Arm-in-Arm Response Regulator Dimers Promote Intermolecular Signal Transduction

ABSTRACT

Bacteriophytochrome photoreceptors (BphPs) and their cognate response regulators make up two-component signal transduction systems which direct bacteria to mount phenotypic responses to changes in environmental light quality. Most of these systems utilize single-domain response regulators to transduce signals through unknown pathways and mechanisms. Here we describe the photocycle and autophosphorylation kinetics of RtBphP1, a red light-regulated histidine kinase from the desert bacterium Ramlibacter tataouinensis. RtBphP1 undergoes red to far-red photoconversion with rapid thermal reversion to the dark state. RtBphP1 is autophosphorylated in the dark; this activity is inhibited under red light. The RtBphP1 cognate response regulator, the R. tataouinensis bacteriophytochrome response regulator (RtBRR), and a homolog, AtBRR from Agrobacterium tumefaciens, crystallize unexpectedly as arm-in-arm dimers, reliant on a conserved hydrophobic motif, hFWAhL (where h is a hydrophobic M, V, L, or I residue). RtBRR and AtBRR dimerize distinctly from four structurally characterized phytochrome response regulators found in photosynthetic organisms and from all other receiver domain homodimers in the Protein Data Bank. A unique cacodylate-zinc-histidine tag metal organic framework yielded single-wavelength anomalous diffraction phases and may be of general interest. Examination of the effect of the BRR stoichiometry on signal transduction showed that phosphorylated RtBRR is accumulated more efficiently than the engineered monomeric RtBRR (RtBRR_mon) in phosphotransfer reactions. Thus, we conclude that arm-in-arm dimers are a relevant signaling intermediate in this class of two-component regulatory systems.

IMPORTANCE BphP histidine kinases and their cognate response regulators comprise widespread red light-sensing two-component systems. Much work on BphPs has focused on structural understanding of light sensing and on enhancing the natural infrared fluorescence of these proteins, rather than on signal transduction or the resultant phenotypes. To begin to address this knowledge gap, we solved the crystal structures of two single-domain response regulators encoded by a region immediately downstream of that encoding BphPs. We observed a previously unknown arm-in-arm dimer linkage. Monomerization via deletion of the C-terminal dimerization motif had an inhibitory effect on net response regulator phosphorylation, underlining the importance of these unusual dimers for signal transduction.

INTRODUCTION

Bacteria utilize two-component systems (TCSs) to monitor and respond to diverse signals in the environment, including the spectrum and intensity of visible (Vis) light. Light of a specific wavelength can control directed responses, as in the case of phototaxis (1, 2), or can control generalized stress responses (3). Bacteriophytochrome photoreceptors (BphPs) are soluble cytoplasmic red light-sensing modules, often histidine kinases (HKs), encoded by a region adjacent to a single-domain response regulator (SDRR) to which phosphate is transferred in a light-regulated fashion (4–6). While the mechanism of red light reception by the sensory domains of BphPs is the topic of much active research and recent work has added to our understanding of the mode of intramolecular signal transduction by the sensory domains of phytochrome (7–9), the effects of light on phosphate flux through entire pathways are less well studied. Few phenotypic responses have been conclusively attributed to BphP TCSs, and those that are known suggest that the physiological responses controlled by these photoreceptors are as diverse as the environments that bacteria inhabit (3, 10–14). Certainly critical to connecting red light sensing to appropriate cellular responses are the bacteriophytochrome response regulator (BRR) proteins. Thus, we characterized two BRRs structurally and placed them in context of the biochemical activity of a cognate BphP from the Ramlibacter tataouinensis red light-sensing TCS (RtBphP). This TCS was revealed by annotation of the genome sequence of this chemotrophic desert microbe, the genome of which encodes a high number of putative light-sensing proteins (15).

Typical BphPs and cyanobacterial phytochromes (Cphs) are dimeric proteins that are capable of maximal absorbance of red light in the dark (Pr [phytochrome red]) state and that convert to a far-red light-absorbing (Pfr [phytochrome far red]) state after exposure to red light. The canonical domain architecture is Per-Arnt-Sim (PAS)–cGMP phosphodiesterase-adenylate cyclase-FhlA (GAF)–phytochrome (PHY)-specific HK (16–18). The PAS domain is the site of covalent chromophore attachment in BphPs, which utilize biliverdin IXα (BV), whereas in Cphs, phycocyanobilin is covalently linked to the GAF domain (18). In all cases, the GAF domain amino acids surround the chromophore. Signal transduction is initiated by a light-driven isomerization of a double bond in this tetrapyrrole (19), and the PHY domain transduces conformational changes to the HK domain. The HK catalyzes ATP hydrolysis and transfers the γ-phosphate to a conserved histidine residue. The transfer is assumed to be to the sister protomer in the dimer, on the basis of comparison of the sequence to the sequences of other trans-acting HKs (20, 21). The kinetics of autophosphorylation of a Cph and subsequent phosphotransfer to its cognate response regulator (RR) have been measured for one example but have not been reported in detail for nonphotosynthetic systems. Psakis et al. (22), following an early qualitative report of light regulation of both kinase and phosphotransferase activities (19), reported the K_m and k_cat values for the Cph1/Rcp1 TCS from Synechocystis sp. strain PCC 6803 and demonstrated that the efficiency of both Cph1 autophosphorylation and phosphotransfer to the RR were 4-fold higher in the Pr state. Other groups' results support the finding that kinase activity is higher in the Pr state than in the Pfr state for BphPs as well. Giraud et al. reported that autophosphorylation by Rhodopseudomonas palustris BphP2 (RpBphP2) was reduced by 83% in the Pfr state (6). Phosphotransfer to the cognate RR Rpa3017 was also more efficient in the Pr state. Karniol and Vierstra reported that Agp1 from Agrobacterium tumefaciens had 2-fold greater autophosphorylation activity in the Pr state and 10-fold greater phosphotransfer to the RR1 response regulator from Agrobacterium tumefaciens (AtRR1; here called AtBRR) in the Pr state than in the Pfr state (5). The lack of kinetic parameters for BphPs prompted us to characterize the autophosphorylation activities of RtBphP1 as the earliest light-regulated step in this TCS signaling cascade.

