This review deals with the biophysical aspects of flavin-based photosensors, comprising cryptochromes, LOV (Light, Oxygen and Voltage) and BLUF (Blue Light sensing Using FAD) proteins. Special emphasis is given to structural issues, photocycle quantum yields and energetics, mechanism of the light-triggered reactions, early stages in signal transduction and oligomeric states of the light sensing protein modules. For BLUF and LOV domains important parallels are emerging, despite their different α/β fold arrangement, whereas there is increasing evidence for a mechanicistic and functional splitting of the cryptochrome family.
Photoreceptor proteins binding riboflavin (RB) derivatives as chromophores have emerged during the last 15 years and rapidly become a very active research field. In plants, blue light has long been known to regulate several responses, such as phototropism, stomatal opening, inhibition of hypocotyl growth, induction of flowering, circadian rhythms and light-dependent transcriptional regulation (1). The search for the “cryptic” plant blue-light sensor received a major contribution in 1993 with the identification of cryptochrome 1 (cry 1) (2), a few years later followed by phototropins (phot) (3) and photoactivated adenylyl cyclase (PAC) (4). The flavins bound to these photoreceptors are flavin mononucleotide (FMN) or flavin adenine dinucleotide (FAD), also ubiquitous cofactors involved in a variety of enzymatic reactions. They have a central role in aerobic metabolism through their ability to catalyze one- and two-electron transfer reactions (eT), and they can function as electrophiles and nucleophiles, with covalent intermediates of flavin and substrate frequently being involved in catalysis (5,6). In this review, after a brief summary on the general properties of flavin photosensors, I will focus on their photobiophysical aspects, such as structural and mechanicistic parameters, determination of quantum yields (Φ) and energetics, and early stages in signal transduction.
General features of flavin photosensors
Cry are two-chromophore proteins, very similar in structure to photolyases (PHR), enzymes that catalyze the light-dependent repair of UV-induced cyclobutane pyrimidine dimers (CPD) or (6-4) pyrimidine-pyrimidone photoproducts (7). In PHR, the absorption of a photon by the antenna chromophore that can be a pterin (5,10-methenyltetrahydrofolate, MTHF), a deazaflavin (8-hydroxy-8-demethyl-5-deazariboflavin, 8-HDF) (8) or even FMN or FAD (9,10), is followed by resonance energy transfer (ET) FADH- (fully reduced FAD) chromophore that in turn splits the DNA lesion by a catalytic, nonreductive eT (11,12). Cry lack DNA repair activity, with the exception of the so-called cry-DASH (Drosophila, Arabidopsis, Synechocystis and Homo cryptochrome) (13) that are able to split CPD, with high specificity toward single-stranded DNA (ssDNA) (14). Cry-DASH are represented in vertebrates and plants (15) as well as in the bacterial world (13,16–18). As a matter of convenience and historical reasons cry-DASH are listed here among cry proteins, although there is up to now no definite evidence for their role as photosensors and they could be solely specialized PHR (vide infra). As a whole, cry proteins have been identified in plants, animals and bacteria and are certainly the most ubiquitous flavin binding photosensors (7).
A few years after the identification of cry, plant phot came into the scene (3). Phot bind two FMN chromophores within a pair of PAS (PerArntSims) (19) domains arranged in tandem, later renamed LOV (Light, Oxygen and Voltage), based on their structural similarity with PAS domains binding heme or sensing voltage (20,21). Phot-LOV domains (ca 110 aa) bind oxidized FMN as chromophore in the dark, absorbing maximally at ca 450 nm (LOV447) (22,23). Blue-light illumination triggers a photocycle involving the reversible formation of a blueshifted FMN-cysteine C(4a)-thiol adduct (LOV390) that slowly reverts to LOV447 in the dark (22–26). LOV390 is normally formed via the μs decay of the redshifted FMN triplet state referred to as LOV660, but direct formation of the adduct from the singlet state has also been reported (27). The LOV paradigm has been conserved among distant phyla and is widely represented in fungi, bacteria and archaea. Phot (ca 1000 aa) are light-activated Ser/Thr kinases and are the sole LOV proteins that possess two such domains organized in tandem in the N-terminal part. All prokaryotic, fungal and even other plant LOV proteins bear a single LOV domain, associated with a variety of effector domains, such as kinases, transcriptional regulators, phosphodiesterases, constituting modular systems presumably switchable by light (18,28,29). In a few cases the bacterial proteins are built only from the LOV core and flanking extension, without any associated effector domain (referred to as short-LOV) (18). Phot signaling and other LOV proteins have been recently reviewed (30–32).
A third type of flavin photosensor has been identified in Euglena gracilis as the PAC responsible for phototaxis in this protist (4). PAC is able to respond to blue light owing to FAD-binding BLUF (Blue Light sensing Using FAD) domains, α/β modules comparable in size with the LOV (100–110 aa) (33). A BLUF domain is also the light sensing unit of AppA, a blue light and redox sensor involved in the regulation of photosynthesis genes in Rhodobacter sphaeroides (34). In the dark, BLUF domains show typical features of an oxidized flavin and their very peculiar photocycle involves the reversible formation of a ca 10 nm redshifted intermediate (BLUFRed), without apparent variations in the flavin redox state (34). To date genes coding for BLUF proteins can be solely detected in Euglenoids and Bacteria. The vast majority of them are short-BLUF proteins, consisting of the BLUF core and flanking regions, with the important exception of Euglenoid PAC, R. sphaeroides AppA and Escherichia coli YcgF (33).
During the first phase in the identification of cry, LOV and BLUF proteins, mutant organisms, defective or altered in a given response, played an essential role (1). Later on genome sequencing and annotation projects revealed that many prokaryotes are equipped with putative flavin photosensors, in particular LOV and BLUF proteins, that preserve the same photochemical reactions as their eukaryotic counterparts (18,35,36). In a sort of reverse process with respect to the one that marked the identification of plant blue-light sensors, we are now faced with the challenge of ascribing a light-regulated physiological role for these proteins which are now found in an increasing number of microorganisms. Some information is now available for a small set of these novel potential blue-light sensors: (1) YtvA from Bacillus subtilis is a 261 aa protein that bears a photoreactive LOV domain (37) associated with a sulfate transporter antisigma-factor antagonist (STAS) domain that confers to YtvA the ability of binding triphosphate nucleotides (NTP), such as GTP and ATP (38). YtvA acts as a positive regulator for the general stress transcription factor σB, specifically within the environmental branch (39), apparently in a blue–light-regulated way (40); (2) cry1, from Vibrio cholerae [Vccry1, a cry-DASH type protein (17)] possesses single-stranded (ss) DNA-specific photolyase activity (14). This function is not crucial for the repair of genomic DNA, but may play a very important role in the repair of ssDNA from transferable elements such as plasmids (14). (3) The Slr1694 BLUF protein from Synechocystis sp. PCC 6803 has been identified to be involved in phototaxis (41,42), but its action appears to be integrated with that of other photoreceptors (43). A review on the role of several photosensors in cyanobacteria has been recently published (44); (4) AppA from R. sphaeroides (450 aa) is a light and redox regulator for the expression of photosynthesis, built of a light sensing BLUF module and a redox sensing C-terminal domain (34,45,46). At low oxygen tension, AppA binds to the repressor protein PpsR, whereas under fully aerobic conditions PpsR is released and binds to the promoter of certain photosynthesis genes, repressing their transcription (45,46). These responses are light independent, but at intermediate oxygen concentration light determines whether AppA releases the repressor PpsR, in that under high light conditions PpsR binds to DNA and photosynthesis is repressed (47,48).
