Effect of sequential C‐terminal tryptophans on green fluorescent protein fluorescence

The effect of the addition of sequential C‐terminal tryptophan residues on the fluorescence intensity of GFP was investigated. Tandem repeats of six tryptophan residues markedly decreased fluorescence intensity. This phenomenon is likely to occur because of the inhibition of GFP folding, resulting in insolubility. Exploiting this phenomenon, we constructed a cloning vector that facilitates the identification of recombinant colonies of Escherichia coli by the activation of GFP.

Green fluorescent protein (GFP) fluoresces only when irradiated by UV light. It was isolated and purified from the jellyfish Aequorea by Shimomura et al. in 1962 [1]. Since then, it has proven indispensable in biochemistry studies. Commonly, GFP is utilized as a fusion protein for the visualization of intracellular localization [2,3] and as an indicator of gene activation, in so-called 'reporter assays' [4]. The fluorescence intensity of GFP varies with the type of peptide fused to the C terminus, and we have previously reported on the relationship between GFP fluorescence intensity and such fused polypeptides [5,6]. We found that the fluorescence intensity of GFP fused to polypeptides increases with the product of the average hydrophilicity of the fused segment multiplied by the number of fused amino acids. The ability to arbitrarily turn GFP fluorescence on or off is considered to be very useful. In this study, the effect of a small number of tryptophan (Trp) residues added to the GFP C terminus was investigated. A series of these inhibited GFP folding, resulting in insolubility and loss of fluorescence. Applying this phenomenon, we constructed a cloning plasmid vector whose fluorescence is turned on when foreign DNA is inserted into the cloning site.

Materials and methods pS147PGFP as a basic vector
Trp-appended GFP (pGFP-Trp n , where n indicates the number of added amino acid residues) was constructed by modification of pS147PGFP [7]. The plasmid map of pS147PGFP is shown in Fig. 1A. pS147PGFP was constructed by insertion of the S147PGFP gene (S147PGFP) between the EcoRI and HindIII restriction sites of a pkk223-3 expression vector (GE Healthcare, Little Chalfont, UK). To enhance fluorescence, the 147th residue (where the methionine start codon is defined as the first residue), a serine, was replaced by proline (Fig. 1B). The peaks of excitation and emission were approximately 400 and 520 nm, respectively (Fig. 1C).

Chemicals and instruments
Oligonucleotides for inverse PCR or site-directed mutagenesis were synthesized by Thermo Fisher Scientific (Carlsbad, CA, USA). Inverse PCR was performed using the KOD Plus Mutagenesis Kit (Toyobo, Osaka, Japan). Site-directed mutagenesis was performed using Pfu turbo DNA polymerase (Agilent Technologies, Santa Clara, CA, USA). Plasmids were purified from cells using the PureLink Quick Plasmid Miniprep Kit (Invitrogen, Carlsbad, CA, USA). DpnI was purchased from Invitrogen. PCR was performed using an iCycler (Bio-Rad, Hercules, CA, USA). Sequencing was performed with a 3130 genetic analyzer (Thermo Fisher Scientific). Fluorescence of colonies was evaluated using a 3UV transilluminator (UVP, Upland, CA, USA). The turbidity and fluorescence intensity of Escherichia coli in liquid medium were measured using a SPECTRAFluor Plus microplate reader (Tecan, M€ annedorf, Switzerland). Centrifugation was performed using a Centrifuge 5415R (Eppendorf, Hamburg, Germany) equipped with a F45-24-11 fixed-angle rotor. SDS/PAGE gels were stained using a SeePico CBB stain kit (Benebiosis, Seoul, Korea).

Measurement of fluorescence intensity
Escherichia coli harboring pGFP-Trp n were cultured in 10 mL of LB medium containing ampicillin (final concentration, 50 lgÁmL À1 ) at 37°C overnight (approximately 20 h) and cells were harvested from 1 mL of the medium. After washing with PBS (137 mM NaCl, 2.7 mM KCl, 1.5 mM KH 2 PO 4 , 8.1 mM NaHPO 4 , pH 7.4), the cells were resuspended in 1 mL of PBS. Turbidity of the suspensions was measured at 620 nm and fluorescence intensity was measured at an excitation of 405 nm and emission of 535 nm.

