Communicated by Winslow R. Briggs, Carnegie Institution of Washington, Stanford, CA, July 11, 2008 (received for review February 26, 2008)
Phytochromes are red-light photoreceptors that regulate light responses in plants, fungi, and bacteria via reversible photoconversion between red (Pr) and far-red (Pfr) light-absorbing states. Here we report the crystal structure at 2.9 Å resolution of a bacteriophytochrome from Pseudomonas aeruginosa with an intact, fully photoactive photosensory core domain in its dark-adapted Pfr state. This structure reveals how unusual interdomain interactions, including a knot and an “arm” structure near the chromophore site, bring together the PAS (Per-ARNT-Sim), GAF (cGMP phosphodiesterase/adenyl cyclase/FhlA), and PHY (phytochrome) domains to achieve Pr/Pfr photoconversion. The PAS, GAF, and PHY domains have topologic elements in common and may have a single evolutionary origin. We identify key interactions that stabilize the chromophore in the Pfr state and provide structural and mutational evidence to support the essential role of the PHY domain in efficient Pr/Pfr photoconversion. We also identify a pair of conserved residues that may undergo concerted conformational changes during photoconversion. Modeling of the full-length bacteriophytochrome structure, including its output histidine kinase domain, suggests how local structural changes originating in the photosensory domain modulate interactions between long, cross-domain signaling helices at the dimer interface and are transmitted to the spatially distant effector domain, thereby regulating its histidine kinase activity.
Light is a major environmental stimulus for both prokaryotes and eukaryotes. The phytochrome superfamily contains a diverse set of red light photoreceptors, members of which regulate a wide range of physiologic processes, such as seed germination, floral induction, and phototaxis in plants, fungi, and bacteria. Plant phytochromes and bacteriophytochromes (Bph) use a linear tetrapyrrole (bilin) as a chromophore and photoconvert between red-absorbing (Pr) and far-red-absorbing (Pfr) states (1, 2). Like many naturally occurring signaling molecules, Bphs possess a modular domain architecture in which 3 N-terminal domains, denoted PAS (Per-ARNT-Sim), GAF (cGMP phosphodiesterase/adenyl cyclase/FhlA), and PHY (phytochrome), form the photosensory core domain (PCD). They are covalently linked to the C-terminal histidine kinase (HK) output or effector domain that transduces a light signal into a chemical signal via autophosphorylation of a histidine residue in the HK domain. The PAS and GAF domains together form the chromophore binding domain (CBD), responsible for chromophore incorporation. Recent crystallographic studies of the CBDs from Deinococcus radiodurans DrBphP and Rhodopseudomonas palustris RpBphP3 reveal the structures of the PAS and GAF domains and their biliverdin-IXα (BV) chromophore in the Pr state (3–5). However, as previously shown in many other phytochromes (6–9), these shorter CBD constructs lack the PHY domain and hence do not undergo full photoconversion between the Pr and Pfr states. The photoconversion mechanism, involving rapid 15Z anti to 15E anti isomerization of the C15 = C16 double bond between rings C and D of the bilin chromophore, has been extensively explored by static and time-resolved spectroscopic studies on various phytochromes and Bphs (10–15). However, the molecular basis of reversible Pr/Pfr photoconversion and signal transduction mechanisms in phytochromes remains unclear, largely owing to the absence of structural information on fully photoactive constructs containing the PHY and HK domains, and on constructs in the Pfr state.
We have determined the crystal structure of a fully photoactive photosensory core domain from Pseudomonas aeruginosa bacteriophytochrome (PaBphP-PCD) in its dark-adapted Pfr state (Fig. 1A). This construct consists of the PAS, GAF, and PHY domains and includes residues 1–497 together with 2 linker residues and 6 additional histidine tag residues at the C terminus. The dark-adapted Pfr state is confirmed by the UV-visible absorption spectra in solution (top trace of Fig. 2D) and in crystals, which is consistent with spectra of Agp1 and Agp2 reconstituted with a synthetic bilin chromophore sterically locked in the 15Ea configuration (14, 16). The PaBphP-PCD crystals exhibited similar spectral behavior at room temperature: photoconversion between the Pr and Pfr states and dark reversion to the Pfr state (data not shown).
