Titration and Conditional Knockdown of the prfB Gene in Escherichia coli: Effects on Growth and Overproduction of the Recombinant Mammalian Selenoprotein Thioredoxin Reductase

Abstract

Release factor 2 (RF2), encoded by the prfB gene in Escherichia coli, catalyzes translational termination at UGA and UAA codons. Termination at UGA competes with selenocysteine (Sec) incorporation at Sec-dedicated UGA codons, and RF2 thereby counteracts expression of selenoproteins. prfB is an essential gene in E. coli and can therefore not be removed in order to increase yield of recombinant selenoproteins. We therefore constructed an E. coli strain with the endogenous chromosomal promoter of prfB replaced with the titratable P_BAD promoter. Knockdown of prfB expression gave a bacteriostatic effect, while two- to sevenfold overexpression of RF2 resulted in a slightly lowered growth rate in late exponential phase. In a turbidostatic fermentor system the simultaneous impact of prfB knockdown on growth and recombinant selenoprotein expression was subsequently studied, using production of mammalian thioredoxin reductase as model system. This showed that lowering the levels of RF2 correlated directly with increasing Sec incorporation specificity, while also affecting total selenoprotein yield concomitant with a lower growth rate. This study thus demonstrates that expression of prfB can be titrated through targeted exchange of the native promoter with a P_BAD-promoter and that knockdown of RF2 can result in almost full efficiency of Sec incorporation at the cost of lower total selenoprotein yield.

Studies involving titration of expression of essential Escherichia coli genes must be carefully conducted and naturally cannot be performed using a direct gene knockout procedure. It was shown early that the P_BAD promoter is suitable for titration or conditional knockdown of essential genes when used in plasmid-driven complementation of chromosomal gene deletions (13). Recently, additional approaches for chromosomal integration of P_BAD aiming at titration and knockdown of essential genes have been described (25, 26). One caveat in low-level titration of the transcriptional activity of P_BAD (in contrast to knockdown with glucose) is that it generally occurs by an on-or-off effect in individual cells, due to arabinose-induced induction of the araE transporter, which is needed for uptake of arabinose as an inducer of the P_BAD promoter (28). This effect, eliminating the possibility of controlled expressional titration on the cellular level, can be circumvented by use of the BW27783 strain, which has araE constitutively expressed (20). We therefore selected the BW27783 strain as an E. coli host in this work, where we wished to study the impact of the essential (15, 18) prfB gene on selenoprotein expression, utilizing replacement of its native chromosomal promoter with P_BAD.

Protein translation occurs at the ribosome up to the point of termination of translation, which requires release factor 2 (RF2) as one of two polypeptide chain release factors. Thus, when the protein translation complex reaches an in-frame termination codon in the mRNA and this is exposed at the A site, binding of either RF1, at UAG or UAA codons, or RF2, at UGA or UAA codons, may occur, thereby catalyzing release of the completed polypeptide chain as aided by RF3 (30).

The two genes encoding RF1 and RF2, i.e., prfA and prfB, respectively, share high homology and gene structure. As a result, RF1 and RF2 exhibit similar domain structures and polypeptide chain release mechanisms (27, 33). However, RF2, in contrast to RF1, is expressed in a highly regulated manner involving both an internal regulatory +1 frameshift (8, 9) and a unique conformational change that affects the activity of RF2 in relation to its cellular expression levels (30). The frameshift event during RF2 translation, interestingly, occurs at the position of an in-frame UGA termination codon located at codon position 26, the switch of which is facilitated by the codon context (22). This is believed to function as an autoregulatory feature, since full-length RF2 will recognize the in-frame UGA codon in the RF2 mRNA as a termination codon, especially if RF2 is expressed at high levels, thereby prematurely terminating the translation of additional RF2 molecules (34). If, on the other hand, low levels of RF2 are expressed, then the frameshift will occur at a higher rate, leading to production of more RF2 (11, 23). Moreover, studies using an overexpression plasmid with a strong promoter in control of RF2 transcription showed that the frameshift then occurred more often, probably due to a higher overexpression of the mRNA levels than of the resulting RF2 protein levels, thereby facilitating frameshift rather than UGA termination (30). It is clear that the expression of RF2 is a highly and intricately regulated process, which is also of interest in relation to selenoprotein synthesis.