Exploiting blue light through an ancient chromophore
Riboflavin and RB derivatives are biosynthesized by plants, fungi, bacteria and archaea, from GTP and D-ribulose-5-phosphate as precursors, via a complex and phylogenetically ancient enzymatic pathway (49,50). It has been suggested that the RB structure, related to nucleotides, may have originated during the time of the hypothetical ancient “RNA world,” where RB and its derivatives may have worked as RNA cofactors or allosteric elements, only later adapted by protein enzymes (51). Light excitation of flavins results in a charge redistribution and an altered redox potential, giving rise to a variety of possible photochemical reactions among potentially dangerous photosensitization effects, via eT and ET reactions occurring mostly from the triplet state (52). The triplet state of RB and FMN is in fact formed with high yield (ΦT), with measured values ranging between 0.38 (53) and 0.6 (54,55) in aqueous solution. A recent ab initio theoretical study on the flavin isoalloxazine ring has shown that both the lowest singlet (S1) and triplet (T1) excited states correspond to ππ* transitions and that the efficient population of T1 is mediated by singlet-triplet crossing with a state of nπ* type (56). The efficient population of T1, mainly responsible for flavin photoreactivity is therefore an intrinsic property of the isoalloxazine ring. FAD represents an exception, in that in neutral solution it exists mainly in a stacked conformation in a stacked arrangement of the isoalloxazine ring and the adenine moiety, that undergoes photo-induced intramolecular eT, thereby quenching the singlet excited state and impairing formation of 3FAD (57 and references therein). The stacked conformation is stabilized by hydrogen bonds (HB) and the equilibrium between the stacked and unstacked form is thus pH dependent, conferring to FAD the typical pH-dependent fluorescence (57). Accordingly, photoexcited RB and FMN are powerful photosensitizers in that they generate singlet oxygen (1O2) with high quantum yield (ΦΔ = ca 0.5 in aqueous solution) via ET to molecular oxygen (58,59), whereas FAD has a low ΦΔ = ca 0.07 in water (58). Furthermore, whereas the reduction potential of flavins is about −0.3 V in neutral aqueous solution, the redox potential of the triplet state is shifted to 1.7 V, and can oxidize several electron-rich biological substrates, e.g. aromatic amino acids (aa) as well as external donors (60 and references therein). The yield and lifetime of the flavin radical thus formed are considerably smaller in the presence of oxygen (60,61), always representing a competing factor both in ET and eT reactions from excited states. The flavin-binding cavity of photosensors must therefore insure that photosensitizing reactions are minimized and they achieve this by several methods, essentially by quenching the triplet state with efficient reactions or by ultrafast deactivation of the singlet excited state via specific eT reaction pathways.
Small Photosensing Units: The LOV And BLUF Domains
LOV and BLUF domains represent minimal modules (ca 100–110 aa) able to sense and react to blue light via a bound flavin. An alignment of representative proteins is shown in Fig. 1, together with the secondary structural elements derived from deposited structures.
LOV and BLUF domains are α/β folds, with an βAβBαCαDαEαFβGβHβI and β1α1β2β3α2β4β5α3 arrangement, respectively [the nomenclature of the secondary elements is from recent literature (62,63)]. LOV is a typical PAS fold (19), where the extended central anti-parallel β-sheet is built by two distinct portions (βAβB and βGβHβI), forming a sort of solid basis that anchors the helical connector (αCαDαEαF) (Fig. 2).
The helical connector with its intervening loops is in fact the most variable part within PAS domains (19) and even within LOV domains (Fig. 1). It determines to a large extent the co-factor binding specificity and the protein reactivity (21). In LOV domains, the helical connector closes like a roof the FMN cavity and contains the substrate cysteine, some residues that interact with the ribityl moiety of FMN, and E960 (on Dα), engaged in a conserved salt bridge with K1001 (Gβ-Hβ loop) (the numbering is from Acphy3-LOV2) (21,64). The pleated β-sheet builds a quite hydrophobic pavement for the face of the isoalloxazine ring opposite to the reactive cysteine, at the same time bearing polar residues that form a well-defined HB network with C(2) = O, N3 and C(4) = O, comprising N998 on Gβ, N1008 on Hβ and Q1029 on Iβ (21,64). The rigidity of such a cavity, resembling a closed hand, is well illustrated by the high fluorescence anisotropy (0.3) of bound FMN (65) and by the fact that it seems to respond minimally to the formation of the photoadduct (25,64,66). Nevertheless it offers in principle a versatile sensing/transmitting system: virtually all the secondary structure elements bear residues that interact with the chromophore and could convey the signal from the inner cavity to the surface of the domain via different pathways, preferentially activated via fine tuning of intracavity/intradomain interactions. This idea finds support in recent results from molecular dynamics (MD) simulations, highlighting the different activation pathways of LOV1 and LOV2 in phot (67) (vide infra). At a further level, interdomain interactions could amplify or on the contrary block signal transmission conveyed to the LOV domain surface through the different pathways, resulting for example in activation of an effector domain or regulation of protein activity.
In BLUF domains the β1α1β2β3α2β4β5 arrangement is unique among flavin-binding proteins and bears similarity with the ferredoxin fold (68–70). The two central helices, α1 and α2, build two walls that flank the flavin ring along its main axis, while the two main players within the photocycle, Y21 and Q63, are localized on β1 and β3, respectively. The β-sheet is here partially organized in parallel strands, basically running perpendicular to the isoalloxazine ring. Overall, the sequence of secondary structure elements seems to wrap around the chromophore, that in fact establishes the majority of interactions with the inner strands and helices. As a consequence, the flavin within a BLUF domain is buried in a sort of “organized coil,” whereas within an LOV domain it is hosted in a quite rigid “floor–and–roof”-like cavity.
Sensing and reacting to blue light: the signaling state
The LOV paradigm is characterized by the formation of the metastable FMN-cysteine C(4a) thiol adduct (LOV390), usually via the decay of LOV660 (3FMN) that is built on the ns timescale (71). Typically LOV660 decays with 1–2 μs time constant (37,72–76). In LOV390 the isoalloxazine ring is two electron reduced and the spectral features of the dark state, namely the sharply vibrationally resolved absorption band in the blue region and the bright green fluorescence, are lost. The only aa known to be essential for this photocycle is the reactive cysteine, referred to as Cys966 from the numbering of phy3-LOV2, the first crystallized LOV domain (phy3 is a phytochrome-phot hybrid photoreceptor from the fern Adiantum capillus-veneris) (21).
In BLUF proteins the formation of the signaling state BLUFRed occurs instead from the excited singlet state of the FAD chromophore, within 1 ns (77,78), whereas the triplet state is formed with very low efficiency (77). The first photochemical investigation of BLUF-based light-induced reactions has been performed with His-tagged AppA, expressed in E. coli, binding FAD with a 1:1 stoichiometry (34). Dark-adapted AppA exhibits the two spectral features typical of an oxidized flavin, centered at 365 and 445 nm and vibrationally resolved. Both features undergo a redshift upon light absorption, becoming centered at 371 and 460 nm, respectively. These spectral changes are reversible in the dark, with t1/2 = 15 min (25°C) (34). All the BLUF proteins investigated to date show an AppA-like photocycle, with some differences in the dark state absorption maxima and in the magnitude of the redshift interval and the recovery lifetime, ranging from a few seconds to several minutes (a quite large collection of these parameters is reported in Ref. 79).
