Structure-guided Engineering Enhances a Phytochrome-based Infrared Fluorescent Protein*

Background: Engineered variants of the phytochrome photoreceptor are infrared fluorescent proteins. Results: Based on crystal structures, side chain substitutions near the chromophore were combined with monomerization of truncated phytochrome to yield an enhanced fluorophore. Conclusion: Amino acid changes that increase fluorescence discourage photoproduct formation. Significance: This improved infrared phytofluor provides long-wavelength excitation for high signal to noise in tissue and whole animals. Phytochrome is a multidomain dimeric red light photoreceptor that utilizes a chromophore-binding domain (CBD), a PHY domain, and an output module to induce cellular changes in response to light. A promising biotechnology tool emerged when a structure-based substitution at Asp-207 was shown to be an infrared fluorophore that uses a biologically available tetrapyrrole chromophore. We report multiple crystal structures of this D207H variant of the Deinococcus radiodurans CBD, in which His-207 is observed to form a hydrogen bond with either the tetrapyrrole A-ring oxygen or the Tyr-263 hydroxyl. Based on the implications of this duality for fluorescence properties, Y263F was introduced and shown to have stronger fluorescence than the original D207H template. Our structures are consistent with the model that the Y263F change prevents a red light-induced far-red light absorbing phytochrome chromophore configuration. With the goal of decreasing size and thereby facilitating use as a fluorescent tag in vivo, we also engineered a monomeric form of the CBD. Unexpectedly, photoconversion was observed in the monomer despite the lack of a PHY domain. This observation underscores an interplay between dimerization and the photochemical properties of phytochrome and suggests that the monomeric CBD could be used for further studies of the photocycle. The D207H substitution on its own in the monomer did not result in fluorescence, whereas Y263F did. Combined, the D207H and Y263F substitutions in the monomeric CBD lead to the brightest of our variants, designated Wisconsin infrared phytofluor (Wi-Phy).

The absorption of a photon by a photoreceptor can be the initial trigger for a light-driven molecular signaling pathway. This trigger leads to a conformational change within the photoreceptor, resulting in the activation or deactivation of an output domain used to transduce a signal. Phytochromes, found in plants, fungi, and bacteria, are red/far-red photoreceptors that sense the light environment through a linear tetrapyrrole chromophore. Biliverdin IX␣ (BV) 2 is the chromophore typically found in bacterial phytochromes (BphPs), whereas cyanobacterial phytochromes and plant phytochromes bind the further reduced phycocyanobilin and phytochromobilin, respectively (1). Within the phytochrome structure, this light-absorbing chromophore is covalently attached via a thioether linkage to a Cys and cradled in a pocket composed of residues within the hallmark GAF (cGMP phosphodiesterase/adenyl cyclase/FhlA) domain. In BphPs, the Cys is found in an N-terminal attachment region, and this is separated in the primary sequence from the GAF domain by an intervening PAS (Per/ARNT/Sim) domain; together, these three structural elements are referred to as the chromophore-binding domain (CBD). In canonical phytochromes, the CBD is trailed by the PHY domain, which contributes to the complete enclosure of the bilin (2,3). Following the PHY domain is an output module, which, in many cases, is a histidine kinase (4). Absorption of a photon by the redabsorbing state (Pr) leads to the excited Pr* state with a strained C15ϭC16 bond. Isomerization about this bond leads to the second excited state, Lumi-R*, followed by relaxation to form the first photoproduct, Lumi-R (5,6). A proton is then transiently released, possibly from BV and possibly to the tightly bound pyrrole water or the main chain carbonyl oxygen of an adjacent residue, to form Meta-R c . Subsequent proton reuptake and protein conformational changes lead ultimately to the farred-absorbing form (Pfr).