Another hallmark of BV-binding BphPs is their reverse photoconversion from the Pfr state to the Pr state upon exposure to far-red light (18). Many are stable in both states and can be switched to the alternative ground state with the appropriate light stimulus. BphPs can also thermally revert to the Pr state in the absence of light with a reversion half-life that is characteristic of the particular phytochrome and can be temperature dependent. An example of a poorly stable Pfr form is Agp1, which is relatively unstable at room temperature (23) and displays accelerated thermal reversion at 30°C (24). Here we describe the reversion kinetics for RtBphP1 and show that RtBphP1 and Agp1 share rapid thermal reversion.

Most putative BRRs are annotated as CheY-like SDRRs, which consist of a receiver domain with no obvious regulatory domain, such as a DNA-binding motif. For cyanobacterial Cph systems, three such RRs have been characterized by X-ray crystallography: Rcp1 from Synechocystis sp. PCC 6803 (25) (PDB accession number 1I3C) and RcpA and RcpB, both of which are from Calothrix sp. strain PCC 7601 (26) (PDB accession numbers 1K66 and 1K68, respectively). This cyanobacterial set has recently been augmented by the structure of Rpa3017 (27). These proteins share an overall topology with other SDRRs. They consist of an internal parallel five-strand β sheet with a hydrophobic character surrounded by five α helices. A conserved aspartate residue protrudes from the terminus of the third β strand (β3) and serves as the phosphoacceptor site. Receiver domains can dimerize by one of four known structural arrangements, named for the α-helix and β-strand numbers involved in packing (28). The Rcp RRs and Rpa3017 all crystallized as unusual inverted α4-β5-α5 homodimers, and Benda et al. observed dimers for RcpA and RcpB irrespective of the phosphorylation state (26). The same work noted the conservation of consecutive Phe and Trp residues in a C-terminal extension in phytochrome RRs not found in other SDRRs. These residues fold as an α helix to form a solvent-exposed hydrophobic patch which interacts with the sister protomer via an aromatic cluster. Here we report the crystal structure of dimeric BRRs from two chemotrophic bacterial species which share the conserved FW key residues yet form homodimers through a distinct crossover linkage which, we propose, represents a novel category of RRs. These and arm-in-arm RR dimers may be common to a set of BphP TCSs from nonphotosynthetic bacteria.

We investigate the effects of these arm-in-arm dimers on signal transduction and underline their importance to intermolecular signal transduction in the BphP-BRR TCS.

MATERIALS AND METHODS

Cloning, expression, and purification of BphPs and BRRs.

R. tataouinensis strain TTB310 was grown as described previously (29), and genomic DNA was extracted using a Wizard genomic DNA kit (Promega, Madison, WI). The genes for RtBphP1 and R. tataouinensis BRR (RtBRR) were amplified by PCR from R. tataouinensis genomic DNA with primers encoding BamHI and HindIII restriction enzyme sites (primers RtBphP1 F [5′-CGAAGGATCCATGAACCTTCCGCCGCCTGACCTGG-3′], RtBphP1 R [5′-CGAAAAGCTTTTAAGCATGGTTCCTGTCCTCTTTCCTCTTGGGCGGC-3′], RtBRR F [5′-GGTTGGATCCATGCTTAAACCCATCTTGCTTGTCGAGGACGACAAGC-3′], and RtBRR R [5′-CCTTAAGCTTTGCTTCGTAGCGGCGCATGGCCTTCATGGACC-3′]). These sites were used to clone the genes into plasmid pJ414 carrying an N-terminal hexahistidine tag and tobacco etch virus (TEV) protease site (DNA2.0, Menlo Park, CA) or pET23a carrying a C-terminal hexahistidine tag (EMD Millipore, Billerica, MD), creating pJ414RtBphP1 and pET23aRtBRR_His, respectively. The gene for monomeric RtBRR (RtBRR_mon) was amplified from pJ414RtBRR using the original forward primer and a reverse primer (primer R [5′-GGTTAAAAGCTTTTAGCCCAGGTCGGCGATGGCGGCG-3′]) and then religated into empty pJ414, which resulted in the 5′ truncation of 21 codons. The AtBRR gene was codon optimized for expression in Escherichia coli, synthesized, and cloned into pJ414, resulting in pJ414AtBRR (DNA2.0, Menlo Park, CA).

Vectors with sequence-confirmed inserts were transformed into E. coli BL21(DE3) (AtBRR) or BL21-Codon Plus(DE3)-RP (R. tataouinensis genes) (Agilent Technologies, Santa Clara, CA). Overexpression was induced with 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) at an optical density at 600 nm of 0.5 to 0.8 and carried out for 16 to 18 h at 18°C. Cell pellets from 2-liter cultures were resuspended in 30 ml lysis buffer (30 mM Tris, pH 8.0, 300 mM NaCl, 10 mM imidazole) and were lysed in a French press. The RtBphP1 lysate was clarified by centrifugation (30 min at a 39,190 relative centrifugal force) and was incubated in the dark for 1 h on ice with 200 μl of 20 mM BV HCl in dimethyl sulfoxide (Frontier Scientific, Logan, UT).

All chromatography steps were run at 4°C on an Akta Explorer chromatograph (GE Healthcare, Pittsburgh, PA). RtBphP1, RtBRR_mon, and AtBRR were first enriched on a Ni-nitrilotriacetic acid (NTA) column (Qiagen, Hilden, Germany), followed by buffer exchange on a 50,000-molecular-weight-cutoff (MWCO) or 10,000-MWCO filter (Amicon; EMD Millipore, Billerica, MD) before digestion with TEV protease, which had been purified in-house as previously described (30), at 8°C for 16 to 18 h. Cleaved species were collected from the flowthrough fraction of a second Ni-NTA column, and the buffer was changed to 30 mM Tris, pH 8.0 (RtBphP1), or 30 mM Tris, pH 8.0, 100 mM NaCl, 5% glycerol (BRRs). The proteins were further purified and analyzed for stoichiometry on size-exclusion columns (Superdex 200 for RtBphP1 or Superdex 75 for BRRs; GE Healthcare, Pittsburgh, PA). The 189-kDa dimer fraction of RtBphP1 was isolated for use in all experiments. Of the proteins in this study, only dimeric His-tagged RtBRR (RtBRR_His) retained the C-terminal hexahistidine tag and thus underwent two-step purification with Ni-NTA affinity purification followed by Superdex 75 size exclusion.

