Crystal Structure of Phototoxic Orange Fluorescent Proteins with a Tryptophan-Based Chromophore

Phototoxic fluorescent proteins represent a sparse group of genetically encoded photosensitizers that could be used for precise light-induced inactivation of target proteins, DNA damage, and cell killing. Only two such GFP-based fluorescent proteins (FPs), KillerRed and its monomeric variant SuperNova, were described up to date. Here, we present a crystallographic study of their two orange successors, dimeric KillerOrange and monomeric mKillerOrange, at 1.81 and 1.57 Å resolution, respectively. They are the first orange-emitting protein photosensitizers with a tryptophan-based chromophore (Gln65-Trp66-Gly67). Same as their red progenitors, both orange photosensitizers have a water-filled channel connecting the chromophore to the β-barrel exterior and enabling transport of ROS. In both proteins, Trp66 of the chromophore adopts an unusual trans-cis conformation stabilized by H-bond with the nearby Gln159. This trans-cis conformation along with the water channel was shown to be a key structural feature providing bright orange emission and phototoxicity of both examined orange photosensitizers.

The absorbance spectra of KillerOrange and mKillerOrange possess two overlapped bands with maxima at approximately 455 nm and 514 nm (Fig 2). Excitation at these wavelengths produces weak cyan (λ em~4 80 nm) and bright orange (λ em~5 55 nm) fluorescence, respectively. Most likely, the shorter and longer wavelength forms correspond to the CFP-like [16][17][18] and mHoneyDew-like [19] chromophores, respectively. Unlike parental KillerRed, which is toxic under green/orange light illumination, KillerOrange develops phototoxicity under blue/cyan light. The new orange variants expand the palette of genetically encoded photosensitizers and in combination with KillerRed they would make a useful pair for independent simultaneous control of two cell populations [15].
Here we present 1.81 and 1.57 Å crystal structures of two new orange-emitting photosensitizers with tryptophan-based chromophore (Gln65-Trp66-Gly67), dimeric KillerOrange and monomeric mKillerOrange (Fig 1) and discuss the results of the complementary mutagenesis experiments carried out to elucidate both the orange emission and the phototoxicity of these proteins.

Protein Expression, Purification, and Crystallization
The plasmids KillerOrange/pQE-30 and mKillerOrange/pQE-30 were transformed into E. coli XL1 Blue cells. The proteins were expressed in 3L of LB supplemented with 100 mg/L ampicillin by overnight incubation at 37°C. No induction by IPTG was applied since promoter leakage was sufficient for effective expression. Cells were resuspended in phosphate buffer (pH 7.4, PanEco, Russia), and lysed by sonication. Supernatant clarified by centrifugation was applied to a Talon metal-affinity resin (Clontech, USA) and washed with 10 column volumes of phosphate buffer saline (pH 7.4; 10 mM phosphate buffer, 137 mM NaCl, 2.7 mM KCl). The target protein was then eluted with the same phosphate buffer containing 250 mM of imidazole. Final purification was achieved by size-exclusion chromatography on a Superdex 200 HiLoad (16/ 60) (Amersham, USA). Protein concentration and buffer exchange were performed in 10 kDa molecular-weight cutoff concentration units (VivaScience, USA).
Crystals of KillerOrange and mKillerOrange were grown by hanging drop vapor diffusion method at 20°C. Each drop consisted of 2 μl of~22 mg/ml protein solution in 20 mM Tris pH 8.0, 200 mМ NaCl, mixed with an equal amount of the respective reservoir solution. For KillerOrange, it contained 0.02M Na acetate pH 5.0, 1% (w/v) gamma-PGA (Na + form, LM), 3% (w/v) PEG 4K; (where PGA-LM is poly-gamma-glutamic acid low molecular weight polymer). For mKillerORange, it contained 0.09M citric acid pH 3.5 22.5% PEG 3350. The crystals reached their final size in two weeks. X-ray Data collection, structure solution, and crystallographic refinement Diffraction data were collected from a single crystal flash-cooled in a 100 K nitrogen stream. Prior to cooling, the crystal was transferred to a cryoprotecting solution containing 20% of glycerol and 80% of reservoir solution. The data were collected at a wavelength of 1 Å with a MAR300 CCD detector at the SER-CAT beamline 22ID (Advanced Photon Source, Argonne National Laboratory, Argonne, IL) and were processed with HKL2000 [20]. Crystal structures of both phototoxic orange FPs were solved by the molecular replacement method with MOLREP [21,22], using the coordinates of KillerRed monomer (PDB ID: 3GB3; [23]) as a search model. Structure refinement was performed with REFMAC5 [24], alternating with manual revision of the model using COOT [25]. Crystallographic data and refinement statistics are given in Table 1. The coordinates and structure factors of the KillerOrange and mKil-lerOrange were deposited in the Protein Data Bank under accession codes 4ZFS and 4ZBL, respectively.