Following structure determination of the dimeric Deinococcus radiodurans CBD (DrCBD) (7,8), Wagner et al. probed the chromophore-binding pocket for amino acid positions impor-tant for proper photochromicity. If the evolutionarily selected energy transduction pathway starting with the absorption of a photon is disrupted by variations in amino acids of the photoreceptor, light energy leads to an excited state but cannot result in productive rearrangement of the chromophore or conformational changes within the protein. Instead, fluxes through nonproductive pathways, including fluorescence, are increased. Thus, several substitutions were identified that reduced photoconversion and/or increased fluorescence in the D. radiodurans BphP (DrBphP) (9). In particular, strong red fluorescence was observed for the DrBphP-D207H variant. DrCBD-D207H was subsequently optimized for increased brightness with 12 additional substitutions leading to IFP1.4, a monomeric DrCBD with increased red fluorescence albeit undesired blue shifting of absorption and emission maxima (10). As an important application demonstration, IFP1.4 has been shown to work as an imaging tool in whole mice. Intrinsically fluorescent phytochromes are also known and help explain the origins of fluorescence. For example, fluorescence of the atypical BphP3 from Rhodopseudomonas palustris (referred to as P3) has been elegantly demonstrated by Kennis and co-workers (11) to be the result of a long (ϳ300 ps) excited state life time. The source of the long life time is three stabilizing polar interactions between protein side chains and the Pr conformation of the BV D-ring, which are not observed in canonical BphPs (11,12). These investigators further showed that increased fluorescence in two P3 variants analogous to DrBphP-D207H and DrBphP-Y263F was due to further increased excited state life times (13).
Here, we report three-dimensional structures and near-infrared light-excitable fluorescence properties of DrCBD-D207H. On the basis of the interpretation of these structures, we introduced the substitution Y263F for increased fluorescence. Additionally, we engineered a monomeric DrCBD and discovered unexpected photochemical properties.
Protein Purification-DrCBD and IFP1.4 constructs were transformed into chemically competent BL21(DE3) cells (Invitrogen). Cells were grown at 37°C in LB medium supplemented with ampicillin (0.1 mg/ml final concentration). When cells reached A 600 ϳ 0.6, expression was induced with the addition of isopropyl ␤-D-thiogalactopyranoside to a 1 mM final concentration. After 4 h at 20°C, cells were harvested at 6200 ϫ g for 30 min and then resuspended and lysed in 25 ml of 30 mM Tris (pH 8.0), 50 mM NaCl, 5 mM imidazole, and 1 mM tris(2-carboxyethyl)phosphine in a French pressure cell at 1000 p.s.i. Lysates were cleared for 45 min at 40,000 ϫ g and then incubated on ice in the dark with a final concentration of 0.16 mM BV (stock solution prepared by dissolution in Me 2 SO; Frontier Scientific Inc., Logan, UT) for 1 h. (IFP1.4 required two overnight incubations, prior to and following Ni 2ϩ affinity purification, to obtain an equivalent specific absorbance as the DrCBD variants.) Proteins were affinity-purified under green light using nickel-nitrilotriacetic acid resin (Qiagen, Valencia, CA). Further purification was performed using a hydrophobic interacting phenyl-Sepharose column (GE Healthcare) to separate apoand holophytochrome. Ammonium sulfate was added to the protein at a final concentration of 0.35 M prior to injection onto the phenyl-Sepharose column, which was pre-equilibrated with 30 mM Tris (pH 8.0) and 0.3 M ammonium sulfate. Phytochrome was eluted during a 20-ml gradient to 30 mM Tris (pH 8.0). To maximize holophytochrome content, only fractions with specific absorbances of ϳ2.0 were kept (14). Finally, purified protein was dialyzed against 30 mM Tris (pH 8.0) and stored at Ϫ80°C.