Dedicated UGA codons can encode either termination of translation (through RF2) or translational elongation with insertion of selenocysteine (Sec), a highly reactive selenium-containing amino acid present in selenoproteins (5, 17). Incorporation of Sec requires the presence of a secondary structure in the mRNA, called a Sec insertion sequence (SECIS) element, which in E. coli is located directly downstream of the designated Sec-encoding UGA codon. The SECIS element binds a Sec incorporation complex, consisting of a dedicated elongation factor, SELB, binding to a Sec-specific selenocysteinylated tRNA, Sec-tRNA^Sec, thereby forming a quaternary complex with GTP and catalyzing Sec insertion at the position of the dedicated UGA codon (14). Thus, an mRNA that encodes a selenoprotein will during the translation process expose its Sec-encoding UGA codon in the A site of the ribosome, thereby potentially allowing it either to bind RF2 or to support Sec incorporation by the Sec-tRNA^Sec-SELB-GTP complex; however, the latter occurs only if a SECIS element is present at the correct downstream position. This results in a competition between RF2-mediated termination and SELB-mediated Sec incorporation at the site of the Sec-encoding UGA and is therefore another stoichiometric obstacle that counteracts the possibility of using E. coli for production of recombinant selenoproteins (31). Overexpression of the recombinant rat selenoprotein thioredoxin reductase is possible in E. coli by using introduction of a bacterial-type SECIS element into the gene and concomitant overexpression of the bacterial selA, selB, and selC genes (4), constituting major parts of the selenoprotein synthesis machinery (5). That approach resulted in about 20 to 25% Sec incorporation efficiency with a standard protocol for recombinant expression at high yield (4). After optimization of the growth conditions, where protein expression was induced at 24°C in late exponential phase, the Sec incorporation reached up to 50% efficiency (24), probably in part due to slower bacterial growth and thereby lower RF2 levels (2, 24). In the present study, we therefore wished to ask if the efficiency of RF2 could be deliberately diminished in a selenoprotein-overexpressing E. coli background, thereby potentially increasing the Sec incorporation efficiency.

MATERIALS AND METHODS

Chemicals, reagents, enzymes, strains, plasmids, and primers.

DTNB [5′-5′-dithio-bis(2-nitrobenzoic acid); Ellman's reagent], NADPH, and common chemicals were from Sigma. Polymerases, restriction enzymes, and other enzymes for DNA work were from either Invitrogen or Fermentas unless stated otherwise. The λDE3 lysogenization kit was from Novagen. Rabbit polyclonal antibodies against mammalian thioredoxin reductase (TrxR) were from Upstate. Rabbit polyclonal antibodies against RF2 were raised by Agrisera (Umeå, Sweden) under ethical permit number A97-04, using the synthetic peptide RIQDLTERSDVLRGYC(-COOH) as the antigen, corresponding to amino acid residues 9 to 24 of the N terminus of RF2 with a C-terminally added Cys residue for conjugation of adjuvant. The bacterial strains and plasmids used are listed in Table 1. All primers used (Table 2) were from Invitrogen and were synthesized with high-pressure liquid chromatography purification. Keto-Diabur-Test 5000 glucose sticks from Roche were used for verification of a maintained high glucose concentration in the growth medium of the turbidostatic culture.

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Major features	Resistance	Reference(s) or source
E. coli strains
BW27783	Chromosomally encoded araE under control of a Pcp8 promoter		20
ORa(cm)	BW27783 with prfB under control of pBAD; Cmr	Cmr	This work
ORaa	ORa(cm) with FLP recombinase scar site		This work
ORaa(DE3)	ORaa lysogenized with DE3		This work
BL21(DE3)	General protein expression host		Invitrogen
Plasmids
pBAD/Myc-His A	Contains PBAD promoter	Ampr	Invitrogen
pGEM-T	Cloning vector used for all sequencing steps	Ampr	Promega
pKD46	Coplasmid for expression of Red recombinase	Ampr	10
pKD3	Plasmid carrying FRT-flanked adjacent Cmr genea	Cmr	10
pCP20	Plasmid for expression of FLP recombinase	Cmr Ampr	10
pET-TRSTER	Plasmid for overexpression of mammalian TrxR under control of the T7lac promoter (with E. coli SECIS element)	Kanr	4
pSUABC	Coplasmid for overexpression of selA, selB, and selC under control of their endogenous promoters	Cmr	3, 4, 29

^aFRT, FLP recombinase recognition target sequence.

TABLE 2.

