The Photocycle of a Flavin-binding Domain of the Blue Light Photoreceptor Phototropin*

,

Near-UV blue light regulates a variety of different responses in higher plants. These include phototropism, the inhibition of hypocotyl elongation, the expression of various genes, and stomatal opening. Phot1 (nph1), the recently discovered blue light receptor, is a member of the phototropin receptor family (1). Phot1 is a plasma membrane-associated flavoprotein that functions as the primary photoreceptor mediating phototropic plant movement (2)(3)(4). Phot1 has two 12.1-kDa flavin-binding domains, LOV1 and LOV2, within its N-terminal region and a typical serinethreonine protein kinase domain at the C-terminal region. Heterologous expression studies have shown that phot1 binds FMN 1 as a chromophore and undergoes autophosphorylation in re-sponse to light treatment. It has therefore been proposed that this receptor functions as a light-activated serine/threonine kinase (4). The isolated LOV domains from oat phot1 expressed in Escherichia coli have been shown to undergo a cyclic photoreaction upon the absorption of light; LOV1 recovers with a half-time of 11.5 s, whereas LOV2 recovers with a half-time of 27 s (5). In addition, the quantum efficiencies for photoproduct (adduct) formation for LOV1 and LOV2 are ϳ0.045 and 0.44, respectively (5). The ground forms of the LOV domains have major absorption peaks at 360 and 450 nm with the 450 peak having vibronic structure at 425 and 475 nm. Upon absorption of light, the chromophore bleaches 2 in the 450 nm region generating a species that absorbs maximally at 390 nm. This intermediate has been assigned as a flavin-cysteinyl adduct between the protein and the C(4a) carbon of the FMN chromophore. This adduct breaks down spontaneously, returning the protein to its ground form. A LOV2 mutant (LOV2C39A) in which the cysteine that forms the adduct has been mutated to alanine does not undergo this photoreaction (5).
Recently the crystal structure of the LOV2 domain from the fern Adiantum capillus-veneris phy3 (6) was solved to 2.7-Å resolution (7). Phy3 is a chimeric photoreceptor with homology to phytochrome at its N-terminal end and an almost complete phototropin at its C-terminal end. Its LOV2 domain shares a 70% sequence homology to the oat phot1 LOV2 (6). The structure indicates that the FMN molecule is held noncovalently within a chromophore-binding pocket. It places the sulfur of cysteine 39 at ϳ4.2 Å from the C(4a) carbon of the FMN chromophore. These observations are consistent with the lightinduced formation of an FMN-cysteinyl adduct.
Here, we characterize the photocycle of the LOV2 domain of Avena sativa (oat), phot1. We have identified a new intermediate state in this photocycle and present evidence that Cys 39 exists as a thiolate in the ground-state chromoprotein. We propose a photocycle scheme for this domain of the photoreceptor consistent with these observations. In addition we show that the LOV2C39A mutant undergoes a truncated photocycle in which this early intermediate reverts to the pigment's ground form.

EXPERIMENTAL PROCEDURES
Light-induced Absorption Changes at Long Times-Difference spectra in the 1-100 s region were collected on a Hewlett Packard 8452A diode array spectrometer. The optical path length was 1 cm. The blue light pulse was provided by a white light camera strobe flash (1 ms, ϳ100 mJ pulse) filtered through Corning Glass filters 3-73 and 4-96, and a Corning 100-nm band pass filter with maximum transmission at 400 nm. Control of data acquisition and flash were automated with software written in LabVIEW (National Instruments, Austin, TX). Temperature was not controlled but was measured to be 20 Ϯ 2°C.
Light-induced Absorption Changes at Short Times-Difference spectra in the 30-ns to 1-ms time window were collected on an instrument described previously (8). In brief, a dye laser pumped by the third harmonic of a Nd:Yag laser provided a 10-ns, 80-J/mm 2 light pulse at 477 nm. The optical path length was 2 mm. A fresh sample was provided for each laser flash, allowing the averaging of absorbance data of several samples. The temperature for all measurements was 20°C. Light used to probe absorbance was polarized linearly to the magic angle (54.7°) relative to the laser polarization axis to prevent rotational diffusion artifacts (9).
