Variations in protein/flavin hydrogen bonding in a LOV domain produce non-Arrhenius kinetics of adduct decay

Abstract

Light Oxygen Voltage (LOV) domains utilize a conserved blue light-dependent mechanism to control a diverse array of effector domains in biological and engineered proteins. Variations in the kinetics and efficiency of LOV photochemistry fine tune various aspects of the photic response. Characterization of the kinetics of a key aspect of this photochemical mechanism in EL222, a blue-light responsive DNA binding protein from Erythrobacter litoralis HTCC2594, reveals unique non-Arrhenius behavior in the rate of dark state cleavage of the photochemically-generated adduct. Sequence analysis and mutagenesis studies establish that this effect stems from a Gln to Ala mutation unique to EL222 and homologous proteins from marine bacteria. Kinetic and spectroscopic analyses reveal that hydrogen bonding interactions between the FMN N1, O2 and ribityl hydroxyls with the surrounding protein regulate photocycle kinetics and stabilize the LOV active site from temperature-induced alteration in local structure. Substitution of residues interacting with the N1-O2 locus modulates adduct stability, structural flexibility and sequestration of the active site from bulk solvent without perturbation of light-activated DNA binding. Together, these variants link non-Arrhenius behavior to specific alteration of an H-bonding network, while affording tunability of photocycle kinetics.

The ability of Light, Oxygen, Voltage (LOV) domains to couple environmental stimuli to alterations in cellular signaling has led to the development of bioengineered tools to manipulate biological pathways in vitro and in vivo (1–5). These systems use a conserved signaling mechanism based on blue light dependent formation of a covalent cysteine-flavin adduct (6). Bond formation is coupled to reorientation of protein elements adjacent to the LOV domain, regulating their participation in protein:protein interactions or modulating the activity of kinases, phosphodiesterases, DNA binding or other diverse effector families (6–11). Recently, it has been demonstrated that LOV domains can be engineered to confer light-dependent control to non-photosensitive enzymes (e.g. histidine kinase (5), GTPase (3)) and DNA binding proteins (2). The utility of both native and engineered LOV proteins has spurred interest in characterizing and manipulating their basic photochemical mechanisms, including regulation of photocycle kinetics and signal transduction (6, 12–15).

Studies which have generated LOV variants with altered photocycle kinetics can largely be divided into two camps: Those which have mutated conserved residues in or near the flavin-binding pocket (e.g. the adduct-forming active site Cys) (12, 16–18) and those which have identified naturally varying residues that direct the diverse range of photocycle kinetics observed in natural systems (15, 19, 20). Whereas the former method has provided keen insight into the biophysics of adduct formation (12, 21, 22), as well as the development of in vivo tools (1), these variants can alter the signaling mechanism of LOV domains (12, 18). Alternatively, the latter strategy has identified naturally-occurring residue selections that can regulate adduct stability with minimal effects on the signaling mechanism (15).

Mutational analyses of conserved residues within LOV domains have revealed mechanisms of adduct formation, regulation of adduct formation kinetics and yield, as well as factors regulating adduct stability. For instance, initial studies of photocycle-altering variants focused on the conserved Cys moiety that is directly involved in adduct formation. Time-resolved UV-visible absorbance and EPR measurements demonstrated rapid formation of a flavin triplet state that readily formed a radical intermediate in Cys→Ala and Cys→Ser variants (21, 22). In addition, random mutagenesis targeted at identifying rapid cycling phenotypes indicated that steric interactions with the active site Cys can appreciably affect the rate of adduct decay (16). Complementary targeted mutation of conserved active site residues confirmed the importance of Cys and flavin steric contacts (13, 17, 18), H-bonding interactions within the flavin core (12), as well as solvent and gas access to the active site as a means of kinetic control (17). While these variants can tune adduct lifetime over two orders of magnitude, they can also destabilize flavin binding and alter signaling pathways (12, 18).

In parallel with this approach, detailed analyses of naturally-occurring variants have identified sites that tune adduct-state lifetime over four orders of magnitude with minimal effects on the photochemical mechanism (15). In addition to steric effects at the active site Cys, this study demonstrated that the primary mechanisms mediating diverse LOV adduct-state decay kinetics include electronic and steric alteration of the flavin reduction potential as well as solvent access to the active site. Notably, crystallographic studies of these variants revealed no alteration in protein structure or function despite the significant alteration in photocycle kinetics. Similarly, comparisons of the sequences of two homologous bacterial LOV proteins with distinct photocycle kinetics identified a key Arg residue interacting with the ribityl backbone that dictates a fast or slow cycling phenotype (19). This study demonstrated that changes to elements outside of the LOV domain can have over 100-fold effects on photocycle kinetics as well, potentially by allosteric modification of LOV structure.