Primers used in this study

Primer	5′-to-3′ sequencea
Primer1a	AATCAGACCATGGTTGAAATTAATCCGG (NcoI)
Primer1b	CTCTTTCTCTCGAGCGTAGTCAAAGATACCCCCTAAG (XhoI)
Primer2a	TCACCCGGGAAACCAATTGTCCATATTGC (XmaI)
Primer3a	GTGTAGGCTGGAGCTGCTTC
Primer3b	TTAGCCCCCGGGCATATGAATATCCTCCTTAG (XmaI)
Primer4a	gagcgtggattgggtacaatcccgctcttatcaccgcaGTGTAGGCTGGAGCTGCTTC
Primer4b	cttcttccagacgctctttcttggcgtcgtagtcaaagatacc
pctrl1	CTGGCCATTTTAGCGTCATCTTCTC
pctrl2	AGACATCCGGCTGTTCCAGCTCGGCG

^aUnderlined nucleotides introduce the restriction sites as indicated in parentheses. Boldface indicates primer site 1 (10). Lowercase indicates parts homologous to prfB, used for recombination (H1 and H2 for Primer4a and Primer4b, respectively; see Fig. 1B).

Growth conditions.

All bacterial cultures were maintained in LB medium (10g NaCl, 10 g peptone, and 5 g yeast extract per liter of medium [pH 7], with addition of 15 g agar per liter for production of LB plates). Adequate antibiotics were used depending upon the choice of plasmids (for pSUABC, 34 μg/ml chloramphenicol; for pETTRS_TER, 50 μg/ml kanamycin; for pKD3, 34 μg/ml chloramphenicol; and for pCP20, 100 μg/ml ampicillin). In all experiments designed for expression of TrxR using pET-TRS_TER and pSUABC, the medium was supplemented with 5 μM selenite and 100 μg/ml l-Cys, with production induced using 100 μM IPTG (isopropyl-β-d-thiogalactopyranoside) (4, 24). Strains with prfB under P_BAD control were always maintained in at least 0.00001% (0.67 μM) arabinose for sufficient expression from the prfB gene to support growth or with the addition of either glucose for repression or arabinose for titration at concentrations and conditions as discussed below.

Replacement of the native prfB promoter with P_BAD.

The chromosomal promoter replacement was performed in strain BW27783 (20, 21), using the Red recombinase-based technique described by Datsenko and Wanner (10) with the P_BAD promoter described by Guzman et al. (13) as cloned from the pBAD Myc/HisA plasmid (Invitrogen). The approach that was employed is shown schematically in Fig. 1. First the initial region of the prfB gene, including the first 85 nucleotides of the RF2 open reading frame (encompassing the internal frameshift), was amplified by PCR using primers Primer1a and Primer1b, with BW27783 genomic DNA as the template (PCR 1). Overhanging adenine nucleotides were attached to this PCR product (106 nucleotides [nt]) by using Taq polymerase and then ligated into a pGEM-T vector according to the protocol from the manufacturer (Promega). This introduced restriction sites that were then cleaved with NcoI and XhoI. This DNA fragment was purified on an agarose gel and ligated into the pBAD Myc/HisA vector (Invitrogen) between the NcoI and XhoI restriction sites (placing the first part of the prfB gene under control of the P_BAD promoter in the plasmid). Subsequently, a second PCR was performed using primers Primer2a and Primer1b with the ligation product as the template, which amplified the P_BAD promoter together with the initial part of prfB (PCR 2). This PCR product (424 nt) ligated into a pGEM-T vector was called pGEMTPCR2a. Another PCR was performed using primers Primer3a and Primer3b with the pKD3 vector as the template (PCR 3); the product (1026 nt) ligated into a pGEM-T vector was called pGEMTPCR3C. The two ligated vectors pGEMTPCR3C and pGEMTPCR2a were cleaved with NdeI and XmaI, liberating products of 4,062 and 439 nucleotides, respectively, which were purified on agarose gels and ligated with each other. A subsequent PCR was performed directly on the ligation mixture, using primers Primer4a and Primer4b (PCR 4). Throughout all of these PCR steps (PCR 1 to PCR 4), the PCR products ligated into pGEM-T were sequenced and confirmed as being correct as planned. After PCR 4, the amplified product was digested with DpnI and introduced by electroporation into BW27783 cells, which were made electrocompetent according to the method of Datsenko and Wanner and also transformed with the temperature-sensitive plasmid pKD46 for expression of Red recombinase (10). These bacteria electroporated with the PCR 4 product were subsequently grown in chloramphenicol and arabinose (0.2%) at 37°C for loss of pKD46 (temperature sensitive) and maintenance of RF2 expression (arabinose addition). The resulting chloramphenicol-resistant strain with the P_BAD promoter guiding prfB transcription was called ORa(cm). The ORa(cm) strain was TSS transformed (7) with the temperature-sensitive pCP20 plasmid encoding FLP recombinase and then grown at 30°C in ampicillin and arabinose to allow loss of the chloramphenicol resistance gene. It was subsequently shifted to growth at 37°C in the absence of antibiotics for loss of pCP20, making a new strain, called ORaa, lacking chloramphenicol resistance. The integrity of the constructed artificial prfB gene was verified by colony PCR using primers pctrl1 and pctrl2 complementary to the flanking genomic regions outside of the recombination site, and the products of this PCR were used for sequence verification (Fig. 1C and D).