Mutagenesis and photophysical characterization
The KillerOrange mutant variants were obtained by site-directed mutagenesis using selfassembly cloning [26]. SuperNova-encoding plasmid was kindly provided by Prof. Takeharu Nagai. Synthetic DNA oligonucleotides for mutagenesis were purchased from Evrogen (Russia). PCRs were carried out using PTC-100 thermal cycler (MJ Research, USA). PCR products and products of digestion were purified by gel electrophoresis and extraction of DNA using Cleanup Standard Kit (Evrogen, Russia).
Mutants were cloned into a pQE-30 vector (Qiagen, USA). Small-scale expression was performed in E. coli XL1 Blue strain (Invitrogen, USA) grown at 37°C on Petri dishes with LB agar (100 mg/ml ampicillin), with no IPTG induction. After cell sonication, mutant proteins were purified using TALON metal affinity resin (Clontech, USA). Absorption and excitation-emission spectra of the purified proteins were recorded with Varian Cary 100 UV/VIS spectrophotometer and Varian Cary Eclipse fluorescence spectrophotometer, respectively. Quantum yield for orange emission of mutant proteins was determined by direct comparison with that of Kill-erOrange (assuming quantum yield of KillerOrange to be 0.42 [15]). The extinction coefficients were determined as following. To determine the concentration of mature (chromophore-containing) protein, aliquots of protein were denatured in 5M NaOH and their absorption spectra were measured. Under these conditions, proteins with tryptophan-based chromophores show an absorption peak at 463 nm with extinction coefficient of 46000 M -1 cm -1 enabling determination of the protein concentration [15]. The extinction coefficient of the native protein was calculated from the absorption spectra of the aliquots of the same concentration in PBS pH 7.5.
The phototoxicity of KillerOrange and its mutants was evaluated by bacterial cell killing test described in [15]. Briefly, we mixed E. coli cells expressing EGFP with E. coli cells expressing the tested mutant, illuminated half of the suspension with a strong blue light, while keeping another half in the dark. We then plated each aliquot on Petri dishes, let the colonies grew

Overall structure
The crystal asymmetric unit of KillerOrange contains two dimers and one monomer forming a dimer with the symmetry related subunit, while the crystal asymmetric unit of mKillerOrange contains one monomer. The dimeric structure of KillerOrange comprises two monomers related by a non-crystallographic 2-fold symmetry axis that are positioned at an angle of~120°r elative to each other. The contact area (~1400 Å 2 per monomer) between the subunits is stabilized by symmetry related van der Waals contacts, two weak hydrogen bonds between the side chain of His148 (A/D) and the carbonyl of Glu146 (D/A), and two strong peripheral salt bridges between Glu99 (A/D) and Arg158 (D/A). Two aromatic side chains of Phe162 from each subunit make a stabilizing stack (at~3.6 Å) shielded by side chains of four nearest His148/176 (A/D) and two Leu160 (D/A) residues at the interface center near the symmetry axis. In Killer-Orange dimer, the irregular C-terminal tail (residues 224-230) sticks to the surface of the interacting subunits, contributing to the interface stabilization. The contact surface between the monomers of the adjacent dimers is smaller (~735 Å 2 per monomer) than within a dimer and is stabilized by two salt bridges, two hydrogen bonds, and a few hydrophobic contacts. Similar to parental SuperNova [2], destabilization of central hydrophobic patch in the dimeric Killer-Orange by Leu160Thr and Phe162Thr replacements ensures a monomeric state of mKillerOrange.
The principal structural fold of the monomer in KillerOrange/mKillerOrange, shared with all members of the GFP family, is an 11-stranded β-barrel having loops from both sides and the chromophore (matured from Gln65-Trp66-Gly67 sequence) embedded in the middle of the internal α-helix wound around the β-barrel axis. The structure shows existence of a pore formed by the backbone of Ile142, Leu143, Pro144, Ile199, Ile200, and Thr201. The pore is filled by four water molecules connecting the indole moiety of the chromophore with the protein exterior. (Fig 3). The pore was found in the earlier reported photosensitizers KillerRed and SuperNova [2,13,14] and in numerous non-phototoxic fluorescent proteins, such as Tur-boGFP (variant of ppluGFP2, Pontellina plumata) [27]; wild type zGFP506, zYFP538, and zRFP574 (Zoanthus) [28,29]; and eqFP578 (Entacmaea quadricolor) [30]. The authors of [27] suggested that the pore affords an additional oxygen supply to the chromophore and accelerates its maturation.
Similar to their red progenitors, KillerOrange and mKillerOrange, show the presence of a water-filled channel extending along the β-barrel axis (Fig 3). A continuous chain of seven hydrogen-bonded water molecules in the channel covers the distance of~20 Å from catalytic Glu218 in the chromophore area to Pro192 at the bottom of β-barrel. Each water molecule forms hydrogen bonds with the preceding and following water molecules and with the proximal amino acid residues lining the channel. This water-filled channel is a distinct feature of GFP-like photosensitizers and was suggested to be a key structural element providing their phototoxicity [13,14]. In the channel of mKillerOrange, we found a glycerol molecule, presumably coming from the cryoprotecting solution and replacing three out of seven water molecules present in KillerOrange.