Importantly, studies of diverse LOV proteins suggest that their photocycle kinetics may be perturbed by structural changes in the surrounding protein. Initial studies of plant phototropin LOV1 and LOV2 domains revealed distinct differences in cycling kinetics, with LOV1 domains generally having longer-lived adducts than their LOV2 counterparts (20). FTIR, circular dichroism (CD) and NMR experiments indicate that LOV2 domains undergo a partial disordering of local structure following photoexcitation (20, 23–25), whereas LOV1 domains do not exhibit the same flexibility. These differences have primarily been attributed to a single residue at the re-face of the flavin, which differs between LOV1 (Leu) and LOV2 (Phe) (20). Similarly, Laser Induced OptoAcoustic Spectroscopy (LIOAS) data revealed that mutations that alter photocycle kinetics and yield can also change the magnitude of light dependent volume changes (12). However, the role of conformational fluctuations in mediating adduct state stability is still largely unknown.

We have obtained some insight into these issues with our studies of a bacterial LOV-HTH (LOV-helix-turn-helix) protein EL222 (26). Identified in the marine bacterium Erythrobacter litoralis HTCC2594, this protein contains a photosensitive LOV domain and a DNA-binding HTH domain effector. Structural and functional characterization of EL222 established that photochemical changes within the EL222 LOV domain lead to separation of the LOV and HTH domains and subsequent light-induced DNA binding (26). Herein, we characterize the kinetics of adduct cleavage in EL222 and identify a new natural LOV variant that alters the lifetime of the photoinduced Cys/flavin adduct, while maintaining native protein conformation and signaling mechanisms. Moreover, we demonstrate unique non-Arrhenius behavior indicating temperature dependent conformational fluctuations that also affect adduct stability.

Materials and Methods

Cloning and Protein Purification

Rate-altering variants were produced using the QuikChange protocol (Stratagene) within the context of an N-terminally truncated LOV-HTH construct of EL222 (14–222) with an N-terminal His₆ tag (26). Mutations were verified by DNA sequencing (UT Southwestern Sequencing Facility). All LOV-HTH (14–222) and LOV only (14–144) EL222 variants were expressed in E. coli BL21(DE3) cells. Cells were induced with 0.1 mM Isopropyl-Thio-Galactoside (IPTG) for 22 hr at 18°C. Cells were harvested, pelleted and stored in 100 mM NaCl, 50 mM Tris (pH 8.0) buffer.

Proteins were purified by Ni-NTA chromatography at 4°C with gradient elution from 10 to 500 mM imidazole in 100 mM NaCl, 50 mM Tris (pH 8.0) buffer. The His₆ tag was removed via incubation with His₆-tagged TEV protease (27) overnight at 4°C. An additional round of Ni-NTA chromatography was conducted to remove His₆-TEV and the His₆ tag. EL222 samples were then subjected to Superdex S75 size-exclusion chromatography with 100 mM NaCl, 50 mM Tris(pH 8.0) buffer for all subsequent analysis.

UV-Visible Absorbance Spectroscopy and Kinetics

UV-visible absorbance spectroscopy of EL222 variants was conducted on a Varian Cary 50 spectrophotometer. Protein samples were at a concentration of approximately 40 μM and measured using a 1 cm pathlength cuvette at temperatures between 285 and 328 K.

Kinetics measurements of EL222 dark state recovery were obtained by measuring the absorbance at 450 nm after excitation with a camera flash. Absorbance measurements were made at a sampling frequency varied to provide approximately 30 measurements per half-life, ensuring sufficient data coverage while not continuously illuminating the sample. Solvent isotope effect (SIE=k_H2O/k_D2O) measurements were made similarly, using a 99% D₂O-containing form of the buffer.

