- Open Access
Small, synthetic, GC-rich mRNA stem-loop modules 5′ proximal to the AUG start-codon predictably tune gene expression in yeast
© Lamping et al.; licensee BioMed Central Ltd. 2013
- Received: 5 February 2013
- Accepted: 10 July 2013
- Published: 29 July 2013
A large range of genetic tools has been developed for the optimal design and regulation of complex metabolic pathways in bacteria. However, fewer tools exist in yeast that can precisely tune the expression of individual enzymes in novel metabolic pathways suitable for industrial-scale production of non-natural compounds. Tuning expression levels is critical for reducing the metabolic burden of over-expressed proteins, the accumulation of toxic intermediates, and for redirecting metabolic flux from native pathways involving essential enzymes without negatively affecting the viability of the host. We have developed a yeast membrane protein hyper-expression system with critical advantages over conventional, plasmid-based, expression systems. However, expression levels are sometimes so high that they adversely affect protein targeting/folding or the growth and/or phenotype of the host. Here we describe the use of small synthetic mRNA control modules that allowed us to predictably tune protein expression levels to any desired level. Down-regulation of expression was achieved by engineering small GC-rich mRNA stem-loops into the 5′ UTR that inhibited translation initiation of the yeast ribosomal 43S preinitiation complex (PIC).
Exploiting the fact that the yeast 43S PIC has great difficulty scanning through GC-rich mRNA stem-loops, we created yeast strains containing 17 different RNA stem-loop modules in the 5′ UTR that expressed varying amounts of the fungal multidrug efflux pump reporter Cdr1p from Candida albicans. Increasing the length of mRNA stem-loops (that contained only GC-pairs) near the AUG start-codon led to a surprisingly large decrease in Cdr1p expression; ~2.7-fold for every additional GC-pair added to the stem, while the mRNA levels remained largely unaffected. An mRNA stem-loop of seven GC-pairs (∆G = −15.8 kcal/mol) reduced Cdr1p expression levels by >99%, and even the smallest possible stem-loop of only three GC-pairs (∆G = −4.4 kcal/mol) inhibited Cdr1p expression by ~50%.
We have developed a simple cloning strategy to fine-tune protein expression levels in yeast that has many potential applications in metabolic engineering and the optimization of protein expression in yeast. This study also highlights the importance of considering the use of multiple cloning-sites carefully to preclude unwanted effects on gene expression.
- mRNA stem-loops
- Post-transcriptional regulation of gene expression
- Inhibition of translation initiation
- Yeast 43S preinitiation complex
- Regulation of translation
- Negative interactions of multiple cloning-sites
Recent advances in synthetic biology and bioinformatics together with exponentially growing biological databases and the -omics revolution, especially transcriptomics and metabolomics, have increased the importance of yeast to industrial biotechnology [1, 2]. Saccharomyces cerevisiae is a key eukaryotic model organism for fundamental molecular biology research, it was the first eukaryotic organism to have its entire genome sequenced , and it is also a common industrial microorganism used extensively in food and beverage production. These factors, together with its genetic tractability and its ability to grow at low pH, have made S. cerevisiae an attractive microorganism to be used as a chemical factory. Gibson et al., 2008, have assembled the entire Mycoplasma genitalium genome in yeast , and Shao et al., 2009, used transformation-associated recombination to assemble entire metabolic pathways in one single step in yeast . The list of non-natural biological compounds successfully produced by S. cerevisiae is diverse and ranges from protein drugs to fine and commodity chemicals , advanced biofuels , the large family of benzylisoquinoline alkaloids  and many other secondary metabolites with a wide range of pharmacological activities  including the successful production of high levels of artemisinin , a highly effective antimalarial.
Despite these significant advances in synthetic biology major challenges in the design of optimal metabolic pathways remain. To obtain maximal yield, pathway flux needs to be optimized, and the accumulation of toxic intermediates and the metabolic burden on the host minimized. Therefore, one of the key challenges of pathway engineering is the regulation of individual pathway enzymes for optimal activity . Control of expression still relies heavily on regulatable, often plasmid-based, expression systems, but their use is largely limited to research and development only. Both regulatable promoters and plasmids require expensive synthetic media for their stable maintenance and controlled function (i.e. addition of inducers or repressors). In addition, many inducers exhibit disadvantageous pleiotropic effects [9, 10] that affect other aspects of the cell’s biology and/or physiology that are often not well characterized and may lead to misinterpretations of the induced effects or negatively affect expression of foreign genes. Thus, an ideal production host requires expression modules stably integrated into the genome with each enzyme expression level individually optimized in a way that does not depend on regulatable promoters and the use of complex synthetic media. Alper et al., 2005, provided an elegant solution by creating constitutive promoter libraries in Escherichia coli and S. cerevisiae that drove a wide (~1000-fold) dynamic range of protein production [11, 12]. However, the lack of well-characterized promoters still provides a significant hurdle for pathway engineering in yeast.
Here we describe how we discovered a way to tune protein production predictably in yeast. This was revealed during the development of a novel system for the constitutive expression of exceptionally high levels of functional heterologous membrane proteins in S. cerevisiae. The expression system consists of plasmid pABC3 and derivative plasmids and the S. cerevisiae hosts AD1-8u- and its close relative AD∆ [13, 14]. Both strains are deleted in seven ABC transporters, which makes them exquisitely sensitive to a wide range of xenobiotics [15, 16], and the transcription factor PDR3. They also contain the gain-of-function mutant transcription factor Pdr1-3p, that drives the hyper-expression of heterologous ORFs from single-copy genes stably integrated at the genomic PDR5 locus [13–15, 17]. This system has several advantages over other, plasmid-based, expression systems: i) significant cost savings for large-scale protein production, as there is no need for expensive synthetic media for the maintenance of plasmids; ii) robust, highly reproducible, phenotypes and homogenous cell populations; and iii) improved homogeneity of the expressed protein. The objective of this study was to develop a strategy to down-regulate these constitutively over-expressed proteins in a way that avoids the potential disadvantages of existing regulatory systems. When we tried to improve the cloning efficiency of large (7~kb) expression modules by replacing all hexamer cutting sites of the multiple cloning-site (MCS) of pSK-PDR5-PPUS  with rare 8 bp cutting sites we noticed that inclusion of the GC-rich Sfi I site reduced protein expression levels ~8-fold. This forced us to use a single AT-rich 8 bp cutting site (Pac I) for efficient cloning of the 5′ end of heterologous ORFs in plasmid pABC3  - thus indicating, as shown by Crook et al., 2011 , that MCSs can be far from benign cloning tools.