Sample Preparation-Samples were prepared as outlined previously (5). In brief, the LOV domains were expressed in E. coli and purified by calmodulin affinity chromatography. The LOV2 domain used for these experiments was derived from oat phot1. The LOV2C39S mutant was made following the same procedure used for the LOV2C39A mutant (5). FMN solutions were made by dissolving FMN (Sigma) in the same buffer as that used for the protein preparation. D 2 O Exchange-A LOV2 sample was divided into two aliquots and lyophilized in the dark to a dry powder (20 h). One sample was then resuspended in D 2 O (Aldrich); the second (control) sample of LOV2 was resuspended in H 2 O.
Analysis-All data were transferred to a personal computer for analysis using programs written in a Matlab environment (The Mathworks, Natick, MA). Absorption difference spectra taken at different delays following light excitation pulse were arranged in the columns of a data matrix. The data matrix was then subjected to singular value decomposition (SVD) followed by global exponential fitting (10,11). The global exponential fitting analysis assumes that the dark reactions following light excitation are first order processes. Kinetic changes are decomposed into a sum of exponential components. The exponents contain the apparent rate constants for the observed kinetic changes, and the amplitudes (pre-exponential factors) at different wavelengths represent the spectral changes associated with the exponential process and are called the b-spectra (10,12).
Fluorescence Titrations-Concentrations of protein stock solutions were determined with a Hewlett Packard 8452A diode array spectrometer using ⑀ 447 (LOV2) ϭ 13,800 M Ϫ1 cm Ϫ1 (5). Stock protein was diluted with buffer (5 mM Tris, 10 mM NaCl, pH 8 for acid titrations, pH 6 for base titrations) to a concentration of about 1 M. Corrected fluorescence excitation spectra were recorded between 300 and 515 nm with constant stirring on a Spex Fluorolog fluorometer, and fluorescence emission was monitored at 535 nm. The fluorometer was equipped with a lid containing holes fitted for a pH electrode and syringes so that pH adjustments could be made and monitored without opening the sample compartment. The pH was changed in a stepwise manner using 0.5-1 M HCl or NaOH and was monitored with a Corning Digital 110 meter and a Beckman Futura (model 511063, Fullerton, CA) semi-micro AgCl combination electrode. The total volume change over the course of the titrations was a maximum of 3%, precluding significant volume effects.

RESULTS
Nanosecond Laser Flash Spectroscopy of LOV2-The long recovery times of the LOV2 photocycle (ϳ200 s) precluded the use of simple signal averaging techniques to acquire transient laser-induced absorption changes over the full time range. We therefore obtained the data in two time scales using different instrumentation. For the short times (30 ns-100 s), we used the gated diode array flowing a new 1-l sample into the cuvette for every laser flash and recorded absorption data at about 720 wavelengths at selected delay times of 0.03, 0.13, 0.33, 1, 3, 10, and 100 s. For extended time intervals (0.5-200 s) we used the Hewlett Packard diode array with blue-filtered narrow-band flash lamp excitation. The LOV2 absorption difference spectra at short times, averaged over eight laser flashes, are shown in Fig. 1. The spectra show a bleach of the 450 nm peak and a transient increase of absorption in the green/red and near-UV regions. This transient state relaxes within 10 s into a metastable intermediate state that decays into the original ground state in tens of seconds (Fig. 2). LOV1 shows spectroscopically similar transitions with a much lower quantum efficiency (data not shown).
The above data show that the spectral features at 10 and 100 s are nearly identical to those at 500 ms, indicating that the long-lived state decays in seconds. The recording time gap (100 s-500 ms) does not therefore influence our kinetic analysis in this data set. The difference in pulse width of the excitation sources could have complicated the analysis of the data if the millisecond lamp flashes, which overlap in time with the metastable-state life span, caused second photon hits on this intermediate. Fortunately, the first absorption difference spectrum recorded in the long time window (500 ms) is nearly identical to  the spectrum recorded at 100 s, indicating that there are no additional spectral transitions within the time gap, and second photon-induced photochemistry did not occur. This was apparent when we merged both data sets, adjusted for identical amounts of bleaching at 447 nm, as shown in Fig. 3.
