We were able to identify in the E. coli K12 genome 223 of 246 LUCA proteins described by Kyrpides et al. 15(see additional file 1: SummaryOfResults.xls). Of 223 LUCA genes represented in E. coli, we successfully cloned 221 (99.1%) into pQTEV2 (Fig 1). In small-scale, expression of 209 His-tagged proteins (94.6%) was detected in the cell extracts. Of these 209 proteins, 5 proteins could not be purified under native conditions with Ni-NTA agarose (proteins 198, 199, 205, 208, 209). 204 proteins (97.6%) could be purified. To obtain maximum amounts of protein, 161 proteins were expressed at 37°C and 43 proteins at 25°C (data not shown). 111 proteins (54.4%) were purified in high yields, 66 (32.4%) in moderate, and 27 (13.2%) in low yields. The estimation of protein yields is illustrated in Fig. 2. Together with the genes coding for the 5 proteins that were not purified, the genes of 27 proteins with low yields were subjected to Gateway (Invitrogen) recombination cloning in order to improve solubility by fusing the target proteins to MBP-His₇, GST-His₇, and NusA-His₆.

Of 48 ECHH genes, we successfully cloned 47 (97.9%) into pQTEV2 (Fig 3). Expression of 45 His-tagged proteins (95.7%) was detected in the bacterial lysate. To obtain maximum amounts of protein, 40 proteins were expressed at 37°C and 5 proteins at 25°C (data not shown). All of these 45 proteins could be purified with Ni-NTA agarose. 36 proteins (80%) were purified in high yields, 7 (15.5%) in moderate, and two (4.5%) in low yields. These two genes were also subjected to Gateway recombination cloning.

In total, 14 LUCA or ECHH proteins could not be expressed in E. coli SCS1 (see additional file 1: SummaryOfResults.xls). Some proteins could not be expressed at all. Other proteins could be expressed, but revealed low yields in small-scale purification or could not be purified. The distribution of these proteins into functional groups is illustrated in Table 1. Interestingly, 53% of transport proteins present in either protein sets were not expressed. 33% of ribosomal proteins were purified with low yields. 187 enzymes (87.4%), 15 ribosomal proteins (62.5%), but only 3 transport proteins (21.4%) were purified in soluble form with at least moderate yields.

Corresponding genes of the 34 proteins that were not purified or displayed low yields under native conditions were sub-cloned into Gateway destination vectors in order to fuse the target proteins to MBP-His₇, GST-His₇, and NusA-His₆, respectively. The resulting 102 fusion proteins were overexpressed in small-scale and purified by His-tag affinity to Ni-NTA agarose. Per target protein, the fusion partner was determined that most efficiently improved the yield after purification. Table 2 illustrates yields of 34 target proteins fused to the His₇-tag and to MBP-His₇, GST-His₇, or NusA-His₆. Yields of 30 target proteins improved with fusion to MBP-His₇ (80%), NusA-His₆ (16.7%) or GST-His₇ (3.3%). Yields of 4 proteins including two membrane proteins were not affected.

Four E. coli proteins were treated with polymyxin B coated magnetic beads in order to analyze efficiency of this protocol to reduce potential LPS related stimulation of CD4+ T cells. We compared the frequencies of CD40L+/IFN-γ+ T cells resulting from whole blood stimulations with protein solutions before vs. after treatment with polymyxin B beads. Equal amounts of proteins were applied, i.e. 5 μg per ml blood. Treatment of proteins 98, 251, and 256 with LPS removal beads reduced the frequency of stimulated CD4+ T cells. Except for protein 98, analyzed in blood of donor M, all donors showed decreased frequencies. Especially, stimulations with protein 256 revealed a drastic decrease in all donors. In contrast, treatment of protein 253 with LPS removal beads did not reduce the frequency of stimulated CD4+ T cells. In donor M, the frequency of double-positive cells stimulated with protein 253 even increased (Fig 4A). Since we observed variable loss of protein after incubation with LPS removal beads, different volumes of protein solutions had to be taken in subsequent immunological assays to provide 5 μg of protein per assay. This may have increased the amount of other stimulatory bacterial contaminants, such as muramyl dipeptide 23 and lipoproteins 24. Therefore, an additional analysis of the frequencies after stimulation with equal volumes of treated vs. untreated protein solutions was done (Fig. 4B). Without exception, treatment with LPS removal beads resulted in decreased frequencies.

In addition, LPS contents of protein solutions 251 and 96 were quantified with the Limulus amoebocyte lysate (LAL) test. Both protein solutions were treated with LPS removal beads. The LPS content in the protein solution 96 was below the detection limit of 0.05 EU/ml (5 pg/ml), whereas 0.08 EU/ml (8 pg/ml) was measured in protein solution 251.

