Dynamic Structural Changes Underpin Photoconversion of a Blue/Green Cyanobacteriochrome between Its Dark and Photoactivated States*

Background: Phytochromes are photochromic bili-proteins vital to microbial and plant photoperception. Results: NMR spectroscopy generated three-dimensional structures of the photosensing module from a cyanobacterial variant in the dark and photoactivated states. Conclusion: Photoconversion involves thioether bond rupture, bilin isomerization and sliding, and increased protein disorder. Significance: Combined with crystallographic models, these paired NMR structures provide an unprecedented view into photoconversion of a phytochrome-type photoreceptor. The phytochrome superfamily of photoreceptors exploits reversible light-driven changes in the bilin chromophore to initiate a variety of signaling cascades. The nature of these alterations and how they impact the protein moiety remain poorly resolved and might include several species-specific routes. Here, we provide a detailed picture of photoconversion for the photosensing cGMP phosphodiesterase/adenylyl cyclase/FhlA (GAF) domain from Thermosynechococcus elongatus (Te) PixJ, a member of the cyanobacteriochrome clade. Solution NMR structures of the blue light-absorbing dark state Pb and green light-absorbing photoactivated state Pg, combined with paired crystallographic models, revealed that the bilin and GAF domain dynamically transition via breakage of the C10/Cys-494 thioether bond, opposite rotations of the A and D pyrrole rings, sliding of the bilin in the GAF pocket, and the appearance of an extended region of disorder that includes Cys-494. Changes in GAF domain backbone dynamics were also observed that are likely important for inter-domain signal propagation. Taken together, photoconversion of T. elongatus PixJ from Pb to Pg involves complex structural changes within the GAF domain pocket that transduce light into a mechanical signal, many aspects of which should be relevant to others within the extended phytochrome superfamily.

The phytochrome superfamily of photoreceptors exploits reversible light-driven changes in the bilin chromophore to initiate a variety of signaling cascades. The nature of these alterations and how they impact the protein moiety remain poorly resolved and might include several species-specific routes. Here, we provide a detailed picture of photoconversion for the photosensing cGMP phosphodiesterase/adenylyl cyclase/FhlA (GAF) domain from Thermosynechococcus elongatus (Te) PixJ, a member of the cyanobacteriochrome clade. Solution NMR structures of the blue light-absorbing dark state Pb and green light-absorbing photoactivated state Pg, combined with paired crystallographic models, revealed that the bilin and GAF domain dynamically transition via breakage of the C10/Cys-494 thioether bond, opposite rotations of the A and D pyrrole rings, sliding of the bilin in the GAF pocket, and the appearance of an extended region of disorder that includes Cys-494. Changes in GAF domain backbone dynamics were also observed that are likely important for inter-domain signal propagation. Taken together, photoconversion of T. elongatus PixJ from Pb to Pg involves complex structural changes within the GAF domain pocket that transduce light into a mechanical signal, many aspects of which should be relevant to others within the extended phytochrome superfamily.
Biological light perception is mediated by a collection of photoreceptors that initiate various signaling cascades upon photoactivation. One influential set in plants and microorganisms is the phytochromes, a superfamily of bilin (or open-chain tetrapyrrole)-containing chromoproteins that employ photoreversible conformational changes in the bilin to monitor the ambient light environment (1,2). The bilin is cradled within a signature cGMP phosphodiesterase/adenylyl cyclase/FhlA (GAF) 3 domain through numerous electrostatic and van der Waals interactions and fixed by a cysteine-based thioether linkage involving the C3 side chain on pyrrole ring A (see Fig. 2).
Phytochromes assume two photointerconvertible states, typically a dark Pr state that absorbs red light, and a metastable Pfr state that absorbs far red light and is generated only upon red light illumination. These unique spectral properties are created by a number of conserved chromophore/GAF domain contacts, including an extensive hydrogen-bonding network involving a centrally positioned "pyrrole water" and the four pyrrole ring nitrogens (3)(4)(5)(6)(7)(8). It has long been assumed that the Pr 3 Pfr photocycle is triggered by a Z to E isomerization of the bilin at the C15AC16 double bond that flips the D pyrrole ring (see Refs. 9 and 10), but structural and biochemical studies indicate that other routes might also exist for some phytochrome subtypes, including light-driven rotation of the A pyrrole ring (11) and light-induced changes in bilin protonation (12). These light-driven alterations in the chromophore then induce conformational changes within the protein moiety that ultimately allow phytochromes to act as long-lived photoswitches.
