The crude extract was prepared with E. coli BL21 Rosetta2 cells (Novagen) grown at 37°C in 2xYT medium up to OD600 = 1.5, according to Kigawa et al 21 and Liu et al 9 . All the following steps were performed either on ice or at 4°C except for the pre-incubation at 37°C. The cells were washed twice and resuspended in S30 buffer A (50 mM Tris, 60 mM potassium glutamate, 14 mM magnesium glutamate, pH 7.7, 2 mM DTT). The cells were broken with a bead-beater (Biospecs Products Inc, mini bead-beater-1) using 0.1 mm glass beads. The extract was clarified by centrifugation at 30000 g for 25 minutes. The clear supernatant was pre-incubated 80 minutes at 37°C followed by a centrifugation at 30000 g for 10 minutes. The clear supernatant was dialyzed against S30 buffer B (5 mM Tris, 60 mM potassium glutamate, 14 mM magnesium glutamate, pH 8.2, 1 mM DTT) for 3 hours with a Slide-A-Lyzer cassette (MWCO 10 kDa, Pierce Biotechnology). The crude extract was stored at -80°C after dialysis. A typical concentration of 27-30 mg/ml of proteins in the crude extract was measured by Bradford assay. The crude extract is stable at least 1 year at -80°C.

The cell-free reactions were composed of 33% crude extract (9-9.5 mg/ml of proteins), the other 66% contained the reaction buffer and the plasmid. The reaction buffers were prepared with slight modifications according to Sitaraman and co-workers for 3-PGA 4 , Kim and Swartz for PEP 3 and Kim and co-workers for CP 22 . 3-PGA and PEP are natural E. coli substrates, therefore no enzyme was added to the reaction. Creatine phosphate is not a natural substrate for E. coli. Creatine kinase was added according to Kim and co-workers 22 .

All the plasmids used in this study originate from the plasmid pBEST-Luc (Promega). The list and sequences of the different regulatory parts are reported in the additional file 1 (List on page 1). Luc refers to Firefly luciferase [GenBank: CAA59281.1], eGFP to the enhanced green fluorescent protein [GenBank: CAD97424.1], eGFP-Del6 to the enhanced green fluorescent protein truncated and modified in N-terminal, eGFP-Del6-229 to the enhanced green fluorescent protein truncated and modified in N- and C-terminal, PtacI to the consensus E. coli sigma factor 70 promoter 23 , OR2-OR1-Pr to the lambda repressor Cro promoter [GenBank: J02459.1], UTR1 to the untranslated region containing the T7 g10 leader sequence for highly efficient translation initiation 24 [GenBank: M35614.1] and T500 to the transcription terminator 25 . The plasmids were prepared according to the standard molecular cloning procedures. Picogreen (Invitrogen) was used to measure plasmid concentrations. The E. coli strain KL740 was purchased from the Coli Genetic Stock Center (Yale) [CGSC#: 4382].

A Wallac Victor III platereader (PerkinElmer) was used to measure eGFP, eGFP-Del6 and eGFP-Del6-229 expression (kinetics and end-point measurements). Pure recombinant eGFP (Clontech) was used for calibration. Luc expression was measured with a custom-built luminometer described previously 12 . Pure Luc and Luc assay kit (Promega) were used for calibration and measurements. Polyacrylamide gel electrophoresis was carried out according to the standard procedures and labeled with SimplyBlue safestain (Invitrogen).

The crude extract was prepared according to Kigawa et al 21 and Liu et al 9 with some modifications. The usual buffers S30 A and S30 B used for the preparation of cell-free systems were composed of magnesium glutamate and potassium glutamate instead of the acetate form of the ions 26 . DTT only was used for the washing and the resuspension steps (S30 buffer A) at a concentration of 2 mM instead of the usual combination of 2-mercaptoethanol and DTT. The optimum pH for the S30 buffer A and the S30 buffer B was 7.7 and 8.2 respectively. The cells were broken with a mini bead-beater. This lysis method is more reproducible than the other methods 21 and this technique is also more affordable. Typically, 4.5 g of cells were resuspended in 4.05 ml of S30 buffer A before adding 23 g of dry beads. The mixture was homogenized before bead beating twice at 4600 rpm for 30 seconds with a pause of 30 seconds in between. The mixture was centrifuged at 30000 g for 25 minutes followed by a pre-incubation of the clear supernatant at 37°C for 80 minutes. The extract was clarified again at 30000 g for 10 minutes before 3 hours of dialysis. The entire preparation, from cell culture to crude extract storage, takes 12 hours and does not present any particular constraints compared to other procedures. The cells do not need to be frozen at any time during the process. The crude extract can be thawed once without any loss of activity. Two liters of cell culture give approximately 6 ml of crude extract with a final protein concentration of 27-30 mg/ml.