Monomer Design and Analytical Ultracentrifugation-Monomeric DrCBD (DrCBD mon ) was designed by inspection of the biologically relevant dimer interface and by the use of the alanine-scanning mutagenesis software Robetta (15) and the KFC server, a web-based tool used for the prediction of "hotspots" in protein-protein interfaces (16). The DrCBD mon was subjected to equilibrium ultracentrifugation in a Beckman Coulter XL-A analytical ultracentrifuge. Protein was dialyzed against 30 mM Tris and 125 mM NaCl (pH 8.0), and the dialysate was used for protein dilution. 100-l samples were placed in double-sector charcoal-filled Epon centerpieces with ϳ108 l of dialysate as reference. Concentration gradients were recorded at 4°C at both 280 and 390 nm for three concentrations of protein and nine speeds. Equilibrium was judged as superimposable gradients at 280 nm collected 2-3 h apart. There was no evidence of loss of material after 16,800 or 20,000 rpm data collection based on superimposable 12,000 rpm data collected before and after these higher speed runs. Data were fit to various models using a program written for IGOR Pro (WaveMetrics, Inc., Lake Oswego, OR). For analysis, the density and extinction coefficient of the aqueous solvent were assumed to be 1 g/ml and 14,900 M Ϫ1 cm Ϫ1 , respectively. The partial specific volume and extinction coefficient of the apoprotein were calculated from the sequence to be 0.737 ml/g and 30,940 M Ϫ1 cm Ϫ1 , respectively, and the molecular mass of the holoprotein including the covalently attached chromophore was calculated to be 37,610 Da.
Crystallization and Data Collection-Proteins were concentrated to ϳ20 mg/ml in 30 mM Tris (pH 8.0) and crystallized by hanging drop vapor diffusion with drops containing a 1:1 mixture of protein and reservoir solutions. DrCBD-D207H crystals formed when reservoir solutions contained 0.1 M sodium citrate (pH 5.6), 14 -20% (v/v) PEG 4000, and 20% (v/v) isopropyl alcohol. The cryoprotectant was 8% (v/v) glycerol or 10 or 20% (v/v) ethylene glycol in reservoir solution, resulting in the HiRes, H:Y263, and H:Aring data sets, respectively. Crystals were dipped into cryoprotectant no longer than 5 s and flashcooled in liquid N 2 . Data for the DrCBD-D207H phytochrome crystals were collected either at LS-CAT beamline 21-ID-D at the Advanced Photon Source (Argonne, IL) for the HiRes and H:Aring structures or on an in-house Bruker HI-STAR rotating anode with an R6000 CCD detector for the H:Y263 structure. DrCBD mon -D207H/Y263F (Wi-Phy) crystals were identified in a JCSGϩ Suite (Qiagen) screen, and the optimized reservoir solution contained 0.1 M phosphate citrate (pH 4.2), 3% (v/v) PEG 1000, 20% (v/v) ethanol, and 6% (v/v) glycerol. Wi-Phy crystals were vitrified directly from their mother liquor. Wi-Phy data were also collected at LS-CAT beamline 21-ID-D. Each crystal analyzed was in space group C2 with one molecule per asymmetric unit.
Each data set was integrated and scaled using HKL2000. The DrCBD (Protein Data Bank code 2O9C) (8) was used as a search model for molecular replacement using Phaser (17) for the HiRes, H:Y263, and Wi-Phy structures. For these three searches, the resolution range, log likelihood gain for the correct model, and R-factor of the correct solution were 4.0 -19.4, 3.5-27.5, and 2.5-28.5 Å; 3172, 2191, and 1973; and 40.7, 38.5, and 46.1%, respectively. The HiRes refined phases were used as starting phases for the H:Aring structure. For all structures, model building with Coot (18) was alternated with structure refinement using REFMAC5 as implemented in CCP4 6.1.13 (19). For the HiRes structure, atomic displacement parameters (i.e. B-factors) were refined anisotropically, whereas the other structures were refined isotropically with TLS (20). The PAS and GAF domains were used to define two TLS groups. Hydrogens were added in the riding position for all structures. The coordinates and structure factors have been deposited in the Protein Data Bank (HiRes, 3S7O; H:Aring, 3S7P; H:Y263, 3S7N; and Wi-Phy, 3S7Q).
