Temperature-Scan Cryocrystallography Reveals Reaction Intermediates in Bacteriophytochrome

Abstract

Light is a fundamental signal that regulates important physiological processes such as development and circadian rhythm in living organisms. Phytochromes form a major family of photoreceptors responsible for red light perception in plants, fungi and bacteria¹. They undergo reversible photoconversion between red-absorbing (Pr) and far-red-absorbing (Pfr) states, thereby ultimately converting a light signal into a distinct biological signal that mediates subsequent cellular responses². Several structures of microbial phytochromes have been determined in their dark-adapted Pr or Pfr states^3-7. However, the structural nature of initial photochemical events has not been characterized by crystallography. Here we report the crystal structures of three intermediates in the photoreaction of Pseudomonas aeruginosa bacteriophytochrome (PaBphP). We employed cryo-trapping crystallography to capture intermediates, and followed structural changes by scanning the temperature at which the photoreaction proceeds. Light-induced conformational changes in PaBphP originate in ring D of the bilin chromophore, and E to Z isomerization about the C₁₅=C₁₆ double bond between rings C and D is the initial photochemical event. As the chromophore relaxes, the twist of the C₁₅ methine bridge about its two dihedral angles is reversed. Structural changes extend further to rings B and A, and to the surrounding protein regions. These data suggest that absorption of a photon by the Pfr state of PaBphP converts a light signal into a structural signal via twisting and untwisting of the methine bridges in the linear tetrapyrrole within the confined protein cavity.

Cryo-trapping and time-resolved room-temperature experiments are two main experimental strategies to study the structures of intrinsically short-lived reaction intermediates⁸. To establish the molecular mechanism of Pfr/Pr photoconversion, we generated and cryotrapped intermediates between the reactant Pfr state and product state(s) in fully photoactive crystals of the photosensory core module (PCM)⁵ of PaBphP. We followed progress of the reaction by applying a “trap-pump-trap-probe” strategy at variable pump temperatures (Fig. 1, Fig. S1, Methods Summary). Here, temperature mimics time: the higher the pump temperature, the greater the structural relaxation and the further a reaction proceeds along its pathway.

Fig. 1.

Trap-pump-trap-probe experiment. (a) Ribbon diagram of the PaBphP-PCM dimer. The BV chromophore is colored in cyan. (b) Experimental difference (F_light-F_dark) map at 130K (contoured at +/-5σ, where σ is the standard deviation of difference densities across the entire map). Strong positive (green) and negative (red) densities with peak signal greater than +/-12σ are clustered near the chromophores of the eight monomers (A-H) in the asymmetric unit. (c) Dark stripes of a mounted crystal correspond to segments from which X-ray datasets were collected. (d) A ball-and-stick representation of the chromophore in the Pfr state (PDB 3NHQ). The α-face of a pyrrole ring is defined when atom numbering follows a clockwise direction, with β defined as the opposite face. The α-face of an entire bilin chromophore is defined according to Rockwell et al.³⁰.

We collected diffraction data from six crystals at 10 pump temperatures between 100 and 180K (Table S1), and calculated difference (F_light-F_dark) electron density maps from 14 light datasets and six reference dark datasets (Fig. S1a). In all maps, strong and highly significant difference electron densities are exclusively concentrated at the chromophore binding sites embedded in the GAF domain (Fig. 1b). Singular value decomposition (SVD) analysis⁹ of difference densities within a 5Å radius of the aligned chromophores revealed three major, independent, light-induced structures (Fig. S1c-d). Since difference densities are largely consistent among the eight monomers in the asymmetric unit and between different crystals illuminated at the same temperature, we averaged these densities by applying non-crystallographic symmetry (NCS), and focus here on the principal features common to the eight monomers.

Difference densities vary markedly with pump temperature (Fig. 2). At the lowest temperatures they appear near ring D, which suggests that light-induced structural changes originate in ring D. As the temperature rises, difference densities expand to ring C and eventually to ring B and ring A, thus exhibiting a systematic structural progression as a function of temperature. Based on representative difference maps at 110, 130 and 173K, we modeled three light-induced chromophore structures, denoted L1, L2 and L3, respectively (Fig. 3a,b). We further refined the initial models of the L1, L2 and L3 structures jointly in real space against the NCS-averaged, SVD-filtered difference maps at all pump temperatures, and determined their relative concentrations at each temperature (Fig. 3g).

Fig. 2.

Difference maps (F_light-F_dark) at pump temperatures between 100 and 180K (contoured at +/- 4σ). The BV chromophore (cyan) is shown in the Pfr state. Positive (green) and negative (red) densities represent structural changes associated with formation of photoproduct state(s) and loss of the parent Pfr state, respectively. No significant difference densities with signal greater than +/-4σ are detected at longer range, beyond the 5Å radius around the chromophore.