Two aa are essential for the accomplishment of BLUF photocycle, referred to as Tyr21 and Gln63 from AppA numbering (see Fig. 1). Even if the exact orientation of Gln63 is presently one of the most questioned topics in the field (vide infra), it is widely accepted that the light-induced redshift in the absorption spectrum is caused by HB rearrangements around the flavin chromophore, mainly strengthening of HB at the C(4) = O position, and that Tyr21 and Gln63 are both required to obtain BLUFRed (63,70,80–86). The requirement is so strict that the relatively harmless Q50N mutation in Tll0078 (corresponding to Q63 in AppA) shifts the photocycle to one more similar to LOV domains, with the formation of a redshifted species decaying on the short μs timescale, possibly a flavin triplet state (86). This points to the fact that Q63 must have a very precise orientation in the dark state, notwithstanding the fact that different crystal structures exhibit two possible rotamers of this residue (Fig. 3) (63,68,69,87), even within different subunits of the same crystal (70). In one orientation the lateral amide group of Q63 establishes HB with the hydroxyl group of Y21 (68) and subunit 2HFNd (70), while in the alternative one it is oriented toward C(4) = O of the flavin ring and Y21 is H-bonded to the Q63 carbonyl group (63,69,70,87). We will refer to these orientations as Q63Nε–Y21 and Q63Nε–C(4) = O, respectively.
Q63Nε–Y21 is associated with a rotameric conformation of W104, located on the β5 strand, defined as “in” because of its proximity to the flavin ring (68). With Q63Nε–C(4) = O, W104 may assume an “out” orientation facing away from the flavin ring, with a considerable conformational variability (63,69,70,87). Despite the fact that BLUF proteins are crystallized more often in the Q63Nε–C(4) = O configuration, there is considerable evidence from advance spectroscopy that Q63Nε–Y21 represents the actual dark state. The nuclear magnetic resonance (NMR) solution structure of AppA-BLUF in the dark shows that W104 is in the “in” rotameric configuration, within the core of the protein, with its indole NH able to hydrogen bond with the Q63 side chain carbonyl oxygen (88). Furthermore, there is increasing evidence from NMR and Fourier transform infrared spectroscopy (FTIR) spectroscopy that in BLUFRed a strong HB between the flavin C(4) = O and Q63 side chain amide is established, after Q63 has “flipped” by 180° with respect to the dark state (83,89–91). In accord with this, redshift of the absorbance spectra has been observed for flavins in apolar nonprotic solvents, in the presence of proton donors (92). Furthermore, a strong HB is established between Y21 and the Q63 side chain carbonyl group, as revealed by NMR data (91). Q63Nε–C(4) = O may thus be representative of the lit state BLUFRed and be, for reasons still to be clarified, the structure most often crystallized even when crystals are grown in the dark.
The peculiar BLUF binding cavity, the “organized coil,” may also partially account for the variability observed among the structures of BLUF domains. The different rotameric conformations of W104, with its bulky indole ring, seem to be able to affect adjacent elements, from the configuration of Q63 to the structure of β5 that partially unfolds in the “out” state, bringing M106 close to the flavin ring. In this way, a conformational variability of the “peripheric”β5 strand may result in a conformational variability of the binding cavity, probably resembling the dark and lit state(s) of the receptor and conceivably involved in signal transduction.
The energy and yield scenario
As mentioned above RB and FMN form the lowest excited triplet state with high quantum yield ΦT reported to be around 0.6 by means of different techniques (54,55). Accordingly, the photosensitized formation of singlet oxygen is quite high with ΦΔ = ca 0.5 (58,93). This clearly shows that ΦT > 0.5, not compatible with ΦT = 0.38 for RB and ΦT = 0.25 for FMN as obtained by means of picosecond laser double-pulse excitation with time-resolved fluorescence detection (27,53). The latter results are also in disagreement with photocalorimetric determinations (vide infra). The flavin T1 lies at an energy level ET = ca 200 kJ mol−1 as demonstrated by phosphorescence spectroscopy (94–97), ET measurements (98), photocalorimetric studies (37) and quite in agreement with recent theoretical calculations (56). The LOV cavity has little effect both on ET (37,99,100) and ΦT (37,71,74–76).
As for the value of Φ390, the formation quantum yield for LOV390, the literature data are sometimes contrasting. Steady-state optical measurements, e.g. the rate of FMN fluorescence decay that is certainly proportional to Φ390, can only provide relative values (23,75,76). This technique has nevertheless provided valuable information, e.g. the fact that in phot1 LOV2 has a higher Φ390 than LOV1 (23).
A classical approach for the measurement of a photocycle quantum yield makes use of the comparative method (actinometry) and laser flash photolysis (101). This technique follows in real time the appearance of the transient species during the photocycle and it is the most reliable one for the measurements of Φi, provided that the experiments are conducted within the linear region of amplitude vs laser fluence, in order to avoid artifacts due to double photon excitation. In this way we have evaluated ΦT and Φ390 in the LOV protein YtvA from Bacillus subtilis (37) (Table 1).
Table 1. Energetics and quantum yields of the photocycle intermediates in LOV proteins.
E390 (kJ mol−1)
*YtvA from Bacillus subtilis (35). †YtvA isolated LOV domain (75). ‡PpSB2-LOV = Pseudomonas putida Sensory Box 2, a short-LOV full-length protein (76). §PSPTO2896 = Pseudomonas syringae pv tomato hybrid LOV kinase/response regulator (Z. Cao, V. Buttani, A. Losi and W. Gärtner, unpublished). ||Losi et al. (74). ¶Picosecond excitation at 400 nm (27). #Continuous lamp excitation at 428 nm (181). **Blue light fs transient absorption and ps fluorescence (109). ††Steady-state photobleaching (22), for similar determinations with other plant LOV domains see Ref. (23).
The combination of time-resolved optical and photocalorimetric techniques offers the unique possibility to evaluate the energy level (Ei) of the photocycle intermediates together with molar structural volume changes (ΔVi), in a relatively simple and straightforward way (102). In the case of LOV proteins, laser-induced optoacoustic spectroscopy (LIOAS) has been extremely informative, given that LOV390 is formed within few after μs laser excitation (37,73) (72), i.e. within the time window of this technique (103). Deconvolution fitting procedures allow discriminating between the subns formation of the triplet state (“prompt” component) and its μs decay into LOV390, providing directly the product ΦiEi where i is any detectable transient species. By using the optically determined values of ΦT it was possible to estimate the triplet energy content of free FMN and LOV proteins as ET = ca 200 kJ mol−1 (37,75), in excellent agreement with phosphorescence measurements (97,99). In the same way, we determined optically the values of Φ390 and directly obtained the energy content of the adduct, E390. A collection of photophysical and thermodynamic parameters as measured by means of the combined optical-photocalorimetric (ns 450 nm laser excitation) approach is reported in Table 1. For comparison, values determined with other methods are also shown.