Electrophoresis
The cells harvested from 10 mL LB medium after the culture (see above) were disrupted by sonication in 1 mL of PBS. The sonicate was centrifuged (13 362 g, for 10 min at 4°C) and 10 lL of supernatant or pellet mixed with 10 lL of sample buffer [125 mM Tris/HCl, pH 6.8, 10% (v/v) 2-mercaptoethanol, 4% (w/v) SDS, 10% (w/v) sucrose, 0.01% (w/v) bromophenol blue]. Samples were subjected to SDS/PAGE ether with or without boiling in the presence of mercaptoethanol. The polyacrylamide concentrations of the stacking and separation gels were 6% and 12%, respectively. Electrophoresis was performed at a constant voltage of 150 V with running buffer of 25 mM Tris, 192 mM glycine and 0.1% SDS. The fluorescing band of GFP in nonboiled samples was observed under UV-A (365 nm) irradiation using a transilluminator, and the gel then stained with Coomassie Brilliant Blue (CBB). In the case of boiled samples, gels were immediately stained with CBB after electrophoresis. Densitometry data of the bands were obtained using IMAGEJ software [11].

Effect of C-terminal Trps on GFP fluorescence
Escherichia coli colonies harboring pGFP-Trp 1 , pGFP-Trp 2 or pGFP-Trp 3 and growing on agar fluoresced under UV irradiation. Conversely, attenuation of fluorescence was observed with pGFP-Trp 4 , pGFP-Trp 5 or pGFP-Trp 6 ( Fig. 2A). In addition, the fluorescence intensity of the cells in liquid medium decreased as the number of Trp codons increased (Fig. 2B). Fluorescence was barely observed in pGFP-Trp 6 . To investigate the effect of hydrophobicity on fluorescence, pGFP-Leu n (n = 2, 4, 6) and pGFP-Gly n (n = 2, 4, 6) were constructed. The hydrophobicity scales of Trp, Leu and Gly are À3.4, À1.8 and 0.0, respectively [12]. Fluorescence was observed in E. coli harboring all these plasmids, except for pGFP-Leu 6 (Fig. 2C,D). Thus, fluorescence of GFP was attenuated as the number of hydrophobic acid residues at the C terminus increased, and more than five hydrophobic amino acids extinguished the fluorescence almost entirely. This result is consistent with our previous hypothesis that the fluorescence intensity of GFP-fused polypeptides decreases with the product of the average hydrophilicity of the fused segment multiplied by the number of fused amino acids [5], since we also suggested that this rule could be applied not only to peptides but also to small numbers of amino acid residues.

Factors decreasing fluorescence intensity
We considered the following factors in the loss of fluorescence in E. coli harboring pGFP-Trp 6 : (a) the codons added at 3 0 end of GFP decrease translation efficiency, (b) Trp residues inhibit the folding of GFP, and (c) the residues attenuate fluorescence by acting as a 'quencher' that absorbs excitatory or emitted light. To investigate hypothesis (a), the CTG codons of Leu of pGFP-Leu n (n = 2, 4, 6) were replaced with TTA, since Trp has no synonymous codons, and the fluorescence intensities of E. coli harboring these plasmids were measured. The codon usages of CTG and TTA are 46.9& and 15.2&, respectively [13]. If translation efficiency and the fluorescence intensity were related to each other, then the fluorescence intensity of the E. coli harboring pGFP-Leu(CTG) n would be expected to be higher than that of pGFP-Leu(TTA) n . However, there was no significant difference between them (Fig. 2C,D). Therefore, it appeared that the codons added at the 3 0 end of GFP did not markedly affect its expression. To investigate hypothesis (b), sonicates of E. coli harboring pGFP-Trp n were analysed by SDS/ PAGE. GFP is stable in 1% SDS, and its migration can be traced in real time using UV illumination [14]. In native SDS/PAGE, fluorescence was attenuated, and the migration distance of the fluorescing band increased with the number of Trp residues (Fig. 3A). The increased migration may have been due to the binding of more SDS to Trp residues as the Trp chain was extended in the different constructs. In addition, the attenuation of fluorescence associated with Trp extension corresponded with the staining intensity of CBB (Fig. 3B). However, in denaturing SDS/PAGE, GFP-Trp n bands were observed close to 25 kDa (the molecular mass of GFP is 27 kDa), regardless of the number of Trps added (Fig. 3C). These results indicate that non-fluorescent GFP migrates at a different position to fluorescent GFP (Fig. 3A,B) migration positions suggest that a structural change had occurred in non-fluorescent GFP from the addition of Trps at the C terminus. To investigate this structural change in detail, the sonicate was separated into supernatant (soluble fraction) and pellet (insoluble fraction) by centrifugation and each boiled product was subjected to SDS/PAGE. Bands of GFP containing added Trp were observed in both the soluble and insoluble fraction (Fig. 3D), with those containing five or six added Trps being more abundant in the insoluble material (Fig. 3E). These results suggest that multiple Trps at the C terminus sterically inhibit correct folding of the GFP and result in insolubility. Therefore, it seemed likely that the additional Trps did not reduce fluorescence because of quenching [hypothesis  addition of a peptide, His-tag or heterologous protein negatively affects the solubility of the target protein of interest [15][16][17][18][19]. The C-terminal Trp chain may have a negative effect on the formation of the 11th strand of GFP, which is essential for fluorescence [20]. However, we did not attempt to clarify the detailed mechanism by which multiple C-terminal Trps reduce GFP solubility in this study. This will be addressed in a future experiment appending Trp chains to a robustly folded (superfolder) GFP [19] or self-assembly experiments with the 1-10th strands of GFP and 11th strand appending the Trp chain [20].

