Learn more: PMC Disclaimer | PMC Copyright Notice
Engineering Escherichia coli for Increased Productivity of Serine-Rich Proteins Based on Proteome Profiling
Abstract
Variations in proteome profiles of Escherichia coli in response to the overproduction of human leptin, a serine-rich (11.6% of total amino acids) protein, were examined by two-dimensional gel electrophoresis. The levels of heat shock proteins increased, while those of protein elongation factors, 30S ribosomal protein, and some enzymes involved in amino acid biosynthesis decreased, after leptin overproduction. Most notably, the levels of enzymes involved in the biosynthesis of serine family amino acids significantly decreased. Based on this information, we designed a strategy to enhance the leptin productivity by manipulating the cysK gene, encoding cysteine synthase A. By coexpression of the cysK gene, we were able to increase the cell growth rate by approximately twofold. Also, the specific leptin productivity could be increased by fourfold. In addition, we found that cysK coexpression can improve the production of another serine-rich protein, interleukin-12 β chain, suggesting that this strategy may be useful for the production of other serine-rich proteins as well. The approach taken in this study should be useful in designing a strategy for improving recombinant protein production.
Escherichia coli has been the most widely used host for the production of various recombinant proteins. The yield of a recombinant protein is generally proportional to both the final cell density and the specific protein productivity (25). Therefore, various strategies have been developed to cultivate E. coli to high densities and/or to optimize expression systems for increased specific protein productivity. However, overexpression of recombinant proteins results in a rapid stress response and changes in E. coli metabolism (10, 33). These cellular responses may lead to plasmid instability, ribosome destruction, growth inhibition, or even cell lysis (6, 11, 21), all of which negatively affect recombinant protein production.
Recent advances in genomics and transcriptomics have allowed many functional genomic-based approaches to be taken toward understanding global metabolic changes caused by genotypic and/or environmental changes (26, 32, 36, 38, 39). Similarly, proteome profiling can also be employed to examine the changes in the expression levels of many proteins under particular conditions to be compared (3, 29, 39). Even though transcriptome profiling is truly useful, as it can generate a full spectrum of data on the expression levels of all of the genes at the mRNA level, there is growing evidence showing poor correlation between mRNA and protein abundances for a number of genes (2, 14). However, proteome profiling is presently limited by the fact that many fewer proteins than genes can be identified on a two-dimensional (2D) gel. Nonetheless, it is possible to identify protein spots showing altered expression levels, which may help us to understand metabolic and physiological changes and thereby to design metabolic and cellular engineering strategies. Several groups have carried out proteome analysis of E. coli cells overproducing recombinant proteins and have described metabolic changes under these conditions (19, 33). However, the results obtained from these studies have not been extended to actual engineering of cells to achieve enhanced recombinant protein production.
Here we report profiling of the proteome of recombinant E. coli during the overproduction of human leptin, identification of a target gene to be manipulated, and engineering of metabolic pathways to achieve increased leptin productivity. In order to understand how the overproduction of foreign proteins influences the synthesis of critical cellular proteins and thus to identify relevant target genes to be engineered, proteome analyses were performed (i) in the presence of the control plasmid only, (ii) before and after induction, and (iii) on inclusion body (IB) fractions.
MATERIALS AND METHODS
Bacterial strains and plasmids.
The bacterial strains and plasmids used in this study are shown in Table Table1.1. E. coli XL1-Blue was used as a host strain for cloning and maintenance of plasmids. E. coli BL21(DE3) was used as a host strain for the production of recombinant proteins. For the expression of the E. coli BL21(DE3) cysK gene, pAC104CysK (Fig. (Fig.1A)1A) was constructed as follows. The forward primer 5′-GCGAATTCATGAGTAAGATTTTTGAAGATAA-3′ was designed to contain an EcoRI site (underlined) immediately upstream of the start codon (ATG) of the cysK gene. The reverse primer 5′-GCGAATTCTATATACTGTTGCAATTCTTTCTC-3′ was designed to contain an EcoRI site (underlined). PCR was performed in a PCR thermal cycler (Takara Shuzo Co., Shiga, Japan) by using the High Fidelity PCR system (Boehringer, Mannheim, Germany) according to the manufacturer's instruction. The PCR product was digested with EcoRI prior to being cloned into the same restriction enzyme site in p10499A (30). The resulting plasmid, p104CysK, was digested with EcoRV and ScaI, and the fragment containing the cysK expression module was cloned into the EcoRV site of pACYC184 to yield pAC104CysK. The cysK gene is constitutively expressed by using this plasmid. Plasmids used for the production of human leptin and human granulocyte colony-stimulating factor (G-CSF) were pEDOb5 and pEDCSFmII, respectively, and were described previously (17, 18).
Schematic diagrams of plasmids pAC104CysK (A) and pEDIL-12p40 (B). The 966-bp PCR product which encodes the E. coli BL21(DE3) cysK gene was digested with EcoRV and ScaI and cloned into pACYC184 at the EcoRV site to make pAC104CysK. The 918-bp PCR product which encodes mature human interleukin-12 β chain was digested with AseI and BamHI and cloned into pET21c at the NdeI and BamHI sites to make pEDIL-12p40.
TABLE 1.