Spectroscopy-To ascertain the photoconversion capabilities of our variants, we performed absorption scans after 1) dark incubation (Pr), 2) 15 min of irradiance with 700 nm light (Pfr), and 3) 15 min of irradiance with 750 nm light (reversion to Pr). The 700 and 750 nm light were provided by a Fostec ACE light source fitted with either a 700-or 750-nm interference filter with a Ϯ5-nm bandwidth (Andover Corp., Salem, NH). Fluence rates were determined with an International Light Technologies RPS900 spectroradiometer. The light source-to-sample distance was adjusted so that irradiances of 140 mol/m 2 /s were used for both 700 and 750 nm light. Absorption wave-length scans in 1-nm steps from 250 to 800 nm were performed on a Beckman Coulter DU640B spectrophotometer.
Fluorescence measurements were taken on a Tecan Infinite M1000 Monochromator-based plate reader with a bandwidth of 5 nm. Excitation/emission scans were run in Greiner FLUO-TRAC 200 96-well flat-bottom black microplates. Excitation scans were run with emission monitored at 728 nm (705 nm for IFP1.4), and emission scans were run with excitation monitored at 695 nm (676 nm for IFP1.4 and 645 nm for Cy5), both in 5-nm steps. Each sample was measured in triplicate. Instrument calibration took into account the wavelength dependence of the detector efficiency, allowing for the use of a comparative method to determine the quantum yield of variants using Cy5 as a reference. Each emission spectrum was corrected for the effect of constant-bandwidth measurements by multiplying each intensity by 2 . Then the integral of the emission spectrum was used for quantum yield ( F ) determinations by comparison with Cy5-N-hydroxysuccinimidyl ester (Combinix, Irvine, CA) in Tris (pH 8.0), the quantum yield of which is 0.27 in PBS (21). Emission integrations of Cy5 in PBS and 30 mM Tris (pH 8.0) were within Ϯ5% of one another. Absorption at 695 nm (645 nm for Cy5 and 676 nm for IFP1.4) and integration of the area under the emission spectrum of each sample at four concentrations were measured. Absorbance was plotted against emission integration for each DrCBD sample and the Cy5 standard, and the slope of a best fit line for each was determined. The slope of the DrCBD sample in comparison with the slope and known quantum yield of Cy5 was used to calculate the quantum yield of each sample. Because scans were truncated at 800 nm, reported quantum yields are an underestimate of the true quantum yields.
To ascertain the extinction coefficient (⑀) for each variant using the Beer-Lambert law (A ϭ ⑀cl), holoprotein concentration was first estimated based on the assumption that absorbance at 388 nm was due to bound BV and that the ⑀ 388 of the holoprotein was equal to that of free BV, 39,900 M Ϫ1 cm Ϫ1 (10). This concentration was then used along with the measured absorbance at 695 nm to determine the extinction coefficient. Brightness, defined as the product of F and ⑀, is reported as a percentage of DrCBD brightness.

RESULTS
Crystal Structure of a Near-infrared Fluorescent Phytochrome-DrCBD-D207H is fluorescent in the near-infrared with excitation and emission maxima at 700 and 720 nm, respectively (Fig. 1A). To gain insight into the fluorescent nature of DrCBD-D207H, high-resolution crystal structures were obtained. The DrCBD-D207H structure was refined to 1.24 Å resolution, the highest resolution phytochrome structure to date ( Table 1). The protonated states of each pyrrole ring nitrogen and His-207 were confirmed by the high-resolution electron density map (supplemental Fig. 1, A and B). Because the thioether linkage that makes up the covalent bond between Cys-24 of the DrCBD and the BV chromophore is sensitive to radiation damage (supplemental Fig. 1A) (7,12), we also carried out a more gentle data collection on a separate crystal, resulting in a 1.7 Å resolution structure with an unambiguously linked chromophore (Table 1 and supplemental Fig. 1C).
With the exception of the chromophore linkage, no changes were noted between the two structures (supplemental Fig. 1D). This observation validates the hypothesis that the protein structure itself drives the positioning of the chromophore, and thioether formation requires only proper orientation of the tetrapyrrole within the pocket. We make specific reference to the 1.7 Å resolution structure in the following descriptions.