FIG. 1.

Construction and verification of the ORaa strain. The construction of a host strain enabling RF2 titration (ORaa) is described in the text and schematically shown here. (A) Targeted E. coli genomic region, with the transcribed parts of the recJ and prfB/lysS operons shown as boxes. The positions of the H1 and H2 homologous regions are shown, and the frameshift in the RF2 open reading frame is indicated by an asterisk. In the box, the nucleotide sequence of the relevant region of the prfB gene is shown in lowercase letters, with the translated initial part of RF2 also given in one-letter amino acid code. The additional thymidine in the gene, leading to an in-frame UGA frameshift site in the RF2 mRNA, is shown in parentheses. The region constituting H1 corresponds to the nucleotides between the two number signs (#), while the H2 region is indicated by plus signs. Indicated by arrows are also the two sites where Primer1a and Primer1b annealed in the PCR 1 amplification and the restriction sites introduced by this primer pair. (B) Flow scheme, showing templates and primers used in PCR 1 to PCR 4 together with the ligation steps and schematic drawings depicting the resulting products or chromosomal replacements. The P_BAD promoter regions are indicated by hatched boxes. FRT, FLP recombinase target sequence. (C) Analysis of the products of PCR performed with genomic DNA from BW27783 (lane 1), with five clones of ORaa (lane 2a to 2e), and with ORa(cm) (lane 3), using primers pctrl1 and pctrl2. The PCR products were analyzed on a 1.5% agarose gel with size markers in the first lane. The expected sizes of the amplification products were 250 bp for BW27783 (endogenous prfB promoter), 1.5 kbp for ORa(cm) (artificial P_BAD promoter preceded by a chloramphenicol resistance cassette), and 0.5 kbp for ORaa [the promoter from the ORa(cm) strain with the chloramphenicol resistance removed], which agreed with sizes of the PCR products. (D) Actual DNA sequence determined from one of the PCR products for ORaa shown in panel C. In the sequence, the FLP scar, the ribosomal binding site (RBS), the NcoI site introduced by Primer1a, the initial part of the prfB gene encompassing the RF2 ATG start codon (underlined), the position of the frameshift (asterisk and frameshift thymidine in parentheses), and the P_BAD promoter region as derived from the pBAD Myc/HisA vector (boldface) are indicated.

A clone of the sequence-verified ORaa strain was subsequently lysogenized with phage DE3 lysogenic as described in the Novagen λDE3 lysogenization kit in order to enable its use as a host for pET vector-guided recombinant protein expression. The resulting ORaa(DE3) strain was finally transformed with the pETTRS_TER and pSUABC plasmids for TrxR1 production (4, 24).

Assessment of viability upon RF2 knockdown.

For assessment of viability upon RF2 knockdown (see Fig. 3), overnight cultures of BW27783, ORaa, or ORaa(DE3) transformed with pETTRS_TER, with addition of pSUABC, were grown in LB medium supplemented with 0.0001% arabinose and antibiotics in case of the transformed ORaa(DE3). These stationary-phase cultures were diluted 1:100 in 2 ml of the same medium, and growth was continued with vigorous shaking at 37°C for 6 hours. Next, 0.2% glucose was added and growth was continued for an additional 3 hours, whereupon an equal amount of each strain (corresponding to 0.01 μl culture at an optical density at 600 nm (OD₆₀₀) of 1) was plated onto LB plates with either 0.01% arabinose or 0.2% glucose (and antibiotics for the transformed strain), which were then incubated at 37°C for the indicated time periods. After 72 h, ORaa grown on glucose plates was scraped and plated out on new glucose or arabinose plates, and growth was continued at 37°C for another 24 or 96 h (see Fig. 3).

FIG. 3.

