-lacZ (▴) or pRuof_CGU-lacZ (◊), was pulse labeled with [³⁵S]methionine for 2 min at time 0 (lane 1), 5 min (lane 2), 10 min (lane 3) and 20 min (lane 4) after addition of dip (final concentration 250 μM). The proteins were resolved in a 12% SDS–polyacrylamide gel. Only the relevant part of the autoradiographs of one representative experiment is shown. In all plasmids (bottom), codon 24 of uof (open bar) is fused to the eigth codon of the lacZ gene (black bar). Mutations introduced in uof at the mRNA level are underlined. (C) Translational efficiencies of the uof₅-lacZ and uof₇-lacZ fusion genes in an miaB+ (bars on the left) and an miaB− strain (bars on the right) harboring plasmids pRuof₅-lacZ (gray bar) and pRuof₇-lacZ (white bar). The experiment was performed in duplicate as described in Materials and methods. The values obtained with the uof₅-lacZ construct were set to 100%. The error bars represent s.d. The steady-state levels of the uof₅-lacZ and uof₇-lacZ mRNAs, respectively, are shown below the bars. 5S RNA was used as a loading control. The plasmid constructs used in these experiments are schematically shown at the bottom. The open and black bars represent the uof and the lacZ gene, respectively.

The only iron–sulfur (Fe–S) enzyme involved in translation is MiaB (Pierrel et al, 2002), which participates in methylthiolation of adenosine 37 (A37) of almost all tRNAs that read codons starting with U (UNN), except tRNA_I^Ser and tRNA_V^Ser (Grosjean et al, 1985). A37, next to the anticodon, stabilizes the codon–anticodon interaction by increasing the stacking energy (Lim, 1997), whereas the modification of A37 affects the accuracy of codon reading in the ribosomal A-site (Bouadloun et al, 1986). In addition, it was shown that the ms²io⁶A-37 hypermodification in Salmonella typhimurium tRNAs contributes to the decoding efficiency (Esberg and Björk, 1995). As shown in Figure 2, uof comprises, at position six, the serine codon UCA, which is decoded by the MiaB-dependent tRNA_IV^Ser. In addition, the AGA codon downstream of UCA in uof is decoded by the rare tRNA₄^Arg. It is known that the presence of AGA codons close to the initiation codon can lead to ribosome stalling and eventually release of mRNA (Gao et al, 1997; Cruz-Vera et al, 2004). As iron limitation results in a lack of methylthiolation of A37 (Buck and Griffiths, 1982), we asked whether the observed decrease of Fur de novo synthesis in the ryhB− strain upon iron depletion (Figure 5A) could be attributed to the consecutive 5′-…UCA₆AGA₇…-3′ codons (Figure 2), that is, whether this codon arrangement in uof renders uof-fur translation responsive to iron.

We first tested whether iron depletion affects expression of uof in the absence of RyhB. Pulse-labeling experiments for 20 min after addition of dip revealed that the de novo translation of the uof-lacZ gene was reduced in the ryhB− strain by ∼25–30% when it contained the authentic 5′-…UCA₆AGA₇…-3′ codons in uof (Figure 5B). However, when UCA₆ in uof was replaced by the serine codon UCU, which is cognate for the MiaB independent tRNA_V^Ser, no dip-dependent decrease in the de novo synthesis of the UofΦLacZ protein was observed (Figure 5B). We noted that the steady-state levels of uof-fur mRNA derived from strain DJ480ΔryhB (pRuof-lacZ) and DJ480ΔryhB (pRuof_UCU-lacZ) after 20 min of addition of dip were comparable (data not shown), which suggested that the observed decrease in the UofΦLacZ synthesis in DJ480(pRuof-lacZ) is linked to translation. Comparably, the conversion of AGA codon 7 to the canonical arginine codon CGU abolished dip-dependent decrease of UofΦLacZ synthesis (Figure 5B).