Bacterial strains and plasmids used in this study
| Strain or plasmid | Relevant characteristics | Reference or source |
|---|---|---|
| E. coli strains | ||
| XL1-Blue | supE44 hsdR17 recA1 endA1 gyrA96 thi relA1 | Stratagenea |
| BL21 (DE3) | lac F′[proAB+laclqlacZΔM15 Tn10(Tetr)] F−ompT hsdSB (rB− mB−) gal dcm (DE3) | Novagenb |
| Plasmids | ||
| p10499A | 4.20 kb, Apr, gntT104promoter | Lab stock |
| p104CysK | 5.20 kb, Apr, gntT104promoter | This study |
| pACYC184 | 4.30 kb, Cmr, Tcr | New England Biolabsc |
| pAC104CysK | 6.20 kb, Cmr Tcr, gntT104promoter | This study |
| pEDCSFmII | 5.90 kb, human G-CSF gene, T7 promoter | 18 |
| pEDOb5 | 5.75 kb, Apr, human obese gene, T7 promoter | 17 |
| pUC18/p40 | 3.67 kb, Apr, human IL-12p40 gene | BCRd |
| pEDIL-12p40 | 6.1 kb, Apr, human IL-12p40 gene, T7 promoter | This study |
For the expression of the mature interleukin-12 β chain gene, plasmid pIL-12p40 (Fig. (Fig.1B)1B) was constructed as follows: The forward primer 5′-GGCTAGCATTAATGATATGGGAACTGAAGAAAGAT-3′ was designed to contain an AseI site (underlined) immediately upstream of the codon for isoleucine (ATA), which is the first amino acid of the mature interleukin-12 β chain. The reverse primer 5′-GCCGGATCCTTATTAACTGCAGGGCACAGA-3′ was designed to contain a BamHI site (underlined) and tandem TAA stop codons immediately after the final serine codon (AGT). The PCR product was digested with NdeI and BamHI and was cloned into the same restriction sites of pET21c. The interleukin-12 β chain gene was expressed from the strong T7 promoter by induction with isopropyl-β-d-thiogalactopyranoside (IPTG) (Sigma Chemical Co., St. Louis, Mo.). All DNA manipulations, including restriction digestion, ligation, and agarose gel electrophoresis, were carried out as described by Sambrook et al. (34).
Fed-batch culture conditions.
Fed-batch cultures were carried out in a 6.6-liter jar fermentor (Bioflo 3000; New Brunswick Scientific Co., Edison, N.J.) containing 1.8 liters of R/2 medium plus 20 g of glucose per liter. The R/2 medium (pH 6.8) contains, per liter, 2 g of (NH4)2HPO4, 6.75 g of KH2PO4, 0.85 g of citric acid, 0.7 g of MgSO4 · 7H2O, and 5 ml of a trace metal solution. The trace metal solution contains, per liter of 5 M HCl, 10 g of FeSO4 · 7H2O, 2.25 g of ZnSO4 · 7H2O, 1 g of CuSO4 · 5H2O, 0.5 g of MnSO4 · 5H2O, 0.23 g of Na2B4O7 · 10H2O, 2 g of CaCl2 · 2H2O, and 0.1 g of (NH4)6MO7O24. A seed culture was prepared in a 1-liter flask containing 200 ml of R/2 medium. Except for periods when the pH increased due to glucose depletion, it was kept at 6.8 by adding 28% (vol/vol) ammonia water. The dissolved oxygen concentration was kept at 40% of air saturation by automatically increasing the agitation speed to 1,000 rpm and by changing the percentage of pure oxygen. A nutrient feeding solution was added by using the pH-stat (with high limit) feeding strategy (25). The feeding solution contained 800 g of glucose per liter and 20 g of MgSO4 · 7H2O per liter. When the pH rose to a value of 0.08 greater than its set point (pH 6.8) due to the depletion of glucose, the appropriate volume of feeding solution was automatically added to increase the glucose concentration in the culture broth to 0.7 g/liter. Cell growth was monitored by measuring the optical density at 600 nm (OD600) (DU series 600 spectrophotometer; Beckman, Fullerton, Calif.). For the expression of leptin, G-CSF, and interleukin-12 β chain, cells were induced with 1 mM IPTG at an OD600 of 30 or 90.
One-dimensional gel electrophoresis.
For protein quantification, cells at the same concentrations were harvested by centrifugation at 3,500 × g for 5 min at 4°C. Protein samples were analyzed by electrophoresis on sodium dodecyl sulfate-12% (wt/vol) polyacrylamide gels as described by Laemmli (22). The gels were stained with Coomassie brilliant blue R250 (Bio-Rad, Hercules, Calif.), and the protein bands were quantified with a GS-710 calibrated imaging densitometer (Bio-Rad).
2D gel electrophoresis and peptide mass fingerprinting.
2D gel electrophoresis experiments were carried out with a Protean II xi 2-D cell (Bio-Rad) by procedures described previously (39). Culture broth was centrifuged for 5 min at 3,500 × g and 4°C. The pellet was washed four times with TE solution (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]) and suspended in double-distilled water, followed by four cycles of sonication (each for 10 s at 10% of maximum output with high-intensity ultrasonic liquid processors [Sonics & Material Inc., Newtown, Conn.]). By centrifugation of the cell extract at 10,000 × g and 4°C for 20 min, the supernatant containing soluble proteins and the pellet containing inclusion bodies were obtained. The pellet was washed twice with 1% (vol/vol) Triton X-100 and once with double-distilled water to remove contaminants. After protein quantification by the Bradford assay with bovine serum albumin as a standard (8), protein samples (200 μg) were dried by vacuum centrifugation, resuspended in 340 μl of isoelectric focusing denaturation buffer {9 M urea, 0.5% 3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate [CHAPS], 10 mM dithiothreitol, 0.2% [wt/vol] Bio-lyte pH 3 to 10, 0.001% [wt/vol] bromophenol blue), and carefully loaded on immobilized pH gradient strips (pH 3 to 10; 17 cm) (Bio-Rad). The loaded immobilized pH gradient strips were rehydrated for 12 h and focused at 20°C for 15 min at 250 V, followed by 60,000 V · h with a Protean isoelectric focusing cell (Bio-Rad). The strips were exchanged in equilibration buffer and then were placed on sodium dodecyl sulfate-12% polyacrylamide gels prepared by the standard protocol (22). Protein spots were visualized with a silver staining kit (Amersham Biosciences, Uppsala, Sweden), and the stained gels were scanned with a GS-710 calibrated imaging densitometer (Bio-Rad). Melanie II software (Bio-Rad) was used to identify spots and to quantify spot densities on a volume basis (i.e., integration of spot optical intensity over the spot area). Matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) mass spectrometer analysis was carried out as described previously (15).