Investigation of AUG start-codon scanning has shown that the yeast ribosomal 43S preinitiation complex (PIC) is very sensitive to interfering hairpins in their 5′ UTRs [19–24], which, as will be demonstrated in this article, explains why the GC-rich Sfi I site was so detrimental for protein production. Although it has long been established that mRNA stem-loops in the 5′ UTR near the AUG start codon inhibit protein expression in eukaryotes, and their inhibitory activities appeared largely independent of gene context [19, 21, 22, 24], no attempts have been made to exploit this intrinsic feature of the eukaryotic translation machinery to regulate protein production in yeast. The only systematic study of the effects of the stability, size, sequence and position of a set of different mRNA stem-loop constructs, expressed in COS-7 cells, found that stems of identical stability but with increasing GC content from (52% to 92%) diminished the expression of a GFP reporter by over 18-fold  indicating that it is not only the thermodynamic stability of the stem-loop per se but also, and perhaps more importantly, its GC-content that determines its degree of inhibition of protein expression/translation in mammalian cells.
In this study we created 17 systematically modified mRNA stem-loop constructs in front of the C. albicans multidrug efflux pump reporter Cdr1p ORF which revealed minimal features necessary for effective repression of protein expression in yeast. Stem-loops of mixed A/U- and G/C-pair containing stems inhibited Cdr1p expression less predictably. However, Cdr1p expression controlled by mRNA stem-loops comprising stems containing only GC-pairs was highly predictable and decreased exponentially with the number of GC-pairs in the stem. Even the smallest stem-loop stem of 3 GC-pairs inhibited Cdr1p expression by ~50%. Additional fine-tuning of expression could be achieved by varying the size of the loop. The degree of translation inhibition by individual mRNA stem-loop modules appeared an intrinsic feature that was independent of: i) sequence context; ii) the host yeast strain; iii) the growth medium; iv) the pH and carbon source of the growth medium; v) steady-state mRNA levels; and vi) they also appeared independent of the growth stage of cells. The stable and predictable tuning of protein expression, with a large dynamic range, from a single promoter by well-defined, small, GC-rich mRNA stem-loops near the AUG start codon provides a simple and very powerful tool for optimal pathway engineering and synthetic biology in yeast. It also provides important clues for an improved understanding of the molecular mechanism of AUG start codon scanning of the yeast 43S PIC.
An Sfi I cloning site 5′ proximal to the ATG start-codon severely affects gene expression levels in yeast
Effects of PDR5 5′ UTR and transcription terminator on Pdr5p expression in yeast strains AD and AD-sec6-4
PDR5 5′ UTR
We therefore created plasmid pABC3, without the Sfi I site, which then enabled efficient cloning and maximal levels of expression of heterologous ORFs . Clearly, as also demonstrated by Crook et al., 2011 , MCSs are not just benign and convenient cloning tools but they can dramatically affect protein production in yeast, and sites with high GC content 5′ proximal to the ATG start codon should be used with caution. As the ability to reduce gene expression could be of use to us, we investigated the effect of the Sfi I site on Cdr1p expression.
Inhibition of PDR5 expression by the Sfi I cloning site is independent of its position, the host in which it is expressed, its sequence context, and growth conditions
Effects of PDR5 5′ UTR on Cdr1p and Cdr2p expression
PDR5 5′ UTR++
Effects of growth conditions on Pdr5p, Cdr1p, and Cdr2p expression
The Sfi I-site in the 5′ UTR near the ATG start codon inhibits translation and increases steady-state mRNA levels
These results clearly demonstrated that the Sfi I-site 5′ proximal to the AUG start codon inhibited translation of Pdr5p in the presence of a ~3-fold increase in PDR5 mRNA levels that was independent of: i) steady-state mRNA levels; and ii) the growth stage of cells.
The Sfi I mRNA stem-loop provides a strong physical barrier for the yeast translation initiation machinery
To analyze the inhibitory effect of the Sfi I mRNA stem-loop on the expression of Cdr1p in more detail, and to ascertain whether we could predictably tune expression by modifying mRNA stem-loop structures positioned at −4 relative to the AUG start codon, we created Cdr1p-expressing yeast strains with 17 different, systematically modified, GC-rich mRNA stem-loops near the AUG start codon. A detailed description of the strategy employed to create these different mRNA species is given in the Materials and Methods section, and a schematic illustration can be found in Additional file 2: Figure S2.
Effects of modifying the core Sfi I stem-loop sequence (GGCCGCTCGGGCC; modifying the size of the stem and the loop) at position −4 to the ATG start codon on the expression of Cdr1p
TCCGCTCGAGGCC AA GCTCG A GGCCT AAAATG
TCCGCTCGTTCGAAAGGCC AA GGCCT AAAATG
TCCGCTCGTTCAGGCC GCTCG GGCCT AAAATG
TCCGCTCGTTCGATT CCGCTCGGG CCAAAATG
TCCGCTCGTTCGAGGCCGCTCGGGC G AAAATG
TCCGCTCGTTCC GGCCGCTCGGGCC G AAAATG
TCCGCTCGAAAC GGCCGCTCGGGCC G AAAATG
TCCGCTAAAGC GGCCGCTCGGGCC GC AAAATG
TCCGAAACGC GGCCGCTCGGGCC GCG AAAATG
TCCGCTCGTTCA GGCCGCTCGGGCC T AAAATG
TCCGCTCGAAAT GGCCGCTCGGGCC A AAAATG
TCCGCTAAATA GGCCGCTCGGGCC TA AAAATG
TCCGAAATTA GGCCGCTCGGGCC TAA AAAATG
TCCGCTAAAGA GGCCGCTCGGGCC TC AAAATG
TCCGAAATGA GGCCGCTCGGGCC TCA AAAATG
TCCGAAACGA GGCCGCTCGGGCC TCG AAAATG
These results were consistent with an Sfi I site at −4 that forms a GC-rich mRNA stem-loop which provides a strong physical barrier for the yeast ribosomal AUG start codon scanning machinery.