E390 lies between 133 and 180 kJ mol−1 for LOV proteins, that is the adduct stores a large percentage of the E00 energy (ca 246 kJ mol−1) (Table 1). This indicates that the protein has responded with minimal conformational changes to the formation of LOV390 (102). This is in sharp contrast to another well-known photosensing PAS domain, the photoactive yellow protein (PYP) (104). The putative signaling state of PYP stores only about 20% of the E00 energy due to a partial light-induced unfolding (105). The high value of E390 in LOV proteins drives the completion of the photocycle and the return to the dark state that would otherwise be virtually impossible given the intrinsic stability of photochemically formed C(4)a-alkylated flavin adducts in solution (106). This inherent stability has been also confirmed by theoretical calculations (107), that provide a negative energy content for the FMN-C(4)a-Cys adduct (−53.55 kJ mol−1). A value more in line with the experimental ones has been obtained by the group of Schulten, that calculated E390 = 36.5 kJ mol−1 (100). Different to that observed for the triplet state, the protein microenvironment has a profound effect on the energetics of LOV390, suggesting that the results of theoretical calculations must be taken with great care, as long as they favor a given reaction mechanism on the basis of energy minima.
The LIOAS data are not compatible with the values measured for ΦT by means of picosecond laser double-pulse excitation with fluorescence detection, e.g. 0.2 for Cr-phot-LOV1, 0.25 for FMN (27,53). From the LIOAS measured values of ΦTET (where ET is the energy level of the triplet state) FT < 0.48 would imply that ET > E00, which is obviously impossible (37,75,76). The data obtained by means of ps excitation also point to Φ390 ≥ ΦT, implying that the adduct can be formed via other states than the FMN triplet, a feature especially evident for Crphot-LOV2 (27) (Table 1). In that case the authors propose that LOV390 can be formed from the triplet state or from a charge transfer complex, [FMN−…D+], directly from the singlet excited state and that in Crphot-LOV2 this route is very efficient (108). This situation would be unique to Crphot-LOV2: in other LOV2 proteins, no transient species other than the LOV660 have been detected (73,109). This peculiarity of Crphot-LOV2 could be due to the presence of redox active aa in the vicinity of FMN, e.g. Cys216 on strand Aβ, normally an apolar aa (Val or Ile) in the other LOV domains (29). This aspect, important also for the identification of the mechanicistic details in the LOV photocycle, has to be verified with the aid of site-specific mutagenesis.
For BLUF proteins we have, up to now, no information on the energy levels of the putative signaling state BLUFRed, but we have quite a large collection of published quantum yields (ΦRed). One distinct feature of BLUF domains is that they do not exhibit absolute chromophore preference and they can bind RB, FMN or FAD (79,110–113). Nonetheless, it is generally believed that the in vivo active chromophore is FAD (4,34). In the case of AppA-BLUF the chromophore composition does not have remarkable effects on the photocycle, as the light-driven conformational changes are very similar in samples containing FAD, RB or FMN (110,112). Cofactor specificity can be achieved by incubating the protein with a given flavin chromophore (110). The value of ΦRed = 0.24 (77) also remains unchanged (111). The same is true for Slr1694, for which ΦRed = 0.63 independent on chromophore composition (113). There is no such information for Tll0078 [ΦRed = 0.29 (114)], while the heterologously expressed BLUF2 domain of E. gracilis PACα subunit (indicated as EgPACαF2) has ΦRed = 0.28–0.32 also independent of the specific flavin bound (82). An exception is the short-BLUF protein BlrB from R. sphaeroides—in this case the value of ΦRed = 0.4 in a sample that contains RB, FMN and FAD, and ΦRed = 0.9 for a sample containing solely FAD (79). The reasons for this different behavior with respect to other BLUF proteins remain unexplained (79).
On the way to the photoproduct: exploring the radical pathway
The LOV paradigm Although the elucidation of photochemical reactions is definitely the most advanced area in the field of LOV and BLUF photosensing, the mechanicistic details for the formation of the respective photoproducts are still a matter of debate. The long-lived LOV390 and BLUFRed are quite easily detected with conventional techniques, but the elucidation of the reaction intermediates is complicated by their ultrafast conversion at room temperature and by the difficulties in interpreting the spectroscopic data in a straightforward way, even when reaction intermediates can be trapped at low temperature. The possible mechanism of LOV390 formation has been recently reviewed (31) and we will only highlight some points here. For the majority of LOV proteins, the precursor of the LOV390 is the triplet state LOV660 but this reaction requires, for symmetry reasons, a spin inversion before it can proceed. Such process can be accomplished only with the formation of a radical pair before the establishment of the covalent FMN-Cys bond (115), and not, as initially proposed, via a concerted mechanism directly from LOV660 (21). The ability to undergo light-induced redox reactions has been demonstrated with mutated LOV domains, in which the reactive cysteine has been substituted. In Asphot1LOV2-C450A blue light induces formation of a long-lived flavin neutral radical (FMNH·), protonated on N(5) (26,116). The radical species induces a nuclear-spin polarization during NMR experiments, indicating the participation of an internal redox-active amino acid (A) to form the geminate radical pair 3[FMN•−...A•+], most probably a tryptophan (116). Crphot-LOV1 C57M forms instead a metastable photoadduct, decaying thermally into a stable flavin radical where FMN is covalently bound to the protein moiety (117). A detailed electron paramagnetic resonance (EPR) and electron-nuclear double resonance (ENDOR) analysis has identified the flavin radical as neutral, carrying and alkyl substitution at N(5) (118). The question is, of course, if such photoredox reactions are relevant for the wild type (WT) proteins, carrying the reactive cysteine, for which there is no direct evidence radical species are formed as precursor of the LOV390.
An early study with ultrafast spectroscopy showed that LOV660 appears on the fs time scale and it was suggested that this species corresponds to a N(5) protonated flavin triplet where the proton comes from the SH group of the reactive cysteine (71). FTIR studies at low temperature pointed instead to the fact that in LOV660 the cysteine is still protonated (119). Also, the demonstration that the methionine residue can form the adduct confirms that an initial proton transfer to N(5) is not required (118). A time-resolved EPR study at low temperature proposed that after triplet formation an eT step generates a radical pair in the form 3[FMN•−...SH•+], whose recombination would give a zwitterionic complex and then the final LOV390 observed at room temperature (115). The proposed mechanism agrees with the abovementioned FTIR investigation, that nevertheless could not definitely establish if the reactive cysteine is in the SH or SH•+ state (119).
Recent theoretical calculations favor instead a concerted H+ and eT mechanism to form a biradical and exclude eT from the cysteine to the flavin to generate a zwitterionic species, because a minimum of energy for such species could not be located (107). Nevertheless, given that a proper energy level cannot be calculated for the final adduct due to the effects of the protein environment (see previous paragraph), it is not possible to exclude a given mechanism solely on this basis. Further experimental verification is therefore required to write the final word on the LOV paradigm.
The BLUF paradigm Strong evidence for photoinduced eT (to form FAD•–) followed by H+ transfer (to form the semireduced species FADH•) has been obtained by means of ultrafast absorption spectroscopy with Slr1694, a short-BLUF protein from Synechocystis sp. PCC 6803 (120). In the mechanism proposed, eT from Y21 to FAD is followed by proton transfer, ultimately leading to Q63 (there Q50) flipping and to the establishment of new HB. A 180° flipping of Q63 in the lit state was first proposed by Anderson et al. (68) and supported by spectroscopic, mutagenesis and computational studies (81,85,91). In particular, FTIR showed that Q63 is not H-bonded to FAD C(4) = O in the dark (85) and NMR data are consistent with the formation of a strong HB between the carbonyl group of Q63 and the hydroxyl proton of Y21 (91).
