mRNA Secondary Structure Modulates Translation of Tat-Dependent Formate Dehydrogenase N

Abstract

Formate dehydrogenase N (FDH-N) of Escherichia coli is a membrane-bound enzyme comprising FdnG, FdnH, and FdnI subunits organized in an (αβγ)₃ configuration. The FdnG subunit carries a Tat-dependent signal peptide, which localizes the protein complex to the periplasmic side of the membrane. We noted that substitution of the first arginine (R₅) in the twin arginine signal sequence of FdnG for a variety of other amino acids resulted in a dramatic (up to 60-fold) increase in the levels of protein synthesized. Bioinformatic analysis suggested that the mRNA specifying the first 17 codons of fdnG forms a stable stem-loop structure. A detailed mutational analysis has demonstrated the importance of this mRNA stem-loop in modulating FDH-N translation.

The respiratory formate dehydrogenase N (FDH-N) enzyme of Escherichia coli is a seleno-molybdoenzyme that is synthesized when the bacterium grows anaerobically with nitrate as exogenous electron acceptor. FDH-N can comprise up to 10% of the total membrane protein (9). Together with nitrate reductase-A, it forms a respiratory chain transferring electrons from formate to nitrate and results in the generation of a protonmotive force (10). FDH-N has a number of cofactors, including bis-molybdopterin guanine dinucleotide cofactor, selenocysteine, and a single [4Fe-4S] cluster. Consequently, synthesis of this enzyme requires careful control. Transcription of both the fdnGHI operon, which encodes FDH-N, and narGHJI, which encodes nitrate reductase-A, is coordinately controlled. Expression of both operons is maximal anaerobically in the presence of nitrate and is controlled by the transcription factors Fnr and NarL (8).

The high-resolution X-ray structure of FDH-N has revealed that it adopts an (αβγ)₃ “trimer-of-trimers ” architecture, with the active site of the enzyme located in the periplasm (13). FDH-N is translocated across the membrane by the twin arginine translocation (Tat) pathway (21). The Tat translocase is dedicated to the transport of prefolded proteins, which bear an N-terminal signal peptide with the conserved S/T-R-R-x-F-L-K twin arginine motif (2, 3). Translocation of FDH-N is mediated by virtue of a Tat signal peptide on the FdnG subunit. Stanley et al. have shown previously that the FdnG signal peptide is able to mediate export of the reporter proteins β-lactamase and chloramphenicol acetyltransferase to the periplasm in a Tat-dependent fashion (24). During the course of these fusion studies we noted that point mutations in the first or second arginine codons of the twin arginine motif resulted in a dramatic overproduction of the fusion protein. The mRNA specifying the first 17 codons of the fdnG gene is predicted to fold into a stable stem-loop structure, and we demonstrate here that this stem-loop mediates translational control of FDH-N synthesis.

As shown in Table 1, substitution of Arg₅, either conservatively for Lys or nonconservatively for Ser, resulted in a marked increase (up to 60-fold) in fusion protein synthesis, regardless of the nature of the reporter protein. In contrast, substitution of Lys₁₀ for Glu had no significant effect on levels of fusion protein (Table 1). No significant differences in fusion protein synthesis were noted in a tat mutant background, suggesting that these observations were not directly related to operation of the Tat pathway. When the mfold program (25) was used, the mRNA covering the start codon and signal peptide-coding region of fdnG could be folded into a stem-loop structure (Fig. 1B). The ΔG value (at 37°C and physiological pH) associated with this structure was calculated to be −12.6 kCal/mol, suggesting that the folded mRNA would be relatively stable. Such a folded mRNA structure would be consistent with the results seen in Table 1, since mutations at codon 5, which fall within one arm of the predicted stem, would be expected to disrupt the fold, whereas the mutation at codon 10, which is located within the putative loop region, would not be expected to disrupt the structure.

TABLE 1.

Total cellular activity of reporter proteins fused to the wild-type or amino acid-substituted FdnG signal sequence^a

Signal sequence	Activity (μmol min−1g−1)
	β-Lactamaseb	Chloramphenicol acetyltransferasec	β-Galactosidased
FdnG wild type	1.50	0.70	3.90
R5S mutation	34.0	42	NDe
R5K mutation	60	ND	112
K10E mutation	1.85	0.90	ND

^aSite-specific mutations were introduced into plasmids pVJS2213 [Φ(fdnG-lacZ)_K44], pVJS2245 [Φ(fdnG-cat)_K44], and pVJS2248 [Φ(fdnG-bla)_K44] (24) by the use of standard procedures (20). Plasmids were transformed into MC4100 (6), and cells were cultured anaerobically in Luria-Bertani medium (20) supplemented with 0.2% glucose and 0.4% potassium nitrate.