RF2 knockdown has a bacteriostatic effect in the absence of antibiotics. Overnight cultures of BW27783, ORaa, or ORaa(DE3) transformed with pET-TRS_TER and pSUABC were established in LB medium with addition of 0.01% arabinose, diluted 1:100 in the same medium, and grown for another 6 h, whereupon 0.2% glucose was added. Three hours later an equal amount of culture (corresponding to 0.1 μl of culture at an OD₆₀₀ of 1) was spread on LB plates with addition of sugar as indicated (0.01% arabinose or 0.2% glucose). BW27783 and ORaa were grown in the absence of antibiotics, while ORaa(DE3) transformed with the two plasmids was grown with the addition of kanamycin and chloramphenicol to maintain plasmid propagation. As shown, growth was evident after 24 h in all cases except for ORaa or ORaa(DE3) transformed with pET-TRS_TER and pSUABC when grown on glucose. These plates were therefore kept for an additional 48 h (72 h total incubation), after which they were again photographed, scraped, and spread on two new plates each (as indicated), which were continuously incubated and photographed after 24 h and another 72 h (96 h total incubation). Note that in the absence of antibiotics ORaa displayed very slow growth on glucose, which was regained on arabinose, while the doubly transformed ORaa(DE3) grown on kanamycin and chloramphenicol displayed no detectable growth on glucose.

Turbidostatic fed-batch culture system.

A turbidostatic fed-batch system was established to allow continuous addition of glucose during maintained growth (see Fig. 4). The ORaa(DE3) or the BL21(DE3) strains, both transformed with pET-TRS_TER and pSUABC, were cultivated in 200 ml LB medium with the addition of 34 μg/ml chloramphenicol (for pSUABC), 50 μg/ml kanamycin (for pETTRS_TER), 5 μM selenite, 100 μg/ml l-Cys, and 0.0001% arabinose at 25°C. A magnetic stirrer set at 250 rpm was used for oxygenation. When an OD₆₀₀ of 0.4 was reached, 1% glucose was added to the culture of ORaa(DE3) for repression of the P_BAD promoter and knockdown of RF2 levels, with BL21(DE3) being treated in the same manner for comparison purposes. Fresh medium containing 1% glucose, 100 μg/ml l-Cys, 5 μM selenite, and antibiotics was continuously added to the culture by using a pump. The flow of medium addition was set at an adjusted rate every 30 min in line with the monitored growth rate, as assessed by OD₆₀₀, so that the cell density in the culture was kept constant. A second pump was set at the same rate to lead away culture, thus keeping the volume of the turbidostatic culture constant and at the same time allowing samples to be taken from the withdrawn culture. Samples of 5 ml were split into two aliquots (1 ml and 4 ml). The smaller, 1-ml sample was centrifuged, and the bacterial pellet was lysed by resuspension in Tris-EDTA and incubation with 50 μg/ml lysozyme for 1 hour on ice followed by sonication. The sample was then cleared by centrifugation, and RF2 levels in the supernatant were analyzed by Western blotting. The larger, 4-ml sample of the culture was induced with IPTG (0.1 mM) to initiate TrxR expression, whereupon it was incubated for another 4 hours at 25°C with vigorous shaking. The bacteria were then harvested and analyzed for protein expression, either with the enzymatic DTNB assay for TrxR activity using the crude lysate as described earlier (4, 24) or with Western blotting for quantification of both RF2 and total TrxR levels. For Western blotting, crude bacterial lysates corresponding to 50 μl at an OD₆₀₀ of 1 were added to each lane of 4 to 12% sodium dodecyl sulfate-polyacrylamide gels, with transfer to nitrocellulose membranes. These were probed with polyclonal rabbit antibodies against RF2 (1:1,500) or mammalian TrxR (1:5,000), and after washing the membranes were incubated with horseradish peroxidase-conjugated goat anti-rabbit antibodies and subsequently visualized with Western Lightning Plus chemiluminescence reagent from Perkin-Elmer Life Sciences (see Fig. 5F). The amount of TrxR was determined by comparison to a known quantity of recombinant TrxR produced as described before (24), while arbitrary units for RF2 were determined as the measured total intensity of the RF2 band.

FIG. 4.

Scheme of the turbidostatic culture system. As described further in the text, a turbidostatic culture system was established in order to maintain continuous addition of glucose, follow culture growth, and enable recombinant protein induction in the same culture but at different time points after glucose addition. As schematically shown, to a culture volume of 200 ml (oxygenated by a magnetic stirrer), fresh medium was added with a pump, using another pump for removal of culture set at the same speed. The pump flow (Q) was continuously adjusted to the bacterial growth, so that the OD₆₀₀ was maintained at ≈0.4. For this, the OD₆₀₀ was measured every 30 min, after which the turbidostatic culture was diluted with fresh medium to an OD₆₀₀ of 0.4, with removal of volume so that the culture was kept at 200 ml. Between OD₆₀₀ measurements, the pumps were used to maintain constant glucose addition. With Keto-Diabur-Test sticks it was verified that the glucose concentration was kept at high levels throughout the fermentation. Samples (5 ml) were taken at different time points and used for RF2 level determinations and IPTG addition, allowing 4 h of production of recombinant TrxR, which was subsequently analyzed together with RF2 as described in the text. The growth rate in the turbidostatic culture was determined using the equation given in the figure, resulting in a calculation of the percentage of change in biomass per time unit.