BphP spectrophotometry and Pfr-state half-life determination.

UV-Vis spectra and single-wavelength measurements of RtBphP1 were collected on a Beckman Coulter DU-640B spectrophotometer (Pasadena, CA). All experiments in the dark state were carried out under dim green light. The spectra of dark-adapted protein which was protected from light for ≥3 h (Pr state) or illuminated with red light for 1 min (mixed Pr state/Pfr state) were collected. Proteins were illuminated with light from a Fostec ACE source filtered with a 700- ± 5-nm-band-pass filter (Andover Corp., Salem, NH), which delivered irradiance of 140 μmol/m²/s, as previously described (31). Illuminated-state measurements were limited to spectra recorded no less than 10 s after light stimulus removal, by which time thermal reversion was under way. The dark-reversion spectra of an illuminated RtBphP1 sample were recorded for a time course of 10 s to 60 min after light stimulus removal. Difference spectra were calculated as the illuminated absorbance minus the Pr-state absorbance and plotted for each time point.

The extinction coefficient (ε₇₀₈) of RtBphP1 in the Pr state was calculated by solving the equation c₂₈₀ = A₂₈₀/ε₂₈₀, where c₂₈₀ is the concentration of protein based on the absorbance at 280 nm, using the theoretical ε₂₈₀. Assuming that c₂₈₀ is equal to c₇₀₈, we then measured the A₇₀₈ and solved the equation ε₇₀₈ = A₇₀₈/c₇₀₈. This yielded a ε₇₀₈ of 98,937 M⁻¹ cm⁻¹ for Pr-state RtBphP1. Notably, this simplified method does not take into account contamination by apophytochrome, which was estimated to be 22% using the extinction coefficient for BV (39,900 M⁻¹ cm⁻¹) at 388 nm (32). Pfr-state half-lives were determined by measuring the A₇₀₈ for the Pr state on a dark-adapted RtBphP1 sample, followed by a 1-min illumination with 700-nm light and then measurement of the A₇₀₈ every 30 s for 1 h at 24°C. Three independent experiments were conducted, and a biexponential curve, y = Aexp(bt) + Cexp(dt), where t is time and y is the sum of two exponential functions with constants A and C and exponents bt and dt, respectively, was fit to the mean data using SigmaPlot dynamic curve fitting (Systat Software, San Jose, CA). The mean and standard deviation (SD) of ln(2)/b and ln(2)/d yielded the reported first and second half-lives of the Pfr state, respectively.

RtBRR_His and AtBRR X-ray crystal structure determination and protein interface analysis.

Purified RtBRR_His and AtBRR were concentrated to 6.3 mg/ml and 30.0 mg/ml, respectively, in 3,000-MWCO centrifugation filters (EMD Millipore, Darmstadt, Germany), and the buffer was changed to 30 mM Tris, pH 8.0, 100 mM NaCl, 5% glycerol. Hanging drops (33) were set up with 1 μl BRR and 1 μl reservoir solution. RtBRR_His crystallized with a reservoir solution of 200 mM zinc chloride, 100 mM sodium cacodylate, pH 6.5, 10 mM magnesium chloride, and 10% isopropanol. AtBRR crystallized with a reservoir solution of 100 mM Tris, pH 8.5, 200 mM magnesium chloride, and 20% polyethylene glycol 8000. Three-dimensional, birefringent crystals (multiple diamond-shaped crystals for RtBRR_His, single large rods for AtBRR) grew at 19°C within 1 week. For data collection, crystals were cryoprotected in the mother liquor with 15% glycerol for 1 min before vitrification in liquid nitrogen.

Diffraction data were collected at the Advanced Photon Source (Argonne National Laboratory, Argonne, IL) on LS-CAT beamline 21-ID-D on a Rayonix MX300 detector (RtBRR_His) and on LS-CAT beamline 21-ID-F on a Rayonix MX225 detector (AtBRR). The data were integrated and scaled with the HKL2000 program (34). For RtBRR_His, Phaser-MR software (35) was used for initial phasing by molecular replacement by use of the structure with PDB accession number 1I3C as the search model. The structure revealed an interesting dimethylarsenate-zinc interaction with an ordered C-terminal His tag. Even though we collected data at the peak wavelength for Se (0.97910 Å; LS-CAT 21-ID-D beamline) and not As (1.0417 Å) or Zn (1.2837 Å), we were able to use the anomalous signal and the protein sequence in the AutoSolve program (36) to independently phase the reported structure. The measurability of the anomalous signal was 0.0619 at a 2.58-Å resolution (37), with peaks being found for one As ion and three Zn²⁺ ions. The experimentally phased electron density map revealed all but the N-terminal 6 amino acids of the 166 amino acids in the protein. The cacodylate sits on a 2-fold rotation axis and was refined at half occupancy. For AtBRR, Phaser-MR software was used for phasing by molecular replacement using RtBRR_His as the search model. Model fitting was done in the Coot program (38), and refinement was done using the Phenix.refine program (37).

Coordinate files for AtBRR and RtBRR_His were submitted to the Proteins, Interfaces, Surfaces, and Assemblies (PISA) web server (39) to predict biologically relevant dimer interfaces from the crystal structures; the same analysis was performed on BRR structures from cyanobacteria: Rcp1 (25) and RcpA and RcpB (26). Grand average of hydropathicity (GRAVY) scores were computed for the hFWAhL motif (where h is a hydrophobic M, V, L, or I residue) from AtBRR, RtBRR, Rcp1, RcpA, RcpB, Deinococcus radiodurans A0049, and R. palustris Rpa3017 using the ExPasy ProtParam tool (40). Surface electrostatics for RtBRR (without cloning tag residues) were generated using the APBS server (41, 42) and visualized in the PyMOL molecular graphics system.

Autophosphorylation kinetics and phosphotransfer profiling.