We evaluated both mono- and biexponential fitting to extract kinetic parameters from these decays. While monoexponential fitting adequately modeled most decays (τ~31 s at 298 K; Figures 1A, S1; Table S1), we noted non-random residuals at the beginnings of some recoveries that suggested a minor fast component might be present. Biexponential fitting revealed these two processes (τ~28 s and ~4.7 s at 298 K), with significant sample-to-sample variation in amplitude between 5–20% of the recovery. Coupling this variation with instrumental limitations on sampling rate which complicate precisely fitting this faster process at high temperature, we opted to use monoexponential fitting to determine rate (k) and time constants (1/k) from our kinetic data. Critically, comparisons of Arrhenius plots of the rates extracted from both approaches showed the same non-Arrhenius behavior for rates from mono-exponential fits and the slow component of the bi-exponential fits (Figure S1), establishing the validity of this approach. Additional commentary on this point is provided in the Supplementary Information.

Figure 1.

Effects of temperature on EL222 dark state recovery kinetics and structure. A) Dark state recovery kinetics of WT EL222 (14–222) at 297 K as monitored by increased absorbance at 450 nm post-illumination. We noted that these recovery curves can be fit by either mono- (black) or biexponential curves (red), with some datasets fitting slightly better to biexponential curves at initial timepoints (insets show both initial recovery points and residuals; additional discussion in Supplementary Materials). B) Non-Arrhenius behavior of dark state recovery of WT EL222 (14–222). At low temperatures (< 318 K), Arrhenius behavior is observed with a E_a of 63 kJ/mol; at approximately 318 K, we observed a transition to a second linear regime with a higher E_a of 115 kJ/mol. C) Circular dichroism spectra of EL222 as a function of temperature. The shape of the CD spectra is temperature independent; minor loss of intensity reflects precipitation at high temperature. D) At 328 K noticeable loss of vibronic structure in the FMN absorbance is noted (red) compared to WT at 278 K (black). Incubation back at 278 K for ten minutes (including cooling time) results in partial return of vibronic structure (blue).

To estimate errors in parameters extracted from Arrhenius and Eyring analyses, we recorded kinetic data from each of three separate samples of WT EL222(14–222), independently calculating thermodynamic parameters from each of these datasets. Averages and standard deviations of the mean were calculated for E_a, ΔH^‡ and ΔS^‡ for these data (Table 1). Analyses for other proteins were acquired from single samples, which we assume to have comparable errors to the WT EL222(14–222) given comparable data quality.

Table 1.

Kinetic and Thermodynamic Parameters for EL222 Dark State Reversion

variant	rate (1/s)	τ (s)	Ea (kJ/mole)	ΔH‡ (kJ/mole)	ΔS‡ (kJ/mole, 318K)
WT (14–222)	0.034	29	63±2, 115±4b	61±3, 107±5	−22±10, 26±10
WT (14–144)a	0.045	22	75	72	−9.1
A79Q (14–222)	0.0044	227	100, 138	97, 135	12, 51
A79R (14–222)	0.37	2.7	64, 108	63, 109	−13, 56
A79T (14–222)c	0.11	8.9	68, 141	70, 124	−11, 39

Kinetic parameters (rate, τ) were measured at 298 K, while thermodynamic parameters E_a, ΔH^‡ and ΔS^‡ are determined by fits to eight temperature points (285–308 K) for the low temperature regime and the four highest temperature points (319–328 K) for the high temperature regime (values indicated by underline).

^aWT (14–144) shows only linear Arrhenius behavior, so only one set of thermodynamic parameters are reported.

^bError is reported as the standard deviation of the mean for three independent measurements (Materials and Methods), which we assume to be applicable to all datasets given similar data quality among all.

^cThermodynamic parameters for A79T at high temperature may be imprecise due to difficulties with fitting the curved Arrhenius and Eyring plots exhibited by this variant.

CD Spectroscopy

CD spectroscopy measurements of EL222 variants were conducted using an Aviv model 62DS spectrometer. Samples were incubated for 10 min at the stated temperatures prior to data collection. Spectra were obtained between 260 to 190 nm in 1 nm steps, with 3 s signal averaging at every wavelength. Spectra were corrected for buffer and averaged over 3 scans.

Nuclear Magnetic Resonance

Solution ¹⁵N/¹H HSQC spectra (28) were collected on 200–300 μM samples on a Varian 800 MHz spectrometer equipped with a cryogenically-cooled, triple resonance probe, with samples at 298 K. Light state samples were collected as described previously (26). All data was processed using NMRPipe (29) and analyzed using NMRViewJ (30).