To verify whether the sub-sequence 5′-…UCA₆AGA₇…-3′ is sufficient for iron-responsive decoding of uof, two plasmids, pRuof₅-lacZ and pRuof₇-lacZ (Figure 5C) were constructed, in which either uof codon 5 or codon 7, was abutted to the 8th codon of the lacZ gene. Similarly, as observed in the experiments shown in Figure 5B, addition of dip resulted in a decreased synthesis of the Uof₇ΦLacZ but not of that of the Uof₅ΦLacZ fusion protein in strain DJ480ΔryhB− (data not shown). Translational efficiencies of the uof₅-lacZ and uof₇-lacZ genes obtained with plasmids pRuof₅-lacZ and pRuof₇-lacZ were comparable in the miaB+ strain, showing that the sub-sequence 5′-…UCA₆AGA₇…-3′ has no general negative effect on expression of the corresponding lacZ fusion genes under iron-replete conditions (Figure 5C). To test whether the decrease in uof₇-lacZ expression upon dip addition is linked to MiaB, we next compared the translational efficiencies of the uof₅-lacZ and uof₇-lacZ genes in a miaB− strain. As shown in Figure 5C, in the mid-logarithmic phase, the translational efficiency of the uof₇-lacZ gene was ∼20% reduced when compared with that obtained with the uof₅-lacZ gene. As the steady-state levels of the uof₅-lacZ and uof₇-lacZ mRNAs were indistinguishable (Figure 5C), this experiment supported the idea that decoding of the sub-sequence 5′-…UCA₆AGA₇…-3′ is influenced by MiaB.

Under-modification of A37 appears to affect ribosome movement only to a measurable extend when consecutive iron-responsive UNN codons are present in a reading frame (Landick et al, 1990) or, as indicative from Figure 5B, when an UNN codon is adjacent to at least one rare codon. Hence, we analyzed a number of E. coli mRNAs, including sodB, sdhD, acnA, fumA and bfr that encode iron-containing enzymes and are negatively regulated by RyhB (Massé and Gottesman, 2002; Massé et al, 2005) for the presence of ‘iron-responsive codons'. This analysis revealed that consecutive UNN codons and/or ‘iron-responsive/rare codon arrangements' are predominantly found at or close to the 5′-end of these genes (data not shown). To gain further support for iron-responsive decoding, we finally tested whether the synthesis of FeSOD (sodB), aconitase A (acnA) and malate dehydrogenase (mdh) is affected in the ryhB− strain upon addition of the iron chelator. In contrast to the sodB and acnA genes, the mdh gene, encoding the non-iron-containing malate dehydrogenase does not contain iron-responsive codons in the 5′-coding region (Supplementary Figure S4). The synthesis of the corresponding FeSODΦLacZ and AcnAΦLacZ proteins decreased upon iron limitation in the absence of RyhB, whereas synthesis of the MdhΦLacZ protein continued unabated (Supplementary Figure S4). These results underlined our hypothesis of iron-responsive decoding and could suggest a link between iron availability and synthesis of iron-containing proteins, which are non-essential under iron-limiting conditions.

Discussion

Originally, translational coupling has been suggested to provide a mechanism to ensure equimolar synthesis of proteins encoded by a polycistronic transcript (Oppenheim and Yanofsky, 1980). In majority of examples, ribosomes terminating on the proximal gene usually relieve an inhibitory mRNA structure occluding the downstream rbs, as it appears to be the case for the fur gene (see Figure 2). This can either lead to de novo translation initiation or translational restarts (Ivey-Hoyle and Steege, 1992). Translation of fur required upstream translation of uof, and uof and fur are translated with the same ratio (see Figure 3C). Therefore, translational restarts at fur rather than de novo initiation would explain the high-efficiency translational coupling between uof and fur. Although it remains elusive as to why fur translation in E. coli K12 is regulated by means of translational coupling, it seems worth noting that an identical genetic overlap between uof and fur is found in a pathogenic E. coli strain as well as in several other enteric bacteria (Figure 6). A BLAST search did not reveal any homology of the peptide encoded by uof, nor could any probable function be assigned by bioinformatic analyses. However, as uof expression is apparently regulated by iron, at present we cannot exclude the possibility that the encoded basic peptide has a role in iron metabolism.