Autophosphorylation reactions were carried out at room temperature under green safe lights or over a panel of 700-nm light-emitting diodes (230 μmol/m²/s). The reaction mixtures contained 5 μM RtBphP1 in kinase buffer (150 mM MES [morpholineethanesulfonic acid; pH 7.5], 150 mM KCl, 5% glycerol, 5 mM MnCl₂, 5 mM MgCl₂, 2.5 mM dithiothreitol [DTT]), and the reactions were started with the addition of ATP (6.25 to 1,000 μM cold ATP with 0.03 to 0.15% [γ-³²P]ATP [6,000 Ci/mmol, 150 mCi/ml]) and were stopped after 10 s with an equal volume of 2× sodium dodecyl sulfate (SDS) loading buffer (100 mM Tris-HCl, pH 6.8, 4% [wt/vol] SDS, 0.2% bromophenol blue, 20% vol/vol glycerol, 200 mM DTT). Samples were not heated or vortexed prior to 15% SDS-PAGE to minimize the loss of phosphorylated species. After Coomassie staining, RtBphP1 band slices were cut and added to 4 ml of Bio-Safe II scintillation fluid (Research Products International, Mt. Prospect, IL) before the number of counts per minute was recorded with a Packard Tri-Carb 2100TR liquid scintillation analyzer (Perkin-Elmer, Waltham, MA). Initial rates (millimoles of ³²P-labeled RtBphP1 per second) were plotted versus the ATP concentration to derive V_max, K_m, and k_cat values (22). The means and standard deviations from five independent experiments are reported. Previous experiments carried out at a constant ATP concentration and various RtBphP1 concentrations were used to ensure that this assay was done in the linear range of activity.

Radioactive phosphotransfer reactions were carried out at room temperature under green safe lights. The reaction mixtures contained 5 μM RtBphP1 plus 15 μM RR in kinase buffer. First, the autophosphorylation reaction was run for 30 s by addition of ATP (1,000 μM, 0.05% [γ-³²P]ATP) to RtBphP1, and then phosphotransfer was initiated by adding the RR for an additional 30 s. The reactions were stopped with an equal volume of 2× SDS loading buffer. Samples were run on 20% SDS-polyacrylamide gels, and the gels were stained prior to exposure to phosphor screen (exposure time, 2 to 16 h) and imaged on a GE Typhoon FLA-9000 imager. Phosphor band intensities, including background estimation and normalization for the BphP protein amount on the basis of Coomassie staining, were quantified with ImageJ software (version 1.49; W. S. Rasband, National Institutes of Health, Bethesda, MD). The values reported are the means and standard deviations from three independent experiments. A one-way analysis of variance determined that there was a statistically significant difference between groups. A Tukey's honestly significant difference test determined the statistically significant difference between individual groups.

Phosphotransfer reaction mixtures for Phos-tag acrylamide gels contained 5 μM RtBphP1 plus 15 μM RR in kinase buffer with or without 1 mM ATP and were incubated for 0, 10, 30, or 60 min before the addition of 3× SDS loading buffer to stop the reaction. Potentially phosphorylated proteins were placed on ice after the reactions were stopped and separated on 12% SDS-polyacrylamide gels with 100 μM Phos-tag acrylamide (Wako Pure Chemicals, Osaka, Japan) within 1 h. Gels were run at 4°C until the dye front ran off and then stained overnight with Sypro Ruby before they were imaged on a GE Typhoon FLA-9000 imager. The band intensity (I) minus the background intensity was quantified using ImageJ software. The percentage of RR phosphorylated was calculated as (I_upper × 100)/(I_upper + I_lower) for each lane, where I_upper is I for the upper band and I_lower is I for the lower band.

Protein structure accession numbers.

Coordinate and structure factor files for RtBRR_His and AtBRR were deposited in the Protein Data Bank under accession numbers 5IC5 and 5BRJ, respectively.

RESULTS

Photoproperties of RtBphP1.

In order to begin characterization of the red light-sensing TCS from R. tataouinensis, we cloned, expressed, and purified full-length RtBphP1. Size-exclusion chromatography separated the RtBphP1 molecules into four populations with distinct sizes, corresponding to 771, 300, 189, and 108 kDa (Fig. 1A). The predominant 189-kDa dimer fraction had a molar proportion of BV to BphP of 0.78, as determined by the relative absorbance and extinction coefficients of RtBphP1 (ε = 98,937 at 708 nm) and free BV (ε = 39,900 at 391 nm). This fraction was used to carry out all dark reversion, autophosphorylation, and phosphotransfer experiments.

FIG 1.

Photoproperties of RtBphP1. (A) Gel filtration of RtBphP1. Peaks with absorbances of both 280 nm and 700 nm (peaks A, B, C, and D) correspond to RtBphP1 oligomers. Peak C is the dimer fraction used in all experiments. Peaks A and B correspond to higher-order oligomers, and peak D corresponds to monomers. Numbered circles represent size standards: 1, thyroglobulin (669 kDa); 2, ferritin (440 kDa); 3, catalase (232 kDa); 4, aldolase (158 kDa). (B) Visible spectra of RtBphP1 in the fully dark-adapted, Pr state (black line) after 1 min of illumination with 700-nm light (red line) and products recovered over an interval of 5 to 60 min (gradient of red to dark gray). (C) Difference spectra (illuminated absorbance minus dark-state absorbance) of the data from panel B. (D) Time course and biexponential modeling of Pfr-state thermal decay at 24°C. The mean data and derived half-lives from three independent experiments are shown. mAu, milliabsorbance units; Au, absorbance units; ΔAu, change in the number of absorbance units; t_1/2, half-life.

Zinc-binding fluorescence assays (data not shown) demonstrated that RtBphP1 covalently binds the added BV chromophore. UV-Vis spectroscopy verified that RtBphP1 is a functional red light-sensing BphP with a Pr dark ground state (λ_max = 708 nm) (Fig. 1B). Exposure of fully dark-adapted RtBphP1 to 700-nm red light for 1 min induced a relatively unstable Pfr state (λ_max = ∼750 nm) which thermally reverted to the Pr state (Fig. 1B and C) with an unexpected biphasic behavior and half-lives of 0.7 and 20.4 min (Fig. 1D).

RtBphP1 is a light-regulated autokinase.