Numerous signaling outputs are possible; the most common initiates two component-type phosphorelays (1,2).
In addition to the canonical red/far red light-absorbing phytochromes, photosynthetic bacteria are populated with variants with peak absorptions that span near UV light and much of the visible spectrum despite incorporating the same or similar bilin (13). This diversity is particularly widespread in cyanobacteria, where a unique phytochrome subclade, called cyanobacteriochromes (CBCRs), assembles with phycocyanobilin (PCB) to acquire reversible end states ranging from violet/blue to green/ red (e.g. Refs. 12 and 14 -16). CBCRs possess two key properties helpful to investigations on phytochrome photochemistry. In many cases the GAF domain alone is sufficient for their unique absorption and photoconversion, and for at least some, photoexcitation of the dark state leads to near complete conversion to the activated state (14,15,17). These collective properties are not typical among more canonical phytochromes for which additional domains are often photochemically essential (i.e. Per/Arndt/Sim (PAS) and PHY domains) and where ground state contamination is unavoidable due to spectral overlap (2). Expanding studies on CBCRs revealed that multiple photocycles are possible, including the use of a second thioether link to spectrally tune light absorption (15)(16)(17)(18).
One CBCR that has attracted considerable interest is PixJ from the thermotolerant cyanobacterium Thermosynechococcus elongatus (Te), which photointerconverts between a blue light-absorbing dark state Pb (428 nm maximum) and a green light-absorbing photoactivated state Pg (529 nm maximum). Its hypsochromically shifted spectra are generated by an isomerase activity intrinsic to the GAF domain that transposes the C4AC5 double bond within the bound PCB to the C2-C3 position, thus generating the bilin variant phycoviolobilin (PVB) (19,20). In addition to the conventional linkage involving the C3 1 carbon and Cys-522, the Pb state of TePixJ links PVB via a second thioether bond between the C10 carbon and Cys-494 (15,20,21). The end result is a bisected bilin -conjugation system that yields a blue light-absorbing system involving the C and D pyrrole rings and a UV-absorbing system derived from the B ring.
Comparisons between recent x-ray crystallographic structures of Pb and Pg also implied that blue-light irradiation drives D-ring rotation via Z-to E-isomerization of the C15AC16 double bond and concomitant rupture of the C10/Cys-494 linkage to change the geometry of C10 from tetrahedral to planar (21,22). However, interpretation of the paired crystal models is challenged by potential constraints imposed upon TePixJ(GAF) in the crystalline state, especially with regard to a non-native disulfide bridge between GAF domains within the asymmetric unit of the Pg crystal (22), and a general lack of information regarding the dynamics of the GAF domain.
To overcome these challenges, we determined the solution structures of TePixJ(GAF) for both the Pb and Pg states using high resolution NMR spectroscopy. By comparing our NMR models to those previously generated by x-ray crystallography (21,22), a more complete picture of photoconversion for a CBCR GAF domain is now possible. It shows that transformation between Pb and Pg includes Z to E photoisomerization about the C15AC16 double bond, a photolabile thioether link-age, sliding of the bilin within the GAF domain pocket concomitant with extensive changes in chromophore/protein contacts, and multiple effects on protein disorder, which collectively should be relevant to how TePixJ signals to downstream effectors. The paired dark (Pb) and photoactivated state (Pg) structures shown here likely reveal common themes for phytochrome photochemistry as well as emphasize the striking diversity within the superfamily.
Resonance Assignments and Secondary Structure Calculations-Data were processed with the NMRPipe package (24). For the respective Pb and Pg states, 91 and 84% of the backbone resonances were assigned manually using the PIPP/STAPP software (25). The TALOSϩ program (26) was used to provide 147 and 136 pairs of / backbone torsion angle restraints for the Pb and Pg states, respectively, and to identify the secondary structural elements, which were confirmed by local NOEs. The protein and chromophore NOE statistics for Pb and Pg are presented in Tables 1 and 2, respectively. Hydrogen bond restraints were inferred initially for ␣-helices and later for ␤-strands when the level of structural refinement allowed their unambiguous alignment within the ␤-sheet. Distance restraints of 1.9 and 2.9 Å per involved pair of residues were used to represent hydrogen bonds for H N -O and N-O, respectively (27).