The burning question of Q63 side chain orientation in the dark state is obviously of paramount importance, because the reaction mechanism proposed above would hold only for the Q63Nε–Y21 configuration (120). For the mutated AppA-BLUF C20S (87) and in BlRB (69), a light-driven flipping in the opposite direction has been proposed (87).
Some further suggestions on this topic have come from experiments performed with Q63 L and W104A mutants of AppA, which are locked in a lit-similar state, namely they are much less efficient than WT-AppA in inhibiting the DNA binding activity of PpsR in the dark (121). The authors propose that the HB between W104 and the Q63 carbonyl group is necessary to “lock” the dark state in a light-sensitive configuration, thus supporting the “in” hypothesis for W104 and the Q63Nε–Y21 configuration for the dark state. Light absorption would then induce Q63 flipping, break the W104-Q63 HB and change the β-sheet conformation, eventually promoting association between the BLUF domain and the C-terminal tail (121). Although these data indirectly support the “in” conformation for W104, they do not solve the problem of Q63 side chain orientation and of the photocycle mechanism, because W104A and W104F-mutated AppA-BLUF still retain the light-induced redshift (84,122). It appears rather that Q63 is needed to lock W104 in the “in” orientation, so that the bulky indole ring, once set free from the flipping of Q63, is able to perturb the β-sheet structure. In other words Q63 seems necessary for the photocycle to occur and W104 could efficiently couple it to signal transduction (vide infra).
Spectroscopic measurements performed with the Slr1694 and Tll0078 proteins at low temperature offer now a more comprehensive view of the BLUF photocycle. At 5 K an intermediate called I is formed, only a few nm redshifted with respect to the parent state (114,123). By increasing the temperature a further redshifted J intermediate is formed, and, at room temperature, the final state BLUFRed (114,123). This sequential chain of events seems to correspond to a gradual modification of HB at C(4) = O, most probably with Q50. No charge separate state could nevertheless be trapped at low temperature, suggesting that the photoinduced eT process is barrierless.
The quaternary structure of light sensing modules: dimers and more
LOV domains PAS domains, to which the LOV domains belong, are known to have diverse functions, including co-factor binding, activation/regulation of transcription, protein–protein interactions and dimerization (19 and references therein). Just to give an example in the field of photoreceptors, we recently learnt that PAS2 is responsible for dimerization of phytochrome A, a crucial feature for the action of this photoreceptor in vivo (124). The scenario is still quite confused for LOV domains and, most of all, the physiological significance of the dimerization, observed in some cases, is unclear.
It was shown by gel filtration chromatography that Asphot1-LOV1 has a tendency to dimerize, whereas LOV2 is monomeric (125). This has led to the suggestion that LOV1 is responsible for phot dimerization, providing a possible functional role for the tandem organization of LOV domains in phot (125). Although the dimeric state of full-length phot has not been definitely proved, a construct comprising the N-terminal half of Asphot1 (aa 1–525) and therefore LOV1 and LOV2 was found to be dimeric in solution (125). Dimeric states have been detected by means of small-angle X-ray scattering (SAXS) for the LOV domain of FKF1 (126) and phot LOV1 domains (127). The SAXS experiments showed that Atphot1-LOV2 is a dimer in contrast to other data (125,128) whereas Atphot2-LOV2 is monomeric (127). The LOV domain of WC-1 from Neurospora has also been shown to homodimerize in vitro (129). In the isolated LOV domain of YtvA (YtvA-LOV) the dimerization is very stable and from circular dichroism (CD) spectroscopy and docking simulations we have strong evidences that the LOV-LOV interface largely comprises the central β-scaffold (66). In all the studies on LOV proteins mentioned in this subparagraph, no significant changes in the dimerization state have been observed by varying the light conditions.
Experiments of time-resolved thermal grating and transient lens are uniquely able to determine the change in the molecular diffusion coefficient (D) of photoexcited molecules in a time-resolved way (130). Such experiments with Atphot1-LOV2 indicated that for concentrations <100 μm the protein is monomeric in the dark but D undergoes a transient light-induced decrease, occurring with a time constant of ca 40 ms at 50 μm, suggesting that there is a dimerization following the formation of the adduct (128). The data also indicated that the dimer is heterogeneous, with one molecule in the dark and the other in the lit state, LOV2450+LOV2390. Furthermore, the small intrinsic bimolecular association rate constant, well below the diffusion limit, indicates that the dimer is formed only under a very precise orientation of the partner monomers (128). On the other hand, in the high concentration range (>100 μm), a fraction of the protein is dimeric in the dark state and undergoes light-induced dissociation with a time constant of 300 μs. The authors propose that this apparent discrepancy might be explained by two possibilities: (1) the conformations of the ground-state dimer LOV2450+LOV2450 and the hetero-dimer created by the phototransfomation LOV2450+LOV2390 could be different or (2) the transient conformational fluctuations of the phototransformed LOV2390 could cause the driving force of the photodissociation. Phot2-LOV2 is instead monomeric and exhibits only minor conformational changes upon light excitation (130). As mentioned above, we have no hints on the functional significance of LOV-LOV dimerization, not to speak of light-induced changes in the dimerization state. Surely these studies, in the absence of structural data for full-length LOV proteins, are helping in defining the molecular surface that on the LOV core establishes intraprotein interactions and, most probably, is involved in light-to-signal transduction or/and regulation. Furthermore, LOV1-LOV2 heterodimerization could play a role in the signaling/regulation of full length phototropin, as suggested by data obtained with tandem LOV1-LOV2 constructs (23,131,132).
BLUF domains AppA-BLUF5–125 has been suggested to form a dimer in the dark state (133) in good agreement with gel filtration analysis that measured a molecular weight (MW) = 35kDa for AppA-BLUF1–156 (80). The decrease in D upon light excitation, indicates the formation of a tetrameric state, with second-order rate constant k∼2.5 × 105m−1 s−1, well below the diffusional limit, again pointing to the fact that the dimers must come in favorable orientation in order to form the complex (133). By gel filtration chromatography an increase in the apparent MW to 37 kDa in the lit state of AppA-BLUF1–156 is also indicative of a sharp conformational change, but it is not compatible with the formation of a tetramer. One has to keep in mind that this could be related to the low concentration of proteins in the excited state, a problem that is not relevant in photothermal methods, for which only laser-excited molecules are monitored. From dimers within the crystal units, we have also a good hint about the dimer interface of AppA-BLUF, conceivably built by the hydrophobic β-scaffold surface (68,69). This is similar to the putative LOV-LOV interface in YtvA (66), and accounts for the stability of the BLUF dimer in aqueous solution. Finally we note that the short-BLUF protein Slr1694 exists in solution in oligomeric form (trimer or tetramer), suggested to be important for its function in vivo (134).