Next, we quantified protein loss by the LPS removal procedure. For this, protein concentrations of 34 different protein preparations were determined before and after treatment with LPS removal beads. More than 61% of the analyzed proteins revealed a protein loss less than 50%. Only 5.9% of the proteins revealed a protein loss of 80–90% (Table 3). Since only few proteins showed major loss by the LPS removal procedure, we applied the procedure to all protein preparations of this study.

In order to produce a minimum of 500 μg of purified soluble protein, expression culture volumes were scaled up. By small-scale expression and purification, we obtained 177 LUCA proteins with high and moderate yields (Fig 1). However, in 35 proteins we did not achieve the minimum of 500 μg of purified protein. 26 of 32 Gateway LUCA fusion proteins revealed high or moderate yields when expressed in small-scale. Of these, 19 were expressed and purified in large-scale as fusions to MBP-His₇ (Fig 1). Of 43 ECHH proteins with high and moderate yields when expressed in small-scale, 9 proteins could not be produced in sufficient amounts. In addition, we obtained 2 ECHH proteins from Gateway fusions to MBP-His₇. Thus, we were able to produce 36 ECHH proteins with a minimum of 500 μg (Fig 3).

In order to correlate codon usage of proteins with success of expression, the codon adaptation index (CAI)25, a global indicator which compares codon usage of individual proteins with codon usage of a set of reference proteins, was determined for all proteins. As shown in Fig 5A., all proteins with a CAI of > 0.6 could be overexpressed in our system. In addition, we studied individual codons most rarely used in E. coli in a representative set of proteins. This analysis revealed that some rare codons are overrepresented in non-expressed proteins (Fig 5B.). Of particular interest might be the rare isoleucine codon ATA which occured 3 times or 15 times more frequently in non-expressed proteins than in weakly or highly expressed proteins, respectively (Fig 5B).

The 48 proteins that could not be expressed or yielded no protein or insufficient quantities of protein after purification as non-fusion proteins (see Table 1) were subjected to an in-silico analysis for the presence of predicted transmembrane domains. In addition, an arbitrarily chosen protein was included in this study. The results of this analysis are given in the additional file 1 (SummaryOfResults.xls). All 14 non-expressed proteins contained at least 2 predicted transmembrane helices (proteins 210–221, 269, 270). The two proteins that could be expressed but not be purified even as fusion proteins contained 5 or 7 predicted transmembrane helices, respectively (proteins 208, 209). On the other hand, at least 3 proteins containing 1, 2, or 8 predicted transmembrane helices, respectively, could be expressed and purified in sufficient quantities using our approach (proteins 202, 163, 258).

E. coli K12 representatives of the LUCA protein functions as described by Kyrpides et al. 15 were identified in the NCBI protein database. In addition, the protein databases Biocyc.org and Brenda were screened for LUCA proteins present in the E. coli genome. Moreover, the 246 LUCA proteins represented in M. jannaschii were used as queries in BlastP searches against the E. coli K12 protein sequences. By these methods, we identified 223 LUCA genes in the E. coli K12 genome. Using the BlastP algorithm 37 for comparison of E. coli K12 proteins with human protein equivalents, E. coli proteins with highest homologies to human proteins were identified. LUCA proteins were excluded from this list. Thus, 48 ECHH proteins with highest homologies to human proteins and not represented in the LUCA set of proteins were identified.

Competent E. coli cells were prepared with CaCl₂ and MnCl₂ as described previously 38. The competent cells were frozen in liquid nitrogen and stored at -80°C. Electro-competent E. coli cells were prepared as described previously 39. The competent cells were frozen in liquid nitrogen and stored at -80°C.