Three-dimensional Structure Calculations and Refinements-15 N T 2 measurements with 1.2 mM [U-13 C, 15 N]Te-PixJ(GAF) (see below) yielded average values around 55 ms for the rigid part of the molecule, indicating that the protein behaved as a monomer. In support, the line widths of 1 H, 15 N HSQC crosspeaks were independent of protein concentration from 0.3 to 1.3 mM. Peak intensities in three-dimensional NOESY spectra were assigned using the PIPP/STAPP package and converted into a continuous distribution of 2847 and 2444 approximate interproton distance restraints for Pb and Pg, respectively, with a uniform 40% distance error applied to take into account spin diffusion. In addition, 163 and 62 chromophore distance restraints (Pb and Pg, respectively) were extracted from NOESY spectra of the labeled chromophore/unlabeled protein samples. Structure calculations and refinements made use of the torsion angle molecular dynamics and the internal variable dynamics modules of XPLOR-NIH (28). The Dundee PRODRG2 Server (29) was used to generate the PVB topology and parameters. A separate structure calculation run (100 structures) was used to identify and generously constrain side chain dihedral angles if these were consistent with a unique rotameric state in Ͼ90% of the structures. The magnitude and rhombicity of the alignment tensors and the agreement with NMR and x-ray structures are presented in Table 3. PyMol (Delano Scientific, LLC) and VMD-XPLOR (30) were used to prepare the structure figures.

RESULTS AND DISCUSSION
Solution NMR Analysis of TePixJ(GAF)-Here, we examined a TePixJ(GAF) fragment encompassing residues 430 -591 and bearing the Cys-555 to Ala mutation, which was assembled recombinantly with PCB. Previous studies showed that this preparation is highly stable, fully photochromic, and amenable to photostate-specific structural characterization by NMR spectroscopy after isotopic labeling of either the bilin or protein moieties (15,21). To counteract Pg 3 Pb thermal reversion (21), the Pg samples were irradiated continuously with low fluence rate blue light during data collection. Pg retention and the absence of photoconversion intermediates were ensured by periodic inspection of 1 H, 15 N HSQC spectra collected during and after irradiation.
By using an assortment of two-and three-dimensional NMR spectroscopic methods, including measurement of RDCs, we solved the solution structures of TePixJ(GAF) in its Pb and Pg states (PDB ID codes 2M7U and 2M7V, respectively). Statistical support for the two structures is found in Tables 1 and 2, and agreement of the structures with experimental RDCs is shown  in Table 3. Like other phytochrome-type GAF domains (2), TePixJ(GAF) consists of an arched six-stranded ␤-sheet (␤1-␤6) in a 3-2-1-6-5-4 ␤-strand sequence, which together with ␣-helices ␣2 and ␣4 create the bilin binding pocket, and two anti-parallel ␣-helices (␣1 and ␣5) that extend to other domains and might promote dimerization (Fig. 1, A-C). With the exception of local changes near strand ␤3 in the Pg state (see below), the respective protein folds of the Pb and Pg states retained high similarity to each other and to the pair deduced by x-ray crystallography (PDB ID codes 4GLQ and 3VV4, respectively (21, 22)) ( Fig. 1, B-D) Despite their overall similarity, close comparison of the Pb and Pg structures generated by x-ray crystallography identified several significant differences in backbone positions (Fig. 1D). These were confirmed in the NMR structures (Fig. 1C) and supported by fitting of the Pb and Pg state experimental RDCs to the NMR and x-ray crystallographic models (Table 3). For example, the R free factor of Pb RDCs was lower when fitted to the x-ray Pb structure than when fitted to the NMR or x-ray structures of Pg, whereas no significant differences in R free were observed when the NMR and crystallographic Pg models were compared and vice versa. Taken together, it appears that photoconversion of TePixJ, like that seen for more canonical phy-

Pb Pg
Pb/Pg (NMR) Pb/Pg (X-ray)  tochromes (10,11), does not appear to globally alter the GAF domain but acts to more subtly reposition existing structural elements. The RDC cross-validations also implied that the NMR and x-ray crystallographic structures are near equivalent in accuracy.