Light-driven micromachines: passing the signal to effector domains
LOV proteins Given that no structures for full length LOV proteins have been solved, the molecular mechanism by which the different effector domains are activated is not yet clarified. The light-triggered enhancement of phot self-phosphorylation must rely on conformational changes that propagate from the LOV core and modify somehow intraprotein interactions. Similar to phot, it has recently been reported that in bacterial LOV kinases, light excitation increases the level of phosphorylation (36). X-ray data point to minor changes in the overall structure of LOV domains upon formation of the adduct, restricted to the vicinity of the chromophore (25,64). Actually the most spectacular change concerns the lateral chain of a conserved glutamine on Iβ (Q1029 in Acphy3-LOV2, Q120 in Crphot-LOV1), that in the dark state establishes a HB with C(4) = O by using Nε and flips by 180° after formation of the adduct, loosing the HB (25,64). FTIR light–dark difference spectra indicated that protein conformational changes are induced by light activation in Acphy3LOV2, restricted within the turn fraction at low temperature and propagated mainly to the β-sheet at 295 K (135). Similarly, the central β-sheet has been proposed to be the surface through which Ytva-LOV transmits the light-induced conformational changes to the STAS domain (66,136). In YtvA, Trp103, localized on Hβ and conserved within the LOV series, is involved in intraprotein interactions (137), most probably between the LOV core and the LOV-STAS linker region (65,76). Also, the nearby Glu105 mediates the transmission of conformational changes from the LOV core to the GTP-binding cavity of the STAS domain (136). The possible involvement of the β-scaffold in LOV signaling first emerged during NMR investigations of a LOV2 extended construct comprising 40 aa downstream the LOV core, in the full protein representing the linker region (J-linker) between LOV2 and the kinase domain (62). In the dark state the J-linker is organized as an amphipatic helix (Jα-linker) underneath the central β-scaffold, establishing multiple interactions with this part of the protein via its hydrophobic half. After the formation of the photoadduct the Jα-linker becomes unstructured and this is supposed to be the trigger that activates the kinase domain (62,138). This idea has been challenged by the observation that in phot2 the Jα-linker is not needed either for LOV2-kinase interaction or for light-driven phosphorylation of a heterologous substrate (139). The linker could nevertheless be necessary for the self-phosphorylation reaction or/and for regulation of self- vs hetero-phosphorylation. Larger conformational changes in Atphot1-LOV2-linker (aa 441–661 of A. thaliana phot1) than in Atphot1-LOV2 have been detected by means of SAXS (127). A strong reduction of D for Atphot1-LOV2-linker upon photoexcitation has been detected and, with the support of CD measurements, this was interpreted as the unfolding of the Jα-linker following the formation of LOV390 (140). The associated kinetic constants for the decrease in D suggested a model in which the dissociation of the linker from the LOV2 core, occurring with a time constant of 300 μs, is followed by unfolding of Jα with a time constant of 1 ms (140). A similar scenario, although with differences in the time constants and magnitude of D change, is depicted for the Atphot2-LOV2 linker (130). Interestingly the first time constant of 300 μs is similar to that observed for the light-induced monomerization of phot1-LOV2 in the high concentration range (128), indicating that the same mechanism underlies the detachment of the linker or of one LOV2 monomer. This also suggests that the molecular surface involved on the LOV2 core is, at least partially, overlapped in the two cases, namely the central β-scaffold that is known to interact with the Jα-linker in the dark (62).
The above depicted scenario bears important parallels and also differences to the bacterial protein YtvA. In that case the isolated YtvA-LOV core is dimeric even at a concentration of <5 μm (66). The N-terminal cap and the J-linker (the region between the LOV core and the STAS domain) do not prevent dimerization (66) (A. Losi, unpublished), whereas the full protein is monomeric, suggesting a competitive interface involved in LOV-LOV dimerization and LOV–STAS interaction (66). From CD spectroscopy and docking simulations we have evidence that this surface largely comprises the central β-scaffold (66,136). We nevertheless did not observe a light-induced unfolding of the J-linker. The linker region is anyway likely to be important in transferring the light-induced conformational changes signal from the LOV core to the STAS domain (65,136).
In the recently published structure of the Neurospora crassa photoreceptor VIVID, a short LOV protein, the N-terminal cap is partially helical and is in extensive contact with the central β-scaffold (141), indeed largely superimposed to the postulated dimerization surface (136). Although the light-induced structural changes in the crystal are very small, VIVID undergoes a large light-driven change in the hydrodynamic radius in solution, attributed to an increased disorder of the N-terminal helix preceding Aβ (named aα) (141). Given that a J-linker is missing in VIVID, the N-terminal cap could take over its function during activation of the protein and promote the switching from a compact to an unpacked conformation.
Related to the postulated involvement of the β-scaffold in signal transmission are recent studies on the functional role of the “flipping”glutamine Q1029 (phy3 numbering). Substitution of this residue with a leucine strongly impairs light-driven conformational changes in the central β-sheet, most probably due to the lack of HB with the FMN ring (142). Most importantly it was reported that in Atphot1 the corresponding Q575L mutation attenuates light-induced self-phosphorylation (143). We note nevertheless that this effect could be due to a decrease in the photocycle quantum yield, which was not quantified, in that the spectroscopic properties of the chromophore are also altered in the mutated protein (blueshift of the absorption spectrum and lower ΦF) (143). MD simulations failed to detect appreciable light-driven conformational changes in this protein region (67). Recent studies on VIVID confirm the role of the “flipping”glutamine (here Gln182) in coupling photochemical changes to signal transmission. In fact the Q182L mutation does not alter the spectral properties of the protein but the abovementioned change in the hydrodynamic radius is not observed (141).
Do LOV domains all function with the same mechanism during signal transmission, e.g. via the central β-scaffold? An alternative mechanism has been proposed that involves the salt bridge formed by E51 (on Dα) and K92 (Gβ-Hβ loop, Crphot-LOV1 numbering) on the surface of the LOV core, linked to the inner FMN cavity via a conserved volume of aa (28). Interestingly MD simulations suggest that light-induced strengthening of the E-K salt bridge is a characteristic of LOV1. In particular the Gβ-Hβ loop containing Lys92 is much more mobile in the dark state (67). In the dark, K92 establishes a salt bridge with D93 (a LOV1 specific aa), set free upon light activation. In LOV2 instead, the E-K salt bridge is stable both in the dark and in the light state, with conformational changes occurring mainly within the Hβ-Iβ loop and the adjacent regions of the central β-sheet that becomes more mobile (67). Consistently, mutation of the E-K salt bridge in LOV2 does not affect light-driven self-phosphorylation of phot1 (143). These two different activation mechanisms between LOV1 and LOV2 are triggered by modifications of the HB around FMN, transferred to the domain surface via different intradomain pathways, most probably initiated by N99 in the case of LOV1 and by Q1029 in LOV2 (67).
Bacterial LOV domains share similarities in sequence with both LOV1 and LOV2; in general they are more similar to LOV2 in the Aβ-Eα region and to LOV1 in the remaining part of the domain, e.g. in the majority of them the LOV1 aa D93 is conserved (18,29). In YtvA the E56Q mutation that breaks the E-K salt bridge influences negligibly the overall secondary structure and the light-induced conformational changes, although it has some effects on the fluorescence of W103 (65). The consequences of the E105L mutation are more profound. In WT YtvA a light-driven change in the turns region is detected, missing in YtvA-LOV (66,144) and in YtvA-E105 L (136). Furthermore, in YtvA light excitation promotes a conformational change that travels from the LOV core to the GTP-binding cavity within the STAS domain, a coupling that is abolished in YtvA-E105 L, although the binding of GTP is not impaired (136). These data suggest that the pathway of activation for Ytva-LOV may be more similar to LOV2.
