Structural Basis of a Phototropin Light Switch

See allHide authors and affiliations

Science  12 Sep 2003:
Vol. 301, Issue 5639, pp. 1541-1544
DOI: 10.1126/science.1086810


Phototropins are light-activated kinases important for plant responses to blue light. Light initiates signaling in these proteins by generating a covalent protein–flavin mononucleotide (FMN) adduct within sensory Per-ARNT-Sim (PAS) domains. We characterized the light-dependent changes of a phototropin PAS domain by solution nuclear magnetic resonance spectroscopy and found that an α helix located outside the canonical domain plays a key role in this activation process. Although this helix associates with the PAS core in the dark, photoinduced changes in the domain structure disrupt this interaction. We propose that this mechanism couples light-dependent bond formation to kinase activation and identifies a signaling pathway conserved among PAS domains.

Molecular switches are integral to any sensory system, converting input stimuli into signals that propagate to downstream components. PAS (Per-ARNT-Sim) domains are used through all kingdoms of life to combine these detection and transduction roles required of a switch. This is achieved by regulating protein-protein interactions with bound organic cofactors, providing a way to sense environmental stimuli including light (1). One class of PAS photosensors is the light-oxygen-voltage (LOV) domains (2), which bind flavin chromophores (3). In plants, LOV domains are critical to the function of phototropins, a group of light-activated kinases that regulate phototropism, stomatal opening, chloroplast relocation, and leaf opening (4, 5). Phototropins consist of a serine/threonine kinase domain preceded by two LOV domains (LOV1 and LOV2) (fig. S1), with LOV2 playing the dominant role in kinase activation (6). Photoactivation converts a noncovalent protein-FMN complex (LOV2D447) into a covalent one (LOV2S390) by forming a cysteinyl-flavin bond, which is stable for tens of seconds before returning to the ground state (7, 8). Although this photocycle is well characterized, the mechanism by which this signal activates the kinase remains poorly understood. Biophysical approaches to define this pathway have produced conflicting data, because crystallographic studies of LOV domains show minimal light-induced conformational changes (9, 10), whereas spectroscopic data suggest that more pronounced structural rearrangements occur upon illumination (1113). We resolve these conflicts by demonstrating that structural elements outside of the PAS domain core undergo large light-induced conformational changes and thus provide a pathway for propagating structural changes to the kinase domain.

We used nuclear magnetic resonance (NMR) spectroscopy to characterize the dark-state solution structure of the Avena sativa (oat) phototropin1 LOV2 domain (AsLOV2) (14). This construct (residues 404 to 560) contains the central PAS structural elements as well as a 40–amino acid C-terminal extension. Using uniformly 15N- and 13C-labeled protein samples and standard triple resonance methods (15), we assigned the chemical shifts of 94% of the amino acid residues. We determined that the secondary structure of AsLOV2 closely matches that of the crystal structure of the Adiantum capillus-veneris phy3 LOV2 domain (phy3LOV2) (3), based on patterns of backbone chemical shifts, 1H-1H nuclear Overhauser effects (NOEs), and 2H-exchange protection (Fig. 1A). Examination of more than 330 medium- and long-range NOEs (|ij| ≥ 2) further confirmed that both LOV domains adopt similar tertiary structures around their internal FMN cofactors, as expected from their 73% sequence identity.

Fig. 1.

Structural model of the AsLOV2 dark state in solution. (A) AsLOV2 secondary structure as determined by solution NMR spectroscopy. (Top panel) 13Cα chemical-shift index analysis (28), smoothed by a three-residue weighted running average. Deviations greater than ±0.7 ppm (red dashed lines) indicate significant changes from random coil; similar elements were identified with more comprehensive analyses with TALOS (17, 29). (Middle panels) Schematic summary of residues exhibiting short-range NOEs characteristic of secondary-structure elements [β sheet: strong Hαi–HNi+1 NOEs; α helix: HNi–HNi+2 NOEs, supported by Hαi–HNi+3, Hαi–HNi+4 (17)]. 2H-exchange data are summarized as bars for significantly protected residues, shaded to indicate protection factor (black, f > 106; gray, f = 105 to 106; white, f < 105). (Bottom panel) Comparison of structural elements observed in AsLOV2 (black) and phy3LOV2 (red) (3). (B) Helical wheel projection of Jα helix, generated assuming a turn of 100°/residue for residues 526 to 543. Boxes indicate residues with NOEs to the PAS core. (C) Model of AsLOV2 as determined from NMR-based constraints and homology modeling. Yellow lines connect the Cα positions of residues showing NOEs between the core PAS domain and the C-terminal Jα helix.