All other characterized phytochromes assemble as homodimers, with studies on several indicating a head-to-head orientation (2). Prior crystal structures of TePixJ(GAF) gave conflicting results, with one crystalline form packing as a head-to-tail dimer and the other two packing as head-to-head dimers, both via helices ␣1 and ␣5 (21,22). The crystal structure of Pg also included a surprising Cys-494 -Cys-494 disulfide bridge between separate GAF domains within the asymmetric unit (22). In contrast, our solution models of TePixJ(GAF) were monomeric based on several NMR parameters, including line widths of 1 H, 15 N HSQC cross-peaks at various concentrations, RDCs, and 15 N transverse relaxation (T 2 ) measurements (data not shown and see Fig. 7). Even though the full-length photoreceptor is likely a dimer, we concluded that the disulfide bridge and the associated solvent projection of residues 493-496 in the Pb model are crystallization artifacts, and that the contacts between sister GAF domains are insufficient to maintain the dimer by themselves even at high protein concentrations.
Bilin Movements during Photoconversion-The transition from Pb to Pg induced several distinct structural changes in the chromophore detectable by NMR. For the bilin as Pb, 83-13 C and 81-15 N NOE-derived distance constraints allowed relatively precise determination of its conformation and orientation within the GAF domain cavity (Fig. 2). As in the crystal model (21), the Pb chromophore was identified as a rubinoid derivative of PVB that has sp 3 hybridization at the C10 position and a Z configuration at the C15AC16 double bond (Fig. 3). In fact, the diagnostic C4 -C5 single bond for PVB was easily detected in the 13 C, 1 H HSQC spectra of 13 C, 15 N-labeled chromophore/unlabeled protein samples by the methylene signal at the C5 position (5.6 ppm/38.1 ppm). The composite five lowest energy models also revealed a substantially contorted chromophore, with the planes of the A and D rings oriented by 46°a nd 124°out of plane relative the B and C-ring planes, respectively, and the B and C rings oriented out of plane from each other by 34° (Fig. 3E). Cys-494 was modeled in close proximity to the bridge between the B and C rings by numerous NOE contacts (Fig. 2), which supported a second thioether linkage in Pb involving C10 and this cysteine (21).
Assignment of the 1 H, 15 15 N chromophore signals became vanishingly weak, most likely due to solvent exchange, and the chromophore appeared more mobile as evidenced by the resolution of substantially fewer chromophore 13 C, 1 H-and 1 H, 15 N NOE constraints (62 and 0, respectively), even after increasing the number of spectral scans to the practical limit (Figs. 2 and 3C). Depleted NOE constraints were especially evident for the A-C pyrrole rings. For the A ring, this ambiguity generated a varied ensemble of acceptable positions in the lowest energy solution structures, which is likely reflec-tive of the true solution state(s) of the photoreceptor (Fig. 3D). Comparisons of the crystal structures for TePixJ(GAF) supported the proposed break of the C10/Cys-494 link followed by reformation of the C10AC11 double bond (21,22). For our NMR studies, the lack of NOE signals unambiguously assigned to C4-H, C10-H, and C15-H prevented direct detection of this break (15). However, the break resulting in the Pg state was evident by the loss of all NOE contacts between Cys-494 and PVB and migration of Cys-494 away from C10 (Figs. 1A and 2). Retention of PVB as opposed to its possible re-isomerization back to PCB after photoconversion (16) was supported by the easily detected C5 methylene signal in Pg ( 1 H ϭ 4.2 ppm, 13 C ϭ 38.1 ppm) and by its NOE contact to the chromophore C3 2 methyl. Taken together, we conclude that photoconversion of TePixJ to Pg restores PVB, thus generating a bathochromically shifted -conjugated system that now encompasses the B, C, and D rings (Fig. 3A).