Figure 6.

Established and inferred post-transcriptional control of fur in different bacterial species. (A) E. coli K12-like conserved regulation based on the presence of an uof (gray arrow), an RyhB homolog and ‘iron-responsive/rare codon arrangements' (hatched rectangles) in E. coli 0157:H7 (GenBank NC_000913.2), four Shigella species (S. dysenteriae, GenBank CP00034; S. boydii, CP000036.1; S. sonnei, CP000038.1; S. flexneri, AE014073.1) and in S. enterica (GenBank NC_004631). (B) Inferred direct negative regulation of fur by RyhB orthologs and ‘iron-responsive/rare codon arrangements' (hatched rectangles) in P. aeruginosa (Wilderman et al, 2004) and V. cholerae (Davis et al, 2005). (C) Putative downregulation of the fur gene by a cis-antisense RNA (wavy line) in Anabaena species PCC7120 (Hernandez et al, 2006).

The in vitro and in vivo experiments presented in Figure 4 showed that RyhB diminishes translation of uof in an Hfq-dependent manner. As uof and fur translations are coupled, to our knowledge, this is the first report on indirect translational regulation by a trans-encoded ncRNA. The sequence of RyhB is conserved in E. coli 0157:H7, Shigella flexneri (Oglesby et al, 2005) as well as in other Shigella species (Figure 6). In addition, an RyhB homologue with the potential to base-pair with the translation initiation region of the short reading frame upstream of fur gene is present in S. enterica (Figure 6). Thus, the molecular mechanisms underlying control of fur expression appear to be conserved in several enteric bacteria. In addition, the Fur-regulated RyhB orthologs of Pseudomonas aeruginosa (Wilderman et al, 2004) and Vibrio cholerae (Davis et al, 2005) display complementarity with the translation initiation region of the respective fur genes (B Večerek, unpublished), and thus could similarly downregulate fur mRNA translation (Figure 6). Yet another variation of the post-transcriptional regulation of fur appears to be operative in the Anabaena species PCC 7120. Here, a cis-antisense RNA has been implicated in the post-transcriptional regulation of fur (Figure 6; Hernandez et al, 2006).

RyhB apparently counterbalances the increased transcription of uof-fur mRNA (Figure 1) under iron-deplete conditions, that is, in the absence of metallo-Fur. One of the best-characterized mechanisms underlying the functional inactivation of a target mRNA by an ncRNA is the RyhB-mediated decay of the E. coli sodB mRNA. Recent work (Massé et al, 2003; Afonyushkin et al, 2005) has shown that RNase E is involved in the stability control of sodB mRNA upon translational inhibition by RyhB. We observed that the entire uof-fur transcript is stabilized in RNase E mutants under non-permissive conditions (I Moll and B Večerek, unpublished). The stability of uof-fur mRNA was reduced in the presence of RyhB (Figure 1), and elimination of the uof translation initiation signals, that is, inhibition of fur translation (see Figure 3C), resulted in strongly diminished steady levels, as well as in a reduced stability of the uof_UA-CUGfur mRNA in vivo (Supplementary Figure S2). Hence, a simple model for RyhB-mediated negative regulation of fur mRNA (Figure 7) would include translational inhibition of uof-fur translation, endonucleolytic RNase E cleavages in the untranslated uof-fur mRNA, followed by 3′ → 5′ decay of intermediate mRNA fragments.

Figure 7.