Unexpectedly, our data indicated that residues C-terminal to the β2α4β3 PAS core of AsLOV2 form an amphipathic α helix not previously observed in any PAS-domain structure (Fig. 1, A and B). This helix, designated Jα, is about 20 residues long as established by patterns of chemical shifts and short-range NOEs and is moderately flexible given the lack of measurable protection from 2H exchange. An alignment of the AsLOV2 and phy3LOV2 constructs suggests that phy3 also contains an analogous Jα helix, although its truncation in the fragment used for crystallization disorders the C-terminal 19 residues (3) (Fig. 1A). Expanding this alignment to other phototropin LOV2 domains reveals that this pattern of hydrophobic and hydrophilic residues is conserved (fig. S1), suggesting that Jα plays an integral role in their structure and function.

To establish the orientation of the Jα helix with respect to the PAS core, we obtained 49 unambiguous distance constraints between these regions from three-dimensional (3D) isotope-edited NOESY spectra. Combining these constraints with a model of the AsLOV2 PAS core built from the phy3LOV2 coordinates, we generated a low-resolution structure of the LOV2D447 form of AsLOV2 (Fig. 1C). In this structure, the Jα helix lies on the solvent-exposed face of the Gβ, Hβ, and Iβ strands of the central β sheet, burying hydrophobic surfaces on both sides. This orientation places the Jα helix immediately underneath the FMN cofactor, suggesting a model for photoactivation in which covalent-adduct formation causes structural changes that propagate across to the conserved Jα helix.

To test this model, we characterized the light-induced changes in AsLOV2 by recording NMR spectra in the dark and lit states (Fig. 2, A and B). Comparisons of 15N/1H and 13C/1H heteronuclear single-quantum coherence (HSQC) spectra showed that illumination produced widespread changes in backbone and side-chain chemical shifts. Additionally, intense cross peaks with 1H chemical shifts between 7.8 and 8.4 parts per million (ppm) appeared in the 15N/1H HSQC lit-state spectrum, suggesting that part of the protein unfolds in a light-dependent manner. We confirmed the specificity of these changes by acquiring the same spectra on C450S AsLOV2, a point mutant that folds normally but cannot proceed through the photocycle because it lacks the essential cysteine residue. No changes were seen upon illumination of this sample (fig. S2), indicating that the spectral differences observed in AsLOV2 are a direct consequence of the formation of the cysteinyl-flavin adduct.

Fig. 2.

Light-induced changes in the NMR spectra of AsLOV2. (A) 15N/1H HSQC spectra of AsLOV2 in the dark (black) and lit (red) states. All lit-state spectra shown were acquired under pulsed illumination (50 mW, 472.7 nm). (B) Methyl region of 13C/1H HSQC spectra of AsLOV2 in the dark (black) and lit (red) states. (C) Minimum chemical-shift difference from 3D HNCO spectra obtained in the dark and lit states of AsLOV2 {Δδmin = [Δδ(1H)2 + 0.17*Δδ (15N)2 + 0.39*Δδ (13C)2]1/2} (16). (D) Δδmin mapped onto the model structure of AsLOV2 based on the view shown in Fig. 1D, with a color gradient encoding the magnitude of the shift change. No data are available for residues colored in gray (e.g., prolines or overlapping peaks).

We then mapped the magnitudes of light-dependent chemical-shift changes onto our AsLOV2 structure. For maximum resolution, we compared dark- and lit-state 3D HNCO spectra, using the minimum chemical-shift difference method (16) to quantitate the spectral changes (Fig. 2, C and D). Within the PAS core, large differences were observed for residues close to the FMN chromophore, including those in the Eα helix and Gβ, Hβ, and Iβ strands. Significant perturbations were also observed at sites in the Jα helix, over 15 Å away from the FMN chromophore. Thus, structural changes initiated by the formation of the cysteine-FMN adduct propagate through the domain to disrupt the structure of the Jα helix.

To characterize the lit-state structure, we used several approaches, including limited proteolysis (Fig. 3A). Chymotrypsin treatment of AsLOV2 under illumination resulted in its rapid and complete cleavage into a minimal PAS-domain fragment (residues 409 to 530). In contrast, proteolysis was markedly slowed in the dark, completing within 60 minutes. Similar results were obtained with trypsin, showing light-enhanced proteolysis to a fragment including residues 411 to 521 (17). These data establish that photoactivation causes structural changes sufficient to completely destabilize the Jα helix while maintaining the overall PAS core.

Fig. 3.

Examination of structural changes upon AsLOV2 photoactivation as identified by proteolysis and 2H exchange. (A) Limited proteolysis of AsLOV2 in dark and lit states with chymotrypsin. The digestion reaction was stopped after 10, 30, and 60 min. Mass spectrometric analysis indicated that chymotrypsin cleaved the full-length domain (filled triangle, residues 404 to 560) to a protease-resistant fragment (open triangle, residues 409 to 530). (B) Results of NMR-based deuterium-exchange measurements analyzed by loss of 15N/1H HSQC peak intensity of AsLOV2 in 99% D2O. These experiments were done in both the dark (black) and lit (gray) states.

