ABSTRACT
Selenoprotein synthesis in Escherichia coli strictly depends on the presence of a specific selenocysteine insertion sequence (SECIS) following the selenocysteine-encoding UGA codon of the respective mRNA. It is recognized by the selenocysteine-specific elongation factor SelB, leading to cotranslational insertion of selenocysteine into the nascent polypeptide chain. The synthesis of three different selenoproteins from the gram-positive anaerobe Eubacterium acidaminophilum in E. coli was studied. Incorporation of 75Se into glycine reductase protein B (GrdB1), the peroxiredoxin PrxU, and selenophosphate synthetase (SelD1) was negligible in an E. coli wild-type strain and was fully absent in an E. coli SelB mutant. Selenoprotein synthesis, however, was strongly increased if selB and selC (tRNASec) from E. acidaminophilum were coexpressed. Putative secondary structures downstream of the UGA codons did not show any sequence similarity to each other or to the E. coli SECIS element. However, mutations in these structures strongly reduced the amount of 75Se-labeled protein, indicating that they indeed act as SECIS elements. UGA readthrough mediated by the three different SECIS elements was further analyzed using gst-lacZ translational fusions. In the presence of selB and selC from E. acidaminophilum, UGA readthrough was 36 to 64% compared to the respective cysteine-encoding UGC variant. UGA readthrough of SECIS elements present in Desulfomicrobium baculatum (hydV), Treponema denticola (selD), and Campylobacter jejuni (selW-like gene) was also considerably enhanced in the presence of E. acidaminophilum selB and selC. This indicates recognition of these SECIS elements and might open new perspectives for heterologous selenoprotein synthesis in E. coli.
The 21st amino acid selenocysteine (Sec) is present in a variety of different eukaryotic, archaeal, and bacterial proteins (11, 25, 26, 29, 50). The insertion of selenocysteine into growing peptide chains is directed by an opal (UGA) codon (8, 51), which in the standard genetic code signals termination of translation. Selenoprotein synthesis in Escherichia coli has been elucidated and intensively studied by Böck and coworkers (6) and requires the products of the genes selA, selB, selC, and selD. selC encodes a selenocysteine-specific tRNASec that is first charged with serine by seryl-tRNA synthetase. It is converted to selenocysteyl-tRNASec by selenocysteine synthase, the gene product of selA. A low-molecular-weight selenium donor, selenophosphate, is required for this reaction. It is synthesized from ATP and selenide by selenophosphate synthetase, SelD (41).
Recoding of UGA into selenocysteine requires a special mRNA motif, the selenocysteine insertion sequence (SECIS) (22). In mRNAs encoding bacterial selenoproteins, it is located at the 3′ site of the UGA codon (52). This mRNA structure is recognized by the bacterial selenocysteine-specific elongation factor SelB, which functions homologously to the standard elongation factor EF-Tu (9, 24). SelB binds GTP and selenocysteyl-tRNASec and recognizes the SECIS element, thereby donating the selenocysteyl-tRNASec in close proximity to the UGA codon at the ribosomal A site. Binding of bacterial SelB to the SECIS element is conferred in a highly sequence- and structure-specific manner by a C-terminal extension (domain IV) that is not present in EF-Tu (10, 19, 24, 48).
In eukaryotes and archaea the SECIS element is located in the 3′ untranslated region of the selenoprotein mRNA (5, 22, 37). Consequently, selenocysteine can be inserted at any UGA codon present in the coding region of the mRNA, irrespective of the adjacent sequence.
SECIS elements from distantly related bacteria show only little sequence conservation. In fact, the selenocysteine-specific elongation factor from E. coli is highly specific for the SECIS element of the three formate dehydrogenase isoenzymes, which are the only selenoproteins present in this organism (17, 28). Therefore, recombinant selenoprotein production is difficult, with the main barrier being a proper interaction between E. coli SelB and a given selenoprotein mRNA (12, 13, 21, 43). Insertion of selenocysteine at a specific site into a protein can be achieved only by using a hairpin structure which strongly resembles the fdhF SECIS element at a proper distance downstream of a UGA codon (3, 7, 16, 18). However, sequence requirements for the SECIS element would usually result in amino acid changes in the respective protein.
Amino acid-fermenting gram-positive anaerobic bacteria contain a variety of selenoproteins with different functions in the same organism. Eight different functional selenoproteins have been identified in the amino acid-fermenting Eubacterium acidaminophilum, with up to three gene copies for some of them (1, 2). Putative SECIS elements following the UGA codon do not show significant sequence homology to each other or to the E. coli SECIS element. Only one selB gene copy is present in the genome, being part of a selDABC cluster (15).
In the present study, the synthesis of three different selenoproteins from E. acidaminophilum in E. coli was analyzed. When selB and selC from E. acidaminophilum are provided on a second plasmid, selenoproteins are formed. To our knowledge this is the first report on the synthesis of selenoproteins from gram-positive bacteria in E. coli.