FLC resistance levels are an accurate measure for the amounts of Cdr1p expressed
Cdr1p expression levels reduce exponentially with the number of GC-pairs in mRNA stem-loop stems containing only GC-pairs
Surprisingly, the MICFLC values decreased exponentially with the number of GC-pairs in mRNA stem-loop constructs containing only GC-pairs (Figure 4A). Even the two smallest 2 and 3 GC-pairs stem constructs 5 and 4 fit well (R2 = 0.99) onto the black exponential trend lines (Figure 4A). This was even more remarkable given the 50% margin of error that was intrinsic to the way MICFLC values were determined (see ‘error’ bars in Figure 4).
There was also a clear exponential relationship between the thermodynamic stabilities of these constructs and their MICFLC values (R2 = 0.98; dashed grey trend lines in Figure 4B). However, in this case the variance decreased when the data point for the smallest, 2 GC-pairs, stem-loop construct 5 was excluded (R2 = 0.999; black trend lines in Figure 4B).
The exponential dependency of Cdr1p expression levels (MICFLC) for cells with mRNA stem-loops containing only GC-pairs with stems ≥3 GC-pairs on either: i) the number of GC-pairs (MICFLC(GC)); or ii) the thermodynamic stabilities (MICFLC(∆G)) of these constructs could be expressed as the formulae shown at the bottom of Figure 4 (x is the number of GC-pairs). The dashed green trend line in Figure 4A represents the calculated trend line for these constructs (assuming a MICFLCmax = 200 mg/l). It matched the experimentally determined results (black trend line) exceptionally well.
We conclude that even an mRNA stem-loop of 3 GC-pairs is biologically active and able to provide a relatively strong physical barrier for the yeast 43S PIC.
Sfi I mRNA stem-loop stems of mixed AU/GC-pair stems
Eliminating the single AU-pair of the original Sfi I stem-loop construct 1 to form construct 9 caused an unexpected ~2-fold reduction of Cdr1p expression (Table 4). Adding 1, 2, or 3 additional AU-pairs to construct 9 also led to unpredictable results: one extra AU-pair (construct 10) caused a ~4-fold reduction in expression rather than the 2-fold increase observed for construct 1 (construct 10 had the same number of AU/GC-pairs as construct 1 but the additional AU pair was arranged in an inverted fashion and the three nucleotides 5′ proximal to the stem-loop of construct 1 (TTC) were replaced with AAA in construct 10; Table 4), two extra AU-pairs (construct 13) had no apparent effect, while three additional AU-pairs (construct 18) gave only a ~2-fold reduction of Cdr1p expression compared with construct 9 (Table 4). Clearly, the presence of additional AU-pairs in GC-rich stem-loops (constructs 1, 10, 12, 13, 16, 17 and 18; Table 4) led to less predictable Cdr1p expression levels and their effects appeared to be dependent on the surrounding mRNA sequence unlike mRNA stem-loops containing only GC-pairs whose inhibitory effects remained unaffected by the surrounding mRNA sequence (e.g. constructs 2 and 11 had identical MICFLC values although they contained the same nucleotide variations as constructs 1 and 10; Table 4).
A practical application of GC-rich mRNA stem-loops for high-throughput drug screening
We demonstrated how such a screen may be used to distinguish between a strong, target-specific, inhibitor of efflux pump Cdr1p, enniatin , and RC21v2 , a weaker Cdr1p-specific D-octapeptide inhibitor (Figure 5). CSM agar plates contained FLC at concentrations of ¼ the MICFLC of the test strains so that each strain was able to grow and cells accumulated similar amounts of FLC. The assumption that the test strains accumulate similar amounts of intracellular FLC is based on the fact that the FLC drug target Erg11p is located inside the cell in the endoplasmic reticulum and that, while the test strains differ in the amounts of Cdr1p expressed, they express the same, or very similar, amounts of Erg11p and therefore require the same, or very similar, intracellular concentrations of FLC to inhibit its essential function. Conducting the experiment in this way ensured that Cdr1p inhibition was directly dependent on the amount of inhibitor used. Two-fold increasing amounts of enniatin or RC21v2 were put on filter disks and the disks were placed onto plates seeded with a lawn of either of the two strains (Figure 5). After incubating the plates at 30°C for two days growth inhibitory zones appeared, the sizes of which were used as an indication of the level of Cdr1p inhibition. wt-CDR1 expressing cells required ~8-16 times more enniatin than CDR1-construct 10 expressing cells whereas both Cdr1p-expressing strains required similar amounts of the weaker inhibitor RC21v2 to inhibit cell growth to the same degree (Figure 5).
Our host strain AD∆ is deleted in seven ABC transporters  and therefore exquisitely sensitive to many xenobiotics [14, 15]. The overexpression of Cdr1p led to a ~400-fold increase in FLC resistance. We exploited this large dynamic range of FLC susceptibilities as a very sensitive and robust tool to analyze the effects of varying mRNA stem-loops 5′ proximal of the AUG start-codon on the efficiency of Cdr1p translation in yeast.
A number of studies have shown that small, GC-rich, mRNA stem-loops placed into 5′ UTRs of yeast genes have strong inhibitory effects on their expression levels [19–24]. This effect was exhibited at the level of translation (mRNA levels were mostly unaffected and varied no more than 2–4 fold [19, 22, 24]) and was largely independent of gene context and the promoter used. The inhibitory effects of individual mRNA stem-loops were comparable to some of our Cdr1p stem-loop constructs (Table 4): e.g. i) a −10.5 kcal/mol mRNA stem-loop (GAATTCCC ATCTTGGGAATTC; stem nucleotides are in italics) positioned 21 nt upstream of the AUG start-codon of the GCN4-lacZ reporter plasmid reduced the β-galactosidase activity to 13% ; and ii) a −8.5 kcal/mol mRNA stem-loop (TGAATTCG TTAACGAATTCA) right next to the AUG start codon of the CYC1 gene (this construct was integrated into the CYC1 locus) reduced iso-1-cytochrome c expression to 10% . Most other mRNA stem-loops tested were of higher stabilities (<−20 kcal/mol) and inhibited reporter gene expression (e.g. endogenous CYC1 and HIS4 genes or the plasmid-encoded chloramphenicol acetyl transferase (cat) reporter) to <1% or completely undetectable levels causing histidine auxotrophy for some HIS4 constructs [19, 21, 24].
