Enhancing solubility of deoxyxylulose phosphate pathway enzymes for microbial isoprenoid production
© Zhou et al.; licensee BioMed Central Ltd. 2012
Received: 19 June 2012
Accepted: 7 November 2012
Published: 14 November 2012
Recombinant proteins are routinely overexpressed in metabolic engineering. It is well known that some over-expressed heterologous recombinant enzymes are insoluble with little or no enzymatic activity. This study examined the solubility of over-expressed homologous enzymes of the deoxyxylulose phosphate pathway (DXP) and the impact of inclusion body formation on metabolic engineering of microbes.
Four enzymes of this pathway (DXS, ISPG, ISPH and ISPA), but not all, were highly insoluble, regardless of the expression systems used. Insoluble dxs (the committed enzyme of DXP pathway) was found to be inactive. Expressions of fusion tags did not significantly improve the solubility of dxs. However, hypertonic media containing sorbitol, an osmolyte, successfully doubled the solubility of dxs, with the concomitant improvement in microbial production of the metabolite, DXP. Similarly, sorbitol significantly improved the production of soluble and functional ERG12, the committed enzyme in the mevalonate pathway.
This study demonstrated the unanticipated findings that some over-expressed homologous enzymes of the DXP pathway were highly insoluble, forming inclusion bodies, which affected metabolite formation. Sorbitol was found to increase both the solubility and function of some of these over-expressed enzymes, a strategy to increase the production of secondary metabolites.
Isoprenoids, a large family of natural compounds including many plant based pharmaceuticals such as artemisinin  and paclitaxel , are produced by the deoxyxylulose phosphate (DXP) pathway and/or the mevalonate (MVA) pathway in nature . The current industrial isoprenoid production methods include direct extraction from plants and semi-synthesis using plant metabolites . These processes are all restricted by the supply of specific plant materials, which are often affected by unpredictable factors including variations in weather and market fluctuations . To improve sustainability and production capacity of isoprenoids, heterologous biosynthesis from economical carbon sources in microbes has been intensively studied in the past decade [2, 5]. To date, there have been many successful reports of using the DXP pathway to produce isoprenoid nutraceuticals (e.g. lycopene [6–10] and other carotenoids [11, 12]) and pharmaceuticals (e.g. artemisinin precursors [13, 14] and paclitaxel precursors ).
To increase carbon flux through the DXP pathway, the enzymes involved were overexpressed [2, 6–14] to catalyze the bio-transformations of the DXP metabolites in vivo. So far, four enzymes (DXS, IDI, ISPD and ISPF) have been identified to be rate limiting based on a series of empirical studies [9, 15]. The expression levels of these four enzymes have also been semi-empirically optimized for paclitaxel precursor production . It is well accepted that some recombinant proteins can form insoluble aggregates (termed as inclusion bodies), generally regarded to be functionally inactive . However, the extents to which the over-expressed recombinant endogenous DXP enzymes forming inclusion bodies and their impacts on the flux through the pathway have yet to be systematically investigated.
This study examined the solubility status of all the DXP enzymes when overexpressed and attempted to demonstrate the importance of protein solubility in the production of secondary metabolites. Computational prediction was initially explored to evaluate the solubility status and empirical verifications were carried out in E. coli. An unanticipated and critical observation is that many DXP enzymes (DXS, ISPA, ISPG and ISPH) were found to be highly insoluble. Interestingly, the enzymes IDI, ISPD and ISPF, thought to be rate-limiting and hence useful for the enhancement of isoprenoids production , were found to be highly soluble. From these observations, it is now necessary to reevaluate the use of the other highly insoluble DXP enzymes for enhancing isoprenoid production. Attempts were also made to optimize the solubility of the insoluble enzymes and to examine the enhancements in isoprenoid production.
Solubility of over-expressed recombinant DXP pathway enzymes
Solubility of the enzymes in the DXP pathway predicted in silico and determined in vitro
Enzymatic activity of insoluble recombinant DXS
Improving solubility of DXS enhanced the production of DXP
To demonstrate that improved solubility of DXS results in enhanced production of DXP (committed metabolic intermediate in the DXP pathway), cells grown in sorbitol were lysed and the extracts quantified by LC-MS. It was found that concentrations of DXP were significantly higher in sorbitol treated cells as compared to control cells (Figure 4 C). Addition of sorbitol did not alter the production of DXP with cells overexpressing a nonfunctional DXS (Figure 4 C, the construction and characterization of the nonfunctional dxs was described in Additional file 3), indicative that the effect of sorbitol on cells overexpressing functional enzyme was likely be due to the increase in the concentration of soluble DXS. A parallel increase in the concentrations of MEP and MEC (Figure 1) were also observed (Additional file 4), suggesting that sorbitol increased the flux through the entire DXP pathway in cells overexpressing DXS.