^bActivity levels were determined from crude cell extracts and are expressed as micromoles of PADAC hydrolyzed per minute per gram of cells (19).

^cActivity levels were determined from crude cell extracts and are expressed as micromoles of thionitrobenzoate produced per minute per gram of cells (23).

^dActivity levels were determined from crude cell extracts and are expressed as micromoles of o-nitrophenol produced per minute per gram of cells (17).

^eND, not determined.

FIG. 1.

The mRNA specifying the fdnG signal peptide coding region can be folded into a stem-loop structure. The position of the initiation codon is indicated in bold type; the Shine-Dalgarno ribosome-binding sequence is underlined. Numbering is shown relative to the first base of the start codon, which is designated +1. (A) Portion of the DNA from plasmid pVJS2248 (24) showing the first part of the fdnG signal peptide-coding region. To aid clarity, the amino acids are numbered with subscripts. The amino acids of the twin arginine motif are doubly underlined. The clone carries a 487-bp insert that covers the entire fdnG promoter region and fdnG sequence as far as an engineered BamHI site in the DNA after codon 44 (24). (B) The mRNA covering the region shown in panel A was folded using the program mfold 3 (25), which was accessed through the world-wide web (http://www.bioinfo.rpi.edu/applications/mfold). The codon substitutions used during this study are V₃P (GTC→CCC), R₅K (CGC→AAG), R₅S (CGC→AGC), R₅Rhigh (CGC→CGT), R₅Rlow (CGC→AGG), R₆K (AGA→AAG), R₆Rhigh (AGA→CGT), F₉L (TTT→CTG), K₁₀E (AAA→GAA), I₁₁V (ATC→GTC), I₁₁I (ATC→ATA), I₁₁T (ATC→ACT), C₁₂A (TGC→GCC), A₁₃S (GCG→AGC), G₁₄K (GGC→AAA), and R₆K/I₁₁T (AGA→AAG and ATC→ACT).

To test whether a hairpin in the fdnG mRNA was controlling expression at a translational level, we constructed a number of additional mutations in codons 3 to 14 of fdnG. As shown in Fig. 2A, substitutions that were predicted to severely disrupt the mRNA fold (R₅S [CGC→AGC], R₅K [CGC→AAG], R₅Rhigh [CGC→CGT], and R₅Rlow [CGC→AGG], where the natural Arg codon was replaced with Arg codons of higher and lower usage) (18) led to significant upregulation of β-galactosidase activity. Mutations in codon 6 that were predicted to have more modest effects on the mRNA secondary structure gave less-dramatic increases in β-galactosidase activity. As predicted by the model, substitutions of codons that fell within the putative loop region (Fig. 2B) did not have significant effects on the activity of β-galactosidase, with the exception of the I11V mutation, which resulted in a marked decrease in the activity of β-galactosidase. Interestingly, this substitution is predicted to result in a significant increase in the stability of the stem-loop (from −12.6 to −16.4 kCal/mol) due to the formation of an extra base pair at the top of the stem. The results of substitutions in the putative second arm of the stem (Fig. 2C) were also consistent with the premise that a stem-loop structure was controlling translation of fdnG-lacZ. A plot of ΔG value for the structures associated with the wild-type sequence and each of the mutations against the observed β-galactosidase activity (Fig. 2E) shows a linear relationship. Indeed, a straight line can be drawn through the points with a correlation factor (R²) of 0.82, providing strong support for the model. As a final test, we looked at the effect of introducing a compensatory mutation into the second arm of the stem, to restore base pairing, with a mutation in the first arm of the stem that disrupts the stem-loop structure. When introduced singly, both the AGA→AAG mutations at codon 6 and the ATC→ACT mutations at codon 11 led to increased β-galactosidase activity (Fig. 2D). However, when these two sets of mutations were combined, β-galactosidase activity was restored to the wild-type level, indicating that the two mutations had a compensatory effect.

FIG. 2.