FIG. 5.

Relationship between RF2 levels, growth rate, and recombinant selenoprotein production in a turbidostatic culture system. A turbidostatic culture system was established as described in the text and in Fig. 4 with either BL21(DE3) (filled symbols) or ORaa(DE3) (open symbols), both transformed with pET-TRS_TER and pSUABC. (A) Growth rates in the turbidostatic cultures, calculated as described in Fig. 4. (B) RF2 levels as determined by Western blotting at the time of harvest, 4 hours after the IPTG induction. At the time of addition of IPTG, the RF2 levels were approximately the same (not shown). (C) Total recombinant TrxR levels after 4 hours of production, as measured by Western blotting detecting both UGA-truncated and Sec-containing TrxR. (D) Sec-dependent TrxR activity in crude extracts. (E) Specific enzymatic activity of the produced TrxR, as calculated from the data given in panels C and D. (F) Illustration of the Western blot determination of TrxR and RF2 levels in ORaa(DE3) after 4 hours of protein expression. For TrxR (upper panel), a known amount of pure recombinant enzyme was used as standard (rightmost two lanes) to enable calculation of absolute yield, whereas the RF2 levels (lower panel) were determined as arbitrary units as described in the text. Both Western blots (for TrxR and RF2) were obtained using the same sample of bacterial extract from each time point and with the same amount of total protein added to each lane (total protein lysate corresponding to 50 μl culture at an OD₆₀₀ of 1).

RESULTS

Effects of RF2 titration on E. coli growth.

To be able to assess the impact of different RF2 levels on growth and selenoprotein expression in E. coli, we first set out to construct a strain with a chromosomal insertion of the P_BAD promoter guiding prfB transcription. The procedure for construction of this strain, named ORaa, and its corresponding DE3 lysogen, ORaa(DE3), was based upon placing the initial part of the prfB open reading frame under control of the P_BAD promoter in an expression vector, with subsequent introduction of this construct at the corresponding position in the chromosome by recombination. This results in removal of the endogenous promoter of prfB and its replacement with P_BAD. The method is described at detail in Materials and Methods and shown schematically in Fig. 1.

During growth in 0.1% arabinose, the P_BAD promoter is highly induced (13). We therefore first analyzed the effect of high prfB overexpression on bacterial growth, comparing ORaa with its parental strain BW27783. Until mid-exponential phase the two strains had comparable growth rates, and the ORaa strain had only about two- to three-times-higher levels of RF2 than BW27783 (Fig. 2). Beyond mid-exponential phase, however, the parental BW27783 strain displayed decreasing RF2 levels, while ORaa maintained an overexpression of RF2 in the arabinose-containing medium, resulting in about sevenfold-higher RF2 levels than in BW27783 and a lower growth rate up to stationary phase (Fig. 2).

FIG. 2.

Effects of RF2 overexpression on growth. BW27783 (squares and white bars) and ORaa (circles and black bars) were inoculated (1:100) in 100 ml LB medium supplemented with 0.1% arabinose to an OD₆₀₀ of about 0.05, and the cultures were then grown with vigorous shaking at 37°C. Growth (OD₆₀₀) was followed (circles and squares, left y axis), and RF2 levels were determined by Western blotting (bars, right y axis). Note the endogenous down regulation of RF2 beyond mid-exponential phase in the parental BW27783 strain and the overexpression of RF2 in ORaa with its slightly lower growth rate beyond mid-exponential phase, as further discussed in the text.

We next analyzed the effect of RF2 knockdown on growth. For this we used ORaa as well as ORaa(DE3) transformed with pETTRS_TER and pSUABC, with the parental BW27783 strain as a control. No apparent growth of ORaa or ORaa(DE3) transformed with pET-TRS_TER and pSUABC could be detected after 24 h in the presence of glucose. This was expected, since the RF2 gene is known as an essential gene (15, 18). The growth of the two strains was, however, comparable to that of BW27783 in the presence of 0.01% arabinose. Notably, after continued incubation for another 48 h, tiny colonies appeared on the ORaa plate with addition of glucose, suggesting a very low growth rate. When those colonies were seeded on arabinose plates, growth again became rapid, while when the colonies were maintained on glucose, it remained very slow, suggesting that the knockdown of RF2 had a bacteriostatic but not bactericidal effect. With the ORaa(DE3) strain transformed with two plasmids and thus grown in the additional presence of kanamycin and chloramphenicol, no detectable growth was seen on glucose even after 96 h. These results are summarized in Fig. 3.