Improvement of ERG12 solubility and overproduction of mevalonate phosphate
This study addressed an important and often overlooked issue of the solubility of over-expressed recombinant homologous or heterologous enzymes in metabolic engineering. Specifically, we investigated the solubility status of overexpressed DXP enzymes and a heterologous enzyme of the MVA pathway and the impacts on the production of critical precursor metabolites (DXP or MVAP), which are building blocks of all the isoprenoids. It was unexpected that four out of nine enzymes in DXP pathway (DXS, ISPA, ISPG and ISPH) were highly insoluble, despite being endogenous enzymes. Overexpression of DXS resulted in the accumulation of highly insoluble and non-functional (< 1% activity of the equivalent soluble form) enzyme. This observation cautions against the assumption that overexpression of an enzyme necessarily confers higher enzymatic activity. Interestingly, the combinatorial screening study  based on this contentious assumption identified three rate-limiting DXP enzymes (IDI, ISPD and ISPF), which incidentally were found to be highly soluble (Figure 2). It is thus not unreasonable to speculate that the previously thought to be ‘non rate-limiting’ enzymes found to be insoluble in this study, may serve to enhance the productions of isoprenoids when expressed in soluble forms.
Using DXS as a model enzyme, four commonly used fusion partners, trxA , nusA , slyD  and malE  were fused at the N-terminus of DXS in the attempt to increase solubility. The use of these fusion partners did not significantly increase the solubility of DXS (Additional file 5). The effectiveness of the fusion partners in enhancing protein solubility is largely protein-dependent and unpredictable [26, 34]. Cysteine residues on surface of DXS (C32, C330 and C457), may form non-specific disulfide bonds and result in protein aggregation . Site-directed mutagenesis of these residues also did not improve solubility (Additional file 6), suggesting that the aggregation of dxs protein was not due to disulfide bond mediated interactions.
Osmolytes have been shown to improve overexpressed proteins in E. coli. Sorbitol at high concentrations significantly improved DXS solubility and the production of the metabolic product (DXP) in E. coli, indicating that solubility of recombinant enzymes is an important factor in the production of secondary metabolites. Consistent with this suggestion was that ERG12, another model enzyme, also showed improved solubility and secondary metabolite production in the presence of sorbitol. The reason why the metabolic intermediates (DXP, MEP and MEC etc.) instead of final product (lycopene etc.) were used as read-out for characterization of DXS was that rate limiting step (ISPG) existed between the intermediates and the isoprenoid products (Zhou et al. PLoS One, In Press, Additional file 7). It is worthy to note that the use of sorbitol is acceptable as a proof-of-concept but may not be routinely used in microbial fermentation due simply to the cost involved . An alternative is to modify the host microbes (such as manipulation of cellular protein folding system [27, 35]) to render these proteins more soluble for industrial applications.
In this study, about half of the nine DXP proteins (DXS, ISPG, ISPH and ISPA) were found to be highly insoluble when overexpressed in E. coli. Insoluble DXS, the committed enzyme of the DXP pathway, showed significantly less enzymatic activity when compared to the equivalent amount of soluble enzyme in vitro. High concentration of sorbitol successfully increased the solubility of DXS and resulted in a parallel increase in the metabolic product (DXP). The strategy also improved both solubility and secondary metabolite production of ERG12, a critical enzyme in the mevalonate pathway. This study highlighted the importance of protein solubility in metabolic engineering of microbes for the overproduction of isoprenoids.