To further probe the conformational changes that displace the Jα helix, we measured 2H-exchange rates in the dark and lit states (Fig. 3B). Globally, illumination lowers protection factors ∼10-fold, indicating that the lit state is destabilized by ∼1.4 kcal/mol (18), much less than the 5 to 15 kcal/mol typically attributed to the net energetic stability of a small protein domain (19). However, similar secondary-structure elements are protected from exchange under both conditions, confirming that no gross structural deformations occur in the PAS core. Therefore, the light-dependent conformational changes in the domain must be sufficiently destabilizing to displace the Jα helix without globally unfolding the PAS core.

Because illumination leads to selective unfolding of the C-terminal Jα helix, we created a truncated protein lacking the last 24 residues of AsLOV2 (AsLOV2ΔJα, residues 404 to 536). This deletion removes a critical portion of the C-terminal Jα helix, producing a construct equivalent to the one used in the crystal structures of phy3LOV2 (3). Comparison of 15N/1H HSQC spectra recorded in the dark state of AsLOV2 and AsLOV2ΔJα reveals significant chemical-shift differences for residues located in the core of each protein (Fig. 4A). Further removal of C-terminal residues leaving only the PAS core (AsLOV2Δ, residues 404 to 521) produced no additional spectral changes in dark-state spectra. These results indicate that the Jα helix has a critical interaction with the PAS core that is removed when the helix structure is disrupted by truncation.

Fig. 4.

The PAS core adopts a common lit-state structure in AsLOV2 fragments lacking the Jα helix. (A) Overlay of the dark state 15N/1H HSQC spectra of AsLOV2 (black), AsLOV2ΔJα (blue), and AsLOV2Δ (orange). (B) An overlay of the corresponding lit-state spectra of AsLOV2 (red) and AsLOV2ΔJα (green). Peaks in all spectra have been labeled with residue assignments obtained from AsLOV2, using 15N SCOTCH data (30) to obtain lit-state assignments. All peaks shown are from the central PAS core domain (residues 414 to 516). (C) Expansion of the effects of light and protein truncation on the Trp491 side-chain Hϵ1imine resonance. The color scheme is the same as that used in (A) and (B).

Intriguingly, whereas the Jα helix affects the PAS core structure in the dark state, illumination completely eliminates this interaction. This was established by the marked similarity of lit 15N/1H HSQC spectra of AsLOV2 and AsLOV2ΔJα (Fig. 4B), indicating identical PAS structures regardless of the presence of the Jα helix. This finding is reinforced by data from Trp491, located at the hinge between the PAS core and the Jα helix (Fig. 1C). In AsLOV2ΔJα, this residue is only modestly affected by photoactivation (Fig. 4C) and has shifts comparable to that of the solvent-exposed Trp491 of AsLOV2Δ (17). In contrast, the corresponding peak of AsLOV2 has a completely different location in the dark state, becoming solvent-exposed only upon illumination and the consequent displacement of Jα. Together, these data demonstrate that the differences between the dark states of AsLOV2 and AsLOV2ΔJα are eliminated upon the light-induced displacement of the Jα helix from the PAS core.

Our observations of light-induced conformational changes that convert AsLOV2 from a “closed” dark state into an “open” lit state suggest a direct mechanism to convey photodetection to the kinase domain (fig. S3). Functional LOV2 domains are essential for biological activity, as shown by the requirement of the critical cysteine residue for phototropin activation (6). Combined with our demonstration that such mutations prevent AsLOV2 from under-going photoinduced conformational changes, there is a clear correlation between kinase activation and the structural changes that we observed. Although AsLOV2 could serve as a lit-state activator or dark-state inhibitor to achieve the observed photoinduced response, we favor the latter mechanism on the basis of parallels with FixL (20, 21) and PAS kinase (22, 23), which use PAS domains to inhibit their kinases in the latent state.

Finally, our work reveals similarities to the PYP class of PAS blue-light photoreceptors (PYP) (24). Both proteins cover the otherwise solvent-exposed face of the central β sheet with α helices from outside the PAS core (fig. S3). Because these helices are not protected from 2H exchange in either AsLOV2 or PYP (25), this suggests that they are flexible and primed for activation in the dark state. Photoactivation alters the structures of internal chromophores, producing conformational changes that destabilize each protein and unfold the external helical segments (26, 27). These parallels between photosensors with radically different cofactors and photochemistries suggest that there are unifying signal transduction mechanisms among these proteins. In a broader sense, these themes likely extend to other PAS domains despite the diverse cofactors they use and the wide range of stimuli they sense.

Supporting Online Material

Materials and Methods

Figs. S1 to S3


References and Notes

View Abstract

Navigate This Article