We successfully exploited the intrinsic nature of small GC-rich mRNA stem-loop modules 5′ proximal to the AUG start-codon of yeast genes to stably and predictably tune gene dosage from a single promoter without the need for inducers. This discovery, and the general lack of well-characterized promoters for gene expression in yeast, makes GC-rich mRNA stem-loop modules an important tool for regulating protein expression in yeast. They could be of value for i) the titration of minimal expression levels required for essential genes; ii) the elucidation of gene function; or iii) the determination of the precise impact of the gene dosage on a desired phenotype . They could help identify the rate-limiting step and optimize the expression levels for genes in novel metabolic pathways by modifying the expression modules for each gene. Also they could be used to down-regulate expression levels of essential genes of competing endogenous biological pathways, which can lead to dramatically reduced levels of a target metabolite. One example would be the successful synthesis of artemisinic acid, precursor of the antimalarial artemisinin, in yeast that required down-regulation of the essential gene ERG9[8, 43]. Another application of the GC-rich mRNA stem-loop modules is the optimization of heterologous membrane protein expression in yeast as, often, high expression levels can lead to their misfolding and/or mislocalization [44, 45]. As AUG start-codon scanning is a universal eukaryotic feature it is possible that this strategy can be applied in many other eukaryotic hosts as well .
Strains and culture conditions
Saccharomyces cerevisiae strains used in this study
MAT a, leu2-3, leu2-112, his 4–519, can1
G. R. Fink, MIT, MA, USA
MAT a, ura3-52, leu2-3, 112, his 4–619, sec6-4, GAL
AD124567u- = AD/wt-PDR5
MAT α, PDR1-3, ura3, his1, ∆yor1::hisG, ∆snq2::hisG, ∆pdr10::hisG, ∆pdr11::hisG, ∆ycf1::hisG, ∆pdr3::hisG
AD1-8u- = AD
MAT α, PDR1-3, ura3, his1, ∆yor1::hisG, ∆snq2::hisG, ∆pdr10::hisG, ∆pdr11::hisG, ∆ycf1::hisG, ∆pdr5::hisG, ∆pdr15::hisG, ∆pdr3::hisG
AD1-8u-, ∆pdr5:: pABC3 (empty vector cassette)
MAT α, PDR1-3, ura3, his1, ∆yor1::hisG, ∆snq2::hisG, ∆pdr10::hisG, ∆pdr11::hisG, ∆ycf1::hisG, ∆pdr5::hisG, ∆pdr15::hisG, ∆pdr3::hisG, sec6-4::200
AD1-8u-, ∆pdr5:: pABC1-SfiI-PacI-PDR5
AD1-8u-, ∆pdr5:: pABC1-SfiI-PDR5
AD1-8u-, ∆pdr5:: pABC3-PDR5
AD1-8u-, sec6-4::200, PDR5
AD1-8u-, sec6-4::200, SfiI-PacI-PDR5
AD1-8u-, sec6-4::200, SfiI(−18)-PDR5
AD1-8u-, sec6-4::200, PacI-PDR5
AD1-8u-, sec6-4::200, ∆pdr5:: pABC1-SfiI-PacI-PDR5
AD1-8u-, sec6-4::200, ∆pdr5:: pABC3-PDR5
AD1-8u-, ∆pdr5:: pABC1-SfiI-CDR2A (A allele of C. albicans 10261)
AD1-8u-, ∆pdr5:: pABC3-CDR2A
AD1-8u-, ∆pdr5:: pABC1-SfiI-CDR2B (B allele of C. albicans 10261)
AD1-8u-, ∆pdr5:: pABC3-CDR2B
AD1-8u-, ∆pdr5:: pABC1-SfiI-CaCDR1A (A allele of C. albicans 10261)
AD∆, ∆pdr5:: pABC3-CaCDR1A
AD∆/SfiI(−4)-CDR1 = AD∆/construct1-CDR1
AD∆, ∆pdr5:: construct1
AD∆/constructs(2 and 4–18)-CDR1
AD∆, ∆pdr5:: constructs(2 and 4–18)
Molecular biology reagents, restriction and modifying enzymes were from New England Biolabs (Beverly, MA) or from Roche Diagnostics N.Z. Ltd. (Auckland, New Zealand). Lyophilized desalted DNA oligonucleotides listed in Additional file 3: Table S1 were purchased from Sigma-Aldrich Pty. Ltd. (Sydney, Australia). PCR and DNA fragments were purified using kits from Qiagen Pty. Ltd. (Clifton Hill, Victoria, Australia). Genomic DNA (gDNA) was isolated from individual yeast colonies by using the Y-DER™ Yeast DNA Extraction Reagent Kit from Pierce (Rockford, IL) and downscaling the recommended protocol 50-fold. Yeast cells were transformed using the alkali-cation yeast transformation kit from Bio 101 with slight modifications for AD1-8u- as described previously . Plasmids and entire transformation cas-settes PCR-amplified from the gDNA of different yeast strains (Table 5) were verified by DNA sequencing using the DYEnamic ET Terminator Cycle Sequencing kit v 3.1 (Amersham Pharmacia Biotech, UK) and analyzed at the Micromon DNA Sequencing Facility (Monash University, Melbourne, Australia). For standard PCR reactions (95°C for 5 min followed by cycles of: 95°C for 20 sec; 55°C for 10 sec; and 68°C for 1 min/kb of PCR fragment) the high fidelity KOD+ DNA polymerase was used (Toyobo, Osaka, Japan or Novagen, San Diego, CA). For site-directed mutagenesis of plasmids the Chameleon® site-directed mutagenesis kit (Stratagene, La Jolla, CA) was employed. ExoSAP treatment was used to eliminate unwanted DNA oligomer primers from PCR reactions. In short, a 5 μl portion of the PCR reaction was incubated at 37°C with 0.2 μl ExoSAP-IT® (Affymetrix, Santa Clare, CA) for 15 min and the enzyme was heat inactivated at 80°C for 30 min. Small aliquots (0.1 – 1 μl) were then used as DNA templates for DNA sequencing or PCR.
Fluconazole (FLC, Diflucan; aqueous solution) was purchased from Pfizer Laboratories Ltd. (Auckland, New Zealand) and enniatin was purchased from Sigma-Aldrich New Zealand Ltd. (Auckland, New Zealand). D-octapeptide RC21v2 is a Cdr1p-specific inhibitor of FLC transport by ABC efflux pump Cdr1p .