Crystal structure of wild-type PaBphP-PCD. (A) Ribbon diagram of the dimeric PaBphP-PCD structure. The PAS, GAF, and PHY domains of one monomer are highlighted in yellow, green, and blue, respectively. Helices in the GAF and PHY domains are identified by letters (A–E). (B) The PAS, GAF, and PHY domains are integrated via extensive interdomain interactions and converge on the chromophore binding site (cyan). (C) Accessory structure elements (gray) decorate the common cores of the PAS, GAF, and PHY domains and are spatially clustered near the chromophore (cyan) and as helical bundles at the dimer interface. (D) The core of the PAS, GAF, and PHY domains contains an antiparallel β sheet with strands in the spatial order of 2–1–5–4–3 and a variable connector between strands 2 and 3 that contains helix C. (E) Both core and accessory elements are highlighted in a topologic diagram of the PaBphP-PCD structure.
Residues and interactions in the chromophore binding pocket. (A) The BV chromophore of PaBphP-PCD in the Pfr state adopts the 15Ea configuration (cyan) as compared with the 15Za configuration (gray) in the Pr state of the RpBphP3-CBD structure. The superposition is based on the structural alignment of the PAS and GAF domains of the PaBphP-PCD and RpBphP3-CBD (PDB ID 2OOL) structures. (B) At the interface between the GAF and PHY domains in PaBphP-PCD, conserved residues Tyr-250, Asp-194, Ser-459, and Arg-453 and the chromophore form extensive hydrogen bond interactions (red dotted lines) to stabilize ring D in the 15Ea configuration. (C) The side chains of Tyr-250 and Gln-188 directly interact with the carbonyl group of ring D. (D) UV-visible absorption spectra of the dark-adapted (solid) and light-illuminated (dash) states in PaBphP-PCD wild type (WT) and mutants. Half-times of dark reversion are indicated in parentheses.
We solved 2 crystal structures in 2 different space groups. The crystal structure of a single mutant Q188L was determined by multiple wavelength anomalous dispersion in space group P65 and was then used as a search model to solve the crystal structure of the wild type in space group C2221 by molecular replacement (Materials and Methods). Diffraction from wild-type crystals was anisotropic, extending to 2.7 Å resolution along the c axis and ≈3.3 Å along the a and b axes [supporting information (SI) Table S1]. The PaBphP-PCD wild-type structure contains 8 molecules, each containing 1 BV chromophore and 16 water molecules in the asymmetric unit. The final model was refined at 2.9 Å resolution with an R-factor of 0.221 and free R-factor of 0.268. In this report, we focus on the PaBphP-PCD wild-type structure; the PaBphP-PCD Q188L mutant structure will be discussed in detail separately (X.Y., J.K., and K.M., unpublished data).
The 8 PaBphP-PCD molecules in the asymmetric unit comprise 4 parallel, head-to-head dimers (Fig. 1A and Fig. S1). The tertiary structures of the PAS and GAF domains and their disposition in the dimers are similar to those in the RpBphP-CBD and DrBphP-CBD structures (3–5). The core of the novel PHY domain forms a typical α/β fold in which a central 5-stranded, antiparallel β sheet has a single helix packed on one side and a helical bundle on the other (Fig. 1A). It bears a significant resemblance to the core of the GAF domain, which confirms a prediction made in the PFAM database for the Cph2 family (17). Both the GAF and PHY domains contain a large segment (approximately 44 residues) inserted in their core fold containing 1 helix (D) in an extended loop but at different topologic locations (Fig. 1E). The PAS, GAF, and PHY domains make extensive interdomain contacts via these large inserts (Fig. 1 B and C). The PAS domain penetrates the GAF domain to form a knot in which its N-terminal extension passes through the large insert in the GAF domain (3–5). The large insert in the PHY domain forms an extended arm that shields the chromophore hosted by the GAF domain and constitutes 90% of the buried surface area between the GAF and PHY domains.