The main feature of the nanosecond transient species is the absorption at longer wavelengths (500 -700 nm), showing broad peaks at 510 and 660 nm. Various flavin photoproducts are known to show absorption in this wavelength range, including charge transfer states, triplet states, and neutral flavosemiquinones (13,14). To explore the nature of the early photoproduct, we generated the triplet and flavo-semiquinone states of FMN in aqueous solution by laser flash excitation (15,16) and compared their spectral features with those of the intermediate states of LOV2. Although charge transfer bands of flavoenzymes also absorb maximally at wavelengths longer than that of the ground state (17,18), they were excluded from consideration because the LOV2C39A mutant, which cannot form a charge transfer complex, formed an almost identical transient species (see below). Fig. 4 compares the LOV2 absorption difference spectra at 30 ns and 500 ms with the absorption difference spectra of the FMN triplet and semiquinone forms. Clearly, the early transient only fits well to the spectral features of well established triplet-state difference spectra of flavins in aqueous solution (15,16).
Global Kinetic Analysis of LOV2-Transient absorption spectra are the algebraic sum of absorption spectra of all coexisting intermediate states and are not very informative. The spectra of the individual intermediate states can be obtained from the transient data by using global kinetic analysis, which involves singular value decomposition and exponential fitting. Global exponential fitting of the short time data gave a single decay time constant of 2 s. Adding more exponential decay time constants did not improve the fit significantly. The residuals (difference between spectral data and calculated data from exponential fitting) using one exponential were already within the signal-to-noise ratio in the data. The b-spectra b 1 and b 0 from the exponential fit are shown in Fig. 1B. In general, the b-spectra reflect spectral changes associated with the exponential process and contain useful information about the kinetic scheme. If the decays of kinetic components are well separated in time, the b-spectra can be interpreted as difference spectra between decaying and forming intermediates. By assuming a given kinetic scheme, we calculated the spectra of the intermediates from the b-spectra. Because only one apparent rate was found, a linear scheme with two intermediates is sufficient to account for the data at early times. The observation of an isosbestic point around 420 nm for this transition is consistent with a two-state system. As expected, the calculated difference spectrum of the first light-activated state that we designate LOV2 660 L (see Fig. 4A) 3 coincides with the earliest transient and has a bleach in the 450 nm region that is twice as large as that observed in the subsequent species, which we designate LOV2 390 S . This can be explained by including a back-reaction from the LOV2 660 L species to the ground-state LOV2 447 D . Because global exponential fitting produces only one rate constant, this assumption implies that LOV2 660 L decays to LOV2 390 S and back to the ground state simultaneously as follows.
S 390 SCHEME 1 3 We use the following nomenclature for intermediate spectra. Each intermediate contains a base name with a superscript and subscript. The superscripts we chose are D for dark, L for light-activated, and S for signaling (we have kept them in alphabetical order). The subscripts denote the absorption maximum of the intermediate in its longest wavelength absorption band. For this paper we chose LOV2 as a base name (e.g. LOV2 447 D ). We suggest that as new intermediates are assigned they be named following this convention.   4B), presumably the FMN triplet state, fits well with the triplet state spectrum measured for lumiflavin in solution (19). Because the ground-state spectrum added was obtained in the Hewlett Packard spectrophotometer at much higher spectral resolution than that attainable in our flash spectrometer, we applied a 12-point Savitsky-Golay smoothing to the spectrum to have comparable spectral resolution. The molar extinction coefficients are calculated relative to the published value for LOV2, ⑀ 447 (LOV2) ϭ 13,800 M Ϫ1 cm Ϫ1 (5). 4 LOV2C39A Forms LOV2 660 L , Which Decays Back to the Ground State-Because it was shown earlier that the sitespecific mutant LOV2C39A is apparently photochemically inactive (5), thus implicating Cys 39 as the reactive protein side chain, we carried out laser flash photolysis studies on this mutant. Absorption spectra were collected in the 10 ns to 1 ms time window and are the average of 6 laser flashes. The unexpected result is that LOV2C39A shows absorbance changes at early times nearly identical to those observed for LOV2, but at late times the system returns to the original ground state and does not form the LOV2 390 S metastable species as shown in Fig.  5. Global exponential fitting gives satisfactory residuals (data not shown) with a fit to one time constant of 72 s. The quality of the analysis and the amplitude of residuals for the LOV2C39A data set were comparable with those of the wildtype pigment. We see formation of only one species, which is similar to that found in LOV2 designated LOV2 660 L . As has been indicated previously, the LOV2C39A protein does not form the LOV2 390 S intermediate. The major difference between the native pigment and the LOV2C39A mutant is, therefore, that all of the early LOV2 660 L species decays back into the ground state because it cannot form the cysteine thiol adduct.