RESULTS AND DISCUSSION
Fed-batch culture and proteome profiles.
pH-stat fed-batch culture of E. coli BL21(DE3) was first carried out. E. coli BL21(DE3) was grown to an OD600 of 148 (74 g [dry weight] of cells/liter), with an apparent specific growth rate of 0.49 h−1 (Fig. (Fig.2A).2A). When the OD600 reached 30, sample S1 was taken for proteome analysis. pH-stat fed-batch culture of recombinant E. coli BL21(DE3)(pEDOb5) was carried out next. The apparent specific growth rate was 0.15 h−1, which is less than a half of that obtained with the strain without the plasmid. Just before induction at an OD600 of 30, sample S2 was taken. The dry cell weight and the maximum leptin content at 8 h after induction were 24 ± 0.5 g/liter and 41% ± 1.8% of the total protein, respectively (Fig. (Fig.2C).2C). At this point, sample S3 was taken. The proteome profiles of S1, S2, and S3 were analyzed and compared. From over 1,000 spots on the 2D gels, we identified 88 total proteins by comparing them with the E. coli SWISS 2D polyacrylamide gel electrophoresis database (http://kr.expasy.org/cgi-bin/map2/def?ECOLI) or by MALDI-TOF mass spectrometry analysis (Table (Table2).2). Considering the error range of detection, we accepted quantitative differences of above a 1.5-fold change as an increase or of below a 0.6-fold change as a decrease.
Time profiles of cell growth and recombinant protein production during the fed-batch cultures. (A) Cell densities (OD600) of E. coli BL21(DE3) (○) and E. coli BL21(DE3)(pACYC184) (▿); (B) cell density (OD600) of E. coli BL21(DE3)(pAC104CysK); (C to H) cell density (OD600) (□), dry cell weight (DCW) (○), and recombinant protein content (▴) without (C, E, and G) and with (D, F, and H) cysK coexpression for the production of leptin (C, D, E, and F) and interleukin-12 β chain (G and H). The dashed lines indicate the time of induction. S1, S2, S3, S4, and S5 are the sampling points for proteome analyses.
TABLE 2.
Proteins identified from 2D electrophoresis
| Spot no. | Protein name | Method for identification | AC Swiss Prot no. | pI/molecular mass (kDa) | Protein description | Protein level ratio (fold changea)
| |||
|---|---|---|---|---|---|---|---|---|---|
| S2/S1 | S3/S2 | S4/S2 | S5/S2 | ||||||
| 1 | AcnB | Gel match | P36683 | 5.24/75.9 | Aconitate hydratase 2 | 0.66 | 1.5 (Δ) | 0.61 (▿) | 1.6 (Δ) |
| 2 | FusA or EF-G | Gel match | P02996 | 5.21/75.5 | Elongation factor G | 0.57 (▿) | 1.3 | 0.97 | 0.78 |
| 3 | AsnS | Gel match | P17242 | 5.64/92.8 | Asparaginyl-tRNA synthetase | 1.4 | 1.4 | 3.3 (ΔΔΔ) | 1.4 |
| 4 | AceF | Gel match | P06959 | 5.01/77.5 | Dihydrolipoamide acetyltransferase component of pyruvate dehydrogenase complex | 1.5 (Δ) | 1.8 (Δ) | 1.6 (Δ) | 12 (ΔΔΔ) |
| 5 | DnaK | Gel match | P04475 | 4.81/69.6 | Chaperone protein DnaK | 0.87 | 1.5 (Δ) | 1.5 (Δ) | 1.8 (Δ) |
| 6 | PtsI | Gel match | P08839 | 4.78/59.8 | Phosphoenolpyruvate-protein phosphotransferase | 1.0 | 1.0 | 0.89 | 0.85 |
| 7 | NusA | Gel match | P03003 | 4.55/61.1 | N utilization substance protein A (NusA protein) | 3.3 (ΔΔΔ) | 0.52 (▿) | 0.55 (▿) | 0.47 (▿▿) |
| 8 | MopA or GroEL | Gel match | P06139 | 4.85/56.7 | 60-kDa chaperonin (GroEL protein) | 1.1 | 1.5 (Δ) | 1.6 (Δ) | 1.9 (Δ) |
| 9 | Tig | Gel match | P22257 | 4.83/51.0 | Trigger factor | ||||
| 10 | HtpG | Gel match | P10413 | 5.06/65.6 | Chaperone protein HtpG (heat shock protein HtpG) | 0.83 | 1.1 | 1.5 (Δ) | 1.8 (Δ) |
| 11 | AtpD | Gel match | P00824 | 4.90/47.7 | ATP synthase beta chain | 0.68 | 1.5 (Δ) | 1.0 | 0.86 |
| 12 | Icd or IcdA | Gel match | P08200 | 5.02/46.0 | Isocitrate dehydrogenase | 0.96 | 0.94 | 1.1 | 1.0 |
| 13 | GlnA | Gel match | P06711 | 5.25/53.8 | Glutamine synthetase | 0.88 | 0.94 | 1.0 | 0.99 |
| 14 | IlvC | Gel match | P05793 | 5.26/52.0 | Ketol-acid reductoisomerase | 0.91 | 0.69 | 0.89 | 0.71 |
| 15 | GlpK | Gel match | P08859 | 5.30/50.6 | Glycerol kinase (glycerokinase) | 1.0 | 1.1 | 0.95 | 0.58 |
| 16 | Eno | Gel match | P08324 | 5.34/46.5, 5.29/46.2 | Enolase (2-phosphoglycerate dehydratase) | ||||
| 17 | TufA or EF-Tu | Gel match | P02990 | 5.32/44.6 | Elongation factor Tu | 0.60 (▿) | 0.29 (▿▿▿) | 0.95 | 1.2 |
| 18 | FabD or TfpA | Gel match | P25715 | 5.37/44.8 | Malonyl coenzyme A-acyl carrier protein transacylase | 1.3 | 0.90 | 1.4 | 1.0 |