Bacteria strains and plasmids
Bacteria strains and plasmids used in this study were summarized in Additional file 8. All the DXP genes were amplified from E. coli genomic DNA and cloned into the modified pBAD-B (Invitrogen), pET-11a (Stratagene) and pQE30 (Qiagen) plasmids with 6xhis tag, SacI, XhoI restriction enzyme sites. Fusion partners (trxA, nusA, malE and slyD) were amplified from E. coli genomic DNA and cloned into pBAD-dxs with NcoI and SacI sites. Erg12 was amplified from S. cerevisiae genomic DNA and cloned into the modified pBAD-B plasmid with 6xhis tag, SacI and XhoI restriction enzyme sites. Dxs mutants R398A , C32A, C330A, C457A and C32A-C330A-C457A were generated according the ‘megaprimer’ protocol . Primers used in this study were summarized in Additional file 8. All the pET-11a, pBAD-B and pQE30 based plasmids were transformed into E. coli BL21-Gold (DE3), E. coli DH10B and E. coli M15 respectively. pAC-LYC was co-transformed with all the plasmids except pBAD-erg12.
E. coli growth and induction of protein expression
A colony was picked from agar plate, inoculated into 2xPY medium (20g/L Peptone, 10g/L Yeast extract, and 10g/L NaCl, pH=7) containing proper antibiotics, and incubated overnight. Ten microliter aliquots of overnight grown cell culture were inoculated into 1mL 2xPY medium in 14mL Falcon tube. Cells were grown at 37°C/300rpm till OD595 reached the range of 0.5~1.0. The cells were then induced with 1mM IPTG (E. coli BL21-Gold (DE3) and E. coli M15) or 10mM L-arabinose (E. coli DH10B) and grown at 37°C or 20°C for indicated time before collected for protein solubility assay or metabolite assay. Additives (sorbitol, betaine, phosphate, HEPES, mevalonate etc.) were also fed to cell culture upon induction if necessary.
Prediction and quantification of protein solubility
The revised WH algorithm  was used for prediction of protein solubility. Protein solubility was experimentally quantified by centrifugation  as described below. At 24h after induction, cell suspension equivalent to 1mL OD595=1.0 cells, was centrifuged, and the pellet was resuspended in 100uL B-PERII reagent (Pierce). The mixtures were vortexed at room temperature for 10min, and centrifuged at 16,000g for 10min. The supernatant containing soluble cell lysates, and the pellets (resuspended in 100uL 2% w/v SDS) containing insoluble cell lysates were analyzed by SDS-PAGE. The SDS-PAGE gel was visualized by staining with instant blue (Gentaur), and image of the gel was processed and quantified by the software Quantity One (Bio-Rad). Protein solubility was defined as the quantity of the target protein in soluble cell lysates over that in total cell lysates (soluble cell lysates + insoluble cell lysates). Because ERG12 protein cannot be separated from an abundant endogenous protein on SDS-PAGE, it was detected by western blot analysis with anti-6xhis tag antibody (Penta-his Ab, Qiagen).
In vitro quantification of dxs activity
The DH10B strain – araBAD promoter system was used to produce DXS at 20°C, whose catalytic activity was characterized in vitro. At 24h after induction, insoluble cell lysates were prepared as described above except that they were well resuspended in 100uL NPI-10 (50mM NaH2PO4, 300mM NaCl, 10mM imidazole, pH=8) instead of 2% w/v SDS. 1uL the mixture was then incubated in 20uL in vitro reaction solution containing 40mM Tris (pH=6), 10mM pyruvate, 20mM DL-glyceraldehyde 3-phosphate, 1mM thiamine diphosphate, 12.5 mM MgCl2 and 5 mM β-mercaptoethanol. The reaction was terminated by 1mL acidic extraction solution (acetonitrile/methanol/water 40:40:20, 100mM formic acid) after 2h incubation at 37°C, and formation of DXP was quantified by SPE UPLC-MS.