Model for post-transcriptional regulation of fur in E. coli K12. In the presence of iron (+Fe), metallo-Fur functions as a weak repressor of its own mRNA (De Lorenzo et al, 1988). As a consequence, transcription of fur mRNA leads to a certain level of metallo-Fur synthesis, which among many other genes represses ryhB transcription (Massé and Gottesman, 2002). In the absence of iron (−Fe), Fur is inactive, RyhB is synthesized and binds to and inhibits translation of multiple targets, including a number of mRNAs encoding iron-binding proteins as well as of uof. This results in downregulation of uof translation and degradation of a proportion of uof-fur mRNA (see Figure 1). Another level of regulation entails under-modification of tRNA_IV^Ser caused by iron scarcity and the presence of the ‘iron-responsive/rare codon arrangement' in uof. Iron-responsive decoding may serve as a back-up mechanism for RyhB-mediated regulation, and may cause a blockage of ribosomal movement and/or a release of ribosomes. The mechanisms governing post-transcriptional control of fur expression lead to a constant level of Fur synthesis under iron-replete and iron-deplete conditions. Inactive apo-Fur could potentially poison metallo-Fur (see text).

The iron–sulfur protein MiaB methylthiolates i⁶A37 in certain tRNAs (Esberg et al, 1999; Pierrel et al, 2002). The conversion of i⁶A37 to ms²i⁶A37 does not occur during growth of E. coli in iron-restricted medium (Buck and Griffiths, 1982; Buck and Ames, 1984). Under these conditions, the activity of two enzymes encoded by the trp operon was shown to be upregulated 2–3-fold (Buck and Griffiths, 1982). These effects were later ascribed to an increased slowing or stalling of ribosomes at the two tandem Trp codons in the leader sequence (Landick et al, 1990). In our studies, iron limitation resulted in a reduced expression of the ‘wild-type' uof-lacZ gene, but not when the UCA₆ codon was replaced by UCU or when the rare AGA₇ codon was replaced by the frequently used CGU codon. The latter two variation controls showed that iron depletion does not have a general negative effect on the expression of the uof-lacZ gene. In addition, as the combination of the ‘iron-responsive' (UCA₆) and ‘rare codon' (AGA₇), but not the presence of either codon alone, resulted in a decrease in translation, effects related to plasmid-mediated overexpression of the fusion genes, that is, to a shortage of either tRNA species, can most likely be excluded. Moreover, when compared with the uof₅-lacZ mRNA, the presence of the sub-sequence -5′-…UCA₆AGA₇…-3′ in uof₇-lacZ mRNA had no effect on expression in a miaB+ strain, but decreased expression in an miaB− strain (Figure 5C). As the steady-state levels of the uof₅-lacZ and uof₇-lacZ mRNAs were indistinguishable in either strain (Figure 5C), this phenomenon is most likely linked to MiaB and occurs at the level of translation.

Surprisingly, iron-responsive decoding was observed rather rapidly (between 5 and 10 min; see Figure 5B) after addition of the iron chelator. Although there are no literature data on MiaB concerning the loss of the [Fe–S]²⁺ cluster upon iron removal, it has been reported that aconitase B, which contains four [Fe–S]²⁺ clusters, is very rapidly (t_1/2=<5 min) demetallated upon addition of dip (Varghese et al, 2003). As on the other hand the pool of tRNAs containing the ms²i⁶A modification is known to be very low under aerobic growth conditions (Buck and Ames, 1984), a rapid demetallation of MiaB upon iron shortage would be expected to lead concomitantly to a quick depletion of methylthiolated tRNAs. In turn, this could explain why iron-responsive decoding occurs rather rapidly after iron deprivation.