| 19 | LeuC | Gel match | P30127 | 5.42/44.4, 5.95/51.6 | 3-Isopropylmalate dehydratase large subunit | 1.2 | 0.60 (▿) | 0.80 | 0.59 (▿) |
| 20 | SdhA | Gel match | P10444 | 5.74/63.7 | Succinate dehydrogenase flavoprotein subunit | 0.7 | 1.9 (Δ) | 1.63 (Δ) | 1.6 (Δ) |
| 21 | OppA | Gel match | P23843 | 5.93/56.1 | Periplasmic oligopeptide-binding protein | 1.0 | 0.37 (▿▿) | 1.0 | 0.96 |
| 22 | TrpD | Gel match | P00904 | 6.08/55.9 | Anthranilate synthase component II; anthranilate | 1.0 | 0.37 (▿▿) | 0.71 | 1.1 |
| 23 | GuaB or GuaR | Gel match | P06981 | 6.01/55.0 | Inosine-5′-monophosphate dehydrogenase | 1.2 | 1.2 | 1.2 | 1.3 |
| 24 | AtpA | Gel match | P00822 | 5.84/53.1 | ATP synthase alpha chain | 1.3 | 0.85 | 0.69 | 0.82 |
| 25 | LpdA or Lpd | MALDI-TOF | P00391 | 5.81/53.6 | Dihydrolipoamide dehydrogenase | 1.3 | 1.0 | 0.69 | 0.91 |
| 26 | DppA | Gel match | P23847 | 5.69/52.1 | Periplasmic dipeptide transport protein | 1.0 | 0.60 (▿) | 1.0 | 0.95 |
| 27 | GlyA | MALDI-TOF | P00477 | 6.04/45.9, 5.94/46.1 | Serine hydroxymethyltransferase (serine methylase) | 0.60 (▿) | 0.39 (▿▿▿) | 1.3 | 0.58 (▿) |
| 28 | CarA or PyrA | Gel match | P00907 | 5.91/44.0 | Carbamoyl-phosphate synthase small chain | 0.55 (▿) | 0.60 (▿) | 1.1 | 0.94 |
| 29 | FbaA or Fba | MALDI-TOF | P11604 | 5.55/40.6 | Fructose-bisphosphate aldolase class II | 0.80 | 0.69 | 0.72 | 0.75 |
| 30 | SerC or PdxF | Gel match | P23721 | 5.34/40.2 | Phosphoserine aminotransferase | 0.85 | 0.90 | 0.89 | 0.72 |
| 31 | SucC | Gel match | P07460 | 5.30/42.3 | Succinyl-coenzyme A synthetase beta chain | 0.81 | 1.1 | 1.1 | 1.1 |
| 32 | LivJ | Gel match | P02917 | 5.28/41.9 | Leu/Ile/Val-binding protein | 1.0 | 0.6 (▿) | 1.1 | 0.46 (▿▿) |
| 33 | Pgk | Gel match | P11665 | 5.07/41.9, 5.02/41.7 | Phosphoglycerate kinase | 0.60 (▿) | 1.1 | 1.7 (Δ) | 0.79 |
| 34 | MalE | Gel match | P02928 | 5.08/41.1 | Maltose-binding periplasmic protein | 1.3 | 1.5 (Δ) | 0.32 (▿▿) | 0.44 (▿▿) |
| 35 | LivK | Gel match | P04816 | 5.00/41.4 | Leucine-specific binding protein | 1.4 | 0.6 (▿) | 0.73 | 0.95 |
| 36 | RfaD or HtrM | Gel match | P17963 | 4.85/36.8 | ADP-l-glycero-d-manno-heptose-6-epimerase | 1.3 (ΔΔΔ) | 0.08 (▿▿▿) | 0.44 (▿▿) | 0.16 (▿▿▿) |
| 37 | PotD | Gel match | P23861 | 4.77/35.8 | Spermidine/putrescine-binding periplasmic protein | 2.6 (ΔΔ) | 0.84 | 0.36 (▿▿) | 1.1 |
| 38 | TalB | Gel match | P30148 | 5.01/35.8 | Transaldolase B | 1.3 | 0.60 (▿) | 1.5 (Δ) | 0.49 (▿▿) |
| 39 | Tsf or EF-Ts | Gel match | P02997 | 5.15/33.6 | Elongation factor Ts | 1.0 | 0.60 (▿) | 0.83 | 0.63 |
| 40 | Mdh | Gel match | P06994 | 5.55/35.5 | Malate dehydrogenase | 0.92 | 0.99 | 0.86 | 0.76 |
| 41 | CysK | Gel match | P11096 | 5.81/36.0 | Cysteine synthase A | 0.59 (▿) | 0.44 (▿▿) | 12 (ΔΔΔ) | 11 (ΔΔΔ) |
| 42 | ManX or PtsL | Gel match | P08186 | 5.17/26.1 | Phosphotransferase system, mannose-specific IIAB component | 0.80 | 0.50 (▿) | 1.3 | 1.4 |
| 43 | AroG | Gel match | P00886 | 6.12/39.4 | Phospho-2-dehydro-3-deoxyheptonate aldolase | 0.81 | 0.75 | 0.68 | 0.69 |
| 44 | Sbp | MALDI-TOF | P06997 | 6.49/49.5 | Sulfate-binding protein | 0.57 (▿) | 0.77 | 1.5 (Δ) | 1.0 |
| 45 | GapA | Gel match | P06977 | 6.58/36.3 | Glyceraldehyde 3-phosphate dehydrogenase A | 1.1 | 0.49 (▿▿) | 1.5 (Δ) | 0.49 (▿▿) |
| 46 | PyrB | Gel match | P00479 | 6.13/35.3 | Aspartate carbamoyltransferase catalytic chain | 0.69 | 0.87 | 1.4 | 1.4 |
| 47 | FkpA | Gel match | P45523 | 7.08/33.2 | FKBP-type peptidyl-prolyl cis-trans isomerase FkpA | 1.0 | 0.14 (▿▿▿) | 1.2 | 0.58 (▿) |
| 48 | SodA | Gel match | P00448 | 6.44/22.9 | Superoxide dismutase [Mn] | 0.50 (▿) | 0.75 | 1.0 | 0.35 (▿▿) |
| 49 | GpmA or Gpm | MALDI-TOF | P31217 | 5.86/28.4 | Phosphoglycerate mutase 1 | 1.1 | 0.6 (▿) | 0.41 (▿▿) | 0.36 (▿▿) |
| 50 | Udp | Gel match | P12758 | 5.86/27.9 | Uridine phosphorylase | 0.73 | 0.81 | 1.6 (Δ) | 1.3/PICK> |
| 51 | YadK | Gel match | P37016 | 5.55/28.4 | Protein YadK | 1.3 | 1.0 | 1.3 | 0.88 |
| 52 | TpiA or Tpi | Gel match | P04790 | 5.57/26.9 | Triosephosphate isomerase | 1.2 | 0.50 (▿) | 1.5 (Δ) | 0.53 (▿) |
| 53 | Bla | Gel match | P00810 | 5.46/28.9 | β-Lactamase | ∞ (ΔΔΔ) | 1.1 | 1.3 | 0.43 (▿▿) |
