Expression of PsBphP1 as a full-length dimer assembled with biliverdin (BV)Full-length P. syringae BphP1 (Fig. 1a) assembled covalently with BV and modified with an N-terminal 6His tag followed by a tobacco etch virus (TEV) protease cleavage sequence, was synthesized by simultaneous expression of the tagged polypeptide with the Synechocystis heme oxygenase (HO)−1 in Escherichia coli BL21(DE3) cultures supplemented with 5-aminolevulinic acid to boost BV synthesis19,20,28. Purification and removal of the tag by TEV protease yielded homogeneous preparations of photointerconvertible bili-protein as judged by SDS-PAGE and UV-vis absorption spectroscopy (Supplementary Fig. 1a, b). The final PsBphP1 samples used here differed from wild type only by inclusion of a single N-terminal glycine remnant from the TEV protease recognition site.Fig. 1: High resolution 3D structure of the PSM from the P. syringae (Ps) BphP1 dimer as Pr.a Domain organization of PsBphP1. Shown are the positions of BV, the PAS, GAF, and PHY domains, and the NTE, knot lasso (KL) and hairpin (HP) features, portions of the S-helix linker, and the DHp and CA domains within the HK bidomain of the OPM. The numbers delineate domain boundaries within the polypeptide. Cys16, which forms a thioether linkage with BV is identified by the red circle. His530, which provides the phosphoacceptor site in the DHp, is indicated by the arrow. The length of the NTE was extended for clarity. b Relative positions of BV as Pr in the PsBphP1 dimer (PDB 8U4X) with that determined previously by X-ray crystallography of a PSM fragment from D. radiodurans (Dr) BphP (PDB 4Q0J29). The nitrogen, oxygen, and sulfur atoms are in blue, red, and yellow, respectively. The A-D pyrrole rings are labeled. The 32 carbon that forms the thioether linkage with the apoprotein and the 182 vinyl carbon are labeled for reference. c Cryo-EM model of the BV-binding pocket from the A protomer (sticks) of PsBphP1 superposed with the EM map (gray mesh). BV and the NTE, GAF domain, and PHY hairpin carbon atoms are in cyan, black, green, and orange, respectively. d Surface-rendered 2.8-Å resolution view of the 3D EM map of the dimeric PsBphP1 PSM. EM density was rendered at 5 σ. EM density within 3 Å of a modeled atom was color-coded by domain with NTE, PAS, GAF, PHY, S helix, and BV colored in black, blue, green, yellow/orange, teal/light cyan, and red, respectively; those at greater distances were rendered white. Positions of the helical spines (HS), HP, NTE, and KL features are highlighted. Dimensions of the dimeric PSM are indicated where the width excludes NTE and hairpin contributions. e Orthogonal cartoon views of the dimeric PSM model of PsBphP1 as Pr. Features are colored as in panel (d). BV is shown in red sticks. C, C-terminus.Initial characterizations by cryo-EM revealed that PsBphP1 preparations as Pr or Pfr tend to form aggregates in buffered solutions containing NaCl at various concentrations even in the presence of assorted detergents, additives, pH values, and/or cryo-EM grid treatments, which obfuscated high-resolution structural studies. Fortunately, a screen of various buffers revealed that moderate concentrations of KSCN produced monodispersed PsBphP1 samples. Analysis of preparations in 175 mM KSCN indicated normal Pr/Pfr absorption, an indistinguishable spectral change ratio for the Pr/Pfr difference spectra, and near equivalent Pfr→Pr thermal reversion rates (Supplementary Fig. 1b, c). Measures of autokinase activity using [δ-32P]-ATP19 showed faster Pfr-dependent kinetics with PsBphP1 samples in KSCN than those in NaCl, indicating that KSCN increases the activity of the output module (Supplementary Fig. 1d). Negative stain EM also showed a similar ensemble of 2D particle images compared to samples dissolved in solutions containing NaCl (Supplementary Fig. 2). From these results, we found that PsBphP1 retains strong HK activity and near native photocycles in the presence of KSCN, thus leading us to conclude that KSCN at the concentrations used did not appreciably disrupt the native 3D structure of PsBphP1 dimers and thus was suitable for cryo-EM.Cryo-EM models of PsBphP1 as PrFrom PsBphP1 preparations dissolved in a HEPES-NaOH (pH 7.5) buffer containing 175 mM KSCN, we generated cryo-EM maps of Pr and Pfr using either dark-adapted samples containing solely Pr or samples irradiated to steady state levels of Pfr (~80% of total Phy pool) with red light immediately before freezing. As shown by the work flows of both samples, we collected a set of highly resolved 2D class averages for both