Our NMR analyses implied that the D pyrrole ring is more rigid than the A-C rings in the Pg state and fully supported photoisomerization about the C15AC16 bridge to generate the 15E configuration (Fig. 3, D and E). Photostate-dependent changes in the unambiguously assigned NOE contacts between ring D and the protein (Fig. 2) supported an averaged 148°flip embodied in our five lowest energy structures, which was consistent with large displacements of the methyl signals at the 13C 1 and 17C 1 positions (Fig. 3B). By comparison, superposition of the Pb and Pg crystal structures estimated a 165°D-ring flip (21,22). This 15Z to E isomerization is in agreement with the 15E photoproduct of PVB when bound to cyanobacterial photosynthetic antennae protein ␣-phycoerythrocyanin (31) and with the photoconversion mechanism proposed for a canonical phytochrome assembled with PCB (9). Photoconversion in solution also induced an ϳ71°swivel of the A ring based on the average of the five lowest energy NMR structures (Fig. 3,  D and E). The significance of the A ring movement is unclear, but its rotational freedom might optimize photoinduced sliding of PVB upon rotation of the D ring while still retaining the 3C 1 thioether linkage (see below).
Photoinduced Changes in the Chromophore Pocket-After Pb 3 Pg photoconversion, structural changes in PVB drive numerous alterations in chromophore/protein contacts. Both the solution and crystal structures show extensive displacement of the bilin that is likely induced by migration of the flipped D-ring (21,22). This migration can best be seen by the alignment of rings C and D with individual strands of the ␤-sheet (Fig. 4). In Pb, the D ring straddles the grooves between the side chains of the ␤1-␤6 and ␤5-␤6 strand pairs, whereas in Pg it slides into the ␤1-␤2 groove and the space between ␤2 and the displaced ␤3 strand. In Pb, the C-ring methyl and propionate groups sits atop the side chain ridges formed by strands ␤5 and ␤4, respectively, and slide into the grooves between the ␤1-␤6 and ␤5-␤6 strand pairs in Pg. Notably, Yang et al. (32) detected similar sliding of the bilin within the GAF pocket of the Pseudomonas aeruginosa bathyphytochrome BphP from comparison of its Pfr structure with that for the Q188L mutant, which appears to generate a Pr/Pfr-mixed state. These observations suggest that bilin sliding is a general mechanism for both CBCRs and more canonical phytochromes.
As with other structurally defined phytochromes, TePixJ includes an aspartate (Asp-492) nestled within the crescent configuration of the A-C rings. In canonical phytochromes, the main-chain carbonyl of this aspartate interacts with all three pyrrole nitrogens through the pyrrole water intermediary (3)(4)(5)(6)(7)(8), but in the crystallographic structure of TePixJ as Pb, the carboxylate group of Asp-492 is modeled to bind directly to the pyrrole nitrogens (Figs. 4 and 5; Ref. 21). Rotation of ring D during photoconversion dissolves the hydrogen bonds between its carbonyl and Asn-535 and His-553 to now allow a new contact between its pyrrole nitrogen and the carboxylate group of Asp-492, which is repositioned toward the D ring. This new contact apparently serves as an island of stability in Pg given that our NMR analyses showed increased mobility in the residues flanking Asp-492 (see below).
The combined movements upon photoconversion break direct contact of Asp-492 with the A-C rings. However, crystallographic data showed that the hydrogen-bond lattice is maintained indirectly upon entry of two water molecules into the void (22). The presence of these waters might explain in part the loss of bilin 1 H, 15 N NOE signals upon photoconversion that could arise from the new solvent exchange properties of the Pg state (Fig. 2). In the crystallographic structures, the positions of the waters are further stabilized by their hydrogen bonding with the carboxylate of Glu-497, the carbamoyl side chain of Gln-526, and the main chain carbonyl of Pro-493 (Fig. 5). Although Gln-526 moves little during photoconversion, bilin sliding allows its carbamoyl to hydrogen bond with the B-ring nitrogen. Given their central positions, it is not surprising that most Asp-492 and Gln-526 substitutions in TePixJ have abnormal blue/green photochemistry (15,21). The Q526H mutant in particular fails to form Pg, likely due to repulsion of the histidine imidazole group by the PVB pyrrole nitrogens. Collectively, the NMR and crystallographic data also indicate that the A ring is loosely held in the GAF cavity with the ring rotating and becoming more mobile as Pg (Fig. 3; Refs. 21 and 22). The three-dimensional structures of the Pb (gray) and Pg states (pink) were determined by x-ray crystallography (21,22). The relevant ␤-strands and amino acids are shown along with key hydrogen bond contacts (dashed lines). Cys-494 is located by an asterisk. Cyan, nitrogens; red, oxygens; orange and yellow, side-chain carbon and sulfur atoms of Cys-494 and Cys-522, respectively.