Previous studies have shown that hypomodification of A37 can cause +1 frameshifts (Urbonavicius et al, 2001). However, β-galactosidase activity conferred by the −1 and +1 out-of frame uof-lacZ fusions did not increase upon addition of the iron chelator (B Večerek, unpublished), which argued against a frameshifting event occurring at the sub-sequence 5′-…UCA₆AGA₇…-3′. Therefore, the observed effects can either be reconciled with ribosome stalling (Landick et al, 1990; Figure 7) or with mRNA/peptidyl-tRNA drop-off at the sub-sequence -5′-…UCA₆AGA₇…-3′ (Gao et al, 1997; Cruz-Vera et al, 2004). Nevertheless, as uof and fur are translationally coupled, either event would have the same consequence, that is, a decrease in Fur synthesis. Iron responsive expression of uof-fur may not be limited to E. coli K12. Assuming a similar codon usage and perhaps tRNA modification function in Gram-negative bacteria (Daniels et al, 1998), it is noteworthy that apart from the four Shigella species and E. coli 0157:H7 (Figure 6), comprising at the sequence level identical uof-fur entities, ‘iron-responsive/rare codon arrangements' are also present at the beginning of the short putative reading frame upstream of fur in S. enterica as well as in the 5′-coding region of the fur genes of P. aeruginosa and V. cholerae (Figure 6).

In the model shown in Figure 7, RyhB-dependent silencing of uof concurrently with fur under iron-deplete conditions counterbalances fur mRNA synthesis transcribed in default of metallo-Fur. As a result, Fur synthesis remains steady upon iron depletion (see Figure 5A). This finding may be explained in terms of the iron-sparing model (Massé et al, 2005), which predicts a RyhB-mediated redistribution of intracellular iron upon repression of non-essential iron-containing proteins to essential iron-containing enzymes during iron limitation. Hence, an increase in Fur levels under these conditions could be adverse, as Fur could compete for intracellular iron with the essential iron-containing enzymes. Consequently, this scenario could interfere with sufficient iron uptake.

Iron depletion in the absence of RyhB resulted first in an increase and then a decrease of Fur translation (Figure 5A). As fur mRNA levels continued to increase under these conditions (Figure 1A), the decrease in the de novo synthesis of Fur can most likely be attributed to iron-responsive decoding. As mentioned above, MiaB inactivation may occur rapidly after iron depletion. Hence, we hypothesize that iron-responsive decoding of uof-fur, and by inference that of other genes (see Supplementary Figure S4), serves in addition to RyhB as a back-up mechanism to downregulate the synthesis of Fur and other non-essential iron-containing proteins during iron limitation (Figure 7).

Why does Fur synthesis occur (Figure 7) upon iron limitation? Possibly, ongoing Fur synthesis could ensure a rapid conversion of inactive apo-Fur, synthesized during iron limitation, into metallo-Fur, and thus a quick response to reach iron homeostasis. However, any model for fur regulation should address the cellular abundance of Fur (Zheng et al, 1999). Although to our knowledge there is no direct experimental evidence, it is generally assumed that Fe²⁺ dissociates from metallo-Fur when iron becomes scarce. In light of this, we would like to suggest another possible role for the observed Fur synthesis under iron-limiting conditions. Fur binds as a dimer to DNA, and iron binding to Fur appears to trigger conformational changes, which are required for a tight interaction with the ‘Fur-boxes' on DNA (Gonzalez de Peredo et al, 2001). However, as apo-Fur dimers have been detected, iron does not seem to be a pre-requisite for dimerization of Fur (Adrait et al, 1999). Taking this in consideration, it is conceivable that apo-Fur present or synthesized upon iron limitation (Figure 7) binds to residual metallo-Fur by dimerization, which in turn could lead to non-functional dimers, and thus, to an accelerated uptake of iron.