The overall DrCBD-D207H structure, including the PAS and GAF domains and the intervening figure-of-eight knot, deviates little from the previously published 1.45 Å resolution structure of the DrCBD (8), with a root mean square deviation of 0.65 Å over all C␣ atoms (Fig. 1B). N-and C-terminal tags, the first four residues of the native protein sequence, and an interdomain linker at residues 131-134 are disordered. The architecture of the chromophore-binding pocket is preserved, with the propionate groups of the B-and C-rings positioned via charge interactions and a hydrogen bonding network and with a hydrophobic environment surrounding the D-ring (Fig. 1C) (7,8). The A-ring is puckered (8), with the C2 1 methyl group adopting an S stereochemistry by pointing toward the ␣-facial side of BV (Fig. 1C). The A-ring is tipped slightly farther out of the plane formed by the B-and C-rings than in the DrCBD structure (Fig. 1C), which is consistent with minor torsion angle deviations about the methine bridge observed by resonance Raman spectroscopy on DrBphP-D207H (9) as well as heterogeneous Pr chromophore structures reported for several phytochromes (22,23).
Asp-207 is part of a highly conserved motif also consisting of Ile-208 and Pro-209. Thus, replacement of Asp-207 in this DIP motif with histidine had potential structural consequences. In these DrCBD-D207H structures, the positions of Ile-208 and Pro-209 are invariant compared with the DrCBD (Fig. 1C). Also unaltered are the main chain atoms of residue 207, and thus, the hydrogen bonding network among the carbonyl oxygen of residue 207, the nitrogens of the A-, B-, and C-rings of BV, and the pyrrole water is maintained (Fig. 1D).
In the DrCBD, the carboxylate of Asp-207 and the hydroxyl of Tyr-263 were noted as being in hydrogen bonding distance (8). This hydrogen bond is missing in these DrCBD-D207H structures, with the distance between the His-207 N␦1 and the Tyr-263 hydroxyl now 4.1 Å (Fig. 1D). In comparison with its position in the DrCBD, Tyr-263 migrates away from position 207 toward His-290, and this position is supported by interaction with a water molecule 3.0 Å away. More importantly, in DrCBD-D207H, a new hydrogen bond is formed between the N␦1 of the histidine imidazole and the carbonyl oxygen of the A-ring, thus fixing the A-ring and histidine side chain orientations ( Fig. 1D and supplemental Fig. 1B). The position of His- 207 is further stabilized by a water molecule 2.9 Å from N⑀2. Therefore, in the DrCBD-D207H variant, the A-ring position is stabilized by three potential hydrogen bonds: the first from the A-ring nitrogen to the backbone carbonyl of residue 207, the second from the A-ring nitrogen to the pyrrole water, and the third from the A-ring carbonyl to the histidine N␦1. The last of these is unique to this DrCBD variant.
Alternative His-207 Conformation-During the course of this work, several DrCBD-D207H data sets were collected on crystals grown and cryopreserved under a slightly different set of conditions, leading to structures equivalent to those discussed above except that the imidazole ring of His-207 is rotated ϳ90°( Fig. 2A and Table 1). This novel rotamer prevents the hydrogen bond of His-207 with the A-ring carbonyl but restores a hydrogen bond between residue 207 and Tyr-263 (3.3 Å from the His-207 N␦1 to the Tyr-263 hydroxyl). On the other side of His-207, a water molecule is present 3.0 Å from the His-207 N⑀2 (Fig. 2B). No other appropriate bonding partner is within 4 Å of the His-207 imidazole nitrogens. To distinguish the two DrCBD-D207H structures, we refer to them henceforth based on the orientation of the His-207 imidazole: H:Aring or H:Y263. An additional distinguishing feature of the H:Y263 structure is the gauche(Ϫ) rotamer of Cys-24. Observations of either the gauche(ϩ) or gauche(Ϫ) Cys-24 rotamer have been noted in the other structurally characterized phytochromes (3). Clearly, His-207 has some freedom of rotation, and this side chain may not always lock the BV A-ring position, as might have been interpreted from the H:Aring structure alone.  DrCBD-D207H and DrCBD-Y263F Are Independently Fluorescent-We reasoned that Tyr-263 acts as an opposing force to a strictly A-ring interaction of His-207 and considered the possibility that eliminating the His-207-Tyr-263 hydrogen bond with a Phe substitution would increase the probability of the orientation of His-207 toward the A-ring and, in so doing, increase fluorescence synergistically. This substitution also maintains the hydrophobic bulkiness contributed by a benzene ring. The contributions of His-207 and Phe-263 to the fluorescence quantum yield ( F ) and brightness ( F ϫ ⑀) were assessed for each amino acid individually and in combination ( Table 2). The combination of DrCBD-Y263F with DrCBD-D207H had little effect on either measure compared with DrCBD-D207H alone. However, an unanticipated result was the strongly increased fluorescence of the single Y263F substitution compared with the DrCBD ( ϭ 0.037 versus 0.018). In fact, the Y263F substitution alone resulted in the fluorophore with the highest quantum yield and brightness, somewhat surprising as the D207H variation was assumed to be the major causal factor in increased fluorescence for the DrBphP (9).