Pr and Pfr which enabled construction of informative 3D maps (Supplementary Figs. 3–5). From 3,233,515 picked particles, 1,151,696 were sufficiently similar to produce high quality 2D class averages for Pr. The combined averages then generated an EM density map of the PSM at 2.9-Å resolution which was further improved to 2.8-Å resolution through symmetry expansion and local refinement of 1,666,098 matching protomers (Supplementary Fig. 3b; Supplementary Table 1). The resulting data allowed construction of a head-to-head dimeric model encompassing residues 12-459 and 466-516 of protomer A and residues 11-459 and 466-516 of protomer B (Fig. 1d, e; Supplementary Table 1). While this 3D map provided high resolution for the PSMs, the HK bidomains and much of the S-helices were poorly visible probably due to high mobility.The resulting high-resolution Pr model conformed well with prior PSM structures of BphPs11,12,14,15,22,29,30,31 (Fig. 1d, e). Besides the covalent connection, the PAS and GAF domains were tightly linked non-covalently via hydrophobic interactions within the figure-of-eight knot created by a lasso loop extending from the GAF domain to tether the N-terminal extension (NTE) projecting from the PAS domain, which was further solidified by a noncovalent contact involving Ile27 at the knot center. BV was covalently linked via a thioether linkage to Cys16 within the NTE. As expected from other homodimeric Pr models, BV in both protomers assumed a ZZZssa configuration (Fig. 1b), and was fixed within the GAF domain pocket by an extensive network of conserved amino acid contacts to all four pyrrole rings and the B- and C-ring propionates (Fig. 1c; Supplementary Fig. 6a). Notable examples were the side-chain positions of: (i) Asp203 within the DIP motif and Tyr259 which move as a pair in hairpin binding as Pr and Pfr; (ii) His256 and the main chain carbonyl of Asp203 that help fix the pyrrole water at the center of the A-C pyrrole rings; (iii) Tyr172 and Phe199 near the D pyrrole ring that reorient during photoconversion; and (iv) Arg250 and Arg218 that bind the B-ring propionates as Pr and Pfr, respectively12,13,14,22,23,31.While most of these contacts are conserved among bacterial, cyanobacterial, and plant Phys1, one notable exception in PsBphP1 was Gly286. This residue is generally a histidine in canonical Phys, which hydrogen bonds with the D-ring carbonyl through its imidazole group as Pr but then switches to provide an important anchor for the C-ring propionate in Pfr after the chromophore isomerizes14,32,33 (Supplementary Fig. 6a). This substitution probably explains the slight positional difference of the bilin in the binding pocket by comparison to its D. radiodurans ortholog DrBphP, with the absence of this histidine/propionate connection also possibly explaining the increased rate of Pfr→Pr thermal reversion for PsBphP1 compared to that of DrBphP (see Supplementary Fig. 1c and ref. 14).The GAF and PHY domains in PsBphP1 as Pr were linked covalently via the intervening helical spine and noncovalently through the hairpin extension from the PHY domain contacting the GAF domain near the chromophore (Fig. 1d). Its antiparallel structure was clearly β-stranded in both protomers as expected for Pr14,15,22,25,26 (Supplementary Fig. 7a), and makes prominent contacts with the GAF domain using WGG and PRxSF motifs within the hairpin (WSG and PRTSF in PsBphP1), and the DIP motif and an adjacent α-helix at residues 254-263 within the GAF domain (Fig. 1c; Supplementary Figs. 6a and 7a). As it exits the PAS/GAF bidomain, the NTE also assumed an α-helical fold that contacts the hairpin.Improved resolution of the HK bidomain was then enabled by 3D variability analysis (3DVA), which isolated particle images with better congruity (Fig. 2; Supplementary Fig. 3c). These refinements culminated in a 3.3-Å resolution EM map that partially illuminated the HK bidomains, including an outline of the CA domain positions, and extended modeling of the sister DHp domain helices to Asn533 and Arg534 in protomers A and B, respectively, to now include the predicted phosphoacceptor histidines (His530) (Fig. 2a). Although the EM density for helix α2 was readily apparent, its signals became too diffuse for exact modeling shortly after residues 533/534, thus muddling the connecting turn between DHp helices α1 and α2 so that the amino acid registry was lost. Fortunately, map density for helix α2 then reemerged for 17–20 additional residues along with a rough outline of the mobile CA domains, which were placed on opposite sides of the DHp four-helix