Results

EL222 exhibits a typical LOV domain photocycle (26) with the photochemical formation of a cysteinyl-flavin C4a adduct leading to a significant decrease in visible absorbance above 400 nm. Upon ceasing illumination, the adduct spontaneously decays with a time constant of approximately 30 s at 298 K that we observed with either mono- or bi-exponential fitting of the kinetic traces (Figure 1A, Table 1; Supplementary Table S1). We observed a fast component in biexponential fits of certain traces (τ~4.7 s at 298K in Figure 1A), but this minor component varied by sample between 5–20% of the total recovery amplitude and was difficult to precisely characterize at high temperature. As such, we conducted all further analyses using monoexponential fitting. Analysis of the temperature dependence of the dark state recovery kinetics revealed unusual non-Arrhenius behavior, with two distinct states exhibiting different linear Arrhenius dependencies (Figure 1B, Table 1). As such, the two regimes differ in their relative activation barriers and corresponding entropic and enthalpic contributions to adduct stability. At temperatures below 318 K, EL222 displays an activation barrier of 63±2 kJ/mole. Above 318 K, we observed an abrupt transition to a second linear regime with a 1.8-fold elevated activation energy of 115±4 kJ/mole (Figure 1B, Table 1). Eyring plot analysis reveals that the two regions differ in their enthalpic and entropic contributions (Table 1), with a more favorable entropy of activation in the high temperature regime than the low temperature region (26±10 kJ/mole vs. –21±10 kJ/mole at 318 K respectively).

While LOV domains typically exhibit linear Arrhenius behavior for the temperature dependence of adduct scission (12), non-linear Arrhenius behavior has occasionally been observed in biochemical and chemical systems. This can range from curved Arrhenius behavior, as seen in the decay of the pB signaling state in photoactive yellow protein (31), to abrupt transitions between two linear Arrhenius regimes as occasionally observed in a variety of enzymes (32). These are often interpreted as involving a structural transition that alters or inactivates the catalytic mechanism. While these transitions most commonly link a higher-activity state at low temperature with a lower-activity high temperature state, we strikingly observed that adduct scission in EL222 occurred more rapidly at high temperature than expected from low temperature data. We suggest that structural or mechanistic perturbation of LOV chemistry in EL222 could stem from a variety of sources, including disruption of the FMN binding pocket via local alteration in secondary structure or a global dislocation of the HTH domain from the LOV core (which may alter the LOV β-sheet, immediately adjacent to the flavin itself).

Solvent isotope effect (SIE) experiments were conducted to determine whether there is a change in the adduct scission mechanism at high temperature. Interestingly, EL222 displays a high SIE of approximately 4 between 285 and 328 K (Supplemental Figure 2). The temperature independence of the SIE suggests that proton abstraction from N5 is rate limiting (15, 16) at all temperatures, eliminating an in the LOV mechanism as the source of the non-Arrhenius behavior. Rather, the unusual break in the linear Arrhenius dependency likely results from temperature-dependent structural perturbation.

To evaluate the role of the HTH domain in the non-Arrhenius behavior of EL222, we examined the temperature dependence of the kinetics of adduct-scission in the isolated LOV domain (residues 14–144, compared to the LOV-HTH 14–222 construct). We observed only one linear regime in the isolated LOV domain (Supplemental Figure 3), but we note that the construct irreversibly aggregates above 318 K preventing data collection at these higher temperatures. The truncated protein also demonstrated minor alteration in the energetics of adduct scission with a 8 kJ/mole higher energy of activation compared to a LOV-HTH construct (Table 1). The increase in energy of activation is offset by a more favorable entropy of activation (−9.1 kJ/mole at 298 K). The minor elevation in activation barrier may indicate that, whereas the HTH domain stabilizes the solubility of EL222 at elevated temperatures, it may slightly destabilize the light-state adduct. Thus, we must consider the effect of alteration in local structure on the non-Arrhenius behavior.

To test this possibility, we used CD and UV/visible absorbance spectra as probes of secondary structure and local flavin environment. Near-UV CD spectra of dark-state EL222 indicated limited change in local secondary structure at temperatures between 297 and 325 K (Figure 1C). Notably, we observed some loss of signal at elevated temperature, consistent with minor aggregation observed in UV-visible absorbance spectra above 321 K. Notably, UV-visible absorbance spectra also identified a partially reversible loss of FMN vibronic structure above 318 K (Figure 1D). The decrease in vibronic structure and lack of secondary structural changes are consistent with an increase in the plasticity of the active site that may be coupled to increased solvent access at increased temperatures. Our data further indicate that most, but not all, protein in these samples recovers sufficiently to rebind FMN after such heat treatment but we have not investigated this in detail.