The cores of the PAS, GAF, and PHY domains share a common topology: each has an antiparallel β sheet whose 5 strands occur in the spatial order 2–1–5–4–3 with various connectors of different lengths, but all include helix C between strands 2 and 3 (Fig. 1 C and D).Accessory structural elements decorate these common cores and, although scattered in the primary structure, are spatially localized in 2 regions. The first is located near the chromophore binding pocket, where the N-terminal extension of the PAS domain and the arm of the PHY domain interact with the large insert of the GAF domain, and conserved residues from all 3 domains interact directly with the chromophore (discussed below). The second region comprises helical bundles at the dimer interface, in which a single long helix directly connects the GAF and PHY domains. The similar topologies in the PAS, GAF, and PHY domains are associated with only a low level of identity in primary structure (PAS-GAF and PAS-PHY: ≈25%; GAF-PHY: ≈10%; PAS-GAF-PHY: <5%) (18, 19). We propose that the PAS, GAF, and PHY domains originated in a precursor to the PAS sensory module and became specialized in phytochromes via accessory structural elements that confer Pr/Pfr photoconversion and facilitate dimerization.
Conformational changes in the chromophore binding pocket and structural variation of the helical bundle at the dimer interface. (A) In the chromophore binding pocket of PaBphP-PCD, Tyr-190 and Tyr-163 (green) near ring D adopt side chain conformations distinct from the corresponding residues Phe-212 and Tyr-185 (light gray) of RpBphP3-CBD in the Pr state. As ring D flips, the 2 side chains rearrange. (B) Superposition of the 8 monomers (4 shown in green, 4 in dark gray) in the asymmetric unit shows 2 distinct locations for the GAF-hA helix, resulting in 4 tertiary heterodimers. (C) Quaternary structural variation at the dimer interface, as shown by angles between the GAF-hE helices of one monomer when dimeric structures are aligned according to the PAS and GAF domains of the other monomer. Structures compared are DrBphP-CBD (1ZTU, pink; 2O9C, gray), RpBphP3-CBD (2OOL, yellow), PaBphP-PCD-WT (3C2W, cyan), and PaBphP-PCD-Q188L (G.X. Yang, J.K., and K.M., unpublished results; blue).
The BV chromophore is covalently attached to Sγ of Cys-12 in the PAS domain via the vinyl group of ring A. The electron density in the simulated-annealing omit map establishes the 5-syn/10-syn/15-anti configuration for the BV chromophore (Fig. 2A and Fig. S2). Limited by anisotropic diffraction and moderate resolution (Table S1), the electron density alone could not unambiguously distinguish between the 15Ea and 15Za configurations of ring D. However, hydrogen bonding interactions between ring D and the surrounding residues clearly favor the 15Ea over the 15Za configuration (Fig. 2B). This interpretation of the 5Zs/10Zs/15Ea configuration is consistent with the assignment of the 15Ea and 15Za configurations to the Pfr and Pr states, respectively, but does not support the 5Za configuration of ring A proposed for the Pfr state (14).
On the basis of our RpBphP3-CBD structure, we predicted that a surface patch in the GAF domain consisting of conserved residues Asp-216 from the characteristic phytochrome PASDIP sequence, Tyr-272, and the 15Ea pocket of RpBphP3 forms part of the interface with the PHY domain (5). This prediction is now confirmed by the PaBphP-PCD structure. At the interface of the GAF and PHY domains, there are extensive interactions among conserved residues Asp-194 and Tyr-250 of the GAF domain [corresponding to Asp-216 and Tyr-272 of RpBphP3 (Fig. S5)], Ser-459 and Arg-453 from the arm of the PHY domain, and the chromophore. Specifically, Asp-194, Tyr-250, and Ser-459 interact directly with the pyrrole nitrogen of ring D of the chromophore in the 15Ea configuration (Fig. 2B). The carboxyl group of Asp-194 is sandwiched between ring D and the side chains of Ser-459 and Arg-453. The PHY arm is largely diverse in sequence and rich in glycine and proline but contains a stretch of conserved residues spanning residues 453–463. Within this stretch, Ser-459 is located in helix D, and Arg-453 protrudes from a sharp turn immediately N-terminal to helix D. To explore the roles of these residues in stabilizing the Pfr state, we subjected them to site-specific mutagenesis. Single alanine substitutions at any of Asp-194, Ser-459, and Arg-453 disrupted the Pfr dark state of wild type and caused each to adopt the Pr dark state (Fig. 2D). Remarkably, the D194A mutant undergoes no photoconversion from the Pr state, is completely incapable of forming the Pfr state, and is intensely fluorescent (Fig. S3) (20). The S459A and R453A mutants display limited light-induced photoconversion from the Pr to the Pfr state. Consistently, a shorter construct containing only the PAS and GAF domains of PaBphP adopts Pr instead of Pfr as the dark-adapted state (data not shown). These results establish that interactions among Asp-194, Arg-453, Ser-459, and the chromophore at the interface of the GAF and PHY domains are critical for engaging the PHY domain and forming the Pfr state.