| 54 | TrpA | Gel match | P00928 | 5.30/28.7 | Tryptophan synthase alpha chain | 1.2 | 0.56 (▿) | 0.80 | 0.58 (▿) |
| 55 | SspA or Ssp | Gel match | P05838 | 5.24/26.6 | Stringent starvation protein A | 2.4 (ΔΔ) | 0.72 | 1.2 | 0.17 (▿▿▿) |
| 56 | HisJ | Gel match | P39182 | 5.05/28.6 | Histidine-binding periplasmic protein | 1.5 (Δ) | 0.77 | 1.2 | 0.50 (▿) |
| 57 | FliY | Gel match | P39174 | 5.01/26.2, 5.11/25.8 | Cystine-binding periplasmic protein | 2.2 (ΔΔ) | 0.38 (▿▿) | 0.47 (▿▿) | 0.17 (▿▿▿) |
| 58 | HdhA or HsdH | Gel match | P25529 | 5.17/25.0 | 7-Alpha-hydroxysteroid dehydrogenase | 1.3 | 1.5 (Δ) | 0.99 | 0.89 |
| 59 | Upp or UraP | Gel match | P25532 | 5.29/23.8 | Uracil phosphoribosyltransferase | 1.1 | 0.45 (▿▿) | 1.5 (Δ) | 0.66 |
| 60 | GrpE | Gel match | P09372 | 4.68/25.5 | GrpE protein (HSP-70 cofactor) | 1.7 (Δ) | 0.80 | 0.93 | 1.5 (Δ) |
| 61 | AccB or FabE | Gel match | P02905 | 4.57/22.0 | Biotin carboxyl carrier protein of acetyl-coenzyme A carboxylase | ∞ (ΔΔΔ) | 1.1 | 0.99 | 0.74 |
| 62 | AhpC | Gel match | P26427 | 5.01/21.5 | Alkyl hydroperoxide reductase C22 protein | 4.6 (ΔΔΔ) | 0.66 | 0.95 | 0.80 |
| 63 | Crr | Gel match | P08837 | 4.57/20.0, 4.68/18.9 | Phosphotransferase system, glucose-specific IIA component | 1.3 | 0.51 (▿) | 0.94 | 1.4 |
| 64 | DksA | Gel match | P18274 | 4.90/18.7 | DnaK suppressor protein | 1.7 (Δ) | 0.82 | 1.0 | 0.85 |
| 65 | SgaH | Gel match | P39304 | 5.14/19.5 | Probable hexulose-6-phosphate synthase | 9.4 (ΔΔΔ) | 0.47 (▿▿) | 0.54 (▿) | 0.48 (▿▿) |
| 66 | AroK | Gel match | P24167 | 5.30/17.9 | Shikimate kinase I | 2.1 (ΔΔ) | 1.0 | 1.4 | 0.80 |
| 67 | SodB | Gel match | P09157 | 5.53/22.1 | Superoxide dismutase [Fe] | 1.0 | 0.70 | 1.0 | 0.33 (▿▿) |
| 68 | PpiB | Gel match | P23869 | 5.51/17.7 | Peptidyl-prolyl cis-trans isomerase B | 1.1 | 0.73 | 0.93 | 0.29 (▿▿▿) |
| 69 | RplI | Gel match | P02418 | 6.20/19.8, 6.17/15.7 | 50S ribosomal protein L9 | 0.60 (▿) | 1.0 | 1.1 | 1.1 |
| 70 | YfiA | Gel match | P11285 | 6.16/15.2 | Protein YfiA | 0.74 | 0.42 (▿▿) | 0.58 (▿) | 0.72 |
| 71 | YbdQ | Gel match | P39177 | 6.08/15.5 | Unknown protein from 2D polyacurylamide gel electrophoresis | 0.88 | 1.0 | 1.3 | 0.69 |
| 72 | RbfA | Gel match | P09170 | 6.00/15.6 | Ribosome-binding factor A | 0.69 | 1.4 | 0.93 | 1.1 |
| 73 | RplU | Gel match | P02422 | 6.71/10.3 | 50S ribosomal protein L21 | 0.96 | 0.98 | 0.86 | 0.43 (▿) |
| 74 | Hns | Gel match | P08936 | 5.45/15.6 | DNA-binding protein H-NS (histone-like protein HLP-II) | 4.0 (ΔΔΔ) | 0.81 | 0.55 (▿) | 0.64 |
| 75 | Ndk | Gel match | P24233 | 5.59/15.2 | Nucleoside diphosphate kinase | 1.0 | 0.92 | 1.4 | 0.90 |
| 76 | AtpC | Gel match | P00832 | 5.48/14.8 | ATP synthase epsilon chain | 0.97 | 1.5 (Δ) | 0.62 | 0.52 (▿) |
| 77 | RpsF | Gel match | P02358 | 5.31/15.8, 5.15/15.8, 5.26/15.8 | 30S ribosomal protein S6 | 1.4 | 0.53 (▿) | 1.1 | 0.5 (▿) |
| 78 | Bcp | Gel match | P23480 | 5.02/15.8 | Bacterioferritin comigratory protein | 1.3 | 0.60 (▿) | 0.99 | 1.2 |
| 79 | GreA | Gel match | P21346 | 4.68/15.9 | Transcription elongation factor GreA | 1.6 (Δ) | 0.58 (▿) | 0.38 (▿▿) | 0.97 |
| 80 | GroES or MopB | Gel match | P05380 | 5.15/15.6 | 10-kDa chaperonin (GroES protein) | 1.4 | 0.65 | 0.67 | 1.1 |
| 81 | UspA | Gel match | P28242 | 5.14/15.1 | Universal stress protein A | 1.5 (Δ) | 1.7 (Δ) | 0.53 (▿) | 1.1 |
| 82 | YjgF | Gel match | P39330 | 5.29/13.0 | Protein YjgF | 2.3 (ΔΔ) | 1.1 | 0.81 | 1.4 |
| 83 | TrxA or TsnC | Gel match | P00274 | 4.67/11.5 | Thioredoxin I | 2.6 (ΔΔ) | 1.2 | 1.1 | 1.3 |
| 84 | HdeB | Gel match | P26605 | 4.85/11.2 | Protein HdeB (10K-L protein) | 8.9 (ΔΔΔ) | 1.0 | 0.36 (▿▿) | 0.90 |
| 85 | IbpA | MALDI-TOF | P29209 | 5.57/15.8 | 16-kDa heat shock protein A | ∞ (ΔΔΔ) | ∞ (ΔΔΔ) | ||
| 86 | IbpB | MALDI-TOF | P29210 | 5.19/16.1 | 16-kDa heat shock protein B | ∞ (ΔΔΔ) | ∞ (ΔΔΔ) | ||
| 87 | Leptin | MALDI-TOF | P41159 | 5.88/18.6 | Obese protein | ∞ (ΔΔΔ) | ∞ (ΔΔΔ) | ||
| 88 | OmpF | MALDI-TOF | P02931 | 4.61/36.2 | Outer membrane protein F (porin OmpF) | ∞ (ΔΔΔ) | ∞ (ΔΔΔ) | ||
Effect of plasmid presence.