MATERIALS AND METHODS
Materials.The molecular mass marker for polyacrylamide gel electrophoresis (PAGE) was obtained from Sigma (Taufkirchen, Germany). Vectors pASK-IBA3 and pASK-IBA5, Strep-Tactin, and the respective horseradish peroxidase conjugate for detection of the Strep-tag II were received from IBA (Göttingen, Germany). Vector pACYC184 was from New England Biolabs (Frankfurt, Germany). Restriction enzymes were obtained from New England Biolabs and MBI-Fermentas (St. Leon-Roth, Germany), and oligonucleotides were from Invitrogen (Karlsruhe, Germany). [75Se]selenite (176 mCi/mmol) came from the Risø National Laboratory Isotope Division (Roskilde, Denmark).
Bacterial strains, media, and growth conditions.The strains used in this study were E. coli XL1-Blue MRF′ from Stratagene and E. coli WL81300 (44). Cultures were grown aerobically at 37°C in LB medium containing ampicillin (100 μg/ml) and chloramphenicol (40 μg/ml) if required. Cultures used for studying the heterologous expression of selenoproteins were supplemented with 2 μM sodium selenite.
For β-galactosidase assays, cells were grown in TP medium (1% tryptone, 0.5% yeast extract, 0.5% glycerol, 100 mM potassium phosphate [pH 6.5], 1 mM MgSO4, 0.1 mM CaCl2, 0.4 μM H3BO3, 30 nM CoCl2, 10 nM CuSO4, 10 nM ZnSO4, 80 nM MnCl2, 10 μM FeCl3) supplemented with 2 μM sodium selenite and the antibiotic concentrations given above.
General DNA techniques.Basic recombinant DNA methods were carried out as described previously (38). Enzymes were used according to the recommendations of the manufacturer. Plasmids were prepared using the E.Z.N.A. plasmid miniprep kit I (Peqlab, Erlangen, Germany) or the plasmid midikit (Qiagen, Hilden, Germany). Sequencing reactions were performed on the ABI Prism 377 automated laser fluorescence sequencer (Applied Biosystems, Langen, Germany).
Construction of plasmids.To clone grdB1 and selD1 into the expression vectors pASK-IBA3 and pASK-IBA5, respectively, the genes were amplified with Pwo DNA polymerase from plasmid pDGS1 (45) and chromosomal DNA from E. acidaminophilum (selD1). All primers contained a BsaI restriction endonuclease site, and the amplified DNA fragments were cloned into the respective sites of the vector, generating translational fusions with an C-terminal (grdB1) or N-terminal (selD1) Strep-tag II peptide. The resulting plasmids were designated pIBA3B and pSD1N, respectively. The exchange of the selenocysteine codon UGA in both genes for the cysteine codon UGC was performed following the QuikChange site-directed mutagenesis kit protocol (Stratagene, Heidelberg, Germany). Briefly, plasmids pIBA3B and pSD1N were used as templates in a PCR with an oligonucleotide pair carrying the respective mutation in the middle. The amplification was performed over 16 cycles with Pwo DNA polymerase. The reaction mixture was incubated with DpnI to degrade the methylated template DNA and then transformed into E. coli XL1-Blue MRF′. The mutant plasmids were controlled by sequencing the inserts and designated pIBA3MBU and pMUD15. Construction of the pASK-IBA3 derivatives pPRXU (prxU) and pPRXU47C (prxU, cysteine mutant) were described previously (40).
The exchange of specific nucleotides in the postulated SECIS elements of grdB1, prxU, and selD1 was performed by site-directed mutagenesis as described above. The plasmids pIBA3B, pIBA3MBU, pPRXU, pPRXU47C, pSD1N, and pMUD15 were used as templates in a PCR with oligonucleotide pairs (37 to 46 nucleotides) carrying the respective mutations (see Fig. 4). Mutant plasmids were controlled by sequencing the insert and were designated pGBMS11 and pGBML2 (grdB1, mutants MS and ML), pPUMS1 and pPUML2 (prxU, mutants MS and ML), and pSDMS11 and pSDML2 (selD1, mutants MS and ML). Plasmids pGBMS11, pPUMS1, and pSDMS11 were used as templates to generate plasmids pGBMR3, pPUMR3, and pSDMR1 (mutants MR).
For construction of plasmid pASBC4, a fragment containing selB and selC from E. acidaminophilum was amplified by PCR from plasmid pTN2 (15) using Pwo DNA polymerase. Restriction sites for BamHI (forward primer) and SalI (reverse primer) were introduced, and the PCR product was cloned into the respective sites of the plasmid vector pACYC184 (New England Biolabs, Frankfurt, Germany).
To generate fusions of the postulated SECIS elements of grdB1, prxU, and selD1 to the glutathione S-transferase (GST) gene (gst) and the β-galactosidase gene (lacZ), a region spanning 10 nucleotides upstream and 50 nucleotides downstream of the respective UGA codon was amplified from plasmid pIBA3B, pPRXU, or pSD1N using Taq DNA polymerase. The resulting fragment was cloned into the HindIII and ApaI restriction sites of the plasmid vector pSKAGS (46). The gst-′lacZ fusions of the (postulated) SECIS elements of hydV (Desulfomicrobium baculatum), selD (Treponema denticola), and the selW-like gene (Campylobacter jejuni) were generated in the same way, except that the respective gene regions were amplified from synthetic oligonucleotides.