Our standard procedure employed PCR cloning of E. coli genes 30 into pQTEV2 (Fig 6) using restriction sites BamHI and NotI and a genomic DNA preparation from E. coli K12 from the German strain collection (DSMZ 5695). In case of intragenic BamHI sites, this enzyme was replaced by BglII. A second choice was use of type II enzymes BpiI or Eco31I. PCR primers were delivered in 96-well microplate format (Eurogentec). 10 μM stocks of forward and reverse PCR primers were rearranged by a Speedy pipetting robot (Zinsser) in corresponding plate positions of two polystyrene microtitre plates 30. PCR reactions were performed with Expand High Fidelity PCR kit (Roche Applied Science) according to the manufacturer's instructions. The PCR products were analyzed and quantified by agarose gel electrophoresis. Subsequently, the PCR products were purified with magnetic beads (Genopure ds kit, Bruker) according to the manufacturer's instructions. The PCR products were digested o/n with restriction enzymes with 10 units of each enzyme per reaction. After an additional purification step with magnetic beads, ligation reactions were set up taking into account visual estimates of PCR product concentrations. PCR product and water were added to make 6.5 μl in a well of a Thermowell 96-well plate, followed by the addition of 2 μl of linearized pQTEV2 (16 ng), 0.5 μl T4 DNA ligase (400 units/μl, NEB), and 1 μl 10× ligase buffer (NEB). The plate was covered with a sealing sheet and the samples were incubated at 20°C for 1 h, followed by heat inactivation (65°C, 20 min). Chemically competent E. coli SCS1 cells 30 were transformed with 5 μl of the ligation reaction in a Thermowell 96-well plate using standard heat shock procedure (42°C, 45 sec). Cells were recovered for 30 min at 37°C with 1 ml of 2xYT broth 16, supplemented with 20 mM MgCl₂ and 20 mM glucose. The cultures were plated on individual 2xYT agar plates, containing 100 μl/ml ampicillin and were incubated o/n at 37°C. Six colonies per transformation were picked with sterile tooth picks into individual wells of a polystyrene microtitre plate, containing 200 μl of stock medium: 2xYT broth, 1xHMFM, 100 μl/ml ampicillin, and 2% glucose (10xHMFM: solution "a": 5 mM MgSO₄.4H₂0, 20 mM tri-Sodium citrate.2H₂O, 85 mM (NH₄)₂SO₄, 45% glycerol; solution "b": 0.66 M KH₂PO₄, 1.3 M K₂HPO₄; four parts of solution "a" were combined with one part of solution "b"). The plate was incubated o/n at 37°C, sealed and stored at -80°C. To identify successfully cloned inserts, 6 clones per transformation were screened by colony PCR with primers pQE65 (5'-TGAGCGGATAACAATTTCACACAG-3') and pQE276 (5'-GGCAACCGAGCGTTCTGAAC-3') which bind to internal vector sequences adjacent to the multiple cloning site. After analysis of the PCR reactions on agarose gels, two positive clones per transformation were sub-cultured for protein expression in individual wells of a polystyrene microtitre plate, containing 200 μl of stock medium.

We utilized the Gateway (Invitrogen) recombination technology 40 to fuse the target protein with fusion partners, such as MBP, GST, or NusA. These fusion partners may enhance solubility of target proteins 32. We therefore constructed three Gateway destination vectors for expression of target proteins fused to MBP-His₇, GST-His₇, and NusA-His₆ (pD-MAL1, pD-GEX1, and pD-Nus1, respectively). All three destination vectors contained the sequence coding for a His-tag, enabling purification of either protein fusion by His-tag affinity to Ni-NTA agarose. In order to create entry clones, we performed BP recombination reactions between pQTEV2 containing the gene of interest and the Gateway donor vector pDONR221, followed by the transformation of E. coli XL1-blue cells, according to the manufacturer's instructions (Invitrogen). Positive entry clones were confirmed by colony PCR screening with primers M13Forward and M13Reverse which bind to internal vector sequences adjacent to the att sites. In the LR reaction, plasmid preparations of entry clones were incubated with the destination vectors pD-MAL1, pD-GEX1, and pD-Nus1, respectively, in order to generate E. coli SCS1 expression clones for MBP-His₇ or GST-His₇ fusion proteins. Electro-competent E. coli BL21 (DE3) cells were transformed with LR reactions of pD-Nus1. Positive expression clones were confirmed by colony PCR screening. pD-MAL1 constructs were screened with primers M13Forward (5'-Gcggatcctacctgacgcttt-3') and malE (5'-GGTCGTCAGACTGTCGATGAAGCC-3'). pD-GEX1 constructs were screened with primers pGEX5' (5'-GGGCTGGCAAGCCACGTTTGGTG-3') and pGEX3' (5'-CCGGGAGCTGCATGTGTCAGAGG-3'). pD-Nus1 constructs were screened with primers Nus-tag (5'-AAGCCGGAGCACTGATTATGG-3') and Colidown (5'-TTCACTTCTGAGTTCGGCATGG-3').

To identify clones that provided sufficient amounts of His-tagged proteins, small-scale expression and purification was performed as described previously 16. Protein expression was induced with 1 mM IPTG at 25°C for 6 hrs and at 37°C for 3 hrs. Overexpression of the recombinant protein was analyzed in the cell lysate. Purification under native conditions via affinity chromatography with Ni-NTA agarose (Qiagen) was performed to assess yields of purified proteins. Cell lysates and purified proteins were analyzed by SDS-PAGE with 15% polyacrylamide separation gels as described 41. Expression levels in cell lysates and yields of purified proteins were subjectively attributed to one of four categories: 3 (high), 2 (moderate), 1 (low), 0 (not purified).

