STRUCTURAL BASIS FOR THE PHOTOCONVERSION OF A PHYTOCHROME TO THE ACTIVATED FAR-RED LIGHT-ABSORBING FORM

Abstract

Phytochromes are a collection of bilin-containing photoreceptors that regulate numerous photoresponses in plants and microorganisms through their ability to photointerconvert between a red light-absorbing, ground state Pr and a far-red light-absorbing, photoactivated state Pfr^1,2. While the structures of several phytochromes as Pr have been determined^3-7, little is known about the structure of Pfr and how it initiates signaling. Here, we describe the three-dimensional solution structure of the bilin-binding domain as Pfr using the cyanobacterial phytochrome from Synechococcus OSB’. Contrary to predictions, light-induced rotation of the A but not the D pyrrole ring is the primary motion of the chromophore during photoconversion. Subsequent rearrangements within the protein then affect intra- and interdomain contact sites within the phytochrome dimer. From our models, we propose that phytochromes act by propagating reversible light-driven conformational changes in the bilin to altered contacts between the adjacent output domains, which in most phytochromes direct differential phosphotransfer.

The biological perception of light is mediated by a collection of photoreceptors that couple light absorption to specific signaling cascades. One influential set is the phytochromes, a superfamily of dimeric chromoproteins that absorb light via a bound bilin (or linear tetrapyrrole) chromophore^1,2,8. The bilin is buried within an N-terminal cGMP phosphodiesterase/adenyl cyclase/FhlA (GAF) domain whose contacts with the chromophore generate much of the unique photochromic behavior of Phys. Typically, the GAF domain is preceded by a Per/Arndt/Sim (PAS) domain and followed by a Phy-associated (PHY) domain and an output module, which often includes a histidine kinase domain that initiates a two-component phosphorelay. By photointerconversion between Pr and Pfr, phytochromes act as light-regulated switches for measuring the fluence, direction, duration and color of the ambient light environment⁸.

Despite intensive study, we know little about how phytochromes acquire their unique photochromic behavior and how Pfr then initiates signal transmission. Recently, we and others provided important insights by determining the structure of the bilin-binding photosensory domain as Pr^3-7. These models showed that the bilin is cradled within the GAF domain crevice, revealed a figure-of-eight knot that connects the PAS and GAF domains, identified a dimerization contact between adjacent GAF domains in the homodimer, and discovered a hairpin projection from the PHY domain that helps seal the chromophore pocket from the solvent. Unfortunately, these models have not fully illuminated how Pfr is generated. A long held notion is that the initial photochemistry involves a Z to E isomerization of the C15=C16 methine bridge which concomitantly rotates the D pyrrole ring^9-13. Specific protein conformational changes have also been proposed from the structural analyses of an unusual phytochrome variant that prefers Pfr as the ground state, but whether these movements pertain to canonical phytochromes remains speculative^7,14.

To better understand photoconversion, we used NMR spectroscopy to generate companion high resolution Pr and Pfr structures of the GAF domain from the phytochrome SyB-Cph1 obtained from the thermotolerant cyanobacterium Synechococcus OSB’. This fragment efficiently assembles with its native chromophore phycocyanobilin (PCB) to generate a chromoprotein with near full Pr/Pfr photochromicity^4,15. NMR spectra were collected without illumination with the chromoprotein as Pr and during continuous red light irradiation, which produced an equimolar mixture of Pr and Pfr. By comparing the results to our previous SyB-Cph1(GAF) Pr structure⁴, we generated a highly refined solution structure of Pfr (Protein Data Bank (PDB) code 2KLI) and an improved solution structure of Pr (PDB code 2KOI) with structured backbone root mean square deviations of 0.44 Å and 0.30 Å, respectively.