Recombinant gene expression and analysis.Heterologous gene expression was induced at an optical density at 550 nm of 0.5 by adding anhydrotetracycline to a final concentration of 0.2 μg/ml. Cells were harvested after 3 h by centrifugation (2 min, 10,000 × g) and were lysed for 5 min at 95°C in sodium dodecyl sulfate (SDS) sample buffer (62.5 mM Tris-HCl [pH 6.8], 25% [vol/vol] glycerol, 2.5% [vol/vol] β-mercaptoethanol, 0.005% [wt/vol] bromophenol blue, 1% [wt/vol] SDS), using 100 μl per ml culture and unit of optical density at 550 nm. Proteins of whole-cell lysates were separated by SDS-PAGE and subsequently blotted onto polyvinylidene difluoride (PVDF) membranes. The recombinant gene products containing the Strep-tag II octapeptide were visualized by the binding of Strep-Tactin conjugated to horseradish peroxidase. For 75Se labeling, the medium was supplemented with l-cysteine (30 μg/ml), and [75Se]selenite (2 μM, 0.3 μCi/ml) was added immediately after induction of heterologous gene expression with anhydrotetracycline. Cells were harvested after 1 and 3 h by centrifugation (2 min, 10,000 × g), and cell pellets were washed twice with medium. SDS-PAGE of whole-cell lysates was performed as described above. Gels were stained with Coomassie blue (0.2% [wt/vol] Serva blue R250, 0.05% [wt/vol] Serva blue G250, 42.5% [vol/vol] ethanol, 5% [vol/vol] methanol, 10% [vol/vol] acetic acid), dried, and exposed to a PhosphorImager screen (Molecular Dynamics, Krefeld). The intensity of radioactive bands was quantified using the ImageQuant software (Amersham Biosciences, Buckinghamshire, United Kingdom). All values were corrected for the intensity of the major labeled tRNA band in the respective lane to compensate for differences in the amount of added radioactivity and gel loading. Alternatively, bands were excised from Coomassie blue-stained gels, and radioactivity was measured using a liquid scintillation counter (LS 6500; Beckman, Palo Alto, CA).
Nucleotide and amino acid sequence analysis.RNA secondary structures were calculated using the mfold server available at http://www.bioinfo.rpi.edu/applications/mfold/old/rna/ (53). Amino acid sequences were analyzed using the tools available from the Expasy Molecular Biology Server (http://www.expasy.ch .).
Expression and purification of PrxU.Expression of prxU was performed in E. coli XL1-Blue MRF′ as described above. Three hours after induction with anhydrotetracycline, cells from a 500-ml culture were harvested by centrifugation and were frozen at −20°C. Cells pellets were suspended in 5 ml buffer W (50 mM Tris-HCl [pH 7.6], 150 mM NaCl, 1 mM EDTA, 3 mM dithiothreitol) containing 1 mg/ml lysozyme and 0.1 μM phenylmethylsulfonyl fluoride and were incubated on ice for 30 min. Cells were broken by two passages through a French pressure cell (20K-cell; SLM Instruments, Rochester, NY) at 1,260 lb/in2. After addition of 5 μg/ml DNase I and 25 μg/ml avidin, lysates were incubated on ice for 30 min and were subsequently centrifuged at 60,000 × g for 20 min at 4°C. The supernatants were centrifuged once again using the same parameters. The resulting crude extract was applied to a 1-ml Strep-Tactin Sepharose column equilibrated with buffer W. After collection of the flowthrough, the column was washed six times with 1 ml buffer W. The elution of bound Strep-tag II-containing protein was achieved by repetitive addition of 0.5 ml buffer W containing 2.5 mM α-desthiobiotin (Sigma). The purified protein PrxU was >95% pure as judged by SDS-PAGE. Prior to mass spectrometry, PrxU was desalted by four gel filtration steps using protein desalting spin columns (Pierce, Rockford, IL) according to the instructions of the manufacturer.
Mass spectrometry.The spectra were acquired with an ESI-Q-TOF 2 mass spectrometer (Micromass, Manchester, United Kingdom) equipped with a nanospray source. The samples were injected via a PicoTip (New Objective, Cambridge, MA) with a syringe pump (Harvard Apparatus, MA) at a flow rate of 300 nl/min. The MaxEnt1 algorithm was used for deconvoluting the data to single-charge state.