To examine the influence of plasmid presence on proteome variation, we compared the proteome of S2 with that of S1 (Fig. 3A and B). The levels of about 33% (29 proteins) of the total identified proteins were altered in the presence of the plasmid (Table (Table2).2). Among the 20 proteins that showed higher expression levels were stationary-phase-responsive proteins, including GrpE, HdeB, AhpC, SspA, UspA, and Hns. These proteins are known to protect cells from stressful conditions such as heat shock, ethanol, H2O2, and NaCl. The sigma factor σS, encoded by the rpoS gene, is a central regulator for the expression of many stationary-phase-responsive genes (23). However, the above-mentioned proteins, except for HdeB, are not regulated by σS. These stationary-phase-responsive proteins are known to be induced under conditions of slow growth (16). Therefore, it seems that the presence of a plasmid itself resulted in slower growth, as shown by the observation that the growth rate of E. coli BL21(DE3)(pEDOb5) was less than half of that of E. coli BL21(DE3) and consequently induced expression of the stationary-phase-responsive proteins. To examine if this is truly due to the presence of a plasmid itself, we carried out pH-stat fed-batch culture of E. coli BL21(DE3)(pACYC184). The apparent specific growth rate was 0.22 h−1, which is less than a half of that of E. coli BL21(DE3) without the plasmid (Fig. (Fig.2A).2A). These results suggest that plasmid presence itself can induce stationary-phase-responsive proteins during fed-batch culture. Other than these proteins, the levels of 14 proteins (AceF, NusA, RfaD, PotD, TalB, Bla, HisJ, FliY, AccB, DksA, SgaH, AroK, GreA, and YjgF) increased in the presence of the plasmid (Table (Table22).
2D gel electrophoresis of samples for Fig. Fig.2.2. The soluble fractions of samples S1 (A), S2 (B), S3 (C), S4 (E), and S5 (F), as well as the insoluble fraction of sample S3 (D), were analyzed. Identified proteins shown by numbers corresponding to those in Table Table2.2. Proteins showing increased and decreased levels are indicated by circles and rectangles, respectively.
The levels of nine proteins (EF-G, EF-Tu, GlyA, CarA, Pgk, CysK, Sbp, SodA, and RplI) decreased in the presence of the plasmid. Since the levels of protein elongation factors (EF-G and EF-Tu) and ribosomal proteins (RplI) decreased, the protein translational activity seems to be negatively affected in the presence of the plasmid. This result is consistent with previous reports showing that plasmid presence decreased the levels of proteins involved in translation as well as ribosomal subunit pools (4, 7). The levels of most of the identified proteins involved in amino acid biosynthesis, including GlnA, IlvC, LeuC, TrpA, TrpD, SerC, and AroG, did not change much in the presence of the plasmid. However, the level of shikimate kinase I (AroK), which is involved in a common pathway of aromatic amino acid biosynthesis, increased by twofold in the presence of the plasmid. On the other hand, the levels of GlyA and CysK, which are involved in the synthesis of serine family amino acids, decreased in the presence of the plasmid. Taken together, these results suggest that the presence of the plasmid negatively affects the cellular capacity of protein synthesis and biosynthesis of serine family amino acids. At present, it is not clear why the levels of enzymes for serine family amino acid synthesis decreased in the presence of the plasmid.
Effect of leptin overproduction.
In order to investigate the effect of overproducing leptin on the host cell physiology, we compared the proteomes of S2 and S3 (Fig. 3B, C, and D). Overproduction of leptin altered the levels of 47% (41 proteins) of the total identified proteins (Table (Table2).2). In agreement with previous studies (10, 19, 33), the levels of heat shock proteins such as DnaK and GroEL increased upon overproduction of leptin. Also, relatively large amounts of the small heat shock proteins IbpA and IbpB, which are good indicators for the formation of IBs (1, 24), were associated with the IB fraction (Fig. (Fig.3B).3B). Production of these proteins is known to be regulated by σ32 (12). This indicates that overproduction of recombinant protein (leptin) acts as a stress to the cells. The levels of nine proteins (AcnB, AceF, AtpD, SdhA, MalE, HdhA, AtpC, UspA, and OmpF) increased after leptin production (Table (Table22).