Chromophore area
Posttranslational modification of the chromophore-forming sequence Gln65-Trp66-Gly67 (QWG) in KillerOrange/mKillerOrange results in a chromophore consisting of a five-membered imidazolinone heterocycle and an aromatic nine-membered indole ring. In both proteins, the nearest environment of the QWG chromophore (within 3.9 Å) is nearly identical and consists of 19 residues with the only difference at position 177 occupied by Leu and Phe in Kill-erOrange and mKillerOrange, respectively (Figs 1, 4 and 5). The chromophore forms nine direct hydrogen bonds with its immediate environment, including those with the catalytic Glu218 and Arg94. The side chain of Trp66 of the chromophore in KillerOrange and mKiller-Orange adopts an unusual trans-cis conformation, not seen for other Trp-based chromophores before. It is described by the χ1/χ2 torsion angles of 175°/16°and -178°/26°, respectively (rotation around C α -C β and C β -C γ bonds defined by dihedral angles N-C α -C β -C γ and C α -C β -C γ -   Fig 6). A higher resolution structure of mKillerOrange also shows the traces (~10-15%) of alternative cis-cis conformation with χ1/χ2 of -5°/-8°. In both structures, the indole ring of Trp66 in a dominant trans-cis conformation is stabilized by hydrogen bond between the indole nitrogen and the side chain of Gln159 and shows a noticeable deviation from the chromophore planarity (χ2~16-26°). Conversely, in the minor cis-cis conformation  the indole ring is nearly coplanar with the chromophore imidazolinone ring and is stabilized by a hydrogen bond with Thr201.

Structure-based site-directed mutagenesis
Based on the obtained crystal structures we designed site-directed mutagenesis experiments aimed at identifying of the residues responsible for phototoxicity of KillerOrange and mKiller-Orange. Gln159 and Thr201 were identified as sites affecting the equilibrium of trans-cis and cis-cis conformations of Trp66. Also, in the earlier study of KillerRed [14], Glu68 located right after the chromophore was shown to influence the protein phototoxicity. To test these chromophore-adjacent residues, we have prepared the mutants with the following single substitutions: The χ1 andχ2 torsion angles (grad) around C α -C β and C β -C γ bonds (defined by dihedral angles presented by atoms N-C α -C β -C γ and C α -C β -C γ -C δ , respectively). c [18]. d [17].
doi:10.1371/journal.pone.0145740.t002 Gln159Gly or Glu68Gln in KillerOrange, and Thr201Ala in mKillerOrange. Unfortunately, all three mutations resulted in almost immature proteins that impeded any conclusion regarding the specific role these residues play in (m)KillerOrange phototoxicity and fluorescence.
We have also applied a site-directed mutagenesis to block the water-filled channel claimed to be crucial for ROS generation [13,14]. Ile199Phe and Ile199Leu replacements resulted in a considerable loss of phototoxicity (Fig 7), while the spectral properties of the corresponding mutants were rather similar to those of KillerOrange (Table 3). Surprisingly, replacement of Ile199 with a positively charged bulky Lys resulted in a low-phototoxic protein with predominantly cyan emission (Fig 2C, Table 3). The other "water channel" mutants Ala82Ser and Ile199Met/Arg were dim and exhibited a very low maturation rates, preventing direct comparison of their phototoxicity with that of the parental KillerOrange.