Assay of β-galactosidase activity.Bacterial cells transformed with the gst-′lacZ fusion constructs were grown at 37°C in TP medium supplemented with 2 μM sodium selenite. At an A600 of 0.3 the expression of the fusion genes was induced by addition of 50 μM isopropyl-β-d-thiogalactopyranoside (IPTG). The cultures were grown for 2 h at 37°C and were subsequently chilled on ice for 5 min. Cells were harvested by centrifugation at 10,000 × g for 2 min at 4°C and frozen at −20°C. β-Galactosidase activity was determined using a slightly modified protocol described previously (32). Briefly, cell pellets corresponding to 0.1 to 1.0 ml culture were suspended in 950 μl Z buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4, 50 mM β-mercaptoethanol) and were incubated at room temperature for 10 min. Permeabilization of the cells was achieved by adding 50 μl 0.25% (wt/vol) cetyltrimethylammonium bromide and 0.5% (wt/vol) sodium deoxycholate in Z buffer. After 10 min at room temperature, 100 μl o-nitrophenyl-β-d-galactopyranoside solution (4 mg/ml in Z buffer) was added. The reaction was allowed to proceed until a light yellow color appeared and was subsequently stopped with 500 μl 1 M Na2CO3. Enzyme activity was determined as described previously (32).
Detection of GST fusion proteins.GST in cells expressing gst-′lacZ fusions was detected with an anti-GST peroxidase conjugate antibody (Sigma) using the GST Western blotting kit (Roche Diagnostics).
RESULTS
Coexpression of E. acidaminophilum selB and selC increases the amounts of GrdB1, PrxU, and SelD1 synthesized in E. coli.Three different selenoproteins from E. acidaminophilum were heterologously expressed in E. coli. GrdB1 (47 kDa) is the substrate-specific selenocysteine-containing subunit of a multicomponent enzyme complex called glycine reductase (1, 45). PrxU (22 kDa) is a selenocysteine-containing peroxiredoxin (40). SelD1 (36 kDa) is a selenocysteine-containing selenophosphate synthetase (15). Secondary structures that might act as SECIS elements are present in the respective mRNAs but do not show any sequence conservation (see Fig. 4). The genes were cloned downstream of the tetA promoter into the expression vectors pASK-IBA3 and pASK-IBA5, respectively, to give C-terminal (grdB1 and prxU) or N-terminal (selD1) fusions to the short peptide Strep-tag II.
The respective plasmids were transformed into E. coli XL1-Blue MRF′ and E. coli WL81300. The latter is a selB-deficient mutant unable to form selenocysteine-containing proteins. The synthesis of GrdB1, PrxU, and SelD1 in E. coli was studied at 3 h after addition of anhydrotetracycline either in the presence of a control plasmid (pACYC184) or in the presence of plasmid pASBC4, harboring the genes selB (encoding selenocysteine-specific elongation factor SelB) and selC (tRNASec) from E. acidaminophilum. Protein synthesis was monitored by Western blotting (Fig. 1).
Western blot detection of GrdB1, PrxU, and SelD1 after heterologous expression in E. coli. SDS lysates of whole cells were separated on a 12.5% SDS-polyacrylamide gel, followed by transfer onto a PVDF membrane. Strep-tag II-containing proteins were visualized using Strep-Tactin conjugated to horseradish peroxidase. Arrows mark the expected sizes of the proteins (48 kDa for GrdB1, 23 kDa for PrxU, and 37 kDa for SelD1). When indicated, selB and selC from E. acidaminophilum were coexpressed by providing these genes on plasmid pASBC4. E. coli WL81300 is a ΔselB mutant. An unspecific band detected in case of SelD1 was obtained after a longer exposure time compared to GrdB1 and PrxU.
In the absence of selB and selC from E. acidaminophilum, GrdB1 was produced in the E. coli wild-type strain, but no protein was synthesized in the selB mutant WL81300. When selB and selC from E. acidaminophilum were coexpressed (pASBC4), GrdB1 synthesis was strongly increased. PrxU was produced in the E. coli wild-type strain but also in the mutant strain, in the absence of selB and selC, indicating a suppression mechanism. However, again, coexpression of selB and selC from E. acidaminophilum strongly increased the amount of PrxU in both strains. SelD1 was not detectable in either of the two E. coli strains, but also in this case the protein became visible when selB and selC were coexpressed.
selB and selC from E. acidaminophilum promote selenium incorporation.Next, recombinant selenoprotein synthesis was analyzed by incorporation of 75Se, present as sodium selenite in the medium. Samples were taken at 1 and 3 h after induction (Fig. 2). As expected, no selenoproteins were synthesized in the selB-deficient strain E. coli WL81300 in the absence of sel genes from E. acidaminophilum. Labeled bands of low molecular mass in these extracts are due to the synthesis of selenium-containing tRNAs (47).
Labeling of recombinantly synthesized GrdB1, PrxU, and SelD1 by 75Se. E. coli strains were grown in LB medium in the presence of [75Se]selenite and were harvested at 1 or 3 h after addition of anhydrotetracycline. Autoradiograms of 12.5% SDS-polyacrylamide gels after electrophoretic separation of whole-cell lysates are shown. Arrows indicate the migration positions of proteins of the expected sizes (48 kDa for GrdB1, 23 kDa for PrxU, and 37 kDa for SelD1) and the selenium-containing tRNAs. When indicated, the genes selB and selC from E. acidaminophilum were provided on plasmid pASBC4.