Effects of RF2 titration on production of a recombinant selenoprotein.

Because RF2 is believed to counteract SELB-mediated Sec insertion, we next wished to assess whether titration of RF2 could affect the yield or specificity of Sec incorporation during overproduction of recombinant selenoprotein. For this we used the ORaa(DE3) strain for production of the rat selenoprotein thioredoxin reductase (with the pET-TRS_TER plasmid), using concomitant overproduction of the selA, selB, and selC genes (with the pSUABC plasmid) and comparing this RF2-titratable system with the corresponding BL21(DE3)-based production system described previously (4, 16, 24).

A turbidostatic fed-batch fermentor system was used in order to enable growth with addition of glucose under steady-state conditions; a schematic description of this experimental setup is shown in Fig. 4. The system enabled induction of recombinant selenoprotein expression for 4 h in a background of different RF2 levels, utilizing IPTG-induced production at different time points after initiating the constant addition of glucose. A prerequisite for the interpretation of our analysis was the fact that the Sec residue in mammalian thioredoxin reductase is needed for its enzymatic activity and is located at the penultimate position at the carboxyl-terminal end of the 55-kDa protein, making it possible to use the specific activity of the recombinant enzyme as a measure of Sec incorporation, as shown earlier (4, 16, 24). At the start of the turbidostatic culture conditions, the ORaa(DE3) strain had a lower growth rate than BL21(DE3), and upon addition of 1% glucose it dropped to a very low growth rate, while BL21(DE3) showed relatively constant growth throughout the experiment (Fig. 5A). At the beginning of the turbidostatic culture, the RF2 levels in ORaa(DE3) were about twice those in BL21(DE3), dropping to about the same level as in BL21 (DE3) after 3 h, and finally were knocked down to very low levels beyond 6 hours after addition of glucose. In contrast, BL21(DE3) had virtually constant RF2 levels throughout the experiment (Fig. 5B).

When measuring the total yield in production of recombinant thioredoxin reductase (TrxR1) (combined amount of UGA-truncated and full-length Sec-containing protein) by using ORaa(DE3) or BL21(DE3) grown under these conditions, we found that at the start of the culture, ORaa(DE3) gave about a threefold-higher total yield than BL21(DE3). When the knockdown of RF2 had resulted in a lower growth rate, this also correlated with a lower total yield in the recombinant protein production, so that at 8 hours after glucose addition it became lower than in BL21(DE3) (Fig. 5C). The amount of enzymatically active TrxR1 selenoprotein increased in ORaa(DE3) during the first 3 hours of growth after glucose addition, whereupon it slowly decreased to the same production level as in BL21(DE3) when induced 10 hours after addition of glucose, i.e., finally resulting in about 1 mg selenoprotein per liter per OD₆₀₀ unit produced in 4 hours (Fig. 5D). Concomitantly with the decreasing total RF2 levels in ORaa(DE3), however, it was evident that the ratio of Sec insertion to UGA truncation increased, as reflected by the continuous increase in the specific activity of the protein produced upon knockdown of RF2 (Fig. 5E). The levels of RF2 and total TrxR protein (Fig. 5B and C) were determined using Western blotting, as illustrated in Fig. 5F.

DISCUSSION

In this study we have found that the expression of RF2, an essential protein in E. coli (18), could be titrated using chromosomal replacement of its endogenous promoter with a synthetic P_BAD promoter. We furthermore found that such titration or knockdown had a significant impact on both growth characteristics and yield of selenoprotein production.

The endogenous native expression of RF2 in E. coli is known to be well regulated, including several posttranscriptional control events (8, 9). We found that overexpression of RF2 by induction of the P_BAD promoter with 0.1% arabinose led to only about two- to sevenfold-higher levels of RF2 than in the parental strain (Fig. 2), which was much less than the several hundredfold induction of the P_BAD promoter typically obtained using this arabinose concentration (13). The finding of rather modest RF2 overexpression correlates with an earlier attempt to overexpress RF2 by using a multicopy plasmid guiding expression of RF2 from the endogenous prfB promoter, which resulted in about fivefold-increased RF2 activity (6). The explanation for the small effect on overexpression in our study is likely the autoregulatory function of the +1 frame shift at codon 26 in the RF2 mRNA, as has been suggested earlier (8, 9).