Construction of plasmids pABC1 and pABC3
Plasmid pABC1 (Figure 1A) is a pSK-PDR5-PPUS  derivative based on the high copy number plasmid pBluescriptIISK(+) (Stratagene). To ensure the efficient termination of highly expressed genes, the S. cerevisiae PGK1 transcription terminator was PCR amplified as a Hin dIII/Bam HI fragment from AD1-8u- gDNA and used to replace the Hin dIII/Bam HI PDR5 terminator fragment of pSK-PDR5-PPUS immediately 3′ of the PDR5 promoter to generate pSK-PDR5-PGK1. Further improvements (creation of a multiple cloning site with additional unique cloning sites upstream and downstream of the transformation cassette) of pSK-PDR5-PGK1 by site-directed mutagenesis led to the creation of vector pABC (precursor of pABC1).
Plasmid pABC3, the cloning vehicle that we routinely use for the overexpression of membrane proteins in yeast [14, 17], was derived from plasmid pABC1 as previously described . In short, we used site-directed mutagenesis to introduce a unique Eco RI site at the 3′ end of the URA3 marker and replaced Sac I of pABC1 with Xho I creating plasmid pABC2. In a final step vector pABC3 was created by reverting the Sfi I/Ava I sites of pABC2 to the wildtype PDR5 sequence to maximize translation efficiency in yeast and a second Asc I site was created at the 3′ end of the transformation cassette for ease of cassette excision .
DNA sequences of pSK-PDR5-PPUS, pSK-PDR5-PGK1, pABC, pABC1 and pABC2 were submitted to GenBank under accession numbers JN581374-78, respectively.
Creation of PDR5 over-expressing strains that had only their 5′ UTR modified
AD and AD/sec6-4 strains that overexpressed wt-PDR5 or PDR5 with their 5′ UTR modified to contain either the Sfi I-, the Sfi I-Pac I- or the Pac I-site 5′ proximal to the ATG start codon of PDR5 were created by transforming AD and AD/sec6-4 with four different DNA fragments that contained that part of the promoter and ~1/3 (1163)bp) of the ORF of PDR5 that was deleted in both strains . To ensure proper integration of these DNA fragments via homologous recombination into the genomic PDR5 locus of AD and AD/sec6-4 >200 bp additional PDR5 sequence was included on either end. The DNA fragments that were used to create AD/ and AD/sec6-4/PDR5 were PCR-amplified from gDNA of AD/wt-PDR5 using the primer pair pd5f/pd8r. The DNA fragments that were used to create AD/SP-, /S-, and /P-PDR5 and AD/sec6-4/SP-, /S-, and /P-PDR5 were obtained by digesting 2 μg of plasmids pABC1-SP-PDR5, pABC1-S-PDR5, and pABC3-PDR5, respectively, with Asc I and Sal I and gel purifying the resulting ~2.5 kb DNA fragments. Positive transformants were selected on CSM plates containing 20 μg/ml FLC, a concentration that was high enough to prevent growth of AD and AD/sec6-4 but low enough for any of the expected recombinant yeast strains to grow. Three independent transformants were verified for each individual construct for proper integration at the PDR5 locus by PCR from purified gDNA and by DNA sequencing.
Creation of an mRNA stem-loop library near the AUG start-codon of CDR1
1–10 ng of pABC3-CDR1A  were used as DNA template to amplify 18 pairs of PCR fragments to create 17 different yeast strains (AD∆/constructs(1,2,4-18)-CDR1; Table 4 and Additional file 2: Figure S2). Strains with weaker stem-loops (AD∆/constructs(1,2,4-8)-CDR1) were created using strategy 1, as illustrated in Additional file 2: Figure S2A. In a first step, two DNA fragments (the 3′ part of the PDR5 promoter and the 5′ part of the CDR1 ORF) were amplified with primers pd5f/pSfiM-(1–8)r and pSfiM-(1–8)f/Rev-3, respectively. Each pair of DNA fragments had identical stem-loop sequences near their 3′ (PDR5 fragment) and 5′ (5′-CDR1 fragment) ends, respectively (highlighted light blue in Additional file 1: Figure S2 and underlined sequences in Additional file 3: Table S1). Portions of PCR amplicons were treated with ExoSAP-IT® to eliminate DNA oligomer primers before mixing 1 μl (~40 ng) of each pair of DNA fragments (construct pairs 1–8) and amplifying the fused PCR products with the outside primer pair pd5f/Rev-3 in a second PCR step. The eight fused PCR products were column purified and used to transform AD∆ as described below. Using this approach it was impossible to amplify the fused PCR product for AD∆/construct3-CDR1 (core Sfi I stem-loop extended with three extra GC-pairs; see primers pSfiM-3f/r in Additional file 3: Table S1). This strong stem-loop of seven GC-pairs inhibited the fusion of the two overlapping PCR fragments as illustrated in Additional file 2: Figure S2B (top left).
In order to create larger stem-loops with additional AU- and GC-pairs (including AD∆/construct15-CDR1 to replace the planned but not obtainable AD∆/construct3-CDR1) we developed an alternative cloning strategy (Additional file 2: Figure S2B). Primer pairs pSfiM-(9–18)(f/r) were designed so that their 5′ ends contained a core Sfi I site (bold type face; Additional file 3: Table S1) that was extended on either side with one, two, or three extra nucleotides (underlined in Additional file 3: Table S1). The sequences for primers pSfiM-(9–18)f were extended with three additional As followed by ~20 bp of the CDR1A ORF. This design ensured that the positions of the stem-loops of constructs 9–18 were always at −4 relative to the AUG start-codon. The sequences of primers pSfiM-(9–18)r were each extended with an additional three Ts so that each sequence of the Sfi I constructs(9–18) was flanked by three A nucleotides (Additional file 3: Table S1) to minimize secondary structure around the predicted stem-loops. An additional ~25 bp of the wild-type PDR5 promoter sequence was added to the reverse primers to ensure the amplification of PCR products. The PDR5 promoter and the 5′-CDR1 fragments were then amplified by PCR from pABC3-CDR1A with pd5f/pSfiM-(9–18)r and pSfiM-(9–18)f/Rev-3 primer pairs, respectively. A portion (~200 ng) of each PCR fragment was digested with Sfi I, the digested fragments were then gel purified and dissolved in 50 μl water. Corresponding pairs of Sfi I-digested PCR fragments (5 μl; ~20 ng) were mixed, ligated, and ~2 ng of each ligation mix (mixes 9–18) used as a DNA template for PCR amplification of ligated products using primers pd5f/Rev-3 (Additional file 2: Figure S2B). A single PCR fragment was obtained in all cases. This was possible for two reasons: i) the two DNA fragments of each pair of PCR fragments could only ligate at their Sfi I digested ends (grey) leading to only three possible ligation products (Additional file 2: Figure S2B) as their blunt ends were not 5′-phosphorylated (non-phosphorylated primers were used to amplify the fragments); and ii) only fragment three of the ligation mix (PDR5-Sfi I-5′-CDR1; Additional file 1: Figure S2B) could be amplified because the other two ligation products were inverted repeats that spontaneously form an intramolecular double strand after strand separation at 94°C.