In the E. coli wild-type strain XL1-Blue MRF′ in the absence of selB and selC from E. acidaminophilum, only minute amounts of selenium-containing proteins were present. Obviously, E. coli SelB could not properly interact with the respective SECIS mRNAs. However, all three selenoproteins were synthesized in the E. coli wild-type and selB mutant strains in the presence of selB and selC from E. acidaminophilum (pASBC4). In the case of GrdB1, additional selenium-labeled protein bands corresponding to masses of 36 kDa and less were found. These were most likely caused by a fragmentation of GrdB1, as previously observed (14, 45).
UGA suppression in prxU leads to tryptophan insertion.PrxU was strongly produced in both E. coli strains even in the absence of sel genes from E. acidaminophilum, indicating UGA suppression (Fig. 1). However, selenium incorporation is also indicated (Fig. 2). To specify which PrxU variants were produced in the presence and absence of selB and selC, the respective proteins were purified by Strep-tag II affinity chromatography, and the masses were determined by mass spectrometry. When prxU was expressed in E. coli XL1-Blue MRF′ in the absence of genuine selB and selC, the corresponding protein had a mass of 23,672 Da, which exactly matches a polypeptide containing tryptophan instead of selenocysteine (Fig. 3A). When PrxU was produced in the presence of pASBC4, a protein of this mass was still present, but two additional masses appeared (Fig. 3B). The mass of 23,634 Da points to a selenocysteine-containing protein (23,636 Da, assuming a protonated selenol). The lack of two hydrogens might be due to the presence of an intramolecular selenide-sulfide (Se-S) bond (selenenylsulfide). The third mass (23,573 Da) would correspond to a serine-containing variant which might arise either due to cotranslational insertion of this amino acid or due to a loss of selenium during sample processing (see Discussion).
Analysis of purified PrxU by mass spectrometry. The peroxiredoxin gene prxU was expressed in E. coli XL1-Blue MRF′ in the absence (A) or presence (B) of selB and selC from E. acidaminophilum and was purified as described in Materials and Methods. After desalting, the protein was subjected to mass spectrometry.
Mutations in the postulated SECIS structures of grdB1, prxU, and selD1 reduce selenium incorporation.Selenocysteine incorporation requires a special secondary structure in the mRNA (the SECIS element), which in bacteria directly follows the UGA codon. Putative SECIS elements present in grdB1, prxU, and selD1 are shown in Fig. 4A. mRNA folding was calculated using the mfold prediction algorithm (53). To see whether these computationally calculated mRNA secondary structures in fact act as SECIS elements, mutations were introduced into the apical loop and the adjacent stem. Respective regions of the E. coli fdhF SECIS structure have been shown to directly interact with SelB (17, 36, 48).
Effect of mutations in the postulated mRNA secondary structures on selenocysteine incorporation. (A) Wild-type and mutated forms (MS, ML, and MR) of the postulated stem-loop structures (SECIS) downstream of the selenocysteine codon UGA in mRNAs of grdB1, prxU, and selD1. The selenocysteine codon is marked in bold, and altered nucleotides are boxed. (B) Incorporation of 75Se into the corresponding proteins. Wild-type (WT) and mutant (MS, ML, and MR) grdB1, prxU, and selD1 were expressed in E. coli WL81300 in the presence of plasmid pASBC4, containing selB and selC from E. acidaminophilum. Cells were grown in the presence of [75Se]selenite and were harvested at 3 h after induction with anhydrotetracycline. Autoradiograms of 12.5% SDS-polyacrylamide gels after electrophoretic separation of whole-cell lysates are shown. Arrows indicate the expected size of the proteins (48 kDa for GrdB1, 23 kDa for PrxU, and 37 kDa for SelD1).
The SECIS-mutated grdB1, prxU, and selD1 genes were expressed in E. coli WL81300 (ΔselB) in the presence of pASBC4, carrying selB and selC from E. acidaminophilum, as described above. The amounts of 75Se-labeled proteins obtained from these mutant genes were compared to those obtained from the wild-type genes by phosphorimager exposure and densitometric analysis (Fig. 4B).
The replacement of three nucleotides in the upper region of the stem was accompanied by a loss of three base pairings (mutant MS, Fig. 4A). This mutation strongly reduced the amounts of 75Se-labeled proteins, to 7% (grdB1), 46% (prxU), or 8% (selD1), compared to expression of the respective wild-type genes. When base pairing was restored by additional mutations in the opposite strand (mutant MR), selenoprotein synthesis was regained up to 94% (grdB1), 85% (prxU), and 59% (selD1).
The exchange of two nucleotides in the loop region of the postulated hairpin structure (mutant ML) reduced the incorporation of 75Se into the respective proteins to 15% (grdB1), 56% (prxU), and 36% (selD1). Similar results were obtained in a second experiment where the radioactivity of 75Se was determined by liquid scintillation of the respective excised bands from a polyacrylamide gel (not shown). In summary, all mutations in the putative SECIS structures strongly diminished selenoprotein synthesis, indicating that these elements play a critical role during UGA decoding.