In relation to our use of the ORaa strain and its DE3 lysogen to knock down RF2 expression, it is important to note that the promoter of the prfB gene regulates the dual prfB/lysS operon (19). lysS encodes a lysyl-tRNA synthetase, and if lysS had encoded the sole lysyl-tRNA synthetase in E. coli, then our approach for RF2 knockdown likely would not have been possible, since Lys residues are clearly needed for translational elongation of virtually all polypeptide chains. However, lysS was previously shown not to be essential in E. coli, although it is normally constitutively expressed (12), because the alternative lysU isoform is induced upon demand (32). We therefore believe that during the knockdown experiments described here, lysU likely substituted for a lack of lysS. However, it cannot be excluded that the adverse effects on growth or total yield of recombinant protein seen upon the repression of the prfB promoter were due to a combination of both RF2 and lysS repression. Further experiments are needed to determine the individual impact of a lack of these proteins, but it seems unlikely that a reduction of LysS should contribute to the effects on Sec insertion. The specific activity of the TrxR1 produced at the lowest RF2 levels reached about 40 U/mg, which reflects a full selenium content (16). This suggests that Lys residues were incorporated into TrxR1 during the production, at least to such an extent that its activity was not adversely affected.

In contrast to lysS, which is dispensable for growth (12, 19), RF2 has earlier been shown to be an essential protein by using a conditionally lethal mutant of the prfB gene (18) and in a system using complementary expression of RF2 from a plasmid (15, 18). Total deletion of the prfB gene should therefore be lethal. Knocking down the RF2 levels in the nontransformed ORaa strain studied here was found to have a bacteriostatic and not a bactericidal effect, while the transformed ORaa(DE3) strain grown in kanamycin and chloramphenicol could not be recovered on arabinose, although the plasmids it was transformed with carried the corresponding resistance cassettes. The molecular mechanisms explaining these phenotypes are unclear. It is possible that the promoter was not completely shut down in the presence of glucose, perhaps resulting in low (nondetectable by Western blotting) expression of RF2 in ORaa, which was sufficient for maintaining bacterial survival but with extremely slow growth. Such low RF2 levels, however, may not be compatible with the resistance to kanamycin and chloramphenicol in the transformed strain, with both compounds additionally adding strain to the ribosomal machinery.

We found earlier that the specific activity of recombinant TrxR increased when it was produced in late exponential phase, and we postulated that the reason could be a decrease in RF2 levels (24). That interpretation agrees well with the results in the present study, confirming that RF2 normally decreases in late exponential phase (see BW27783 in Fig. 2) and that the specific activity of the recombinant thioredoxin reductase increased when RF2 was repressed with glucose in ORaa(DE3).

By what mechanism did the knockdown of RF2 lead to the drastically slower growth rate? One reason could be a ribosomal pausing effect at in-frame UGA termination codons, leading to general ribosomal stacking and less efficient overall protein synthesis. Significant ribosomal pausing has, interestingly, been shown to occur specifically at Sec-encoding UGA codons in E. coli (29). During that pause, RF2 may bind at the UGA exposed in the A site, thus competing with SELB-mediated elongation (1, 29), and likely more so at selenoprotein overproduction (4). As found here, lowering the RF2 levels led to a reduced growth rate and lower total yield of recombinant protein production, while it clearly increased the specificity of Sec insertion. This suggests that in the virtual absence of RF2, or at least at very low RF2 levels, the ribosome may pause for a very long time at the position of the Sec-encoding UGA, allowing SELB to catalyze Sec insertion and thereby not producing the truncated protein species that are otherwise commonly detected during recombinant selenoprotein production in a background of higher RF2 levels (4, 24).

To conclude, we have here shown that it was possible to introduce a P_BAD promoter in place of the endogenous chromosomal promoter of the essential RF2 protein, encoded by the prfB/lysS operon. Using this tool, we have found that RF2 knockdown was bacteriostatic and not bactericidal, that the autoregulatory mechanisms maintaining RF2 levels seem to be highly efficient during high transcription of the prfB gene, and that RF2 knockdown clearly increases the specificity of Sec insertion into an overproduced recombinant selenoprotein at the expense of total yield, using production of mammalian TrxR1 as a model system. The ORaa and ORaa(DE3) strains developed here should constitute good tools for further studies of RF2 function in E. coli protein translation.