Fig. 3.

Light-induced structural changes. (a) Representative difference (F_light-F_dark) maps at 110, 130 and 173K. Arrows indicate viewpoints of Fig. 3c and 3d. (b) Stereoview of the superposition of the chromophore conformations in the Pfr (cyan), L1(magenta), L2(yellow) and L3 (blue) structures. (c) Sideview of difference map at 130K (contoured at +/-3σ) shows a twist in the C₁₅ methine bridge. (d) Sideview of difference map at 173K (contoured at +/-2σ) indicates a β–facial shift of rings B/A. Arrows in Figs. 3c, 3d mark the α-face of the chromophore. (e) Difference densities at 173K (contoured at +/-2σ) near the PXSDIP motif (residues 191-196) at the GAF-PHY interface. (f) Differences densities associated with the side chains of His277, Tyr163 and Tyr190 (map contours: +/-3σ at 173K). (g) Relative concentrations of the L1, L2 and L3 structures as a function of pump temperature.

In the L1 structure represented at 110K, strong positive and negative densities are roughly aligned in the plane of ring D of the bilin chromophore (Fig. 3a). Negative densities span the pyrrole nitrogen of ring D and the side chain of a highly conserved Asp194 from the PXSDIP sequence motif, indicating the rupture of a key hydrogen bond that stabilizes ring D in the Pfr state⁵. The corresponding positive densities between His277 and ring D identify formation of the L1 structure in which the chromophore has isomerized about the C₁₅=C₁₆ double bond to adopt the 15Za configuration, and ring D is significantly shifted in its plane towards His277. As a result, L1 exhibits a smaller and strained C₁₄-C₁₅-C₁₆ bond angle in the methine bridge between rings C and D compared to the “stretched” bond angle in the Pfr state (Fig. S3) and to those of the Pr state in D. radiodurans BphP and R. palustris BphP3^{4, 10}.

In the L2 structure represented at 130K, strong difference densities near ring C suggest that ring C moves towards its α-face, and that its propionate side chain breaks the hydrogen bonds with His277, Tyr163 and Ser275 present in the Pfr and L1 structures to form a new hydrogen bond with Ser261 (Fig. 3a, Fig. 4). Difference densities associated with rings C and D (Fig. 3c) suggest that counter-twist occurs across the C₁₅ methine bridge in forming L2, in which ring D assumes a β-facial disposition relative to ring C. This partially relaxes strain in the C₁₅ methine bridge while the C₁₀ methine bridge between rings B and C is concomitantly twisted in the opposite direction (Fig. S5b).

Fig. 4.

Light-induced molecular events in PaBphP. (Upper panel) The chromophores are in ball-and-stick representation, with their surrounding residues shown as van der Waals spheres. Green dotted lines indicate potential interactions with each cryo-trapped structure. (Lower panel) Schematic representation of changes in relative disposition of the four pyrrole rings of the BV chromophore, in which pyrrole rings A, B, C and D (boxes) are linearly connected by methine bridges (circles). The α- and β-faces of the chromophore are denoted by arrows.

In the L3 structure represented at 173K, structural changes extend to rings A and B (Fig. 3a). Difference densities sandwiching these rings (Fig. 3d) suggest that they move as a unit slightly towards the β-face of the chromophore while the C₅ methine bridge retains its direction of twist. Motion is accompanied by small twists in the C₅ and C₁₀ methine bridges (Fig. S4b). Rings B and C are more coplanar in L3 than in L2, consistent with a more relaxed C₁₀ methine bridge in L3.

These data demonstrate that E to Z isomerization about the C₁₅=C₁₆ double bond between rings C and D is the initial structural event in the Pfr to Pr photoreaction of PaBphP. This is consistent with findings in many phytochromes and BphPs^{2, 11, 12}; but contrasts with a recent report based on room temperature NMR spectroscopy, in which rotation of ring A occurs between the Pr and Pfr states of an unusual, knotless phytochrome consisting of only the GAF domain¹³. In PaBphP, when ring D flips into the 15Za configuration within a protein cavity that is still optimized to accommodate the 15Ea configuration of the Pfr state, a highly distorted and strained C₁₅ methine bridge is generated in L1. As L1 evolves to L2, counter-twist across the C₁₅ methine bridge partially relieves the strain. Motion of ring C results in a C₁₀ methine bridge twisted in the opposite direction in L2, further relaxation of which leads to the L3 structure. These data suggest a reaction trajectory that proceeds in the order Pfr →L1→ L2→L3 (Fig. 4).