The effects of mutations in the stem and loop regions of the fdnG mRNA on expression of fdnG-lacZ. Constructs were transformed into strain MC4100-P (4), which carries the pcnB allele, providing more careful control of plasmid copy number (14, 15). (A) Constructs carrying mutations in the codons forming the first arm of the stem; (B) constructs carrying mutations in codons that fall in the loop region; (C) constructs carrying mutations in codons forming the second arm of the stem; (D) constructs carrying either individual single codon substitutions or a combined double-codon substitution that is predicted to have a compensatory effect. All strains were cultured anaerobically in Cohen and Rickenberg medium (7) supplemented with 0.2% glucose and 0.4% nitrate. β-Galactosidase enzyme activity is shown in Miller units (17), and the bars represent standard errors of the means (n = 3 to 4). (E) Correlation between stem-loop stability (measured as ΔG) and β-galactosidase enzyme activity. Free energies for each codon mutation were calculated with the program mfold3 (25) using refined energy parameters (16) and with the mRNA sequence covering codons 1 to 17 inclusive. WT, wild type.

In addition to the substantial transcriptional regulation already reported (1), our findings strongly suggest that a further, possibly up to 10-fold, level of control of fdnGHI expression was exerted at the translational level. To confirm the physiological relevance of these findings, we used published methods (11) to introduce two sets of mutations (R₅Rhigh [CGC→CGT] and R₅Rlow [CGC→AGG]) in the fdnG gene in the chromosome. In each case we made a substitution that retained an Arg codon at position 5 so that the cellular location of FDH-N would not be compromised. To circumvent possible translational errors associated with use of rare arginine codons, such as AGG, in E. coli (see, e.g., reference 5), we additionally introduced into each strain plasmid pUt-AGA/AGG, which carries the genes encoding tRNA^Arg_UCU and tRNA^Arg_UCC (22). In the presence of this plasmid, but not in its absence, we saw an approximate fourfold increase in FDH-N activity from the strain with the R₅Rlow chromosomal replacement of fdnG (Fig. 3A) and a corresponding overproduction of the FdnG polypeptide (Fig. 3C). This result strongly suggests that the translational control of fdnG expression is of physiological significance, since we have clearly demonstrated that the cell has the capacity to synthesize, assemble, and export fourfold-higher levels of FDH-N. Interestingly, as shown in Fig. 3B the fourfold overproduction of FDH-N activity was associated with a marked increase in NAR activity. This suggests that there might be a further level of coordinate control of FDH-N and NAR synthesis not previously observed.

FIG. 3.

The stem-loop structure regulates FDH-N activity in vivo. FDH-N (A) and NAR (B) enzyme assays were performed with crude cell extracts derived from strains MC4100, NRS9 (as MC4100; FdnG R₅R [CGC→CGT]), and CMP1 (as MC4100; FdnG R₅R [CGC→AGG]) or the same strains carrying plasmid pUt-AGA/AGG (22). Cells were grown anaerobically in Luria-Bertani medium supplemented with 0.2% glucose and 0.4% potassium nitrate, crude extracts were prepared, and enzyme activities were assayed exactly as described previously (21). Assays were performed in triplicate with results differing by not more than 15% of the mean. (A) FDH-N enzyme activity (measured as the rate of phenazine methosulfate-linked formate dehydrogenase activity) (9). An activity level of 100% is that measured from the crude cell extract of the wild-type (WT) strain and corresponds to 33 nmol of formate oxidized/min/mg of protein. (C) NAR enzyme activity, measured according to the method of Jones and Garland (12). An activity level of 100% is that measured in the crude cell extract of the wild-type strain and corresponds to 598 nmol of nitrate reduced/min/mg of protein. (C) Detection of FdnG polypeptide in whole cells after Western blot analysis with anti-FdnG antibodies (24). Equivalent quantities of cells were loaded in each lane. The fold overexpression of FdnG polypeptide relative to that detected in the wild-type sample is shown above each lane.

In conclusion, we have demonstrated that there is control of FDH-N synthesis at the translational level. Interestingly, the stem-loop structure seen here for the E. coli mRNA is also predicted to be conserved in the fdnG coding regions of several other bacteria, including Salmonella enterica serovar Typhimurium, Haemophilus influenzae, and Pseudomonas aeruginosa. Notably, however, the stem-loop structure is apparently not conserved over the similar coding region of E. coli fdoG. While the mechanism underlying modulation of FDH-N translation is presently unclear, it may potentially serve as a quality control feature ensuring that only the fully assembled complex is transported to the periplasm.