Both His-523 and Trp-499 show subtle but potentially important movements during photoconversion. In the Pb crystal structure (21), the imidazole ring of His-523 participates in an unusual cationinteraction with ring B, whereas Trp-499 is displaced away from the chromophore and fixed via a hydrogen bond between its indole NH and the B-ring propionate (Fig. 5). Upon rupture of the Cys-494 linker and sliding of PVB, both amino acids assume a closer and more parallel orientation relative to the bilin, suggesting a role for this displacement in finetuning Pg absorption by -stacking interactions (Ref. 22 and Fig. 5). Our NMR data support these movements and show Trp-499 to be relatively mobile in Pg as evidenced by the uncertainty in the indole ring position. The mobility of the Trp-499 indole ring might play a role in Pg 3 Pb photoconversion and thermal reversion, as the position of this side chain in Pg should impede access of the bilin C10 for nucleophilic attack by the Cys-494 thiol.
Finally, NOE-derived interatomic distances lend additional credence to the complex set of interaction changes between the bilin propionates and several surrounding amino acids (His-498, Trp-499, Arg-507, Gln-509, Tyr-503, and Asn-535) that might lock the bilin in place after sliding. Surprisingly, the string of solvent-excluded interactions extending across much of the GAF domain ␤-sheet (including His-553, Leu-530, Arg-459, Glu-476, Pro-483, and Ala-485) were only subtly affected by photoconversion in the ensemble of crystallographic and NMR structures after the electrostatic contact between the D ring and His-553 was broken by the 15Z to E isomerization (Fig. 5).
Photostate-dependent Changes in GAF Domain Architecture-Comparisons of the Pb and Pg solution structures identified a number of changes in the GAF domain upon photoconversion. Prominent main chain differences were seen at residues 485-506 and 512-522. In Pg, the region spanning residues 493-501 exhibits intermediate conformational exchange, resulting in complete broadening of most of its NMR signals. In fact, no Pg NOEs were detected in this region, resulting in a less structured NMR bundle (Fig. 1A). Residues 512-522 include Cys-522, which provides the thioether link with the C3 ethylidene side chain of ring A. Comparisons between the Pb and Pg crystal models suggested that this stretch undergoes a conformational change during photoconversion ( Fig. 1D; Refs. 21 and 22), a possibility implied by the observations that the L519S mutant has a highly unstable Pg state (21). Supported by several assigned NOE contacts (Fig. 2), residues 512-522 in our paired NMR models exhibited increased flexibility among the lowest energy structures and a downward displacement of helix ␣4 and proximal loop bundles toward the chromophore (Figs. 1C and  6, A and B).
In Pb, residues 485-506 encompass strand ␤3, helix ␣2, and a type II ␤-turn motif formed by residues 485-488 that connects strand ␤3 with a conserved loop motif (Figs. 1, A and C,  and 6A). Strikingly, overlay of the solution models revealed that upon Pb 3 Pg photoconversion the ␤3 strand (residues 488 -490) becomes largely unstructured, and in agreement with the paired crystal structures, part of the ␤-turn (residues 485-488) converts to a 3 10 -helix (Fig. 6A). This unusual change is likely driven by collision of the repositioned D ring with Ile-490 and is accompanied by disposition of the entire region toward helix ␣4 (Fig. 6, A and B). Importantly, visualization of these changes by both NMR and x-ray crystallography verified that they were not generated by the unexpected disulfide bridge in the Pg crystal structure (22) nor by artifacts in crystal packing, the latter of which have been observed previously in other crystal-derived structures (e.g. Refs. 33 and 34).
The net effect of these alterations to strand ␤3 is to pinch together the newly formed 3 10 -helix in Pg and helix ␣4 by as much as 6 Å. Notably, 3 10 -helix formation is accompanied by breakage of a potential salt-bridge between Arg-486 and Glu-445 (Fig. 6B). Loss of this interaction in Pg could release helix ␣1, with the acquired freedom then facilitating signal transmission to other domains. Interestingly, pinching of a comparable region was detected by NMR for the GAF domain of SyBCph1 during its Pr 3 Pfr photoconversion (11), suggesting that this structural rearrangement might be common among phytochromes.