Monomeric CBD as a Template for the Fluorescent Phytochrome-Like all other phytochromes, the DrBphP forms a biologically relevant dimer and, as such, is not an ideal candidate for a genetically encoded fluorophore. The dimer interface in the DrBphP consists of a six-helix bundle contributed by GAF domain helices (Fig. 3A) (8). To improve the usefulness of any fluorescent DrCBD variant for creation of small fusion proteins with biotechnology value, we engineered the dimer interface of the DrCBD to remove stabilizing interactions. Leu-311 and Leu-314 create ideal zipper interactions, which were discouraged by replacing them with negatively charged Glu residues. Phe-145 forms a favorable van der Waals interaction, so it was made smaller via replacement with Ser. The monomeric nature of this triple variant was verified by analytical ultrasedimentation (Fig. 3B).
The D207H and Y263F substitutions were put into the DrCBD mon background, both singularly and in combination. All monomeric variants had approximately the same absorption and fluorescence emission wavelength maxima as the wildtype DrCBD (Table 2). IFP1.4, with its blue-shifted absorption and emission maxima, was included for comparative purposes. As in the dimeric DrCBD, Y263F alone increased DrCBD mon fluorescence. D207H alone did not lead to fluorescence in the DrCBD mon background. In contrast to the dimeric DrCBD case, among DrCBD mon variants, the double substitution DrCBD mon -D207H/Y263F did further increase the quantum yield and brightness over the Y263F substitution alone. Given its improved properties and potential applicability, we have designated DrCBD mon -D207H/Y263F as Wisconsin infrared phytofluor (Wi-Phy).
The photoconversion potential for each DrCBD and DrCB-D mon variant was also tested (Fig. 4). The DrCBD has a defective photocycle and stalls at a deprotonated Meta-R c -like intermediate because of its missing PHY domain (9). Consequently, although it displayed the same loss of the Pr peak at 700 nm as the full-length DrBphP when irradiated, the DrCBD did not have a concomitant increase in the Pfr peak at 750 nm (Fig. 4A), i.e. it did not photoconvert. The most unexpected result of our photoconversion tests was the effect of monomerization on the extent of photochromicity. DrCBD mon spectra show not only  loss of the Pr peak at 700 nm but also the appearance of a Pfr peak at 750 nm upon irradiation (Fig. 4B). The extent of photoconversion is correlated with fluorescence. The fluorescent phytochrome variants DrCBD-D207H, DrCBD-Y263F and DrCBD-D207H/Y263F (Fig. 4A) and DrCBD mon -Y263F and DrCBD mon -D207H/Y263F (Fig. 4B) showed little-to-no change in UV-visible spectra upon illumination. In contrast, DrCBD mon -D207H, the least fluorescent variant compared with the wild-type DrCBD, displayed a modest increase in absorbance at 750 nm ( Fig. 4B and Table 2).