To relate these cryo-trapped L1, L2 and L3 structures to spectroscopic intermediates in the Pfr to Pr photoreaction^14-18, we measured visible absorption spectra on a crushed PaBphP-PCM crystal at temperatures between 100 and 180K using the X-ray experimental protocol (Fig. S5). Difference absorption spectra between illuminated and reference “dark” states show significant loss of the Pfr state, indicated by a negative peak at 768nm, and formation of blue-shifted photoproducts (Fig.S5b). SVD analysis revealed two significant basis difference absorption spectra, with blue-shifted peaks between 650 and 700nm. The basis spectrum with a blue-shifted peak at 684nm is largely populated between 100 and 150K, in agreement with the temperature dependence of the population of the L2 structure in X-ray data (Fig. S5c, Fig. 3g). Blue-shifted absorption peaks are consistent with the distorted tetrapyrrole conjugated system observed in L2 and L3.

The L1, L2 and L3 structures likely arise from early molecular events in the Pfr to Pr reaction of PaBphP. One or more may correspond to the Lumi-F spectroscopic intermediate detected on the femto- to pico-second time scale by room temperature time-resolved spectroscopy and characterized by blue-shifted absorption peaks^{18,14, 19}. These structures are consistent with strong FTIR difference bands attributed to the carbonyl group of ring D resulting from 15Ea/15Za isomerization and the B-C methine stretching in cryotrapped Lumi-F of CphA¹⁸. L1 may be identified with early intermediates in Cph1, in which the C₁₅-H out-of-plane (HOOP) mode was detected by ultrafast Raman spectroscopy²⁰. The properties of L1 are also consistent with the NMR spectroscopic data on the cryotrapped Lumi-F of Cph1, which indicated that the C₁₄-C₁₅-C₁₆ angle is distorted following 15Ea/15Za isomerization¹⁹. Furthermore, our temperature-scanning range (100-180K) coincides with the temperature range in which Lumi-F intermediates were trapped in several phytochrome systems ^{11, 18, 19}. Formation of late Meta-F intermediates would require higher pump temperatures ¹⁹ to permit more extensive structural relaxation in the protein. However, at temperatures >180K the cryoprotectant solution undergoes a glass transition which causes severe deterioration in crystal diffraction ^5,6.

In both the L2 and L3 structures, the twist of the dihedral angles in the C₁₅ methine bridge is quite distinct from those of the chromophore structures in the Pr and Pfr states (Figs. 3, 4 and S4b). Both the Pr and Pfr crystal structures of BphPs ^{3-5, 10} exhibit an α-facial disposition of ring D relative to ring C, which corresponds to a negative rotation of the red absorbance band in the circular dichroism (CD) spectra of BV-containing bacteriophytochromes such as PaBphP, DrBphP and Agp1^{21, 22}. Evidence for such counter-twist events in the methine bridges during photoconversion has been presented in Cph1 and Agp1^{19, 22}. We propose that the chromophore of PaBphP generates structural signals in response to light via subtle and local twist or counter-twist motions in the methine bridges within the confined protein cavity (Fig. 4), which alter specific interactions between the pyrrole rings and their immediate protein surroundings. It remains to be seen to what extent our findings in PaBphP apply to other members of the diverse and expanding phytochrome superfamily.

The protein moiety also plays an important role. First, following prompt E/Z isomerization, steric clashes between ring D and the side chain of Asp194 lead to initial strain in L1 that drives subsequent relaxation events. Second, the side chains of His277, Tyr163 and Tyr190 surrounding ring D also move slightly but concertedly to accommodate ring D as the chromophore evolves from Pfr to the L3 structure (Fig. 3f). Third, Ser261 appears to stabilize L2 and L3 via hydrogen bonds to the propionate group of ring C (Fig. 3a, 4). The single point mutant S261A inhibits formation of the Pr state upon illumination and significantly accelerates dark reversion²³ (Fig. S6b). This provides further evidence that L2 and L3 are authentic photoproducts on the productive trajectory towards the product Pr state. Fourth, more extensive structural changes occur in L3 near the highly conserved PXSDIP sequence motif at the interface of the GAF and PHY domains. This segment and the side chain of Tyr190 are pushed away from the chromophore as rings B/A and ring D move towards the β-face of the chromophore (Fig. 3e). These movements expand the chromophore cavity to allow further relaxation, and may trigger further structural perturbations between the GAF domain and the arm of the PHY domain. As the reaction proceeds further, Tyr163 and Tyr190 may adjust their side chain rotamers to reshape the ring D pocket to accommodate the product Pr state²³.

To examine light dependence of the HK activity in PaBphP, we conducted HK assays on wild type and the S261A mutant of full-length PaBphP under light and dark conditions (Fig. S6) and found that autophosphorylation of PaBphP is light-dependent, in contrast to earlier observations²⁴. As in RpBphP2²⁵ and Cph1²⁶, PaBphP exhibits higher levels of autophosphorylation in the Pr state (here, the light state) than in the Pfr state. S261A exhibits a photoconversion phenotype that greatly prefers the Pfr state, and shows significantly reduced HK activity compared to WT. We predict that small light-induced structural changes, exemplified by the cryo-trapped intermediate structures L1, L2 and L3, propagate from the chromophore binding pocket of the N-terminal photosensory domains to the C-terminal HK domain via further tertiary and/or quaternary rearrangements²⁷, which ultimately affect the HK activity and convert a light signal into a biological signal.