bundle and near the connecting turn between DHp helices α1 and α2 (Fig. 2c, d).Fig. 2: Maps of P. syringae (Ps) BphP1 as Pr shown in low contour inform the relative positions of the S-helices and HK bidomains in the dimer.a Orthogonal cartoon views of full-length PsBphP1 as Pr (PDB 8U8Z) modeled from a 3.5-Å resolution map that allowed extension of the structure into the S-helix/DHp regions connecting the PSMs to the HK bidomains. Color scheme is the same as in Fig. 1d, with the purple coloring added for the DHp domains. His530 predicted to participate in the HK phosphorelay is shown as red spheres. b An unsharpened EM density map of full-length PsBphP1 contoured at 2 σ reveals a more complete dimer. c Positioning of the protomers as cartoon models in the map from (b) showing the relative positions of the PSM and HK regions. The position of the loop between the Hα1 and Hα2 regions of the DHp domains was extended from those in (a) using the predicted AlphaFold structure of PsBphP1 (AF-Q885D3-F1). In the absence of sufficiently resolved CA domains, we docked a predicted cartoon model (shown in magenta) of the CA domain developed in RoseTTA-fold34. ATP-binding pocket in the CA domains is outlined by the dashed black ovals. d Orthogonal cartoon views of a predicted model for the DHp domains superposed with an unsharpened map of PsBphP1 as Pr (gray mesh) contoured at 4 σ. The S-helix is colored in teal/light cyan and the DHp domains in purple. His530 is shown as red spheres. The CA domains were omitted for clarity.To better appreciate the relationship of the DHp and CA domains, especially with respect to the ATP-binding pocket in the CA domains and the phosphoacceptor histidine (His530) in the DHp domains, we generated a hypothetical structure based on the unsharpened volume map (Fig. 2b, c). Here, we overlaid the S-helix and DHp domain specified by AlphaFold for PsBphP1 (https://alphafold.ebi.ac.uk/entry/Q885D3) onto our 3.3-Å resolution model, which allowed us to exploit this predicted model to position the connecting turn between DHp helices α1 and α2. Helix α2 placement was then adjusted manually into its EM density while retaining the relative amino acid registry of the two helices. Subsequently, we modeled the DHp four-helix bundle with left-handed helix α1 to α2 transitions and extended helices α2 as polyalanines to the end of the observable EM density for the DHp domains (Fig. 2d). Although the handedness of the turn was too ambiguous to model solely by the cryo-EM map, data described below for the Pfr state unambiguously identified the turn as left-handed. While not completely resolved, the short length of the DHp-CA domain linker and the resting positions of the CA domains along-side the DHp domains allowed predictions of the DHp(helix α2)-CA domain connectivity. Here, the CA domains associated non-covalently with the DHp domains via helix α1 of the DHp domain from its own protomer and helix α2 of the DHp domain from its sister protomer (Fig. 2c).To help place the CA domain structure in the lower resolution model, we employed RoseTTAFold34 to generate its predicted model ab initio. This domain model showed strong congruity with other CA or CA-like domains found in the PDB database as analyzed by the DALI Protein Structure Comparison Server35. Notably, one of the strongest matches was the 3D model for the ADP-bound CA domain determined empirically by X-ray crystallography from the Thermotoga maritima HK853 transmitter histidine kinase (PDB 4JAV36) (Supplementary Fig. 8). As can be seen in Fig. 2c, the EM densities for both CA domains within PsBphP1 as Pr, though weak, mimicked the overall shape of the predicted CA domain model when docked manually, with their expected placements near the DHp-CA domain transitions (Fig. 2c). Inspection of this assembly led us to conclude that the ATP γ-phosphate upon binding within the ATP-binding pockets as Pr would sit near the helix α1-α2 transitions at the top of the DHp domains, and thus reside ~30 Å away from the phosphoacceptor histidines (His530) (Fig. 2c; Supplementary Fig. 8). Given this distance, we predict that these positions would sequester ATP-binding and autophosphorylation from each other in PsBphP1 as Pr, and thus dampen the kinase activity for this conformer as seen in vitro19,20,28.Cryo-EM models of PsBphP1 as PfrWe then defined by cryo-EM a 3D model of the full-length PsBphP1 dimer as Pfr by analyzing red light-irradiated samples. Here, 4,796,622 particle images were extracted and winnowed down to 1,029,141 after 2D classification, which revealed three similar