Does LOV2C39A Exist As a Thiol (-SH) or a Thiolate(-S Ϫ )?-The typical pK for free thiol is around 8.5, but the values observed in proteins depend on local interactions such as ion pair formation or polarity of the environment. For example, the reaction mechanism of some cysteine-containing flavoenzymes active in the oxidoreduction of disulfide bonds involves the formation of flavin C(4a) thiol adducts as enzymatic intermediates, and the reactive species in those systems appears to be the thiolate ion with pK values well below 8.5 (17). Flavin fluorescence yield is also strongly affected by electrostatic environment and can be fully quenched in the presence of a neighboring thiolate, as was observed in the study of a mercuric ion reductase mutant (17). The close proximity of Cys 39 to FMN in LOV2 (7) allowed us to probe the ionization state of Cys 39 by monitoring pH-dependent perturbations in fluorescence yield. We monitored the pH-dependent fluorescence yield of aqueous FMN, LOV2, LOV2C39A, and LOV2C39S.
The titration of an aqueous solution of FMN (inset, Fig. 6) shows the typical fluorescence decrease that accompanies the deprotonation of N(3) with a pK of about 10 and a similar decrease in fluorescence in the acid region with an apparent pK around 1.7. Flavin fluorescence and absorption spectra are known to be affected by the ionization state of its N(3) atom. Deprotonation of N(3)H is accompanied by a significant decrease in fluorescence and a shift of the near-UV absorption band to shorter wavelengths. The pK of N(3) can be modulated by hydrogen bonding and the surrounding electrostatic environment. The mechanism of fluorescence quenching of flavins in solution at low pH is attributed to collisional quenching by protons (20). Riboflavin, which does not contain a phosphate 4 We are currently studying a construct that contains both photochemically active LOV1 and LOV2 domains. Studies also involve the photochemistry of the entire photoreceptor protein. Clearly a notation is required to indicate the photoactive domain within a given chromoprotein construct. We suggest for these proteins and their photochemical intermediates the following notation. We shall call phot1 and phot2, P1 and P2, respectively, and the LOV1 and LOV2 domains will be abbreviated to L1 and L2. For example when referring to the ground state of the LOV2 domain of a construct containing both LOV1 and LOV2 from nph1, the notation will be (L1L2P1) LOV2 447 D . For studies on LOV2 in the native photoreceptor, the notation would be (P1) LOV2 447 D . To simplify photocycle diagrams, we suggest only the ground state be fully described and the intermediates referred to simply as the active domain (i.e. LOV2 660 L ). In this study we refer to the wild-type chromopeptide as LOV2 and refer to mutants of LOV2 by adding the mutation at the end (e.g. LOV2C39A) (5). group bound to its ribityl chain, shows the same fluorescence titration profile as FMN, indicating that fluorescence yield is not significantly affected by phosphate ionization state. Similar high and low pH fluorescence titration curves have been reported for flavins and flavinyl peptides (21). The other "titratable" groups on the chromophore, such as N(1), N(5), C(2)O, and C(4)O, have pK values of less than or equal to zero and therefore do not contribute to the observed effect (14).
Fluorescence base titration of LOV2 and the two mutants LOV2C39A and LOV2C39S are similar to that of FMN, all showing the decrease in fluorescence associated with N(3) ionization. The reversibility of these titrations in the basic region shows that the N(3) group is readily accessible to bulk protons, presumably via a proton pathway provided by protein residues or structural water, which is present in the three-dimensional crystal structure of Adiantum Phy3 LOV2. In addition, the reversible nature of the titrations shows that the flavin chromophore is stable in the binding pocket at alkaline pH.