SPE UPLC-MS quantification of DXP and MVAP
Concentration of DXP and MVAP in cell culture was quantified by SPE UPLC-MS. At 5h after induction, 50uL cell suspension was sampled and diluted in 1mL acidic extraction solution (acetonitrile/methanol/water 40:40:20, 100mM formic acid) and centrifuged at 16,000g for 1min. Supernatant was loaded to a cartridge holding 11 mg LC-NH2 resin (Sigma) that was activated by 200uL acidic extraction solution. The cartridge was centrifuged at 2,800g for 1min, and eluted with 100uL 1% w/v NH4OH that was subsequently neutralized by 0.75uL acetic acid. The eluate was analyzed by UPLC (Waters ACQUITY UPLC) – MS (Bruker micrOTOF II) as described below. Aqueous solution (A) containing 15 mM acetic acid and 10mM tributylamine and methanol (B) were used as mobile phase with a UPLC C18 column (Waters CSH C18 1.7μm 2.1x 50mm). The elution was done at 0.15 mL/min with gradient (start: 100% A, 1.8min: 100% A, 3.1min: 60% A, 4.9min: 60% A, 5.4min: 10% A, 9.5min: 10% A, 10min: 100% A). Electrospray ionization was used and (TOF)MS was operated to scan 50–800 m/z in negative mode with -500V end plate voltage and 4500V capillary voltage. Nebulizer gas was provided in 1bar, drying gas temperature was 9mL/min, and dry gas temperature was 200°C. Sample injection volume was 5μL. A range of m/z was extracted from the acquired data for DXP (213.0170±0.03, eluted at 5.6min) or MVAP (227.0315±0.03, eluted at 6.7min). The integrated area of signal peak at its retention time then was calculated for the metabolites with the software provided by the manufacturer. Based on the integrated area of signal, concentration of DXP and MVAP were determined by interpolating from a standard dilution of the intermediates prepared in biological matrix.
Financial support from Singapore-MIT Alliance (Flagship Research Project) is acknowledged.
- Ro DK, Paradise EM, Ouellet M, Fisher KJ, Newman KL, Ndungu JM, Ho KA, Eachus RA, Ham TS, Kirby J, et al: Production of the antimalarial drug precursor artemisinic acid in engineered yeast. Nature. 2006, 440 (7086): 940-943. 10.1038/nature04640.View ArticleGoogle Scholar
- Ajikumar PK, Xiao WH, Tyo KE, Wang Y, Simeon F, Leonard E, Mucha O, Phon TH, Pfeifer B, Stephanopoulos G: Isoprenoid pathway optimization for Taxol precursor overproduction in Escherichia coli. Science (New York, NY. 2010, 330 (6000): 70-74. 10.1126/science.1191652.View ArticleGoogle Scholar
- Hunter WN: The non-mevalonate pathway of isoprenoid precursor biosynthesis. J Biol Chem. 2007, 282 (30): 21573-21577. 10.1074/jbc.R700005200.View ArticleGoogle Scholar
- Ajikumar PK, Tyo K, Carlsen S, Mucha O, Phon TH, Stephanopoulos G: Terpenoids: opportunities for biosynthesis of natural product drugs using engineered microorganisms. Mol Pharm. 2008, 5 (2): 167-190. 10.1021/mp700151b.View ArticleGoogle Scholar
- Westfall PJ, Pitera DJ, Lenihan JR, Eng D, Woolard FX, Regentin R, Horning T, Tsuruta H, Melis DJ, Owens A, et al: Production of amorphadiene in yeast, and its conversion to dihydroartemisinic acid, precursor to the antimalarial agent artemisinin. Proc Natl Acad Sci USA. 2012, 109 (3): E111-E118. 10.1073/pnas.1110740109.View ArticleGoogle Scholar
- Kim SW, Keasling JD: Metabolic engineering of the nonmevalonate isopentenyl diphosphate synthesis pathway in Escherichia coli enhances lycopene production. Biotechnol Bioeng. 2001, 72 (4): 408-415. 10.1002/1097-0290(20000220)72:4<408::AID-BIT1003>3.0.CO;2-H.View ArticleGoogle Scholar
- Alper H, Jin YS, Moxley JF, Stephanopoulos G: Identifying gene targets for the metabolic engineering of lycopene biosynthesis in Escherichia coli. Metab Eng. 2005, 7 (3): 155-164. 10.1016/j.ymben.2004.12.003.View ArticleGoogle Scholar