RtBphP1 acts as a light-regulated autokinase, as evidenced by radiolabeled phosphorylation reactions and their associated Michaelis-Menten kinetic constants (Fig. 2). The protein is autophosphorylated in the dark, and this activity is modestly suppressed by 700-nm light. The initial rate of autophosphorylation was slow and was retarded by 2-fold under red light (k_cat in the dark was 2.0 × 10⁻⁵ ± 0.1 × 10⁻⁵ s⁻¹, whereas k_cat under red light was 0.9 × 10⁻⁵ ± 0.3 × 10⁻⁵ s⁻¹). Red light acts as a noncompetitive inhibitor of autokinase activity, as indicated by unchanged K_m values for ATP (for the Pr state and the Pr/Pfr mixture, K_m = 12.9 ± 3.5 μM and 13.6 ± 6.0 μM, respectively), despite the modulated k_cat.

FIG 2.

Red light-regulated autophosphorylation of RtBphP1. Michaelis-Menten plots from five independent experiments are shown. The initial rate of formation (V₀; mean ± SD) of ³²P-labeled RtBphP1 versus the ATP concentration are compared for the activity in the dark state (black) or in the red illuminated state (gray).

AtBRR and RtBRR_His are novel arm-in-arm dimers.

In order to probe intermolecular signal transduction by BphPs, we solved the crystal structure of RtBRR, the putative cognate RR of RtBphP1, in the nonphosphorylated state. Crystal growth and phasing were made possible by the inclusion of a C-terminal 14-residue tag ending in hexahistidine, which is well ordered and coordinated by zinc and cacodylate [(CH₃)₂AsO₂H] (Fig. 3A). In fact, the zinc and As atoms provided a strong anomalous signal, and we were able to phase the RtBRR_His structure with a Zn²⁺/As single-wavelength anomalous dispersion data set. This technique should be of general use to solve protein structures that have a histidine tag by crystallizing the cacodylate/zinc/His complex, since the number of structures with His tags in the Protein Data Bank (PDB) is rapidly increasing (43). This strategy is similar to that used in a new general method that has been proposed for coating the surface of a protein with zinc atoms to solve the phase problem (44). AtBRR phases were obtained by molecular replacement with RtBRR_His as the search model (Fig. 3B). RtBRR_His was refined to a resolution of 1.9 Å with R_work and R_free values of 18% and 20%, respectively. AtBRR was refined to a resolution of 1.9 Å with R_work and R_free values of 20% and 24%, respectively (Table 1).

FIG 3.

Electron density maps and models for RtBRR_His and AtBRR. (A) Ordered hexahistidine tags from two RtBRR_His molecules (one [left] with C in gray, one [right] with C in orange, both with O in red, and both with N in blue) across the crystallographic 2-fold axis are coordinated by cacodylate (As [purple]) and zinc ions (slate spheres). Gray mesh, F_o electron density map phased with anomalous signal contoured at 1σ; red mesh, anomalous difference map contoured at 3σ. (B) Detail of the AtBRR phosphoacceptor site including an Mg ion (green sphere) coordinated by side chain and main-chain atoms and three ordered water molecules (red spheres). The electron density shown is a 2mF_o-DF_c simulated annealing composite omit map (m, figure of merit; D, σ_A weighting factor) contoured at 1σ.

TABLE 1.

X-ray data collection and structure determination statistics

Parametera	Value(s) forb:
	RtBRRHis (PDB accession no. 5IC5)	AtBRR (PDB accession no. 5BRJ)
Data collection statistics
Wavelength (Å)	0.9791	0.9787
Resolution (Å)	38.3–1.83 (1.85–1.83)	35.0–1.92 (1.95–1.92)
Space group	P4122	P4122
a, b, c unit cell dimensions (Å)	47.7, 47.7, 193.5	41.0, 41.0, 187.7
Completeness (%)	99.8 (98.1)	98.7 (97.0)
No. of unique reflections/no. of measured reflections	20,790/20,756	13,163/12,959
No. of anomalous reflections	37,570
Redundancy	1.8 (1.8)	27.2 (26.4)
<I/σI>	22.9 (1.1)	53.2 (23.0)
Wilson B value (Å2)	33.0	23.7
Rsym (%)	3.4 (66.4)	5.1 (15.5)
Refinement statistics
Resolution (Å)	38.36–1.90 (1.93–1.90)	30.9–1.92 (1.99–1.92)
No. of reflections/no. of anomalous reflections	18619/33646	12959
Rwork/Rfree (%)	17.5/21.3 (31.5/40.4)	19.6/24.0 (21.7/27.4)
RMSD
Bond length (Å)	0.02	0.07
Bond angle (°)	1.64	1.03
Ramachandran statistics (%)
Allowed	98.7	96.3
Generously allowed	1.2	3.7
No. of atoms
Protein	1,375	1,107
Ligand	9	1
Water	123	135
<B factor> (Å2)
Protein	38.4	26.8
Ligand	37.3	20.2
Water	52.8	36.6

^aR_sym = ΣΣj|Ij − <I>|ΣIj, where Ij is the intensity measurement for reflection j and <I> is the mean intensity for multiply recorded reflections. R_work and R_free = Σ||F_o| − |F_c||/|F_o|, where the working and free R factors are calculated by using the working and free reflection sets, respectively. For R_free, 5 to 10% of the total reflections were held aside throughout the refinement.

^bThe data for the highest-resolution bin are indicated in parentheses.

As expected, the basic topology of RtBRR_His is similar to that of the archetypal SDRR, E. coli CheY (PDB accession number 3CHY; root mean square deviation [RMSD] = 1.3 Å over 70 C-α atoms within secondary structure elements) (45). The common structural features are a hydrophobic five-stranded parallel β sheet core (2-1-3-4-5) surrounded by five α helices (Fig. 4A). The sequence-conserved phosphoaccepting aspartate (D64 in RtBRR, D65 in AtBRR) is situated at the C-terminal end of β3, faces into the solvent, and is accessible for interaction with the cognate BphP HK (Fig. 4A). Both BRR structures had clear electron density for magnesium ions, which are required for the phosphoryl transfer reaction (46), in the phosphorylation site (Fig. 3B).

FIG 4.