The backbone conformation of the SyB-Cph1 GAF domain as Pfr is similar to that as Pr, indicating that the overall shape of this domain does not change dramatically during photoconversion (Fig. 1). However, photoinduced movements were obvious for the bilin and a number of amino acid side chains. In contrast to our previous report⁴, the refined Pr structure showed that the PCB A pyrrole ring is nearly perpendicular to the B and C rings, with the A-ring carbonyl now pointing away from the thioether linkage to Cys138 (Fig. 2a,b). Upon photoconversion to Pfr, the orientations of the B, C and D rings are unchanged. Instead, we found in the ensemble of Pfr conformers that the A ring becomes nearly co-planar with the B and C rings, implying a ~90° rotation around the C4=C5 bridge during photoconversion (Fig. 2b). The thioether linkage to PCB is also contorted, which is supported by the fact that the Cys138 β carbon displays the largest chemical shift change during photoconversion (-4.6 ppm (Suppl. Fig. 1)). Most NMR signals from PCB exhibited considerable broadening in Pfr, suggesting increased mobility relative to the more rigid Pr state (Fig. 2c,d and Suppl. Fig. 2).

Figure 1. Three-dimensional overlay of SyB-Cph1(GAF) Pr and Pfr solution structures.

The unstructured first 30 amino acids and C-terminal 6His-tag are not shown. a, Aligned Pr (grey) and Pfr (magenta) superimpositions of the protein backbone from their respective 20 lowest energy conformers. The PCB chromophore is shown in grey for Pr and cyan for Pfr. b, Same as a, but only the lowest energy conformer for Pr (grey) and Pfr (magenta) are shown. The α helices and β strands are labeled.

Figure 2. Rotation of the A ring of the PCB chromophore during Pr to Pfr photoconversion.

a, Schematic diagram of PCB with numbered carbons for reference. b, Five lowest energy NMR conformers of PCB shown as a bundle in Pr (grey, top), Pfr (cyan, middle), and Pr/Pfr superimposition (bottom). The lines illustrate the direction of the A-ring carbonyl for each conformer. The indicated angle represents the Pr to Pfr rotation of the A ring for the lowest energy conformers. c, Two-dimensional ¹H-¹⁵N heteronuclear single quantum coherence spectrum of SyB-Cph1(GAF) with ¹³C incorporated into PCB carbons 2¹ ,3², 7¹, 8², 12², 13¹, 17¹, and 18² as Pr (blue) and following saturating red light irradiation (red, mixture of Pr and Pfr). d, Two-dimensional nuclear Overhauser effect spectrum of the sample in c showing ¹Hx/¹Hy crosspeaks from Pr (blue) and the mixture of Pr and Pfr (red). Only the C3¹ and C7¹ methyl carbons attached to PCB rings A and B (circled) show chemical shift changes upon photoconversion to Pfr whereas the chemical shifts for the C13¹, C17¹, C18² methyl carbons attached to the C and D rings are unaffected.

Although prior studies proposed that the D ring rotates during phototransformation^9-11,13, our NMR analyses of SyB-Cph1(GAF) failed to detect significant chemical shift changes for this ring during photoconversion. For example, various NMR spectra for the D-ring C17¹ and C18² methyls, amide, the pyrrole nitrogen, and C18 failed to detect Pfr signals distinct from Pr, nor did the immediate neighboring C13¹ methyl of the C ring, whereas differences in and around the environment of the A ring were obvious (Fig. 2c,d, Suppl. Fig. 2, and ref.15).

Rotation of the A ring of SyB-Cph1(GAF) is accompanied by conformational changes of several amino acids proximal to PCB, including Asp86, Tyr142, Phe82, Tyr54, His139, His169, Arg101 and Val100. Previous structural studies of Pr showed that the N^δ1 nitrogen of His139 contributes to a complex hydrogen bond network, involving the A-C ring nitrogens and a centrally positioned pyrrole water which together participate in the protonation cycle of the bilin during photoconversion, whereas the N^ψ1 nitrogen of His169 hydrogen bonds with the C19 carbonyl oxygen to stabilize the D ring^4-6,16,17. In Pfr, both these interactions are disrupted; the imidazole rings of His139 and His169 are rotated away from the pyrrole water and the D ring, respectively (Fig 3a-c). The position of His169 in Pfr is stabilized by displacement of strand β6 toward strand β1, leading to the formation of a new set of hydrogen bonds involving His170 with Tyr176 and Thr48 (Suppl. Fig. 3). Collectively, these changes likely alter the environment of the pyrrole water and thus the bilin photocycle^16-19, a possibility supported by our observations that the Pfr forms of Tyr176-Phe, His169-Ala, and Thr48-Ala mutants thermally revert more rapidly back to Pr (Suppl. Fig. 4).