Neither FMN nor the two thiol-free mutants showed fluorescence perturbations in the 4 -9 pH range (Fig. 6). A fluorescence change would have been expected for LOV2 in that range indicating deprotonation of the Cys 39 thiol group. Strikingly, the titration of LOV2 does not show any changes in fluorescence in the 4.5-9.5 pH range, indicating that no change in the Cys 39 ionization state is taking place within this pH range.
In the acid region the fluorescence intensities of both LOV2 mutants decrease as observed for free FMN. Proton collisional fluorescence quenching of protein-bound FMN and free FMN (that is slowly released from the protein) contribute to the observed decrease in fluorescence yield. In contrast, the LOV2 protein shows a marked fluorescence increase below pH 4. We interpret this fluorescence increase to reflect the protonation of the Cys 39 thiolate with an apparent pK less than 4. This low pK explains why no fluorescence perturbation was observed in the 4 -9 pH range and suggests that Cys 39 is ionized (i.e. S Ϫ ) in LOV2 under physiological conditions. The calculation of a precise value for the Cys 39 pK is complicated by slow release of FMN from the protein at low pH. This release of chromophore also partially masks the maximally observable fluorescence increase of protein-bound FMN due to thiolate protonation. Maximal fluorescence levels as high as that observed for LOV2C39A were attained at low pH in some experiments. We conducted fast titrations, in which the pH was rapidly changed from 7 to 2.8 and returned to 7 within 60 s, to explore the observed low pH fluorescence changes in more detail. The pH-induced fluorescence changes of LOV2 were almost fully reversible, and chromophore loss was slow (data not shown). Chromophore loss was faster for LOV2C39A than for LOV2 and slower for LOV2C39S than for LOV2. This progression parallels the decrease in polarity and hydrogen bonding potential of the groups at low pH, e.g. -OH Ͼ -SH Ͼ methyl. Chromophore release in LOV2 can also be effected by the addition of iodoacetamide (known to react exclusively with thiolate (22)) at neutral pH, whereas there was no effect on the LOV2C39A mutant (data not shown).
The assignment of Cys 39 as a thiolate is also supported by the relative fluorescence levels observed for LOV2 and the two mutants in the 4 -10 pH range. These fluorescence levels reflect the electrostatic environment of the chromophore. The LOV2C39A mutant (containing nonpolar alanine) shows the highest fluorescence yield followed by LOV2C39S (containing polar serine) and the LOV2 protein, which contains a negatively charged thiolate and shows the lowest fluorescence yield (Fig. 6). A similar effect was noted in thioredoxin reductase, in which a neighboring serine quenched flavin fluorescence to a greater extent than a neighboring protonated thiol (23).
The UV-Visible absorption spectra of LOV2 wild type and the two mutants, LOV2C39A and LOV2C39S, are shown in Fig. 7. The absorption spectra of the three proteins are nearly identical in the blue spectral region but show key differences in the near-UV region. These bands are strongly affected by electron redistribution in the chromophore, such as those associated with deprotonation of N(3) (24). These bands may also reflect changes in the electrostatic environment of the chromophore as well as hydrogen bonding (24,25). The LOV2 protein spectrum shows a single maximum at about 375 nm with a slight shoulder at shorter wavelengths, whereas spectra for both the LOV2C39A and the LOV2C39S mutants have two nearly equal bands. This is consistent with Cys 39 perturbing the local chromophore electrostatic environment to a greater extent than the corresponding group in either LOV2C39S, which contains the polar -OH group or the nonpolar -CH 3 group in LOV2C39A. Because the -SH group is less polar than -OH, the Cys 39 would be expected to affect the electrostatic environment significantly in its negatively charged thiolate form. Although not specifically mentioned in the original cited publications, we have noticed that analogous effects occur in both lipoamide dehydro- genase (18) and in a mutant of the flavoenzyme mercuric ion reductase. Mutating three of the four cysteine residues to alanine in mercuric ion reductase yielded an absorption spectrum with a single near-UV peak (17) similar to that observed for LOV2. The remaining cysteine, Cys 140 , of the mutated mecuric reductase was readily titrated and found to have a pK of about 6.7 (17). If the protein was taken to lower pH (such that the thiol was fully protonated), or if Cys 140 was mutated to serine, the near-UV peak became a double band, confirming the effect of a thiolate negative charge on this near-UV flavin band. D 2 O Exchange-To investigate the possible role of a proton transfer reaction in the rate-limiting step of the back-reaction, we measured the back-reaction rate after exchanging D 2 O for H 2 O. The back-reaction in D 2 O is three times slower than in H 2 O as shown in Fig. 8. The ground-state absorption spectrum of LOV2 in D 2 O is unaltered; the light-induced spectral changes are also the same as shown in Fig. 2. Usually slowing down of a reaction in D 2 O is indicative of hydrogen chemical bond breakage, alteration of hydrogen bonds, proton transfer reaction, and proton diffusion processes. Some of these factors (or a combination of them) are clearly involved in the ratelimiting step of the back-reaction of the LOV2 photocycle.