Methods Summary

Structural heterogeneity is intrinsic to all dynamic processes, and often challenges accurate interpretation of both cryo-trapping experiments at a single temperature and time-resolved crystallographic data ^{28, 29}. This work presents a temperature-scan and analytical strategy to resolve structural heterogeneity and determine distinct, homogeneous structural species.

We apply a “trap-pump-trap-probe” strategy to photoactive crystals of PaBphPPCM. We first trap the dark-adapted Pfr state (trap dark) by freezing the crystal (grown at 293K in the dark) in liquid nitrogen under safety green/blue light, and collected a reference crystallographic “dark” dataset at 100K for each crystal. The same crystal was then uniformly illuminated under white light for 10-15 minutes at elevated temperatures to generate reaction intermediates (pump), then cryo-cooled to 100K (trap light) to collect a “light” dataset from a fresh crystal volume (probe) (Fig. 1b). We calculated 14 (F_light-F_dark) difference maps from six crystals that cover 10 pump temperatures between 100 and 180K. Since each map contains eight monomers in the asymmetric unit, we obtained 112 independent difference maps for a PaBphP monomer. We spatially aligned these maps on a reference monomer C, and subjected difference densities within a 5Å radius of the aligned chromophores to singular value decomposition (SVD)⁹ that yielded 112 noise-reduced difference maps (Fig. S1c-d). Atomic models for the L1, L2 and L3 structures were initially built based on representative difference densities at 110, 130 and 173K, respectively. These models were further refined jointly in real space (together with a fixed reference Pfr conformation) against SVD-filtered difference maps at all pump temperatures. Based on least squares fitting between the calculated and observed difference densities, we also obtained relative concentrations of L1, L2 and L3 at each temperature. Fig. S1a summarizes methods and software used in X-ray data reduction and analysis.

Methods

Purification, mutagenesis and crystallization

PaBphP-PCM containing residues 1-497 of P. aeruginosa PA4117 was expressed, purified, crystallized and cryo-protected in the dark as described⁵. Site-directed mutagenesis was carried out using the QuikChange Site-directed Mutagenesis Kit (Stratagene). Both wild type and the S261A mutant of the full-length PaBphP were purified using the same protocol for PaBphP-PCM.

Data collection and processing

Diffraction data were collected at LSCAT 21-IDG beamline, BioCARS 14-BMC and 14-IDB beamlines of Advanced Photon Source. Temperature was controlled by a cryostream cooler (Oxford Cryosystems). Illumination was applied to a mounted crystal using two unfiltered fiber optic lights from different directions for 10-15min, while the crystal spinned around the spindle axis of the goniometer at a speed of 30°/s. All diffraction images were processed using HKL2000³¹. Light and dark data sets from the same crystal were scaled using ScaleIt in CCP4³² (Table S1). A reference “dark” structure in the Pfr state was refined at 2.55Å resolution using Phenix³³ (Table S2), and was used to calculate phases in generating difference Fourier maps by FFT in CCP4. Difference maps were masked and aligned using NCSMASK and MAPROT of CCP4, and were further subjected to SVD⁹ analysis in DynamiX (Fig. S1). Real space refinement of the L1, L2 and L3 structures, including the chromophore and several adjacent residues (Gln188-Asp194, His277, Ser261, Tyr163 and the covalent chromophore anchor Cys12), was carried out against NCS-averaged, SVD-filtered difference maps using DynamiX (Table S3). All modeling building was carried out using Coot³⁴. Structure figures were generated using PyMol (http://pymol.org)

UV-visible spectroscopy

Absorption spectra of proteins in solution were recorded at room temperature with Shimadzu UV-1650 PC spectrophotometer. Visible spectra of crystals were measured using a microspectrophotometer Xspectra (4DX-ray Systems) at BioCARS.

Histidine kinase autophosphorylation assay

The HK assay was modified based on published protocols^{26, 35}. Reactions under light (continuous illumination with unfiltered fiber optical light) or dark (samples were protected from light in a covered box during dark incubation and reaction; handled under dim room light) were started by adding 10μl reaction buffer (50mM Tris HCl, pH 8.0; 2mM MgCl₂; 2mM MnCl₂; 0.1M KCl; 10% ethylene glycol; 80μM ATP and ³²P-γ-ATP at 0.3mCi/ml) to each 10μl protein sample at the concentration of 1mg/ml. Reactions were stopped using standard stop buffer containing 0.1M EDTA and 0.1M DTT.