The cell-free reaction contained all the usual components of bacteriophage systems with minor modifications. Almost all the components in the reaction were optimized. The results for DTT, tRNA, and amino acids are reported in the additional file 1 (Figures S1, S2 and S3). The most important parameters for the optimization of the cell-free reactions were the magnesium concentration, the potassium concentration and the plasmid construction. The magnesium and the potassium concentrations were adjusted with the plasmids pBEST-UTR1-Luc and pBEST-UTR1-eGFP. The reporter genes were transcribed with a PtacI promoter 23 . PtacI, a consensus E. coli sigma factor 70 promoter, is one of the strongest E. coli promoters. The untranslated region of the plasmid pBEST-Luc was replaced by the untranslated region called UTR1 in this work. The UTR1 sequence includes the T7 gene 10 leader RNA, which contains a strong ribosome binding site 24 . The lac operator overlapping the -10 part of the promoter was conserved. With this modification, the expression was increased by a factor of five to ten on average in any of the buffers. The protein production was characterized with the three most common ATP regeneration systems used in cell-free studies: creatine phosphate (CP), phosphoenolpyruvate (PEP) and 3-phosphoglyceric acid (3-PGA). In each of these buffers, we found that the concentration of magnesium had to be precisely adjusted in order to maximize protein production (Figure 1A). This result was similar to the results obtained with cell-free systems using bacteriophage RNA polymerases. The best expression was obtained with 3-PGA, followed by PEP and CP. In the three buffers, the protein production was relatively independent of the potassium concentration over a wide range of concentrations (Figure 1B). The best protein production was also obtained with 3-PGA, followed by PEP and CP. For eGFP, the best expression was obtained with the 3-PGA buffer although the results obtained with the PEP buffer were close to 3-PGA (Figure 1C). All the following experiments were carried out with the 3-PGA buffer, which gave a greater protein production than PEP and CP buffers. In the 3-PGA buffer, the protein production was independent of the nucleotide concentrations over a wide range of concentrations (additional file 1, Figure S4.

The plasmid concentration is another important parameter in a cell-free reaction. Because the cell-free system described in this study works with the endogenous RNA polymerase and sigma factor 70, it was interesting to determine the quantity of template required to get maximum protein production and to compare it to the amount of plasmid required in bacteriophage-based transcription systems. The best protein production was obtained with a plasmid concentration between 5 nM and 15 nM for Luc (Figure 2A) and eGFP (Figure 2B). Surprisingly, the amount of plasmid required for maximum protein production was comparable to cell-free systems using bacteriophage RNA polymerase for transcription 2 3 4 5 6 . The protein production was linear with respect to the plasmid concentration below 2 nM. At a plasmid concentration of 5 nM, the protein production was 90% of the maximum for Luc and 80% of the maximum for eGFP. In our best extract preparation, a maximum Luc production of 10 μM could be measured, which corresponds to 0.6 mg/ml of active proteins. This protein yield, comparable to bacteriophage cell-free systems, was unexpected because the endogenous E. coli transcription machinery is much less efficient. However, the maximum production of eGFP was much smaller than Luc, with only 0.1 mg/ml produced (4 μM). Addition of pure E. coli RNA polymerase saturated with sigma factor 70 did not increase the protein production (additional file 1, Figure S5). This led us to the conclusion that expression could be optimized with different DNA regulatory parts and that eGFP was poorly translated. Furthermore, the amplification of plasmids with E. coli promoters can be problematic because of potential over-expression of the recombinant proteins. The plasmids pBEST-UTR1-Luc and pBEST-UTR1-eGFP were leaky during plasmid amplification despite the lac operator.

The maximum production of Luc was obtained with the plasmid pBEST-UTR1-Luc, which contains the PtacI consensus promoter and UTR1 (Figure 2A). No other modifications increased Luc production. A maximum of 0.6 mg/ml (10 μM) active Luc was measured by luminescence. eGFP production with the plasmid pBEST-UTR1-eGFP was significantly lower with only a few micromolars of eGFP produced (Figure 2B). This was far below the expected amount compared to Luc with the same plasmid. We found that translation initiation of eGFP did not work efficiently in our extract. To resolve this problem, the N-terminal sequence of eGFP was truncated according to the work of Li et al 27 and modified by silent mutations to remove ribosome binding site-like sequences. The new protein was named eGFP-Del6. Compared to pBEST-UTR1-eGFP, the fluorescence intensity at the end of the reaction was approximately 3 times greater with the plasmid pBEST-UTR1-eGFP-Del6 (Table 1). This increase, which is a real increase of protein production as opposed to an increase of the reporter fluorescence, was confirmed later by polyacrylamide gel electrophoresis (Figure 4B). The C-terminal sequence of eGFP was also modified 27 . A 20% increase in protein production was measured with this new reporter, named eGFP-Del6-229 (Table 1).