Protein-Chromophore Interactions in the LOV2 Ground
State-The data presented here strongly suggest that the ground state of LOV2 contains the thiolate form of Cys 39 . This hypothesis is supported by the following observations: (a) the constancy of the fluorescence intensity over a wide pH range (4.5-9.5); (b) the lower fluorescence yield in the wild type versus either of the Cys 39 mutants, suggesting the presence of a quenching charged species; (c) the reversible fluorescence increase for the wild type at low pH, as expected if protonation removes such a charged group, and the fact that neither of the cysteine-deficient mutants shows this low pH titration profile; (d) the near-UV spectral differences between LOV2 and both Cys 39 mutants.
We considered the possibility of hydrogen bonding between the Cys 39 sulfur and flavin N(5) because the distance, measured from the crystal structure, is about 3.6 Å. This distance is within the range of typical hydrogen bonds. However, we cannot fully account for all the above observations with arguments based solely on such an interaction.
A thiolate must be stabilized by direct interactions such as ion pair formation with a proton donor/acceptor group (22,26,27), hydrogen bonding (28,29), and charge transfer complexes (to flavin (17)) and/or by indirect stabilization such as helix dipole interactions (30). We find no evidence for the expected long wavelength absorption typical of flavin-sulfur charge transfer complexes. A charge pair formation by partial or total proton transfer, with or without involvement of hydrogen bonding interactions, from the thiol to a neighboring group(s) seems the most plausible. The crystal structure of LOV2 (phy3 LOV2) (7) shows no basic residues (arginine, lysine, etc.) in the immediate proximity of Cys 39 . At longer distances basic residues are plentiful, but it is not clear which one (if any) could be involved in the photocycle reaction. It has been shown in DsbA, a protein containing a reactive thiolate with a pK value around 3.5, that site-specific replacement of all nearby charged residues did not modify the low cysteine thiol pK (31). The mechanism for thiolate stabilization in DsbA remains unclear. We therefore cannot identify on a structural basis which amino acid side chain stabilizes the thiolate and acts as the proton donor/acceptor group in the photocycle scheme. However, the crystal structure shows a water molecule in the vicinity of the FMN chromophore (7). It has been demonstrated that in bacteriorhodopsin structural water is required to stabilize the protein and participates in intramolecular proton transfer reactions during the proton-pumping photocycle (32,33). This water itself could be either the proton donor/acceptor group, or it could act as a bridge between the cysteine and another remote amino acid side chain that donates/accepts the proton. It might also be possible that no single amino acid side group acts as the proton donor/acceptor but rather that a group of side chains and structural water provides an electrostatic environment that is modified by protein conformational changes associated with the LOV2 photocycle. In bacteriorhodopsin several protein groups and structural water contribute to the protonated Schiff base counterion (34,35) and to the proton transfer reaction associated with the ion pumping process (35).