Residues Tyr-250, Gln-188, and Ser-459 are within hydrogen bonding distance of the carbonyl substituent of ring D (Fig. 2C). Both the Y250F and Q188L mutants retain the Pfr ground state and are able to undergo light-induced photoconversion to the Pr state, but their rates of spontaneous reversion to the Pfr state are significantly slower than wild type (Fig. 2D). The amino group of Gln-188 apparently contributes to stabilizing the 15Ea configuration. It is noteworthy that those Bphs whose dark-adapted state is Pfr typically contain Gln (in PaBphP, RpBphP5, and AtBphP2) (5, 21–24) or Asn (in BrBphP and RpBphP1) (25) at the position corresponding to Gln-188 (see Fig. 2c of ref. 5).
Understanding the signal transduction mechanism in bacteriophytochromes raises several questions. What conformational changes occur upon prompt isomerization of the chromophore? How do these, initially local, structural changes propagate through the photosensory domains to the spatially remote effector domain, where they regulate its HK activity? The pair of aromatic residues Tyr-190 and Tyr-163 surrounding ring D in the PaBphP-PCD structure adopts side chain conformations that differ substantially from their counterparts in other structures (3–5). The side chain of the Tyr-190 is deeply buried underneath ring D in the PaBphP-PCD structure (Fig. 3 A), but the side chain of the equivalent Phe-212 in RpBphP3-CBD lies on the surface of the GAF domain and shields ring D from solvent (5). In the RpBphP3-CBD structure the side chain of Tyr-185 occupies the same cavity, and its equivalent Tyr-163 in the PaBphP-PCD structure adopts a new side chain conformation, where its hydroxyl group interacts directly with the propionate group of ring C (Fig. 3A). Sequence alignment shows that all phytochromes contain aromatic residues at both positions: either Tyr or Phe at the position of Tyr-190 but only Tyr at the position of Tyr-163. Similar displacements of aromatic side chains around ring D have also been observed in the crystal structures of α-phycoerythrocyanin with its phycoviolobilin chromophore in the Z- and E- configurations (26). In Cph1, the equivalent Y176H mutant is unable to undergo photoconversion and is intensely fluorescent, with an emission maximum of approximately 650–670 nm (8). Saturation mutagenesis on Tyr-176 of Cph1 indicated that its hydroxyl group is important for efficient Pr to Pfr photoconversion (27). We propose that Tyr-163 and Tyr-190 undergo concerted side chain rotamer changes upon light-induced isomerization, to facilitate a snug fit of ring D in its cavity and thereby stabilize the 15Za or 15Ea configurations.
How could structural changes originating in isomerization of the BV chromophore be coupled to the HK domain? In the PaBphP-PCD structure, the C-terminal GAF-hE helix is collinear with the N-terminal helix A in the PHY domain (PHY-hA); together they form a continuous helix of 12 turns spanning residues 286–331 that directly connects the GAF and PHY domains. Such long helices with ≈40 residues are widely found in signaling proteins, where they connect modular domains (28). Structural perturbation, either specifically via these long “signaling helices” or more generally (29), has been proposed as a general mechanism by which helices can transmit a signal from one domain to another (28). In the PaBphP-PCD structure, the long signaling helix forms a substantial portion of the parallel dimer interface. Furthermore, the PHY-hA helix displays an alternating pattern of glutamic acid and arginine residues, and oppositely charged Glu and Arg residues interact across the dimer interface (Fig. S4).