Discussion
The phototoxicity of the GFP-like proteins is usually very low-the only known GFP-like photosensitizers with substantial phototoxicity reported to date are red-emitting KillerRed and its monomeric variant SuperNova with tyrosine-based chromophore. Replacement of the chromophore Tyr66 in KillerRed with Trp, despite a significant change of spectral properties, had little effect on phototoxicity of its (m)KillerOrange successors. Chromophores in red and orange photosensitizers have a nearly identical nearest amino acid environment with a similar  interaction pattern showing the major difference in the area of the aromatic side chains of the chromophores.
In red photosensitizers, cis conformation of Tyr66 is stabilized by direct and water-mediated hydrogen bonds with Asn145 and Thr201, respectively. In orange photosensitizers, Trp66 adopts a dominant trans-cis and a minor cis-cis conformations stabilized by hydrogen bond with Gln159 and Thr201, respectively. The trans-cis isomer of (m)KillerOrange chromophore is the first example of such conformation among the known Trp-based chromophores. All earlier reported Trp-based chromophores adopted either cis-cis or cis-trans conformation and exhibited either cyan or green emission in case of protonated [17,18] and anionic Trp66 [31], respectively (Fig 6, Table 2). Ile199Lys replacement in KillerOrange resulted in~25 nm blue shift of absorption band maxima accompanied by significant decrease of the protein phototoxicity and raise of cyan emission (Figs 2 and 7, Table 3), presumably due to isomerization of Trp66 from dominant trans-cis to minor cis-cis conformation. We suggest that trans-cis conformation of the chromophore Trp66 in KillerOrange is a key structural feature for appearance of bright orange emission and phototoxicity.
The water-filled channel connecting the chromophore with exterior of the β-barrel was identified as a prerequisite feature for phototoxicity of GFP-based photosensitizers. The continuous chain of seven hydrogen bonded water molecules extending along the channel axis has been suggested to transport oxygen and ROS between the solvent and the chromophore [2,13,14,[32][33][34]. Alternatively, highly ordered water molecules could play a role of "electron wire", conducting electrons from the external redox species, which was a subject of a number of publications, e.g., [35,36].
The results of mutagenesis demonstrated that interruption of the string of hydrogenbonded water molecules in the channel by introduction of bulky residues, Ile199Phe and Ile199Leu, results in decreased phototoxicity of the mutants, presumably, due to decelerated ROS production. All tested mutants I199Phe/Leu/Lys possessed remarkably higher quantum yield of orange fluorescence compared to the parental protein ( Table 3), suggesting that bulky residues at position 199 could restrict the freedom of the chromophore movements suppressing its non-radiative transition.

Conclusions
Here, we presented the X-ray structures of dimeric KillerOrange and monomeric mKillerOrange, the only known to date GFP-photosensitizers with Trp-based chromophore. Same as for its red progenitors, KillerRed and SuperNova, to exhibit phototoxicity KillerOrange and its monomer require the presence of the water-filled channel connecting the chromophore to the β-barrel exterior and enabling transport of ROS. Even a partial blockage of the channel decelerates production of ROS reducing the protein phototoxicity. Specifically for KillerOrange and its monomer, never reported before dominant trans-cis conformation of the chromophore Trp66, stabilized by Hbond with the adjacent Gln159, provides orange fluorescence of both photosensitizers and enables their phototoxicity. The third feature, worth mentioning in conclusion, is the position #199 occupied by Ile. Replacement of Ile199 by bulky Phe, Leu or Lys decreases phototoxicity and results in almost doubled quantum yields of 0.75-0.87. The structural and mutagenesis data presented here provide a solid base for further investigation of the mechanisms of ROS generation and for development of new, more efficient genetically encoded photosensitizers. grant 14-25-00129. Experiments were partially carried out using the equipment provided by the IBCH core facility (CKP IBCH). Diffraction experiments were carried out at beamline 22-ID of the Southeast Regional Collaborative Access Team (SER-CAT), located at the Advanced Photon Source, Argonne National Laboratory, USA. Use of the APS was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. W-31-109-Eng-38. The project has been supported in part with Federal funds from the National Cancer Institute, National Institutes of Health under contract No. HHSN261200800001E, by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U. S. Government.