The LOV2 Photocycle and Its Mechanism-We have identified only one transient species preceding the formation of the metastable state LOV2 390 S in LOV2. We chose the superscript S to indicate a possible signaling state of the protein, as has been found for the longest lived intermediates of other sensory systems (36). Upon absorption of a photon, LOV2 presumably is raised to its singlet excited state, which decays in times shorter than our time resolution (ϳ30 ns) to form an intermediate that we call LOV2 660 L . As stated earlier, the spectral properties of   (5), the singlet excited state must undergo singlet/triplet intersystem crossing with an efficiency as high as 88%. The LOV2 390 S intermediate spectrally resembles the mercuric ion reductase thiol-C(4a) adduct (17). Based on this similarity it was previously proposed that the LOV2 390 S intermediate involves formation of a covalent bond between the LOV2 protein and the FMN chromophore (5). Fig. 9 shows the kinetic scheme for the LOV2 photocycle and the details of the proposed mechanism. Because the ground state of the pigment is thought to contain an ionized Cys 39 , we have to postulate the existence of a proton donor/acceptor group in the protein that actively participates in the reaction mechanism. The proton held by this group is donated to the flavin N(5) during the formation of the postulated C(4a)-thiol adduct. This latter bond is presumably formed by nucleophilic attack of the sulfide of Cys 39 on the C(4a) carbon of FMN. This reaction should be strongly favored because the well known charge redistribution occurring in the triplet state polarizes the N(5)-C(4a) double bond (37,38). The carbon acquires in this process a fractional positive charge and the nitrogen a fractional negative charge increasing significantly the N(5) pK (38).
We propose the following mechanism. The increase in basicity causes the N(5) to become protonated by a proton-donating group in the protein. Upon protonation of N(5), the N(5)-C(4a) double bond becomes a single bond leaving a very reactive C(4a) carbo-cation that undergoes attack by the cysteine sulfide, resulting in formation of the long lived intermediate LOV2 390 S . In addition, the distance between the C(4a) carbon and the Cys 39 thiolate is expected to change upon creation of a tetrahedral sp 3 carbon from a planar sp 2 . A decrease in distance could be important in the progress of this reaction. The C(4a) adduct contains an asymmetric chiral C(4a) center that may be responsible for the large light-induced chromophore circular dichroism changes observed previously (5). The proton transfer reaction therefore initiates the process. In the absence of the large pK shift induced by light activation this reaction cannot occur. In the model the sharp increase in the pK a of N(5) is the "molecular switch" that drives the photoreaction.
Our mechanism for the formation of the flavin-thiol adduct is based largely on that proposed by Miller et al. (17) for mercuric ion reductase in which a thiolate (generated by reduction of a cystine S-S bond) binds spontaneously to the flavin C(4a). The main difference is that in LOV2 the thiolate is the stable chromoprotein form, and modulation of the N(5) pK by light activation promotes the formation of the flavin-cysteine adduct. A mechanism using similar arguments but not involving a stable Cys 39 thiolate has been recently proposed for LOV2 (7).
The back-reaction must involve a reversal of this scheme. The ionized group that donated the proton in the forward reaction is now a base. In a base-catalyzed reaction, a proton from N(5) is abstracted by this group, resulting in the formation of LOV2 ? B . In this configuration the flavin and the protein are presumably in strained conformations and return to a lower energy state by breaking the S-C(4a) bond and regenerating the C(4a)-N(5) double bond. The back-reaction is three times slower in D 2 O as in H 2 O, indicating that formation or breakage of bonds involving hydrogen atoms and/or proton transfers are rate-limiting steps during this back-reaction. Interestingly, the yield of LOV2 390 S in D 2 O is nearly identical to that in H 2 O, suggesting that such processes are not rate-lim-iting in the forward reaction. The back-reaction time constant measured here is significantly slower than that reported previously (5). Because the kinetics of back-reaction are temperature-dependent (slower by about a factor of 7.5 for a 25°C change (5)), a small difference in temperature may explain the observed differences.
At long times, the LOV2C39A mutant showed an apparent absence of photochemical activity (5). Our work shows that this mutant is competent in the early photochemical steps, but the absence of Cys 39 precludes the formation of the LOV2 390 S adduct. The triplet state of this mutant decays back to the ground state with a time constant of 72 s, more than one order of magnitude longer than LOV2. The presence of a charged sulfur must significantly enhance the rate of triplet decay presumably by increasing the probability of spin flipping.
In other photoreceptor systems that undergo photocyclic reactions such as visual pigments and bacterial sensory rhodopsins, it has been shown that the longest lived metastable intermediates in the photocycles function as signaling states (36). We tentatively assign to the signaling state role the phototropin metastable adduct state, LOV2 390 S , and label it with S as superscript. It is clear that the phototropic response is the result of a complex signal transduction system. The signal provided by the initial photochemical product must be coupled to the rest of the transduction system by a primary signal transducer that has yet to be identified.