SECIS elements of grdB1, prxU, and selD1 mediate UGA readthrough with different efficiencies.If these short mRNA secondary structures are critical and sufficient for UGA decoding as selenocysteine, then these elements should also allow UGA readthrough in the context of other translational fusions. Fragments containing 10 nucleotides upstream and 50 nucleotides downstream of the selenocysteine codon (UGA) in grdB1, prxU, and selD1 were cloned in frame into the gst-′lacZ gene fusion of plasmid pSKAGS (Fig. 5A). The respective fusion constructs were transformed into E. coli XL1-Blue MRF′ containing either plasmid pASBC4 or the empty vector pACYC184. Fusion protein synthesis was monitored by assaying β-galactosidase activity after induction of fusion gene expression by IPTG (Fig. 5B). For each construct, a cysteine (UGC)-encoding variant served as a control. LacZ activity determined for the wild-type (UGA) fusion was only 12% (grdB1), 2% (prxU), or 4% (selD1) of that for the respective cysteine variant (100%) in the absence of selB and selC from E. acidaminophilum. Upon coexpression of genuine selB and selC, β-galactosidase activity for the UGA fusions reached 60% (grdB1), 64% (prxU), or 36% (selD1), whereas the β-galactosidase activity of the cysteine variants (UGC) was not significantly changed (106%, 98%, and 92%, respectively). Western blotting was done with an anti-GST antibody that detects both the product of translational termination at UGA (27 kDa) and the full-length fusion protein (145 kDa). As shown in Fig. 5C, no full-length fusion protein was detected when expression was performed in the absence of selB and selC from E. acidaminophilum. In the presence of pASBC4, the full-length protein was present, as well as the prematurely terminated protein.
UGA readthrough in the presence of SECIS elements of grdB1, prxU, and selD1. (A) Translational fusions were constructed encompassing the GST gene (gst); the UGA codon and SECIS-containing region of grdB1, prxU, or selD1; and the β-galactosidase gene (lacZ). The inserts were cloned into the HindIII (H) and ApaI (A) sites of the vector pSKAGS. The expression of the fusions is under control of the lac promoter (Plac). As a control, similar fusions were constructed containing a cysteine (UGC) instead of a selenocysteine (UGA) codon. (B) β-Galactosidase activity in cell lysates after expression of the fusion constructs in E. coli XL1-Blue MRF′ in the presence or absence of pASBC4. Enzyme activities were determined 2 h after induction with IPTG. Values are given as a percentage of the value for the respective cysteine mutant in the absence of pASBC4 (selB and selC from E. acidaminophilum). Under these conditions, 100% readthrough corresponded to 11,212 Miller units β-galactosidase activity for grdB1, 8,232 units for prxU, and 10,471 units for selD1. Bars represent the means and standard deviations obtained with three different clones which were measured in duplicate. (C) Immunodetection of the GST-LacZ fusion proteins produced in E. coli XL1-Blue MRF′. Whole-cell lysates were separated by PAGE and transferred onto a PVDF membrane. Recombinant proteins were visualized using an anti-GST antibody. F, full-length product (145 kDa); T, termination product (27 kDa). The 60-kDa protein band is probably due to an unspecific cross-reaction with the E. coli protein extract.
Coexpression of E. acidaminophilum selB and selC mediates UGA readthrough in selenoprotein mRNAs from other bacterial species.Since SelB from E. acidaminophilum interacts with a variety of different selenoprotein mRNAs present in this organism, it might also enable UGA readthrough in SECIS elements from other bacterial species. Three different SECIS elements were chosen and derived from the selenoprotein-encoding genes hydV from D. baculatum and selD from Treponema denticola, and the selW-like gene from Campylobacter jejuni (Fig. 6A). DNA fragments containing 10 nucleotides upstream and 50 nucleotides downstream of the selenocysteine codon (UGA) were inserted into plasmid pSKAGS (see above). UGA readthrough was measured in E. coli XL1-Blue MRF′ in the absence or presence of pASBC4 and was compared to readthrough of the respective E. acidaminophilum prxU fusion construct.
UGA readthrough in the presence of SECIS elements from various bacterial species. (A) Putative SECIS elements downstream of the selenocysteine codon (UGA) in mRNAs of hydV from D. baculatum, selD from T. denticola, and the selW-like gene from C. jejuni. The selenocysteine codon is marked in bold. Translational fusions were constructed encompassing the GST gene (gst); the UGA codon and SECIS-containing regions; and the β-galactosidase gene (lacZ). The inserts were cloned into the HindIII and ApaI sites of the vector pSKAGS. The expression of the fusions is under control of the lac promoter (Plac). (B) β-Galactosidase activity in cell lysates after expression of the fusion constructs in E. coli XL1-Blue MRF′ in the presence or absence of pASBC4. Enzyme activities were determined 2 h after induction with IPTG. Values are given as a percentage of that for the respective gst-′lacZ fusion of E. acidaminophilum prxU, which was analyzed as a standard (100%) in each experiment. Bars represent the means and standard deviations obtained with three different clones which were measured in duplicate.