Materials and methods

Bacterial strains and plasmids

The E. coli strains DJ480, DJ480ΔryhB (Massé and Gottesman, 2002), MC4100 and its derivative AM111 (MC4100hfq−) (Tsui et al, 1994), as well as TX3346miaB− (Esberg et al, 1999), have been described. E. coli strain H1941Δfur was obtained from Dr K Hantke, University of Tübingen. Cells were grown at 37°C in M9 medium supplemented with 0.2% glucose, 2 mM MgSO₄, 0.1 mM CaCl₂ and 10 μg/ml thiamine. For growth of TX3346, the M9 medium was additionally supplemented with 0.05% casamino acids (Difco). Ampicillin (100 μg/ml), kanamycin (25 μg/ml) or chloramphenicol (15 μg/ml) was added to the medium where appropriate.

The construction of plasmids serving as templates for in vitro mRNA synthesis (see Figure 3B) and that of plasmids used to monitor uof-lacZ, fur-lacZ, acnA-lacZ, sodB-lacZ and mdh-lacZ expression, respectively, as well as that of plasmid variants (see Figures 3, 4 and 5) is described in Supplementary data. Transcription of the fusion genes was controlled by the lacpo instead of the natural fur promoter in plasmids used to monitor in vivo uof-lacZ or fur-lacZ expression upon addition of dip. This was performed to study regulatory events at the post-transcriptional level, that is, to avoid overlapping regulatory effects resulting from iron depletion on expression of the respective fusion genes. The copy numbers of all isogenic plasmids used in the in vivo experiments were found to be identical (not shown).

RNA preparation for in vitro studies

For synthesis of full-length fur mRNA and variants thereof, the plasmids pUuof_AUGfur, pUuof_UA-CUGfur, pUuof₅fur, pUuof₇₃fur or pUuof_AUGfur-inv were linearized with EcoRI and used as templates for in vitro transcription with T7 RNA polymerase (Promega). In vitro transcription of plasmid pUuof_AUGfur linearized with DraI generated fur242 mRNA, comprising nucleotides −183 to +59, which was used for toeprinting (see below). RyhB, RyhBΔ2 and RyhB-inv RNAs were transcribed with T7 polymerase using plasmids pURyhB, pURyhBΔ2 and pURyhB-inv (Večerek et al, 2003) cleaved with DraI. RyhBΔ3 was transcribed using plasmid pURyhBΔ3 (see Supplementary data), after cleavage with SmaI. DsrA RNA (Sledjeski et al, 1996) was transcribed in vitro from PCR templates.

Determination of fur mRNA steady-state levels and stability

To determine the steady-state levels of fur mRNA, cells were grown in M9 medium to an OD₆₀₀ of 0.5, when dip (250 μM final concentration) was added. Four-milliliter aliquots were withdrawn at time 0 (before addition of dip) and 2, 4, 8 and 16 min thereafter, and total RNA was isolated using standard procedures. To determine fur mRNA stability, the strains DJ480 and DJ480ΔryhB were grown in M9 medium at 37°C to an OD₆₀₀ of 0.5. Then, rifampicin (200 μg/ml) was added and 4-ml aliquots were withdrawn at time 0, 1, 2 and 4 min after addition of rifampicin for total RNA isolation. fur mRNA levels were determined by primer extension with the AMV reverse transcriptase (Promega) using 5 μg of total RNA primed with the fur-specific 5′-end labelled oligonucleotide E29 (5′-GTCATGCGGAATC-3′), which is complementary to nts −8 to +5 of fur mRNA. In addition, the 5S rRNA levels (loading control) were determined using primer R25 (5′-ATGCCTGGCAGTTCCCTACT-3′). The samples were separated on an 8% polyacrylamide–8 M (PAA–8 M) urea gel. The gels were dried and exposed to a Molecular Dynamics PhosphoImager for quantitation.

In vitro translation and coupled in vitro transcription/translation

Full-length fur mRNA and variants thereof (4 pmol) were translated in vitro using an S30 extract for circular templates (Promega), which contains Hfq (Večerek et al, 2003). Plasmid pRuof-lacZ was used as a template for coupled transcription/translation in the presence or absence of increasing concentrations of RyhB and DsrA RNAs, as specified in Figure 4B. Upon incubation for 20 min at 37°C in the presence of 6 μM ¹⁴C-lysine (309 mCi/mmol, Amersham Pharmacia Biotech), translation was stopped by the addition of four volumes of 90% acetone. The samples were resuspended in SDS–protein sample buffer and loaded onto a 12% SDS–polyacrylamide gel. Gels were dried and exposed to a Molecular Dynamics PhosphoImager for visualization.