Twenty-eight proteins were repressed by the overproduction of leptin. The levels of proteins involved in protein synthesis, such as protein elongation factors (EF-Tu and EF-Ts) and 30S ribosomal protein (RpsF), decreased. It was interesting that the protein elongation factor EF-Tu was associated with IBs (Fig. (Fig.3D).3D). This coincided with the decreased level of soluble EF-Tu, indicating that the association of EF-Tu with IBs may be one reason for the observed reduction in overall protein synthesis. This in turn shows that the overproduction of recombinant proteins negatively influences the capacity of the cellular translational machinery (37).
We also observed decreased levels of enzymes involved in the biosynthesis of serine family (GlyA and CysK), aromatic family (AroG, TrpA, and TrpD), and branched-chain family (LeuC) amino acids following induction of leptin overproduction. This suggests that the overproduction of leptin leads to an imbalance in free amino acid stocks and translational machinery (5, 13, 31). The levels of enzymes involved in the synthesis of serine family amino acids were most strongly downregulated by leptin overproduction. This can be explained by the fact that leptin contains much more serine (11.6% of total amino acids) than common E. coli proteins do (average of 5.6%). It should be remembered that the presence of a plasmid itself decreased the levels of GlyA and CysK by 40%. Leptin production resulted in a further decrease of their levels by ca. 60%.
Interestingly, the levels of OmpF and MalE were upregulated after overproduction of leptin, while those of GlyA, OppA, LivJ, and LivK were downregulated. All of these proteins are regulated by leucine-responsive regulatory protein (Lrp). Lrp activates transcription of the ompF and malE genes but represses that of the glyA, oppA, livJ, and livK genes (9, 28). Some metabolic pathways affected by Lrp are shown in Fig. Fig.4A.4A. Lrp activates 3-phosphoglycerate dehydrogenase (encoded by serA), which converts the glycolytic intermediate 3-phosphoglycerate to serine, while it represses serine hydroxymethyltransferase (encoded by glyA) and serine deaminase (encoded by sdaA) of the serine degradation pathway. It seems that leptin could not be produced immediately after induction due to the insufficient pool of available serine (Fig. (Fig.4A).4A). As time passed, Lrp activated the pathways for the biosynthesis of serine and other amino acids. This led to an increase in the amino acid pool required for protein synthesis and consequently to leptin production that reached the maximum level in 8 h.
Metabolic pathway of recombinant E. coli overproducing leptin, quantified from the S3/S2 (A), S4/S2 (B), and S5/S2 (C) protein level ratios. Boldface arrows indicate that the protein levels were increased by more than 1.5-fold, while dotted arrows indicate that the protein levels were decreased to less than 0.6-fold. Some metabolic pathways affected by Lrp are also indicated in panel A. Reactions regulated by Lrp enzymes are marked + and − for positive and negative regulation, respectively. Abbreviations: 6PGL, 6-phosphogluconolactone; 6PG, 6-phosphogluconate; 2K3D6PG, 2-dehydro-3-deoxy-6-phosphogluconate; R5P, ribose-5-phosphate; RL5P, ribulose-5-phosphate; X5P, xylulose-5-phosphate; S7P, sedoheptulose-7-phosphate; E4P, erythrose-4-phosphate; F6P, fructose-6-phosphate; F1,6P2, fructose-1,6-diphosphate; G3P, glyceraldehydes-3-phosphate; DHAP, dihydroxyacetone-phosphate; G1,3P2, glycerate-1,3-diphosphate; 3PG, 3-phosphate-glycerate; 2PG, 2-phosphate-glycerate; PEP, phosphoenolpyruvate; PYR, pyruvate; AcCoA, acetyl coenzyme A; CIT, citrate; ICT, isocitrate; α-KG, α-ketoglutarate; Suc-CoA, succinyl coenzyme A; SUC, succinic acid; FUM, fumaric acid; MAL, malic acid; OAA, oxaloacetate.
Enhanced productivity of human leptin by cysK coexpression.
We made an effort to search for a target protein from proteome analysis to shorten the long period required to reach the maximum leptin content. Developing a strategy for increasing the serine pool is difficult due to the complex nature of metabolic pathway regulation. During growth on glucose, 15% of the carbon assimilated in E. coli is converted to serine or its intermediates (35). As described above, the expression of the glyA gene was negatively regulated by Lrp during leptin production. The decreased GlyA has a positive effect on increasing the serine pool and consequently on leptin production. These results led us to examine the effects of the cysK gene on the biosynthesis of serine family amino acids. When the cysK gene was coexpressed in leptin-producing cells, the dry cell weight and the maximal leptin content obtained 2 h after induction were 26 ± 0.8 g/liter and 39% ± 1.1% of total protein, respectively (Fig. (Fig.2D),2D), which are nearly identical to the levels obtained in a strain without cysK coexpression. However, it should be noted that the maximal leptin content was reached at only 2 h after induction by coexpression of the cysK gene, compared to the 8 h required without cysK coexpression. Therefore, cysK coexpression increased leptin productivity by fourfold. A similar phenomenon was observed when cells were induced at an OD600 of 90 (Fig. 2E and F). When the cysK gene was coexpressed, the dry cell weight and the maximal leptin content obtained 3 h after induction were 51 ± 0.5 g/liter and 35% ± 1.5% of total protein, respectively, which are nearly identical to the levels obtained at 8 h without cysK coexpression.
Metabolic pathways affected by cysK coexpression.