Cell-free expression with E. coli sigma factor 70 promoters can pose significant problems of toxicity due to a potential over-expression of the recombinant protein during plasmid amplification. Although showing some signs of toxicity, the plasmids pBEST-UTR1-Luc and pBEST-UTR1-eGFP-Del6-229 were stable during amplification. A more general solution to this problem was found by switching to the strong Lambda phage promoter Pr flanked by the two operators OR2 and OR1. The E. coli strain KL740, which over-expresses the temperature sensitive Lambda repressor Cl857, was used to efficiently repress and stabilize the plasmids during cell growth at temperatures below 30°C. Potential problems due to increased background expression during plasmid amplification could also be solved by changing the origin of replication. The plasmid pBEST-Luc has a ColE1 origin of replication (few hundreds copies per cell), which can be replaced by a P15A (10-12 copies per cell) or a PSC101 (2-5 copies per cell) origin of replication. Other tightly regulated transcription systems and strains could be also used 28 .

In the case of eGFP-Del6-229, expression with the OR2-OR1-Pr promoter was slightly greater than with the PtacI promoter (Table 1). A 20% increase in eGFP-Del6-229 production was observed with the addition of the strong transcription terminator T500 25 . For eGFP-Del6-229, the best expression was obtained with the plasmid pBEST-OR2-OR1-Pr-UTR1-eGFP-Del6-229-T500 at a concentration of 10 nM (Table 1). In the best conditions, a concentration of 0.65 mg/ml of active eGFP-Del6-229 was measured. The protein was synthesized at the same rate whether transcribed from a PtacI or a OR2-OR1-Pr promoter (Figure 3). However, the expression lasted a longer time with the lambda promoter. The production of eGFP with the original plasmid pBEST-eGFP was more than 100 times smaller than with the plasmid pBEST-OR2-OR1-Pr-UTR1-eGFP-Del6-229-T500.

The high protein production for both Luc and eGFP-Del6-229 was confirmed by polyacrylamide gel electrophoresis (Figure 4A and 4B). SDS PAGE analysis showed that eGFP-Del6-229 production was greater than eGFP production. This was essentially due to the modification of the N-terminal of the protein (Table 1). An amount of 0.75 mg/ml for both Luc and eGFP-Del6-229 could be estimated on gel whereas the amount of active reporter proteins measured by luminescence was 0.60 mg/ml (Luc) and by fluorescence 0.65 mg/ml (eGFP-Del6-229). The plasmid maps for pBEST-UTR1-Luc and pBEST-OR2-OR1-Pr-UTR1-eGFP-Del6-229-T500 can be found in the additional file 1 (Figure S6).

The protein productions measured in this work were comparable to the protein productions obtained with cell-free systems using bacteriophage RNA polymerases. Expression of GFP with a T7 transcription, characterized and quantified in two cell-free systems by Iskakova and co-workers 11 , was comprised between 0.3 and 0.7 mg/ml. A maximum of 0.15 mg/ml of active Luc was measured in two different cell-free systems 12 14 . Furthermore, the protein production obtained with commercial cell-free systems based on bacteriophage transcription is comprised between 0.5 and 1 mg/ml.

With the CP buffer, the maximum Luc and eGFP-Del6-229 production was only 60% of the amount obtained with 3-PGA. With the PEP buffer, the maximum Luc and eGFP-Del6-229 production was 60% and 100% respectively of the amount obtained with 3-PGA. It is important to note that Luc production was neither greater nor smaller with the different regulatory parts tested with eGFP. The production of Luc with the plasmids pBEST-UTR1-Luc and pBEST-OR2-OR1-Pr-UTR1-Luc-T500 were identical. The plasmid pBEST-OR2-OR1-Pr-UTR1-Luc-T500 was easier to amplify in the E. coli strain KL740.

The protein production in cell-free reactions depends on the temperature of incubation. For Luc, the best protein production was obtained in the temperature range 20-25°C (Figure 5A). This result was in agreement with previous studies on the reporter's activity that showed a net decrease of luminescence above room temperature 29 30 31 . In the case of eGFP, the best protein production was obtained at 32°C (Figure 5B). At this temperature, the expression of eGFP-Del6-229 was twice greater than eGFP. The best production of eGFP-Del6-229 was obtained at 29°C. At 37°C and 42°C, expression of eGFP and eGFP-Del6-229 were comparable. Additional tests were performed to determine if the temperature range 29-32°C was optimum for gene expression or optimum for thermo-stability of the proteins. Cell-free reactions were carried out at 29°C and 32°C for both eGFP and eGFP-Del6-229 during 8 hours followed by 1 hour of incubation at 37°C (Figure 5C). No change in fluorescence was observed. This result confirmed that the production of fluorescent proteins was maximum in the temperature range 29-32°C. Similar results have been obtained for the wild type GFP 11 . In the best conditions, eGFP production never reached more than 25% of the maximum obtained for eGFP-Del6-229. The expression of eGFP and eGFP-Del6-229 was also tested in E. coli using the same plasmid (pBEST-UTR1 with a PtacI promoter). The expression of eGFP-Del6-229 was slightly better than eGFP (additional file 1, Figure S7).