Comparison of the crystal structures of Bphs reveals tertiary and quaternary plasticity of the helical bundles at the dimer interface found in all structures. When the 8 monomers in the asymmetric unit of the wild-type PaBphP-PCD structure are superposed, significant differences in location of the GAF-hA helix are observed between the 2 monomers in each dimer (Fig. 3B). That is, the chemically identical monomers exhibit 2 distinct tertiary structures in the GAF-hA helix and form a structural heterodimer. A heterodimeric structure may result from “induced fit” at the dimer interface to allow tighter dimeric association and eliminate steric clashes. When we model the 2 structural homodimers formed by each tertiary structure, they display either significantly less buried surface area or severe steric clashes at the dimer interface (Fig. S1D). In addition, when dimeric Bphs structures are aligned according to the PAS and GAF domains of one monomer, there is wide variation in the relative orientation of the other monomer as shown by angles between the GAF-hE helices (Fig. 3C). We propose that tertiary and quaternary plasticity of extended helical bundles at the dimer interface plays an important role in transmitting signals over long distances.
Using the PaBphP-PCD structure as the dimeric scaffold, we modeled a full-length PaBphP structure on the basis of sequence and structural alignment between the C terminus of PaBphP-PCD and the N terminus of a homologous sensor HK structure (30, 31). Secondary structure predictions strongly indicate a long continuous helix in the linker region between the PHY and HK domains. In our model, the PHY-hE helix is directly fused to and collinear with the N-terminal helix of the HK domain in which the phospho-acceptor histidine residue is located (Fig. 4), thereby further extending the central helical bundle at the dimer interface into the HK domain. The side chain of the phospho-acceptor His in one monomer is orientated toward the ATP binding site of the kinase domain from the other monomer, consistent with the geometry of in trans phosphorylation proposed for phytochromes (32, 33). In addition, the radius of gyration (Rg ≈46 Å) and maximum particle diameter (Dmax ≈152 Å) calculated from this model [CRYSOL (34)] is consistent with those dimensions derived from the experimental SAXS data on full-length R. palustris bacteriophytochrome RpBphP2 in the Pr state (Rg ≈52 Å and Dmax ≈155 Å) that suggested a structure with strongly anisotropic axial dimensions (35).
A domain architecture model of the full-length dimeric PaBphP based on the PaBphP-PCD dimer structure and the sensor HK structure (PDB accession ID 2C2A) (PAS, GAF, and PHY in green; HK in blue; BV in cyan).
We propose that signals generated by light-induced 15Za/15Ea isomerization from the photosensory domain are transmitted to the extended central helical bundle at the dimer interface, where they effect tertiary and quaternary structural changes that position the phospho-acceptor His in the catalytic site of the HK domain and thus regulate phosphorylation in trans of bacteriophytochromes.
The full-length PaBphP gene (PaBphP-FL, residues 1–728) vector was constructed from the PCR-amplified gene product of P. aeruginosa strain PA01 genomic DNA (American Type Culture Collection) using primers (5′- CAGCCATATGACGAGCATCAC CCCGGTTAC and 5′ CATCCTCGAGTCAGGACGAGGAGCCGGTCTC [Integrated DNA Technologies]), which was then ligated into expression vector pET28a (Novagen) via restriction sites NdeI/XhoI. The photosensory core domain coding region (residues 1–497) of PaBphP (PaBphP-PCD) was PCR-amplified from the pET28a vector containing the PaBphP-FL gene and inserted into the expression vector pET24a (Novagen) using the restriction sites NdeI/XhoI. PaBphP-PCD was coexpressed with heme oxygenase (pET11a carrying the R. palustris hmuO gene, kindly provided by Dr. Carl Bauer of Indiana University) in BL21(DE3). Se-Met proteins were prepared using the methionine synthesis inhibition protocol (36). The purification procedure for both wild-type and mutants proteins was as described (5). Site-directed mutagenesis was carried out using the QuikChange Site-Directed Mutagenesis Kit (Stratagene).