The β-galactosidase activity of a gst-′lacZ fusion containing the SECIS structure of the C. jejuni selW-like gene was negligible in the absence of pASBC4 but was strongly increased upon coexpression of selB and selC from E. acidaminophilum (Fig. 6B). In contrast, gst-′lacZ fusions containing the SECIS regions of selD from T. denticola and hydV from D. baculatum gave considerable β-galactosidase activity, indicating UGA readthrough. When selB and selC from E. acidaminophilum were coexpressed, the activity increased about four- or twofold, respectively, and reached a level similar to that for the gst-′lacZ fusion of prxU.
DISCUSSION
The recognition of SECIS elements by the selenocysteine-specific elongation factor SelB plays a key role in the UGA-directed insertion of selenocysteine into bacterial proteins. In the case of E. coli SelB, this interaction is highly specific (4, 18, 28). The exchange of only one nucleotide in the loop region of fdhF mRNA, which encodes one of the three formate dehydrogenase selenoproteins, abolishes in vivo selenocysteine incorporation. Shortening or extending the stem of the hairpin also had detrimental effects. Therefore, recombinant synthesis of selenoproteins cannot be accomplished, since E. coli SelB does not recognize SECIS elements from other organisms, unless these secondary structures resemble that of fdhF mRNA. However, this high selectivity of SelB cannot be a rule for all bacterial species, since selenocysteine occurs in functionally very different bacterial proteins even in the same organism.
Our study shows that the barrier to recombinant selenoprotein synthesis in E. coli can be overcome by coexpression of the genuine selB gene, resulting in selenocysteine incorporation. A similar approach was described previously (43). Those authors studied the translation of hydV mRNA from D. baculatum, encoding an [NiFeSe]-hydrogenase in E. coli. In contrast to the observations described here, coexpression of the selB gene from this deltaproteobacterium did not result in an insertion of selenocysteine into the respective protein, although the formation of a complex between D. baculatum SelB, selenocysteyl-tRNASec, and the hydV SECIS element was observed in vitro. A structural incompatibility between this complex and the E. coli ribosome was considered to be the reason for the lack of selenoprotein synthesis in vivo. This hindrance does not seem to occur in the case of SelB from E. acidaminophilum. Previous studies had revealed that selB from E. acidaminophilum is able to complement the defect of an E. coli selB mutant, thereby restoring the ability of this strain to produce the selenoprotein formate dehydrogenase H (15). However, complementation was more effective in the presence of selC from the gram-positive anaerobe. Therefore, we provided E. acidaminophilum selC together with selB on plasmid pASBC4 in the present work.
The incorporation of selenium into heterologously synthesized GrdB1, PrxU, and SelD1 was strictly dependent on the presence of SelB from E. acidaminophilum. Thus, the expression system described here enabled us to further characterize the postulated SECIS elements of the respective genes. Especially, the compensation of the negative effect of base changes in the stem of the apical hairpin by additional mutations in the opposite strand provides evidence that at least the upper parts of the putative structures fold in vivo according to the calculated conformation. Furthermore, the sequence of the basal part of the upper hairpin is not critical for recognition by SelB, provided that base pairing is possible. This is in accordance with the properties of the fdhF SECIS from E. coli (28). Previously, an alternative SECIS element was postulated for grdB1 (15), showing a lower calculated minimum free energy than the structure presented here. However, this secondary structure would not be consistent with the results of the mutational analysis. Furthermore, this structure does not match the bacterial SECIS consensus model which was developed recently from 100 known bacterial selenoprotein sequences (49).
Due to the astonishing variety of selenoproteins in E. acidaminophilum, it is obvious that SelB from this organism shows a lower specificity toward a given SECIS element than the E. coli protein does. Mutations that were introduced into the apical loops of the SECIS elements of grdB1, prxU, and selD1 had only a moderate effect on the incorporation of selenocysteine. In contrast, virtually all examined single-base changes in the loop of the E. coli fdhF SECIS element resulted in a loss of functionality (17, 23). Likewise, screening of randomized libraries of SECIS variants did not reveal functional structures with mutations in the loop (39).
In contrast to GrdB1 and SelD1, the peroxiredoxin PrxU was produced in E. coli in large amounts even in the absence of genuine SelB and tRNASec. The corresponding protein was not labeled with 75Se, indicating a selenium-independent UGA readthrough. In the presence of selB and selC from E. acidaminophilum, labeling of the protein with 75Se provided clear evidence for selenocysteine incorporation. However, it remained to be elucidated whether the whole protein or only a part of it contained this amino acid at the position encoded by UGA. The purified protein revealed masses corresponding to both a selenoprotein and a tryptophan-containing polypeptide. However, the major component with a mass of 23,573 Da did not match one of these two variants. The questionable mass would match exactly the replacement of the selenol group by a hydroxyl group. Since seryl-tRNASec is an intermediate in the formation of selenocysteyl-tRNASec, a cotranslational incorporation of serine into PrxU might be considered. However, a tRNA charged with serine is not recognized by SelB, at least in E. coli (9). Whether a comparably strong discrimination exists for E. acidaminophilum SelB still needs to be proven. GrdB1 was produced exclusively as a selenoprotein in E. coli XL1-Blue MRF′ in the presence of genuine selB and selC (14). The respective peptide obtained after proteolytic fragmentation exhibited the typical distribution pattern of selenium isotopes. A peptide that would point to insertion of serine was not detected.