Role of Active Site Residues in Modulating Non-Arrhenius Behavior

To our knowledge, non-Arrhenius behavior of LOV photocycle kinetics has not previously been reported, leading us to compare the EL222 sequence to other LOV domains to look for possible contributors. We found that the terminal Gln residue found in the canonical LOV active site sequence (GRNCRFLQ; (33)) is changed in EL222 to an Ala (GRNCRFLA) at position A79 in EL222. Examination of other LOV sequences indicated that mutations at the conserved Gln position are unique to EL222 and closely related LOV-HTH homologs from other marine bacteria (Figure 2A), showing that this site can accommodate Arg, Thr and Cys replacements. Notably, this is specific to the LOV-HTH proteins, as other LOV domains from the same bacteria contain the canonical Gln residue.

Figure 2.

Structural and functional effects of natural sequence variants at the A79 position. A) Conserved LOV residues (blue) include variation at a conserved glutamine residue as seen in five LOV-HTH proteins (above dashed line) from Erythrobacter litoralis (EL222), Novosphingobium aromaticivorans Saro_1231 (N. ar), Novosphingobium nitrogenifigens (N. ni), Sphingopyxis alaskensis Sala_1000 (Sp A.), and Sphingomonas (Sphin); these are aligned with LOV domains from selected proteins without HTH domains (below dashed line) including the three other E. litoralis LOV proteins (EL346, EL362, EL368), Avena sativa phototropin 1 LOV2 (AsLOV2), Arabidopsis thaliana phototropin 1 LOV1 (AtLOV1) and Neurospora Vivid (VVD) A conserved GRNCRFLQ is altered in five bacterial LOV-HTH proteins (orange). B) Kinetic traces of adduct cleavage post-illumination as monitored by absorbance at 450 nm at 298 K, with all curves normalized to the same asymptotic A₄₅₀. Two variants accelerate the rate of adduct decay (A79R, A79T) compared to wildtype, while A79Q stabilizes the adduct. C) The conserved Gln residue in most LOV domains forms a H-bonding network (highlighted with the white circle) with groups on the FMN ring and ribityl chain (Gln79 interacting with ring N1 and O2, plus a ribityl hydroxyl as seen in VVD (10). D) The Ala79 in EL222 cannot form H-bonds to either N1 or O2 (26). Note that Gln27 from another molecule in the crystallographic unit forms an H-bond to the backbone hydroxyl. E) Modeling suggests that artificial introduction of a Thr at position 79 in EL222 (grey) can H-bond to O2, but not the N1 or backbone hydroxyl groups.

Examination of the structures of EL222 and other LOV domains reveals that these substitutions alter hydrogen bonding interactions between the protein and the N1-O2 positions on the isoalloxazine ring (Figures 2C–E). Specifically, the terminal Gln residue hydrogen bonds with multiple positions on the isoalloxazine ring (3.5 Å Gln Nε FMN N1; 3.0 Å Gln Ne – FMN O2 as observed in VVD (10)) with an additional interaction to the FMN O4′ ribityl hydroxyl group (3.0 Å) (Figure 2C). The naturally-occurring Ala in EL222 both removes hydrogen bonding groups at this position and truncates the sidechain, potentially increasing solvent access to the flavin chromophore (4.5 Å Ala Cβ - FMN O2; Figure 2D). In contrast, models of an A79T substitution in EL222 suggest that it could form a Thr Oγ1 – FMN O2 H-bond (~3.1 Å, Fig 2E); however, this would be unable to interact with N1 or O4. Similar models of an A79R mutation suggest that the longer Arg side chain could be capable of interacting with either O2 or N1, but this would require significant alteration of the local structure to avoid steric clashes with the FMN ribityl side chain. These naturally occurring variants provide an opportunity to probe the effects of H-bonding interactions at the N1 and O2 positions of the flavin ring and to evaluate the effect of mutation of the canonical Gln residue on structure, function and photocycle kinetics.

By making these A79 variants in EL222 and characterizing aspects of their photocycle, we found that this position substantially influences the LOV photocycle in a residue-dependent manner that depends on sidechain H-bonding capabilities (Figure 2). Restoring the canonical Gln residue decreases the rate of adduct scission ten-fold (Table 1), while maintaining structure and function consistent with WT protein (Figure 2C, Supplemental Figure 1) as indicated by solution NMR spectra and electromobility shift assay (EMSA) analyses of the WT and A79Q proteins. Conversely, A79R and A79T variants increased the rate of adduct scission, with the positively-charged A79R variant (τ=2.7 s) undergoing dark reversion faster than the neutral A79T (τ=8.9 s) (Table 1). Moreover, similar to WT EL222 all A79 variants exhibited non-Arrhenius behavior with two regimes differing in thermodynamic properties (Figure 3, Table 1), but with markedly different changes in the height of activation barriers and transition temperature variations depending on the A79 substitution.