A conspicuous feature of the GAF domain ␤-sheet is a prominent ridge that bisects the sheet perpendicular to the ␤-strands. It is formed by a kink in the ␤-sheet that begins with the Pb-state-dependent ␤-turn motif at residues 485-488 (Figs. 1 and 6A) and is accentuated by a ␤-bulge within strand ␤2 created by an unpaired residue (Val-474). The ridge is propagated through Val-462 of strand ␤1 and Leu-550 of strand ␤6, which is adjacent to the potentially disruptive Pro-539 of strand ␤5. Comparison of the Pb and Pg crystal structures revealed distinct bending of the sheet at this ridge in a photostate-dependent manner, which was most prominent for residues 462-474 of strands ␤1-␤2 and residues 539 -550 of strands ␤5-␤6 relative to helices ␣1 and ␣5 (Figs. 1 and 6C). Surprisingly, the remainder of the ␤-sheet (except for strand ␤3) exhibits little photostate-dependent movement. Although still evident in the paired NMR structures (Fig. 1), the magnitude of the ␤-sheet displacement was less, possibly due to limited NOE signals available to precisely position helices ␣1 and ␣5.
Photoconversion Effects on Backbone Dynamics-With respect to the 15 N backbone dynamics of the GAF domain, our NMR analyses revealed subtle photostate-dependent changes in longitudinal (T 1 ) and transverse (T 2 ) relaxation times, and steadystate 1 H, 15 N NOE measurements. Despite the greater overall rigidity of Pb as compared with Pg, several residues in Pb had unusually dynamic features (Fig. 7). Leu-530 near ring D exhib-ited exchange dynamics on the s-ms time scale (shorter T 2 values than average without an increase in T 1 (Fig. 7)), and residues 556 -561, which form a loop between strand ␤6 and helix ␣5, and Val-532, which is proximal to the D pyrrole ring, displayed dynamics on the ps-ns time scale (T 1 and T 2 larger and heteronuclear NOE smaller than average; Fig. 7).
Photoconversion to Pg annulled these dynamic features and introduced new ones to the NMR models. These included residues involved in photo-induced movement of the ␤-sheet (Fig.  7) such as 471-474 and 493-500, which showed slow conformational dynamics, and Asp-466, which exhibited dynamics on the ps-ns time scale. Several residues near the chromophore showed conformational dynamics in the Pg state, including s-ms motions for Ile-490 and Leu-527, and ps-ns motions for Gln-488 and Val-489. Strikingly, these Pb and Pg dynamic features occur at opposite sides of Arg-459, an almost invariant residue among phytochromes and part of the buried hydrogen-bonding network connecting the bilin to the protein (Fig. 5). For Pb, both the crystalline and NMR-derived structures showed that Arg-459 is well anchored through electrostatic interactions with the Glu-476 carboxylate, the main-chain carbonyls of Pro-483 and Leu-530, and the His-553 imidazole. The non-degenerate, welldispersed N H signals for Arg-459 as Pb, which are consistent with a salt bridge, disappeared in Pg, thus suggesting increased conformational freedom (data not shown). In support of these observations, we previously demonstrated that Arg-459 is required to stabilize the Pb state but is not required for photoconversion or maintenance of the Pg state (15).
Effects of Photoconversion on Signal Transmission-The differences between the paired Pb and Pg structures of TePixJ-(GAF) provide clues as to how light energy is transduced into a mechanical signal. Presumably TePixJ signaling is initiated by photo-induced structural changes in the bilin, which then reverberate through the GAF domain, the up and/or downstream histidine kinase/adenylate cyclase/methyl accepting/ phosphatase (HAMP) domains, and ultimately to the methylaccepting chemotaxis protein signaling (MCP) domain. In this context, light-triggered repositioning the GAF domain relative to the HAMP domain(s) might be facilitated through effects on the connecting GAF domain helices ␣1 and ␣5. Consequently, it is easy to imagine that any motion affecting these helices is crucial to signal transduction within the TePixJ dimer.