- Kang MJ, Lee YM, Yoon SH, Kim JH, Ock SW, Jung KH, Shin YC, Keasling JD, Kim SW: Identification of genes affecting lycopene accumulation in Escherichia coli using a shot-gun method. Biotechnol Bioeng. 2005, 91 (5): 636-642. 10.1002/bit.20539.View ArticleGoogle Scholar
- Yuan LZ, Rouviere PE, Larossa RA, Suh W: Chromosomal promoter replacement of the isoprenoid pathway for enhancing carotenoid production in E. coli. Metab Eng. 2006, 8 (1): 79-90. 10.1016/j.ymben.2005.08.005.View ArticleGoogle Scholar
- Jin YS, Stephanopoulos G: Multi-dimensional gene target search for improving lycopene biosynthesis in Escherichia coli. Metab Eng. 2007, 9 (4): 337-347. 10.1016/j.ymben.2007.03.003.View ArticleGoogle Scholar
- Wang C, Oh MK, Liao JC: Directed evolution of metabolically engineered Escherichia coli for carotenoid production. Biotechnol Prog. 2000, 16 (6): 922-926. 10.1021/bp000124f.View ArticleGoogle Scholar
- Lee PC, Mijts BN, Schmidt-Dannert C: Investigation of factors influencing production of the monocyclic carotenoid torulene in metabolically engineered Escherichia coli. Appl Microbiol Biotechnol. 2004, 65 (5): 538-546.Google Scholar
- Martin VJ, Pitera DJ, Withers ST, Newman JD, Keasling JD: Engineering a mevalonate pathway in Escherichia coli for production of terpenoids. Nat Biotechnol. 2003, 21 (7): 796-802. 10.1038/nbt833.View ArticleGoogle Scholar
- Chang MC, Eachus RA, Trieu W, Ro DK, Keasling JD: Engineering Escherichia coli for production of functionalized terpenoids using plant P450s. Nat Chem Biol. 2007, 3 (5): 274-277. 10.1038/nchembio875.View ArticleGoogle Scholar
- Harker M, Bramley PM: Expression of prokaryotic 1-deoxy-D-xylulose-5-phosphatases in Escherichia coli increases carotenoid and ubiquinone biosynthesis. FEBS Lett. 1999, 448 (1): 115-119. 10.1016/S0014-5793(99)00360-9.View ArticleGoogle Scholar
- Lee KK, Jang CS, Yoon JY, Kim SY, Kim TH, Ryu KH, Kim W: Abnormal cell division caused by inclusion bodies in E. coli; increased resistance against external stress. Microbiol Res. 2008, 163 (4): 394-402. 10.1016/j.micres.2008.03.004.View ArticleGoogle Scholar
- Wilkinson DL, Harrison RG: Predicting the solubility of recombinant proteins in Escherichia coli. Biotechnology (N Y). 1991, 9 (5): 443-448. 10.1038/nbt0591-443.View ArticleGoogle Scholar
- Davis GD, Elisee C, Newham DM, Harrison RG: New fusion protein systems designed to give soluble expression in Escherichia coli. Biotechnol Bioeng. 1999, 65 (4): 382-388. 10.1002/(SICI)1097-0290(19991120)65:4<382::AID-BIT2>3.0.CO;2-I.View ArticleGoogle Scholar
- Agostini F, Vendruscolo M, Tartaglia GG: Sequence-based prediction of protein solubility. J Mol Biol. 2012, 421 (2-3): 237-241. 10.1016/j.jmb.2011.12.005.View ArticleGoogle Scholar
- Tartaglia GG, Pechmann S, Dobson CM, Vendruscolo M: A relationship between mRNA expression levels and protein solubility in E. coli. J Mol Biol. 2009, 388 (2): 381-389. 10.1016/j.jmb.2009.03.002.View ArticleGoogle Scholar
- Hirose S, Kawamura Y, Yokota K, Kuroita T, Natsume T, Komiya K, Tsutsumi T, Suwa Y, Isogai T, Goshima N, et al: Statistical analysis of features associated with protein expression/solubility in an in vivo Escherichia coli expression system and a wheat germ cell-free expression system. J Biochem. 2011, 150 (1): 73-81. 10.1093/jb/mvr042.View ArticleGoogle Scholar
- Niwa T, Ying BW, Saito K, Jin W, Takada S, Ueda T, Taguchi H: Bimodal protein solubility distribution revealed by an aggregation analysis of the entire ensemble of Escherichia coli proteins. Proc Natl Acad Sci USA. 2009, 106 (11): 4201-4206. 10.1073/pnas.0811922106.View ArticleGoogle Scholar