Figure 3.

Effect of A79 variants on non-Arrhenius behavior in adduct scission rates. A) The A79R variant retains non-Arrhenius behavior similar to WT with a stronger temperature dependence at high T. Open (closed) symbols are data assigned to the high (low) temperature regime, with lines obtained by linear least-squares fitting to the four highest (lowest) temperatures. B) The A79T variant has non-Arrhenius behavior with a more gradual transition than seen in WT and A79R variants, indicative of a mixed population of the two states at intermediate temperatures (half-filled symbols). C) A79Q exhibits only minor deviation from typical Arrhenius behavior, even up to 331 K.

Among the variants surveyed, A79T demonstrated the largest discrepancy in activation energies between the two structural states, with a 141 kJ/mole activation energy above 318 K vs. 68 kJ/mole at lower temperatures (Tables 1, 2). Conversely, the long-lived A79Q variant had the smallest difference between the two states (E_a=138 kJ/mole vs. 99.5 kJ/mole), complicating observation of the transition between the two states somewhere above 321 K in this protein. Interestingly, the A79T variant also has non-Arrhenius behavior, with a very gradual transition between the two states in contrast with the abrupt change seen in A79R and WT EL222 (Figure 3). In this regard, A79T likely exists as a mixture of the two states, whose equilibrium population is dictated by temperature. Thus, A79 variants capable of H-bonding to O2 seem to stabilize the structural plasticity observed in EL222.

Examination of the temperature-dependent kinetic properties also indicates that the H-bonding properties to O2 and N1 correlate with the relative energy barriers and the observed kinetics. Not surprisingly, the rapid cycling A79R variant demonstrates the lowest activation barrier (64 kJ/mole) during dark state recovery, whereas the long-lived A79Q has the highest activation barrier (100 kJ/mole). On the other hand, the rapid cycling A79T variant has an activation barrier (68 kJ/mole) similar to WT EL222 (63 kJ/mole). Examination of the relative enthalpic and entropic contribution reveals that A79T has a more favorable entropic contribution (−10 kJ/mole) vs. WT (−20 kJ/mole) at 298 K. Similar to A79T, the A79Q variant also exhibits a significantly more favorable entropic contribution (11 kJ/mole at 298 K), although it is not large enough to offset the activation barrier. Thus, H-bonding interactions at N1-O2 entropically stabilize the light-state adduct.

The origins of the activation energy stabilization could result from changes in solvent accessibility, steric hindrance of active site residues, perturbation of the electronic state of the FMN chromophore or alterations of an H-bonding network involving the N1, O2 and N3 positions of the flavin ring. To better clarify the mechanism of photocycle perturbation in WT EL222 and A79 variants, we analyzed the proteins for their sensitivity to base catalysis and structural or electronic alteration via NMR analysis.

Effect of H-bonding Network and Electronic Stabilization

The N3 position of the flavin ring is protonated in both the dark- and light-state of LOV proteins and is readily observable in ¹⁵N-¹H HSQC spectra with characteristic far downfield chemical shifts on both nuclei. Moreover the N3-H is involved in a network of H-bonding interactions involving several protein sidechains and the O2 and O4 carbonyls and N1 positions of the flavin ring (Figures 2C–E, and 4A). Thus, the N3-proton functions as a probe for alterations in H-bonding interactions at the N1-O2 position. Introduction of A79R, A79T and A79Q variants all result in an upfield shift of the N3-H (Figure 4A), consistent with increased electron density on the N3 proton and increased H-bonding interactions at N1-O2. This is consistent with weakening of the hydrogen bond between the N3 and Asn107 residue, perhaps from an increase in H-bonding interactions at the neighboring N1-O2 position. Moreover, with the exception of WT EL222 where A79 is incapable of H-bonding to N1-O2, the degree of chemical shift directly correlates with alteration in adduct state stability.

Figure 4.