Wi-Phy Crystal Structure at 1.8 Å Resolution-With monomeric D207H/Y263F exhibiting the greatest fluorescence and to confirm the elimination of the six-helix bundle as a dimer interface, Wi-Phy was crystallized, and its structure was refined to 1.8 Å resolution (Table 1). In previous DrCBD structures, the biological dimer axis formed the crystallographic 2-fold axis. Because of disruption of three favorable interactions, the Wi-Phy crystal packing differs significantly, with no involvement of the GAF domain helical bundle ( Fig. 5A and supplemental Fig.  2). Likely due to this difference in quaternary structure, the Wi-Phy structure has a slight variation in the first GAF domain helix, but otherwise, it is not significantly different from other DrCBD structures, with a 0.97 Å root mean square deviation over all C␣ atoms compared with the DrCBD and with BV and pyrrole water positions equivalent to other structures presented here. With respect to His-207, Wi-Phy and H:Y263 both have the imidazole ring in a plane almost parallel to the chromophore (Figs. 2 and 5B). Notably, these two structures also have similar unit cell dimensions and volume (Table 1) and thus water content.
The DrCBD mon A-ring methyl group (C2 1 ) has been modeled with equal occupancy in two enantiomeric conformations (Fig. 5B). A similar observation was reported for P3 (12). The initial attack by Cys-24 on the A-ring vinyl side chain C3 2 position to form the thioether bond is resolved by the addition of a proton to C2. This leads to formation of the chiral center and loss of A-ring planarity. The two observed positions of C2 1 suggest that the proton can originate from either above or below the chromophore. Although chemistry allows either enantiomer, in vivo, the presence of the PHY domain or presentation of the chromophore by heme oxygenase (24) could restrict the orientation.

DISCUSSION
Obtaining fluorophores with emission in the infrared region of the electromagnetic spectrum has become the new frontier in fluorophore research. At these wavelengths, most autofluorescence, as well as scattering and/or absorption from lipids, water, and heme, is avoided, allowing imaging through thick tissue and even in live animals (25). Bacterial phytochromes are uniquely positioned to capitalize on this biophysical opportunity with their intrinsic near-infrared absorption capabilities. Although phytochrome function in vivo would be compro-  mised if absorbed energy were radiated via fluorescence, scientists can manipulate phytochromes to create highly fluorescent phytofluors (26) with desirable properties for imaging studies. Several endeavors into this realm have been made. Incorporation of the non-native tetrapyrrole phycoerythrobilin into plant or bacterial phytochromes can lead to intensely fluorescent proteins (26,27). A Y176H change in Cph1 renders it fluorescent (28,29), and the equivalent substitution in Arabidopsis thaliana phytochromes A and B confers fluorescence and a constitutive red light phenotype (28,30). The DrCBD-D207H variant we reported (9) was exploited in the creation of the brighter albeit somewhat blue-shifted IFP1.4, which has been fused to a liver-specific protein and used for imaging in live mice (10). Most recent is the utilization of R. palustris BphP2 (referred to as P2) as a near-infrared phytofluor, also initially based on the equivalent D207H variation (31).
To increase fluorescence, the pathways for relaxation of the transient excited states formed when a photon is absorbed by the chromophore must be biased away from photoconversion and excited state proton transfer, thereby increasing the excited state life time. This might be accomplished in three ways: 1) by stabilizing the chromophore Pr conformation, thus decreasing C15ϭC16 isomerization in the excited state; 2) by destabilizing the BV conformation after isomerization of C15ϭC16, thus increasing back-flipping from Lumi-R* to Pr*; or 3) by decreasing the BV deprotonation rate and thus the likelihood of excited state proton transfer (11,32). An example of the first mechanism includes the polar side chain interactions with the D-ring seen in the naturally fluorescent P3 (12). Compared with P2, P3 maintains a prolonged excited state, which is the cause for its increased fluorescence (11). A second example of Pr stabilization leading to fluorescence is the packing around the D-ring apparently introduced by several of the substitutions that led to IFP1.4 (10). Our structural analyses support mechanism 2 or 3 as the main source of both the contributions to fluorescence of D207H in the DrCBD and the origin of fluorescence in Y263F variants of the DrCBD and DrCBD mon .