Figure 3. Light-driven conformational changes for amino acids surrounding the chromophore.

a-c, Lowest energy structures for the PCB binding pocket as Pr (a), Pfr (b), and a superimposition of the two (c) highlighting conformational changes of relevant sidechains. The pyrrole water (pw) was modeled by aligning the available crystal structures for Deinococcus radiodurans BphP (PDB codes 1ZTU and 2O9B⁵,⁶) and Synechocystis Cph1 (PDB code 2VEA³) with the SyB-Cph1(GAF) solution structures. Straight arrows indicate the direction of the A-ring carbonyl of PCB in Pr and Pfr. Curved arrows indicate the directions of His139 and His169 sidechain movements. d-f, Ribbon representation for the region surrounding Arg101 as Pr (d), Pfr (e), and a superimposition of the two (f), highlighting the rotation of Val100 and Arg101 during photoconversion. Phe95 is shown in green, Val100 in yellow, Arg101 in orange, and Gln185 in purple. The 180° rotation of Val100 and Arg101 during Pr to Pfr photoconversion allows Arg101 to contact helix α5 at Gln185. Helix α2 and Phe95 move toward PCB to potentially block Arg101 from swiveling back to its Pr position. In all panels, the A-D pyrrole rings are indicated. Dashed lines highlight potential electrostatic interactions.

A second set of rearrangements during photoconversion involves Phe82, Tyr54, Asp86, and Tyr142 near the A and D rings of PCB (Figs. 3a-c and 4a,b). The Phe82 aromatic ring rotates ~30° to assume a parallel displaced orientation relative to the PCB D ring that could enable hydrophobic π stacking interactions (Fig. 4e,f). Movement of Phe82 eliminates a hydrogen bond between its main chain nitrogen and the hydroxyl of Tyr54, a conserved residue that helps avoid non-productive fluorescence of some Phys during photoexcitation^15,16,20. Mutant analysis shows that both Phe82 and Tyr54 are required for Pfr formation and stability (Suppl. Fig. 4). Rotation of the A-ring nitrogen to its position in Pfr is stabilized by a new hydrogen bond with the main chain oxygen of Asp86. Subsequent motion of the Asp86 side chain then leads to a new hydrogen bond network with the hydroxyl of Tyr142 and the D-ring carbonyl, the importance of which is confirmed by the aberrant photochemistry of a Tyr142-Phe mutant and several Asp86 substitutions (Suppl. Fig. 4 and ref.15). Collectively, these movements help stabilize the D ring (in addition to its contact with Lys52), and decrease the solvent accessibility of the Pfr chromophore (Fig 4c,d). Given that the carboxylate group of Asp86 is predicted to form a double salt bridge with a conserved arginine located in the PHY domain hairpin^3,7, movement of Asp86 likely affects this contact as well.

Figure 4. Conformational rearrangement of Asp86, Tyr142 and Phe82 during Pr to Pfr photoconversion.

a and b, Ribbon representation of the structures of SyB-Cph1(GAF) as Pr (a) and as Pfr (b) illustrating the conformational changes combined with the upshift of helix α2 (green) during photoconversion. The small arrow indicates the orientation of the A-ring carbonyl. The large arrow in b highlights the displacement of helix α2. Proposed hydrogen bonds between Tyr142, Asp86 and the D ring nitrogen are indicated by the dashed lines. c and d, Changes in solvent accessibility for PCB in Pr (c) versus Pfr (d). The green and red areas show the solvent exposed surfaces of helix α2 and the Asp86/Tyr142/Phe82 triad, respectively. e and f, Predicted π-stacking of PCB with Phe82 (red) in the Pr (e) and Pfr states (f). Rotation of the aromatic ring of Phe82 upon Pfr formation generates a parallel displaced orientation that favors π stacking interactions with PCB. In all panels, PCB is shown in cyan, the A-D pyrrole rings are labeled, and the key amino acids are in red.