A79 variants depict alteration in N3 H-bonding interactions and susceptibility to base catalysis. A) The ¹⁵N/¹H chemical shifts of the N3 position of the flavin isoalloxazine ring are altered in similar magnitude and direction in both the light and dark states of WT, A79R and A79T. Notably, the A79Q variant demonstrates larger chemical shift changes. B) Effect of imidazole on the catalyzed rate constant, showing that the rate of adduct scission in A79Q is very weakly dependent on imidazole compared to WT, consistent with reduced solvent access. In contrast, A79T and A79R show more pronounced solvent access with correspondingly stronger effects of imidazole on rate enhancement.

Importantly, additional alterations in chemical environment are evident from the relative ¹H chemical shifts of the N3-bound proton between the A79 variants in the dark and lit states. Whereas the chemical shift changes resulting from introduction of Thr or Arg residues at the Ala79 position are similar in both states, the A79Q variant has significantly greater upfield shifts in both spectra (Figure 4A). This is consistent with stabilization of additional electron density near the N3 nitrogen, altering the electronic environment to stabilize the light-state adduct.

Solvent Accessibility and Rate Catalysis in LOV Domains

Studies of other LOV proteins indicate that solvent access and steric stabilization are primary mediators of adduct-scission kinetics (15–17). While the latter cannot be directly observed experimentally, solvent access can be probed by the ability of imidazole to access the active site and catalyze the recovery process (15). In the absence of an external base, solvent can catalyze adduct reversion. Greater solvent access should result in a decrease in the relative energy of activation (catalytic activation). Indeed, the active site accessibility has a direct correlation with the energy of activation in the A79 series (Fig 4B). Similarly, the strong H-bonding character of the A79Q variant also results in a more closed active site structure, inhibiting solvent and imidazole based catalysis. The A79T and A79R variants exhibit progressively increased kinetic acceleration with imidazole (57-fold and 100-fold relative to A79Q, respectively) that correlates with increases in the activation barrier. Importantly, WT EL222 is four-fold less susceptible to kinetic acceleration with imidazole than A79T and seven-fold less than A79R variants, yet demonstrates an identical activation barrier as A79T.

Discussion

Here we demonstrate that a critical aspect of LOV photophysics in EL222, the lifetime of the photoinduced adduct upon dark state reversion, exhibits unusual non-Arrhenius behavior. This results from increased solvent access, increased flavin vibrational freedom and potentially involves structural changes between the LOV core and HTH domain. Kinetic and sequence analyses reveals that a key H-bonding residue that is conserved in most LOV proteins contributes to non-Arrhenius behavior and regulates adduct stability. Introduction of the Gln conserved in other LOV proteins significantly reduces the non-Arrhenius dependency and stabilizes the light-state adduct. NMR, base catalysis and kinetic studies reveal that the H-bonding property of the A79 position is directly correlated with adduct stability, with Gln residues optimally closing off solvent access, while maintaining three H-bonding contacts, (N1, O2 and ribityl O4^*) that rigidify LOV structural dynamics.

Non-Arrhenius behavior has been observed in a variety of enzymes, where increased temperature most commonly induces a less-active conformation (32). Studies of such enzymes demonstrated that the nature of the deviation from a linear temperature dependency may inform on the difference in mechanism between the two temperature regimes (32, 34, 35). In these cases, the deviation can display concave or convex curvature (31, 35) or sharp transitions between two linear regimes (32, 34). Applying this background to EL222, it is clear that the hydrogen-bonding capabilities of our A79 variants dictate the magnitude and type of non-Arrhenius deviation. The sharp transitions observed in WT, A79R and A79Q indicate the presence of two distinct conformations that differ in reactivity or mechanism. The minor temperature-dependent changes in CD spectra clearly indicate that any structural changes must be confined to loop regions or involve subtle alteration of H-bonding networks. Position 79 is located in the Eα-Fα loop and anchors the FMN in other LOV domains through H-bonding interactions between the Gln side chain and the N1-O2 and O4^* positions. The marked decrease in non-linear Arrhenius behavior in A79Q variants indicates the importance of this H-bonding locus in maintaining fidelity of the FMN pocket. Further, the curved transition in A79T variants is reminiscent of non-Arrhenius behavior in PYP, where the photoactivation process is coupled to temperature-dependent partial unfolding (31). Similarly, a single H-bond to O2 in the EL222 A79T variant stabilizes the LOV core compared to WT; however, this change is insufficient to prevent disruption of Eα-Fα loop contacts with FMN. Rather, increased temperatures result in a gradual transition between the two structural states. Combined, our A79 variants demonstrate a lack of H-bonds to the active site flavin introduces plasticity to the LOV active site and modulates adduct lifetime.