but distinct EM density maps representing unique structural conformations of the photoactivated photoreceptor (Supplementary Figs. 4 and 5; Supplementary Table 1). Two maps resolved to 3.1- and 3.0-Å were dimeric head-to-head arrangements of the PSM with only short S-helix extensions visible beyond the PHY domains (Supplementary Fig. 9d–g). The PAS-GAF regions were intimately connected in the dimer like those in Pr while the paired PHY domains were either moderately or strongly splayed (designated “medial” and “splayed”, respectively). Both the BV and hairpin conformations matched previous views of Phys as Pfr14,15,22,25. Much like the high-resolution Pr EM density map, these two Pfr maps lacked the HK bidomains and much of the S-helices presumably due to high mobility.The third Pfr map resolved to 3.3 Å was most informative and revealed a surprising tetrameric arrangement whereby two full-length PsBphP1 dimers abutted each other through their HK bidomains thus creating an elongated “dimer-of-dimers” (DoD) with the PSMs located at opposite poles (Fig. 3a, b; Supplementary Fig. 5). This arrangement presumably limited the flexibility of the dimers to enable further definition of the S-helices to Ala526 in protomer A and Ala517 in protomer B just proximal to the phosphoacceptor His530 (Fig. 3c, d). This stabilized structure also permitted construction by focused refinement of a PSM EM density map to 3.0-Å resolution (Fig. 3g; Supplementary Table 1), and permitted reasonably accurate modeling of the DHp and CA domains for one of the protomers (Supplementary Fig. 10). Interestingly, detailed SEC analysis of PsBphP1 in the KSCN buffer revealed that this DoD configuration arises from a dynamic equilibrium between dimeric and tetrameric forms. As shown in Supplementary Fig. 11, increasing concentrations of full-length chromoprotein as Pfr progressively assembled tetramers from dimers. For example, the mole fraction of tetramers was 0.52 at 0.117 mg mL−1 and increased to 0.84 at 1.17 mg mL−1.Fig. 3: 3D structure of the full-length PsBphP1 dimer as Pfr based on the dimer-of-dimer (DoD) maps.a 2D class averages generated from a collection of cryo-EM images showing the tetrameric assembly of PsBphP1 into DoDs. b Orthogonal surface-rendered views of a 3D map reconstruction illustrating DoD assembly. C2 symmetry was applied to the map prior to symmetry expansion and local refinement (see Supplementary Fig. 5). c Orthogonal surface-rendered views of a 3D EM map of PsBphP1 dimers as Pfr (PDB 8U62) that was generated by focused refinements of one of the PSMs and the DoD assembly point at the HK region (see b). The EM density was colored and labeled as described in Figs. 1d and 2a. d Orthogonal 3.3-Å resolution cartoon views of a dimer model deconstructed from the DoDs. Domains are colored as in (c). e BV positions after superposition of the GAF domains as Pfr (PDB 8U63) or as Pr (PDB 8U4X). The 32 carbon that participates in the thioether linkage with the apoprotein and the 182 vinyl carbon are labeled for reference. Pyrrole ring D rotates ~161° after Pr→Pfr conversion. f Cryo-EM model of the BV-binding pocket as Pfr (PDB-8U63). The model was derived from the A protomer (sticks) and superposed with the EM map (gray mesh). BV and the NTE, GAF domain, and PHY hairpin carbon atoms are in cyan, black, green, and orange, respectively. g Surface-rendered view of the 3D EM map of the dimeric PsBphP1 PSM as Pfr. The EM density was colored and labeled as in Fig. 1d. Dimensions of the dimeric PSM are indicated where the width excludes NTE and hairpin contributions. h Orthogonal 3.0-Å resolution cartoon views of the dimeric PsBphP1 model as Pfr (PDB-8U63) generated with cryo-EM views encompassing just the PSM (see Supplementary Fig. 8c–f). The various features are colored as in (g). BV is shown in red sticks. Positions of the helical spines (HS), HP, NTE, and KL features are highlighted.Importantly, the PSM regions of all three maps (DoD, medial, and splayed) allowed unambiguous modeling of both protomers as Pfr as judged by sufficiently resolved ZZEssa conformations of the bilins that aligned well for all three maps except for slightly different tilts of the D rings (Supplementary Fig. 9c) and by the presence of obvious α-helical hairpins, which are both signatures of this spectral state (Fig. 3e, f; Supplementary Figs. 7a and 9d–g). Comparison of BV at high-resolution for the Pr model to that of the DoD Pfr model indicated that the cis-to-trans isomerization of the C15 = C16 methine bridge during photoconversion generated a 161° planar flip