In the case of D207H, replacement of the side chain results in a new hydrogen bonding network (Figs. 1D and 2). In the H:Aring rotamer with a hydrogen bond to the A-ring carbonyl, a circuit is completed, which includes the His-207 backbone carbonyl to the pyrrole water, the pyrrole water to the A-ring nitrogen, and finally the A-ring carbonyl to the His-207 imidazole nitrogen (Fig. 1). This localized network undoubtedly has profound effects on the energetics of proton transfer in the environment of the chromophore.
In the Pfr Pseudomonas aeruginosa BphP structure, the conserved DIP aspartate side chain forms a hydrogen bond with the BV D-ring nitrogen to stabilize the Pfr conformation of the bilin (3,33). The same interaction has been suggested for the Pfr structure of the cyanobacterial Cph1 (23,34). In our DrCBD-D207H H:Aring structure, the His-207 rotamer can no longer stabilize this Pfr chromophore conformation (Figs. 2B and 6). Thus, we suggest that the conformer seen in the H:Aring structure is the one that contributes to fluorescence by biasing His-207 away from the D-ring, which may destabilize Lumi-R* relative to Pr* so that the chromophore readily back-flips rather than progressing to the Lumi-R photoproduct.
The increased fluorescence of Y263F may also be interpreted through its effects on Lumi-R* chromophore stability relative to Pr* and/or the likelihood of BV deprotonation via excited state proton transfer. With the predicted loss of two hydrogen bonds with the BV D-ring nitrogen and oxygen atoms in the Pfr state (Fig. 6), Phe at position 263 would discourage the Lumi-R* chromophore configuration. Moreover, the loss of these polar interactions with the D-ring will have a profound effect on the polarization and deprotonation of BV, potentially including an effect on the keto-enol isomerization of the D-ring oxygen. Variations at position 263 in the full-length DrBphP have been previously studied, with a Y263H substitution showing hindered photoconversion and moderate fluorescence (9). In the truncated DrCBD, the non-polar substitution Y263F we have investigated here shows no photoconversion and is fluorescent ( Fig. 4 and Table 2).
Why does the oligomeric nature of a phytochrome impart photochemical consequences? The dimeric DrCBD, truncated at residue 321 and thus without a PHY domain, is unable to photoconvert (Fig. 4) (7). Asp-207 plays a crucial role in the interaction between the GAF and PHY domains in the Pr state by forming a salt bridge with a conserved Arg, as is evident in the PAS-GAF-PHY phytochrome structures for Cph1 and P. aeruginosa BphP (2,3,33). Loss of this salt bridge by removal of the PHY domain also increases the excited state life time and therefore fluorescence in truncated phytochromes (13). However, we have shown here that the addition of the three mutations that render the DrCBD monomeric allows Pfr formation (Fig. 4). Dimerization, at least in the context of the truncated DrCBD, must inhibit photoconversion. This inhibition is overcome in the native structure by the PHY domain, which must channel chromophore conformational changes along the long interdomain helix to the histidine kinase domain. The inhibi-  (33)). D207H contributes to fluorescence through loss of the Pfr D-ring interaction and stabilization of the Pr A-ring conformation. Y263F has a greater influence due to destabilization of the Pfr D-ring conformation. tion is also relieved by the engineered monomerization of the DrCBD, which removes the constraint of helix packing at the dimer interface and allows more degrees of freedom.
The use and manipulation of phytochromes as fluorophores will likely flourish as more phytochrome structures are reported, additional non-canonical phytochromes are discovered, and researchers advance novel fluorescence imaging techniques. Combining Pr-stabilizing influences (such as found in P3 or apparently in IFP1.4) with Y263F would make an interesting study. Although the P2-Y258F variant remains capable of photoconversion (12) and has not been described as fluorescent, the recently reported fluorophore iRFP is P2 with 13 amino acid substitutions, including Y258F (31). Additional engineering to augment quantum yield or to nudge the excitation maxima farther into the infrared would be welcome developments. Studies like these will not only further fluorophore research but also address the distinction between photochemical and signaling properties of phytochrome.