According to protein yields obtained after small-scale expression and purification, culture volumes were adjusted for large-scale expression and purification. Proteins that were purified in high and moderate yields were expressed in 100 ml or 150 ml cultures, respectively. The following expression and purification protocol was applied per 100 ml of bacterial culture: 5 ml pre-culture was grown for 16 hrs at 37°C, followed by the addition of 95 ml of pre-warmed SB medium with ampicillin 16. At the optical density (OD) of 1.5 at 600 nm, protein expression was induced for 4 hrs at 37°C with 1 mM IPTG or 6 hrs at 25°C. Cells were harvested by centrifugation (4000 × g, 4°C, 15 min) in 50 ml polypropylene tubes (BD Falcon). The cell pellet was frozen in liquid nitrogen and stored at -80°C. To purify protein, the cell pellet was thawed on ice for 15 min and resuspended in 3 ml of lysis buffer 16. 450 μl of lysozyme solution (1 mg/ml) was added and incubated on ice for 30 min. 600 μl of benzonase buffer (20 mM Tris, pH 8.0, 10 mM MgCl₂) containing 0.4 unit/μl benzonase grade II (Merck) was added and incubated on ice for 30 min. The lysate was centrifuged at 4900 × g, 4°C, 45 min. The supernatant was transferred to individual cavities of a 24-deepwell plate (Brand) and 600 μl of 50% Ni-NTA agarose (Qiagen) was added. The plate was shaken for 30 min at 10°C and the lysate-bead mixture was transferred to a well of a 96-well filter plate (Macherey & Nagel, No 738655.M). Ni-NTA agarose was washed 6 times with 1.5 ml of washing buffer. Protein was eluted with 600 μl of elution buffer 16. EDTA (final concentration: 0.5 mM) was added to bind residual Ni²⁺ ions in the protein solution. In order to exchange the buffer, 500 μl of the protein solution was applied on a NAP-5 column (GE healthcare) and the protein was eluted with 1 ml of phosphate buffered saline (PBS, pH 7.4).

After changing the buffer, we treated the protein solutions with polymyxin B coated magnetic beads (25 mg/ml; Chemicell, Berlin, Germany) in order to remove LPS. Per protein, 10 mg of LPS removal beads were transferred into 1.5 ml reaction tubes (Eppendorf) and washed three times with 1 ml PBS. To bind LPS, protein solutions were added to the beads and mixed constantly for 30 min at 4°C. Afterwards, the tubes were placed in a magnet and the cleared protein solutions were transferred into fresh reaction tubes.

Protein solutions were filtered with 0.2 μm syringe filters (Pall). Protein concentrations were determined by photometric determination of the OD at 280 nm. In case of protein loss during downstream applications, such as buffer exchange and LPS removal, proteins were expressed in larger volumes (1–2 l).

Aliquots of protein solutions 98, 251, 253, and 256 (see additional file 1: SummaryOfResults.xls) were treated with LPS removal beads. Treated and untreated protein preparations were used for whole blood stimulations as described previously 42. Per ml of blood, 5 μg of protein was used. Fixed and permeabilized cells were stained with following antibodies (BD): CD4-peridinin chlorophyll A protein (PerCP), CD154 (CD40 Ligand)-phycoerythrin (PE), IFN-γ-allophycocyanin (APC). Cells were analyzed using a FACScalibur flow cytometer and Cell Quest software (both BD). CD4+ T cells were gated electronically and were quantified as the frequencies of cells that were double-positive for CD40 Ligand (CD40L) and IFN-γ. Frequencies resulting from stimulations with treated proteins vs. untreated proteins were related to each other. CD40L was shown to be a universal marker for activation of antigen-specific CD4+ T cells 42. We used CD40L up-regulation as a marker for T cell activation upon stimulation with our recombinant E. coli proteins.

Proteins 251 and 96 were treated with polymyxin B beads and LPS contents in the protein solutions were quantified with the Limulus amoebocyte lysate (LAL) test by Profos (Regensburg, Germany).

CAI values 25 were determined using the EMBOSS software package (4.0.0) and the codon usage file Eecoli_high.cut 43; the number of rare codons was established using the NIH Rare Codon Calculator 44; for prediction of transmembrane helices we used the CBS prediction server TMHMM (2.0) 45.