of the BV D ring relative to the GAF domain (Fig. 3e). BV also slid in the binding pocket to allow formation of a hydrogen bond between the D-ring pyrrole nitrogen and the Asp203 carboxylate group (Supplementary Fig. 6b). As mentioned above, PsBphP1 harbors a glycine at residue 286 rather than the highly conserved histidine. As comparable His-to-Gly substitutions largely precludes photoconversion to Pfr in other Phys by removing a C-ring propionate contact in Pfr32,33, we speculate that hydrogen bonding between Ser284 and the C-ring propionate compensates (Supplementary Fig. 6b). Related to this unusual connection, a serine is often found at this position in bathyPhys that use Pfr as the dark-adapted ground state25,31, and has been shown to stabilize the Pfr state indefinitely when introduced into the PSM of the plant PhyB isoform from Arabidopsis thaliana19,32. Another notable change in the GAF domain connections with the B-ring propionate was the use of Arg218 rather than Arg250 to anchor the propionate carboxylate group through a salt bridge (Supplementary Fig. 6b). In a similar vein, the hairpin modified its interactions after BV moved to its final Pfr position with Ser472 in the hairpin PRxSF motif hydrogen bonding to Asp203 and Try259 in the GAF domain (Supplementary Fig. 6a, b).As with the Pr data, resolution of the HK bidomains was possible by pointed scrutiny of the DoD map. Here, the HK regions were analyzed in isolation to produce a 4.1-Å map with adequate features to define the overall shape of both DHp domains and the position of one of the two CA domains (Fig. 4; Supplementary Figs. 5 and 10). EM density for the second CA domain was too diffuse to detect, suggesting it was either not bound to the DHp domains or that its position in the tetramer was displaced by the sister dimer. These maps were then merged with the higher resolution DoD full-length map to produce a composite map of 4.4-Å, and from this map a full-length hypothetical model of one of the dimers from the DoD complex (Fig. 4). Here, modeling was conducted in the same manner as described above for Pr. In this case, we used the AlphaFold model of the DHp domain as a guide to manually extend its fit within the EM density. Placement of the observed CA domain into the density was then possible by manual rotation and translation of the entire CA domain predicted by RoseTTAFold, which oriented the CA domain near the bottom of the DHp and adjacent to the phosphoacceptor histidine.Fig. 4: Possible model of the PsBphP1 dimer as Pfr based on cryo-EM maps of the DoD.Shown are orthogonal views of the PsBphP1 dimer either alone or superposed with a cryo-EM composite map of the region derived from the full-length DoD map and a focused refinement map of the DoD contacts at the HK bidomain (see Supplementary Figs. 5 and 10). The dark gray surface delineates one dimer, while EM density for the opposing dimer was rendered as a white surface. Domains and features are colored as in Fig. 2. His530 is shown in red spheres. ATP-binding pocket of the CA domain in the A protomer is outlined with a dashed black circle. The dashed red circles locate the CA domain from the B protomer, which was added in an arbitrary orientation for completeness, but is absent in the EM density map. As seen in Supplementary Fig. 10, EM density dissipates at the terminus of helix α2 of the DHp domain of protomer B.Based on positioning of ADP within the CA domain of the T. maritima HK853 kinase bound with ADP (PDB 4JAV36), this orientation would place the ATP β-phosphate to within 7 Å of the phosphoacceptor histidine – His530 (See Fig. 4 and Supplementary Figs. 7 and 10). Consequently, we propose that photoconversion to Pfr dissociates both CA domains from their most favored positions in Pr to either become more mobile, or to relocate close to the bottom of the DHp domains. Either scenario would increase the local concentrations of histidine phosphoacceptors in the DHp domains and the catalytic ATP-binding sites in the CA domains to presumably accelerate ATP to histidine phosphotransfer.Possible mechanism for Pr→Pfr photoconversionClose inspection of an alignment of the Pr and Pfr models along with prior studies on other Phys13,14,15,23,25,26,27,37,38, then enabled a possible detailed mechanism for PsBphP1 photoconversion that eventually activates the autokinase activity of its HK bidomains (Fig. 5). After the light-induced Za to Ea flip of the D-pyrrole ring in BV, the PSM undergoes a cascade of conformational changes involving a number of features. As described above and from prior