Perhaps the most dramatic change in the SyB-Cph1(GAF) domain during photoconversion involves movement of Arg101 (Fig. 3d-f). In Pr, Arg101 forms a double salt bridge with the carboxylate of the B-ring propionate, but in Pfr, strand β4 is disrupted, and Arg101 and Val100 swivel approximately 180° to encourage a salt bridge between Arg101 and Glu185 in helix α5 (Fig. 3d-f and Suppl. Fig. 5). Concomitant with this rotation is a 2.6 Å displacement of helix α2 toward the B-ring propionate, thus allowing Phe95 to fill the void left by Arg101. Previous mutagenic analyses of Arg101 and a comparable Arg in Arabidopsis PhyB revealed a critical role for this residue in Pfr stability and signaling^4,22, whereas mutagenic studies of both Gln185 and Arg101 support the importance of their contact in Pfr (Suppl. Fig. 4). For example, the Gln185-Ala and Gln185-Glu mutants of SyB-Cph1 thermally revert from Pfr to Pr slower than wild type, whereas the Gln185-Arg and Arg101-Ala mutants revert much faster, with the Gln185-Arg mutant also displaying aberrant absorption spectra. Collectively, the new Pfr contact between Arg101 and Gln185 appears to adjust the position and/or flexibility of helix α5, as detected by notable chemical shift changes for several neighboring residues (e.g. Val184) and the unusual absence of NMR signals in Pfr from helix α5 (e.g., Gln178, Glu179, Glu180, Leu181, and Gln185), which participates in Phy dimerization^6,7.

Taken together, the structural differences between Pr and Pfr in SyB-Cph1(GAF) combined with the photochemical importance of a number of key conserved residues (Suppl. Fig. 6) offer a possible model for phytochrome photoconversion. An unexpected feature is the substantial rotation of the A ring, presumably driven by a C4=C5 isomerization, and a contortion of its thioether linkage to the protein instead of the former proposal that the D ring rotates^9-13. Either isomerization or relaxation of the strained C4=C5 bridge during Pfr formation could account for the red-shifted absorption spectrum of Pfr by increasing the coupling of the π-conjugation system. Further red-shifting by π-stacking interactions with aromatic residues neighboring the D ring could help explain the Pfr/Pr chemical shift differences reported previously for the D ring²³. Rotation of the A ring is also supported by photochemical studies with sterically locked bilins^24,25 and by prior NMR spectra of phytochrome fragments also containing the PAS and PHY domains^26,27, which strongly suggest that our results with SyB-Cph1(GAF) are not unique to this phytochrome nor artifactually generated by analysis of just the GAF domain. We note that the x-ray crystallographic structures of several Phys as Pr^3-5,7,28 have modeled with a more coplanar configuration for the A ring relative to the B and C rings than seen here for SyB-Cph1 in solution. These differences could reflect subtle variations among Phys, inherent differences in the environment of the chromophore in crystals versus in solution, and/or radiation-induced damage of the Pr bilin during x-ray data collection that could relax the strain of a non-planar A ring^5,28.

We propose that, subsequent to rotation of the A ring, a series of reversible conformational movements occur within the bilin-binding pocket that support the deprotonation/protonation cycle of the bilin, stabilize the Pfr form, and finally adjust several contact sites on the surface of the GAF domain. In particular, movement of the Asp86 and Tyr142 pair could affect the non-covalent interaction of the PHY domain with the GAF domain through its hairpin projection, which could then reorient by a hinge mechanism these domains relative to each other. The swivel of Arg101 to contact Gln185 concomitantly reorients and/or destabilizes helix α5. Given the role of helix α5 in helping sister phytochromes dimerize and in covalently connecting the GAF and PHY domains^3,6,7, even a subtle movement/unfolding of this helix might have profound consequences on intermolecular GAF/GAF dimerization and intramolecular GAF/PHY contacts.