Structure of arm-in-arm BRR dimers. (A) Overview of the RtBRR_His dimer in which one protomer (gray) links with another (orange; β6 and α6 are highlighted in yellow) to form the arm-in-arm dimer (cloning tag and histidine residues are intimated with a dashed line). Green sphere, the Mg ion that marks the phosphoacceptor surface; C, C terminus. (B) Detail of the RtBRR_His arm-in-arm dimer interface. β6 from each monomer participates in an intermolecular antiparallel β sheet; the extensive hydrogen-bonded network also involves N-terminal residues and the loop between α1 and β2. Conserved N29 bridges the quaternary arrangement by hydrogen bonding to main-chain atoms of residues 3 and 136. Conserved P4 positions the N terminus for these interactions. (C) Surface interaction potential for the RtBRR dimer. The hFWAhL dimerization motif and D64 are shown as sticks.

The crystal structure of RtBRR_His revealed an unusual crossover dimer interface geometry that links sister monomers (Fig. 4A). The dimer relies on a C-terminal extension with marked structural deviation from known SDRR structures: a bulky hydrophobic β6 (IFWAVL) that extends from α5 and threads through the adjacent monomer before turning back toward the originating protomer via a proline/glycine-rich sequence (Fig. 4B). Two antiparallel strands, one from each monomer, form an intermolecular β sheet, evidenced by the presence of six main-chain hydrogen bonds. This small sheet is widened by main-chain interactions with the loop between α1 and β2 and stapled in place by the side chain of N29, itself invariant among a group of approximately 150 RRs that share the DLGhFWAhLNEPPP sequence (where h is a hydrophobic M, V, L, or I residue). Conserved P4 positions the N terminus for additional interactions at the interface (Fig. 4B). The RtBRR_His polypeptide ends in a sixth α helix (α6), which packs against α1 and β2 and is not found in most SDRR structures. Models of the surface electrostatic potential of RtBRR reveal a sizable cleft between protomers and an extensive positively charged stripe across the dimer surface, either or both of which might serve as an interaction site for signaling partners (Fig. 4C).

To ensure that the observed RtBRR_His dimer interface was not a consequence of the stabilizing effects of helix-promoting residues or histidine-metal interactions in the cloning tag and to explore the stoichiometry of the BRRs in other BphP systems, we also analyzed the interface in the crystal structure of the native-sequence AtBRR from A. tumefaciens strain F2. This protein contains all three of the noted N-terminal Pro, α1-β2 loop LXN, and C-terminal DLGhFWAhLNEPPP motifs (Fig. 5). AtBRR dimers were linked by the same bulky hydrophobic β6 (VFWALL), although the following proline-rich turn and terminal helix residues could not be modeled due to a weak electron density. Given the stabilizing effect of zinc in the crystal structure of RtBRR_His, size-exclusion chromatography runs for RtBRR_His and AtBRR were performed in metal-free buffer and revealed peaks corresponding to dimers in solution for both species (Fig. 6A and B).

FIG 5.

Multiple-sequence alignment of SDRRs encoded by regions near bacteriophytochromes. The F/Y and W residues conserved in bacteriophytochrome RRs are highlighted (black background with white text). Residues crucial to the arm-in-arm dimer interface observed in RtBRR and AtBRR are also conserved in the RRs of Pseudomonas syringae (PsBRR), Burkholderia glumae (BgBRR), and Rhizobium sp. strain NT-26 (RNBRR) (gray background with black text). SDRRs adopting the inverted 4-5-5 dimer interface (Rcp1, RcpA, RcpB, and Rpa3017; α5 residues are shown with a gray background and white text) lack most or all of these residues, as does the RR from D. radiodurans AA049 (DrAA049). EcCheY, E. coli CheY. Periods, conservation of residues with weakly similar properties; colons, conservation of residues with strongly similar properties; asterisks, fully conserved residues.

FIG 6.

Oligomeric status of BRR proteins determined by gel filtration chromatography (1.4 to 1.6 mg was loaded in each case). Molecular masses, determined on the basis of a standard curve, are reported for the highest-magnitude peak in each panel (the nonlabeled peaks in panel B are 33.1 kDa and 23.9 kDa). The labeled peaks were isolated and used for downstream experiments. (A) RtBRR_His; (B) AtBRR; (C) RtBRR_mon.

Analysis of the packing interactions of RtBRR_His and AtBRR using the PISA server (39) confirmed the dominant nature of this hydrophobic crossover strand in the stabilization of the dimer. The interface has solvent-inaccessible areas of 1,310 Ų (ΔG = −20.7 kcal/mol) and 1,400 Ų (ΔG = −17.9 kcal/mol) per monomer for RtBRR (regardless of the inclusion of tag residues) and AtBRR, respectively, with both hydrophobic packing and hydrogen bonding playing important roles in the interaction (Table 2).

TABLE 2.

Statistics for phytochrome cognate RR dimer interfaces^a

Protein (PDB accession no.)	Crossover motif	Interface area/monomer (Å2)	ΔG (kcal/mol)	Hydropathicity scoreb for motif	No. of residues/monomer in interface	No. of H bonds in interface
RtBRR (5IC5)	IFWAVL	1,309.7	−20.7	2.70	30	16
AtBRR (5BRJ)	VFWALL	1,399.9	−17.9	2.58	30	24
Rcp1 (1I3C)	SFWLET	1,172.5	−13.3	0.12	29	19
RcpA (1K68)	EFWLSY	1,104.9	−15.3	0.02	27	11
RcpB (1K66)	KYWLDI	1,248.4	−13.0	−0.22	33	18
Rpa3017 (4ZYL)	HFWMNT	1,164.9	−7.5	−0.60	33	16

^aStatistics were calculated using the ProtParam tool and PISA.

^bGrand average of hydropathicity (40).

BRR dimers promote signal transduction.

In order to investigate the relevance of the unique arm-in-arm BRR dimers formed by the crossover strand in the phosphorelay from BphPs to BRRs, we engineered a monomeric BRR missing the C-terminal 21 amino acids of the native RtBRR sequence. These amino acids correspond approximately to an extension in the sequence compared to the sequences of canonical SDRRs (Fig. 5). Size-exclusion chromatography verified that RtBRR_mon has a molecular mass in solution of 16 kDa, whereas the molecular mass for the RtBRR_His dimer is 36 kDa (Fig. 6A and C).