structures14,15,16,17,18, the primary event caused by this flip is release and subsequent refolding of the hairpin from β-stranded to α-helical followed by rebinding of the hairpin to the GAF domain. These interactions are mostly hydrophobic through contacts involving residues 467-480 of the hairpin, which includes the entire α-helix, and the face of the GAF domain adjacent to the BV D-ring. Although hydrogen bonding is less represented, the conserved bonds between the PRxSF motif Ser472 of the hairpin and Asp203 and Tyr259 of the GAF domain were observed. The hairpin transition was also coincident with rotation of the NTE helix by ~100°, which is stabilized in part by a salt bridge between Arg466 of the hairpin and Glu10 of the NTE. This bridge implies a potential role for the NTE in Pfr stabilization (Supplementary Figs. 6a and 7a).Fig. 5: Detailed conformational changes and a model for Pr→Pfr photoconversion of full-length PsBphP1.a Straightening the GAF-PHY helical spine kink centered at Thr326 during photoconversion shown by orthogonal views of the Pr (blue) and Pfr (DoD) (yellow) cartoon models. b Rotation of the PHY domain enforced by the β-stranded to α-helical transition of the hairpin shown by orthogonal views of the Pr (blue) and Pfr(DoD) (yellow) cartoon models superposed via the GAF domains. c Light-induced pivot of the helical spine exiting the GAF domain around Val-307. GAF domains of protomer A were superposed for Pr, Pfr (DoD), Pfr (medial), and Pfr (splayed) models of PsBphP1. (left) Cartoon models positioning the helical spine in protomer A. (right) Cartoon models highlighting the positions of the helical spine in protomers B after superposition of the GAF domain of protomers A. The positions of Ala72 highlight the range of motion for the PAS-GAF domain. d Orthogonal cartoon views of the dimer as Pr. Domains/features are colored as in Fig. 1d. e Orthogonal cartoon views of Pfr illustrating movements within the PsBphP1 dimer upon photoconversion. Include are: (i) tugging of the hairpin closer to the GAF domain as the hairpin converts from an anti-parallel β-strand to α-helical, (ii) rotation of the PHY domains with coincident straightening of the helical spine, (iii) pivoting of the GAF domains around V307; (iv) movement of the GAF-PHY helical spines closer to each other (red arrow), and (v) scissoring of the sister S-helices. f Close-up view of the proposed positions of the CA domain as Pr and Pfr showing downward movement and rotation of the ATP-binding pocket closer to His530 (vi). The Pr and Pfr images were derived from the protomer A models described in Figs. 2c and 4. Dashed orange ovals identify the ATP-binding pocket.As shown in Fig. 5b, d, e, the resulting hairpin reconfiguration has two apparent consequences. One is to “pull” the distal exit point for the hairpin closer to the GAF domain (Supplementary Fig. 7a). As an example, the Cα atoms of Gly177 and Ile483 were separated by 15.5 Å in Pr but only 9.5 Å in Pfr. Superposition either of an ensemble of 51 PHY domain 3D models currently available within the PDB database (Supplementary Fig. 12a, b), or only the four PSM structures described here for PsBphP1 (i.e., one Pr and three Pfr models) (Supplementary Fig. 12c, d), revealed that the core PHY domain structure is remarkably rigid with little average structural displacement even among a wide range of Phy relatives with divergent sequences, photostates, and/or crystallization contacts. In fact, PHY domain flexibility was seen only at its entrance and exit points. Thus, any strain imposed on the PHY domains from the hairpin transitions should pass through the PHY domain core and impinge directly on the connecting GAF-PHY helical spine and PHY-OPM S-helices. The cumulative effects are to straighten the helical spine kink found in Pr following by torque on the S-helices, with the end result being a 35° rotation of the whole PHY domain (Fig. 5a, b, d, e).The second consequence of the hairpin reconfiguration is to swivel of the GAF domain protomers relative to each other through a pivot point centered at Val307 (Fig. 5c). Our speculation is that this pivot buffers large motions at the PHY domain from dissociating the dimeric interface at the GAF domains, so that conformational energy can instead be directed toward the HK bidomains. Consequently, the PAS domains, whose function(s) have remained enigmatic, might provide a backstop to limit pivoting, while also supplying interdomain contacts to support PSM dimerization.Finally, rotation of the PHY domains pulls outward the terminal α-helix of the PHY