H-bonding interactions at the N1-O2 locus rigidifies the FMN binding pocket and induces stabilization of the light-state adduct. Previous studies of the LOV protein VVD indicated that the light-state can be stabilized via steric and electronic perturbation of the active site flavin, regulation of solvent access to FMN and via alteration of the electron density at the re-face of the active site flavin (15). The N1 position adjacent to the surface Eα-Fα loop provides a direct route to affect all three of these mechanisms. Disruption of Gln-N1-O4^* interactions in WT and A79T variants provides a solvent accessible surface to the active site Cys and FMN moieties. Indeed, imidazole catalysis reflects increased solvent access and a direct correlation between susceptibility to base catalysis and enzyme kinetics. Further, the presence of a positively charged residue near N1 and H-bonding interactions to the N1 position are direct modulators of flavin reduction potentials in flavo-enzymes (36, 37). These H-bonds favor flavin reduction and stabilization of N5-sulfite complexes (36, 37). Thus, H-bonding interactions at N1-O2 in A79Q variants may support regulation of the flavin redox potential in a similar fashion. The stabilization of charge is reflected in alteration of the N1 chemical shift that is only altered in the presence of the A79Q sidechain and not A79T.

Coupling modulation of the flavin electronic environment with conformational dynamics of LOV proteins may be a conserved phenotype among LOV domains. In VVD and phototropin, residues at the re-face of the flavin modulate structural plasticity and adduct stability (15). In particular, a single Leu→Phe substitution dictates the different properties of LOV1 and LOV2 domains (20). In EL222, we have shown a different mechanism of altering the electronic properties of the flavin that also modulate structural dynamics. However, similar to VVD, these EL222 variants do not affect protein signaling mechanism and further expand the role of flavin H-bonding residues in affecting LOV photophysics.

The effect of flavin-pyrimidine ring H-bonding interactions on electronic modulation is supported by recent studies of rate-altering variants in YtvA (12). There, conserved residues interacting with the O2-N3-O4 locus were analyzed for their effect on triplet formation, adduct formation kinetics and stabilization of the light state adduct. Introduction of negatively charged residues destabilized adduct formation and adduct stability, consistent with H-bonding residues stabilizing increased electron density at the N1 position by forming extensive H-bonding networks at O2-N3-O4. In general, LOV proteins stabilize C4a-adducts in a manner analogous to flavoenzymes, which are tuned for reduction potential through modulation of the H-bonding character at N1 and the O2-N3-O4 positions.

Combined, we have demonstrated a new naturally occurring LOV variant that regulates LOV photocycle kinetics and active site plasticity. Notably, the mechanism of stabilization is similar to those identified in other LOV systems (15, 16); however, the active site variants here modulate adduct stability from 2–300 s at the single A79 site but without grossly affecting structure or signaling mechanisms (Supplemental Figure 1). This begs the question as to why this residue is conserved in LOV domains, except for this particular clade of marine bacterial proteins that all have abandoned the conserved Gln to afford greater dynamic flexibility in the LOV active site and increased enzyme kinetics. Notably, the conservation of signaling mechanism, while altering catalytic turnover makes A79 substitutions ideal for use in optogenetic tools. In addition, detailed kinetic characterization and identification of non-Arrhenius behavior in EL222 facilitates conversion of the light-activated LOV-HTH DNA binding protein into an in vitro tool for regulating gene expression.

Supplementary Material

1_si_001

Supplementary Table S1: Temperature dependence of EL222(14–222) dark state recovery rates as obtained by mono- or bi-exponential fitting.

Supplementary Figure S1: Arrhenius analyses of EL222(14–222) dark state recovery kinetics determined by mono- and bi-exponential fitting.

Supplementary Figure S2: Demonstration of conserved structure and function in A79 variants

Supplementary Figure S3: SIE data and Eyring analysis of EL222(14–222)

Supplementary Figure S4: Arrhenius analysis of EL222(14–144)

Abbreviations

LOV

Light Oxygen Voltage domain

UV

ultraviolet

EPR

Electro Paramagnetic Resonance

FTIR

Fourier Transform Infrared spectroscopy

H-bond

Hydrogen bond

NMR

Nuclear Magnetic Resonance

LIOAS

Laser Induced OptoAcoustic Spectroscopy

HTH

helix-turn-helix

SIE

Solvent Isotope Effect

FMN

Flavin Mononucleotide

PYP

Photoactive Yellow Protein

PAS

Period ARNT Single minded Domain