In order to examine the changes in metabolism caused by cysK coexpression in leptin-producing cells, we analyzed proteomes of S4 (before induction) and S5 (after induction) of the cysK-coexpressing strain (Fig. 2G and H and 3E and F) and compared the profiles with that of S2 (Table (Table2).2). Two beneficial effects of cysK coexpression were evident. First, EF-Tu was no longer trapped in the IBs, and consequently more EF-Tu existed in soluble form (Table (Table2;2; Fig. 3E and F). This may be due to the increased levels of chaperones, including DnaK, GroEL, and HtpG, which helped correct folding of EF-Tu and prevented trapping of EF-Tu within IBs. The availability of more EF-Tu would have strengthened the protein biosynthesis capacity. Second, metabolic fluxes seemed to be altered to allow more efficient production of leptin. Before induction, the levels of TpiA, GapA, Pgk, and AceF were increased in cysK-coexpressing cells compared with control cells, while that of GpmA decreased. As shown in Fig. Fig.4B,4B, TpiA, GapA, and Pgk catalyze the reactions leading to be formation of 3-phosphoglycerate, which is further converted to serine. Through these metabolic changes, the cysK coexpression activated the serine biosynthetic pathway even prior to induction, which led to immediate production of leptin. After induction, the levels of these proteins were similar in cells with and without cysK coexpression (Fig. 4A and C). In summary, cysK coexpression resulted in the central metabolic pathway fluxes being adjusted to allow efficient production of leptin, a serine-rich protein. In addition to this positive effect, cysK coexpression made cells maintain a high protein biosynthetic capacity. All of these factors contributed to achieve high productivity and specific productivity of leptin. Even though the reasons for this phenomenon could not be explained, these results clearly demonstrate the effectiveness of the approach taken to improve the leptin productivity.
Enhanced productivity of another serine-rich protein by cysK coexpression.
To examine whether cysK coexpression can increase the specific productivity of other serine-rich proteins, interleukin-12 β chain (11.1% serine) was selected as another model protein. The amino acid compositions of E. coli proteins show no dramatic deviations from those in the complete protein sequence database, with leucine and alanine being most abundant and cysteine and tryptophan being least abundant (20). We therefore selected G-CSF (18.84% leucine and 11.59% alanine) as a control protein having an amino acids composition similar to that of E coli proteins.
As expected, production of interleukin-12 β chain was improved by cysK coexpression, while production of G-CSF was not affected. When recombinant E. coli BL21(DE3)(pEDIL-12p40) was induced at an OD600 of 30, the dry cell weight and the maximum interleukin-12 β chain content reached at 7 h after induction were 23 ± 0.7 g/liter and 9% ± 0.5% of the total protein, respectively (Fig. (Fig.2E).2E). On the other hand, when BL21(DE3)(pEDIL-12p40)(pAC104CysK) was induced at an OD600 of 30, the dry cell weight and the maximum interleukin-12 β chain content at 2 h after induction were 24 ± 0.8 g/liter and 8% ± 0.3% of the total protein, respectively (Fig. (Fig.2F).2F). The maximum G-CSF content (35% of the total protein) was reached at 2 and 3 h after induction with and without cysK coexpression, respectively. Therefore, cysK coexpression increased interleukin-12 β chain productivity by threefold. These results suggest that cysK coexpression may improve production of any serine-rich protein.
Improved cell growth by cysK coexpression during high-cell-density culture.
Consistent with other studies (4, 7), the introduction of the plasmid caused changes in the levels of cellular proteins and adversely affected cell growth. Interestingly, it was found from proteome profiling that the levels of proteins involved in the biosynthesis of serine family amino acids decreased during the high-cell-density culture of E. coli BL21(DE3). If this is one of the factors negatively affecting cell growth, cysK coexpression should cause recovery of the growth rate. Therefore, we compared the growth rates of wild-type E. coli BL21(DE3), E. coli BL21(DE3)(pACYC184), and E. coli BL21(DE3)(pAC104CysK)(Fig. 2A and B). The presence of the plasmid decreased the growth rate by 50% compared with that of the wild-type strain. Surprisingly, cysK coexpression restored the growth rate of the plasmid-carrying strain to nearly that of the wild type, indicating that the reduced biosynthesis of serine family amino acids caused a lower growth rate during the high-cell-density culture of E. coli BL21(DE3). To see if this is a universal phenomenon in E. coli, we carried out high-cell-density cultures of E. coli W3110 harboring pACYC184 or pAC104CysK. It was found that cysK coexpression did not increase the specific growth rate of recombinant W3110. Therefore, the beneficial effect of cysK coexpression on the growth rate may be specific for E. coli BL21(DE3).
In summary, the growth of recombinant BL21(DE3) and leptin production could be enhanced by cysK coexpression. Coexpression of the cysK gene could increase the biosynthetic flux of serine family amino acids and indirectly repress EF-Tu aggregation by inducing the expression of heat shock proteins, leading to improved cell growth and a three- to fourfold increase in the productivity of serine-rich recombinant proteins. It should be mentioned that the high productivities of serine-rich proteins achieved by cysK coexpression are not solely due to the improved cell growth. As seen above, the growth rate was increased by approximately twofold by cysK coexpression, while the productivities of recombinant proteins increased by three- to fourfold. This is the first report on improving recombinant protein productivity by engineering the metabolic pathways based on the results of proteome analysis. Consequently, we propose a strategy for the rational engineering of metabolic pathways and cellular properties based on the results of proteome profiling. The procedure is to (i) obtain the proteome profiles of E. coli (or recombinant E. coli) under different conditions of interest, (ii) identify potentially limiting enzymes in the biosynthetic pathways, (iii) examine theoretically and/or experientially the possible flux changes that can be achieved by amplifying (or knocking out) the activities of the enzymes identified, (iv) select the final candidate enzymes to be amplified (or knocked out), (v) examine the effects of this metabolic and cellular engineering on achieving the desired objectives, and (vi) repeat steps ii to iv until the objectives are accomplished. This strategy may be extended beyond serine-rich proteins to increase the yield and productivity of other recombinant proteins in industrial bioprocesses.
Acknowledgments
This work was supported by the National Research Laboratory Program (grant 2000-N-NL-01-C-237) of the Ministry of Science and Technology; the Basic Industrial Research Project of the Ministry of Commerce, Industry and Energy; and the Brain Korea 21 project. Hardware for the analysis of proteomes was supported by the IBM-SUR program.