- Rodriguez-Villalon A, Perez-Gil J, Rodriguez-Concepcion M: Carotenoid accumulation in bacteria with enhanced supply of isoprenoid precursors by upregulation of exogenous or endogenous pathways. J Biotechnol. 2008, 135 (1): 78-84. 10.1016/j.jbiotec.2008.02.023.View ArticleGoogle Scholar
- Gonzalez-Montalban N, Garcia-Fruitos E, Villaverde A: Recombinant protein solubility - does more mean better?. Nat Biotechnol. 2007, 25 (7): 718-720. 10.1038/nbt0707-718.View ArticleGoogle Scholar
- Baneyx F, Mujacic M: Recombinant protein folding and misfolding in Escherichia coli. Nat Biotechnol. 2004, 22 (11): 1399-1408. 10.1038/nbt1029.View ArticleGoogle Scholar
- Panavas T, Sanders C, Butt TR: SUMO fusion technology for enhanced protein production in prokaryotic and eukaryotic expression systems. Methods Mol Biol. 2009, 497: 303-317. 10.1007/978-1-59745-566-4_20.View ArticleGoogle Scholar
- de Marco A: Protocol for preparing proteins with improved solubility by co-expressing with molecular chaperones in Escherichia coli. Nat Protoc. 2007, 2 (10): 2632-2639. 10.1038/nprot.2007.400.View ArticleGoogle Scholar
- Xie X, Pashkov I, Gao X, Guerrero JL, Yeates TO, Tang Y: Rational improvement of simvastatin synthase solubility in Escherichia coli leads to higher whole-cell biocatalytic activity. Biotechnol Bioeng. 2009, 102 (1): 20-28. 10.1002/bit.22028.View ArticleGoogle Scholar
- Prasad S, Khadatare PB, Roy I: Effect of chemical chaperones in improving the solubility of recombinant proteins in escherichia coli. Appl Environ Microbiol. 2011, 77 (13): 4603-4609. 10.1128/AEM.05259-11.View ArticleGoogle Scholar
- Zhou S, Grabar TB, Shanmugam KT, Ingram LO: Betaine tripled the volumetric productivity of D(−)-lactate by Escherichia coli strain SZ132 in mineral salts medium. Biotechnol Lett. 2006, 28 (9): 671-676. 10.1007/s10529-006-0033-4.View ArticleGoogle Scholar
- Kim CW, Han KS, Ryu KS, Kim BH, Kim KH, Choi SI, Seong BL: N-terminal domains of native multidomain proteins have the potential to assist de novo folding of their downstream domains in vivo by acting as solubility enhancers. Protein Sci. 2007, 16 (4): 635-643. 10.1110/ps.062330907.View ArticleGoogle Scholar
- Han KY, Song JA, Ahn KY, Park JS, Seo HS, Lee J: Solubilization of aggregation-prone heterologous proteins by covalent fusion of stress-responsive Escherichia coli protein, SlyD. Protein Eng Des Sel. 2007, 20 (11): 543-549. 10.1093/protein/gzm055.View ArticleGoogle Scholar
- Kapust RB, Waugh DS: Escherichia coli maltose-binding protein is uncommonly effective at promoting the solubility of polypeptides to which it is fused. Protein Sci. 1999, 8 (8): 1668-1674. 10.1110/ps.8.8.1668.View ArticleGoogle Scholar
- Kyratsous CA, Silverstein SJ, Delong CR, Panagiotidis CA: Chaperone-fusion expression plasmid vectors for improved solubility of recombinant proteins in Escherichia coli. Gene. 2009, 440 (1-2): 9-15. 10.1016/j.gene.2009.03.011.View ArticleGoogle Scholar
- Tsao CY, Wang L, Hashimoto Y, Yi H, March JC, Delisa MP, Wood TK, Valdes JJ, Bentley WE: LuxS coexpression enhances yields of recombinant proteins in escherichia coli in part through posttranscriptional control of GroEL. Appl Environ Microbiol. 2011, 77 (6): 2141-2152. 10.1128/AEM.02347-10.View ArticleGoogle Scholar
- Xiang S, Usunow G, Lange G, Busch M, Tong L: Crystal structure of 1-deoxy-D-xylulose 5-phosphate synthase, a crucial enzyme for isoprenoids biosynthesis. J Biol Chem. 2007, 282 (4): 2676-2682.View ArticleGoogle Scholar
- Sarkar G, Sommer SS: The "megaprimer" method of site-directed mutagenesis. Biotechniques. 1990, 8 (4): 404-407.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.