Finally, the fused and PCR amplified DNA products obtained by either of these two strategies (Additional file 2: Figure S2A and B) were gel purified, and ~40 ng of each were mixed with 200 ng 3′-CDR1-URA3-PDR5 (PCR amplified from pABC3-CDR1A with primers CaCDR1-3/pAscI-2 and column purified; the remaining 3′ part of the entire CDR1-transformation cassette was identical for all constructs) and used to transform AD∆ (Additional file 2: Figure S2C). The entire CDR1-transformation cassette (~7.5 kb) was divided into two parts (5′ CDR1 and 3′ CDR1) because the smaller (~2.5 kb) 5′ CDR1 constructs required fewer cycles (a combined total of 40–45 cycles for two separate steps of PCR amplification required for either strategy; Additional file 2: Figure S2A and B) of PCR to efficiently amplify, which significantly reduced the rate of amplification errors within positive transformants. Uracil prototroph transformants were selected on CSM-URA plates after incubation at 30°C for 2–3 d. Three independent transformants were verified for each individual construct for proper integration at the PDR5 locus by PCR from purified gDNA and by DNA sequencing.
Northern blot analysis
Total RNA was isolated from S. cerevisiae cells using the hot-phenol extraction method. Usually about 100 ODU (optical density units; defined as the amount of cells corresponding to 1 ml of cells of an OD600 of 1) of cells were harvested by centrifugation for 1 min at 3000 g, the cells were washed once in ice-cold water, and snap frozen in liquid nitrogen and stored at −80°C. Samples (10 μg) of total RNA were separated on 1.2% denaturing agarose gels and stained with ethidium bromide (EtBr). The separated total RNA was photographed, immediately Northern blotted onto nylon+ membranes and further processed according to standard protocols . 32P-radioactively labeled probes for ACT1 and PDR5 were obtained with a random priming kit using PCR-amplified DNA fragments of ACT1 (~800 bp; amplified with pACT1for/pACT1rev) and PDR5 (~1200 bp; amplified with pd7f/pd23r) as DNA templates. Both PCR fragments were amplified from gDNA extracted from AD/wt-PDR5.
Analysis and purification of plasma membrane (PM) proteins
PM fractions of S. cerevisiae cells were prepared as described previously  and protein samples (30 μg) were separated by SDS-PAGE with 8% polyacrylamide gels and stained with Coomassie Blue R250.
Functional analysis of multidrug efflux pump over-expressing yeast strains
The susceptibilities of three independent transformants for each individual construct to the antifungal FLC were measured as described previously .
Screening for inhibitors of C. albicans multidrug efflux pump Cdr1p (chemosensitization assay)
The chemosensitization of yeast strains over-expressing the C. albicans multidrug efflux pump Cdr1p to FLC was carried out as described previously . In brief, a 10 ml YPD overnight culture of cells was diluted 1:20 into CSM medium and incubated at 30°C for a further four hours. Each test strain (OD600nm ~1) was diluted to OD600nm = 0.008 in 5 ml of melted CSM containing 0.6% agarose (50°C) and FLC at 0.25 x the minimum growth inhibitory concentration (MICFLC) of each strain. The cell suspension was poured into a rectangular Omnitray plate (126 by 86 by 19 mm; Nunc, Roskilde, Denmark) that contained 20 ml of CSM solidified with 0.6% agarose and FLC at a concentration of 0.25 x MICFLC of the respective test strain. Whatman 3MM paper disks containing different amounts of the Cdr1p drug pump inhibitor enniatin or RC21v2 were placed on the solidified top agarose and the plates were incubated at 30°C for 48 h.
This work was supported by the Japan Health Sciences Foundation (SH24405 to MN); the Japan Society for the Promotion of Science (S04718, S06741 to EL); and in part by the Foundation for Research Science and Technology of New Zealand (UOOX0607 to RDC).