Application of pGFP-Trp 6 as a cloning vector
If the multiple Trp codons of pGFP-Trp 6 are not expressed, the resulting GFP will fluoresce. Based on this principle, we constructed a cloning vector that can be used to identify recombinant E. coli colonies by GFP fluorescence. This vector has an SmaI recognition site between the GFP and Trp codons. So if a DNA fragment with a stop codon at the 5 0 -end is inserted in the SmaI site, recombinant E. coli colonies will fluoresce upon UV irradiation, making it easy to screen for recombinant. First, we constructed pGFP-SmaI-Trp 6 (Fig. 4A). However, E. coli harboring this construct fluoresced slightly; this small recovery of fluorescence may have been due to two amino acid residues (Gly-Pro) being inserted at the SmaI recognition site (GGGCCC). We therefore constructed pGFP-SmaI-Trp 8 , to offset the effect of the Gly-Pro residue, and E. coli harboring this construct did not fluoresce (Fig. 4B). After digestion of pGFP-SmaI-Trp 8 using SmaI, a DNA that had been amplified by PCR using a primer having a stop codon at the 5 0end was inserted into the plasmid. The recombinant E. coli colonies fluoresced intensely (Fig. 4C). Compared with blue/white selection, the screening for recombinants using fluorescent protein as an indicator does not require isopropyl-b-D-thiogalactopyranoside or 5-bromo-4-chloro-3-indolyl-b-D-galactoside and is easy to perform [21,22]. 'Turn-on' vectors (where recombinants acquire fluorescence) are superior to 'turn-off' vectors (where recombinants lose fluorescence), because recombinant colonies that emit fluorescence are easier to identify than ones that do not [22]. Moreover, even if non-recombinant satellite colonies grow, it is still possible to select recombinants only (Fig. 4D). We have not yet determined the frequency of false positives, so we cannot decisively conclude that it is low. However, when this plasmid was cleaved with SmaI and subjected to selfligation using T4 DNA ligase, fluorescent colonies appeared at a frequency of only 0-1 per 50-100 colonies (data not shown). The 'turn-on' vector that we reported previously was fortuitously obtained by random mutation [22]. However, the 'turn-on' vector developed in this study was intentionally designed based on our finding from the previous 'turn-on' vector that the addition of hydrophobic peptides to GFP attenuates its fluorescence. As long as the positional relationship between GFP and the Trp chain is retained, transformation efficiency of E. coli, ligation efficiency and the false positive (or negative) rate might be improved by changing the plasmid size and/ or cloning site.

Conclusion
A tandem repeat of six Trps at the C terminus of GFP prevents its folding and markedly decreases its fluorescence intensity. Using this phenomenon, we constructed a cloning vector that facilitates identification of recombinant colonies of E. coli by activation of GFP.