Taken together, it is conceivable that such light-induced rearrangements then initiate a cascade of events within the phytochrome dimer that reorient the C-terminal output modules relative to each other and to the photosensory modules. For phytochromes bearing histidine kinase output modules, such light-driven rearrangements could then alter autophosphorylation in trans across the phytochrome dimer. In this manner, phytochromes may resemble the phototropin family of photoreceptors, which couples flavin photochemistry to selective destabilization of a helical contact adjacent to the photosensory domain and finally to activation of the appended output kinase²⁹. Because the PYP family of photoreceptors may work by a similar light-triggered conformational switch³⁰, our model for phytochromes provides further support for the notion that light-induced conformational changes are fundamental for photoactivated signaling.

METHODS SUMMARY

The GAF domain of Synechococcus OSB’ Cph1 bearing a C-terminal 6His tag was expressed and assembled with PCB in Escherichia coli and purified as described previously^4,15. The protein moiety was uniformly labeled by using expression medium containing ¹⁵NH₄Cl and [¹³C]-glycerol and an excess of unlabelled bilin precursor α-aminolevulinic acid (ALA). PCB was selectively labeled by adding [U-¹⁵N]-ALA, [3-¹³C]-ALA, [4-¹³C]-ALA, [1,2-¹³C]-ALA, or [U-¹⁵N;U-¹³C]-ALA to the medium. NMR spectra were collected with a ~50:50 equilibrium mixture of Pr and Pfr that was obtained by continuously irradiating the samples with red light¹⁵. Three-dimensional structures of Pr and Pfr were generated by a suite of NMR analyses using the prior model of SyB-Cph1(GAF) as a reference (PDB code 2K2N)⁴. Statistical support for the Pfr structure is present in Suppl. Table 1. The structures are also supported by the photochemical analyses of various SyB-Cph1(GAF-PHY) mutants generated by site-directed mutagenesis (Suppl. Fig. 4).

Appendix

METHODS

Chromoprotein production

The first 200 amino acids of Synechococcus OSB’ (SyB) Cph1 bearing a C-terminal 6His tag were expressed, assembled with PCB in Escherichia coli, and purified using the dual-plasmid recombinant system described previously³¹,³². Even with its lack of an N-terminal PAS domain, SyB-Cph1 shares a number of photochemical properties with other members of the phytochrome superfamily, including red/far-red light photochromicity, the preference of Pr as the ground state, the extended configuration of the bilin in Pr, the protonation of the bilin as both Pr and Pfr, a similar fold of its GAF domain, and the importance of conserved amino acids lining the bilin-binding pocket³¹,³². Samples used for backbone, sidechain, and nuclear Overhauser effect (NOE) data collection were expressed in medium containing ¹⁵NH₄Cl and [¹³C]-glycerol and excess of the bilin precursor α-aminolevulinic acid (ALA) unlabeled. PCB was selectively labeled by adding [U-¹⁵N]-ALA, [3-¹³C]-ALA, [4-¹³C]-ALA, [1,2-¹³C]-ALA, or [U-¹⁵N;U-¹³C]-ALA (Sigma, St. Louis, MO) to the medium containing unlabeled NH₄Cl and glycerol.

Various site-directed mutants affecting the GAF-PHY fragment of SyB-Cph1 (residues 1-421) bearing a C-terminal 6His tag were generated by the QuickChange method (Stratagene, La Jolla, CA). PCB incorporation was assayed by zinc-induced fluorescence of the chromoproteins following SDS-PAGE³². The zinc-impregnated gels were irradiated with 260 nm light with the fluorescence emission detected in the visible region of the light spectrum. Pr and Pfr absorption and difference spectra, and rates of Pfr to Pr thermal reversion were measured at 55 °C as previously described³².