BLUF proteins The first hypothesis on the mechanism of signal transmission in BLUF proteins came from the group of Schlichting, who proposed that the adenine ring of FAD serves as a hook to anchor the effector/output domain, thereby conveying the light-induced structural changes (69). This idea was challenged by the fact that chromophore substitution (with FMN or RB) does not affect conformational changes in AppA-BLUF (110,112), although things may be different in full length AppA. We note nevertheless that AppA has been reported to be monomeric in solution whereas AppA-BLUF is dimeric and the dimerization interface likely involves the β-scaffold. Therefore as in YtvA (66), there is apparently in AppA a competitive interface for BLUF-BLUF and intraprotein interactions and this interface is at the opposite side of the adenosine moiety, rendering the “adenosine hook” mechanism quite unlikely. The E. coli YcgF protein, built of a BLUF domain and a C-terminal EAL domain [a phosphodiesterase for cyclic diguanylate, c-di-GMP(145)] is also monomeric in solution (146) and the same competition with BLUF dimerization may occur. In YcgF, FTIR difference spectra revealed different and larger light-induced protein conformational changes with respect to YcgF-BLUF, thought to be the proof of signal transmission to the EAL domain (90), although a light-regulated phosphodiesterase activity of this protein remains to be proved. The protein conformational changes in YcgF are similar to those observed for Slr1694, where an effector domain is missing and only two helices follow the BLUF core (83,134), showing that the region C-terminal to the BLUF core could be important in signal transmission.
Recent work highlights the unique role of Trp104 during light sensing in AppA. In AppA-BLUF, the W104A mutation impairs a light-induced structural change in the β-sheet, without appreciable effects on the photocycle (84). In functional essays with full length AppA, the W104A mutation locks the protein in a functionally lit-like state (121). The same is true for the Q63 L mutation that instead inhibits the photocycle (121). It is concluded that the HB established between Q63 and W104 is critical during AppA light sensing, also indirectly supporting the Q63Nε–Y21 and the “in” configuration, respectively, for the two aa.
Decrypting The Cryptochromes
Cry together with PHR form a large protein family, consisting of 55–70 kDa proteins that contain two noncovalently bound prosthetic groups—a photo-redox active FAD and a pterin or, more rarely, a flavin antenna (7,147). All members of this family share a high degree of sequence and structure similarity within the so-called PHR core, where the antenna and the active FAD are bound. PHR and cry exploit light to carry out basically distinct functions—the former harness blue-light energy to break bonds and repair UV-induced photoproducts in DNA, the latter act as blue-light sensors regulating processes ranging from circadian entrainment in animals and plants to plant growth and development (7,148). With the exception of cry-DASH (14), they are not able to repair UVA-produced lesions on DNA (7).
The structure of cry
The crystal structures of three cry proteins have recently been published: (1) the PHR core of Atcry1 (149) (1U3C, 1U3D); (2) the cry-DASH Sll1629 from Synechocystis PCC 6803 (13) (PDB code 1NP7); (3) Atcry3 with bound MTHF, as antenna, and FAD, including the N-terminal cap (150,151) (2IJG; 2J4D, not yet released). The PHR core is built of two parts, an N-terminal α/β subdomain that binds the antenna (only in some cases co-crystallized with the protein) and a 4α-helical C-terminal subdomain binding the FAD chromophore. The α/β domain bears an extended β-scaffold made up of five parallel β-strands, an arrangement reminiscent of flavodoxin, response regulators and small GTP-binding proteins (see Fig. 4 for a sequence alignment with secondary structure nomenclature). The FAD cofactor is deeply buried within the α-helical domain and has a U-shaped conformation with the isoalloxazine and adenine rings in close proximity. A distinct feature of Atcry1 is the absence of the positively charged DNA binding groove typical of PHR (149). Furthermore, it was found that Atcry1 binds ATP stoichiometrically (152) and that the ATP is located in the same active site cavity that binds the CPD in PHR (149), a feature probably related to the self-phosphorylation activity of Atcry1 (152) (vide infra).
The first released structure of a cry-DASH, Sll1629 from Synechocystis PCC 6803, was solved with a bound FAD, but without any antenna chromophore (13). It was suggested that this may be a characteristic of cry-DASH but it soon turned out that this is not the case. In Vccry1, Atcry3 and vertebrate cry-DASH a bound MTHF antenna performs efficient ET to the FAD chromophore (15,153,154). Both published structures of Atcry3 have been crystallized with a bound MTHF, interacting with 12 aa highly conserved within the cry-DASH family (150,151). Interestingly, the binding of MTHF appears stronger than in E. coli PHR, and only one of the interacting residues, E149 in Atcry3, is conserved between the two proteins (Fig. 4) (150,151). Mutation of E149 strongly impairs the binding of MTHF (151). A slightly smaller distance between MTHF and FAD may account for faster ET observed in cry-DASH with respect to PHR (153,154). It was also noted that the cavity hosting the CPD in PHR becomes shallower and wider in cry-DASH and even more polar, thus probably impairing the binding with the pyrimidine dimer (13,150). Finally, in Atcry3 the N-terminal cap, which mediates import into chloroplasts and mitochondria (155), is co-crystallized with the PHR core and is partially organized as a helix, stabilized by intraprotein interactions (150,151).
Light reactions: radical pairs and the splitting of the cry family
PHR and cry show some characteristics in common and also bear important differences. The catalytically active chromophore form in PHR is FADH−, the double electron reduced form of FAD that transfers an electron from the excited state to the DNA lesion (reviewed in Ref. 8). After repair, the e− is transferred back to the flavin chromophore, thereby recovering the active FADH− form. A similar reaction scheme may occur in bacterial cry-DASH during repair of ssDNA (14), but obviously not for those cry that do not repair UVA lesions. Furthermore, there is increasing evidence that the functional form of cry in the dark state is the blue-light absorbing, fully oxidized FAD (156–159).
A second light reaction of PHR is the so-called “photoactivation,” better said photoreduction, a light-driven recovery of the fully reduced chromophore in PHR preparations that contain variable amounts of FAD and FADH• (160,161). In this reaction, light induces eT to the excited flavin through a chain of aromatic amino acids (Trp and/or Tyr residues), thereby reducing FAD and FADH•. This feature pf PHR was indeed previously uncovered by Sancar and coworkers (162), who later presented evidence that “photoactivation” is not relevant in vivo (11). The chain of aromatic, redox active aa is conserved throughout the cry/PHR family (Fig. 4) and actually gave the first hints of how cry may work in vivo. Atcry1 is purified with a bound FAD in the fully oxidized form, but in the presence of an external electron donor, FAD can be photoconverted to the semireduced radical FADH• (163). The reduction of the flavin proceeds via electron extraction from a tryptophan residue, presumably the conserved W324, corresponding to W306 in E. coli PHR (164). Mutation of two tryptophans in the chain (W324 and W400) into redox-inert phenylalanine results in a marked reduction in light-activated autophosphorylation of Atcry1 in vitro and of its photoreceptor function in vivo (165). Thanks to this and subsequent studies, the light-induced formation of a FADH• TRP• radical pair as part of cry activation is gaining increasing consensus. Accordingly, magnetic effects on cry-dependent growth responses have been identified in Arabidopsis thaliana, solely explainable with the formation of a radical pair (166). This mechanism would potentially configure cry as unique magnetophotoreceptors, strengthening an original idea that proposed that radical pairs of cry represent the physical basis of the avian magnetic compass (167). External magnetic fields would modulate the singlet–triplet ratio of the radical pair products and thus subsequent biological reactions that depend on the radical yield. This idea has received support from computer-based calculations on eT and spin dynamics in Atcry1, showing that the spin chemistry of the photoreduction process can be modulated by the presence of a weak magnetic field, in agreement with the radical pair mechanism (168).