domain, which is continuous with the S-helix and helix α1 of the DHp, while straightening inward the GAF-PHY helical spine. As shown in Fig. 5d, e, such movements imbue a scissor motion to this α-helical feature in both protomers which converts the crossover arrangement of the paired S-helices in Pr to the roughly parallel or splayed arrangements in Pfr. As seen from a top view, the S-helix of protomer B in the DoD model undergoes a reasonably rigid and straight transition for the Pfr endstate, while the S-helix of protomer A straightens in Pfr and rotates around the S-helix in protomer B to allow intimate contact between the S-helices and DHp α1 helices (Fig. 5d, e).From our partially resolved HK bidomains (Figs. 2c and 4), we hypothesize for PsBphP1 and related BphPs with transmitter kinase activity (at least those with accelerated activity as Pfr) that this rearrangement translates up into the DHp domains to dislodge the CA domains from their Pr positions, thus allowing their ATP-binding pockets to interact more intimately with the phosphoacceptor His530 residue in the DHp domains as Pfr (Fig. 5f). This proximity should then enhance PsBphP1 autophosphorylation as seen in vitro19,20,28, and possibly improve access to the bound phosphate by the aspartate acceptor in its paired response regulator. The formation of DoDs is potentially telling about the mechanism of autophosphorylation. At concentrations where tetramers form (see Supplementary Fig. 10), the local concentrations of CA and DHp domains from a single dimer should be significantly higher. Consequently, if the CA domains had moderate affinity for the DHp domains at the histidine phosphoacceptor position, the CA domains would easily outcompete neighboring dimers for DHp binding, which is not what was found. It is apparent that a key aspect of the autokinase mechanism is that the CA and DHp domains only transiently contact each other as Pfr, which has special relevance to the results below.
PsBphP1 interacts transiently with its response regulator AlgBThe substantial mobility of the HK bidomains seen with PsBphP1, as well as within other microbial Phys23,24 and transmitter HK bidomains more generally21, led us and others to speculate whether its corresponding response regulator could stabilize this region in a ternary complex and thus might help full structural understanding of microbial Phys. In fact, ref. 23. recently attempted a short cut of this approach through cryo-EM analysis of full-length D. radiodurans (Dr) BphP ectopically fused to its dimeric phosphatase partner DrBphR28,39 via a short C-terminal linker. Here the PSM and parts of the S-helices and DHp region were well resolved, but the CA domains attached to DrBphR were not, as also seen here for our high-resolution cryo-EM maps of PsBphP1.To avoid the inherent challenges that translational fusions present, we attempted to assemble a native PsBphP1/response regulator complex without covalent coupling. At the time of this study, the phosphorylation target(s) of activated PsBphP1 were not known even though it is a robust autokinase as Pfr19,20. No genetic links were available and its operon does not include possible candidates besides the HO for BV synthesis28, unlike other bacterial Phys such as D. radiorurans BphP that include a response regulator in its photoreceptor operon28,39. Fortunately, a survey of other Pseudomonads by Mukherjee et al10. identified AlgB as likely candidate, which was originally discovered in P. aeruginosa as a NtrC-type response regulator whose phosphorylation by an orthologous BphP forms a central node integrating light to quorum sensing and biofilm behavior10. They also noticed that this BphP/AlgB signaling pair is widely distributed among proteobacteria (>150 species), including an ortholog in P. syringae (84% sequence identity for PsAlgB; GenBank: MCF9017822.1).To confirm that P. syringae AlgB works with PsBphP1, we assayed for phosphotransfer using [δ-32P]-ATP. Here, full-length PsAlgB with its phosphoacceptor receiver (REC) domain followed by an AAA-ATPase domain (Fig. 6a), was expressed recombinantly with an N-terminal TEV-protease cleavable 6His tag, which was removed during purification to yield PsAlgB containing a single N-terminal glycine remnant upstream of the initiator methionine (Fig. 6a). As shown in Fig. 6b, c, 2 μM of PsBphP1 preloaded with phosphate as Pfr, readily transferred the phosphate to PsAlgB (1 μM) with an apparent rate constant of 2.8( ± 0.3 SE)10−1 s−1 at 22 °C. The reaction was essentially completed within 10 min while the initial Pfr-specific loading