The questionable mass of PrxU might also point to a loss of selenium from the selenoprotein during sample processing. Ma et al. (30) observed the conversion of one or more selenocysteine residues of rat selenoprotein P to dehydroalanine, probably through the mechanism of oxidation of selenocysteine followed by a syn-β-elimination of selenenic acid. The calculated mass of a PrxU variant containing dehydroalanine instead of selenocysteine would be 23,555 Da. To further substantiate and verify the experimental mass data, peptides were generated from the purified protein, but the corresponding peptide could not be isolated by high-pressure liquid chromatography (not shown).
In the absence of selB and selC from E. acidaminophilum, full-length PrxU is produced almost exclusively as a tryptophan-containing polypeptide at high levels. However, the corresponding gst-′prxU′-′lacZ fusion did not show appreciable β-galactosidase activity in the absence of SelB and tRNASec from E. acidaminophilum, whereas synthesis of full-length PrxU from the vector pASK-IBA3 revealed strong UGA suppression. Thus, this high level of selenium-independent readthrough cannot be explained solely by the features of the sequence surrounding the selenocysteine codon (27, 33, 34). Other factors, such as codon usage, growth conditions, and the expression system itself, might play an important role (31, 35).
UGA decoding of the E. coli fdhF mRNA has been described as a very inefficient process. Using a gst-′fdhF′-′lacZ fusion, Suppmann et al. observed around 4 to 5% readthrough, as determined by measuring the relative amounts of full-length product versus termination product by immunoblotting with an anti-GST antibody (42). Concomitant overproduction of SelB, tRNASec, and SelA increased efficiency to 11%. In the present study relatively high UGA readthrough values were obtained for grdB1, prxU, and selD1. In the presence of genuine SelB and tRNASec, β-galactosidase activities of the gst-′lacZ fusions reached up to 65% of that determined for the respective cysteine (UGC) mutants. This suggests that under certain conditions selenocysteine incorporation in E. coli can be more efficient than previously described for the fdhF gene. A similar observation was reported by Arnér and coworkers, who achieved high-level expression of the rat thioredoxin reductase gene containing an engineered variant of the E. coli SECIS(3, 35).
To our surprise, the grdB1 SECIS fusion construct gave high β-galactosidase activities even in the absence of genuine SelB (Fig. 5B). This can be explained by an internal translation start point, as the inserted grdB1 sequence contains an in-frame ATG triplet downstream of a purine-rich region that can act as ribosome-binding site.
Previous experiments had revealed that SelB from E. acidaminophilum was also able to mediate selenocysteine incorporation into formate dehydrogenase H of E. coli (15). This finding as well as the observed tolerance toward mutations in the examined SECIS elements encouraged us to investigate whether E. acidaminophilum SelB is also able to recognize SECIS elements from other bacterial species. UGA readthrough of the selenocysteine codons of D. baculatum hydV, T. denticola selD, and the C. jejuni selW-like gene in E. coli was significantly increased upon coexpression of selB and selC from E. acidaminophilum, suggesting an incorporation of selenocysteine by the heterologous SelB. To further substantiate these findings, labeling experiments in the presence of 75Se would be required, but these have not been possible thus far due to technical reasons. In any case, the increase of UGA readthrough mediated by SelB from E. acidaminophilum points to a lower sequence specificity of E. acidaminophilum SelB compared to the E. coli enzyme.
It is tempting to speculate that coexpression of the genes selB and selC from E. acidaminophilum might be a straightforward tool not only to produce selenoproteins from this organism in E. coli but also to express selenoprotein-encoding genes from other bacterial species. It might also be used for the insertion of selenocysteine into nonselenoproteins in order to enhance their catalytic activity. Furthermore, selenium isotopes with specific characteristics might be inserted into proteins by this means, which is of interest for different biotechnological applications (20).
ACKNOWLEDGMENTS
This work was supported by grants from the Deutsche Forschungsgemeinschaft (DFG), Bonn.
We thank the DFG for the opportunity to participate with this project in the DFG selenium priority program. We kindly thank A. Böck, Munich, and members of his lab for scientific suggestions and practical support and for the opportunity to carry out some experiments in their lab. A. Böck also provided us with the E. coli selB mutant strain. We also thank Claudia Hammerschmidt, Halle, for technical assistance.
FOOTNOTES
- Received 23 September 2007.
- Accepted 21 December 2007.
- Copyright © 2008 American Society for Microbiology