NMR Data Collection

For Pfr data collection of the SyB-Cph1(GAF) sample, an equilibrium mixture of Pr and Pfr was obtained by irradiating the Shigemi microcell solution with saturating red light and then maintaining this equilibrium during data collection by continuous irradiation with low fluence red light as described³². Unless noted otherwise, all NMR spectra were recorded on 800 and 600 MHz Varian INOVA spectrometers equipped with cryogenic probes. Samples used for backbone, sidechain, and nuclear Overhauser effect (NOE) experiments contained 1.7 mM [U-¹⁵N;U-¹³C]-SyB-Cph1(GAF) assembled with unlabelled PCB in 10 mM deuterated Tris-HCl (pH 8.5) and 0.03% NaN₃ in 93% H₂O/7% D₂O. Protein ¹D_NH and ¹D_Cα_Hα residual dipole couplings (RDCs) were recorded with 1 mM [U-¹⁵N;U-¹³C]-SyB-Cph1(GAF) in anisotropic medium containing an axially stretched (5.4 mm to 4.2 mm) negatively charged acrylamide gel³³ or supplemented with sodium dodecylsulfate-doped ditetradecyl-phosphatidylcholine/dihexyl-phosphatidylcholine bicelles (molar ratio 1:30:10¹⁰). PCB 3-D ¹³C NOE data and isotropic methyl ¹J_CH couplings were collected as previously described³¹. A similar sample in anisotropic medium containing 0.5 mM protein and 15 mg/mL filamentous pf1 phage (ASLA Biotech, Riga, Latvia) was used to measure PCB methyl RDCs (¹D_CH). 1-D ¹⁵N-direct detected, 2-D ¹⁵N-heteronuclear single quantum coherence (HSQC), and ¹⁵N NOE experiments used a 1 mM sample of unlabeled protein assembled with [U-¹⁵N]-PCB³¹.

All NMR data were collected at 25 °C, except for the measurements of bicelle RDCs, which were collected at 33 °C. Distance constraints were obtained from 3-D ¹⁵N-edited NOESY (t_mix=150 ms) and 3-D ¹³C-edited NOESY (t_mix=120 ms) experiments. NH and C^αH^α dipolar couplings were measured from a 3-D HNCO antiphase ¹H-coupled in the ¹⁵N dimension³⁴ and a 3-D HCA(CO)N antiphase ¹H-coupled in the ¹³C dimension, respectively. PCB methyl RDCs were obtained from a J-modulated [¹H-¹³C] HSQC spectrum³⁵ recorded with a 600 MHz Bruker DMX-Avance spectrometer.

Resonance Assignments and Secondary Structure Calculations

¹⁵N T₂ measurements with 1.7 mM [U-¹⁵N; U-¹³C]-SyB-Cph1(GAF) in Pr form yielded uniform values around 45 ms for the rigid part of the molecule³¹. Backbone assignments were obtained and assigned manually as described³¹. The TALOS program³⁶ provided 152 pairs of ϕ/ψ backbone torsion angle restraints and identified the secondary structure, which was confirmed by local NOEs. Hydrogen bond restraints were inferred initially for α-helices and later for β-strands when the level of structural refinement allowed their unambiguous alignment within the β-sheet. Two distance restraints of 1.9 Å and 2.9 Å per involved pair of residues were used to represent hydrogen bonds for H^N-O and N-O, respectively³⁷. To help assign NMR peaks to the Pr and Pfr forms, we classified them into three groups: peaks belonging to atoms with distinct Pfr assignments whose chemical shifts matched those determined previously for Pr, indicating that they correspond to the Pr component of the mixture³¹, peaks belonging to atoms without distinct Pfr assignments whose chemical shifts matched those for Pr, indicating that they correspond to atoms whose environment is unaffected by photoconversion, and peaks with chemical shifts different from those observed for Pr and thus unique to the Pfr form. Backbone signals within the three groups were extended to side chains by reference to NOE and HCCH-TOCSY spectra, and the three-group classification was updated manually.