Dark-adapted RtBphP1 dimers were allowed to autophosphorylate at room temperature before the addition of RtBRR_His, RtBRR_mon, or AtBRR, and the transfer of radiolabeled phosphate was visualized by autoradiography. Phosphotransfer to both the RtBRR_His dimer and the RtBRR monomer was evident (Fig. 7A), indicating that the specificity between the two components was not negatively impacted by the monomer deletion. Specificity for the cognate RR was evident, as RtBphP1 did not transfer phosphate to AtBRR at appreciable levels (Fig. 7A).

FIG 7.

Phosphorylation state of TCS partners during phosphotransfer reactions. (A) After preincubation with [γ-³²P]ATP, RtBphP1 was incubated with the indicated RR as described in Materials and Methods and visualized by Coomassie staining and phosphorimaging of an SDS-polyacrylamide gel. (B) Phosphorylation of RtBphP1 was quantified using the low-exposure image in panel A, the intensity of phosphorylated RtBphP1 was normalized to the intensity of Coomassie-stained BphP bands, and the amount was plotted as a percentage of the basal amount of phosphate remaining. The results from three independent experiments are plotted as the means ± SDs. **, P < 0.01 by Tukey's honestly significant difference test. *P-RtBphP1, phosphorylated RtBphP1. (C) Phosphorylated RtBphP1 (RtBphP1∼P) was incubated with RtBRR_His dimers (first four lanes) or engineered monomers (lanes 5 to 8) and ATP and imaged on a Phos-tag acrylamide gel (see Materials and Methods). RtBRR∼P, phosphorylated RtBRR. (D) Sypro Ruby-stained bands from panel C were quantified with ImageJ software, and the amount was plotted as a percentage of the amount of RR phosphorylated.

To assess the relative efficiency of phosphotransfer from RtBphP1 to each BRR, we independently measured both the loss of phosphorylated HK (HK*) (Fig. 7A and B) and the accumulation of phosphorylated BRR (BRR*) (Fig. 7C and D). In the presence of the RtBRR_His dimer, RtBphP1 was more efficiently dephosphorylated than the phosphorolysis control consisting of HK* only, with 65% of the HK* remaining after 30 s (Fig. 7A and B). However, in the presence of RtBRR_mon, the amount of HK* remaining was actually higher than the basal level (130%). This result may suggest that monomeric RtBRR inhibits phosphotransfer from HK* (Fig. 8, reaction 2), although other steps in the TCS may be affected (Fig. 8). RtBphP1 phosphorylation in the presence of AtBRR was equivalent to that for the buffer-only control, demonstrating phosphotransfer specificity between the RtBphP1 and its cognate BRR.

FIG 8.

Simplified schematic of phosphate movement through the RtBphP1 (HK)-RtBRR (RR) TCS, showing four steps that contribute to signal transduction. One or more of these is likely to be impacted by the stoichiometry of BRR, thus accounting for the accumulation of larger amounts of phosphate on RtBRR_His than RtBRR_mon (putative additional members of this TCS are not taken into account here). * indicates phosphate that is transferred.

In order to examine signal transduction via the accumulation of BRR*, RtBphP1 dimers were incubated with RtBRR_His or RtBRR_mon and an excess of ATP before analysis of the level of RR phosphorylation using Phos-tag acrylamide gels (47–49). Phosphotransfer to both the RtBRR_His dimer and the RtBRR monomer was evident (Fig. 7C), providing confirmation that the specificity between the two components was not impacted by the monomerization. Analysis of the relative levels of phosphorylation of BRRs indicated that RtBRR_His more rapidly accumulated larger amounts of phosphate than RtBRR_mon at all time points tested (Fig. 7D). At the final 60-min time point, RtBRR_His was 38% phosphorylated, whereas RtBRR_mon was 24% phosphorylated. Thus, monomerization of RtBRR most likely inhibits phosphotransfer from HK* (Fig. 8, reaction 2). This experiment does not rule out but deemphasizes the possible major effects of monomerization of RtBRR on other steps in the TCS, including HK autophosphorylation, HK* phosphorolysis, and/or intrinsic BRR* dephosphorylation (Fig. 8, reactions 1, 3, and 4) because in each case the prediction would be a higher level of phosphorylated RtBRR_mon than phosphorylated RtBRR_His remaining.

DISCUSSION

Our characterization of the R. tataouinensis bacteriophytochrome TCS provides the autophosphorylation kinetics of a red light-repressed HK with rapid thermal dark reversion, reveals structural details about a previously unknown arm-in-arm dimer association for the RR, and demonstrates the importance of the arm-in-arm dimer for efficient signal transduction through the TCS.

The thermal reset of RtBphP1 to the dark state, which is kinetically accelerated for signaling, proceeds rapidly (Fig. 1D). This behavior, manifested as weak or unstable absorbance at 750 nm in the Pfr state, is also observed in Agp1 from A. tumefaciens (23), the cognate BphP for AtBRR. This is in contrast to the findings for most other characterized BphPs and Cphs, including D. radiodurans BphP (DrBhpP) and Synechocystis sp. PCC 6803 Cph1, which achieve stable Pfr absorbance states and more slowly decay back to the Pr state, yet can be photoswitched to the Pr state with far-red light. The former class may act as single-state light-sensing switches which in the absence of light are rapidly activated. The advantage of the latter class is that such BphPs can act as two-state light sensors contained in a single enzyme, which would be advantageous in environments where sensing of the ratio of two light wavelength ranges confers a survival advantage. BphPs with an unstable Pfr-state absorbance potentially require a second signal or binding partner to stabilize the state with far-red light, thereby integrating multiple input signals.

RtBphP1 autophosphorylation measurements support the model that red light acts as a noncompetitive inhibitor of the BphP transphosphorylation activity (Fig. 2). BphP HK dimers with covalently attached BV in the sensory domain can bind ATP in the kinase domain with a similar affinity in the Pr and Pfr states. Autophosphorylation proceeds more rapidly in the dark; thus, catalytic steps of ATP hydrolysis and/or transfer of the γ-phosphate to histidine are potentially light regulated. Conformational changes originating in the sensory domains of BphP must be transduced to regulate trans phosphorylation. The PHY domain tongue-refolding mechanism (7) demonstrated for DrBphP likely contributes to the regulation of HK domain activity by repositioning the HK domains on each protomer relative to one other. Perhaps it is surprising that autokinase activity is reduced only 2-fold by red light (in vitro) if R. tataouinensis utilizes this sensor kinase to regulate a process vital to environmental survival. However, ATP hydrolysis and the transfer of phosphate to histidine form but one measurable kinetic step that contributes to overall phosphate flux in the TCS (Fig. 8).