- Kim IK, Roldao A, Siewers V, Nielsen J: A systems-level approach for metabolic engineering of yeast cell factories. FEMS Yeast Res. 2012, 12 (2): 228-248. 10.1111/j.1567-1364.2011.00779.xView ArticleGoogle Scholar
- Siddiqui MS, Thodey K, Trenchard I, Smolke CD: Advancing secondary metabolite biosynthesis in yeast with synthetic biology tools. FEMS Yeast Res. 2012, 12 (2): 144-170. 10.1111/j.1567-1364.2011.00774.xView ArticleGoogle Scholar
- Goffeau A, Barrell BG, Bussey H, Davis RW, Dujon B, Feldmann H, Galibert F, Hoheisel JD, Jacq C, Johnston M, et al: Life with 6000 genes. Science. 1996, 274 (5287): 546, 563-547.View ArticleGoogle Scholar
- Gibson DG, Benders GA, Andrews-Pfannkoch C, Denisova EA, Baden-Tillson H, Zaveri J, Stockwell TB, Brownley A, Thomas DW, Algire MA, et al: Complete chemical synthesis, assembly, and cloning of a Mycoplasma genitalium genome. Science. 2008, 319 (5867): 1215-1220. 10.1126/science.1151721View ArticleGoogle Scholar
- Shao Z, Zhao H: DNA assembler, an in vivo genetic method for rapid construction of biochemical pathways. Nucleic Acids Res. 2009, 37 (2): e16-View ArticleGoogle Scholar
- Peralta-Yahya PP, Keasling JD: Advanced biofuel production in microbes. Biotechnol J. 2010, 5 (2): 147-162. 10.1002/biot.200900220View ArticleGoogle Scholar
- Hawkins KM, Smolke CD: Production of benzylisoquinoline alkaloids in Saccharomyces cerevisiae. Nat Chem Biol. 2008, 4 (9): 564-573. 10.1038/nchembio.105View ArticleGoogle Scholar
- Paddon CJ, Westfall PJ, Pitera DJ, Benjamin K, Fisher K, McPhee D, Leavell MD, Tai A, Main A, Eng D, et al: High-level semi-synthetic production of the potent antimalarial artemisinin. Nature. 2013, 496 (7446): 528-532. 10.1038/nature12051View ArticleGoogle Scholar
- Labbe S, Thiele DJ: Copper ion inducible and repressible promoter systems in yeast. Methods Enzymol. 1999, 306: 145-153.View ArticleGoogle Scholar
- Mumberg D, Muller R, Funk M: Regulatable promoters of Saccharomyces cerevisiae: comparison of transcriptional activity and their use for heterologous expression. Nucleic Acids Res. 1994, 22 (25): 5767-5768. 10.1093/nar/22.25.5767View ArticleGoogle Scholar
- Alper H, Fischer C, Nevoigt E, Stephanopoulos G: Tuning genetic control through promoter engineering. Proc Natl Acad Sci U S A. 2005, 102 (36): 12678-12683. 10.1073/pnas.0504604102View ArticleGoogle Scholar
- Fischer CR, Alper H, Nevoigt E, Jensen KL, Stephanopoulos G, , et al: Tuning genetic control--importance of thorough promoter characterization versus generating promoter diversity. Trends Biotechnol. 2006, 24 (2): 55-56. 10.1016/j.tibtech.2005.12.001View ArticleGoogle Scholar
- Decottignies A, Grant AM, Nichols JW, de Wet H, McIntosh DB, Goffeau A: ATPase and multidrug transport activities of the overexpressed yeast ABC protein Yor1p. J Biol Chem. 1998, 273 (20): 12612-12622. 10.1074/jbc.273.20.12612View ArticleGoogle Scholar
- Lamping E, Monk BC, Niimi K, Holmes AR, Tsao S, Tanabe K, Niimi M, Uehara Y, Cannon RD: Characterization of three classes of membrane proteins involved in fungal azole resistance by functional hyperexpression in Saccharomyces cerevisiae. Eukaryot Cell. 2007, 6 (7): 1150-1165. 10.1128/EC.00091-07View ArticleGoogle Scholar
- Nakamura K, Niimi M, Niimi K, Holmes AR, Yates JE, Decottignies A, Monk BC, Goffeau A, Cannon RD: Functional expression of Candida albicans drug efflux pump Cdr1p in a Saccharomyces cerevisiae strain deficient in membrane transporters. Antimicrob Agents Chemother. 2001, 45 (12): 3366-3374. 10.1128/AAC.45.12.3366-3374.2001View ArticleGoogle Scholar
- Tanabe K, Lamping E, Nagi M, Okawada A, Holmes AR, Miyazaki Y, Cannon RD, Monk BC, Niimi M: Chimeras of Candida albicans Cdr1p and Cdr2p reveal features of pleiotropic drug resistance transporter structure and function. Mol Microbiol. 2011, 82 (2): 416-433. 10.1111/j.1365-2958.2011.07820.xView ArticleGoogle Scholar
- Monk BC, Cannon RD, Nakamura K, Niimi M, Niimi K, Harding DRK, Holmes AR, Lamping E, Goffeau A, Decottignies A: Membrane protein expression system and its application in drug screening. 2002, New Zealand and International patent PCT/NZ02/00163Google Scholar
- Crook NC, Freeman ES, Alper HS: Re-engineering multicloning sites for function and convenience. Nucleic Acids Res. 2011, 39 (14): e92- 10.1093/nar/gkr346View ArticleGoogle Scholar
- Baim SB, Sherman F: mRNA structures influencing translation in the yeast Saccharomyces cerevisiae. Mol Cell Biol. 1988, 8 (4): 1591-1601.View ArticleGoogle Scholar
- Chiu WL, Wagner S, Herrmannova A, Burela L, Zhang F, Saini AK, Valasek L, Hinnebusch AG: The C-terminal region of eukaryotic translation initiation factor 3a (eIF3a) promotes mRNA recruitment, scanning, and, together with eIF3j and the eIF3b RNA recognition motif, selection of AUG start codons. Mol Cell Biol. 2010, 30 (18): 4415-4434. 10.1128/MCB.00280-10View ArticleGoogle Scholar
- Cigan AM, Pabich EK, Donahue TF: Mutational analysis of the HIS4 translational initiator region in Saccharomyces cerevisiae. Mol Cell Biol. 1988, 8 (7): 2964-2975.View ArticleGoogle Scholar
- Oliveira CC, van den Heuvel JJ, McCarthy JE: Inhibition of translational initiation in Saccharomyces cerevisiae by secondary structure: the roles of the stability and position of stem-loops in the mRNA leader. Mol Microbiol. 1993, 9 (3): 521-532. 10.1111/j.1365-2958.1993.tb01713.xView ArticleGoogle Scholar
- Sagliocco FA, Vega Laso MR, Zhu D, Tuite MF, McCarthy JE, Brown AJ: The influence of 5′-secondary structures upon ribosome binding to mRNA during translation in yeast. J Biol Chem. 1993, 268 (35): 26522-26530.Google Scholar