For chromophore assignments, PCB synthesized with ¹⁵N or ¹³C-labelled ALA was used to unambiguously identify almost all protons, the ring D nitrogen, and carbonyl carbons based on NOE contacts among themselves and with unambiguously assigned protein protons. The only exceptions were the CH₂ moieties in the propionate chains, which are known to be mobile and not observable due to conformational broadening of their NMR signals. Protons 5-H, 15-H, 2-H, 3-H, 3′-H, and 18¹-CH₂ were assigned from ¹³C-HSQC and ¹³C-NOE spectra based on multiple NOE contacts with PCB methyls and by direct detection (Suppl. Fig. 2); their assignments were also confirmed by a multitude of intra PCB NOE contacts among themselves and with 2¹, 3², 7¹, 13¹ and 17¹ methyl groups. The proton and carbon assignments are in excellent agreement with NMR assignments of PCB published previously³⁸,³⁹.

3-D Structure Calculations and Refinements

Structure calculations and refinements made use of the torsion angle molecular dynamics and the internal variable dynamics modules of Xplor-NIH⁴⁰. A separate structure calculation run (100 conformers) was used to identify and generously constrain those side chain dihedral angles that exhibited a unique rotameric state in more than 90% of the conformers. PCB topologies and parameters were generated by the Dundee PRODRG2 Server⁴¹. Peak intensities in 3D NOESY spectra of the protein were assigned using the PIPP/STAPP package⁴² and converted into a continuous distribution of 1118 approximate interproton distance restraints, with a uniform 40% distance error applied to take into account spin diffusion. PCB distance restraints were obtained by classifying the NOE peak intensities into 3 bins (strong, medium, and weak) (Suppl. Fig. 7). We attempted to measure RDCs in more than one alignment medium. To accommodate consistent measurements at pH 8.5, hydrolysis-resistant dialkyl analogs⁴³ of the traditional DMPC:DHPC bicelles were exploited⁴⁴. Structural restraints were determined with negatively charged bicelle ¹D_NH and ¹D_Cα_Hα RDC data. We also used as structural constraints the RDCs of the PCB methyls from protein aligned with pf1 phage. The ¹D_NH and ¹D_Cα_Hα RDC sets measured in pf1 were used to obtain the protein alignment tensor in this medium, but not as direct structural constraints.

The final NMR structure for Pfr (PDB code 2KLI) and the further refined structure for Pr (PDB code 2KOI; original code 2K2N) do not include the C-terminal 6His tag and the first 30 amino acids. The Pr structure was improved by adding another set of protein NH and PCB methyl RDCs in negatively charged stretched gels and 187 ¹⁵N and 131 ¹³C additional protein NOEs. We also assigned additional PCB protons (5-H, 15-H, 2-H, 3-H, 3¹-H and 18¹-CH₂) and their associated NOE contacts with the protein and among themselves (44 new PCB NOEs), which resulted in a consistent tilted, out-of-place orientation of ring A in Pr. The new ensemble of refined Pr structures has an improved heavy atom root mean squared deviation over the structured backbone regions (0.30 Å from 0.43 Å) and a slightly improved agreement with all measured RDCs.

From 100 refined conformers, a subset of 20 low-energy conformers was selected to represent the Pr and Pfr solution structures. These Pr and Pfr conformers have a root mean square deviation of 0.30 Å and 0.44 Å for backbone heavy atoms over the most structured regions (residues 31–110, 136–171, and 183–202), respectively. The final set of conformers had no NOE constraint violations greater than 0.5 Å in more than 40% of the calculated models. PROCHECK⁴⁵ analysis of the structured region in the 20 lowest energy Pfr conformers showed that 88.6% of the residues were within the most favored region, 11.3% in the allowed region, and 0.1% in the disallowed region of the Ramachandran map. A summary of the agreement between experimental constraints and calculated Pfr structures is provided in Supplemental Table 1.