of PsBphP1 required ~1 h to saturate (see Supplementary Fig. 1d; ref. 19). As expected10, alanine substitution mutants showed that the initial autophosphorylation of PsBphP1 required the predicted internal phosphoacceptor His530, while subsequent phosphotransfer to PsAlgB required the predicted phosphoreceiver Asp59 in PsAlgB (Fig. 6d). In the latter case, the Asp59-Ala mutant of PsAlgB failed to accept 32P even in the presence of highly labeled, wild-type PsBphP1.Fig. 6: The P. syringae AlgB works downstream of PsBphP1 in phosphotransfer but without a tight interaction.a Domain architecture of PsAlgB. Asp59, predicted phosphoacceptor site by PsBphP110. REC, phosphoacceptor receiver domain. AAA, ATPase associated with various activities. HTH, helix-turn-helix DNA-binding motif. b, c Phosphotransfer kinetics from PsBphP1 to PsAlgB. PsBphP1 was prelabelled as Pfr with [γ−32P]-ATP for 2 h and at t = 0 incubated further with PsAlgB. Samples were subjected to SDS-PAGE and autoradiography (b) and quantified for 32P transfer by densitometric scans of the autoradiograms (c). Data in (c) represents three separate reactions. d Phosphotransfer from PsBphP1 to PsAlgB works via a canonical transmitter HK mechanism. Wild-type PsBphP1 or its His530-Ala mutant were incubated for 2 h as Pfr with [γ−32P]-ATP, and then mixed with a twofold molar excess of either wild type PsAlgB or its Asp59-Ala mutant. Reactions were subjected to SDS-PAGE and autoradiography, followed by staining for protein with Coomassie Blue. PsBphP1 and PsAlgB are indicated by the arrowheads. e–h SEC chromatograms of PsBphP1 as Pr or Pfr with or without PsAlgB and/or 1.5 mM ATP. PsBphP1 at the concentration used (3 mg mL−1) assembles as a dimer as Pr but mostly as a DoD as Pfr (see Supplementary Fig. 11). PsAlgB was added at a concentration of 3 mg mL−1. e Profiles for PsBphP1 as Pfr and PsAlgB either alone or mixed without ATP. f Profiles of PsBphP1 as Pr or Pfr either alone or mixed with PsAlgB and ATP. g Profiles of PsBphP1 as Pr or Pfr alone with ATP. The elution profiles of ATP and ADP are included for comparison. h Profiles of PsBphP1 as Pr or Pfr in the presence of both PsAlgB and ATP. Elution profiles of ATP and ADP alone, or ATP mixed with PsAlgB alone are included for comparison. A closeup of the elution region for ATP is included in (f) and (h).We then tested for assembly of PsAlgB/PsBphP1 complexes by SEC, in this case using the BphR response regulator/phosphatase from D. radiodurans as a control. Unlike PsAlgB, DrBphR is composed on just the REC domain (Supplementary Fig. 13a) but was previously found to be a phosphorylated by PsBphP1 at least in vitro28. As shown in Fig. 6e, PsBphP1 tested at a concentration that would mostly assemble DoDs as Pfr (see Supplementary Fig. 11c) but remain dimeric as Pr, failed to form a stable complex with PsAlgB either as Pr or Pfr as assayed by the lack of an SEC elution shift for either component (Fig. 6e). DrBphR also did not show a mobility shift by SEC when mixed with either of the two PsBphP1 conformers (Supplementary Fig. 13b).Unexpectedly, when PsBphP1 and PsAlgB were mixed in the presence of 1.5 mM ATP, detectable binding was also absent but robust hydrolysis of ATP was discovered, which was seen as a shift in the elution position of ATP to that of ADP. This shift was barely detected when either PsBphP1 or PsAlgB were incubated alone with ATP (Fig. 6f), but occurred when mixing PsAlgB with either the Pr or Pfr conformers of PsBphP1 (Fig. 6g, h), indicating that the two proteins together elicited robust ATPase activity for PsAlgB. For a control, neither the Pr nor Pfr states of PsBphP1 showed this ATP hydrolysis when DrBphR was added instead (Supplementary Fig. 13b). We presumed that the strong and rapid ATPase activity stimulated by PsAlgB was caused by an intrinsic ATPase activity in PsAlgB given that both PsBphP1 and PsAlgB remained phosphorylated for hours when incubated together in the phosphotransfer assays (Fig. 6b). This activity might be a common feature of NtrC-type response regulators harboring an AAA-ATPase domain (Fig. 6a). In sum, it appears that PsBphP1 and PsAlgB work together in a transmitter kinase phosphorelay but through a transient interaction. Taken further, this transience implies that the internal interactions between the DHp and CA domains must be equally ephemeral at least after phosphorylation to ensure that PsAlgB can bind the phosphohistidine intermediate.
Signaling by a bacterial phytochrome histidine kinase involves a conformational cascade reorganizing the dimeric photoreceptor