The arm-in-arm dimer interface observed for RtBRR_His and AtBRR (Fig. 9A and B) differs from the interface observed in three Cph cognate RR structures and R. palustris Rpa3107 (Fig. 9C and D) (PDB accession numbers 1I3C, 1K66, 1K68, and 4ZYL) (25–27) and is distinct from the three other known RR homodimerization modes (28). Although all six structural examples of BphP-associated RRs form dimers mediated by conserved Phe/Tyr and Trp amino acids, conservation of these two aromatic residues alone does not result in equivalent quaternary structures. The hydropathicity of surrounding amino acids and their influence on the local secondary structure determine the quaternary arrangement for these SDRRs as arm-in-arm (Fig. 9A and B) or inverted 4-5-5 dimer (Fig. 9C and D). In the cyanobacterial RRs, charged and polar residues surround FW and result in high hydrophilicity, quantified by grand average of hydropathicity scores of −0.2 to 0.1 (40) (Table 2). The dimerization motif residues form α5 near the C terminus, where Phe and Trp jut out to pack against the sister protomer (Fig. 9C and D). In our newly solved BRR structures, bulky hydrophobic residues surround FW and result in a high hydrophobicity (hydropathicity scores, 2.6 to 2.7) (Table 2); α helices do not form, and instead, hydrophobic residues are buried within the folded core, leaving main-chain atoms to form a hydrogen bond-dominated β-sheet interface (Fig. 9A and B). More than 150 bacterial SDRRs in currently searchable databases carry the extended C-terminal hydrophobic crossover motif DLGhFWAhLNEPPP plus a Pro near the N terminus and the LXN motif between α1 and β2. Notably, all of these predicted arm-in-arm dimer SDRRs are found in nonphotosynthetic bacteria, most of which are plant pathogens or commensals, such as Pseudomonas syringae and Burkholderia glumae (Fig. 5). Future work to elucidate the biological roles of these signaling pathways is needed.

FIG 9.

Uniqueness of arm-in-arm response regulator dimers and the role of FW. (A) Topology of the arm-in-arm dimer seen in RtBRR and AtBRR, characterized by the DLGhFWAhLNEPPP motif, which interacts with the N-terminal segment (P4, marked by P) and α1-β2 connecting loop (L27 and N29, marked by L and N, respectively). (B) Structural view of the motifs in panel A in the context of RtBRR. (C) Topology of the inverted 4-5-5 dimer seen in Rcp1, RcpA, RcpB, and Rpa3017, characterized by hydrophilic residues surrounding the FW key. (D) Structural view of the β5-α5 packing interaction in the context of Rcp1 (PDB accession number 1I3C) (25). Topology diagrams were generated using the Pro-origami system (55).

The BRR dimers possess a larger solvent-inaccessible surface area and greater negative ΔG values of dimerization than the six previous examples (Table 2), which suggest that the arm-in-arm dimers require a substantial input of energy to dissociate and may maintain the arm-in-arm arrangement regardless of the signaling state. Regulation of monomer-dimer equilibrium in response to phosphorylation has been suggested to be a possible mechanism of signal transduction for the Cph1-Rcp1 TCS (25). The possibility of such a transition cannot be ruled out for RtBphP1 and RtBRR; however, the predicted thermodynamics for the dimer interface disfavor such a mechanism. Alternatively, one can consider a model in which phosphorylation promotes conversion between inverted 4-5-5 and arm-in-arm dimers, impacting the quaternary structure arrangement. How phosphorylation affects other, potentially signal-transducing regions of the BRR structure remains an open question. The BRR proteins studied here retain the switch tyrosine essential for chemotaxis signaling in the E. coli CheY system (50), and neither the dimer interface nor the C-terminal 6th helix in RtBRR occlude the position occupied by the FliM α helix, the binding partner for CheY (51). Further structural studies of variants and/or phosphate analog-bound BRRs are warranted to address these questions.

The marked biochemical consequences of disrupting the arm-in-arm dimer interface imply that this oligomeric arrangement is the relevant signal receiver in the TCS in vivo (52). Although we cannot define the precise molecular block, we observed an inhibitory effect on both the dephosphorylation of phosphorylated RtBphP1 and the accumulation of phosphorylated RtBRR in the presence of RtBRR_mon compared to the effect seen in the presence of dimeric RtBRR_His. What new functionalities could be conferred by the novel arm-in-arm dimer compared to the functionalities conferred by a canonical monomeric SDRR? In both experiments conducted, phosphotransfer proceeded to monomeric as well as dimeric forms of RtBRR; thus, HK recognition cannot be carried by the C-terminal dimerization motif. In vivo, additional BRR interactions that couple the TCS to an appropriate cellular response may require the novel surface generated by the dimer interface or may take advantage of the larger molecular size of a dimer compared to that of a monomeric SDRR. The BphP TCS from Rhizobium sp. strain NT-26 has been shown to branch, with the BphP1 serving as a phosphodonor to both the cognate arm-in-arm BRR (Fig. 5) and a hybrid HK containing an RR domain (49); one might postulate that such complex networks of protein interactions could capitalize on the multiple binding sites presented by a dimeric BRR. Alternatively, arm-in-arm BRR dimers may provide a cooperative mechanism for signal transduction by doubling the local concentration of phosphoacceptor sites. This mechanism might increase the efficiency of an RR that acts as a phosphate sink in a multicomponent TCS to fine-tune signaling, as has been proposed for the LovR RR in the blue light-regulated LOV HK pathway (53, 54). A molecular picture of how BRRs couple light sensing by BphPs to phenotypic responses is the next major knowledge gap to be addressed. The signal transduction steps that take place beyond phosphorylation of the SDRR remain one of the most elusive areas of BphP research, and their identification will be aided by knowledge of the biochemistries and structures of the pathway proteins.