Toeprinting analysis

The toeprinting assays on fur242 mRNA were carried out using purified 30S ribosomal subunits and initiator-tRNA, tRNA_f^Met, essentially as described by Hartz et al (1988). Briefly, ³²P-5′end-labelled oligonucleotide complementary to nucleotides +41 to +56 of fur mRNA served as the primer for cDNA synthesis. fur242 mRNA (0.04 pmol) were incubated for 5 min at 37°C in the absence (extension) and presence of 2 pmol 30S subunits (binary complex), and of 2 pmol 30S and 10 pmol tRNA_f^Met (ternary complex), respectively, before addition of MMLV reverse transcriptase and dNTPs. The samples were separated on an 8% PAA–8M urea gel and the extension signals were visualized using a Molecular Dynamics PhosphoImager.

β-galactosidase assays

The different strains were incubated at 37°C in M9 minimal medium. At an OD₆₀₀ of 0.5, dip (250 μM final concentration) was added to the culture. At time 0 and 5, 10 and 20 min after addition of dip, triplicate samples were withdrawn and β-galactosidase activities were determined as described elsewhere, and then averaged.

In vivo synthesis of the FurΦLacZ,UofΦLacZ and Fur proteins

Strains were grown in M9 medium at 37°C to an OD₆₀₀ of 0.5. Then, 250 μM dip (final concentration) was added and 1 ml samples were withdrawn before addition and at different times (see Figures) after addition of the chelator. Pulse labeling was performed by the addition of 1 μl [³⁵S]methionine (10 mCi/ml, Amersham Pharmacia Biotech) for 2 min at 37°C, and then cold methionine (10 mM final concentration) was added. Upon precipitation with 5% TCA, the pellets were washed with cold acetone (90%), resuspended in protein sample buffer and boiled for 5 min before loading onto a 12% SDS–polyacrylamide gel. For determination of the relative de novo synthesis of the FurΦLacZ and UofΦLacZ proteins (Figure 3C, lanes 2 and 6 and Figure 5B) the Phosphoimager signals corresponding to fusion proteins were normalized to the total counts obtained from the corresponding lane.

De novo-synthesized Fur protein (Figure 5A) was visualized with rabbit anti-Fur antibodies upon blotting of the gel-separated labeled proteins onto a nitrocellulose membrane employing standard procedures. The same membrane was then subjected to autoradiography and the signals obtained with anti-Fur antibody were merged with those obtained by autoradiography to identify de novo synthesized Fur protein.

Translational efficiency of uof₅-lacZ and uof₇-lacZ

The MC4100 miaB+ and TX3346miaB⁻ strains carrying plasmids pRuof₅-lacZ and pRuof₇-lacZ, respectively, were cultivated in M9 medium. At an OD₆₀₀ of 0.5, triplicate aliquots were taken for the β-galactosidase assays. In parallel, samples were withdrawn for isolation of total RNA to determine the respective uof₅-lacZ and uof₇-lacZ mRNA levels. β-Galactosidase activities obtained with the pRuof₅-lacZ and pRuof₇-lacZ constructs of two independent experiments were averaged. The steady-state uof₅-lacZ and uof₇-lacZ mRNA levels were determined by primer extension with AMV reverse transcriptase (Promega) using 5 μg of total RNA primed with the lacZ-specific 5′-end labeled probe (5′-GGGAAGGGCGATCGGT-3′). 5S rRNA, the levels of which were determined using primer R25 (see above), was used as loading control.

Supplementary Material

Supplementary data

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4