- Vega Laso MR, Zhu D, Sagliocco F, Brown AJ, Tuite MF, McCarthy JE: Inhibition of translational initiation in the yeast Saccharomyces cerevisiae as a function of the stability and position of hairpin structures in the mRNA leader. J Biol Chem. 1993, 268 (9): 6453-6462.Google Scholar
- Babendure JR, Babendure JL, Ding JH, Tsien RY: Control of mammalian translation by mRNA structure near caps. RNA. 2006, 12 (5): 851-861. 10.1261/rna.2309906View ArticleGoogle Scholar
- Lamping E, Tanabe K, Niimi M, Uehara Y, Monk BC, Cannon RD: Characterization of the Saccharomyces cerevisiae sec6-4 mutation and tools to create S. cerevisiae strains containing the sec6-4 allele. Gene. 2005, 361: 57-66.View ArticleGoogle Scholar
- Holmes AR, Lin YH, Niimi K, Lamping E, Keniya M, Niimi M, Tanabe K, Monk BC, Cannon RD: ABC transporter Cdr1p contributes more than Cdr2p does to fluconazole efflux in fluconazole-resistant Candida albicans clinical isolates. Antimicrob Agents Chemother. 2008, 52 (11): 3851-3862. 10.1128/AAC.00463-08View ArticleGoogle Scholar
- Tsao S, Rahkhoodaee F, Raymond M: Relative contributions of the Candida albicans ABC transporters Cdr1p and Cdr2p to clinical azole resistance. Antimicrob Agents Chemother. 2009, 53 (4): 1344-1352. 10.1128/AAC.00926-08View ArticleGoogle Scholar
- Holmes AR, Tsao S, Ong SW, Lamping E, Niimi K, Monk BC, Niimi M, Kaneko A, Holland BR, Schmid J, et al: Heterozygosity and functional allelic variation in the Candida albicans efflux pump genes CDR1 and CDR2. Mol Microbiol. 2006, 62 (1): 170-186. 10.1111/j.1365-2958.2006.05357.xView ArticleGoogle Scholar
- Abramoff M, Magalhaes PJ, Ram SJ: Image processing with ImageJ. Biophotonics International. 2004, 11 (7): 36-42.Google Scholar
- Zuker M: Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res. 2003, 31 (13): 3406-3415. 10.1093/nar/gkg595View ArticleGoogle Scholar
- Ivnitski-Steele I, Holmes AR, Lamping E, Monk BC, Cannon RD, Sklar LA: Identification of Nile red as a fluorescent substrate of the Candida albicans ATP-binding cassette transporters Cdr1p and Cdr2p and the major facilitator superfamily transporter Mdr1p. Anal Biochem. 2009, 394 (1): 87-91. 10.1016/j.ab.2009.07.001View ArticleGoogle Scholar
- Niimi K, Harding DR, Holmes AR, Lamping E, Niimi M, Tyndall JD, Cannon RD, Monk BC: Specific interactions between the Candida albicans ABC transporter Cdr1p ectodomain and a D-octapeptide derivative inhibitor. Mol Microbiol. 2012, 85 (4): 747-767. 10.1111/j.1365-2958.2012.08140.xView ArticleGoogle Scholar
- Sagliocco FA, Zhu D, Vega Laso MR, McCarthy JE, Tuite MF, Brown AJ: Rapid mRNA degradation in yeast can proceed independently of translational elongation. J Biol Chem. 1994, 269 (28): 18630-18637.Google Scholar
- Balzi E, Wang M, Leterme S, Van Dyck L, Goffeau A: PDR5, a novel yeast multidrug resistance conferring transporter controlled by the transcription regulator PDR1. J Biol Chem. 1994, 269 (3): 2206-2214.Google Scholar
- Hinnebusch AG: Molecular mechanism of scanning and start codon selection in eukaryotes. Microbiol Mol Biol Rev. 2011, 75 (3): 434-467. 10.1128/MMBR.00008-11View ArticleGoogle Scholar
- Jackson RJ, Hellen CU, Pestova TV: The mechanism of eukaryotic translation initiation and principles of its regulation. Nat Rev Mol Cell Biol. 2010, 11 (2): 113-127. 10.1038/nrm2838View ArticleGoogle Scholar
- Sonenberg N, Hinnebusch AG: Regulation of translation initiation in eukaryotes: mechanisms and biological targets. Cell. 2009, 136 (4): 731-745. 10.1016/j.cell.2009.01.042View ArticleGoogle Scholar
- Berthelot K, Muldoon M, Rajkowitsch L, Hughes J, McCarthy JE: Dynamics and processivity of 40S ribosome scanning on mRNA in yeast. Mol Microbiol. 2004, 51 (4): 987-1001. 10.1046/j.1365-2958.2003.03898.xView ArticleGoogle Scholar
- Abaeva IS, Marintchev A, Pisareva VP, Hellen CU, Pestova TV: Bypassing of stems versus linear base-by-base inspection of mammalian mRNAs during ribosomal scanning. EMBO J. 2011, 30 (1): 115-129. 10.1038/emboj.2010.302View ArticleGoogle Scholar
- Mignone F, Gissi C, Liuni S, Pesole G: Untranslated regions of mRNAs. Genome Biol. 2002, 3 (3): REVIEWS0004Google Scholar
- Ringner M, Krogh M: Folding free energies of 5′-UTRs impact post-transcriptional regulation on a genomic scale in yeast. PLoS Comput Biol. 2005, 1 (7): e72- 10.1371/journal.pcbi.0010072View ArticleGoogle Scholar
- 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/nature04640View ArticleGoogle Scholar
- Junge F, Schneider B, Reckel S, Schwarz D, Dotsch V, Bernhard F: Large-scale production of functional membrane proteins. Cell Mol Life Sci. 2008, 65 (11): 1729-1755. 10.1007/s00018-008-8067-5View ArticleGoogle Scholar
- Lee JK, Stroud RM: Unlocking the eukaryotic membrane protein structural proteome. Curr Opin Struct Biol. 2010, 20 (4): 464-470. 10.1016/j.sbi.2010.05.004View ArticleGoogle Scholar
- Nakamoto RK, Rao R, Slayman CW: Expression of the yeast plasma membrane [H+]ATPase in secretory vesicles. A new strategy for directed mutagenesis. J Biol Chem. 1991, 266 (12): 7940-7949.Google Scholar
- Niimi M, Wada S, Tanabe K, Kaneko A, Takano Y, Umeyama T, Hanaoka N, Uehara Y, Lamping E, Niimi K, et al: Functional analysis of fungal drug efflux transporters by heterologous expression in Saccharomyces cerevisiae. Jpn J Infect Dis. 2005, 58 (1): 1-7.Google Scholar
- Sambrook J, Russell DW: Molecular cloning: a laboratory manual. 2001, NY: Cold Spring Harbor Laboratory Press, Cold Spring Harbo, 3Google Scholar
- Niimi K, Harding DR, Parshot R, King A, Lun DJ, Decottignies A, Niimi M, Lin S, Cannon RD, Goffeau A, et al: Chemosensitization of fluconazole resistance in Saccharomyces cerevisiae and pathogenic fungi by a D-octapeptide derivative. Antimicrob Agents Chemother. 2004, 48 (4): 1256-1271. 10.1128/AAC.48.4.1256-1271.2004View ArticleGoogle 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.