Convincing evidence for FAD photoreduction of Atcry1 in living cells has come from very recent studies employing a combined physiological and biophysical approach (159). By means of EPR spectroscopy, it was shown that insect cells expressing Atcry1 and irradiated with blue light accumulate a paramagnetic species, identified as the neutral radical FADH• (159). If green light is given subsequent to blue light, reoxidation of FADH• to FAD in the dark is slowed down and there is formation of the fully reduced species FADH− in a two-photon/two-color process (159). The quantum efficiency of these reactions remains undetermined. A similar mechanism has been described for Atcry2 (169). The neutral semiquinone FADH• is thus configuring as the functional chromophore for the signaling state in plant cry1 and cry2. The formation of FADH• from FAD obviously requires a proton donor. This topic has been elegantly investigated by means of FTIR (170) by using Atcry1, for which an aspartic acid, Asp396, has been proposed to act as H+ donor to the flavin N(5) (170).
A variation of the mechanism proposed for Atcry1 and Atcry2 has been proposed for Drosophila melanogaster cry (Dmcry), a photoreceptor mediating light-entrainment of the circadian clock, for which the signaling state has FAD in the red-absorbing anionic radical FAD•-form (156,171). Stabilization of the buried negative charge, localized/centered around the N(1)-C(2) = O region of the isoalloxazine ring, is thought to be partially achieved via the positively charged Arg271 which is conserved in animal cry (156). The fact that an H+ donor is in this case not needed is underlined by the aa sequence—Atcry1 Asp396 is substituted by Cys416 in Dmcry (see Fig. 4). These recent studies have been very important in defining the properties of plant and animal “classical” crys and have also identified amino acids important for the specific light-activated reactions (see Fig. 4).
What about light reactions in cry-DASH? Purified Atcry3 and vertebrate cry-DASH contain an MTHF antenna and a mixture of oxidized, semireduced and fully reduced FAD (15,151,154). In Atcry3, photo-excitation of FAD results in the reversible formation of the fully reduced species, FADH2, representing a photocycle with a very low yield (ca 0.2%). When the semireduced species FADH• is excited, the fully reduced FADH2 (or FADH−) is formed with a higher quantum efficiency of ca 7%. Finally, prolonged light exposure modifies the re-oxidizable FADH2 into a permanent reduced state (154). Another cry-DASH, Vccry1, is purified with a fully reduced bound FADH−, suggesting that this protein may indeed undergo light reactions similar to that of PHR (17) and consistent with its role as PHR for ssDNA (14). Indeed, even purified Atcry3 contains FADH2 (or FADH−) at the rate of 55% and this could well represent the active chromophore form (154). This feature underscores the similarities between cry-DASH and PHR and prompts uncovering the actual physiological role of this peculiar class of cry.
In summary, cry-DASH seem to work similar to PHR, but with high specificity for ssDNA, with the competent form of the flavin chromophore being most probably FADH− (14,17,154). On the contrary, Atcry1, Atcry2, Dmcry and their homologs, the “canonical” cry proteins, bind oxidized FAD in the dark state and blue-light excitation results in intraprotein eT to form the neutral or anionic radical (FADH• or FAD•−) that can be further reduced in a two-photon process (156,159,164,165,169,170). Nevertheless, we note that blocking intraprotein eT in Danaus plexippus (monarch butterfly) cry1 (Dpcry1) by the W328F mutation does not affect its circadian photoreception function (172), raising questions about the oxidation state of FAD in the dark state of Dpcry1. It is possible that the intraprotein eT chain described above leading to the photoreduction of FAD describes only one of the possible cry photocycles, but we still have no conclusive evidence for its effective role in vivo.
Signaling issues in cry
One of the most intriguing aspects in cry signaling is the regulation of their activity by phosphorylation, in some cases proved to be the result of self-phosphorylation (173). Early works reported that Atcry1 and Hscry1 exhibit blue light stimulated self-phosphorylation dependent upon the presence of the FAD chromophore (152,174) and requiring the photoreduction of FAD from the internal tryptophan chain (165). In a further study, the autokinase activity of these two proteins and of Hscry2 (Homo sapiens cry2) was confirmed, but not found to be dependent on light conditions or on the presence of the flavin chromophore (173). On the other hand, Atcry2, which is known to be phosphorylated upon light exposure in vivo (174), lacked kinase activity (173). Further work is clearly needed to clarify this partially contrasting result, and also to understand the molecular basis of the kinase activity, given that crys do not have any canonical kinase domain, contrary to many LOV proteins.
Another peculiarity of cry signaling is related to the C-terminal tail (50–250), present in all cry proteins with the exception of cry-DASH, and demonstrated to be important for signal transduction/regulation in several cases. Transgenic plants overexpressing the C-terminal extension of either Atcry1 or Atcry2 exhibit a constitutive photomorphogenic phenotype in the dark, suggesting a main role of this protein region in mediating cry interactions with COP1, the main interacting partner of cry (175). In Dmcry, the removal of the C-terminal end renders the protein constitutively active and it was concluded that this protein region promotes repression of signaling, a repression alleviated by light activation (176). A detailed analysis of the C-terminal end in Atcry1 and Hscry2 showed that both are intrinsically disordered in solution, although they share little sequence homology, and that they interact with the PHR core in the dark (177). This interaction induces a stable tertiary structure that undergoes a transition to a more disordered state upon light activation of the protein (177), a feature surprisingly reminiscent of the Jα-linker unfolding in phot (62). Limited proteolysis under light and dark conditions and in silico-analysis also allowed identifying specific amino acid motifs likely to interact with effector partners (177). Likewise, a combined bioinformatic and experimental approach has recently identified likely hotspots for molecular interactions in the C-terminal end of Dmcry (178).
Finally Atcry1 has been reported to form homodimers in solution, mediated by the PHR core, and that dimerization is required for Atcry1 activity in vivo (179). A dimeric crystal structure also characterizes the DASH Atcry3, which is instead monomeric in solution (151). This agrees with the fact that the interface is mostly polar and prone to dissociation in aqueous solution. It is thus likely that the crystallographically observed dimer of Atcry3 does not represent an oligomer that is relevant in vivo.
Biophysical studies are revealing unexpected similarities between LOV and BLUF domains, despite the difference in the topology of their α/β fold and in their photocycle. On the other hand, sharp mechanicistic and functional differences are emerging within the cry family of proteins. In all flavin photosensors analyzed, increasing evidence points to the occurrence of eT reactions preceding the formation of the final photoproduct, in agreement with the most characteristic trait of flavins—the ability to switch among different redox states that can provide the driving force for further reactions. The analysis of oligomeric states of isolated light sensing domains is providing useful information on the “hot” surfaces involved in intraprotein interactions and/or in the early stages of signal transduction. Molecular dynamic simulations have started to nicely complement the experimental data, providing hints on the pathways for light activation in flavin photosensors, likely to be important in signal transduction.
Acknowledgements— I am grateful to Wolfgang Gärtner for his critical reading of the manuscript, and to Robert Bittl and John Kennis for helpful discussion.