One possible avenue is via displacement of the ␤-sheet enabled by the distinct fault propagated by the ␤-bulge in strand ␤2 and release of strand ␤3. Given that helices ␣1 and ␣5 are each physically coupled to a pair of ␤-strands (␤1 and ␤2, and ␤5 and ␤6, respectively), displacement of the ␤ sheet would directly impinge on the mobility and positioning of these helices. Pg formation also stabilizes the loop between strand ␤6 and helix ␣5, which potentially reorients the helix with respect to rest of the GAF domain. Melting and structural reorganization of strand ␤3/␤-turn into a flexible region/3 10 -helix in addition to affecting the ␤1 and ␤2 strands appears to rupture a salt bridge between Arg-486 (just above strand ␤3) and Glu-445 within helix ␣1, which would then add mobility to this helical extension. Interestingly, such a photoinduced release of a rigid domain is reminiscent of the flavin-containing light/oxygen/ voltage (LOV) photoreceptors. Here, blue light-induced cleav-  (21,22). Additionally, Pb 3 Pg photoconversion is coincident with rupture of the Arg-486 -Glu-445 salt bridge, which may serve to loosen constraints on the helix ␣1 position. C, the alternative view provided by the paired crystallographic structures illustrates the sweeping movements of the loops between strands ␤1-␤2 and ␤5-␤6 and correlated motions of helices ␣1 and ␣5. The ectopic cystine disulfide found in the Pg crystal structure is included in C, with a portion of the second GAF domain from the asymmetric unit. Cys-494 is located by an asterisk. Cyan, nitrogens; red, oxygens; orange and yellow, sidechain carbon and sulfur atoms of Cys-494 and Cys-522, respectively.
age of a cysteine-based chromophore linkage within a structurally related PAS domain derepresses a downstream kinase effector domain by relaxing a critical interdomain contact (35). Such rearrangements in TePixJ could affect the GAF domain dimerization contact, loosen constraints on the flanking HAMP domains imposed by the GAF domain as Pb, and/or reorient the HAMP domains relative to the GAF domain via movement or axial rotation of the ␣1/␣5 helical bundles. With respect to the latter possibility, it has been proposed that HAMP domain-containing kinases, in particular, couple ligand binding by the sensory domain to rotation of the HAMP domain helical bundles via a gear box-type mechanism (36). Notably, similar rearrangements could underpin signaling by phytochromes in general. For the red/far red light-absorbing PAS-less phytochrome SyBCph1, we previously showed that upon photoexcitation several bilin-mediated structural changes focus on the GAF domain helical bundles involved in dimerization (11). Clearly, additional insights of the motions underpinning TePixJ signaling will now require more detailed models of photoconversion that include the output HAMP and MCP domains.
Conclusions-The paired solution NMR structures presented here of the CBCR TePixJ GAF domain in the dark and photoactivated states combined with the recently published crystal structures of this same blue-green light-absorbing photosensing module (21, 22) now provide unprecedented atomic details of phytochrome photoconversion. The solution structures confirmed the thioether linkage of the C10 carbon in PVB that underpins the distorted V-like structure of the bilin as Pb, demonstrated blue-light induced breakage of this bond concomitant with Z to E isomerization of the C15AC16 double bond and rotation of the D ring during Pg formation, and supported extensive migration of the bilin within the GAF domain pocket. The NMR structure of Pg also resolved the conformation of the region surrounding Cys-494, which was complicated by the non-native disulfide bridge in the x-ray crystallographic model. In addition, NMR relaxation studies revealed photostate-dependent changes in local dynamics. Key to these events are a number of universally conserved amino acids known to be critical for the photochemistry of TePixJ and others within the phytochrome superfamily (e.g. Asp-492, His-523, Gln-526, Asn-535, Tyr-463, and His-553; Refs. 11, 15, 21, and 37).
Ultimately, these combined changes substantially reorganize a ␤-strand/␤-turn section adjacent to the bilin and cause general bending of the ␤-sheet along a prominent central ridge. Given the proximity of these conformational changes to the helices potentially involved in TePixJ dimerization, we propose that they reorient the adjacent region(s) responsible for signal transmission by modifying intra-and interdomain contacts within the dimer. Collectively, these paired Pb/Pg structures likely reveal commonalities in light perception within the phytochrome super- family and provide informative scaffolds to modify signaling by CBCRs and their use as bilin-based fluorophores (15).