UV-visible spectra of purified wild-type and mutant PaBphP-PCD proteins in solution were recorded at room temperature from 900 to 230 nm with a Shimadzu UV-1650 PC spectrophotometer. Spectra were recorded either in the dark-adapted state or after illumination with fiber optic light at 750 nm (far red) or 690 nm (red) provided by interference filters with a 10-nm bandwidth (Andover). Visible spectra of the PaBphP-PCD crystals were recorded at room temperature with a microspectrophotometer Xspectra (4DX-ray Systems).
Crystals of both the PaBphP-PCD single mutant Q188L and wild type were obtained at 20°C in the dark using the hanging drop vapor diffusion method but under very different crystallization conditions. The wild-type crystals were crystallized with 10 mg/ml protein, 0.45 M ammonium sulfate in 0.1 M Tris·HCl buffer, pH 7.7; and the Q188L crystals with 10 mg/ml protein, 0.5% PEG4000 (wt/vol), and 0.01 M sodium acetate, pH 4.6. Microspectroscopic experiments showed that both crystal forms are photoactive: they undergo reversible Pr/Pfr photoconversion at room temperature and revert to the Pfr state in the dark (data not shown). To minimize light exposure, crystallization trays were wrapped with Al foil and kept in the dark during crystal growth. All observations and handling of crystals were carried out under double-filtered light using a combination of 1 green and 1 blue broad band-pass filter to permit only transmission between 450 and 500 nm, a wavelength range in which PaBphP has minimal absorption (Fig. 2D). In solution studies we have shown that such an illumination protocol with doubled-filtered light can effectively limit light excitations in the PaBphP samples. Dark-adapted crystals were then cryoprotected using 40% glycerol in the mother liquor. Our preliminary microspectroscopic experiments showed no detectable photoactivity at 100 K. All diffraction data used in this study were collected at 100 K at the SBC 19-ID, SBC 19-BM, and BioCARS 14BM-C beam stations of the Advanced Photon Source, Argonne National Laboratory. All images were indexed, integrated, and scaled using HKL2000 or HKL3000 (37).
The PaBphP-PCD Q188L mutant crystals are in space group P65, with 2 molecules in the asymmetric unit that are related by a near-perfect noncrystallographic 2-fold symmetry axis perpendicular to the principal 6-fold crystallographic axis. The crystal structure of the Q188L mutant was determined by the multiwavelength anomalous diffraction method with Solve (38) and Sharp (39) at 2.9 Å resolution and was refined with CNS/Refmac5/PHENIX (40–42) using native data at 2.75 Å resolution. The wild-type PaBphP-PCD crystals are in space group C2221, with 8 molecules in the asymmetric unit packed as 4 dimers. The PaBphP-PCD crystals with a typical size of 300 × 100 × 100 μm diffract to a maximum resolution at 2.7 Å along the long axis c but only to ≈3.3 Å along a/b axes. A partially refined Q188L structure was used as a search model to determine the wild-type structure by the molecular replacement method using Phaser (43), followed by structural refinement at 2.9 Å resolution using PHENIX. Eight molecules in the asymmetric unit were restrained by noncrystallographic symmetry in early refinement cycles; and were independently refined and rebuilt at later refinement stages. Data collection, phasing, and refinement statistics are summarized in Table S1. Coot (44) was used for model building and structural alignment. Structures were illustrated using PyMOL (http://pymol.org).
We thank Ying Pigli and Yuen-Ling Chan for help and advice in cloning and mutagenesis; Yu-Sheng Chen and Vukica Šrajer of BioCARS for assistance in microspectroscopic experiments on crystals; and the staff of the Structural Biology Center and BioCARS at the Advanced Photon Source, Argonne National Laboratory, for beam line access. Supported by National Institutes of Health Grant GM036452 (to K.M.).
Author contributions: X.Y. designed research; X.Y. and J.K. performed research; X.Y. and J.K. contributed new reagents/analytic tools; X.Y. analyzed data; and X.Y. and K.M. wrote the paper.
The authors declare no conflict of interest.
Data deposition: Coordinates and structure factor amplitudes have been deposited in the Protein Data Bank, www.pdb.org (PDB ID code 3C2W).
This article contains supporting information online at www.pnas.org/cgi/content/full/0806718105/DCSupplemental.
